diff -u --recursive --new-file v2.4.0-prerelease/linux/Documentation/Changes linux/Documentation/Changes
--- v2.4.0-prerelease/linux/Documentation/Changes	Mon Jan  1 09:38:34 2001
+++ linux/Documentation/Changes	Mon Jan  1 10:00:04 2001
@@ -52,7 +52,7 @@
 o  Gnu make               3.77                    # make --version
 o  binutils               2.9.1.0.25              # ld -v
 o  util-linux             2.10o                   # fdformat --version
-o  modutils               2.3.21                  # insmod -V
+o  modutils               2.4.0                   # insmod -V
 o  e2fsprogs              1.19                    # tune2fs --version
 o  pcmcia-cs              3.1.21                  # cardmgr -V
 o  PPP                    2.4.0                   # pppd --version
@@ -153,7 +153,7 @@
 Ksymoops
 --------
 
-If the unthinkable happens and your kernel oopses, you'll need a 2.3
+If the unthinkable happens and your kernel oopses, you'll need a 2.4
 version of ksymoops to decode the report; see REPORTING-BUGS in the
 root of the Linux source for more information.
 
@@ -287,11 +287,11 @@
 
 Ksymoops
 --------
-o  <ftp://ftp.kernel.org/pub/linux/utils/kernel/ksymoops/v2.3>
+o  <ftp://ftp.kernel.org/pub/linux/utils/kernel/ksymoops/v2.4>
 
 Modutils
 --------
-o  <ftp://ftp.kernel.org/pub/linux/utils/kernel/modutils/v2.3/modutils-2.3.21.tar.bz2>
+o  <ftp://ftp.kernel.org/pub/linux/utils/kernel/modutils/v2.4/>
 
 Mkinitrd
 --------
diff -u --recursive --new-file v2.4.0-prerelease/linux/Documentation/Configure.help linux/Documentation/Configure.help
--- v2.4.0-prerelease/linux/Documentation/Configure.help	Mon Jan  1 09:38:34 2001
+++ linux/Documentation/Configure.help	Thu Jan  4 13:00:55 2001
@@ -15512,14 +15512,15 @@
   want). The module is called ariadne.o. If you want to compile it as
   a module, say M here and read Documentation/modules.txt.
 
-Ariadne II support
+Ariadne II and X-Surf support
 CONFIG_ARIADNE2
-  If you have a Village Tronic Ariadne II Ethernet adapter, say Y.
+  This driver is for the Village Tronic Ariadne II and the Individual
+  Computers X-Surf Ethernet cards. If you have such a card, say Y.
   Otherwise, say N.
 
   This driver is also available as a module ( = code which can be
-  inserted in and removed from the running kernel whenever you
-  want). The module is called ariadne2.o. If you want to compile it as
+  inserted in and removed from the running kernel whenever you want).
+  The module will be called ariadne2.o. If you want to compile it as
   a module, say M here and read Documentation/modules.txt.
 
 A2065 support
@@ -16993,12 +16994,6 @@
   Select this option to build a kernel for an Itanium prototype system
   with an A-step CPU.  You have an A-step CPU if the "revision" field in
   /proc/cpuinfo is 0.
-
-Enable Itanium A1-step specific code
-CONFIG_ITANIUM_A1_SPECIFIC
-  Select this option to build a kernel for an Itanium prototype system
-  with an A1-step CPU.  If you don't know whether you have an A1-step CPU,
-  you probably don't and you can answer "no" here.
 
 Enable Itanium B-step specific code
 CONFIG_ITANIUM_BSTEP_SPECIFIC
diff -u --recursive --new-file v2.4.0-prerelease/linux/Documentation/DocBook/mousedrivers.tmpl linux/Documentation/DocBook/mousedrivers.tmpl
--- v2.4.0-prerelease/linux/Documentation/DocBook/mousedrivers.tmpl	Mon Jun 19 12:56:07 2000
+++ linux/Documentation/DocBook/mousedrivers.tmpl	Thu Jan  4 12:50:12 2001
@@ -335,7 +335,7 @@
   <para>
     We count off a user and provided that there are still other users need 
     take no further action. The last person closing the mouse causes us to 
-    free up the interrupt. This stopps interrupts from the mouse from using 
+    free up the interrupt. This stops interrupts from the mouse from using 
     our CPU time, and lets us use <function>MOD_DEC_USE_COUNT</function> so 
     that the mouse can now be unloaded.
   </para>
@@ -404,7 +404,7 @@
     play with them.
   </para>
   <para>
-    If a change has occured we also need to wake sleeping processes, so we 
+    If a change has occurred we also need to wake sleeping processes, so we 
     add a wakeup call and a <structname>wait_queue</structname> to use when 
     we wish to await a mouse event.
   </para>
@@ -426,7 +426,7 @@
   <para>
     This is fairly standard poll code. First we add the wait queue to the 
     list of queues we want to monitor for an event. Secondly we check if an 
-    event has occured. We only have one kind of event - the 
+    event has occurred. We only have one kind of event - the 
     <varname>mouse_event</varname> flag tells us that something happened. 
     We know that this something can only be mouse data. We return the flags 
     indicating input and normal reading will succeed.
@@ -476,7 +476,7 @@
   </para>
   <para>
     Next we wait for an event to occur. The loop is fairly standard event
-    waiting in Linux. Having checked that the event has not yet occured, we
+    waiting in Linux. Having checked that the event has not yet occurred, we
     then check if an event is pending and if not we need to sleep. 
   </para>
   <para>
@@ -488,7 +488,7 @@
     Next we sleep until the mouse or a signal awakens us. A signal will 
     awaken us as we have used <function>wakeup_interruptible</function>. 
     This is important as it means a user can kill processes waiting for 
-    the mouse - clearly a desireable property. If we are interrupted we 
+    the mouse - clearly a desirable property. If we are interrupted we 
     exit the call and the kernel will then process signals and maybe 
     restart the call again - from the beginning.
   </para>
diff -u --recursive --new-file v2.4.0-prerelease/linux/Documentation/DocBook/videobook.tmpl linux/Documentation/DocBook/videobook.tmpl
--- v2.4.0-prerelease/linux/Documentation/DocBook/videobook.tmpl	Mon Dec 11 17:59:43 2000
+++ linux/Documentation/DocBook/videobook.tmpl	Thu Jan  4 12:50:12 2001
@@ -486,7 +486,7 @@
   <para>
         We copy the user supplied structure into kernel memory so we can examine it. 
         If the user has selected a tuner other than zero we reject the request. If 
-        they wanted tuner 0 then, suprisingly enough, that is the current tuner already.
+        they wanted tuner 0 then, surprisingly enough, that is the current tuner already.
   </para>
   <para>
         The next two ioctls we need to provide are to get and set the frequency of
@@ -652,7 +652,7 @@
   </para>
   <para>
         The VIDIOCSAUDIO ioctl allows the user to set the audio parameters in the
-        video_audio stucture. The driver does its best to honour the request.
+        video_audio structure. The driver does its best to honour the request.
   </para>
   <programlisting>
 
@@ -812,7 +812,7 @@
         Chroma keying is a technique used by cards to get around this. It is an old
         television mixing trick where you mark all the areas you wish to replace
         with a single clear colour that isn't used in the image - TV people use an
-        incredibly bright blue while computing people often use a paticularly
+        incredibly bright blue while computing people often use a particularly
         virulent purple. Bright blue occurs on the desktop. Anyone with virulent
         purple windows has another problem besides their TV overlay.
   </para>
@@ -1259,7 +1259,7 @@
    </row><row>
         <entry>VIDEO_MODE_NTSC</><>NTSC (US) encoded Television</entry>
    </row><row>
-        <entry>VIDEO_MODE_SECAM</><>SECAM (French) Televison </entry>
+        <entry>VIDEO_MODE_SECAM</><>SECAM (French) Television </entry>
    </row><row>
         <entry>VIDEO_MODE_AUTO</><>Automatic switching, or format does not
                                 matter</entry>
@@ -1269,7 +1269,7 @@
     </table>
     <para>
         The corresponding VIDIOCSCHAN ioctl allows a user to change channel and to
-        request the norm is changed - for exaple to switch between a PAL or an NTSC
+        request the norm is changed - for example to switch between a PAL or an NTSC
         format camera.
   </para>
   <programlisting>
@@ -1332,7 +1332,7 @@
         it make a best effort attempt.
   </para>
   <para>
-        Our depth is 24, as this is in bits. We will be returing RGB24 format. This
+        Our depth is 24, as this is in bits. We will be returning RGB24 format. This
         has one byte of red, then one of green, then one of blue. This then repeats
         for every other pixel in the image. The other common formats the interface 
         defines are
diff -u --recursive --new-file v2.4.0-prerelease/linux/Documentation/DocBook/z8530book.tmpl linux/Documentation/DocBook/z8530book.tmpl
--- v2.4.0-prerelease/linux/Documentation/DocBook/z8530book.tmpl	Sun Mar 12 19:39:47 2000
+++ linux/Documentation/DocBook/z8530book.tmpl	Thu Jan  4 12:50:12 2001
@@ -57,7 +57,7 @@
       <title>Introduction</title>
   <para>
 	The Z85x30 family synchronous/asynchronous controller chips are
-	used on a larg number of cheap network interface cards. The
+	used on a large number of cheap network interface cards. The
 	kernel provides a core interface layer that is designed to make
 	it easy to provide WAN services using this chip.
   </para>
@@ -124,7 +124,7 @@
 	for allocating the interrupt line. The interrupt handler should be
 	set to <function>z8530_interrupt</function>. The device id should
 	be set to the z8530_dev structure pointer. Whether the interrupt can
-	be shared or not is board dependant, and up to you to initialise.
+	be shared or not is board dependent, and up to you to initialise.
   </para>
   <para>
 	The structure holds two channel structures. 
@@ -143,19 +143,19 @@
   </para>
   <para>
 	Repeat the same operation with the B channel if your chip has
-	both channels wired to something useful. This isnt always the
+	both channels wired to something useful. This isn't always the
 	case. If it is not wired then the I/O values do not matter, but
 	you must initialise chanB.dev.
   </para>
   <para>
 	If your board has DMA facilities then initialise the txdma and
 	rxdma fields for the relevant channels. You must also allocate the
-	ISA DMA channels and do any neccessary board level initialisation
+	ISA DMA channels and do any necessary board level initialisation
 	to configure them. The low level driver will do the Z8530 and
 	DMA controller programming but not board specific magic.
   </para>
   <para>
-	Having intialised the device you can then call
+	Having initialised the device you can then call
 	<function>z8530_init</function>. This will probe the chip and 
 	reset it into a known state. An identification sequence is then
 	run to identify the chip type. If the checks fail to pass the
@@ -167,7 +167,7 @@
   <para>
 	Once you have called z8530_init you can also make use of the utility
 	function <function>z8530_describe</function>. This provides a 
-	consistant reporting format for the Z8530 devices, and allows all
+	consistent reporting format for the Z8530 devices, and allows all
 	the drivers to provide consistent reporting.
   </para>
   </chapter>
@@ -191,7 +191,7 @@
 	to the syncppp structures.
   </para>
   <para>
-	The way most drivers approach this paticular problem is to
+	The way most drivers approach this particular problem is to
 	create a structure holding the Z8530 device definition and
 	put that and the syncppp pointer into the private field of
 	the network device. The network device fields of the channels
@@ -330,7 +330,7 @@
   <para>
 	The Z8530 driver is written to be portable. In DMA mode it makes
 	assumptions about the use of ISA DMA. These are probably warranted
-	in most cases as the Z85230 in paticular was designed to glue to PC
+	in most cases as the Z85230 in particular was designed to glue to PC
 	type machines. The PIO mode makes no real assumptions.
   </para>
   <para>
diff -u --recursive --new-file v2.4.0-prerelease/linux/Documentation/IO-mapping.txt linux/Documentation/IO-mapping.txt
--- v2.4.0-prerelease/linux/Documentation/IO-mapping.txt	Tue Oct 31 12:42:25 2000
+++ linux/Documentation/IO-mapping.txt	Thu Jan  4 13:00:15 2001
@@ -1,3 +1,9 @@
+[ NOTE: The virt_to_bus() and bus_to_virt() functions have been
+	superseded by the functionality provided by the PCI DMA
+	interface (see Documentation/DMA-mapping.txt).  They continue
+	to be documented below for historical purposes, but new code
+	must not use them. --davidm 00/12/12 ]
+
 [ This is a mail message in response to a query on IO mapping, thus the
   strange format for a "document" ]
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/Documentation/rtc.txt linux/Documentation/rtc.txt
--- v2.4.0-prerelease/linux/Documentation/rtc.txt	Tue Jan 25 14:13:47 2000
+++ linux/Documentation/rtc.txt	Thu Jan  4 12:50:17 2001
@@ -56,7 +56,7 @@
 exclusive access to the device for your applications.
 
 The alarm and/or interrupt frequency are programmed into the RTC via
-various ioctl(2) calls as listed in ./include/linux/mc146818rtc.h
+various ioctl(2) calls as listed in ./include/linux/rtc.h
 Rather than write 50 pages describing the ioctl() and so on, it is
 perhaps more useful to include a small test program that demonstrates
 how to use them, and demonstrates the features of the driver. This is
@@ -81,7 +81,7 @@
  */
 
 #include <stdio.h>
-#include <linux/mc146818rtc.h>
+#include <linux/rtc.h>
 #include <sys/ioctl.h>
 #include <sys/time.h>
 #include <sys/types.h>
diff -u --recursive --new-file v2.4.0-prerelease/linux/Makefile linux/Makefile
--- v2.4.0-prerelease/linux/Makefile	Mon Jan  1 09:38:34 2001
+++ linux/Makefile	Thu Jan  4 13:48:13 2001
@@ -1,7 +1,7 @@
 VERSION = 2
 PATCHLEVEL = 4
 SUBLEVEL = 0
-EXTRAVERSION = -prerelease
+EXTRAVERSION =
 
 KERNELRELEASE=$(VERSION).$(PATCHLEVEL).$(SUBLEVEL)$(EXTRAVERSION)
 
@@ -145,7 +145,7 @@
 DRIVERS-$(CONFIG_ATM) += drivers/atm/atm.o
 DRIVERS-$(CONFIG_IDE) += drivers/ide/idedriver.o
 DRIVERS-$(CONFIG_SCSI) += drivers/scsi/scsidrv.o
-DRIVERS-$(CONFIG_IEEE1394) += drivers/ieee1394/ieee1394.a
+DRIVERS-$(CONFIG_IEEE1394) += drivers/ieee1394/ieee1394drv.o
 
 ifneq ($(CONFIG_CD_NO_IDESCSI)$(CONFIG_BLK_DEV_IDECD)$(CONFIG_BLK_DEV_SR)$(CONFIG_PARIDE_PCD),)
 DRIVERS-y += drivers/cdrom/driver.o
@@ -188,6 +188,7 @@
 	.tmp* \
 	drivers/char/consolemap_deftbl.c drivers/video/promcon_tbl.c \
 	drivers/char/conmakehash \
+	drivers/char/drm/*-mod.c \
 	drivers/pci/devlist.h drivers/pci/classlist.h drivers/pci/gen-devlist \
 	drivers/zorro/devlist.h drivers/zorro/gen-devlist \
 	drivers/sound/bin2hex drivers/sound/hex2hex \
diff -u --recursive --new-file v2.4.0-prerelease/linux/README linux/README
--- v2.4.0-prerelease/linux/README	Tue Oct 31 12:42:25 2000
+++ linux/README	Tue Jan  2 16:55:26 2001
@@ -1,26 +1,9 @@
-	Linux kernel release 2.3.xx
+	Linux kernel release 2.4.xx
 
-These are the release notes for Linux version 2.3.  Read them carefully,
+These are the release notes for Linux version 2.4.  Read them carefully,
 as they tell you what this is all about, explain how to install the
 kernel, and what to do if something goes wrong. 
 
-Linux version 2.3 is a DEVELOPMENT kernel, and not intended for general
-public use.  Different releases may have various and sometimes severe
-bugs.  It is *strongly* recommended that you back up the previous kernel
-before installing any new 2.3.xx release.
-
-If you need to use a proven and stable Linux kernel, please use 2.0.38
-or 2.2.xx.  All features which will be in the 2.3.xx releases will be
-contained in 2.4.xx when the code base has stabilized again. 
-
-If you decide to use 2.3, it is recommended that you join the kernel mailing
-list.  To do this, e-mail majordomo@vger.kernel.org, and put in the body
-of the message "subscribe linux-kernel" or "subscribe linux-kernel-digest"
-for a daily digest of the mailing list (it is a high-traffic list.)
-
-However, please make sure you don't ask questions which are already answered
-in various files in the Documentation directory.  See DOCUMENTATION below.
-
 WHAT IS LINUX?
 
   Linux is a Unix clone written from scratch by Linus Torvalds with
@@ -63,7 +46,7 @@
    directory where you have permissions (eg. your home directory) and
    unpack it:
 
-		gzip -cd linux-2.3.XX.tar.gz | tar xvf -
+		gzip -cd linux-2.4.XX.tar.gz | tar xvf -
 
    Replace "XX" with the version number of the latest kernel.
 
@@ -72,7 +55,7 @@
    files.  They should match the library, and not get messed up by
    whatever the kernel-du-jour happens to be.
 
- - You can also upgrade between 2.3.xx releases by patching.  Patches are
+ - You can also upgrade between 2.4.xx releases by patching.  Patches are
    distributed in the traditional gzip and the new bzip2 format.  To
    install by patching, get all the newer patch files, enter the
    directory in which you unpacked the kernel source and execute:
@@ -107,7 +90,7 @@
 
 SOFTWARE REQUIREMENTS
 
-   Compiling and running the 2.3.xx kernels requires up-to-date
+   Compiling and running the 2.4.xx kernels requires up-to-date
    versions of various software packages.  Consult
    ./Documentation/Changes for the minimum version numbers required
    and how to get updates for these packages.  Beware that using
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/alpha/kernel/setup.c linux/arch/alpha/kernel/setup.c
--- v2.4.0-prerelease/linux/arch/alpha/kernel/setup.c	Tue Oct 31 12:42:25 2000
+++ linux/arch/alpha/kernel/setup.c	Tue Jan  2 16:45:37 2001
@@ -1089,7 +1089,8 @@
 		       hwrpb->pagesize,
 		       hwrpb->pa_bits,
 		       hwrpb->max_asn,
-		       loops_per_sec / 500000, (loops_per_sec / 5000) % 100,
+		       loops_per_jiffy / (500000/HZ),
+		       (loops_per_jiffy / (5000/HZ)) % 100,
 		       unaligned[0].count, unaligned[0].pc, unaligned[0].va,
 		       unaligned[1].count, unaligned[1].pc, unaligned[1].va,
 		       platform_string(), nr_processors);
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/alpha/kernel/smp.c linux/arch/alpha/kernel/smp.c
--- v2.4.0-prerelease/linux/arch/alpha/kernel/smp.c	Mon Jan  1 09:38:34 2001
+++ linux/arch/alpha/kernel/smp.c	Tue Jan  2 16:45:37 2001
@@ -105,7 +105,7 @@
 static inline void __init
 smp_store_cpu_info(int cpuid)
 {
-	cpu_data[cpuid].loops_per_sec = loops_per_sec;
+	cpu_data[cpuid].loops_per_jiffy = loops_per_jiffy;
 	cpu_data[cpuid].last_asn = ASN_FIRST_VERSION;
 	cpu_data[cpuid].need_new_asn = 0;
 	cpu_data[cpuid].asn_lock = 0;
@@ -601,12 +601,12 @@
 	bogosum = 0;
 	for (i = 0; i < NR_CPUS; i++) {
 		if (cpu_present_mask & (1L << i))
-			bogosum += cpu_data[i].loops_per_sec;
+			bogosum += cpu_data[i].loops_per_jiffy;
 	}
 	printk(KERN_INFO "SMP: Total of %d processors activated "
 	       "(%lu.%02lu BogoMIPS).\n",
-	       cpu_count, (bogosum + 2500) / 500000,
-	       ((bogosum + 2500) / 5000) % 100);
+	       cpu_count, (bogosum + 2500) / (500000/HZ),
+	       ((bogosum + 2500) / (5000/HZ)) % 100);
 
 	smp_num_cpus = cpu_count;
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/i386/kernel/pci-irq.c linux/arch/i386/kernel/pci-irq.c
--- v2.4.0-prerelease/linux/arch/i386/kernel/pci-irq.c	Mon Jan  1 09:38:34 2001
+++ linux/arch/i386/kernel/pci-irq.c	Wed Jan  3 20:45:26 2001
@@ -157,25 +157,7 @@
 {
 	static unsigned char irqmap[16] = { 0, 9, 3, 10, 4, 5, 7, 6, 1, 11, 0, 12, 0, 14, 0, 15 };
 
-	switch (pirq) {
-	case 0x00:
-		return 0;
-	default:
-		return irqmap[read_config_nybble(router, 0x48, pirq-1)];
-	case 0xfe:
-		return irqmap[read_config_nybble(router, 0x44, 0)];
-	case 0xff:
-		return irqmap[read_config_nybble(router, 0x75, 0)];
-	}
-}
-
-static void pirq_ali_ide_interrupt(struct pci_dev *router, unsigned reg, unsigned val, unsigned irq)
-{
-	u8 x;
-
-	pci_read_config_byte(router, reg, &x);
-	x = (x & 0xe0) | val;	/* clear the level->edge transform */
-	pci_write_config_byte(router, reg, x);
+	return irqmap[read_config_nybble(router, 0x48, pirq-1)];
 }
 
 static int pirq_ali_set(struct pci_dev *router, struct pci_dev *dev, int pirq, int irq)
@@ -184,17 +166,7 @@
 	unsigned int val = irqmap[irq];
 		
 	if (val) {
-		switch (pirq) {
-		default:
-			write_config_nybble(router, 0x48, pirq-1, val);
-			break;
-		case 0xfe:
-			pirq_ali_ide_interrupt(router, 0x44, val, irq);
-			break;
-		case 0xff:
-			pirq_ali_ide_interrupt(router, 0x75, val, irq);
-			break;
-		}
+		write_config_nybble(router, 0x48, pirq-1, val);
 		return 1;
 	}
 	return 0;
@@ -202,40 +174,25 @@
 
 /*
  * The Intel PIIX4 pirq rules are fairly simple: "pirq" is
- * just a pointer to the config space. However, something
- * funny is going on with 0xfe/0xff, and apparently they
- * should handle IDE irq routing. Ignore them for now.
+ * just a pointer to the config space.
  */
 static int pirq_piix_get(struct pci_dev *router, struct pci_dev *dev, int pirq)
 {
 	u8 x;
 
-	switch (pirq) {
-	case 0xfe:
-	case 0xff:
-		return 0;
-	default:
-		pci_read_config_byte(router, pirq, &x);
-		return (x < 16) ? x : 0;
-	}
+	pci_read_config_byte(router, pirq, &x);
+	return (x < 16) ? x : 0;
 }
 
 static int pirq_piix_set(struct pci_dev *router, struct pci_dev *dev, int pirq, int irq)
 {
-	switch (pirq) {
-	case 0xfe:
-	case 0xff:
-		return 0;
-	default:
-		pci_write_config_byte(router, pirq, irq);
-		return 1;
-	}
+	pci_write_config_byte(router, pirq, irq);
+	return 1;
 }
 
 /*
  * The VIA pirq rules are nibble-based, like ALI,
- * but without the ugly irq number munging or the
- * strange special cases..
+ * but without the ugly irq number munging.
  */
 static int pirq_via_get(struct pci_dev *router, struct pci_dev *dev, int pirq)
 {
@@ -500,8 +457,16 @@
 	}
 	DBG(" -> newirq=%d", newirq);
 
-	/* Try to get current IRQ */
-	if (r->get && (irq = r->get(pirq_router_dev, dev, pirq))) {
+	/* Check if it is hardcoded */
+	if ((pirq & 0xf0) == 0xf0) {
+		irq = pirq & 0xf;
+		DBG(" -> hardcoded IRQ %d\n", irq);
+		msg = "Hardcoded";
+		if (dev->irq && dev->irq != irq) {
+			printk("IRQ routing conflict in pirq table! Try 'pci=autoirq'\n");
+			return 0;
+		}
+	} else if (r->get && (irq = r->get(pirq_router_dev, dev, pirq))) {
 		DBG(" -> got IRQ %d\n", irq);
 		msg = "Found";
 		/* We refuse to override the dev->irq information. Give a warning! */
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/i386/kernel/process.c linux/arch/i386/kernel/process.c
--- v2.4.0-prerelease/linux/arch/i386/kernel/process.c	Mon Jan  1 09:38:34 2001
+++ linux/arch/i386/kernel/process.c	Thu Jan  4 12:50:17 2001
@@ -32,6 +32,7 @@
 #include <linux/delay.h>
 #include <linux/reboot.h>
 #include <linux/init.h>
+#include <linux/mc146818rtc.h>
 
 #include <asm/uaccess.h>
 #include <asm/pgtable.h>
@@ -257,6 +258,8 @@
  */
 void machine_real_restart(unsigned char *code, int length)
 {
+	unsigned long flags;
+
 	cli();
 
 	/* Write zero to CMOS register number 0x0f, which the BIOS POST
@@ -266,10 +269,12 @@
 	   disable NMIs by setting the top bit in the CMOS address register,
 	   as we're about to do peculiar things to the CPU.  I'm not sure if
 	   `outb_p' is needed instead of just `outb'.  Use it to be on the
-	   safe side. */
+	   safe side.  (Yes, CMOS_WRITE does outb_p's. -  Paul G.)
+	 */
 
-	outb_p (0x8f, 0x70);
-	outb_p (0x00, 0x71);
+	spin_lock_irqsave(&rtc_lock, flags);
+	CMOS_WRITE(0x00, 0x8f);
+	spin_unlock_irqrestore(&rtc_lock, flags);
 
 	/* Remap the kernel at virtual address zero, as well as offset zero
 	   from the kernel segment.  This assumes the kernel segment starts at
@@ -379,13 +384,14 @@
 		pm_power_off();
 }
 
+extern void show_trace(unsigned long* esp);
 
 void show_regs(struct pt_regs * regs)
 {
 	unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
 
 	printk("\n");
-	printk("EIP: %04x:[<%08lx>]",0xffff & regs->xcs,regs->eip);
+	printk("EIP: %04x:[<%08lx>] CPU: %d",0xffff & regs->xcs,regs->eip, smp_processor_id());
 	if (regs->xcs & 3)
 		printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp);
 	printk(" EFLAGS: %08lx\n",regs->eflags);
@@ -407,6 +413,7 @@
 		".previous			\n"
 		: "=r" (cr4): "0" (0));
 	printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n", cr0, cr2, cr3, cr4);
+	show_trace(&regs->esp);
 }
 
 /*
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/i386/kernel/traps.c linux/arch/i386/kernel/traps.c
--- v2.4.0-prerelease/linux/arch/i386/kernel/traps.c	Mon Dec 11 17:59:43 2000
+++ linux/arch/i386/kernel/traps.c	Wed Jan  3 20:45:26 2001
@@ -89,33 +89,18 @@
 
 /*
  * These constants are for searching for possible module text
- * segments. MODULE_RANGE is a guess of how much space is likely
- * to be vmalloced.
+ * segments.
  */
-#define MODULE_RANGE (8*1024*1024)
 
-void show_stack(unsigned long * esp)
+void show_trace(unsigned long * stack)
 {
-	unsigned long *stack, addr, module_start, module_end;
 	int i;
+	unsigned long addr, module_start, module_end;
 
-	// debugging aid: "show_stack(NULL);" prints the
-	// back trace for this cpu.
-
-	if(esp==NULL)
-		esp=(unsigned long*)&esp;
-
-	stack = esp;
-	for(i=0; i < kstack_depth_to_print; i++) {
-		if (((long) stack & (THREAD_SIZE-1)) == 0)
-			break;
-		if (i && ((i % 8) == 0))
-			printk("\n       ");
-		printk("%08lx ", *stack++);
-	}
+	if (!stack)
+		stack = (unsigned long*)&stack;
 
-	printk("\nCall Trace: ");
-	stack = esp;
+	printk("Call Trace: ");
 	i = 1;
 	module_start = VMALLOC_START;
 	module_end = VMALLOC_END;
@@ -138,6 +123,30 @@
 			i++;
 		}
 	}
+	printk("\n");
+}
+
+void show_stack(unsigned long * esp)
+{
+	unsigned long *stack;
+	int i;
+
+	// debugging aid: "show_stack(NULL);" prints the
+	// back trace for this cpu.
+
+	if(esp==NULL)
+		esp=(unsigned long*)&esp;
+
+	stack = esp;
+	for(i=0; i < kstack_depth_to_print; i++) {
+		if (((long) stack & (THREAD_SIZE-1)) == 0)
+			break;
+		if (i && ((i % 8) == 0))
+			printk("\n       ");
+		printk("%08lx ", *stack++);
+	}
+	printk("\n");
+	show_trace(esp);
 }
 
 static void show_registers(struct pt_regs *regs)
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/i386/vmlinux.lds linux/arch/i386/vmlinux.lds
--- v2.4.0-prerelease/linux/arch/i386/vmlinux.lds	Wed Jul  5 13:42:37 2000
+++ linux/arch/i386/vmlinux.lds	Wed Jan  3 20:45:26 2001
@@ -14,6 +14,9 @@
 	*(.gnu.warning)
 	} = 0x9090
   .text.lock : { *(.text.lock) }	/* out-of-line lock text */
+
+  _etext = .;			/* End of text section */
+
   .rodata : { *(.rodata) }
   .kstrtab : { *(.kstrtab) }
 
@@ -25,8 +28,6 @@
   __start___ksymtab = .;	/* Kernel symbol table */
   __ksymtab : { *(__ksymtab) }
   __stop___ksymtab = .;
-
-  _etext = .;			/* End of text section */
 
   .data : {			/* Data */
 	*(.data)
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/Makefile linux/arch/ia64/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/Makefile	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/Makefile	Thu Jan  4 12:50:17 2001
@@ -19,22 +19,28 @@
 EXTRA	=
 
 CFLAGS := $(CFLAGS) -pipe $(EXTRA) -Wa,-x -ffixed-r13 -mfixed-range=f10-f15,f32-f127 \
-	  -funwind-tables
+	  -funwind-tables -falign-functions=32
+# -frename-registers
 CFLAGS_KERNEL := -mconstant-gp
 
 ifeq ($(CONFIG_ITANIUM_ASTEP_SPECIFIC),y)
 	CFLAGS += -ma-step
 endif
+ifeq ($(CONFIG_ITANIUM_BSTEP_SPECIFIC),y)
+	CFLAGS += -mb-step
+endif
 
 ifdef CONFIG_IA64_GENERIC
 	CORE_FILES      :=      arch/$(ARCH)/hp/hp.a	\
 				arch/$(ARCH)/sn/sn.a	\
 				arch/$(ARCH)/dig/dig.a	\
+				arch/$(ARCH)/sn/io/sgiio.o \
 				$(CORE_FILES)
 	SUBDIRS		:=	arch/$(ARCH)/hp		\
 				arch/$(ARCH)/sn/sn1	\
 				arch/$(ARCH)/sn		\
 				arch/$(ARCH)/dig	\
+				arch/$(ARCH)/sn/io	\
 				$(SUBDIRS)
 
 else # !GENERIC
@@ -47,10 +53,7 @@
 endif
 
 ifdef CONFIG_IA64_SGI_SN1
-CFLAGS := $(CFLAGS) -DSN -I. -DBRINGUP -DDIRECT_L1_CONSOLE \
-		-DNUMA_BASE -DSIMULATED_KLGRAPH -DNUMA_MIGR_CONTROL  \
-		-DLITTLE_ENDIAN -DREAL_HARDWARE -DLANGUAGE_C=1 	     \
-		-D_LANGUAGE_C=1
+CFLAGS += -DBRINGUP
         SUBDIRS         :=      arch/$(ARCH)/sn/sn1	\
 				arch/$(ARCH)/sn		\
 				arch/$(ARCH)/sn/io	\
@@ -96,7 +99,7 @@
 
 arch/$(ARCH)/vmlinux.lds: arch/$(ARCH)/vmlinux.lds.S FORCE
 	$(CPP) -D__ASSEMBLY__ -C -P -I$(HPATH) -I$(HPATH)/asm-$(ARCH) \
-		arch/$(ARCH)/vmlinux.lds.S > $@
+		-traditional arch/$(ARCH)/vmlinux.lds.S > $@
 
 FORCE: ;
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/boot/Makefile linux/arch/ia64/boot/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/boot/Makefile	Thu Jun 22 07:09:44 2000
+++ linux/arch/ia64/boot/Makefile	Thu Jan  4 12:50:17 2001
@@ -16,13 +16,11 @@
 	$(CC) $(AFLAGS) -traditional -c -o $*.o $<
 
 OBJECTS	= bootloader.o
-TARGETS =
 
-ifdef CONFIG_IA64_HP_SIM
- TARGETS += bootloader
-endif
+targets-$(CONFIG_IA64_HP_SIM) += bootloader
+targets-$(CONFIG_IA64_GENERIC) += bootloader
 
-all:	$(TARGETS)
+all:	$(targets-y)
 
 bootloader: $(OBJECTS)
 	$(LD) $(LINKFLAGS) $(OBJECTS) $(TOPDIR)/lib/lib.a $(TOPDIR)/arch/$(ARCH)/lib/lib.a \
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/config.in linux/arch/ia64/config.in
--- v2.4.0-prerelease/linux/arch/ia64/config.in	Sun Nov 19 18:44:03 2000
+++ linux/arch/ia64/config.in	Thu Jan  4 12:50:17 2001
@@ -18,7 +18,6 @@
 comment 'General setup'
 
 define_bool CONFIG_IA64 y
-define_bool CONFIG_SWIOTLB y	# for now...
 
 define_bool CONFIG_ISA n
 define_bool CONFIG_EISA n
@@ -41,20 +40,22 @@
 	define_bool CONFIG_ITANIUM y
 	define_bool CONFIG_IA64_BRL_EMU y
 	bool '  Enable Itanium A-step specific code' CONFIG_ITANIUM_ASTEP_SPECIFIC
-	if [ "$CONFIG_ITANIUM_ASTEP_SPECIFIC" = "y" ]; then
-	  bool '   Enable Itanium A1-step specific code' CONFIG_ITANIUM_A1_SPECIFIC
-	fi
 	bool '  Enable Itanium B-step specific code' CONFIG_ITANIUM_BSTEP_SPECIFIC
 	if [ "$CONFIG_ITANIUM_BSTEP_SPECIFIC" = "y" ]; then
 	  bool '   Enable Itanium B0-step specific code' CONFIG_ITANIUM_B0_SPECIFIC
+	  bool '   Enable Itanium B1-step specific code' CONFIG_ITANIUM_B1_SPECIFIC
+	  bool '   Enable Itanium B2-step specific code' CONFIG_ITANIUM_B2_SPECIFIC
+	fi
+	bool '  Enable Itanium C-step specific code' CONFIG_ITANIUM_CSTEP_SPECIFIC
+	if [ "$CONFIG_ITANIUM_CSTEP_SPECIFIC" = "y" ]; then
+	  bool '   Enable Itanium C0-step specific code' CONFIG_ITANIUM_C0_SPECIFIC
 	fi
 	bool '  Force interrupt redirection' CONFIG_IA64_HAVE_IRQREDIR
 	bool '  Enable use of global TLB purge instruction (ptc.g)' CONFIG_ITANIUM_PTCG
 	bool '  Enable SoftSDV hacks' CONFIG_IA64_SOFTSDV_HACKS
 	bool '  Enable AzusA hacks' CONFIG_IA64_AZUSA_HACKS
 	bool '  Enable IA-64 Machine Check Abort' CONFIG_IA64_MCA
-	bool '  Force socket buffers below 4GB?' CONFIG_SKB_BELOW_4GB
-
+	bool '  Enable ACPI 2.0 with errata 1.3' CONFIG_ACPI20
 	bool '  ACPI kernel configuration manager (EXPERIMENTAL)' CONFIG_ACPI_KERNEL_CONFIG
 	if [ "$CONFIG_ACPI_KERNEL_CONFIG" = "y" ]; then
 	  define_bool CONFIG_PM y
@@ -70,13 +71,16 @@
 	  bool '    Enable Itanium B0-step specific code' CONFIG_ITANIUM_B0_SPECIFIC
 	fi
 	bool '  Enable SGI Medusa Simulator Support' CONFIG_IA64_SGI_SN1_SIM n
-        bool '  Enable SGI hack for version 1.0 syngery bugs' CONFIG_IA64_SGI_SYNERGY_1_0_HACKS n
 	define_bool CONFIG_DEVFS_DEBUG y
 	define_bool CONFIG_DEVFS_FS y
 	define_bool CONFIG_IA64_BRL_EMU y
 	define_bool CONFIG_IA64_MCA y
-	define_bool CONFIG_IA64_SGI_IO y
 	define_bool CONFIG_ITANIUM y
+	define_bool CONFIG_SGI_IOC3_ETH y
+	define_bool CONFIG_PERCPU_IRQ y
+	define_int  CONFIG_CACHE_LINE_SHIFT 7
+	bool '  Enable DISCONTIGMEM support' CONFIG_DISCONTIGMEM y
+	bool '	Enable NUMA support' CONFIG_NUMA y
 fi
 
 define_bool CONFIG_KCORE_ELF y	# On IA-64, we always want an ELF /proc/kcore.
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/dig/Makefile linux/arch/ia64/dig/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/dig/Makefile	Fri Apr 21 15:21:23 2000
+++ linux/arch/ia64/dig/Makefile	Thu Jan  4 12:50:17 2001
@@ -12,12 +12,10 @@
 
 all: dig.a
 
-O_TARGET        = dig.a
-O_OBJS          = iosapic.o setup.o
+O_TARGET := dig.a
 
-ifdef CONFIG_IA64_GENERIC
-O_OBJS		+= machvec.o
-endif
+obj-y := setup.o
+obj-$(CONFIG_IA64_GENERIC) += machvec.o
 
 clean::
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/dig/iosapic.c linux/arch/ia64/dig/iosapic.c
--- v2.4.0-prerelease/linux/arch/ia64/dig/iosapic.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/dig/iosapic.c	Wed Dec 31 16:00:00 1969
@@ -1,409 +0,0 @@
-/*
- * Streamlined APIC support.
- *
- * Copyright (C) 1999 Intel Corp.
- * Copyright (C) 1999 Asit Mallick <asit.k.mallick@intel.com>
- * Copyright (C) 1999-2000 Hewlett-Packard Co.
- * Copyright (C) 1999-2000 David Mosberger-Tang <davidm@hpl.hp.com>
- * Copyright (C) 1999 VA Linux Systems
- * Copyright (C) 1999,2000 Walt Drummond <drummond@valinux.com>
- *
- * 00/04/19	D. Mosberger	Rewritten to mirror more closely the x86 I/O APIC code.
- *				In particular, we now have separate handlers for edge
- *				and level triggered interrupts.
- */
-#include <linux/config.h>
-
-#include <linux/kernel.h>
-#include <linux/init.h>
-#include <linux/pci.h>
-#include <linux/smp.h>
-#include <linux/smp_lock.h>
-#include <linux/string.h>
-#include <linux/irq.h>
-
-#include <asm/acpi-ext.h>
-#include <asm/delay.h>
-#include <asm/io.h>
-#include <asm/iosapic.h>
-#include <asm/machvec.h>
-#include <asm/processor.h>
-#include <asm/ptrace.h>
-#include <asm/system.h>
-
-#ifdef	CONFIG_ACPI_KERNEL_CONFIG
-# include <asm/acpikcfg.h>
-#endif
-
-#undef DEBUG_IRQ_ROUTING
-
-static spinlock_t iosapic_lock = SPIN_LOCK_UNLOCKED;
-
-struct iosapic_vector iosapic_vector[NR_IRQS] = {
-	[0 ... NR_IRQS-1] = { -1, -1, -1, -1, -1, -1, -1, -1, -1, -1 }
-};
-
-/*
- * find the IRQ in the IOSAPIC map for the PCI device on bus/slot/pin
- */
-int
-iosapic_get_PCI_irq_vector (int bus, int slot, int pci_pin)
-{
-	int i;
-
-	for (i = 0; i < NR_IRQS; i++) {
-		if ((iosapic_bustype(i) == BUS_PCI) &&
-		    (iosapic_bus(i) == bus) &&
-		    (iosapic_busdata(i) == ((slot << 16) | pci_pin))) {
-			return i;
-		}
-	}
-	return -1;
-}
-
-static void
-set_rte (unsigned long iosapic_addr, int entry, int pol, int trigger, int delivery,
-	 long dest, int vector)
-{
-	u32 low32;
-	u32 high32;
-
-	low32 = ((pol << IO_SAPIC_POLARITY_SHIFT) |
-		 (trigger << IO_SAPIC_TRIGGER_SHIFT) |
-		 (delivery << IO_SAPIC_DELIVERY_SHIFT) |
-		 vector);
-
-#ifdef CONFIG_IA64_AZUSA_HACKS
-	/* set Flush Disable bit */
-	if (iosapic_addr != 0xc0000000fec00000)
-		low32 |= (1 << 17);
-#endif
-
-	/* dest contains both id and eid */
-	high32 = (dest << IO_SAPIC_DEST_SHIFT);	
-
-	writel(IO_SAPIC_RTE_HIGH(entry), iosapic_addr + IO_SAPIC_REG_SELECT);
-	writel(high32, iosapic_addr + IO_SAPIC_WINDOW);
-	writel(IO_SAPIC_RTE_LOW(entry), iosapic_addr + IO_SAPIC_REG_SELECT);
-	writel(low32, iosapic_addr + IO_SAPIC_WINDOW);
-}
-
-static void
-nop (unsigned int irq)
-{
-	/* do nothing... */
-}
-
-static void 
-mask_irq (unsigned int irq)
-{
-	unsigned long flags, iosapic_addr = iosapic_addr(irq);
-	u32 low32;
-
-	spin_lock_irqsave(&iosapic_lock, flags);
-	{
-		writel(IO_SAPIC_RTE_LOW(iosapic_pin(irq)), iosapic_addr + IO_SAPIC_REG_SELECT);
-		low32 = readl(iosapic_addr + IO_SAPIC_WINDOW);
-
-		low32 |= (1 << IO_SAPIC_MASK_SHIFT);    /* Zero only the mask bit */
-		writel(low32, iosapic_addr + IO_SAPIC_WINDOW);
-	}
-	spin_unlock_irqrestore(&iosapic_lock, flags);
-}
-
-static void 
-unmask_irq (unsigned int irq)
-{
-	unsigned long flags, iosapic_addr = iosapic_addr(irq);
-	u32 low32;
-
-	spin_lock_irqsave(&iosapic_lock, flags);
-	{
-		writel(IO_SAPIC_RTE_LOW(iosapic_pin(irq)), iosapic_addr + IO_SAPIC_REG_SELECT);
-		low32 = readl(iosapic_addr + IO_SAPIC_WINDOW);
-
-		low32 &= ~(1 << IO_SAPIC_MASK_SHIFT);    /* Zero only the mask bit */
-		writel(low32, iosapic_addr + IO_SAPIC_WINDOW);
-	}
-	spin_unlock_irqrestore(&iosapic_lock, flags);
-}
-
-
-static void
-iosapic_set_affinity (unsigned int irq, unsigned long mask)
-{
-	printk("iosapic_set_affinity: not implemented yet\n");
-}
-
-/*
- * Handlers for level-triggered interrupts.
- */
-
-static unsigned int
-iosapic_startup_level_irq (unsigned int irq)
-{
-	unmask_irq(irq);
-	return 0;
-}
-
-static void
-iosapic_end_level_irq (unsigned int irq)
-{
-	writel(irq, iosapic_addr(irq) + IO_SAPIC_EOI);
-}
-
-#define iosapic_shutdown_level_irq	mask_irq
-#define iosapic_enable_level_irq	unmask_irq
-#define iosapic_disable_level_irq	mask_irq
-#define iosapic_ack_level_irq		nop
-
-struct hw_interrupt_type irq_type_iosapic_level = {
-	typename:	"IO-SAPIC-level",
-	startup:	iosapic_startup_level_irq,
-	shutdown:	iosapic_shutdown_level_irq,
-	enable:		iosapic_enable_level_irq,
-	disable:	iosapic_disable_level_irq,
-	ack:		iosapic_ack_level_irq,
-	end:		iosapic_end_level_irq,
-	set_affinity:	iosapic_set_affinity
-};
-
-/*
- * Handlers for edge-triggered interrupts.
- */
-
-static unsigned int
-iosapic_startup_edge_irq (unsigned int irq)
-{
-	unmask_irq(irq);
-	/*
-	 * IOSAPIC simply drops interrupts pended while the
-	 * corresponding pin was masked, so we can't know if an
-	 * interrupt is pending already.  Let's hope not...
-	 */
-	return 0;
-}
-
-static void
-iosapic_ack_edge_irq (unsigned int irq)
-{
-	/*
-	 * Once we have recorded IRQ_PENDING already, we can mask the
-	 * interrupt for real. This prevents IRQ storms from unhandled
-	 * devices.
-	 */
-	if ((irq_desc[irq].status & (IRQ_PENDING | IRQ_DISABLED)) == (IRQ_PENDING | IRQ_DISABLED))
-		mask_irq(irq);
-}
-
-#define iosapic_enable_edge_irq		unmask_irq
-#define iosapic_disable_edge_irq	nop
-#define iosapic_end_edge_irq		nop
-
-struct hw_interrupt_type irq_type_iosapic_edge = {
-	typename:	"IO-SAPIC-edge",
-	startup:	iosapic_startup_edge_irq,
-	shutdown:	iosapic_disable_edge_irq,
-	enable:		iosapic_enable_edge_irq,
-	disable:	iosapic_disable_edge_irq,
-	ack:		iosapic_ack_edge_irq,
-	end:		iosapic_end_edge_irq,
-	set_affinity:	iosapic_set_affinity
-};
-
-unsigned int
-iosapic_version (unsigned long base_addr) 
-{
-	/*
-	 * IOSAPIC Version Register return 32 bit structure like:
-	 * {
-	 *	unsigned int version   : 8;
-	 *	unsigned int reserved1 : 8;
-	 *	unsigned int pins      : 8;
-	 *	unsigned int reserved2 : 8;
-	 * }
-	 */
-	writel(IO_SAPIC_VERSION, base_addr + IO_SAPIC_REG_SELECT);
-	return readl(IO_SAPIC_WINDOW + base_addr);
-}
-
-void
-iosapic_init (unsigned long address, int irqbase)
-{
-	struct hw_interrupt_type *irq_type;
-	struct pci_vector_struct *vectors;
-	int i, irq, num_pci_vectors;
-
-	if (irqbase == 0)
-		/* 
-		 * Map the legacy ISA devices into the IOSAPIC data.
-		 * Some of these may get reprogrammed later on with
-		 * data from the ACPI Interrupt Source Override table.
-		 */
-		for (i = 0; i < 16; i++) {
-			irq = isa_irq_to_vector(i);
-			iosapic_pin(irq) = i; 
-			iosapic_bus(irq) = BUS_ISA;
-			iosapic_busdata(irq) = 0;
-			iosapic_dmode(irq) = IO_SAPIC_LOWEST_PRIORITY;
-			iosapic_trigger(irq)  = IO_SAPIC_EDGE;
-			iosapic_polarity(irq) = IO_SAPIC_POL_HIGH;
-#ifdef DEBUG_IRQ_ROUTING
-			printk("ISA: IRQ %02x -> Vector %02x IOSAPIC Pin %d\n",
-			       i, irq, iosapic_pin(irq));
-#endif
-		}
-
-#ifndef CONFIG_IA64_SOFTSDV_HACKS
-	/* 
-	 * Map the PCI Interrupt data into the ACPI IOSAPIC data using
-	 * the info that the bootstrap loader passed to us.
-	 */
-# ifdef CONFIG_ACPI_KERNEL_CONFIG
-	acpi_cf_get_pci_vectors(&vectors, &num_pci_vectors);
-# else
-	ia64_boot_param.pci_vectors = (__u64) __va(ia64_boot_param.pci_vectors);
-	vectors = (struct pci_vector_struct *) ia64_boot_param.pci_vectors;
-	num_pci_vectors = ia64_boot_param.num_pci_vectors;
-# endif
-	for (i = 0; i < num_pci_vectors; i++) {
-		irq = vectors[i].irq;
-		if (irq < 16)
-			irq = isa_irq_to_vector(irq);
-		if (iosapic_baseirq(irq) != irqbase)
-			continue;
-
-		iosapic_bustype(irq) = BUS_PCI;
-		iosapic_pin(irq) = irq - iosapic_baseirq(irq);
-		iosapic_bus(irq) = vectors[i].bus;
-		/*
-		 * Map the PCI slot and pin data into iosapic_busdata()
-		 */
-		iosapic_busdata(irq) = (vectors[i].pci_id & 0xffff0000) | vectors[i].pin;
-
-		/* Default settings for PCI */
-		iosapic_dmode(irq) = IO_SAPIC_LOWEST_PRIORITY;
-		iosapic_trigger(irq)  = IO_SAPIC_LEVEL;
-		iosapic_polarity(irq) = IO_SAPIC_POL_LOW;
-
-# ifdef DEBUG_IRQ_ROUTING
-		printk("PCI: BUS %d Slot %x Pin %x IRQ %02x --> Vector %02x IOSAPIC Pin %d\n", 
-		       vectors[i].bus, vectors[i].pci_id>>16, vectors[i].pin, vectors[i].irq, 
-		       irq, iosapic_pin(irq));
-# endif
-	}
-#endif /* CONFIG_IA64_SOFTSDV_HACKS */
-
-	for (i = 0; i < NR_IRQS; ++i) {
-		if (iosapic_baseirq(i) != irqbase)
-			continue;
-
-		if (iosapic_pin(i) != -1) {
-			if (iosapic_trigger(i) == IO_SAPIC_LEVEL)
-			  irq_type = &irq_type_iosapic_level;
-			else
-			  irq_type = &irq_type_iosapic_edge;
-			if (irq_desc[i].handler != &no_irq_type)
-				printk("dig_irq_init: warning: changing vector %d from %s to %s\n",
-				       i, irq_desc[i].handler->typename,
-				       irq_type->typename);
-			irq_desc[i].handler = irq_type;
-
-			/* program the IOSAPIC routing table: */
-			set_rte(iosapic_addr(i), iosapic_pin(i), iosapic_polarity(i),
-				iosapic_trigger(i), iosapic_dmode(i),
-				(ia64_get_lid() >> 16) & 0xffff, i);
-		}
-	}
-}
-
-void
-dig_irq_init (void)
-{
-	/*
-	 * Disable the compatibility mode interrupts (8259 style), needs IN/OUT support
-	 * enabled.
-	 */
-	outb(0xff, 0xA1);
-	outb(0xff, 0x21);
-}
-
-void
-dig_pci_fixup (void)
-{
-	struct	pci_dev	*dev;
-	int		irq;
-	unsigned char 	pin;
-
-	pci_for_each_dev(dev) {
-		pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
-		if (pin) {
-			pin--;          /* interrupt pins are numbered starting from 1 */
-			irq = iosapic_get_PCI_irq_vector(dev->bus->number, PCI_SLOT(dev->devfn),
-							 pin);
-			if (irq < 0 && dev->bus->parent) { /* go back to the bridge */
-				struct pci_dev * bridge = dev->bus->self;
-
-				/* allow for multiple bridges on an adapter */
-				do {
-					/* do the bridge swizzle... */
-					pin = (pin + PCI_SLOT(dev->devfn)) % 4;
-					irq = iosapic_get_PCI_irq_vector(bridge->bus->number,
-									 PCI_SLOT(bridge->devfn), pin);
-				} while (irq < 0 && (bridge = bridge->bus->self));
-				if (irq >= 0)
-					printk(KERN_WARNING
-					       "PCI: using PPB(B%d,I%d,P%d) to get irq %02x\n",
-					       bridge->bus->number, PCI_SLOT(bridge->devfn),
-					       pin, irq);
-				else
-					printk(KERN_WARNING
-					       "PCI: Couldn't map irq for B%d,I%d,P%d\n",
-					       bridge->bus->number, PCI_SLOT(bridge->devfn),
-					       pin);
-			}
-			if (irq >= 0) {
-				printk("PCI->APIC IRQ transform: (B%d,I%d,P%d) -> %02x\n",
-				       dev->bus->number, PCI_SLOT(dev->devfn), pin, irq);
-				dev->irq = irq;
-			}
-		}
-		/*
-		 * Nothing to fixup
-		 * Fix out-of-range IRQ numbers
-		 */
-		if (dev->irq >= NR_IRQS)
-			dev->irq = 15;	/* Spurious interrupts */
-	}
-}
-
-/*
- * Register an IOSAPIC discovered via ACPI.
- */
-void __init
-dig_register_iosapic (acpi_entry_iosapic_t *iosapic)
-{
-	unsigned int ver, v;
-	int l, max_pin;
-
-	ver = iosapic_version((unsigned long) ioremap(iosapic->address, 0));
-	max_pin = (ver >> 16) & 0xff;
-	
-	printk("IOSAPIC Version %x.%x: address 0x%lx IRQs 0x%x - 0x%x\n", 
-	       (ver & 0xf0) >> 4, (ver & 0x0f), iosapic->address, 
-	       iosapic->irq_base, iosapic->irq_base + max_pin);
-	
-	for (l = 0; l <= max_pin; l++) {
-		v = iosapic->irq_base + l;
-		if (v < 16)
-			v = isa_irq_to_vector(v);
-		if (v > IA64_MAX_VECTORED_IRQ) {
-			printk("    !!! bad IOSAPIC interrupt vector: %u\n", v);
-			continue;
-		}
-		/* XXX Check for IOSAPIC collisions */
-		iosapic_addr(v) = (unsigned long) ioremap(iosapic->address, 0);
-		iosapic_baseirq(v) = iosapic->irq_base;
-	}
-	iosapic_init(iosapic->address, iosapic->irq_base);
-}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/dig/setup.c linux/arch/ia64/dig/setup.c
--- v2.4.0-prerelease/linux/arch/ia64/dig/setup.c	Fri Aug 11 19:09:06 2000
+++ linux/arch/ia64/dig/setup.c	Thu Jan  4 12:50:17 2001
@@ -84,3 +84,14 @@
 	screen_info.orig_video_isVGA = 1;	/* XXX fake */
 	screen_info.orig_video_ega_bx = 3;	/* XXX fake */
 }
+
+void
+dig_irq_init (void)
+{
+	/*
+	 * Disable the compatibility mode interrupts (8259 style), needs IN/OUT support
+	 * enabled.
+	 */
+	outb(0xff, 0xA1);
+	outb(0xff, 0x21);
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/hp/Makefile linux/arch/ia64/hp/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/hp/Makefile	Fri Apr 21 15:21:23 2000
+++ linux/arch/ia64/hp/Makefile	Thu Jan  4 12:50:17 2001
@@ -7,12 +7,10 @@
 
 all: hp.a
 
-O_TARGET	= hp.a
-O_OBJS		= hpsim_console.o hpsim_irq.o hpsim_setup.o
+O_TARGET := hp.a
 
-ifdef CONFIG_IA64_GENERIC
-O_OBJS		+= hpsim_machvec.o
-endif
+obj-y := hpsim_console.o hpsim_irq.o hpsim_setup.o
+obj-$(CONFIG_IA64_GENERIC) += hpsim_machvec.o
 
 clean::
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/hp/hpsim_setup.c linux/arch/ia64/hp/hpsim_setup.c
--- v2.4.0-prerelease/linux/arch/ia64/hp/hpsim_setup.c	Fri Jul 14 16:08:11 2000
+++ linux/arch/ia64/hp/hpsim_setup.c	Thu Jan  4 12:50:17 2001
@@ -63,12 +63,6 @@
 }
 
 void __init
-hpsim_pci_fixup (void)
-{
-}
-
-
-void __init
 hpsim_setup (char **cmdline_p)
 {
 	ROOT_DEV = to_kdev_t(0x0801);		/* default to first SCSI drive */
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/ia32/Makefile linux/arch/ia64/ia32/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/ia32/Makefile	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/ia32/Makefile	Thu Jan  4 12:50:17 2001
@@ -10,7 +10,8 @@
 all: ia32.o
 
 O_TARGET := ia32.o
-O_OBJS	 := ia32_entry.o sys_ia32.o ia32_ioctl.o ia32_signal.o ia32_support.o ia32_traps.o binfmt_elf32.o
+
+obj-y := ia32_entry.o sys_ia32.o ia32_ioctl.o ia32_signal.o ia32_support.o ia32_traps.o binfmt_elf32.o
 
 clean::
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/ia32/binfmt_elf32.c linux/arch/ia64/ia32/binfmt_elf32.c
--- v2.4.0-prerelease/linux/arch/ia64/ia32/binfmt_elf32.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/ia32/binfmt_elf32.c	Thu Jan  4 12:50:17 2001
@@ -9,6 +9,7 @@
 
 #include <linux/types.h>
 
+#include <asm/param.h>
 #include <asm/signal.h>
 #include <asm/ia32.h>
 
@@ -31,6 +32,9 @@
 # define CONFIG_BINFMT_ELF_MODULE	CONFIG_BINFMT_ELF32_MODULE
 #endif
 
+#undef CLOCKS_PER_SEC
+#define CLOCKS_PER_SEC	IA32_CLOCKS_PER_SEC
+
 extern void ia64_elf32_init(struct pt_regs *regs);
 extern void put_dirty_page(struct task_struct * tsk, struct page *page, unsigned long address);
 
@@ -89,8 +93,8 @@
 	
 	/* Do all the IA-32 setup here */
 
-	current->thread.map_base = 0x40000000;
-
+	current->thread.map_base  =  0x40000000;
+	current->thread.task_size =  0xc0000000;	/* use what Linux/x86 uses... */
  
 	/* setup ia32 state for ia32_load_state */
 
@@ -239,6 +243,12 @@
 	if (eppnt->p_memsz >= (1UL<<32) || addr > (1UL<<32) - eppnt->p_memsz)
 		return -EINVAL;
 
+	/*
+	 *  Make sure the elf interpreter doesn't get loaded at location 0
+	 *    so that NULL pointers correctly cause segfaults.
+	 */
+	if (addr == 0)
+		addr += PAGE_SIZE;
 #if 1
 	set_brk(ia32_mm_addr(addr), addr + eppnt->p_memsz);
 	memset((char *) addr + eppnt->p_filesz, 0, eppnt->p_memsz - eppnt->p_filesz);
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/ia32/ia32_entry.S linux/arch/ia64/ia32/ia32_entry.S
--- v2.4.0-prerelease/linux/arch/ia64/ia32/ia32_entry.S	Tue Sep  5 13:50:01 2000
+++ linux/arch/ia64/ia32/ia32_entry.S	Thu Jan  4 12:50:17 2001
@@ -133,7 +133,7 @@
 	data8 sys32_ni_syscall /* sys_stime is not supported on IA64 */  /* 25 */
 	data8 sys32_ptrace
 	data8 sys32_alarm
-	data8 sys32_ni_syscall
+	data8 sys_pause
 	data8 sys32_ni_syscall
 	data8 ia32_utime	  /* 30 */
 	data8 sys32_ni_syscall	  /* old stty syscall holder */
@@ -291,11 +291,43 @@
 	data8 sys_getcwd
 	data8 sys_capget
 	data8 sys_capset	  /* 185 */
-	data8 sys_sigaltstack
+	data8 sys32_sigaltstack
 	data8 sys_sendfile
 	data8 sys32_ni_syscall		  /* streams1 */
 	data8 sys32_ni_syscall		  /* streams2 */
 	data8 sys32_vfork	  /* 190 */
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall	  /* 195 */
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall	  /* 200 */
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall	  /* 205 */
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall	  /* 210 */
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall	  /* 215 */
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall	  /* 220 */
+	data8 sys_ni_syscall
+	data8 sys_ni_syscall
 	/*
 	 *  CAUTION: If any system calls are added beyond this point
 	 *	then the check in `arch/ia64/kernel/ivt.S' will have
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/ia32/ia32_ioctl.c linux/arch/ia64/ia32/ia32_ioctl.c
--- v2.4.0-prerelease/linux/arch/ia64/ia32/ia32_ioctl.c	Fri Jul 14 16:08:11 2000
+++ linux/arch/ia64/ia32/ia32_ioctl.c	Thu Jan  4 12:50:17 2001
@@ -22,81 +22,193 @@
 #include <linux/if_ppp.h>
 #include <linux/ixjuser.h>
 #include <linux/i2o-dev.h>
+#include <../drivers/char/drm/drm.h>
+
+#define IOCTL_NR(a)	((a) & ~(_IOC_SIZEMASK << _IOC_SIZESHIFT))
+
+#define DO_IOCTL(fd, cmd, arg) ({			\
+	int _ret;					\
+	mm_segment_t _old_fs = get_fs();		\
+							\
+	set_fs(KERNEL_DS);				\
+	_ret = sys_ioctl(fd, cmd, (unsigned long)arg);	\
+	set_fs(_old_fs);				\
+	_ret;						\
+})
+
+#define P(i)	((void *)(long)(i))
+
 
 asmlinkage long sys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
 
 asmlinkage long ia32_ioctl(unsigned int fd, unsigned int cmd, unsigned int arg)
 {
+	long ret;
+
+	switch (IOCTL_NR(cmd)) {
+
+	case IOCTL_NR(DRM_IOCTL_VERSION):
+		{
+			drm_version_t ver;
+			struct {
+				int	version_major;
+				int	version_minor;
+				int	version_patchlevel;
+				unsigned int name_len;
+				unsigned int name; /* pointer */
+				unsigned int date_len;
+				unsigned int date; /* pointer */
+				unsigned int desc_len;
+				unsigned int desc; /* pointer */
+			} ver32;
+
+			if (copy_from_user(&ver32, P(arg), sizeof(ver32)))
+				return -EFAULT;
+			ver.name_len = ver32.name_len;
+			ver.name = P(ver32.name);
+			ver.date_len = ver32.date_len;
+			ver.date = P(ver32.date);
+			ver.desc_len = ver32.desc_len;
+			ver.desc = P(ver32.desc);
+			ret = DO_IOCTL(fd, cmd, &ver);
+			if (ret >= 0) {
+				ver32.version_major = ver.version_major;
+				ver32.version_minor = ver.version_minor;
+				ver32.version_patchlevel = ver.version_patchlevel;
+				ver32.name_len = ver.name_len;
+				ver32.date_len = ver.date_len;
+				ver32.desc_len = ver.desc_len;
+				if (copy_to_user(P(arg), &ver32, sizeof(ver32)))
+					return -EFAULT;
+			}
+			return(ret);
+		}
+
+	case IOCTL_NR(DRM_IOCTL_GET_UNIQUE):
+		{
+			drm_unique_t un;
+			struct {
+				unsigned int unique_len;
+				unsigned int unique;
+			} un32;
+
+			if (copy_from_user(&un32, P(arg), sizeof(un32)))
+				return -EFAULT;
+			un.unique_len = un32.unique_len;
+			un.unique = P(un32.unique);
+			ret = DO_IOCTL(fd, cmd, &un);
+			if (ret >= 0) {
+				un32.unique_len = un.unique_len;
+				if (copy_to_user(P(arg), &un32, sizeof(un32)))
+					return -EFAULT;
+			}
+			return(ret);
+		}
+	case IOCTL_NR(DRM_IOCTL_SET_UNIQUE):
+	case IOCTL_NR(DRM_IOCTL_ADD_MAP):
+	case IOCTL_NR(DRM_IOCTL_ADD_BUFS):
+	case IOCTL_NR(DRM_IOCTL_MARK_BUFS):
+	case IOCTL_NR(DRM_IOCTL_INFO_BUFS):
+	case IOCTL_NR(DRM_IOCTL_MAP_BUFS):
+	case IOCTL_NR(DRM_IOCTL_FREE_BUFS):
+	case IOCTL_NR(DRM_IOCTL_ADD_CTX):
+	case IOCTL_NR(DRM_IOCTL_RM_CTX):
+	case IOCTL_NR(DRM_IOCTL_MOD_CTX):
+	case IOCTL_NR(DRM_IOCTL_GET_CTX):
+	case IOCTL_NR(DRM_IOCTL_SWITCH_CTX):
+	case IOCTL_NR(DRM_IOCTL_NEW_CTX):
+	case IOCTL_NR(DRM_IOCTL_RES_CTX):
+
+	case IOCTL_NR(DRM_IOCTL_AGP_ACQUIRE):
+	case IOCTL_NR(DRM_IOCTL_AGP_RELEASE):
+	case IOCTL_NR(DRM_IOCTL_AGP_ENABLE):
+	case IOCTL_NR(DRM_IOCTL_AGP_INFO):
+	case IOCTL_NR(DRM_IOCTL_AGP_ALLOC):
+	case IOCTL_NR(DRM_IOCTL_AGP_FREE):
+	case IOCTL_NR(DRM_IOCTL_AGP_BIND):
+	case IOCTL_NR(DRM_IOCTL_AGP_UNBIND):
+
+	/* Mga specific ioctls */
+
+	case IOCTL_NR(DRM_IOCTL_MGA_INIT):
+
+	/* I810 specific ioctls */
+
+	case IOCTL_NR(DRM_IOCTL_I810_GETBUF):
+	case IOCTL_NR(DRM_IOCTL_I810_COPY):
+
+	/* Rage 128 specific ioctls */
 
-	switch (cmd) {
+	case IOCTL_NR(DRM_IOCTL_R128_PACKET):
 
-	case VFAT_IOCTL_READDIR_BOTH:
-	case VFAT_IOCTL_READDIR_SHORT:
-	case MTIOCGET:
-	case MTIOCPOS:
-	case MTIOCGETCONFIG:
-	case MTIOCSETCONFIG:
-	case PPPIOCSCOMPRESS:
-	case PPPIOCGIDLE:
-	case NCP_IOC_GET_FS_INFO_V2:
-	case NCP_IOC_GETOBJECTNAME:
-	case NCP_IOC_SETOBJECTNAME:
-	case NCP_IOC_GETPRIVATEDATA:
-	case NCP_IOC_SETPRIVATEDATA:
-	case NCP_IOC_GETMOUNTUID2:
-	case CAPI_MANUFACTURER_CMD:
-	case VIDIOCGTUNER:
-	case VIDIOCSTUNER:
-	case VIDIOCGWIN:
-	case VIDIOCSWIN:
-	case VIDIOCGFBUF:
-	case VIDIOCSFBUF:
-	case MGSL_IOCSPARAMS:
-	case MGSL_IOCGPARAMS:
-	case ATM_GETNAMES:
-	case ATM_GETLINKRATE:
-	case ATM_GETTYPE:
-	case ATM_GETESI:
-	case ATM_GETADDR:
-	case ATM_RSTADDR:
-	case ATM_ADDADDR:
-	case ATM_DELADDR:
-	case ATM_GETCIRANGE:
-	case ATM_SETCIRANGE:
-	case ATM_SETESI:
-	case ATM_SETESIF:
-	case ATM_GETSTAT:
-	case ATM_GETSTATZ:
-	case ATM_GETLOOP:
-	case ATM_SETLOOP:
-	case ATM_QUERYLOOP:
-	case ENI_SETMULT:
-	case NS_GETPSTAT:
-	/* case NS_SETBUFLEV: This is a duplicate case with ZATM_GETPOOLZ */
-	case ZATM_GETPOOLZ:
-	case ZATM_GETPOOL:
-	case ZATM_SETPOOL:
-	case ZATM_GETTHIST:
-	case IDT77105_GETSTAT:
-	case IDT77105_GETSTATZ:
-	case IXJCTL_TONE_CADENCE:
-	case IXJCTL_FRAMES_READ:
-	case IXJCTL_FRAMES_WRITTEN:
-	case IXJCTL_READ_WAIT:
-	case IXJCTL_WRITE_WAIT:
-	case IXJCTL_DRYBUFFER_READ:
-	case I2OHRTGET:
-	case I2OLCTGET:
-	case I2OPARMSET:
-	case I2OPARMGET:
-	case I2OSWDL:
-	case I2OSWUL:
-	case I2OSWDEL:
-	case I2OHTML:
-		printk("%x:unimplemented IA32 ioctl system call\n", cmd);
-		return(-EINVAL);
+	case IOCTL_NR(VFAT_IOCTL_READDIR_BOTH):
+	case IOCTL_NR(VFAT_IOCTL_READDIR_SHORT):
+	case IOCTL_NR(MTIOCGET):
+	case IOCTL_NR(MTIOCPOS):
+	case IOCTL_NR(MTIOCGETCONFIG):
+	case IOCTL_NR(MTIOCSETCONFIG):
+	case IOCTL_NR(PPPIOCSCOMPRESS):
+	case IOCTL_NR(PPPIOCGIDLE):
+	case IOCTL_NR(NCP_IOC_GET_FS_INFO_V2):
+	case IOCTL_NR(NCP_IOC_GETOBJECTNAME):
+	case IOCTL_NR(NCP_IOC_SETOBJECTNAME):
+	case IOCTL_NR(NCP_IOC_GETPRIVATEDATA):
+	case IOCTL_NR(NCP_IOC_SETPRIVATEDATA):
+	case IOCTL_NR(NCP_IOC_GETMOUNTUID2):
+	case IOCTL_NR(CAPI_MANUFACTURER_CMD):
+	case IOCTL_NR(VIDIOCGTUNER):
+	case IOCTL_NR(VIDIOCSTUNER):
+	case IOCTL_NR(VIDIOCGWIN):
+	case IOCTL_NR(VIDIOCSWIN):
+	case IOCTL_NR(VIDIOCGFBUF):
+	case IOCTL_NR(VIDIOCSFBUF):
+	case IOCTL_NR(MGSL_IOCSPARAMS):
+	case IOCTL_NR(MGSL_IOCGPARAMS):
+	case IOCTL_NR(ATM_GETNAMES):
+	case IOCTL_NR(ATM_GETLINKRATE):
+	case IOCTL_NR(ATM_GETTYPE):
+	case IOCTL_NR(ATM_GETESI):
+	case IOCTL_NR(ATM_GETADDR):
+	case IOCTL_NR(ATM_RSTADDR):
+	case IOCTL_NR(ATM_ADDADDR):
+	case IOCTL_NR(ATM_DELADDR):
+	case IOCTL_NR(ATM_GETCIRANGE):
+	case IOCTL_NR(ATM_SETCIRANGE):
+	case IOCTL_NR(ATM_SETESI):
+	case IOCTL_NR(ATM_SETESIF):
+	case IOCTL_NR(ATM_GETSTAT):
+	case IOCTL_NR(ATM_GETSTATZ):
+	case IOCTL_NR(ATM_GETLOOP):
+	case IOCTL_NR(ATM_SETLOOP):
+	case IOCTL_NR(ATM_QUERYLOOP):
+	case IOCTL_NR(ENI_SETMULT):
+	case IOCTL_NR(NS_GETPSTAT):
+	/* case IOCTL_NR(NS_SETBUFLEV): This is a duplicate case with ZATM_GETPOOLZ */
+	case IOCTL_NR(ZATM_GETPOOLZ):
+	case IOCTL_NR(ZATM_GETPOOL):
+	case IOCTL_NR(ZATM_SETPOOL):
+	case IOCTL_NR(ZATM_GETTHIST):
+	case IOCTL_NR(IDT77105_GETSTAT):
+	case IOCTL_NR(IDT77105_GETSTATZ):
+	case IOCTL_NR(IXJCTL_TONE_CADENCE):
+	case IOCTL_NR(IXJCTL_FRAMES_READ):
+	case IOCTL_NR(IXJCTL_FRAMES_WRITTEN):
+	case IOCTL_NR(IXJCTL_READ_WAIT):
+	case IOCTL_NR(IXJCTL_WRITE_WAIT):
+	case IOCTL_NR(IXJCTL_DRYBUFFER_READ):
+	case IOCTL_NR(I2OHRTGET):
+	case IOCTL_NR(I2OLCTGET):
+	case IOCTL_NR(I2OPARMSET):
+	case IOCTL_NR(I2OPARMGET):
+	case IOCTL_NR(I2OSWDL):
+	case IOCTL_NR(I2OSWUL):
+	case IOCTL_NR(I2OSWDEL):
+	case IOCTL_NR(I2OHTML):
+		break;
 	default:
 		return(sys_ioctl(fd, cmd, (unsigned long)arg));
 
 	}
+	printk("%x:unimplemented IA32 ioctl system call\n", cmd);
+	return(-EINVAL);
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/ia32/ia32_traps.c linux/arch/ia64/ia32/ia32_traps.c
--- v2.4.0-prerelease/linux/arch/ia64/ia32/ia32_traps.c	Thu Jun 22 07:09:44 2000
+++ linux/arch/ia64/ia32/ia32_traps.c	Thu Jan  4 12:50:17 2001
@@ -119,6 +119,6 @@
 	      default:
 		return -1;
 	}
-	force_sig_info(SIGTRAP, &siginfo, current);
+	force_sig_info(siginfo.si_signo, &siginfo, current);
 	return 0;
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/ia32/sys_ia32.c linux/arch/ia64/ia32/sys_ia32.c
--- v2.4.0-prerelease/linux/arch/ia64/ia32/sys_ia32.c	Sun Nov 19 18:44:03 2000
+++ linux/arch/ia64/ia32/sys_ia32.c	Thu Jan  4 12:50:17 2001
@@ -236,8 +236,6 @@
 
 	if (OFFSET4K(addr) || OFFSET4K(off))
 		return -EINVAL;
-	if (prot & PROT_WRITE)
-		prot |= PROT_EXEC;
 	prot |= PROT_WRITE;
 	front = NULL;
 	back = NULL;
@@ -287,23 +285,20 @@
 	unsigned int poff;
 
 	flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
+	prot |= PROT_EXEC;
 
  	if ((flags & MAP_FIXED) && ((addr & ~PAGE_MASK) || (offset & ~PAGE_MASK)))
  		error = do_mmap_fake(file, addr, len, prot, flags, (loff_t)offset);
- 	else if (!addr && (offset & ~PAGE_MASK)) {
+	else {
  		poff = offset & PAGE_MASK;
  		len += offset - poff;
 
  		down(&current->mm->mmap_sem);
- 		error = do_mmap(file, addr, len, prot, flags, poff);
+ 		error = do_mmap_pgoff(file, addr, len, prot, flags, poff >> PAGE_SHIFT);
   		up(&current->mm->mmap_sem);
 
  		if (!IS_ERR((void *) error))
  			error += offset - poff;
- 	} else {
-  		down(&current->mm->mmap_sem);
-  		error = do_mmap(file, addr, len, prot, flags, offset);
- 		up(&current->mm->mmap_sem);
  	}
 	return error;
 }
@@ -2032,14 +2027,14 @@
 	ret = sys_times(tbuf ? &t : NULL);
 	set_fs (old_fs);
 	if (tbuf) {
-		err = put_user (t.tms_utime, &tbuf->tms_utime);
-		err |= __put_user (t.tms_stime, &tbuf->tms_stime);
-		err |= __put_user (t.tms_cutime, &tbuf->tms_cutime);
-		err |= __put_user (t.tms_cstime, &tbuf->tms_cstime);
+		err = put_user (IA32_TICK(t.tms_utime), &tbuf->tms_utime);
+		err |= __put_user (IA32_TICK(t.tms_stime), &tbuf->tms_stime);
+		err |= __put_user (IA32_TICK(t.tms_cutime), &tbuf->tms_cutime);
+		err |= __put_user (IA32_TICK(t.tms_cstime), &tbuf->tms_cstime);
 		if (err)
 			ret = -EFAULT;
 	}
-	return ret;
+	return IA32_TICK(ret);
 }
 
 unsigned int
@@ -2617,6 +2612,53 @@
 	 *	manipulating the page protections...
 	 */
 	return(sys_iopl(3, 0, 0, 0));
+}
+
+typedef struct {
+	unsigned int	ss_sp;
+	unsigned int	ss_flags;
+	unsigned int	ss_size;
+} ia32_stack_t;
+
+asmlinkage long
+sys32_sigaltstack (const ia32_stack_t *uss32, ia32_stack_t *uoss32,
+long arg2, long arg3, long arg4,
+long arg5, long arg6, long arg7,
+long stack)
+{
+	struct pt_regs *pt = (struct pt_regs *) &stack;
+	stack_t uss, uoss;
+	ia32_stack_t buf32;
+	int ret;
+	mm_segment_t old_fs = get_fs();
+
+	if (uss32)
+		if (copy_from_user(&buf32, (void *)A(uss32), sizeof(ia32_stack_t)))
+			return(-EFAULT);
+	uss.ss_sp = (void *) (long) buf32.ss_sp;
+	uss.ss_flags = buf32.ss_flags;
+	uss.ss_size = buf32.ss_size;
+	set_fs(KERNEL_DS);
+	ret = do_sigaltstack(uss32 ? &uss : NULL, &uoss, pt->r12);
+	set_fs(old_fs);
+	if (ret < 0)
+		return(ret);
+	if (uoss32) {
+		buf32.ss_sp = (long) uoss.ss_sp;
+		buf32.ss_flags = uoss.ss_flags;
+		buf32.ss_size = uoss.ss_size;
+		if (copy_to_user((void*)A(uoss32), &buf32, sizeof(ia32_stack_t)))
+			return(-EFAULT);
+	}
+	return(ret);
+}
+
+asmlinkage int
+sys_pause (void)
+{
+	current->state = TASK_INTERRUPTIBLE;
+	schedule();
+	return -ERESTARTNOHAND;
 }
 
 #ifdef	NOTYET  /* UNTESTED FOR IA64 FROM HERE DOWN */
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/Makefile linux/arch/ia64/kernel/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/kernel/Makefile	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/Makefile	Thu Jan  4 12:50:17 2001
@@ -9,20 +9,20 @@
 
 all: kernel.o head.o init_task.o
 
-obj-y := acpi.o entry.o gate.o efi.o efi_stub.o irq.o irq_ia64.o irq_sapic.o ivt.o		\
-	 machvec.o pal.o pci-dma.o process.o perfmon.o ptrace.o sal.o semaphore.o setup.o	\
-	 signal.o sys_ia64.o traps.o time.o unaligned.o unwind.o
+O_TARGET := kernel.o
 
-obj-$(CONFIG_IA64_GENERIC) += machvec.o
+obj-y := acpi.o entry.o gate.o efi.o efi_stub.o irq.o irq_ia64.o irq_sapic.o ivt.o \
+	 machvec.o pal.o process.o perfmon.o ptrace.o sal.o semaphore.o setup.o	\
+	 signal.o sys_ia64.o traps.o time.o unaligned.o unwind.o
+obj-$(CONFIG_IA64_GENERIC) += machvec.o iosapic.o
+obj-$(CONFIG_IA64_DIG) += iosapic.o
 obj-$(CONFIG_IA64_PALINFO) += palinfo.o
 obj-$(CONFIG_PCI) += pci.o
 obj-$(CONFIG_SMP) += smp.o smpboot.o
 obj-$(CONFIG_IA64_MCA) += mca.o mca_asm.o
 obj-$(CONFIG_IA64_BRL_EMU) += brl_emu.o
 
-O_TARGET := kernel.o
-O_OBJS	 :=  $(obj-y)
-OX_OBJS  := ia64_ksyms.o
+export-objs := ia64_ksyms.o
 
 clean::
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/acpi.c linux/arch/ia64/kernel/acpi.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/acpi.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/acpi.c	Thu Jan  4 12:50:17 2001
@@ -6,6 +6,12 @@
  * 
  * Copyright (C) 1999 VA Linux Systems
  * Copyright (C) 1999,2000 Walt Drummond <drummond@valinux.com>
+ * Copyright (C) 2000 Hewlett-Packard Co.
+ * Copyright (C) 2000 David Mosberger-Tang <davidm@hpl.hp.com>
+ * Copyright (C) 2000 Intel Corp.
+ * Copyright (C) 2000 J.I. Lee <jung-ik.lee@intel.com>
+ *      ACPI based kernel configuration manager.
+ *      ACPI 2.0 & IA64 ext 0.71
  */
 
 #include <linux/config.h>
@@ -36,29 +42,87 @@
 
 void (*pm_idle)(void);
 
+asm (".weak iosapic_register_legacy_irq");
+asm (".weak iosapic_init");
+
+const char *
+acpi_get_sysname (void)
+{
+	/* the following should go away once we have an ACPI parser: */
+#ifdef CONFIG_IA64_GENERIC
+	return "hpsim";
+#else
+# if defined (CONFIG_IA64_HP_SIM)
+	return "hpsim";
+# elif defined (CONFIG_IA64_SGI_SN1)
+	return "sn1";
+# elif defined (CONFIG_IA64_DIG)
+	return "dig";
+# else
+#	error Unknown platform.  Fix acpi.c.
+# endif
+#endif
+
+}
+
 /*
- * Identify usable CPU's and remember them for SMP bringup later.
+ * Configure legacy IRQ information.
  */
 static void __init
-acpi_lsapic(char *p) 
+acpi_legacy_irq (char *p)
 {
-	int add = 1;
-
-	acpi_entry_lsapic_t *lsapic = (acpi_entry_lsapic_t *) p;
+	acpi_entry_int_override_t *legacy = (acpi_entry_int_override_t *) p;
+	unsigned long polarity = 0, edge_triggered = 0;
 
-	if ((lsapic->flags & LSAPIC_PRESENT) == 0) 
+	/*
+	 * If the platform we're running doesn't define
+	 * iosapic_register_legacy_irq(), we ignore this info...
+	 */
+	if (!iosapic_register_legacy_irq)
 		return;
 
+	switch (legacy->flags) {
+	      case 0x5:	polarity = 1; edge_triggered = 1; break;
+	      case 0x7: polarity = 0; edge_triggered = 1; break;
+	      case 0xd: polarity = 1; edge_triggered = 0; break;
+	      case 0xf: polarity = 0; edge_triggered = 0; break;
+	      default:
+		printk("    ACPI Legacy IRQ 0x%02x: Unknown flags 0x%x\n", legacy->isa_irq,
+		       legacy->flags);
+		break;
+	}
+	iosapic_register_legacy_irq(legacy->isa_irq, legacy->pin, polarity, edge_triggered);
+}
+
+/*
+ * ACPI 2.0 tables parsing functions
+ */
+
+static unsigned long
+readl_unaligned(void *p)
+{
+	unsigned long ret;
+
+	memcpy(&ret, p, sizeof(long));
+	return ret;
+}
+
+/*
+ * Identify usable CPU's and remember them for SMP bringup later.
+ */
+static void __init
+acpi20_lsapic (char *p) 
+{
+	int add = 1;
+
+	acpi20_entry_lsapic_t *lsapic = (acpi20_entry_lsapic_t *) p;
 	printk("      CPU %d (%.04x:%.04x): ", total_cpus, lsapic->eid, lsapic->id);
 
 	if ((lsapic->flags & LSAPIC_ENABLED) == 0) {
 		printk("Disabled.\n");
 		add = 0;
-	} else if (lsapic->flags & LSAPIC_PERFORMANCE_RESTRICTED) {
-		printk("Performance Restricted; ignoring.\n");
-		add = 0;
 	}
-	
+
 #ifdef CONFIG_SMP
 	smp_boot_data.cpu_phys_id[total_cpus] = -1;
 #endif
@@ -73,87 +137,234 @@
 }
 
 /*
- * Configure legacy IRQ information in iosapic_vector
+ * Info on platform interrupt sources: NMI. PMI, INIT, etc.
  */
 static void __init
-acpi_legacy_irq(char *p)
+acpi20_platform (char *p)
 {
-	/*
-	 * This is not good.  ACPI is not necessarily limited to CONFIG_IA64_DIG, yet
-	 * ACPI does not necessarily imply IOSAPIC either.  Perhaps there should be
-	 * a means for platform_setup() to register ACPI handlers?
-	 */
-#ifdef CONFIG_IA64_IRQ_ACPI
-	acpi_entry_int_override_t *legacy = (acpi_entry_int_override_t *) p;
-	unsigned char vector; 
-	int i;
+	acpi20_entry_platform_src_t *plat = (acpi20_entry_platform_src_t *) p;
+
+	printk("PLATFORM: IOSAPIC %x -> Vector %x on CPU %.04u:%.04u\n",
+	       plat->iosapic_vector, plat->global_vector, plat->eid, plat->id);
+}
 
-	vector = isa_irq_to_vector(legacy->isa_irq);
+/*
+ * Override the physical address of the local APIC in the MADT stable header.
+ */
+static void __init
+acpi20_lapic_addr_override (char *p)
+{
+	acpi20_entry_lapic_addr_override_t * lapic = (acpi20_entry_lapic_addr_override_t *) p;
+
+	if (lapic->lapic_address) {
+		iounmap((void *)ipi_base_addr);
+		ipi_base_addr = (unsigned long) ioremap(lapic->lapic_address, 0);
+
+		printk("LOCAL ACPI override to 0x%lx(p=0x%lx)\n",
+		       ipi_base_addr, lapic->lapic_address);
+	}
+}
+
+/*
+ * Parse the ACPI Multiple APIC Description Table
+ */
+static void __init
+acpi20_parse_madt (acpi_madt_t *madt)
+{
+	acpi_entry_iosapic_t *iosapic;
+	char *p, *end;
+
+	/* Base address of IPI Message Block */
+	if (madt->lapic_address) {
+		ipi_base_addr = (unsigned long) ioremap(madt->lapic_address, 0);
+		printk("Lapic address set to 0x%lx\n", ipi_base_addr);
+	} else
+		printk("Lapic address set to default 0x%lx\n", ipi_base_addr);
+
+	p = (char *) (madt + 1);
+	end = p + (madt->header.length - sizeof(acpi_madt_t));
 
 	/*
-	 * Clobber any old pin mapping.  It may be that it gets replaced later on
-	 */
-	for (i = 0; i < IA64_MAX_VECTORED_IRQ; i++) {
-		if (i == vector) 
-			continue;
-		if (iosapic_pin(i) == iosapic_pin(vector))
-			iosapic_pin(i) = 0xff;
-        }
-
-	iosapic_pin(vector) = legacy->pin;
-	iosapic_bus(vector) = BUS_ISA;	/* This table only overrides the ISA devices */
-	iosapic_busdata(vector) = 0;
-	
-	/* 
-	 * External timer tick is special... 
+	 * Splitted entry parsing to ensure ordering.
 	 */
-	if (vector != TIMER_IRQ)
-		iosapic_dmode(vector) = IO_SAPIC_LOWEST_PRIORITY;
-	else 
-		iosapic_dmode(vector) = IO_SAPIC_FIXED;
+
+	while (p < end) {
+		switch (*p) {
+		case ACPI20_ENTRY_LOCAL_APIC_ADDR_OVERRIDE:
+			printk("ACPI 2.0 MADT: LOCAL APIC Override\n");
+			acpi20_lapic_addr_override(p);
+			break;
+
+		case ACPI20_ENTRY_LOCAL_SAPIC:
+			printk("ACPI 2.0 MADT: LOCAL SAPIC\n");
+			acpi20_lsapic(p);
+			break;
 	
-	/* See MPS 1.4 section 4.3.4 */
-	switch (legacy->flags) {
-	case 0x5:
-		iosapic_polarity(vector) = IO_SAPIC_POL_HIGH;
-		iosapic_trigger(vector) = IO_SAPIC_EDGE;
-		break;
-	case 0x8:
-		iosapic_polarity(vector) = IO_SAPIC_POL_LOW;
-		iosapic_trigger(vector) = IO_SAPIC_EDGE;
-		break;
-	case 0xd:
-		iosapic_polarity(vector) = IO_SAPIC_POL_HIGH;
-		iosapic_trigger(vector) = IO_SAPIC_LEVEL;
-		break;
-	case 0xf:
-		iosapic_polarity(vector) = IO_SAPIC_POL_LOW;
-		iosapic_trigger(vector) = IO_SAPIC_LEVEL;
-		break;
-	default:
-		printk("    ACPI Legacy IRQ 0x%02x: Unknown flags 0x%x\n", legacy->isa_irq,
-		       legacy->flags);
-		break;
+		case ACPI20_ENTRY_IO_SAPIC:
+			iosapic = (acpi_entry_iosapic_t *) p;
+			if (iosapic_init)
+				iosapic_init(iosapic->address, iosapic->irq_base);
+			break;
+
+		case ACPI20_ENTRY_PLATFORM_INT_SOURCE:
+			printk("ACPI 2.0 MADT: PLATFORM INT SOUCE\n");
+			acpi20_platform(p);
+			break;
+
+		case ACPI20_ENTRY_LOCAL_APIC:
+			printk("ACPI 2.0 MADT: LOCAL APIC entry\n"); break;
+		case ACPI20_ENTRY_IO_APIC:
+			printk("ACPI 2.0 MADT: IO APIC entry\n"); break;
+		case ACPI20_ENTRY_NMI_SOURCE:
+			printk("ACPI 2.0 MADT: NMI SOURCE entry\n"); break;
+		case ACPI20_ENTRY_LOCAL_APIC_NMI:
+			printk("ACPI 2.0 MADT: LOCAL APIC NMI entry\n"); break;
+		case ACPI20_ENTRY_INT_SRC_OVERRIDE:
+			break;
+		default:
+			printk("ACPI 2.0 MADT: unknown entry skip\n"); break;
+			break;
+		}
+
+		p += p[1];
 	}
 
-# ifdef ACPI_DEBUG
-	printk("Legacy ISA IRQ %x -> IA64 Vector %x IOSAPIC Pin %x Active %s %s Trigger\n", 
-	       legacy->isa_irq, vector, iosapic_pin(vector), 
-	       ((iosapic_polarity(vector) == IO_SAPIC_POL_LOW) ? "Low" : "High"),
-	       ((iosapic_trigger(vector) == IO_SAPIC_LEVEL) ? "Level" : "Edge"));
-# endif /* ACPI_DEBUG */
-#endif /* CONFIG_IA64_IRQ_ACPI */
+	p = (char *) (madt + 1);
+	end = p + (madt->header.length - sizeof(acpi_madt_t));
+
+	while (p < end) {
+		
+		switch (*p) {
+		case ACPI20_ENTRY_INT_SRC_OVERRIDE:
+			printk("ACPI 2.0 MADT: INT SOURCE Override\n");
+			acpi_legacy_irq(p);
+			break;
+		default:
+			break;
+		}
+
+		p += p[1];
+	}
+
+	/* Make bootup pretty */
+	printk("      %d CPUs available, %d CPUs total\n",
+		available_cpus, total_cpus);
+}
+
+int __init 
+acpi20_parse (acpi20_rsdp_t *rsdp20)
+{
+	acpi_xsdt_t *xsdt;
+	acpi_desc_table_hdr_t *hdrp;
+	int tables, i;
+
+	if (strncmp(rsdp20->signature, ACPI_RSDP_SIG, ACPI_RSDP_SIG_LEN)) {
+		printk("ACPI 2.0 RSDP signature incorrect!\n");
+		return 0;
+	} else {
+		printk("ACPI 2.0 Root System Description Ptr at 0x%lx\n",
+			(unsigned long)rsdp20);
+	}
+
+	xsdt = __va(rsdp20->xsdt);
+	hdrp = &xsdt->header;
+	if (strncmp(hdrp->signature,
+		ACPI_XSDT_SIG, ACPI_XSDT_SIG_LEN)) {
+		printk("ACPI 2.0 XSDT signature incorrect. Trying RSDT\n");
+		/* RSDT parsing here */
+		return 0;
+	} else {
+		printk("ACPI 2.0 XSDT at 0x%lx (p=0x%lx)\n",
+		(unsigned long)xsdt, (unsigned long)rsdp20->xsdt);
+	}
+
+	printk("ACPI 2.0: %.6s %.8s %d.%d\n",
+		hdrp->oem_id,
+		hdrp->oem_table_id,
+		hdrp->oem_revision >> 16,
+		hdrp->oem_revision & 0xffff);
+
+#ifdef CONFIG_ACPI_KERNEL_CONFIG
+	acpi_cf_init((void *)rsdp20);
+#endif
+
+	tables =(hdrp->length -sizeof(acpi_desc_table_hdr_t))>>3;
+
+	for (i = 0; i < tables; i++) {
+		hdrp = (acpi_desc_table_hdr_t *) __va(readl_unaligned(&xsdt->entry_ptrs[i]));
+		printk("        :table %4.4s found\n", hdrp->signature);
+
+		/* Only interested int the MADT table for now ... */
+		if (strncmp(hdrp->signature,
+			ACPI_MADT_SIG, ACPI_MADT_SIG_LEN) != 0)
+			continue;
+
+		acpi20_parse_madt((acpi_madt_t *) hdrp);
+	}
+
+#ifdef CONFIG_ACPI_KERNEL_CONFIG
+	acpi_cf_terminate();
+#endif
+
+#ifdef CONFIG_SMP
+	if (available_cpus == 0) {
+		printk("ACPI: Found 0 CPUS; assuming 1\n");
+		available_cpus = 1; /* We've got at least one of these, no? */
+	}
+	smp_boot_data.cpu_count = available_cpus;
+#endif
+	return 1;
+}
+/*
+ * ACPI 1.0b with 0.71 IA64 extensions functions; should be removed once all 
+ * platforms start supporting ACPI 2.0
+ */
+
+/*
+ * Identify usable CPU's and remember them for SMP bringup later.
+ */
+static void __init
+acpi_lsapic (char *p) 
+{
+	int add = 1;
+
+	acpi_entry_lsapic_t *lsapic = (acpi_entry_lsapic_t *) p;
+
+	if ((lsapic->flags & LSAPIC_PRESENT) == 0) 
+		return;
+
+	printk("      CPU %d (%.04x:%.04x): ", total_cpus, lsapic->eid, lsapic->id);
+
+	if ((lsapic->flags & LSAPIC_ENABLED) == 0) {
+		printk("Disabled.\n");
+		add = 0;
+	} else if (lsapic->flags & LSAPIC_PERFORMANCE_RESTRICTED) {
+		printk("Performance Restricted; ignoring.\n");
+		add = 0;
+	}
+
+#ifdef CONFIG_SMP
+	smp_boot_data.cpu_phys_id[total_cpus] = -1;
+#endif
+	if (add) {
+		printk("Available.\n");
+		available_cpus++;
+#ifdef CONFIG_SMP
+		smp_boot_data.cpu_phys_id[total_cpus] = (lsapic->id << 8) | lsapic->eid;
+#endif /* CONFIG_SMP */
+	}
+	total_cpus++;
 }
 
 /*
  * Info on platform interrupt sources: NMI. PMI, INIT, etc.
  */
 static void __init
-acpi_platform(char *p)
+acpi_platform (char *p)
 {
 	acpi_entry_platform_src_t *plat = (acpi_entry_platform_src_t *) p;
 
-	printk("PLATFORM: IOSAPIC %x -> Vector %lx on CPU %.04u:%.04u\n",
+	printk("PLATFORM: IOSAPIC %x -> Vector %x on CPU %.04u:%.04u\n",
 	       plat->iosapic_vector, plat->global_vector, plat->eid, plat->id);
 }
 
@@ -161,8 +372,9 @@
  * Parse the ACPI Multiple SAPIC Table
  */
 static void __init
-acpi_parse_msapic(acpi_sapic_t *msapic)
+acpi_parse_msapic (acpi_sapic_t *msapic)
 {
+	acpi_entry_iosapic_t *iosapic;
 	char *p, *end;
 
 	/* Base address of IPI Message Block */
@@ -172,41 +384,31 @@
 	end = p + (msapic->header.length - sizeof(acpi_sapic_t));
 
 	while (p < end) {
-		
 		switch (*p) {
-		case ACPI_ENTRY_LOCAL_SAPIC:
+		      case ACPI_ENTRY_LOCAL_SAPIC:
 			acpi_lsapic(p);
 			break;
 	
-		case ACPI_ENTRY_IO_SAPIC:
-			platform_register_iosapic((acpi_entry_iosapic_t *) p);
+		      case ACPI_ENTRY_IO_SAPIC:
+			iosapic = (acpi_entry_iosapic_t *) p;
+			if (iosapic_init)
+				iosapic_init(iosapic->address, iosapic->irq_base);
 			break;
 
-		case ACPI_ENTRY_INT_SRC_OVERRIDE:
+		      case ACPI_ENTRY_INT_SRC_OVERRIDE:
 			acpi_legacy_irq(p);
 			break;
-		
-		case ACPI_ENTRY_PLATFORM_INT_SOURCE:
+
+		      case ACPI_ENTRY_PLATFORM_INT_SOURCE:
 			acpi_platform(p);
 			break;
-		
-		default:
+
+		      default:
 			break;
 		}
 
 		/* Move to next table entry. */
-#define BAD_ACPI_TABLE
-#ifdef BAD_ACPI_TABLE
-		/*
-		 * Some prototype Lion's have a bad ACPI table
-		 * requiring this fix.  Without this fix, those
-		 * machines crash during bootup.
-		 */
-		if (p[1] == 0)
-			p = end;
-		else
-#endif
-			p += p[1];
+		p += p[1];
 	}
 
 	/* Make bootup pretty */
@@ -214,24 +416,18 @@
 }
 
 int __init 
-acpi_parse(acpi_rsdp_t *rsdp)
+acpi_parse (acpi_rsdp_t *rsdp)
 {
 	acpi_rsdt_t *rsdt;
 	acpi_desc_table_hdr_t *hdrp;
 	long tables, i;
 
-	if (!rsdp) {
-		printk("Uh-oh, no ACPI Root System Description Pointer table!\n");
-		return 0;
-	}
-
 	if (strncmp(rsdp->signature, ACPI_RSDP_SIG, ACPI_RSDP_SIG_LEN)) {
 		printk("Uh-oh, ACPI RSDP signature incorrect!\n");
 		return 0;
 	}
 
-	rsdp->rsdt = __va(rsdp->rsdt);
-	rsdt = rsdp->rsdt;
+	rsdt = __va(rsdp->rsdt);
 	if (strncmp(rsdt->header.signature, ACPI_RSDT_SIG, ACPI_RSDT_SIG_LEN)) {
 		printk("Uh-oh, ACPI RDST signature incorrect!\n");
 		return 0;
@@ -256,7 +452,7 @@
 	}
 
 #ifdef CONFIG_ACPI_KERNEL_CONFIG
-       acpi_cf_terminate();
+	acpi_cf_terminate();
 #endif
 
 #ifdef CONFIG_SMP
@@ -267,23 +463,4 @@
 	smp_boot_data.cpu_count = available_cpus;
 #endif
 	return 1;
-}
-
-const char *
-acpi_get_sysname (void)
-{       
-	/* the following should go away once we have an ACPI parser: */
-#ifdef CONFIG_IA64_GENERIC
-	return "hpsim";
-#else
-# if defined (CONFIG_IA64_HP_SIM)
-	return "hpsim";
-# elif defined (CONFIG_IA64_SGI_SN1)
-	return "sn1";
-# elif defined (CONFIG_IA64_DIG)
-	return "dig";
-# else
-#	error Unknown platform.  Fix acpi.c.
-# endif
-#endif
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/efi.c linux/arch/ia64/kernel/efi.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/efi.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/efi.c	Thu Jan  4 12:50:17 2001
@@ -18,7 +18,6 @@
  * Goutham Rao: <goutham.rao@intel.com>
  * 	Skip non-WB memory and ignore empty memory ranges.
  */
-#include <linux/config.h>
 #include <linux/kernel.h>
 #include <linux/init.h>
 #include <linux/types.h>
@@ -333,6 +332,9 @@
 		if (efi_guidcmp(config_tables[i].guid, MPS_TABLE_GUID) == 0) {
 			efi.mps = __va(config_tables[i].table);
 			printk(" MPS=0x%lx", config_tables[i].table);
+		} else if (efi_guidcmp(config_tables[i].guid, ACPI_20_TABLE_GUID) == 0) {
+			efi.acpi20 = __va(config_tables[i].table);
+			printk(" ACPI 2.0=0x%lx", config_tables[i].table);
 		} else if (efi_guidcmp(config_tables[i].guid, ACPI_TABLE_GUID) == 0) {
 			efi.acpi = __va(config_tables[i].table);
 			printk(" ACPI=0x%lx", config_tables[i].table);
@@ -364,7 +366,7 @@
 #if EFI_DEBUG
 	/* print EFI memory map: */
 	{
-		efi_memory_desc_t *md = p;
+		efi_memory_desc_t *md;
 		void *p;
 
 		for (i = 0, p = efi_map_start; p < efi_map_end; ++i, p += efi_desc_size) {
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/entry.S linux/arch/ia64/kernel/entry.S
--- v2.4.0-prerelease/linux/arch/ia64/kernel/entry.S	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/entry.S	Thu Jan  4 12:50:17 2001
@@ -11,6 +11,17 @@
  * Copyright (C) 1999 Don Dugger <Don.Dugger@intel.com>
  */
 /*
+ * ia64_switch_to now places correct virtual mapping in in TR2 for
+ * kernel stack. This allows us to handle interrupts without changing
+ * to physical mode.
+ *
+ * ar.k4 is now used to hold last virtual map address
+ * 
+ * Jonathan Nickin	<nicklin@missioncriticallinux.com>
+ * Patrick O'Rourke	<orourke@missioncriticallinux.com>
+ * 11/07/2000
+ /
+/*
  * Global (preserved) predicate usage on syscall entry/exit path:
  *
  *	pKern:		See entry.h.
@@ -27,7 +38,8 @@
 #include <asm/processor.h>
 #include <asm/unistd.h>
 #include <asm/asmmacro.h>
-
+#include <asm/pgtable.h>
+	
 #include "entry.h"
 
 	.text
@@ -98,6 +110,8 @@
 	br.ret.sptk.many rp
 END(sys_clone)
 
+#define KSTACK_TR	2
+
 /*
  * prev_task <- ia64_switch_to(struct task_struct *next)
  */
@@ -108,22 +122,55 @@
 	UNW(.body)
 
 	adds r22=IA64_TASK_THREAD_KSP_OFFSET,r13
-	dep r18=-1,r0,0,61	// build mask 0x1fffffffffffffff
+	mov r27=ar.k4
+	dep r20=0,in0,61,3		// physical address of "current"
+	;;
+	st8 [r22]=sp			// save kernel stack pointer of old task
+	shr.u r26=r20,_PAGE_SIZE_256M
+	;;
+	cmp.eq p7,p6=r26,r0		// check < 256M
 	adds r21=IA64_TASK_THREAD_KSP_OFFSET,in0
 	;;
-	st8 [r22]=sp		// save kernel stack pointer of old task
-	ld8 sp=[r21]		// load kernel stack pointer of new task
-	and r20=in0,r18		// physical address of "current"
-	;;
-	mov ar.k6=r20		// copy "current" into ar.k6
-	mov r8=r13		// return pointer to previously running task
-	mov r13=in0		// set "current" pointer
+	/*
+	 * If we've already mapped this task's page, we can skip doing it
+	 * again.
+	 */
+(p6)	cmp.eq p7,p6=r26,r27
+(p6)	br.cond.dpnt.few .map
+	;;
+.done:	ld8 sp=[r21]			// load kernel stack pointer of new task
+(p6)	ssm psr.ic			// if we we had to map, renable the psr.ic bit FIRST!!!
 	;;
+(p6)	srlz.d
+	mov ar.k6=r20			// copy "current" into ar.k6
+	mov r8=r13			// return pointer to previously running task
+	mov r13=in0			// set "current" pointer
+	;;
+(p6)	ssm psr.i			// renable psr.i AFTER the ic bit is serialized
 	DO_LOAD_SWITCH_STACK( )
+
 #ifdef CONFIG_SMP
-	sync.i			// ensure "fc"s done by this CPU are visible on other CPUs
-#endif
-	br.ret.sptk.few rp
+	sync.i				// ensure "fc"s done by this CPU are visible on other CPUs
+#endif 
+	br.ret.sptk.few rp		// boogie on out in new context
+
+.map:
+	rsm psr.i | psr.ic
+	movl r25=__DIRTY_BITS|_PAGE_PL_0|_PAGE_AR_RWX
+	;;
+	srlz.d
+	or r23=r25,r20			// construct PA | page properties
+	mov r25=_PAGE_SIZE_256M<<2
+	;;
+	mov cr.itir=r25
+	mov cr.ifa=in0			// VA of next task...
+	;;
+	mov r25=KSTACK_TR		// use tr entry #2...
+	mov ar.k4=r26			// remember last page we mapped...
+	;;
+	itr.d dtr[r25]=r23		// wire in new mapping...
+	br.cond.sptk.many .done
+	;;
 END(ia64_switch_to)
 
 #ifndef CONFIG_IA64_NEW_UNWIND
@@ -503,7 +550,7 @@
 	;;
 	ld4 r2=[r2]
 	;;
-	shl r2=r2,SMP_LOG_CACHE_BYTES	// can't use shladd here...
+	shl r2=r2,SMP_CACHE_SHIFT	// can't use shladd here...
 	;;
 	add r3=r2,r3
 #else
@@ -542,7 +589,7 @@
 	// check & deliver pending signals:
 (p2)	br.call.spnt.few rp=handle_signal_delivery
 .ret9:
-#if defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_IA64_SOFTSDV_HACKS)
+#ifdef CONFIG_IA64_SOFTSDV_HACKS
 	// Check for lost ticks
 	rsm psr.i
 	mov r2 = ar.itc
@@ -611,14 +658,13 @@
 	mov ar.ccv=r1
 	mov ar.fpsr=r13
 	mov b0=r14
-	// turn off interrupts, interrupt collection, & data translation
-	rsm psr.i | psr.ic | psr.dt
+	// turn off interrupts, interrupt collection
+	rsm psr.i | psr.ic
 	;;
 	srlz.i			// EAS 2.5
 	mov b7=r15
 	;;
 	invala			// invalidate ALAT
-	dep r12=0,r12,61,3	// convert sp to physical address
 	bsw.0;;			// switch back to bank 0 (must be last in insn group)
 	;;
 #ifdef CONFIG_ITANIUM_ASTEP_SPECIFIC
@@ -757,7 +803,7 @@
 
 #endif /* CONFIG_SMP */
 
-#if defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_IA64_SOFTSDV_HACKS)
+#ifdef CONFIG_IA64_SOFTSDV_HACKS
 
 ENTRY(invoke_ia64_reset_itm)
 	UNW(.prologue ASM_UNW_PRLG_RP|ASM_UNW_PRLG_PFS, ASM_UNW_PRLG_GRSAVE(8))
@@ -772,7 +818,7 @@
 	br.ret.sptk.many rp
 END(invoke_ia64_reset_itm)
 
-#endif /* CONFIG_ITANIUM_ASTEP_SPECIFIC || CONFIG_IA64_SOFTSDV_HACKS */
+#endif /* CONFIG_IA64_SOFTSDV_HACKS */
 
 	/*
 	 * Invoke do_softirq() while preserving in0-in7, which may be needed
@@ -1091,7 +1137,7 @@
 	data8 sys_setpriority
 	data8 sys_statfs
 	data8 sys_fstatfs
-	data8 ia64_ni_syscall
+	data8 ia64_ni_syscall			// 1105
 	data8 sys_semget
 	data8 sys_semop
 	data8 sys_semctl
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/fw-emu.c linux/arch/ia64/kernel/fw-emu.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/fw-emu.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/fw-emu.c	Thu Jan  4 12:50:17 2001
@@ -402,7 +402,6 @@
 	sal_systab->sal_rev_minor = 1;
 	sal_systab->sal_rev_major = 0;
 	sal_systab->entry_count = 1;
-	sal_systab->ia32_bios_present = 0;
 
 #ifdef CONFIG_IA64_GENERIC
         strcpy(sal_systab->oem_id, "Generic");
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/head.S linux/arch/ia64/kernel/head.S
--- v2.4.0-prerelease/linux/arch/ia64/kernel/head.S	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/head.S	Thu Jan  4 12:50:17 2001
@@ -74,8 +74,8 @@
 	;;
 
 #ifdef CONFIG_IA64_EARLY_PRINTK
-	mov r2=6
-	mov r3=(8<<8) | (28<<2)
+	mov r3=(6<<8) | (28<<2)
+	movl r2=6<<61
 	;;
 	mov rr[r2]=r3
 	;;
@@ -168,6 +168,11 @@
 	add r19=IA64_NUM_DBG_REGS*8,in0
 	;;
 1:	mov r16=dbr[r18]
+#if defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_BSTEP_SPECIFIC) \
+    || defined(CONFIG_ITANIUM_C0_SPECIFIC)
+	;;
+	srlz.d
+#endif
 	mov r17=ibr[r18]
 	add r18=1,r18
 	;;
@@ -181,7 +186,8 @@
 
 GLOBAL_ENTRY(ia64_load_debug_regs)
 	alloc r16=ar.pfs,1,0,0,0
-#if !(defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_BSTEP_SPECIFIC))
+#if !(defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) \
+   || defined(CONFIG_ITANIUM_B0_SPECIFIC) || defined(CONFIG_ITANIUM_B1_SPECIFIC))
 	lfetch.nta [in0]
 #endif
 	mov r20=ar.lc			// preserve ar.lc
@@ -194,6 +200,11 @@
 	add r18=1,r18
 	;;
 	mov dbr[r18]=r16
+#if defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_BSTEP_SPECIFIC) \
+    || defined(CONFIG_ITANIUM_C0_SPECIFIC)
+	;;
+	srlz.d
+#endif
 	mov ibr[r18]=r17
 	br.cloop.sptk.few 1b
 	;;
@@ -754,7 +765,7 @@
 	mov tmp=ar.itc
 (p15)	br.cond.sptk .wait
 	;;
-	ld1 tmp=[r31]
+	ld4 tmp=[r31]
 	;;
 	cmp.ne p15,p0=tmp,r0
 	mov tmp=ar.itc
@@ -764,7 +775,7 @@
 	mov tmp=1
 	;;
 	IA64_SEMFIX_INSN
-	cmpxchg1.acq tmp=[r31],tmp,ar.ccv
+	cmpxchg4.acq tmp=[r31],tmp,ar.ccv
 	;;
 	cmp.eq p15,p0=tmp,r0
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/ia64_ksyms.c linux/arch/ia64/kernel/ia64_ksyms.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/ia64_ksyms.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/ia64_ksyms.c	Thu Jan  4 12:50:17 2001
@@ -24,9 +24,8 @@
 EXPORT_SYMBOL(strstr);
 EXPORT_SYMBOL(strtok);
 
-#include <linux/pci.h>
-EXPORT_SYMBOL(pci_alloc_consistent);
-EXPORT_SYMBOL(pci_free_consistent);
+#include <asm/hw_irq.h>
+EXPORT_SYMBOL(isa_irq_to_vector_map);
 
 #include <linux/in6.h>
 #include <asm/checksum.h>
@@ -49,14 +48,6 @@
 #include <asm/page.h>
 EXPORT_SYMBOL(clear_page);
 
-#include <asm/pci.h>
-EXPORT_SYMBOL(pci_dma_sync_sg);
-EXPORT_SYMBOL(pci_dma_sync_single);
-EXPORT_SYMBOL(pci_map_sg);
-EXPORT_SYMBOL(pci_map_single);
-EXPORT_SYMBOL(pci_unmap_sg);
-EXPORT_SYMBOL(pci_unmap_single);
-
 #include <asm/processor.h>
 EXPORT_SYMBOL(cpu_data);
 EXPORT_SYMBOL(kernel_thread);
@@ -92,6 +83,9 @@
 #include <asm/uaccess.h>
 EXPORT_SYMBOL(__copy_user);
 EXPORT_SYMBOL(__do_clear_user);
+EXPORT_SYMBOL(__strlen_user);
+EXPORT_SYMBOL(__strncpy_from_user);
+EXPORT_SYMBOL(__strnlen_user);
 
 #include <asm/unistd.h>
 EXPORT_SYMBOL(__ia64_syscall);
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/iosapic.c linux/arch/ia64/kernel/iosapic.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/iosapic.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/kernel/iosapic.c	Thu Jan  4 12:50:17 2001
@@ -0,0 +1,498 @@
+/*
+ * I/O SAPIC support.
+ *
+ * Copyright (C) 1999 Intel Corp.
+ * Copyright (C) 1999 Asit Mallick <asit.k.mallick@intel.com>
+ * Copyright (C) 1999-2000 Hewlett-Packard Co.
+ * Copyright (C) 1999-2000 David Mosberger-Tang <davidm@hpl.hp.com>
+ * Copyright (C) 1999 VA Linux Systems
+ * Copyright (C) 1999,2000 Walt Drummond <drummond@valinux.com>
+ *
+ * 00/04/19	D. Mosberger	Rewritten to mirror more closely the x86 I/O APIC code.
+ *				In particular, we now have separate handlers for edge
+ *				and level triggered interrupts.
+ * 00/10/27	Asit Mallick, Goutham Rao <goutham.rao@intel.com> IRQ vector allocation 
+ *				PCI to vector mapping, shared PCI interrupts.
+ * 00/10/27	D. Mosberger	Document things a bit more to make them more understandable.
+ *				Clean up much of the old IOSAPIC cruft.
+ */
+/*
+ * Here is what the interrupt logic between a PCI device and the CPU looks like:
+ *
+ * (1) A PCI device raises one of the four interrupt pins (INTA, INTB, INTC, INTD).  The
+ *     device is uniquely identified by its bus-, device-, and slot-number (the function
+ *     number does not matter here because all functions share the same interrupt
+ *     lines).
+ *
+ * (2) The motherboard routes the interrupt line to a pin on a IOSAPIC controller.
+ *     Multiple interrupt lines may have to share the same IOSAPIC pin (if they're level
+ *     triggered and use the same polarity).  Each interrupt line has a unique IOSAPIC
+ *     irq number which can be calculated as the sum of the controller's base irq number
+ *     and the IOSAPIC pin number to which the line connects.
+ *
+ * (3) The IOSAPIC uses an internal table to map the IOSAPIC pin into the IA-64 interrupt
+ *     vector.  This interrupt vector is then sent to the CPU.
+ *
+ * In other words, there are two levels of indirections involved:
+ *
+ *	pci pin -> iosapic irq -> IA-64 vector
+ *
+ * Note: outside this module, IA-64 vectors are called "irqs".  This is because that's
+ * the traditional name Linux uses for interrupt vectors.
+ */
+#include <linux/config.h>
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
+#include <linux/string.h>
+#include <linux/irq.h>
+
+#include <asm/acpi-ext.h>
+#include <asm/delay.h>
+#include <asm/io.h>
+#include <asm/iosapic.h>
+#include <asm/machvec.h>
+#include <asm/processor.h>
+#include <asm/ptrace.h>
+#include <asm/system.h>
+
+#ifdef	CONFIG_ACPI_KERNEL_CONFIG
+# include <asm/acpikcfg.h>
+#endif
+
+#undef DEBUG_IRQ_ROUTING
+
+static spinlock_t iosapic_lock = SPIN_LOCK_UNLOCKED;
+
+/* PCI pin to IOSAPIC irq routing information.  This info typically comes from ACPI. */
+
+static struct {
+	int num_routes;
+	struct pci_vector_struct *route;
+} pci_irq;
+
+/* This tables maps IA-64 vectors to the IOSAPIC pin that generates this vector. */
+
+static struct iosapic_irq {
+	char *addr;			/* base address of IOSAPIC */
+	unsigned char base_irq;		/* first irq assigned to this IOSAPIC */
+        char pin;			/* IOSAPIC pin (-1 => not an IOSAPIC irq) */
+	unsigned char dmode 	: 3;	/* delivery mode (see iosapic.h) */
+	unsigned char polarity	: 1;	/* interrupt polarity (see iosapic.h) */
+	unsigned char trigger	: 1;	/* trigger mode (see iosapic.h) */
+} iosapic_irq[NR_IRQS];
+
+/*
+ * Translate IOSAPIC irq number to the corresponding IA-64 interrupt vector.  If no
+ * entry exists, return -1.
+ */
+static int 
+iosapic_irq_to_vector (int irq)
+{
+	int vector;
+
+	for (vector = 0; vector < NR_IRQS; ++vector)
+		if (iosapic_irq[vector].base_irq + iosapic_irq[vector].pin == irq)
+			return vector;
+	return -1;
+}
+		
+/*
+ * Map PCI pin to the corresponding IA-64 interrupt vector.  If no such mapping exists,
+ * return -1.
+ */
+static int
+pci_pin_to_vector (int bus, int slot, int pci_pin)
+{
+	struct pci_vector_struct *r;
+
+	for (r = pci_irq.route; r < pci_irq.route + pci_irq.num_routes; ++r)
+		if (r->bus == bus && (r->pci_id >> 16) == slot && r->pin == pci_pin)
+			return iosapic_irq_to_vector(r->irq);
+	return -1;
+}
+
+static void
+set_rte (unsigned int vector, unsigned long dest)
+{
+	unsigned long pol, trigger, dmode;
+	u32 low32, high32;
+	char *addr;
+	int pin;
+
+	pin = iosapic_irq[vector].pin;
+	if (pin < 0)
+		return;		/* not an IOSAPIC interrupt */
+
+	addr    = iosapic_irq[vector].addr;
+	pol     = iosapic_irq[vector].polarity;
+	trigger = iosapic_irq[vector].trigger;
+	dmode   = iosapic_irq[vector].dmode;
+
+	low32 = ((pol << IOSAPIC_POLARITY_SHIFT) |
+		 (trigger << IOSAPIC_TRIGGER_SHIFT) |
+		 (dmode << IOSAPIC_DELIVERY_SHIFT) |
+		 vector);
+
+#ifdef CONFIG_IA64_AZUSA_HACKS
+	/* set Flush Disable bit */
+	if (addr != (char *) 0xc0000000fec00000)
+		low32 |= (1 << 17);
+#endif
+
+	/* dest contains both id and eid */
+	high32 = (dest << IOSAPIC_DEST_SHIFT);	
+
+	writel(IOSAPIC_RTE_HIGH(pin), addr + IOSAPIC_REG_SELECT);
+	writel(high32, addr + IOSAPIC_WINDOW);
+	writel(IOSAPIC_RTE_LOW(pin), addr + IOSAPIC_REG_SELECT);
+	writel(low32, addr + IOSAPIC_WINDOW);
+}
+
+static void
+nop (unsigned int vector)
+{
+	/* do nothing... */
+}
+
+static void 
+mask_irq (unsigned int vector)
+{
+	unsigned long flags;
+	char *addr;
+	u32 low32;
+	int pin;
+
+	addr = iosapic_irq[vector].addr;
+	pin = iosapic_irq[vector].pin;
+
+	if (pin < 0)
+		return;			/* not an IOSAPIC interrupt! */
+
+	spin_lock_irqsave(&iosapic_lock, flags);
+	{
+		writel(IOSAPIC_RTE_LOW(pin), addr + IOSAPIC_REG_SELECT);
+		low32 = readl(addr + IOSAPIC_WINDOW);
+
+		low32 |= (1 << IOSAPIC_MASK_SHIFT);    /* set only the mask bit */
+		writel(low32, addr + IOSAPIC_WINDOW);
+	}
+	spin_unlock_irqrestore(&iosapic_lock, flags);
+}
+
+static void 
+unmask_irq (unsigned int vector)
+{
+	unsigned long flags;
+	char *addr;
+	u32 low32;
+	int pin;
+
+	addr = iosapic_irq[vector].addr;
+	pin = iosapic_irq[vector].pin;
+	if (pin < 0)
+		return;			/* not an IOSAPIC interrupt! */
+
+	spin_lock_irqsave(&iosapic_lock, flags);
+	{
+		writel(IOSAPIC_RTE_LOW(pin), addr + IOSAPIC_REG_SELECT);
+		low32 = readl(addr + IOSAPIC_WINDOW);
+
+		low32 &= ~(1 << IOSAPIC_MASK_SHIFT);    /* clear only the mask bit */
+		writel(low32, addr + IOSAPIC_WINDOW);
+	}
+	spin_unlock_irqrestore(&iosapic_lock, flags);
+}
+
+
+static void
+iosapic_set_affinity (unsigned int vector, unsigned long mask)
+{
+	printk("iosapic_set_affinity: not implemented yet\n");
+}
+
+/*
+ * Handlers for level-triggered interrupts.
+ */
+
+static unsigned int
+iosapic_startup_level_irq (unsigned int vector)
+{
+	unmask_irq(vector);
+	return 0;
+}
+
+static void
+iosapic_end_level_irq (unsigned int vector)
+{
+	writel(vector, iosapic_irq[vector].addr + IOSAPIC_EOI);
+}
+
+#define iosapic_shutdown_level_irq	mask_irq
+#define iosapic_enable_level_irq	unmask_irq
+#define iosapic_disable_level_irq	mask_irq
+#define iosapic_ack_level_irq		nop
+
+struct hw_interrupt_type irq_type_iosapic_level = {
+	typename:	"IO-SAPIC-level",
+	startup:	iosapic_startup_level_irq,
+	shutdown:	iosapic_shutdown_level_irq,
+	enable:		iosapic_enable_level_irq,
+	disable:	iosapic_disable_level_irq,
+	ack:		iosapic_ack_level_irq,
+	end:		iosapic_end_level_irq,
+	set_affinity:	iosapic_set_affinity
+};
+
+/*
+ * Handlers for edge-triggered interrupts.
+ */
+
+static unsigned int
+iosapic_startup_edge_irq (unsigned int vector)
+{
+	unmask_irq(vector);
+	/*
+	 * IOSAPIC simply drops interrupts pended while the
+	 * corresponding pin was masked, so we can't know if an
+	 * interrupt is pending already.  Let's hope not...
+	 */
+	return 0;
+}
+
+static void
+iosapic_ack_edge_irq (unsigned int vector)
+{
+	/*
+	 * Once we have recorded IRQ_PENDING already, we can mask the
+	 * interrupt for real. This prevents IRQ storms from unhandled
+	 * devices.
+	 */
+	if ((irq_desc[vector].status & (IRQ_PENDING|IRQ_DISABLED)) == (IRQ_PENDING|IRQ_DISABLED))
+		mask_irq(vector);
+}
+
+#define iosapic_enable_edge_irq		unmask_irq
+#define iosapic_disable_edge_irq	nop
+#define iosapic_end_edge_irq		nop
+
+struct hw_interrupt_type irq_type_iosapic_edge = {
+	typename:	"IO-SAPIC-edge",
+	startup:	iosapic_startup_edge_irq,
+	shutdown:	iosapic_disable_edge_irq,
+	enable:		iosapic_enable_edge_irq,
+	disable:	iosapic_disable_edge_irq,
+	ack:		iosapic_ack_edge_irq,
+	end:		iosapic_end_edge_irq,
+	set_affinity:	iosapic_set_affinity
+};
+
+static unsigned int
+iosapic_version (char *addr) 
+{
+	/*
+	 * IOSAPIC Version Register return 32 bit structure like:
+	 * {
+	 *	unsigned int version   : 8;
+	 *	unsigned int reserved1 : 8;
+	 *	unsigned int pins      : 8;
+	 *	unsigned int reserved2 : 8;
+	 * }
+	 */
+	writel(IOSAPIC_VERSION, addr + IOSAPIC_REG_SELECT);
+	return readl(IOSAPIC_WINDOW + addr);
+}
+
+/*
+ * ACPI calls this when it finds an entry for a legacy ISA interrupt.  Note that the
+ * irq_base and IOSAPIC address must be set in iosapic_init().
+ */
+void
+iosapic_register_legacy_irq (unsigned long irq,
+			     unsigned long pin, unsigned long polarity,
+			     unsigned long edge_triggered)
+{
+	unsigned int vector = isa_irq_to_vector(irq);
+
+#ifdef DEBUG_IRQ_ROUTING
+	printk("ISA: IRQ %u -> IOSAPIC irq 0x%02x (%s, %s) -> vector %02x\n",
+	       (unsigned) irq, (unsigned) pin,
+	       polarity ? "high" : "low", edge_triggered ? "edge" : "level",
+	       vector);
+#endif
+
+	iosapic_irq[vector].pin = pin;
+	iosapic_irq[vector].dmode = IOSAPIC_LOWEST_PRIORITY;
+	iosapic_irq[vector].polarity = polarity ? IOSAPIC_POL_HIGH : IOSAPIC_POL_LOW;
+	iosapic_irq[vector].trigger = edge_triggered ? IOSAPIC_EDGE : IOSAPIC_LEVEL;
+}
+
+void __init
+iosapic_init (unsigned long phys_addr, unsigned int base_irq)
+{
+	struct hw_interrupt_type *irq_type;
+	int i, irq, max_pin, vector;
+	unsigned int ver;
+	char *addr;
+	static int first_time = 1;
+
+	if (first_time) {
+		first_time = 0;
+
+		for (vector = 0; vector < NR_IRQS; ++vector)
+			iosapic_irq[vector].pin = -1;	/* mark as unused */
+
+		/* 
+		 * Fetch the PCI interrupt routing table:
+		 */
+#ifdef CONFIG_ACPI_KERNEL_CONFIG
+		acpi_cf_get_pci_vectors(&pci_irq.route, &pci_irq.num_routes);
+#else
+		pci_irq.route =
+			(struct pci_vector_struct *) __va(ia64_boot_param.pci_vectors);
+		pci_irq.num_routes = ia64_boot_param.num_pci_vectors;
+#endif
+	}
+
+	addr = ioremap(phys_addr, 0);
+
+	ver = iosapic_version(addr);
+	max_pin = (ver >> 16) & 0xff;
+	
+	printk("IOSAPIC: version %x.%x, address 0x%lx, IRQs 0x%02x-0x%02x\n", 
+	       (ver & 0xf0) >> 4, (ver & 0x0f), phys_addr, base_irq, base_irq + max_pin);
+
+	if (base_irq == 0)
+		/*
+		 * Map the legacy ISA devices into the IOSAPIC data.  Some of these may
+		 * get reprogrammed later on with data from the ACPI Interrupt Source
+		 * Override table.
+		 */
+		for (irq = 0; irq < 16; ++irq) {
+			vector = isa_irq_to_vector(irq);
+			iosapic_irq[vector].addr = addr;
+			iosapic_irq[vector].base_irq = 0;
+			if (iosapic_irq[vector].pin == -1)
+				iosapic_irq[vector].pin = irq;
+			iosapic_irq[vector].dmode = IOSAPIC_LOWEST_PRIORITY;
+			iosapic_irq[vector].trigger  = IOSAPIC_EDGE;
+			iosapic_irq[vector].polarity = IOSAPIC_POL_HIGH;
+#ifdef DEBUG_IRQ_ROUTING
+			printk("ISA: IRQ %u -> IOSAPIC irq 0x%02x (high, edge) -> vector 0x%02x\n",
+			       irq, iosapic_irq[vector].base_irq + iosapic_irq[vector].pin,
+			       vector);
+#endif
+		  	irq_type = &irq_type_iosapic_edge;
+			if (irq_desc[vector].handler != irq_type) {
+				if (irq_desc[vector].handler != &no_irq_type)
+					printk("iosapic_init: changing vector 0x%02x from %s to "
+					       "%s\n", irq, irq_desc[vector].handler->typename,
+					       irq_type->typename);
+				irq_desc[vector].handler = irq_type;
+			}
+
+			/* program the IOSAPIC routing table: */
+			set_rte(vector, (ia64_get_lid() >> 16) & 0xffff);
+		}
+
+#ifndef CONFIG_IA64_SOFTSDV_HACKS
+	for (i = 0; i < pci_irq.num_routes; i++) {
+		irq = pci_irq.route[i].irq;
+
+		if ((unsigned) (irq - base_irq) > max_pin)
+			/* the interrupt route is for another controller... */
+			continue;
+
+		if (irq < 16)
+			vector = isa_irq_to_vector(irq);
+		else {
+			vector = iosapic_irq_to_vector(irq);
+			if (vector < 0)
+				/* new iosapic irq: allocate a vector for it */
+				vector = ia64_alloc_irq();
+		}
+
+		iosapic_irq[vector].addr     = addr;
+		iosapic_irq[vector].base_irq = base_irq;
+		iosapic_irq[vector].pin	     = (irq - base_irq);
+		iosapic_irq[vector].dmode    = IOSAPIC_LOWEST_PRIORITY;
+		iosapic_irq[vector].trigger  = IOSAPIC_LEVEL;
+		iosapic_irq[vector].polarity = IOSAPIC_POL_LOW;
+
+# ifdef DEBUG_IRQ_ROUTING
+		printk("PCI: (B%d,I%d,P%d) -> IOSAPIC irq 0x%02x -> vector 0x%02x\n",
+		       pci_irq.route[i].bus, pci_irq.route[i].pci_id>>16, pci_irq.route[i].pin,
+		       iosapic_irq[vector].base_irq + iosapic_irq[vector].pin, vector);
+# endif
+		irq_type = &irq_type_iosapic_level;
+		if (irq_desc[vector].handler != irq_type){
+			if (irq_desc[vector].handler != &no_irq_type)
+				printk("iosapic_init: changing vector 0x%02x from %s to %s\n",
+				       vector, irq_desc[vector].handler->typename,
+				       irq_type->typename);
+			irq_desc[vector].handler = irq_type;
+		}
+
+		/* program the IOSAPIC routing table: */
+		set_rte(vector, (ia64_get_lid() >> 16) & 0xffff);
+	}
+#endif /* !CONFIG_IA64_SOFTSDV_HACKS */
+}
+
+void
+iosapic_pci_fixup (int phase)
+{
+	struct	pci_dev	*dev;
+	unsigned char pin;
+	int vector;
+
+	if (phase != 1)
+		return;
+
+	pci_for_each_dev(dev) {
+		pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
+		if (pin) {
+			pin--;          /* interrupt pins are numbered starting from 1 */
+			vector = pci_pin_to_vector(dev->bus->number, PCI_SLOT(dev->devfn), pin);
+			if (vector < 0 && dev->bus->parent) {
+				/* go back to the bridge */
+				struct pci_dev *bridge = dev->bus->self;
+
+				if (bridge) {
+					/* allow for multiple bridges on an adapter */
+					do {
+						/* do the bridge swizzle... */
+						pin = (pin + PCI_SLOT(dev->devfn)) % 4;
+						vector = pci_pin_to_vector(bridge->bus->number,
+									   PCI_SLOT(bridge->devfn),
+									   pin);
+					} while (vector < 0 && (bridge = bridge->bus->self));
+				}
+				if (vector >= 0)
+					printk(KERN_WARNING
+					       "PCI: using PPB(B%d,I%d,P%d) to get vector %02x\n",
+					       bridge->bus->number, PCI_SLOT(bridge->devfn),
+					       pin, vector);
+				else
+					printk(KERN_WARNING
+					       "PCI: Couldn't map irq for (B%d,I%d,P%d)o\n",
+					       bridge->bus->number, PCI_SLOT(bridge->devfn),
+					       pin);
+			}
+			if (vector >= 0) {
+				printk("PCI->APIC IRQ transform: (B%d,I%d,P%d) -> 0x%02x\n",
+				       dev->bus->number, PCI_SLOT(dev->devfn), pin, vector);
+				dev->irq = vector;
+			}
+		}
+		/*
+		 * Nothing to fixup
+		 * Fix out-of-range IRQ numbers
+		 */
+		if (dev->irq >= NR_IRQS)
+			dev->irq = 15;	/* Spurious interrupts */
+	}
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/irq.c linux/arch/ia64/kernel/irq.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/irq.c	Mon Dec 11 17:59:43 2000
+++ linux/arch/ia64/kernel/irq.c	Thu Jan  4 12:50:17 2001
@@ -541,6 +541,18 @@
 	spin_unlock_irqrestore(&desc->lock, flags);
 }
 
+void do_IRQ_per_cpu(unsigned long irq, struct pt_regs *regs)
+{
+	irq_desc_t *desc = irq_desc + irq;
+	int cpu = smp_processor_id();
+
+	kstat.irqs[cpu][irq]++;
+
+	desc->handler->ack(irq);
+	handle_IRQ_event(irq, regs, desc->action);
+	desc->handler->end(irq);
+}
+
 /*
  * do_IRQ handles all normal device IRQ's (the special
  * SMP cross-CPU interrupts have their own specific
@@ -581,8 +593,7 @@
 	if (!(status & (IRQ_DISABLED | IRQ_INPROGRESS))) {
 		action = desc->action;
 		status &= ~IRQ_PENDING; /* we commit to handling */
-		if (!(status & IRQ_PER_CPU))
-			status |= IRQ_INPROGRESS; /* we are handling it */
+		status |= IRQ_INPROGRESS; /* we are handling it */
 	}
 	desc->status = status;
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/irq_ia64.c linux/arch/ia64/kernel/irq_ia64.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/irq_ia64.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/irq_ia64.c	Thu Jan  4 12:50:17 2001
@@ -7,6 +7,9 @@
  *
  *  6/10/99: Updated to bring in sync with x86 version to facilitate
  *	     support for SMP and different interrupt controllers.
+ *
+ * 09/15/00 Goutham Rao <goutham.rao@intel.com> Implemented pci_irq_to_vector
+ *                      PCI to vector allocation routine.
  */
 
 #include <linux/config.h>
@@ -35,38 +38,28 @@
 
 #define IRQ_DEBUG	0
 
-#ifdef CONFIG_ITANIUM_A1_SPECIFIC
-spinlock_t ivr_read_lock;
-#endif
-
 /* default base addr of IPI table */
 unsigned long ipi_base_addr = (__IA64_UNCACHED_OFFSET | IPI_DEFAULT_BASE_ADDR);	
 
 /*
- * Legacy IRQ to IA-64 vector translation table.  Any vector not in
- * this table maps to itself (ie: irq 0x30 => IA64 vector 0x30)
+ * Legacy IRQ to IA-64 vector translation table.
  */
 __u8 isa_irq_to_vector_map[16] = {
 	/* 8259 IRQ translation, first 16 entries */
-	0x60, 0x50, 0x10, 0x51, 0x52, 0x53, 0x43, 0x54,
-	0x55, 0x56, 0x57, 0x58, 0x59, 0x5a, 0x40, 0x41
+	0x2f, 0x20, 0x2e, 0x2d, 0x2c, 0x2b, 0x2a, 0x29,
+	0x28, 0x27, 0x26, 0x25, 0x24, 0x23, 0x22, 0x21
 };
 
-#ifdef CONFIG_ITANIUM_A1_SPECIFIC
-
-int usbfix;
-
-static int __init
-usbfix_option (char *str)
+int
+ia64_alloc_irq (void)
 {
-	printk("irq: enabling USB workaround\n");
-	usbfix = 1;
-	return 1;
-}
-
-__setup("usbfix", usbfix_option);
+	static int next_irq = FIRST_DEVICE_IRQ;
 
-#endif /* CONFIG_ITANIUM_A1_SPECIFIC */
+	if (next_irq > LAST_DEVICE_IRQ)
+		/* XXX could look for sharable vectors instead of panic'ing... */
+		panic("ia64_alloc_irq: out of interrupt vectors!");
+	return next_irq++;
+}
 
 /*
  * That's where the IVT branches when we get an external
@@ -77,42 +70,6 @@
 ia64_handle_irq (unsigned long vector, struct pt_regs *regs)
 {
 	unsigned long saved_tpr;
-#ifdef CONFIG_ITANIUM_A1_SPECIFIC
-	unsigned long eoi_ptr;
- 
-# ifdef CONFIG_USB
-	extern void reenable_usb (void);
-	extern void disable_usb (void);
-
-	if (usbfix)
-		disable_usb();
-# endif
-	/*
-	 * Stop IPIs by getting the ivr_read_lock
-	 */
-	spin_lock(&ivr_read_lock);
-	{
-		unsigned int tmp;
-		/*
-		 * Disable PCI writes
-		 */
-		outl(0x80ff81c0, 0xcf8);
-		tmp = inl(0xcfc);
-		outl(tmp | 0x400, 0xcfc);
-		eoi_ptr = inl(0xcfc);
-		vector = ia64_get_ivr();
-		/*
-		 * Enable PCI writes
-		 */
-		outl(tmp, 0xcfc);
-	}
-	spin_unlock(&ivr_read_lock);
-
-# ifdef CONFIG_USB
-	if (usbfix)
-		reenable_usb();
-# endif
-#endif /* CONFIG_ITANIUM_A1_SPECIFIC */
 
 #if IRQ_DEBUG
 	{
@@ -161,7 +118,10 @@
 		ia64_set_tpr(vector);
 		ia64_srlz_d();
 
-		do_IRQ(vector, regs);
+		if ((irq_desc[vector].status & IRQ_PER_CPU) != 0)
+			do_IRQ_per_cpu(vector, regs);
+		else
+			do_IRQ(vector, regs);
 
 		/*
 		 * Disable interrupts and send EOI:
@@ -169,9 +129,6 @@
 		local_irq_disable();
 		ia64_set_tpr(saved_tpr);
 		ia64_eoi();
-#ifdef CONFIG_ITANIUM_A1_SPECIFIC
-		break;
-#endif
 		vector = ia64_get_ivr();
 	} while (vector != IA64_SPURIOUS_INT);
 }
@@ -194,8 +151,8 @@
 	 * Disable all local interrupts
 	 */
 	ia64_set_itv(0, 1);
-	ia64_set_lrr0(0, 1);	
-	ia64_set_lrr1(0, 1);	
+	ia64_set_lrr0(0, 1);
+	ia64_set_lrr1(0, 1);
 
 	irq_desc[IA64_SPURIOUS_INT].handler = &irq_type_ia64_sapic;
 #ifdef CONFIG_SMP
@@ -217,14 +174,11 @@
 }
 
 void
-ipi_send (int cpu, int vector, int delivery_mode, int redirect)
+ia64_send_ipi (int cpu, int vector, int delivery_mode, int redirect)
 {
 	unsigned long ipi_addr;
 	unsigned long ipi_data;
 	unsigned long phys_cpu_id;
-#ifdef CONFIG_ITANIUM_A1_SPECIFIC
-	unsigned long flags;
-#endif
 
 #ifdef CONFIG_SMP
 	phys_cpu_id = cpu_physical_id(cpu);
@@ -239,13 +193,5 @@
 	ipi_data = (delivery_mode << 8) | (vector & 0xff);
 	ipi_addr = ipi_base_addr | (phys_cpu_id << 4) | ((redirect & 1)  << 3);
 
-#ifdef CONFIG_ITANIUM_A1_SPECIFIC
-	spin_lock_irqsave(&ivr_read_lock, flags);
-#endif
-
 	writeq(ipi_data, ipi_addr);
-
-#ifdef CONFIG_ITANIUM_A1_SPECIFIC
-	spin_unlock_irqrestore(&ivr_read_lock, flags);
-#endif
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/ivt.S linux/arch/ia64/kernel/ivt.S
--- v2.4.0-prerelease/linux/arch/ia64/kernel/ivt.S	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/ivt.S	Thu Jan  4 12:50:17 2001
@@ -6,6 +6,7 @@
  * Copyright (C) 1998-2000 David Mosberger <davidm@hpl.hp.com>
  *
  * 00/08/23 Asit Mallick <asit.k.mallick@intel.com> TLB handling for SMP
+ * 00/12/20 David Mosberger-Tang <davidm@hpl.hp.com> DTLB/ITLB handler now uses virtual PT.
  */
 /*
  * This file defines the interrupt vector table used by the CPU.
@@ -44,23 +45,13 @@
 #include <asm/system.h>
 #include <asm/unistd.h>
 
-#define MINSTATE_START_SAVE_MIN	/* no special action needed */
-#define MINSTATE_END_SAVE_MIN									\
-	or r2=r2,r14;		/* make first base a kernel virtual address */			\
-	or r12=r12,r14;		/* make sp a kernel virtual address */				\
-	or r13=r13,r14;		/* make `current' a kernel virtual address */			\
-	bsw.1;			/* switch back to bank 1 (must be last in insn group) */	\
-	;;
-
+#define MINSTATE_VIRT	/* needed by minstate.h */
 #include "minstate.h"
 
 #define FAULT(n)									\
-	rsm psr.dt;			/* avoid nested faults due to TLB misses... */	\
-	;;										\
-	srlz.d;				/* ensure everyone knows psr.dt is off... */	\
 	mov r31=pr;									\
 	mov r19=n;;			/* prepare to save predicates */		\
-	br.cond.sptk.many dispatch_to_fault_handler
+	br.sptk.many dispatch_to_fault_handler
 
 /*
  * As we don't (hopefully) use the space available, we need to fill it with
@@ -122,15 +113,14 @@
 (p7)	dep r17=r17,r19,(PAGE_SHIFT-3),3	// put region number bits in place
 	srlz.d					// ensure "rsm psr.dt" has taken effect
 (p6)	movl r19=__pa(SWAPPER_PGD_ADDR)		// region 5 is rooted at swapper_pg_dir
-(p6)	shr r21=r21,PGDIR_SHIFT+PAGE_SHIFT-1
-(p7)	shr r21=r21,PGDIR_SHIFT+PAGE_SHIFT-4
+(p6)	shr.u r21=r21,PGDIR_SHIFT+PAGE_SHIFT
+(p7)	shr.u r21=r21,PGDIR_SHIFT+PAGE_SHIFT-3
 	;;
 (p6)	dep r17=r18,r19,3,(PAGE_SHIFT-3)	// r17=PTA + IFA(33,42)*8
 (p7)	dep r17=r18,r17,3,(PAGE_SHIFT-6)	// r17=PTA + (((IFA(61,63) << 7) | IFA(33,39))*8)
 	cmp.eq p7,p6=0,r21			// unused address bits all zeroes?
 	shr.u r18=r16,PMD_SHIFT			// shift L2 index into position
 	;;
-(p6)	cmp.eq p7,p6=-1,r21			// unused address bits all ones?
 	ld8 r17=[r17]				// fetch the L1 entry (may be 0)
 	;;
 (p7)	cmp.eq p6,p7=r17,r0			// was L1 entry NULL?
@@ -145,7 +135,7 @@
 (p7)	ld8 r18=[r21]				// read the L3 PTE
 	mov r19=cr.isr				// cr.isr bit 0 tells us if this is an insn miss
 	;;
-(p7)	tbit.z p6,p7=r18,0			// page present bit cleared?
+(p7)	tbit.z p6,p7=r18,_PAGE_P_BIT		// page present bit cleared?
 	mov r22=cr.iha				// get the VHPT address that caused the TLB miss
 	;;					// avoid RAW on p7
 (p7)	tbit.nz.unc p10,p11=r19,32		// is it an instruction TLB miss?
@@ -153,7 +143,7 @@
 	;;
 (p10)	itc.i r18				// insert the instruction TLB entry
 (p11)	itc.d r18				// insert the data TLB entry
-(p6)	br.spnt.few page_fault			// handle bad address/page not present (page fault)
+(p6)	br.spnt.many page_fault			// handle bad address/page not present (page fault)
 	mov cr.ifa=r22
 
 	// Now compute and insert the TLB entry for the virtual page table.
@@ -183,212 +173,117 @@
 
 	mov pr=r31,-1				// restore predicate registers
 	rfi
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x0400 Entry 1 (size 64 bundles) ITLB (21)
 	/*
-	 * The ITLB basically does the same as the VHPT handler except
-	 * that we always insert exactly one instruction TLB entry.
-	 */
-	/*
-	 * Attempt to lookup PTE through virtual linear page table.
-	 * The speculative access will fail if there is no TLB entry
-	 * for the L3 page table page we're trying to access.
+	 * The ITLB handler accesses the L3 PTE via the virtually mapped linear
+	 * page table.  If a nested TLB miss occurs, we switch into physical
+	 * mode, walk the page table, and then re-execute the L3 PTE read
+	 * and go on normally after that.
 	 */
+itlb_fault:
 	mov r16=cr.ifa				// get virtual address
-	mov r19=cr.iha				// get virtual address of L3 PTE
-	;;
-	ld8.s r17=[r19]				// try to read L3 PTE
+	mov r29=b0				// save b0
 	mov r31=pr				// save predicates
+	mov r17=cr.iha				// get virtual address of L3 PTE
+	movl r30=1f				// load nested fault continuation point
+	;;
+1:	ld8 r18=[r17]				// read L3 PTE
 	;;
-	tnat.nz p6,p0=r17			// did read succeed?
-(p6)	br.cond.spnt.many 1f
+	tbit.z p6,p0=r18,_PAGE_P_BIT		// page present bit cleared?
+(p6)	br.cond.spnt.many page_fault
 	;;
-	itc.i r17
+	itc.i r18
 	;;
 #ifdef CONFIG_SMP
-	ld8.s r18=[r19]				// try to read L3 PTE again and see if same
+	ld8 r19=[r17]				// read L3 PTE again and see if same
 	mov r20=PAGE_SHIFT<<2			// setup page size for purge
 	;;
-	cmp.eq p6,p7=r17,r18
+	cmp.ne p7,p0=r18,r19
 	;;
 (p7)	ptc.l r16,r20
 #endif
 	mov pr=r31,-1
 	rfi
-
-#ifdef CONFIG_DISABLE_VHPT
-itlb_fault:
-#endif
-1:	rsm psr.dt				// use physical addressing for data
-	mov r19=ar.k7				// get page table base address
-	shl r21=r16,3				// shift bit 60 into sign bit
-	shr.u r17=r16,61			// get the region number into r17
-	;;
-	cmp.eq p6,p7=5,r17			// is IFA pointing into to region 5?
-	shr.u r18=r16,PGDIR_SHIFT		// get bits 33-63 of the faulting address
-	;;
-(p7)	dep r17=r17,r19,(PAGE_SHIFT-3),3	// put region number bits in place
-	srlz.d					// ensure "rsm psr.dt" has taken effect
-(p6)	movl r19=__pa(SWAPPER_PGD_ADDR)		// region 5 is rooted at swapper_pg_dir
-(p6)	shr r21=r21,PGDIR_SHIFT+PAGE_SHIFT-1
-(p7)	shr r21=r21,PGDIR_SHIFT+PAGE_SHIFT-4
-	;;
-(p6)	dep r17=r18,r19,3,(PAGE_SHIFT-3)	// r17=PTA + IFA(33,42)*8
-(p7)	dep r17=r18,r17,3,(PAGE_SHIFT-6)	// r17=PTA + (((IFA(61,63) << 7) | IFA(33,39))*8)
-	cmp.eq p7,p6=0,r21			// unused address bits all zeroes?
-	shr.u r18=r16,PMD_SHIFT			// shift L2 index into position
-	;;
-(p6)	cmp.eq p7,p6=-1,r21			// unused address bits all ones?
-	ld8 r17=[r17]				// fetch the L1 entry (may be 0)
-	;;
-(p7)	cmp.eq p6,p7=r17,r0			// was L1 entry NULL?
-	dep r17=r18,r17,3,(PAGE_SHIFT-3)	// compute address of L2 page table entry
 	;;
-(p7)	ld8 r17=[r17]				// fetch the L2 entry (may be 0)
-	shr.u r19=r16,PAGE_SHIFT		// shift L3 index into position
-	;;
-(p7)	cmp.eq.or.andcm p6,p7=r17,r0		// was L2 entry NULL?
-	dep r17=r19,r17,3,(PAGE_SHIFT-3)	// compute address of L3 page table entry
-	;;
-(p7)	ld8 r18=[r17]				// read the L3 PTE
-	;;
-(p7)	tbit.z p6,p7=r18,0			// page present bit cleared?
-	;;
-(p7)	itc.i r18				// insert the instruction TLB entry
-(p6)	br.spnt.few page_fault			// handle bad address/page not present (page fault)
-	;;
-#ifdef CONFIG_SMP
-	ld8 r19=[r17]				// re-read the PTE and check if same
-	;;
-	cmp.eq p6,p7=r18,r19
-	mov r20=PAGE_SHIFT<<2
-	;;
-(p7)	ptc.l r16,r20				// PTE changed purge translation
-#endif
-
-	mov pr=r31,-1				// restore predicate registers
-	rfi
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x0800 Entry 2 (size 64 bundles) DTLB (9,48)
 	/*
-	 * The DTLB basically does the same as the VHPT handler except
-	 * that we always insert exactly one data TLB entry.
-	 */
-	/*
-	 * Attempt to lookup PTE through virtual linear page table.
-	 * The speculative access will fail if there is no TLB entry
-	 * for the L3 page table page we're trying to access.
+	 * The DTLB handler accesses the L3 PTE via the virtually mapped linear
+	 * page table.  If a nested TLB miss occurs, we switch into physical
+	 * mode, walk the page table, and then re-execute the L3 PTE read
+	 * and go on normally after that.
 	 */
+dtlb_fault:
 	mov r16=cr.ifa				// get virtual address
-	mov r19=cr.iha				// get virtual address of L3 PTE
-	;;
-	ld8.s r17=[r19]				// try to read L3 PTE
+	mov r29=b0				// save b0
 	mov r31=pr				// save predicates
+	mov r17=cr.iha				// get virtual address of L3 PTE
+	movl r30=1f				// load nested fault continuation point
 	;;
-	tnat.nz p6,p0=r17			// did read succeed?
-(p6)	br.cond.spnt.many 1f
+1:	ld8 r18=[r17]				// read L3 PTE
 	;;
-	itc.d r17
+	tbit.z p6,p0=r18,_PAGE_P_BIT		// page present bit cleared?
+(p6)	br.cond.spnt.many page_fault
+	;;
+	itc.d r18
 	;;
 #ifdef CONFIG_SMP
-	ld8.s r18=[r19]				// try to read L3 PTE again and see if same
+	ld8 r19=[r17]				// read L3 PTE again and see if same
 	mov r20=PAGE_SHIFT<<2			// setup page size for purge
 	;;
-	cmp.eq p6,p7=r17,r18
+	cmp.ne p7,p0=r18,r19
 	;;
 (p7)	ptc.l r16,r20
 #endif
 	mov pr=r31,-1
 	rfi
-
-#ifdef CONFIG_DISABLE_VHPT
-dtlb_fault:
-#endif
-1:	rsm psr.dt				// use physical addressing for data
-	mov r19=ar.k7				// get page table base address
-	shl r21=r16,3				// shift bit 60 into sign bit
-	shr.u r17=r16,61			// get the region number into r17
 	;;
-	cmp.eq p6,p7=5,r17			// is IFA pointing into to region 5?
-	shr.u r18=r16,PGDIR_SHIFT		// get bits 33-63 of the faulting address
-	;;
-(p7)	dep r17=r17,r19,(PAGE_SHIFT-3),3	// put region number bits in place
-	srlz.d					// ensure "rsm psr.dt" has taken effect
-(p6)	movl r19=__pa(SWAPPER_PGD_ADDR)		// region 5 is rooted at swapper_pg_dir
-(p6)	shr r21=r21,PGDIR_SHIFT+PAGE_SHIFT-1
-(p7)	shr r21=r21,PGDIR_SHIFT+PAGE_SHIFT-4
-	;;
-(p6)	dep r17=r18,r19,3,(PAGE_SHIFT-3)	// r17=PTA + IFA(33,42)*8
-(p7)	dep r17=r18,r17,3,(PAGE_SHIFT-6)	// r17=PTA + (((IFA(61,63) << 7) | IFA(33,39))*8)
-	cmp.eq p7,p6=0,r21			// unused address bits all zeroes?
-	shr.u r18=r16,PMD_SHIFT			// shift L2 index into position
-	;;
-(p6)	cmp.eq p7,p6=-1,r21			// unused address bits all ones?
-	ld8 r17=[r17]				// fetch the L1 entry (may be 0)
-	;;
-(p7)	cmp.eq p6,p7=r17,r0			// was L1 entry NULL?
-	dep r17=r18,r17,3,(PAGE_SHIFT-3)	// compute address of L2 page table entry
-	;;
-(p7)	ld8 r17=[r17]				// fetch the L2 entry (may be 0)
-	shr.u r19=r16,PAGE_SHIFT		// shift L3 index into position
-	;;
-(p7)	cmp.eq.or.andcm p6,p7=r17,r0		// was L2 entry NULL?
-	dep r17=r19,r17,3,(PAGE_SHIFT-3)	// compute address of L3 page table entry
-	;;
-(p7)	ld8 r18=[r17]				// read the L3 PTE
-	;;
-(p7)	tbit.z p6,p7=r18,0			// page present bit cleared?
-	;;
-(p7)	itc.d r18				// insert the instruction TLB entry
-(p6)	br.spnt.few page_fault			// handle bad address/page not present (page fault)
-	;;
-#ifdef CONFIG_SMP
-	ld8 r19=[r17]				// re-read the PTE and check if same
-	;;
-	cmp.eq p6,p7=r18,r19
-	mov r20=PAGE_SHIFT<<2
-	;;
-(p7)	ptc.l r16,r20				// PTE changed purge translation
-#endif
-	mov pr=r31,-1				// restore predicate registers
-	rfi
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x0c00 Entry 3 (size 64 bundles) Alt ITLB (19)
 	mov r16=cr.ifa		// get address that caused the TLB miss
-#ifdef CONFIG_DISABLE_VHPT
+	movl r17=__DIRTY_BITS|_PAGE_PL_0|_PAGE_AR_RWX
+	mov r21=cr.ipsr
 	mov r31=pr
 	;;
-	shr.u r21=r16,61			// get the region number into r21
+#ifdef CONFIG_DISABLE_VHPT
+	shr.u r22=r16,61			// get the region number into r21
 	;;
-	cmp.gt p6,p0=6,r21			// user mode 
-(p6)	br.cond.dptk.many itlb_fault
+	cmp.gt p8,p0=6,r22			// user mode 
 	;;
-	mov pr=r31,-1
-#endif
-	movl r17=__DIRTY_BITS|_PAGE_PL_0|_PAGE_AR_RX
+(p8)	thash r17=r16
 	;;
+(p8)	mov cr.iha=r17
+(p8)	br.cond.dptk.many itlb_fault
+#endif
+	extr.u r23=r21,IA64_PSR_CPL0_BIT,2	// extract psr.cpl
 	shr.u r18=r16,57	// move address bit 61 to bit 4
-	dep r16=0,r16,IA64_MAX_PHYS_BITS,(64-IA64_MAX_PHYS_BITS)	// clear ed & reserved bits
+	dep r19=0,r16,IA64_MAX_PHYS_BITS,(64-IA64_MAX_PHYS_BITS)	// clear ed & reserved bits
 	;;
 	andcm r18=0x10,r18	// bit 4=~address-bit(61)
-	dep r16=r17,r16,0,12	// insert PTE control bits into r16
+	cmp.ne p8,p0=r0,r23	// psr.cpl != 0?
+	dep r19=r17,r19,0,12	// insert PTE control bits into r19
 	;;
-	or r16=r16,r18		// set bit 4 (uncached) if the access was to region 6
+	or r19=r19,r18		// set bit 4 (uncached) if the access was to region 6
+(p8)	br.cond.spnt.many page_fault
 	;;
-	itc.i r16		// insert the TLB entry
+	itc.i r19		// insert the TLB entry
+	mov pr=r31,-1
 	rfi
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x1000 Entry 4 (size 64 bundles) Alt DTLB (7,46)
 	mov r16=cr.ifa		// get address that caused the TLB miss
-	movl r17=__DIRTY_BITS|_PAGE_PL_0|_PAGE_AR_RW
+	movl r17=__DIRTY_BITS|_PAGE_PL_0|_PAGE_AR_RWX
 	mov r20=cr.isr
 	mov r21=cr.ipsr
 	mov r31=pr
@@ -396,29 +291,40 @@
 #ifdef CONFIG_DISABLE_VHPT
 	shr.u r22=r16,61			// get the region number into r21
 	;;
-	cmp.gt p8,p0=6,r22			// user mode
+	cmp.gt p8,p0=6,r22			// access to region 0-5
+	;;
+(p8)	thash r17=r16
+	;;
+(p8)	mov cr.iha=r17
 (p8)	br.cond.dptk.many dtlb_fault
 #endif
+	extr.u r23=r21,IA64_PSR_CPL0_BIT,2	// extract psr.cpl
 	tbit.nz p6,p7=r20,IA64_ISR_SP_BIT	// is speculation bit on?
 	shr.u r18=r16,57	// move address bit 61 to bit 4
-	dep r16=0,r16,IA64_MAX_PHYS_BITS,(64-IA64_MAX_PHYS_BITS) // clear ed & reserved bits
+	dep r19=0,r16,IA64_MAX_PHYS_BITS,(64-IA64_MAX_PHYS_BITS) // clear ed & reserved bits
 	;;
-	dep r21=-1,r21,IA64_PSR_ED_BIT,1
 	andcm r18=0x10,r18	// bit 4=~address-bit(61)
-	dep r16=r17,r16,0,12	// insert PTE control bits into r16
+	cmp.ne p8,p0=r0,r23
+(p8)	br.cond.spnt.many page_fault
+
+	dep r21=-1,r21,IA64_PSR_ED_BIT,1
+	dep r19=r17,r19,0,12	// insert PTE control bits into r19
 	;;
-	or r16=r16,r18		// set bit 4 (uncached) if the access was to region 6
+	or r19=r19,r18		// set bit 4 (uncached) if the access was to region 6
 (p6)	mov cr.ipsr=r21
 	;;
-(p7)	itc.d r16		// insert the TLB entry
+(p7)	itc.d r19		// insert the TLB entry
 	mov pr=r31,-1
 	rfi
-
 	;;
 
 	//-----------------------------------------------------------------------------------
-	// call do_page_fault (predicates are in r31, psr.dt is off, r16 is faulting address)
+	// call do_page_fault (predicates are in r31, psr.dt may be off, r16 is faulting address)
 page_fault:
+	ssm psr.dt
+	;;
+	srlz.i
+	;;
 	SAVE_MIN_WITH_COVER
 	//
 	// Copy control registers to temporary registers, then turn on psr bits,
@@ -430,7 +336,7 @@
 	mov r9=cr.isr
 	adds r3=8,r2				// set up second base pointer
 	;;
-	ssm psr.ic | psr.dt
+	ssm psr.ic
 	;;
 	srlz.i					// guarantee that interrupt collection is enabled
 	;;
@@ -445,36 +351,37 @@
 	mov rp=r14
 	;;
 	adds out2=16,r12			// out2 = pointer to pt_regs
-	br.call.sptk.few b6=ia64_do_page_fault	// ignore return address
+	br.call.sptk.many b6=ia64_do_page_fault	// ignore return address
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x1400 Entry 5 (size 64 bundles) Data nested TLB (6,45)
 	//
-	// In the absence of kernel bugs, we get here when the Dirty-bit, Instruction
-	// Access-bit, or Data Access-bit faults cause a nested fault because the
-	// dTLB entry for the virtual page table isn't present.  In such a case,
-	// we lookup the pte for the faulting address by walking the page table
-	// and return to the continuation point passed in register r30.
-	// In accessing the page tables, we don't need to check for NULL entries
-	// because if the page tables didn't map the faulting address, it would not
-	// be possible to receive one of the above faults.
+	// In the absence of kernel bugs, we get here when the virtually mapped linear page
+	// table is accessed non-speculatively (e.g.,  in the Dirty-bit, Instruction
+	// Access-bit, or Data Access-bit faults).  If the DTLB entry for the virtual page
+	// table is missing, a nested TLB miss fault is triggered and control is transferred
+	// to this point.  When this happens, we lookup the pte for the faulting address
+	// by walking the page table in physical mode and return to the continuation point
+	// passed in register r30 (or call page_fault if the address is not mapped).
 	//
 	// Input:	r16:	faulting address
 	//		r29:	saved b0
 	//		r30:	continuation address
+	//		r31:	saved pr
 	//
 	// Output:	r17:	physical address of L3 PTE of faulting address
 	//		r29:	saved b0
 	//		r30:	continuation address
+	//		r31:	saved pr
 	//
-	// Clobbered:	b0, r18, r19, r21, r31, psr.dt (cleared)
+	// Clobbered:	b0, r18, r19, r21, psr.dt (cleared)
 	//
 	rsm psr.dt				// switch to using physical data addressing
 	mov r19=ar.k7				// get the page table base address
 	shl r21=r16,3				// shift bit 60 into sign bit
 	;;
-	mov r31=pr				// save the predicate registers
 	shr.u r17=r16,61			// get the region number into r17
 	;;
 	cmp.eq p6,p7=5,r17			// is faulting address in region 5?
@@ -482,26 +389,30 @@
 	;;
 (p7)	dep r17=r17,r19,(PAGE_SHIFT-3),3	// put region number bits in place
 	srlz.d
-(p6)	movl r17=__pa(SWAPPER_PGD_ADDR)		// region 5 is rooted at swapper_pg_dir
-(p6)	shr r21=r21,PGDIR_SHIFT+PAGE_SHIFT-1
-(p7)	shr r21=r21,PGDIR_SHIFT+PAGE_SHIFT-4
+(p6)	movl r19=__pa(SWAPPER_PGD_ADDR)		// region 5 is rooted at swapper_pg_dir
+(p6)	shr.u r21=r21,PGDIR_SHIFT+PAGE_SHIFT
+(p7)	shr.u r21=r21,PGDIR_SHIFT+PAGE_SHIFT-3
 	;;
-(p6)	dep r17=r18,r17,3,(PAGE_SHIFT-3)	// r17=PTA + IFA(33,42)*8
+(p6)	dep r17=r18,r19,3,(PAGE_SHIFT-3)	// r17=PTA + IFA(33,42)*8
 (p7)	dep r17=r18,r17,3,(PAGE_SHIFT-6)	// r17=PTA + (((IFA(61,63) << 7) | IFA(33,39))*8)
+	cmp.eq p7,p6=0,r21			// unused address bits all zeroes?
 	shr.u r18=r16,PMD_SHIFT			// shift L2 index into position
 	;;
-	ld8 r17=[r17]				// fetch the L1 entry
+	ld8 r17=[r17]				// fetch the L1 entry (may be 0)
 	mov b0=r30
 	;;
+(p7)	cmp.eq p6,p7=r17,r0			// was L1 entry NULL?
 	dep r17=r18,r17,3,(PAGE_SHIFT-3)	// compute address of L2 page table entry
 	;;
-	ld8 r17=[r17]				// fetch the L2 entry
+(p7)	ld8 r17=[r17]				// fetch the L2 entry (may be 0)
 	shr.u r19=r16,PAGE_SHIFT		// shift L3 index into position
 	;;
+(p7)	cmp.eq.or.andcm p6,p7=r17,r0		// was L2 entry NULL?
 	dep r17=r19,r17,3,(PAGE_SHIFT-3)	// compute address of L3 page table entry
 	;;
-	mov pr=r31,-1				// restore predicates
-	br.cond.sptk.few b0			// return to continuation point
+(p6)	br.cond.spnt.many page_fault
+	br.sptk.many b0				// return to continuation point
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -526,33 +437,19 @@
 	// a nested TLB miss hit where we look up the physical address of the L3 PTE
 	// and then continue at label 1 below.
 	//
-#ifndef CONFIG_SMP
 	mov r16=cr.ifa				// get the address that caused the fault
 	movl r30=1f				// load continuation point in case of nested fault
 	;;
 	thash r17=r16				// compute virtual address of L3 PTE
 	mov r29=b0				// save b0 in case of nested fault
-	;;
-1:	ld8 r18=[r17]
-	;;					// avoid RAW on r18
-	or r18=_PAGE_D,r18			// set the dirty bit
-	mov b0=r29				// restore b0
-	;;
-	st8 [r17]=r18				// store back updated PTE
-	itc.d r18				// install updated PTE
-#else
-	mov r16=cr.ifa				// get the address that caused the fault
-	movl r30=1f				// load continuation point in case of nested fault
-	;;
-	thash r17=r16				// compute virtual address of L3 PTE
+	mov r31=pr				// save pr
+#ifdef CONFIG_SMP
 	mov r28=ar.ccv				// save ar.ccv
-	mov r29=b0				// save b0 in case of nested fault
-	mov r27=pr
 	;;
 1:	ld8 r18=[r17]
 	;;					// avoid RAW on r18
 	mov ar.ccv=r18				// set compare value for cmpxchg
-	or r25=_PAGE_D,r18			// set the dirty bit
+	or r25=_PAGE_D|_PAGE_A,r18		// set the dirty and accessed bits
 	;;
 	cmpxchg8.acq r26=[r17],r25,ar.ccv
 	mov r24=PAGE_SHIFT<<2
@@ -568,70 +465,46 @@
 (p7)	ptc.l r16,r24
 	mov b0=r29				// restore b0
 	mov ar.ccv=r28
-	mov pr=r27,-1
+#else
+	;;
+1:	ld8 r18=[r17]
+	;;					// avoid RAW on r18
+	or r18=_PAGE_D|_PAGE_A,r18		// set the dirty and accessed bits
+	mov b0=r29				// restore b0
+	;;
+	st8 [r17]=r18				// store back updated PTE
+	itc.d r18				// install updated PTE
 #endif
+	mov pr=r31,-1				// restore pr
 	rfi
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x2400 Entry 9 (size 64 bundles) Instruction Access-bit (27)
 	// Like Entry 8, except for instruction access
 	mov r16=cr.ifa				// get the address that caused the fault
+	movl r30=1f				// load continuation point in case of nested fault
+	mov r31=pr				// save predicates
 #ifdef CONFIG_ITANIUM
 	/*
-	 * Erratum 10 (IFA may contain incorrect address) now has
-	 * "NoFix" status.  There are no plans for fixing this.
+	 * Erratum 10 (IFA may contain incorrect address) has "NoFix" status.
 	 */
 	mov r17=cr.ipsr
-	mov r31=pr				// save predicates
 	;;
 	mov r18=cr.iip
 	tbit.z p6,p0=r17,IA64_PSR_IS_BIT	// IA64 instruction set?
 	;;
 (p6)	mov r16=r18				// if so, use cr.iip instead of cr.ifa
-	mov pr=r31,-1
 #endif /* CONFIG_ITANIUM */
-
-#ifndef CONFIG_SMP
-	movl r30=1f				// load continuation point in case of nested fault
 	;;
 	thash r17=r16				// compute virtual address of L3 PTE
 	mov r29=b0				// save b0 in case of nested fault)
-	;;
-1:	ld8 r18=[r17]
-#if defined(CONFIG_IA32_SUPPORT) && \
-    (defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_B0_SPECIFIC))
-	//
-	// Erratum 85 (Access bit fault could be reported before page not present fault)
-	//   If the PTE is indicates the page is not present, then just turn this into a
-	//   page fault.
-	//
-	mov r31=pr				// save predicates
-	;;
-	tbit.nz p6,p0=r18,0			// page present bit set?
-(p6)	br.cond.sptk 1f
-	;;					// avoid WAW on p6
-	mov pr=r31,-1
-	br.cond.sptk page_fault			// page wasn't present
-1:	mov pr=r31,-1
-#else
-	;;					// avoid RAW on r18
-#endif
-	or r18=_PAGE_A,r18			// set the accessed bit
-	mov b0=r29				// restore b0
-	;;
-	st8 [r17]=r18				// store back updated PTE
-	itc.i r18				// install updated PTE
-#else
-	movl r30=1f				// load continuation point in case of nested fault
-	;;
-	thash r17=r16				// compute virtual address of L3 PTE
+#ifdef CONFIG_SMP
 	mov r28=ar.ccv				// save ar.ccv
-	mov r29=b0				// save b0 in case of nested fault)
-	mov r27=pr
 	;;
 1:	ld8 r18=[r17]
-#if defined(CONFIG_IA32_SUPPORT) && \
+# if defined(CONFIG_IA32_SUPPORT) && \
     (defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_B0_SPECIFIC))
 	//
 	// Erratum 85 (Access bit fault could be reported before page not present fault)
@@ -639,15 +512,9 @@
 	//   page fault.
 	//
 	;;
-	tbit.nz p6,p0=r18,0			// page present bit set?
-(p6)	br.cond.sptk 1f
-	;;					// avoid WAW on p6
-	mov pr=r27,-1
-	br.cond.sptk page_fault			// page wasn't present
-1:	
-#else
-	;;					// avoid RAW on r18
-#endif
+	tbit.z p6,p0=r18,_PAGE_P_BIT		// page present bit cleared?
+(p6)	br.sptk page_fault			// page wasn't present
+# endif
 	mov ar.ccv=r18				// set compare value for cmpxchg
 	or r25=_PAGE_A,r18			// set the accessed bit
 	;;
@@ -665,36 +532,42 @@
 (p7)	ptc.l r16,r24
 	mov b0=r29				// restore b0
 	mov ar.ccv=r28
-	mov pr=r27,-1
-#endif
+#else /* !CONFIG_SMP */
+	;;
+1:	ld8 r18=[r17]
+	;;
+# if defined(CONFIG_IA32_SUPPORT) && \
+    (defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_B0_SPECIFIC))
+	//
+	// Erratum 85 (Access bit fault could be reported before page not present fault)
+	//   If the PTE is indicates the page is not present, then just turn this into a
+	//   page fault.
+	//
+	tbit.z p6,p0=r18,_PAGE_P_BIT		// page present bit cleared?
+(p6)	br.sptk page_fault			// page wasn't present
+# endif
+	or r18=_PAGE_A,r18			// set the accessed bit
+	mov b0=r29				// restore b0
+	;;
+	st8 [r17]=r18				// store back updated PTE
+	itc.i r18				// install updated PTE
+#endif /* !CONFIG_SMP */
+	mov pr=r31,-1
 	rfi
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x2800 Entry 10 (size 64 bundles) Data Access-bit (15,55)
 	// Like Entry 8, except for data access
-#ifndef CONFIG_SMP
 	mov r16=cr.ifa				// get the address that caused the fault
 	movl r30=1f				// load continuation point in case of nested fault
 	;;
 	thash r17=r16				// compute virtual address of L3 PTE
+	mov r31=pr
 	mov r29=b0				// save b0 in case of nested fault)
-	;;
-1:	ld8 r18=[r17]
-	;;					// avoid RAW on r18
-	or r18=_PAGE_A,r18			// set the accessed bit
-	mov b0=r29				// restore b0
-	;;
-	st8 [r17]=r18				// store back updated PTE
-	itc.d r18				// install updated PTE
-#else
-	mov r16=cr.ifa				// get the address that caused the fault
-	movl r30=1f				// load continuation point in case of nested fault
-	;;
-	thash r17=r16				// compute virtual address of L3 PTE
+#ifdef CONFIG_SMP
 	mov r28=ar.ccv				// save ar.ccv
-	mov r29=b0				// save b0 in case of nested fault
-	mov r27=pr
 	;;
 1:	ld8 r18=[r17]
 	;;					// avoid RAW on r18
@@ -713,11 +586,20 @@
 	cmp.eq p6,p7=r18,r25			// is it same as the newly installed
 	;;
 (p7)	ptc.l r16,r24
-	mov b0=r29				// restore b0
 	mov ar.ccv=r28
-	mov pr=r27,-1
+#else
+	;;
+1:	ld8 r18=[r17]
+	;;					// avoid RAW on r18
+	or r18=_PAGE_A,r18			// set the accessed bit
+	;;
+	st8 [r17]=r18				// store back updated PTE
+	itc.d r18				// install updated PTE
 #endif
+	mov b0=r29				// restore b0
+	mov pr=r31,-1
 	rfi
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -725,16 +607,14 @@
 	mov r16=cr.iim
 	mov r17=__IA64_BREAK_SYSCALL
 	mov r31=pr		// prepare to save predicates
-	rsm psr.dt		// avoid nested faults due to TLB misses...
 	;;
-	srlz.d			// ensure everyone knows psr.dt is off...
 	cmp.eq p0,p7=r16,r17	// is this a system call? (p7 <- false, if so)
 (p7)	br.cond.spnt.many non_syscall
 
 	SAVE_MIN				// uses r31; defines r2:
 
-	// turn interrupt collection and data translation back on:
-	ssm psr.ic | psr.dt
+	// turn interrupt collection back on:
+	ssm psr.ic
 	;;
 	srlz.i					// guarantee that interrupt collection is enabled
 	cmp.eq pSys,pNonSys=r0,r0		// set pSys=1, pNonSys=0
@@ -746,14 +626,13 @@
 	adds r3=8,r2		// set up second base pointer for SAVE_REST
 	;;
 	SAVE_REST
-	;;			// avoid WAW on r2 & r3
+	br.call.sptk rp=demine_args		// clear NaT bits in (potential) syscall args
 
 	mov r3=255
 	adds r15=-1024,r15			// r15 contains the syscall number---subtract 1024
 	adds r2=IA64_TASK_PTRACE_OFFSET,r13	// r2 = &current->ptrace
-
 	;;
-	cmp.geu.unc p6,p7=r3,r15		// (syscall > 0 && syscall <= 1024+255) ?
+	cmp.geu p6,p7=r3,r15		// (syscall > 0 && syscall <= 1024+255) ?
 	movl r16=sys_call_table
 	;;
 (p6)	shladd r16=r15,3,r16
@@ -788,40 +667,61 @@
 	;;
 	st8 [r16]=r18				// store new value for cr.isr
 
-(p8)	br.call.sptk.few b6=b6			// ignore this return addr 
-	br.call.sptk.few rp=ia64_trace_syscall	// rp will be overwritten (ignored)
+(p8)	br.call.sptk.many b6=b6			// ignore this return addr 
+	br.call.sptk.many rp=ia64_trace_syscall	// rp will be overwritten (ignored)
 	// NOT REACHED
 
+	.proc demine_args
+demine_args:
+	alloc r2=ar.pfs,8,0,0,0
+	tnat.nz p8,p0=in0
+	tnat.nz p9,p0=in1
+	;;
+(p8)	mov in0=-1
+	tnat.nz p10,p0=in2
+	tnat.nz p11,p0=in3
+
+(p9)	mov in1=-1
+	tnat.nz p12,p0=in4
+	tnat.nz p13,p0=in5
+	;;
+(p10)	mov in2=-1
+	tnat.nz p14,p0=in6
+	tnat.nz p15,p0=in7
+
+(p11)	mov in3=-1
+(p12)	mov in4=-1
+(p13)	mov in5=-1
+	;;
+(p14)	mov in6=-1
+(p15)	mov in7=-1
+	br.ret.sptk.many rp
+	.endp demine_args
+
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x3000 Entry 12 (size 64 bundles) External Interrupt (4)
-	rsm psr.dt		// avoid nested faults due to TLB misses...
-	;;
-	srlz.d			// ensure everyone knows psr.dt is off...
 	mov r31=pr		// prepare to save predicates
 	;;
 
 	SAVE_MIN_WITH_COVER	// uses r31; defines r2 and r3
-	ssm psr.ic | psr.dt	// turn interrupt collection and data translation back on
+	ssm psr.ic		// turn interrupt collection
 	;;
 	adds r3=8,r2		// set up second base pointer for SAVE_REST
-	srlz.i			// ensure everybody knows psr.ic and psr.dt are back on
+	srlz.i			// ensure everybody knows psr.ic is back on
 	;;
 	SAVE_REST
 	;;
 	alloc r14=ar.pfs,0,0,2,0 // must be first in an insn group
-#ifdef CONFIG_ITANIUM_A1_SPECIFIC
-	mov out0=r0		// defer reading of cr.ivr to handle_irq...
-#else
 	mov out0=cr.ivr		// pass cr.ivr as first arg
-#endif
 	add out1=16,sp		// pass pointer to pt_regs as second arg
 	;;
 	srlz.d			// make  sure we see the effect of cr.ivr
 	movl r14=ia64_leave_kernel
 	;;
 	mov rp=r14
-	br.call.sptk.few b6=ia64_handle_irq
+	br.call.sptk.many b6=ia64_handle_irq
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -855,7 +755,7 @@
 	// The "alloc" can cause a mandatory store which could lead to
 	// an "Alt DTLB" fault which we can handle only if psr.ic is on.
 	//
-	ssm psr.ic | psr.dt
+	ssm psr.ic
 	;;
 	srlz.i		// guarantee that interrupt collection is enabled
 	;;
@@ -867,7 +767,7 @@
 	;;
 	SAVE_REST
 	;;
-	br.call.sptk.few rp=ia64_illegal_op_fault
+	br.call.sptk.many rp=ia64_illegal_op_fault
 .ret0:	;;
 	alloc r14=ar.pfs,0,0,3,0	// must be first in insn group
 	mov out0=r9
@@ -881,6 +781,7 @@
 	cmp.ne p6,p0=0,r8
 (p6)	br.call.dpnt b6=b6		// call returns to ia64_leave_kernel
 	br.sptk ia64_leave_kernel
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -900,7 +801,7 @@
 	SAVE_MIN
 	;;
 	mov r14=cr.isr
-	ssm psr.ic | psr.dt
+	ssm psr.ic
 	;;
 	srlz.i					// guarantee that interrupt collection is enabled
 	;;
@@ -913,7 +814,7 @@
 	shr r14=r14,16          // Get interrupt number
 	;; 
 	cmp.ne p6,p0=r14,r15
-(p6)    br.call.dpnt.few b6=non_ia32_syscall
+(p6)    br.call.dpnt.many b6=non_ia32_syscall
 
 	adds r14=IA64_PT_REGS_R8_OFFSET + 16,sp	// 16 byte hole per SW conventions
 	adds r15=IA64_PT_REGS_R1_OFFSET + 16,sp
@@ -924,7 +825,7 @@
 	alloc r15=ar.pfs,0,0,6,0	// must first in an insn group
 	;; 
 	ld4 r8=[r14],8          // r8 == EAX (syscall number)
-	mov r15=190		// sys_vfork - last implemented system call
+	mov r15=222		// sys_vfork - last implemented system call
 	;;
 	cmp.leu.unc p6,p7=r8,r15
 	ld4 out1=[r14],8        // r9 == ecx
@@ -961,11 +862,12 @@
 	mov out0=r14                            // interrupt #
 	add out1=16,sp                          // pointer to pt_regs
 	;;			// avoid WAW on CFM
-	br.call.sptk.few rp=ia32_bad_interrupt
+	br.call.sptk.many rp=ia32_bad_interrupt
 .ret1:	movl r15=ia64_leave_kernel
 	;;
 	mov rp=r15
 	br.ret.sptk.many rp
+	;;
 
 #endif /* CONFIG_IA32_SUPPORT */
 
@@ -985,8 +887,8 @@
 	mov r8=cr.iim			// get break immediate (must be done while psr.ic is off)
 	adds r3=8,r2			// set up second base pointer for SAVE_REST
 
-	// turn interrupt collection and data translation back on:
-	ssm psr.ic | psr.dt
+	// turn interrupt collection back on:
+	ssm psr.ic
 	;;
 	srlz.i				// guarantee that interrupt collection is enabled
 	;;
@@ -1000,7 +902,8 @@
 	SAVE_REST
 	mov rp=r15
 	;;
-	br.call.sptk.few b6=ia64_bad_break	// avoid WAW on CFM and ignore return addr
+	br.call.sptk.many b6=ia64_bad_break	// avoid WAW on CFM and ignore return addr
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -1023,7 +926,7 @@
 	// wouldn't get the state to recover.
 	//
 	mov r15=cr.ifa
-	ssm psr.ic | psr.dt
+	ssm psr.ic
 	;;
 	srlz.i					// guarantee that interrupt collection is enabled
 	;;
@@ -1039,7 +942,8 @@
 	adds out1=16,sp				// out1 = pointer to pt_regs
 	;;
 	mov rp=r14
-	br.sptk.few ia64_prepare_handle_unaligned
+	br.sptk.many ia64_prepare_handle_unaligned
+	;;
 
 	.align 1024
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -1055,7 +959,6 @@
 	//
 	// Input:
 	//	psr.ic:	off
-	//	psr.dt:	off
 	//	r19:	fault vector number (e.g., 24 for General Exception)
 	//	r31:	contains saved predicates (pr)
 	//
@@ -1071,7 +974,7 @@
 	mov r10=cr.iim
 	mov r11=cr.itir
 	;;
-	ssm psr.ic | psr.dt
+	ssm psr.ic
 	;;
 	srlz.i					// guarantee that interrupt collection is enabled
 	;;
@@ -1089,7 +992,9 @@
 	movl r14=ia64_leave_kernel
 	;;
 	mov rp=r14
-	br.call.sptk.few b6=ia64_fault
+	br.call.sptk.many b6=ia64_fault
+	;;
+
 //
 // --- End of long entries, Beginning of short entries
 //
@@ -1099,16 +1004,16 @@
 // 0x5000 Entry 20 (size 16 bundles) Page Not Present (10,22,49)
 	mov r16=cr.ifa
 	rsm psr.dt
-#if 1
-	// If you disable this, you MUST re-enable to update_mmu_cache() code in pgtable.h
+	// The Linux page fault handler doesn't expect non-present pages to be in
+	// the TLB.  Flush the existing entry now, so we meet that expectation.
 	mov r17=_PAGE_SIZE_4K<<2
 	;;
 	ptc.l r16,r17
-#endif
 	;;
 	mov r31=pr
 	srlz.d
-	br.cond.sptk.many page_fault
+	br.sptk.many page_fault
+	;;
 
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -1118,7 +1023,8 @@
 	mov r31=pr
 	;;
 	srlz.d
-	br.cond.sptk.many page_fault
+	br.sptk.many page_fault
+	;;
 
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -1128,7 +1034,8 @@
 	mov r31=pr
 	;;
 	srlz.d
-	br.cond.sptk.many page_fault
+	br.sptk.many page_fault
+	;;
 
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -1138,31 +1045,32 @@
 	mov r31=pr
 	;;
 	srlz.d
-	br.cond.sptk.many page_fault
+	br.sptk.many page_fault
+	;;
 
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x5400 Entry 24 (size 16 bundles) General Exception (5,32,34,36,38,39)
 	mov r16=cr.isr
 	mov r31=pr
-	rsm psr.dt		// avoid nested faults due to TLB misses...
 	;;
-	srlz.d			// ensure everyone knows psr.dt is off...
 	cmp4.eq p6,p0=0,r16
 (p6)	br.sptk dispatch_illegal_op_fault
 	;;
 	mov r19=24		// fault number
-	br.cond.sptk.many dispatch_to_fault_handler
+	br.sptk.many dispatch_to_fault_handler
+	;;
 
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x5500 Entry 25 (size 16 bundles) Disabled FP-Register (35)
-	rsm psr.dt | psr.dfh			// ensure we can access fph
+	rsm psr.dfh		// ensure we can access fph
 	;;
 	srlz.d
 	mov r31=pr
 	mov r19=25
-	br.cond.sptk.many dispatch_to_fault_handler
+	br.sptk.many dispatch_to_fault_handler
+	;;
 
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -1204,6 +1112,7 @@
 	;;
 
 	rfi				// and go back
+	;;
 
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -1218,12 +1127,11 @@
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x5a00 Entry 30 (size 16 bundles) Unaligned Reference (57)
-	rsm psr.dt		// avoid nested faults due to TLB misses...
 	mov r16=cr.ipsr
 	mov r31=pr		// prepare to save predicates
 	;;									
-	srlz.d			// ensure everyone knows psr.dt is off
-	br.cond.sptk.many dispatch_unaligned_handler
+	br.sptk.many dispatch_unaligned_handler
+	;;
 
 	.align 256
 /////////////////////////////////////////////////////////////////////////////////////////
@@ -1304,9 +1212,6 @@
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x6a00 Entry 46 (size 16 bundles) IA-32 Intercept  (30,31,59,70,71)
 #ifdef	CONFIG_IA32_SUPPORT
-	rsm psr.dt
-	;;
-	srlz.d
 	mov r31=pr
 	mov r16=cr.isr
 	;;
@@ -1325,7 +1230,7 @@
 	;;
 	mov pr=r31,-1		// restore predicate registers
 	rfi
-
+	;;
 1:
 #endif	// CONFIG_IA32_SUPPORT
 	FAULT(46)
@@ -1334,11 +1239,9 @@
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x6b00 Entry 47 (size 16 bundles) IA-32 Interrupt  (74)
 #ifdef CONFIG_IA32_SUPPORT
-	rsm psr.dt
-	;;
-	srlz.d
 	mov r31=pr
-	br.cond.sptk.many dispatch_to_ia32_handler
+	br.sptk.many dispatch_to_ia32_handler
+	;;
 #else
 	FAULT(47)
 #endif
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/machvec.c linux/arch/ia64/kernel/machvec.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/machvec.c	Sun Aug 13 10:17:16 2000
+++ linux/arch/ia64/kernel/machvec.c	Thu Jan  4 12:50:17 2001
@@ -1,10 +1,12 @@
 #include <linux/config.h>
+
+#ifdef CONFIG_IA64_GENERIC
+
 #include <linux/kernel.h>
+#include <linux/string.h>
 
 #include <asm/page.h>
 #include <asm/machvec.h>
-
-#ifdef CONFIG_IA64_GENERIC
 
 struct ia64_machine_vector ia64_mv;
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/mca.c linux/arch/ia64/kernel/mca.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/mca.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/mca.c	Thu Jan  4 12:50:17 2001
@@ -19,6 +19,7 @@
 #include <linux/irq.h>
 #include <linux/smp_lock.h>
 
+#include <asm/machvec.h>
 #include <asm/page.h>
 #include <asm/ptrace.h>
 #include <asm/system.h>
@@ -365,7 +366,7 @@
 void
 ia64_mca_wakeup(int cpu)
 {
-	ipi_send(cpu, IA64_MCA_WAKEUP_INT_VECTOR, IA64_IPI_DM_INT, 0);
+	platform_send_ipi(cpu, IA64_MCA_WAKEUP_INT_VECTOR, IA64_IPI_DM_INT, 0);
 	ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_NOTDONE;
 	
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/mca_asm.S linux/arch/ia64/kernel/mca_asm.S
--- v2.4.0-prerelease/linux/arch/ia64/kernel/mca_asm.S	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/mca_asm.S	Thu Jan  4 12:50:17 2001
@@ -3,11 +3,10 @@
 //
 // Mods by cfleck to integrate into kernel build
 // 00/03/15 davidm Added various stop bits to get a clean compile
-// 00/03/29 cfleck Added code to save INIT handoff state in pt_regs format, switch to temp kstack,
-//		   switch modes, jump to C INIT handler
 //
-#include <linux/config.h>
-
+// 00/03/29 cfleck Added code to save INIT handoff state in pt_regs format, switch to temp
+//		   kstack, switch modes, jump to C INIT handler
+//
 #include <asm/pgtable.h>
 #include <asm/processor.h>
 #include <asm/mca_asm.h>
@@ -17,14 +16,7 @@
  * When we get an machine check, the kernel stack pointer is no longer
  * valid, so we need to set a new stack pointer.
  */
-#define MINSTATE_START_SAVE_MIN							\
-(pKern) movl sp=ia64_init_stack+IA64_STK_OFFSET-IA64_PT_REGS_SIZE;		\
-	;;
-
-#define MINSTATE_END_SAVE_MIN							\
-	or r12=r12,r14;		/* make sp a kernel virtual address */		\
-	or r13=r13,r14;		/* make `current' a kernel virtual address */	\
-	;;
+#define	MINSTATE_PHYS	/* Make sure stack access is physical for MINSTATE */ 
 
 #include "minstate.h"
 	
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/minstate.h linux/arch/ia64/kernel/minstate.h
--- v2.4.0-prerelease/linux/arch/ia64/kernel/minstate.h	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/minstate.h	Thu Jan  4 12:50:17 2001
@@ -20,6 +20,72 @@
 #define rR1		r20
 
 /*
+ * Here start the source dependent macros.
+ */
+
+/*
+ * For ivt.s we want to access the stack virtually so we dont have to disable translation
+ * on interrupts.
+ */
+#define MINSTATE_START_SAVE_MIN_VIRT								\
+	dep r1=-1,r1,61,3;				/* r1 = current (virtual) */		\
+(p7)	mov ar.rsc=r0;		/* set enforced lazy mode, pl 0, little-endian, loadrs=0 */	\
+	;;											\
+(p7)	addl rKRBS=IA64_RBS_OFFSET,r1;			/* compute base of RBS */		\
+(p7)	mov rARRNAT=ar.rnat;									\
+(pKern) mov r1=sp;					/* get sp  */				\
+	;;											\
+(p7)	addl r1=IA64_STK_OFFSET-IA64_PT_REGS_SIZE,r1;	/* compute base of memory stack */	\
+(p7)	mov rARBSPSTORE=ar.bspstore;			/* save ar.bspstore */			\
+	;;											\
+(pKern) addl r1=-IA64_PT_REGS_SIZE,r1;			/* if in kernel mode, use sp (r12) */	\
+(p7)	mov ar.bspstore=rKRBS;				/* switch to kernel RBS */		\
+	;;											\
+(p7)	mov r18=ar.bsp;										\
+(p7)	mov ar.rsc=0x3;		/* set eager mode, pl 0, little-endian, loadrs=0 */		\
+
+#define MINSTATE_END_SAVE_MIN_VIRT								\
+	or r13=r13,r14;		/* make `current' a kernel virtual address */			\
+	bsw.1;			/* switch back to bank 1 (must be last in insn group) */	\
+	;;
+
+/*
+ * For mca_asm.S we want to access the stack physically since the state is saved before we
+ * go virtual and dont want to destroy the iip or ipsr.
+ */
+#define MINSTATE_START_SAVE_MIN_PHYS								\
+(pKern) movl sp=ia64_init_stack+IA64_STK_OFFSET-IA64_PT_REGS_SIZE;				\
+(p7)	mov ar.rsc=r0;		/* set enforced lazy mode, pl 0, little-endian, loadrs=0 */	\
+(p7)	addl rKRBS=IA64_RBS_OFFSET,r1;		/* compute base of register backing store */	\
+	;;											\
+(p7)	mov rARRNAT=ar.rnat;									\
+(pKern) dep r1=0,sp,61,3;				/* compute physical addr of sp	*/	\
+(p7)	addl r1=IA64_STK_OFFSET-IA64_PT_REGS_SIZE,r1;	/* compute base of memory stack */	\
+(p7)	mov rARBSPSTORE=ar.bspstore;			/* save ar.bspstore */			\
+(p7)	dep rKRBS=-1,rKRBS,61,3;			/* compute kernel virtual addr of RBS */\
+	;;											\
+(pKern) addl r1=-IA64_PT_REGS_SIZE,r1;		/* if in kernel mode, use sp (r12) */		\
+(p7)	mov ar.bspstore=rKRBS;			/* switch to kernel RBS */			\
+	;;											\
+(p7)	mov r18=ar.bsp;										\
+(p7)	mov ar.rsc=0x3;		/* set eager mode, pl 0, little-endian, loadrs=0 */		\
+
+#define MINSTATE_END_SAVE_MIN_PHYS								\
+	or r12=r12,r14;		/* make sp a kernel virtual address */				\
+	or r13=r13,r14;		/* make `current' a kernel virtual address */			\
+	;;
+
+#ifdef MINSTATE_VIRT
+# define MINSTATE_START_SAVE_MIN	MINSTATE_START_SAVE_MIN_VIRT
+# define MINSTATE_END_SAVE_MIN		MINSTATE_END_SAVE_MIN_VIRT
+#endif
+
+#ifdef MINSTATE_PHYS
+# define MINSTATE_START_SAVE_MIN	MINSTATE_START_SAVE_MIN_PHYS
+# define MINSTATE_END_SAVE_MIN		MINSTATE_END_SAVE_MIN_PHYS
+#endif
+
+/*
  * DO_SAVE_MIN switches to the kernel stacks (if necessary) and saves
  * the minimum state necessary that allows us to turn psr.ic back
  * on.
@@ -31,7 +97,6 @@
  *
  * Upon exit, the state is as follows:
  *	psr.ic: off
- *	psr.dt: off
  *	r2 = points to &pt_regs.r16
  *	r12 = kernel sp (kernel virtual address)
  *	r13 = points to current task_struct (kernel virtual address)
@@ -50,7 +115,7 @@
 	mov rCRIPSR=cr.ipsr;									  \
 	mov rB6=b6;		/* rB6 = branch reg 6 */					  \
 	mov rCRIIP=cr.iip;									  \
-	mov r1=ar.k6;		/* r1 = current */						  \
+	mov r1=ar.k6;		/* r1 = current (physical) */					  \
 	;;											  \
 	invala;											  \
 	extr.u r16=rCRIPSR,32,2;		/* extract psr.cpl */				  \
@@ -58,25 +123,11 @@
 	cmp.eq pKern,p7=r0,r16;			/* are we in kernel mode already? (psr.cpl==0) */ \
 	/* switch from user to kernel RBS: */							  \
 	COVER;											  \
-	;; 									                  \
-	MINSTATE_START_SAVE_MIN									  \
-(p7)	mov ar.rsc=r0;		/* set enforced lazy mode, pl 0, little-endian, loadrs=0 */	  \
-(p7)	addl rKRBS=IA64_RBS_OFFSET,r1;		/* compute base of register backing store */	  \
-	;;											  \
-(p7)	mov rARRNAT=ar.rnat;									  \
-(pKern)	dep r1=0,sp,61,3;				/* compute physical addr of sp  */	  \
-(p7)	addl r1=IA64_STK_OFFSET-IA64_PT_REGS_SIZE,r1;	/* compute base of memory stack */	  \
-(p7)	mov rARBSPSTORE=ar.bspstore;			/* save ar.bspstore */			  \
-(p7)	dep rKRBS=-1,rKRBS,61,3;			/* compute kernel virtual addr of RBS */  \
 	;;											  \
-(pKern)	addl r1=-IA64_PT_REGS_SIZE,r1;		/* if in kernel mode, use sp (r12) */		  \
-(p7)	mov ar.bspstore=rKRBS;			/* switch to kernel RBS */			  \
+	MINSTATE_START_SAVE_MIN									  \
 	;;											  \
-(p7)	mov r18=ar.bsp;										  \
-(p7)	mov ar.rsc=0x3;		/* set eager mode, pl 0, little-endian, loadrs=0 */		  \
-												  \
-	mov r16=r1;		/* initialize first base pointer */				  \
-	adds r17=8,r1;		/* initialize second base pointer */				  \
+	mov r16=r1;					/* initialize first base pointer */	  \
+	adds r17=8,r1;					/* initialize second base pointer */	  \
 	;;											  \
 	st8 [r16]=rCRIPSR,16;	/* save cr.ipsr */						  \
 	st8 [r17]=rCRIIP,16;	/* save cr.iip */						  \
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/pal.S linux/arch/ia64/kernel/pal.S
--- v2.4.0-prerelease/linux/arch/ia64/kernel/pal.S	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/pal.S	Thu Jan  4 12:50:17 2001
@@ -52,10 +52,9 @@
 /*
  * Make a PAL call using the static calling convention.
  *
- * in0         Pointer to struct ia64_pal_retval
- * in1         Index of PAL service
- * in2 - in4   Remaining PAL arguments
- * in5	       1 ==> clear psr.ic,  0 ==> don't clear psr.ic
+ * in0         Index of PAL service
+ * in1 - in3   Remaining PAL arguments
+ * in4	       1 ==> clear psr.ic,  0 ==> don't clear psr.ic
  *
  */
 GLOBAL_ENTRY(ia64_pal_call_static)
@@ -69,7 +68,7 @@
 	}
 	;;
 	ld8 loc2 = [loc2]		// loc2 <- entry point
-	tbit.nz p6,p7 = in5, 0
+	tbit.nz p6,p7 = in4, 0
 	adds r8 = 1f-1b,r8
 	;;
 	mov loc3 = psr
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/palinfo.c linux/arch/ia64/kernel/palinfo.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/palinfo.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/palinfo.c	Thu Jan  4 12:50:17 2001
@@ -16,7 +16,6 @@
  *	- as of 2.2.9/2.2.12, the following values are still wrong
  *		PAL_VM_SUMMARY: key & rid sizes
  */
-#include <linux/config.h>
 #include <linux/types.h>
 #include <linux/errno.h>
 #include <linux/init.h>
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/pci-dma.c linux/arch/ia64/kernel/pci-dma.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/pci-dma.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/pci-dma.c	Wed Dec 31 16:00:00 1969
@@ -1,517 +0,0 @@
-/*
- * Dynamic DMA mapping support.
- *
- * This implementation is for IA-64 platforms that do not support
- * I/O TLBs (aka DMA address translation hardware).
- * Copyright (C) 2000 Asit Mallick <Asit.K.Mallick@intel.com>
- * Copyright (C) 2000 Goutham Rao <goutham.rao@intel.com>
- */
-
-#include <linux/config.h>
-
-#include <linux/mm.h>
-#include <linux/pci.h>
-#include <linux/spinlock.h>
-#include <linux/string.h>
-#include <linux/types.h>
-
-#include <asm/io.h>
-#include <asm/pci.h>
-#include <asm/dma.h>
-
-#ifdef CONFIG_SWIOTLB
-
-#include <linux/init.h>
-#include <linux/bootmem.h>
-
-#define ALIGN(val, align) ((unsigned long) (((unsigned long) (val) + ((align) - 1)) & ~((align) - 1)))
-
-/*
- * log of the size of each IO TLB slab.  The number of slabs is command line
- * controllable.
- */
-#define IO_TLB_SHIFT 11
-
-/*
- * Used to do a quick range check in pci_unmap_single and pci_sync_single, to see if the 
- * memory was in fact allocated by this API.
- */
-static char *io_tlb_start, *io_tlb_end;
-
-/*
- * The number of IO TLB blocks (in groups of 64) betweeen io_tlb_start and io_tlb_end.
- * This is command line adjustable via setup_io_tlb_npages.
- */
-unsigned long io_tlb_nslabs = 1024;
-
-/*
- * This is a free list describing the number of free entries available from each index
- */
-static unsigned int *io_tlb_list;
-static unsigned int io_tlb_index;
-
-/*
- * We need to save away the original address corresponding to a mapped entry for the sync 
- * operations.
- */
-static unsigned char **io_tlb_orig_addr;
-
-/*
- * Protect the above data structures in the map and unmap calls
- */ 
-spinlock_t io_tlb_lock = SPIN_LOCK_UNLOCKED;
-
-static int __init
-setup_io_tlb_npages (char *str)
-{
-	io_tlb_nslabs = simple_strtoul(str, NULL, 0) << (PAGE_SHIFT - IO_TLB_SHIFT);
-	return 1;
-}
-__setup("swiotlb=", setup_io_tlb_npages);
-
-/*
- * Statically reserve bounce buffer space and initialize bounce buffer
- * data structures for the software IO TLB used to implement the PCI DMA API
- */
-void
-setup_swiotlb (void)
-{
-	int i;
-
-	/*
-	 * Get IO TLB memory from the low pages
-	 */
-	io_tlb_start = alloc_bootmem_low_pages(io_tlb_nslabs * (1 << IO_TLB_SHIFT));
-	if (!io_tlb_start)
-		BUG();
-	io_tlb_end = io_tlb_start + io_tlb_nslabs * (1 << IO_TLB_SHIFT);
-
-	/*
-	 * Allocate and initialize the free list array.  This array is used
-	 * to find contiguous free memory regions of size 2^IO_TLB_SHIFT between
-	 * io_tlb_start and io_tlb_end.
-	 */
-	io_tlb_list = alloc_bootmem(io_tlb_nslabs * sizeof(int));
-	for (i = 0; i < io_tlb_nslabs; i++)
-		io_tlb_list[i] = io_tlb_nslabs - i;
-	io_tlb_index = 0;
-	io_tlb_orig_addr = alloc_bootmem(io_tlb_nslabs * sizeof(char *));
-
-	printk("Placing software IO TLB between 0x%p - 0x%p\n",
-	       (void *) io_tlb_start, (void *) io_tlb_end);
-}
-
-/*
- * Allocates bounce buffer and returns its kernel virtual address.
- */
-static void *
-__pci_map_single (struct pci_dev *hwdev, char *buffer, size_t size, int direction)
-{
-	unsigned long flags;
-	char *dma_addr;
-	unsigned int i, nslots, stride, index, wrap;
-
-	/*
-	 * For mappings greater than a page size, we limit the stride (and hence alignment)
-	 * to a page size.
-	 */
-	nslots = ALIGN(size, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT;
-	if (size > (1 << PAGE_SHIFT))
-		stride = (1 << (PAGE_SHIFT - IO_TLB_SHIFT));
-	else
-		stride = nslots;
-
-	if (!nslots)
-		BUG();
-
-	/*
-	 * Find suitable number of IO TLB entries size that will fit this request and allocate a buffer
-	 * from that IO TLB pool.
-	 */
-	spin_lock_irqsave(&io_tlb_lock, flags);
-	{
-		wrap = index = ALIGN(io_tlb_index, stride);
-		do {
-			/*
-			 * If we find a slot that indicates we have 'nslots' number of 
-			 * contiguous buffers, we allocate the buffers from that slot and mark the
-			 * entries as '0' indicating unavailable.
-			 */
-			if (io_tlb_list[index] >= nslots) {
-				for (i = index; i < index + nslots; i++)
-					io_tlb_list[i] = 0;
-				dma_addr = io_tlb_start + (index << IO_TLB_SHIFT);
-
-				/*
-				 * Update the indices to avoid searching in the next round.
-				 */
-				io_tlb_index = (index + nslots) < io_tlb_nslabs ? (index + nslots) : 0;
-
-				goto found;
-			}
-			index += stride;
-			if (index >= io_tlb_nslabs)
-				index = 0;
-		} while (index != wrap);
-
-		/*
-		 * XXX What is a suitable recovery mechanism here?  We cannot 
-		 * sleep because we are called from with in interrupts!
-		 */
-		panic("__pci_map_single: could not allocate software IO TLB (%ld bytes)", size);
-found:
-	}
-	spin_unlock_irqrestore(&io_tlb_lock, flags);
-
-	/*
-	 * Save away the mapping from the original address to the DMA address.  This is needed
-	 * when we sync the memory.  Then we sync the buffer if needed.
-	 */
-	io_tlb_orig_addr[index] = buffer;
-	if (direction == PCI_DMA_TODEVICE || direction == PCI_DMA_BIDIRECTIONAL)
-		memcpy(dma_addr, buffer, size);
-
-	return dma_addr;
-}
-
-/*
- * dma_addr is the kernel virtual address of the bounce buffer to unmap.
- */
-static void
-__pci_unmap_single (struct pci_dev *hwdev, char *dma_addr, size_t size, int direction)
-{
-	unsigned long flags;
-	int i, nslots = ALIGN(size, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT;
-	int index = (dma_addr - io_tlb_start) >> IO_TLB_SHIFT;
-	char *buffer = io_tlb_orig_addr[index];
-
-	/*
-	 * First, sync the memory before unmapping the entry
-	 */
-	if ((direction == PCI_DMA_FROMDEVICE) || (direction == PCI_DMA_BIDIRECTIONAL))
-		/*
- 	 	 * bounce... copy the data back into the original buffer
-	  	 * and delete the bounce buffer.
- 	 	 */
-		memcpy(buffer, dma_addr, size);
-
-	/*
-	 * Return the buffer to the free list by setting the corresponding entries to indicate
-	 * the number of contigous entries available.  
-	 * While returning the entries to the free list, we merge the entries with slots below
-	 * and above the pool being returned.
-	 */
-	spin_lock_irqsave(&io_tlb_lock, flags);
-	{
-		int count = ((index + nslots) < io_tlb_nslabs ? io_tlb_list[index + nslots] : 0);
-		/*
-		 * Step 1: return the slots to the free list, merging the slots with superceeding slots
-		 */
-		for (i = index + nslots - 1; i >= index; i--)
-			io_tlb_list[i] = ++count;
-		/*
-		 * Step 2: merge the returned slots with the preceeding slots, if available (non zero)
-		 */
-		for (i = index - 1; (i >= 0) && io_tlb_list[i]; i--)
-			io_tlb_list[i] += io_tlb_list[index];
-	}
-	spin_unlock_irqrestore(&io_tlb_lock, flags);
-}
-
-static void
-__pci_sync_single (struct pci_dev *hwdev, char *dma_addr, size_t size, int direction)
-{
-	int index = (dma_addr - io_tlb_start) >> IO_TLB_SHIFT;
-	char *buffer = io_tlb_orig_addr[index];
-
-	/*
-  	 * bounce... copy the data back into/from the original buffer
-	 * XXX How do you handle PCI_DMA_BIDIRECTIONAL here ?
- 	 */
-	if (direction == PCI_DMA_FROMDEVICE)
-		memcpy(buffer, dma_addr, size);
-	else if (direction == PCI_DMA_TODEVICE)
-		memcpy(dma_addr, buffer, size);
-	else
-		BUG();
-}
-
-/*
- * Map a single buffer of the indicated size for DMA in streaming mode.
- * The PCI address to use is returned.
- *
- * Once the device is given the dma address, the device owns this memory
- * until either pci_unmap_single or pci_dma_sync_single is performed.
- */
-dma_addr_t
-pci_map_single (struct pci_dev *hwdev, void *ptr, size_t size, int direction)
-{
-	unsigned long pci_addr = virt_to_phys(ptr);
-
-	if (direction == PCI_DMA_NONE)
-		BUG();
-	/*
-	 * Check if the PCI device can DMA to ptr... if so, just return ptr
-	 */
-	if ((pci_addr & ~hwdev->dma_mask) == 0)
-		/*
-		 * Device is bit capable of DMA'ing to the
-		 * buffer... just return the PCI address of ptr
-		 */
-		return pci_addr;
-
-	/* 
-	 * get a bounce buffer: 
-	 */
-	pci_addr = virt_to_phys(__pci_map_single(hwdev, ptr, size, direction));
-
-	/*
-	 * Ensure that the address returned is DMA'ble:
-	 */
-	if ((pci_addr & ~hwdev->dma_mask) != 0)
-		panic("__pci_map_single: bounce buffer is not DMA'ble");
-
-	return pci_addr;
-}
-
-/*
- * Unmap a single streaming mode DMA translation.  The dma_addr and size
- * must match what was provided for in a previous pci_map_single call.  All
- * other usages are undefined.
- *
- * After this call, reads by the cpu to the buffer are guarenteed to see
- * whatever the device wrote there.
- */
-void
-pci_unmap_single (struct pci_dev *hwdev, dma_addr_t pci_addr, size_t size, int direction)
-{
-	char *dma_addr = phys_to_virt(pci_addr);
-
-	if (direction == PCI_DMA_NONE)
-		BUG();
-	if (dma_addr >= io_tlb_start && dma_addr < io_tlb_end)
-		__pci_unmap_single(hwdev, dma_addr, size, direction);
-}
-
-/*
- * Make physical memory consistent for a single
- * streaming mode DMA translation after a transfer.
- *
- * If you perform a pci_map_single() but wish to interrogate the
- * buffer using the cpu, yet do not wish to teardown the PCI dma
- * mapping, you must call this function before doing so.  At the
- * next point you give the PCI dma address back to the card, the
- * device again owns the buffer.
- */
-void
-pci_dma_sync_single (struct pci_dev *hwdev, dma_addr_t pci_addr, size_t size, int direction)
-{
-	char *dma_addr = phys_to_virt(pci_addr);
-
-	if (direction == PCI_DMA_NONE)
-		BUG();
-	if (dma_addr >= io_tlb_start && dma_addr < io_tlb_end)
-		__pci_sync_single(hwdev, dma_addr, size, direction);
-}
-
-/*
- * Map a set of buffers described by scatterlist in streaming
- * mode for DMA.  This is the scather-gather version of the
- * above pci_map_single interface.  Here the scatter gather list
- * elements are each tagged with the appropriate dma address
- * and length.  They are obtained via sg_dma_{address,length}(SG).
- *
- * NOTE: An implementation may be able to use a smaller number of
- *       DMA address/length pairs than there are SG table elements.
- *       (for example via virtual mapping capabilities)
- *       The routine returns the number of addr/length pairs actually
- *       used, at most nents.
- *
- * Device ownership issues as mentioned above for pci_map_single are
- * the same here.
- */
-int
-pci_map_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction)
-{
-	int i;
-
-	if (direction == PCI_DMA_NONE)
-		BUG();
-
-	for (i = 0; i < nelems; i++, sg++) {
-		sg->orig_address = sg->address;
-		if ((virt_to_phys(sg->address) & ~hwdev->dma_mask) != 0) {
-			sg->address = __pci_map_single(hwdev, sg->address, sg->length, direction);
-		}
-	}
-	return nelems;
-}
-
-/*
- * Unmap a set of streaming mode DMA translations.
- * Again, cpu read rules concerning calls here are the same as for
- * pci_unmap_single() above.
- */
-void
-pci_unmap_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction)
-{
-	int i;
-
-	if (direction == PCI_DMA_NONE)
-		BUG();
-
-	for (i = 0; i < nelems; i++, sg++)
-		if (sg->orig_address != sg->address) {
-			__pci_unmap_single(hwdev, sg->address, sg->length, direction);
-			sg->address = sg->orig_address;
-		}
-}
-
-/*
- * Make physical memory consistent for a set of streaming mode DMA
- * translations after a transfer.
- *
- * The same as pci_dma_sync_single but for a scatter-gather list,
- * same rules and usage.
- */
-void
-pci_dma_sync_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction)
-{
-	int i;
-
-	if (direction == PCI_DMA_NONE)
-		BUG();
-
-	for (i = 0; i < nelems; i++, sg++)
-		if (sg->orig_address != sg->address)
-			__pci_sync_single(hwdev, sg->address, sg->length, direction);
-}
-
-#else
-/*
- * Map a single buffer of the indicated size for DMA in streaming mode.
- * The 32-bit bus address to use is returned.
- *
- * Once the device is given the dma address, the device owns this memory
- * until either pci_unmap_single or pci_dma_sync_single is performed.
- */
-dma_addr_t
-pci_map_single (struct pci_dev *hwdev, void *ptr, size_t size, int direction)
-{
-        if (direction == PCI_DMA_NONE)
-                BUG();
-        return virt_to_bus(ptr);
-}
-
-/*
- * Unmap a single streaming mode DMA translation.  The dma_addr and size
- * must match what was provided for in a previous pci_map_single call.  All
- * other usages are undefined.
- *
- * After this call, reads by the cpu to the buffer are guarenteed to see
- * whatever the device wrote there.
- */
-void
-pci_unmap_single (struct pci_dev *hwdev, dma_addr_t dma_addr, size_t size, int direction)
-{
-        if (direction == PCI_DMA_NONE)
-                BUG();
-        /* Nothing to do */
-}
-/*
- * Map a set of buffers described by scatterlist in streaming
- * mode for DMA.  This is the scather-gather version of the
- * above pci_map_single interface.  Here the scatter gather list
- * elements are each tagged with the appropriate dma address
- * and length.  They are obtained via sg_dma_{address,length}(SG).
- *
- * NOTE: An implementation may be able to use a smaller number of
- *       DMA address/length pairs than there are SG table elements.
- *       (for example via virtual mapping capabilities)
- *       The routine returns the number of addr/length pairs actually
- *       used, at most nents.
- *
- * Device ownership issues as mentioned above for pci_map_single are
- * the same here.
- */
-int
-pci_map_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nents, int direction)
-{
-        if (direction == PCI_DMA_NONE)
-                BUG();
-        return nents;
-}
-
-/*
- * Unmap a set of streaming mode DMA translations.
- * Again, cpu read rules concerning calls here are the same as for
- * pci_unmap_single() above.
- */
-void
-pci_unmap_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nents, int direction)
-{
-        if (direction == PCI_DMA_NONE)
-                BUG();
-        /* Nothing to do */
-}
-/*
- * Make physical memory consistent for a single
- * streaming mode DMA translation after a transfer.
- *
- * If you perform a pci_map_single() but wish to interrogate the
- * buffer using the cpu, yet do not wish to teardown the PCI dma
- * mapping, you must call this function before doing so.  At the
- * next point you give the PCI dma address back to the card, the
- * device again owns the buffer.
- */
-void
-pci_dma_sync_single (struct pci_dev *hwdev, dma_addr_t dma_handle, size_t size, int direction)
-{
-        if (direction == PCI_DMA_NONE)
-                BUG();
-        /* Nothing to do */
-}
-
-/*
- * Make physical memory consistent for a set of streaming mode DMA
- * translations after a transfer.
- *
- * The same as pci_dma_sync_single but for a scatter-gather list,
- * same rules and usage.
- */
-void
-pci_dma_sync_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction)
-{
-        if (direction == PCI_DMA_NONE)
-                BUG();
-        /* Nothing to do */
-}
-
-#endif /* CONFIG_SWIOTLB */
-
-void *
-pci_alloc_consistent (struct pci_dev *hwdev, size_t size, dma_addr_t *dma_handle)
-{
-	unsigned long pci_addr;
-	int gfp = GFP_ATOMIC;
-	void *ret;
-
-	if (!hwdev || hwdev->dma_mask <= 0xffffffff)
-		gfp |= GFP_DMA; /* XXX fix me: should change this to GFP_32BIT or ZONE_32BIT */
-	ret = (void *)__get_free_pages(gfp, get_order(size));
-	if (!ret)
-		return NULL;
-
-	memset(ret, 0, size);
-	pci_addr = virt_to_phys(ret);
-	if ((pci_addr & ~hwdev->dma_mask) != 0)
-		panic("pci_alloc_consistent: allocated memory is out of range for PCI device");
-	*dma_handle = pci_addr;
-	return ret;
-}
-
-void
-pci_free_consistent (struct pci_dev *hwdev, size_t size, void *vaddr, dma_addr_t dma_handle)
-{
-	free_pages((unsigned long) vaddr, get_order(size));
-}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/pci.c linux/arch/ia64/kernel/pci.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/pci.c	Mon Aug  7 14:31:40 2000
+++ linux/arch/ia64/kernel/pci.c	Thu Jan  4 12:50:17 2001
@@ -1,10 +1,8 @@
 /*
- * pci.c - Low-Level PCI Access in IA64
+ * pci.c - Low-Level PCI Access in IA-64
  * 
  * Derived from bios32.c of i386 tree.
- *
  */
-
 #include <linux/config.h>
 
 #include <linux/types.h>
@@ -44,19 +42,16 @@
  * This interrupt-safe spinlock protects all accesses to PCI
  * configuration space.
  */
-
 spinlock_t pci_lock = SPIN_LOCK_UNLOCKED;
 
-struct pci_fixup pcibios_fixups[] = { { 0 } };
-
-#define PCI_NO_CHECKS		0x400
-#define PCI_NO_PEER_FIXUP	0x800
-
-static unsigned int pci_probe = PCI_NO_CHECKS;
+struct pci_fixup pcibios_fixups[] = {
+	{ 0 }
+};
 
 /* Macro to build a PCI configuration address to be passed as a parameter to SAL. */
 
-#define PCI_CONFIG_ADDRESS(dev, where) (((u64) dev->bus->number << 16) | ((u64) (dev->devfn & 0xff) << 8) | (where & 0xff))
+#define PCI_CONFIG_ADDRESS(dev, where) \
+	(((u64) dev->bus->number << 16) | ((u64) (dev->devfn & 0xff) << 8) | (where & 0xff))
 
 static int 
 pci_conf_read_config_byte(struct pci_dev *dev, int where, u8 *value)
@@ -109,8 +104,7 @@
 	return ia64_sal_pci_config_write(PCI_CONFIG_ADDRESS(dev, where), 4, value);
 }
 
-
-static struct pci_ops pci_conf = {
+struct pci_ops pci_conf = {
       pci_conf_read_config_byte,
       pci_conf_read_config_word,
       pci_conf_read_config_dword,
@@ -120,36 +114,21 @@
 };
 
 /*
- * Try to find PCI BIOS.  This will always work for IA64.
- */
-
-static struct pci_ops * __init
-pci_find_bios(void)
-{
-	return &pci_conf;
-}
-
-/*
  * Initialization. Uses the SAL interface
  */
-
-#define PCI_BUSES_TO_SCAN 255
-
 void __init 
-pcibios_init(void)
+pcibios_init (void)
 {
-	struct pci_ops *ops = NULL;
+#	define PCI_BUSES_TO_SCAN 255
 	int i;
 
-	if ((ops = pci_find_bios()) == NULL) {
-		printk("PCI: No PCI bus detected\n");
-		return;
-	}
+	platform_pci_fixup(0);	/* phase 0 initialization (before PCI bus has been scanned) */
 
 	printk("PCI: Probing PCI hardware\n");
 	for (i = 0; i < PCI_BUSES_TO_SCAN; i++) 
-		pci_scan_bus(i, ops, NULL);
-	platform_pci_fixup();
+		pci_scan_bus(i, &pci_conf, NULL);
+
+	platform_pci_fixup(1);	/* phase 1 initialization (after PCI bus has been scanned) */
 	return;
 }
 
@@ -157,16 +136,15 @@
  *  Called after each bus is probed, but before its children
  *  are examined.
  */
-
 void __init
-pcibios_fixup_bus(struct pci_bus *b)
+pcibios_fixup_bus (struct pci_bus *b)
 {
 	return;
 }
 
 void __init
-pcibios_update_resource(struct pci_dev *dev, struct resource *root,
-			struct resource *res, int resource)
+pcibios_update_resource (struct pci_dev *dev, struct resource *root,
+			 struct resource *res, int resource)
 {
         unsigned long where, size;
         u32 reg;
@@ -181,7 +159,7 @@
 }
 
 void __init
-pcibios_update_irq(struct pci_dev *dev, int irq)
+pcibios_update_irq (struct pci_dev *dev, int irq)
 {
 	pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq);
 
@@ -204,18 +182,16 @@
 	return 0;
 }
 
+void
+pcibios_align_resource (void *data, struct resource *res, unsigned long size)
+{
+}
+
 /*
  * PCI BIOS setup, always defaults to SAL interface
  */
-
 char * __init 
-pcibios_setup(char *str)
+pcibios_setup (char *str)
 {
-	pci_probe =  PCI_NO_CHECKS;
 	return NULL;
-}
-
-void
-pcibios_align_resource (void *data, struct resource *res, unsigned long size)
-{
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/perfmon.c linux/arch/ia64/kernel/perfmon.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/perfmon.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/perfmon.c	Thu Jan  4 12:50:17 2001
@@ -4,18 +4,20 @@
  *
  * Originaly Written by Ganesh Venkitachalam, IBM Corp.
  * Modifications by David Mosberger-Tang, Hewlett-Packard Co.
+ * Modifications by Stephane Eranian, Hewlett-Packard Co.
  * Copyright (C) 1999 Ganesh Venkitachalam <venkitac@us.ibm.com>
  * Copyright (C) 1999 David Mosberger-Tang <davidm@hpl.hp.com>
+ * Copyright (C) 2000 Stephane Eranian <eranian@hpl.hp.com>
  */
 
 #include <linux/config.h>
+
 #include <linux/kernel.h>
 #include <linux/init.h>
 #include <linux/sched.h>
 #include <linux/interrupt.h>
 #include <linux/smp_lock.h>
 #include <linux/proc_fs.h>
-#include <linux/ptrace.h>
 
 #include <asm/errno.h>
 #include <asm/hw_irq.h>
@@ -58,19 +60,51 @@
 #define MAX_PERF_COUNTER	4	/* true for Itanium, at least */
 #define PMU_FIRST_COUNTER	4	/* first generic counter */
 
-#define WRITE_PMCS_AND_START	0xa0
-#define WRITE_PMCS		0xa1
-#define READ_PMDS		0xa2
-#define STOP_PMCS		0xa3
+#define PFM_WRITE_PMCS		0xa0
+#define PFM_WRITE_PMDS		0xa1
+#define PFM_READ_PMDS		0xa2
+#define PFM_STOP		0xa3
+#define PFM_START		0xa4
+#define PFM_ENABLE		0xa5	/* unfreeze only */
+#define PFM_DISABLE		0xa6	/* freeze only */
+/* 
+ * Those 2 are just meant for debugging. I considered using sysctl() for
+ * that but it is a little bit too pervasive. This solution is at least
+ * self-contained.
+ */
+#define PFM_DEBUG_ON		0xe0	
+#define PFM_DEBUG_OFF		0xe1
+
+#ifdef CONFIG_SMP
+#define cpu_is_online(i) (cpu_online_map & (1UL << i))
+#else
+#define cpu_is_online(i)	1
+#endif
 
+#define PMC_IS_IMPL(i)		(pmu_conf.impl_regs[i>>6] & (1<< (i&~(64-1))))
+#define PMD_IS_IMPL(i)  	(pmu_conf.impl_regs[4+(i>>6)] & (1<< (i&~(64-1))))
+#define PMD_IS_COUNTER(i)	(i>=PMU_FIRST_COUNTER && i < (PMU_FIRST_COUNTER+pmu_conf.max_counters))
+#define PMC_IS_COUNTER(i)	(i>=PMU_FIRST_COUNTER && i < (PMU_FIRST_COUNTER+pmu_conf.max_counters))
 
 /*
  * this structure needs to be enhanced
  */
 typedef struct {
+	unsigned long	pfr_reg_num;	/* which register */
+	unsigned long	pfr_reg_value;	/* configuration (PMC) or initial value (PMD) */
+	unsigned long	pfr_reg_reset;	/* reset value on overflow (PMD) */
+	void		*pfr_smpl_buf;	/* pointer to user buffer for EAR/BTB */
+	unsigned long	pfr_smpl_size;	/* size of user buffer for EAR/BTB */
+	pid_t		pfr_notify_pid;	/* process to notify */
+	int		pfr_notify_sig;	/* signal for notification, 0=no notification */
+} perfmon_req_t;
+
+#if 0
+typedef struct {
 	unsigned long pmu_reg_data;	/* generic PMD register */
 	unsigned long pmu_reg_num;	/* which register number */
 } perfmon_reg_t; 
+#endif
 
 /*
  * This structure is initialize at boot time and contains
@@ -78,86 +112,141 @@
  * by PAL
  */
 typedef struct {
-	unsigned long perf_ovfl_val;	/* overflow value for generic counters */
-	unsigned long max_pmc;		/* highest PMC */
-	unsigned long max_pmd;		/* highest PMD */
-	unsigned long max_counters;	/* number of generic counter pairs (PMC/PMD) */
+	unsigned long perf_ovfl_val;	/* overflow value for generic counters   */
+	unsigned long max_counters;	/* upper limit on counter pair (PMC/PMD) */
+	unsigned long impl_regs[16];	/* buffer used to hold implememted PMC/PMD mask */
 } pmu_config_t;
 
-/* XXX will go static when ptrace() is cleaned */
-unsigned long perf_ovfl_val;	/* overflow value for generic counters */
-
 static pmu_config_t pmu_conf;
 
+/* for debug only */
+static unsigned long pfm_debug=1;	/* 0= nodebug, >0= debug output on */
+#define DBprintk(a)	{\
+	if (pfm_debug >0) { printk a; } \
+}
+
 /*
- * could optimize to avoid cache conflicts in SMP
+ * could optimize to avoid cache line conflicts in SMP
  */
-unsigned long pmds[NR_CPUS][MAX_PERF_COUNTER];
+static struct task_struct *pmu_owners[NR_CPUS];
 
-asmlinkage unsigned long
-sys_perfmonctl (int cmd, int count, void *ptr, long arg4, long arg5, long arg6, long arg7, long arg8, long stack)
+static int
+do_perfmonctl (struct task_struct *task, int cmd, int flags, perfmon_req_t *req, int count, struct pt_regs *regs)
 {
-	struct pt_regs *regs = (struct pt_regs *) &stack;
-        perfmon_reg_t tmp, *cptr = ptr;
-        unsigned long cnum;
+        perfmon_req_t tmp;
         int i;
 
         switch (cmd) {
-	      case WRITE_PMCS:           /* Writes to PMC's and clears PMDs */
-	      case WRITE_PMCS_AND_START: /* Also starts counting */
+		case PFM_WRITE_PMCS:          
+			/* we don't quite support this right now */
+			if (task != current) return -EINVAL;
+
+			if (!access_ok(VERIFY_READ, req, sizeof(struct perfmon_req_t)*count)) return -EFAULT;
+
+			for (i = 0; i < count; i++, req++) {
+				copy_from_user(&tmp, req, sizeof(tmp));
+
+				/* XXX needs to check validity of the data maybe */
+
+				if (!PMC_IS_IMPL(tmp.pfr_reg_num)) {
+					DBprintk((__FUNCTION__ " invalid pmc[%ld]\n", tmp.pfr_reg_num));
+					return -EINVAL;
+				}
+
+				/* XXX: for counters, need to some checks */
+				if (PMC_IS_COUNTER(tmp.pfr_reg_num)) {
+					current->thread.pmu_counters[tmp.pfr_reg_num - PMU_FIRST_COUNTER].sig = tmp.pfr_notify_sig;
+					current->thread.pmu_counters[tmp.pfr_reg_num - PMU_FIRST_COUNTER].pid = tmp.pfr_notify_pid;
+
+					DBprintk((__FUNCTION__" setting PMC[%ld] send sig %d to %d\n",tmp.pfr_reg_num, tmp.pfr_notify_sig, tmp.pfr_notify_pid));
+				}
+				ia64_set_pmc(tmp.pfr_reg_num, tmp.pfr_reg_value);
+
+				DBprintk((__FUNCTION__" setting PMC[%ld]=0x%lx\n", tmp.pfr_reg_num, tmp.pfr_reg_value));
+			}
+			/*
+			 * we have to set this here event hough we haven't necessarily started monitoring
+			 * because we may be context switched out
+			 */
+			current->thread.flags |= IA64_THREAD_PM_VALID;
+                	break;
+
+		case PFM_WRITE_PMDS:
+			/* we don't quite support this right now */
+			if (task != current) return -EINVAL;
+
+			if (!access_ok(VERIFY_READ, req, sizeof(struct perfmon_req_t)*count)) return -EFAULT;
+
+			for (i = 0; i < count; i++, req++) {
+				copy_from_user(&tmp, req, sizeof(tmp));
+
+				if (!PMD_IS_IMPL(tmp.pfr_reg_num)) return -EINVAL;
+
+				/* update virtualized (64bits) counter */
+				if (PMD_IS_COUNTER(tmp.pfr_reg_num)) {
+					current->thread.pmu_counters[tmp.pfr_reg_num - PMU_FIRST_COUNTER].val  = tmp.pfr_reg_value & ~pmu_conf.perf_ovfl_val;
+					current->thread.pmu_counters[tmp.pfr_reg_num - PMU_FIRST_COUNTER].rval = tmp.pfr_reg_reset;
+				}
+				/* writes to unimplemented part is ignored, so this is safe */
+				ia64_set_pmd(tmp.pfr_reg_num, tmp.pfr_reg_value);
+				/* to go away */
+				ia64_srlz_d();
+				DBprintk((__FUNCTION__" setting PMD[%ld]:  pmod.val=0x%lx pmd=0x%lx rval=0x%lx\n", tmp.pfr_reg_num, current->thread.pmu_counters[tmp.pfr_reg_num - PMU_FIRST_COUNTER].val, ia64_get_pmd(tmp.pfr_reg_num),current->thread.pmu_counters[tmp.pfr_reg_num - PMU_FIRST_COUNTER].rval));
+			}
+			/*
+			 * we have to set this here event hough we haven't necessarily started monitoring
+			 * because we may be context switched out
+			 */
+			current->thread.flags |= IA64_THREAD_PM_VALID;
+                	break;
+
+		case PFM_START:
+			/* we don't quite support this right now */
+			if (task != current) return -EINVAL;
 
-		if (!access_ok(VERIFY_READ, cptr, sizeof(struct perfmon_reg_t)*count))
-			return -EFAULT;
+			pmu_owners[smp_processor_id()] = current;
 
-		for (i = 0; i < count; i++, cptr++) {
+			/* will start monitoring right after rfi */
+			ia64_psr(regs)->up = 1;
 
-			copy_from_user(&tmp, cptr, sizeof(tmp));
+			/* 
+		 	 * mark the state as valid.
+		 	 * this will trigger save/restore at context switch
+		 	 */
+			current->thread.flags |= IA64_THREAD_PM_VALID;
 
-			/* XXX need to check validity of pmu_reg_num and perhaps data!! */
+			ia64_set_pmc(0, 0);
 
-			if (tmp.pmu_reg_num > pmu_conf.max_pmc || tmp.pmu_reg_num == 0) return -EFAULT;
+                	break;
 
-			ia64_set_pmc(tmp.pmu_reg_num, tmp.pmu_reg_data);
+		case PFM_ENABLE:
+			/* we don't quite support this right now */
+			if (task != current) return -EINVAL;
 
-			/* to go away */
-			if (tmp.pmu_reg_num >= PMU_FIRST_COUNTER && tmp.pmu_reg_num < PMU_FIRST_COUNTER+pmu_conf.max_counters) {
-				ia64_set_pmd(tmp.pmu_reg_num, 0);
-				pmds[smp_processor_id()][tmp.pmu_reg_num - PMU_FIRST_COUNTER] = 0;
+			pmu_owners[smp_processor_id()] = current;
 
-				printk(__FUNCTION__" setting PMC/PMD[%ld] es=0x%lx pmd[%ld]=%lx\n", tmp.pmu_reg_num, (tmp.pmu_reg_data>>8) & 0x7f, tmp.pmu_reg_num, ia64_get_pmd(tmp.pmu_reg_num));
-			} else
-				printk(__FUNCTION__" setting PMC[%ld]=0x%lx\n", tmp.pmu_reg_num, tmp.pmu_reg_data);
-		}
-
-		if (cmd == WRITE_PMCS_AND_START) {
-#if 0
-/* irrelevant with user monitors */
-			local_irq_save(flags);
-
-			dcr = ia64_get_dcr();
-			dcr |= IA64_DCR_PP;
-			ia64_set_dcr(dcr);
-
-			local_irq_restore(flags);
-#endif
+			/* 
+		 	 * mark the state as valid.
+		 	 * this will trigger save/restore at context switch
+		 	 */
+			current->thread.flags |= IA64_THREAD_PM_VALID;
 
+			/* simply unfreeze */
 			ia64_set_pmc(0, 0);
+			break;
 
-			/* will start monitoring right after rfi */
-			ia64_psr(regs)->up = 1;
-		}
-		/* 
-		 * mark the state as valid.
-		 * this will trigger save/restore at context switch
-		 */
-		current->thread.flags |= IA64_THREAD_PM_VALID;
-                break;
-
-	      case READ_PMDS:
-		if (count <= 0 || count > MAX_PERF_COUNTER)
-			return -EINVAL;
-		if (!access_ok(VERIFY_WRITE, cptr, sizeof(struct perfmon_reg_t)*count))
-			return -EFAULT;
+		case PFM_DISABLE:
+			/* we don't quite support this right now */
+			if (task != current) return -EINVAL;
+
+			/* simply unfreeze */
+			ia64_set_pmc(0, 1);
+			ia64_srlz_d();
+			break;
+
+	        case PFM_READ_PMDS:
+			if (!access_ok(VERIFY_READ, req, sizeof(struct perfmon_req_t)*count)) return -EFAULT;
+			if (!access_ok(VERIFY_WRITE, req, sizeof(struct perfmon_req_t)*count)) return -EFAULT;
 
 		/* This looks shady, but IMHO this will work fine. This is  
 		 * the sequence that I could come up with to avoid races
@@ -187,16 +276,31 @@
 		 * is the irq_save/restore needed?
 		 */
 
+		for (i = 0; i < count; i++, req++) {
+			unsigned long val=0;
 
-		/* XXX: This needs to change to read more than just the counters */
-		for (i = 0, cnum = PMU_FIRST_COUNTER;i < count; i++, cnum++, cptr++) {
+			copy_from_user(&tmp, req, sizeof(tmp));
 
-			tmp.pmu_reg_data = (pmds[smp_processor_id()][i]
-				    + (ia64_get_pmd(cnum) & pmu_conf.perf_ovfl_val));
+			if (!PMD_IS_IMPL(tmp.pfr_reg_num)) return -EINVAL;
 
-			tmp.pmu_reg_num = cnum;
+			if (PMD_IS_COUNTER(tmp.pfr_reg_num)) {
+				if (task == current){
+					val = ia64_get_pmd(tmp.pfr_reg_num) & pmu_conf.perf_ovfl_val;
+				} else {
+					val = task->thread.pmd[tmp.pfr_reg_num - PMU_FIRST_COUNTER] & pmu_conf.perf_ovfl_val;
+				}
+				val += task->thread.pmu_counters[tmp.pfr_reg_num - PMU_FIRST_COUNTER].val;
+			} else {
+				/* for now */
+				if (task != current) return -EINVAL;
+
+				val = ia64_get_pmd(tmp.pfr_reg_num);
+			}
+			tmp.pfr_reg_value = val;
 
-			if (copy_to_user(cptr, &tmp, sizeof(tmp))) return -EFAULT;
+DBprintk((__FUNCTION__" reading PMD[%ld]=0x%lx\n", tmp.pfr_reg_num, val));
+
+			if (copy_to_user(req, &tmp, sizeof(tmp))) return -EFAULT;
 		}
 #if 0
 /* irrelevant with user monitors */
@@ -209,11 +313,18 @@
 #endif
                 break;
 
-	      case STOP_PMCS:
+	      case PFM_STOP:
+		/* we don't quite support this right now */
+		if (task != current) return -EINVAL;
+
 		ia64_set_pmc(0, 1);
 		ia64_srlz_d();
-		for (i = 0; i < MAX_PERF_COUNTER; ++i)
-			ia64_set_pmc(4+i, 0);
+
+		ia64_psr(regs)->up = 0;
+
+		current->thread.flags &= ~IA64_THREAD_PM_VALID;
+
+		pmu_owners[smp_processor_id()] = NULL;
 
 #if 0
 /* irrelevant with user monitors */
@@ -225,48 +336,140 @@
 		ia64_psr(regs)->up = 0;
 #endif
 
-		current->thread.flags &= ~(IA64_THREAD_PM_VALID);
-
 		break;
 
+	      case PFM_DEBUG_ON:
+			printk(__FUNCTION__" debuggin on\n");
+			pfm_debug = 1;
+			break;
+
+	      case PFM_DEBUG_OFF:
+			printk(__FUNCTION__" debuggin off\n");
+			pfm_debug = 0;
+			break;
+
 	      default:
+		DBprintk((__FUNCTION__" UNknown command 0x%x\n", cmd));
 		return -EINVAL;
 		break;
         }
         return 0;
 }
 
-static inline void
-update_counters (void)
+asmlinkage int
+sys_perfmonctl (int pid, int cmd, int flags, perfmon_req_t *req, int count, long arg6, long arg7, long arg8, long stack)
 {
-	unsigned long mask, i, cnum, val;
+	struct pt_regs *regs = (struct pt_regs *) &stack;
+	struct task_struct *child = current;
+	int ret;
 
-	mask = ia64_get_pmc(0) >> 4;
-	for (i = 0, cnum = PMU_FIRST_COUNTER ; i < pmu_conf.max_counters; cnum++, i++, mask >>= 1) {
+	if (pid != current->pid) {
+		read_lock(&tasklist_lock);
+		{
+			child = find_task_by_pid(pid);
+			if (child)
+				get_task_struct(child);
+		}
+		if (!child) { 
+			read_unlock(&tasklist_lock);
+			return -ESRCH;
+		}
+		/*
+		 * XXX: need to do more checking here
+		 */
+		if (child->state != TASK_ZOMBIE) {
+			DBprintk((__FUNCTION__" warning process %d not in stable state %ld\n", pid, child->state));
+		}
+	} 
+	ret = do_perfmonctl(child, cmd, flags, req, count, regs);
 
+	if (child != current) read_unlock(&tasklist_lock);
 
-		val = mask & 0x1 ? pmu_conf.perf_ovfl_val + 1 : 0;
+	return ret;
+}
 
-		if (mask & 0x1) 
-			printk(__FUNCTION__ " PMD%ld overflowed pmd=%lx pmod=%lx\n", cnum, ia64_get_pmd(cnum), pmds[smp_processor_id()][i]); 
 
-		/* since we got an interrupt, might as well clear every pmd. */
-		val += ia64_get_pmd(cnum) & pmu_conf.perf_ovfl_val;
+static inline int
+update_counters (u64 pmc0)
+{
+	unsigned long mask, i, cnum;
+	struct thread_struct *th;
+	struct task_struct *ta;
+
+	if (pmu_owners[smp_processor_id()] == NULL) {
+		DBprintk((__FUNCTION__" Spurious overflow interrupt: PMU not owned\n"));
+		return 0;
+	}
+	
+	/*
+	 * It is never safe to access the task for which the overflow interrupt is destinated
+	 * using the current variable as the interrupt may occur in the middle of a context switch
+	 * where current does not hold the task that is running yet.
+	 *
+	 * For monitoring, however, we do need to get access to the task which caused the overflow
+	 * to account for overflow on the counters.
+	 * We accomplish this by maintaining a current owner of the PMU per CPU. During context
+	 * switch the ownership is changed in a way such that the reflected owner is always the 
+	 * valid one, i.e. the one that caused the interrupt.
+	 */
+	ta = pmu_owners[smp_processor_id()];
+	th = &pmu_owners[smp_processor_id()]->thread;
 
-		printk(__FUNCTION__ " adding val=%lx to pmod[%ld]=%lx \n", val, i, pmds[smp_processor_id()][i]); 
+	/*
+	 * Don't think this could happen given first test. Keep as sanity check
+	 */
+	if ((th->flags & IA64_THREAD_PM_VALID) == 0) {
+		DBprintk((__FUNCTION__" Spurious overflow interrupt: process %d not using perfmon\n", ta->pid));
+		return 0;
+	}
+
+	/*
+	 * if PMU not frozen: spurious from previous context 
+	 * if PMC[0] = 0x1 : frozen but no overflow reported: leftover from previous context
+	 *
+	 * in either case we don't touch the state upon return from handler
+	 */
+	if ((pmc0 & 0x1) == 0 || pmc0 == 0x1) { 
+		DBprintk((__FUNCTION__" Spurious overflow interrupt: process %d freeze=0\n",ta->pid));
+		return 0;
+	}
 
-		pmds[smp_processor_id()][i] += val;
+	mask = pmc0 >> 4;
 
-		ia64_set_pmd(cnum, 0);
+	for (i = 0, cnum = PMU_FIRST_COUNTER; i < pmu_conf.max_counters; cnum++, i++, mask >>= 1) {
+
+		if (mask & 0x1) {
+			DBprintk((__FUNCTION__ " PMD[%ld] overflowed pmd=0x%lx pmod.val=0x%lx\n", cnum, ia64_get_pmd(cnum), th->pmu_counters[i].val)); 
+			
+			/*
+			 * Because we somtimes (EARS/BTB) reset to a specific value, we cannot simply use 
+			 * val to count the number of times we overflowed. Otherwise we would loose the value
+			 * current in the PMD (which can be >0). So to make sure we don't loose
+			 * the residual counts we set val to contain full 64bits value of the counter.
+			 */
+			th->pmu_counters[i].val += 1+pmu_conf.perf_ovfl_val+(ia64_get_pmd(cnum) &pmu_conf.perf_ovfl_val);
+
+			/* writes to upper part are ignored, so this is safe */
+			ia64_set_pmd(cnum, th->pmu_counters[i].rval);
+
+			DBprintk((__FUNCTION__ " pmod[%ld].val=0x%lx pmd=0x%lx\n", i, th->pmu_counters[i].val, ia64_get_pmd(cnum)&pmu_conf.perf_ovfl_val)); 
+
+			if (th->pmu_counters[i].pid != 0 && th->pmu_counters[i].sig>0) {
+				DBprintk((__FUNCTION__ " shouild notify process %d with signal %d\n",th->pmu_counters[i].pid, th->pmu_counters[i].sig)); 
+			}
+		}
 	}
+	return 1;
 }
 
 static void
 perfmon_interrupt (int irq, void *arg, struct pt_regs *regs)
 {
-	update_counters();
-	ia64_set_pmc(0, 0);
-	ia64_srlz_d();
+	/* unfreeze if not spurious */
+	if ( update_counters(ia64_get_pmc(0)) ) {
+		ia64_set_pmc(0, 0);
+		ia64_srlz_d();
+	}
 }
 
 static struct irqaction perfmon_irqaction = {
@@ -280,9 +483,13 @@
 {
 	char *p = page;
 	u64 pmc0 = ia64_get_pmc(0);
+	int i;
 
-	p += sprintf(p, "PMC[0]=%lx\n", pmc0);
-
+	p += sprintf(p, "PMC[0]=%lx\nPerfmon debug: %s\n", pmc0, pfm_debug ? "On" : "Off");
+	for(i=0; i < NR_CPUS; i++) {
+		if (cpu_is_online(i)) 
+			p += sprintf(p, "CPU%d.PMU %d\n", i, pmu_owners[i] ? pmu_owners[i]->pid: -1);
+	}
 	return p - page;
 }
 
@@ -308,7 +515,6 @@
 perfmon_init (void)
 {
 	pal_perf_mon_info_u_t pm_info;
-	u64 pm_buffer[16];
 	s64 status;
 	
 	irq_desc[PERFMON_IRQ].status |= IRQ_PER_CPU;
@@ -320,15 +526,13 @@
 
 	printk("perfmon: Initialized vector to %u\n",PERFMON_IRQ);
 
-	if ((status=ia64_pal_perf_mon_info(pm_buffer, &pm_info)) != 0) {
+	if ((status=ia64_pal_perf_mon_info(pmu_conf.impl_regs, &pm_info)) != 0) {
 		printk(__FUNCTION__ " pal call failed (%ld)\n", status);
 		return;
 	} 
-	pmu_conf.perf_ovfl_val = perf_ovfl_val = (1L << pm_info.pal_perf_mon_info_s.width) - 1; 
+	pmu_conf.perf_ovfl_val = (1L << pm_info.pal_perf_mon_info_s.width) - 1; 
 
 	/* XXX need to use PAL instead */
-	pmu_conf.max_pmc       = 13;
-	pmu_conf.max_pmd       = 17;
 	pmu_conf.max_counters  = pm_info.pal_perf_mon_info_s.generic;
 
 	printk("perfmon: Counters are %d bits\n", pm_info.pal_perf_mon_info_s.width);
@@ -347,36 +551,137 @@
 	ia64_srlz_d();
 }
 
+/*
+ * XXX: for system wide this function MUST never be called
+ */
 void
-ia64_save_pm_regs (struct thread_struct *t)
+ia64_save_pm_regs (struct task_struct *ta)
 {
-	int i;
+	struct thread_struct *t = &ta->thread;
+	u64 pmc0, psr;
+	int i,j;
+
+	/*
+	 * We must maek sure that we don't loose any potential overflow
+	 * interrupt while saving PMU context. In this code, external
+	 * interrupts are always enabled.
+	 */
+
+	/*
+	 * save current PSR: needed because we modify it
+	 */
+	__asm__ __volatile__ ("mov %0=psr;;": "=r"(psr) :: "memory");
+
+	/*
+	 * stop monitoring:
+	 * This is the only way to stop monitoring without destroying overflow
+	 * information in PMC[0..3].
+	 * This is the last instruction which can cause overflow when monitoring
+	 * in kernel.
+	 * By now, we could still have an overflow interrupt in flight.
+	 */
+	__asm__ __volatile__ ("rsm psr.up;;"::: "memory");
+	
+	/*
+	 * read current overflow status:
+	 *
+	 * We may be reading stale information at this point, if we got interrupt
+	 * just before the read(pmc0) but that's all right. However, if we did
+	 * not get the interrupt before, this read reflects LAST state.
+	 *
+	 */
+	pmc0 = ia64_get_pmc(0);
 
+	/*
+	 * freeze PMU:
+	 *
+	 * This destroys the overflow information. This is required to make sure
+	 * next process does not start with monitoring on if not requested
+	 * (PSR.up may not be enough).
+	 *
+	 * We could still get an overflow interrupt by now. However the handler
+	 * will not do anything if is sees PMC[0].fr=1 but no overflow bits
+	 * are set. So PMU will stay in frozen state. This implies that pmc0
+	 * will still be holding the correct unprocessed information.
+	 *
+	 */
 	ia64_set_pmc(0, 1);
 	ia64_srlz_d();
+
+	/*
+	 * check for overflow bits set:
+	 *
+	 * If pmc0 reports PMU frozen, this means we have a pending overflow,
+	 * therefore we invoke the handler. Handler is reentrant with regards
+	 * to PMC[0] so it is safe to call it twice.
+	 *
+	 * IF pmc0 reports overflow, we need to reread current PMC[0] value
+	 * in case the handler was invoked right after the first pmc0 read.
+	 * it is was not invoked then pmc0==PMC[0], otherwise it's been invoked
+	 * and overflow information has been processed, so we don't need to call.
+	 *
+	 * Test breakdown:
+	 *	- pmc0 & ~0x1: test if overflow happened
+	 * 	- second part: check if current register reflects this as well.
+	 *
+	 * NOTE: testing for pmc0 & 0x1 is not enough has it would trigger call
+	 * when PM_VALID and PMU.fr which is common when setting up registers
+	 * just before actually starting monitors.
+	 *
+	 */
+	if ((pmc0 & ~0x1) && ((pmc0=ia64_get_pmc(0)) &~0x1) ) {
+		printk(__FUNCTION__" Warning: pmc[0]=0x%lx\n", pmc0);
+		update_counters(pmc0);
+		/* 
+		 * XXX: not sure that's enough. the next task may still get the
+		 * interrupt.
+		 */
+	}
+
+	/*
+	 * restore PSR for context switch to save
+	 */
+	__asm__ __volatile__ ("mov psr.l=%0;;"::"r"(psr): "memory");
+
 	/*
 	 * XXX: this will need to be extended beyong just counters
 	 */
-	for (i=0; i< IA64_NUM_PM_REGS; i++) {
-		t->pmd[i]  = ia64_get_pmd(4+i);
-		t->pmod[i] = pmds[smp_processor_id()][i];
-		t->pmc[i]  = ia64_get_pmc(4+i);
+	for (i=0,j=4; i< IA64_NUM_PMD_COUNTERS; i++,j++) {
+		t->pmd[i] = ia64_get_pmd(j);
+		t->pmc[i] = ia64_get_pmc(j);
 	}
+	/*
+	 * PMU is frozen, PMU context is saved: nobody owns the PMU on this CPU
+	 * At this point, we should not receive any pending interrupt from the 
+	 * 'switched out' task
+	 */
+	pmu_owners[smp_processor_id()] = NULL;
 }
 
 void
-ia64_load_pm_regs (struct thread_struct *t)
+ia64_load_pm_regs (struct task_struct *ta)
 {
-	int i;
+	struct thread_struct *t = &ta->thread;
+	int i,j;
+
+	/*
+	 * we first restore ownership of the PMU to the 'soon to be current'
+	 * context. This way, if, as soon as we unfreeze the PMU at the end
+	 * of this function, we get an interrupt, we attribute it to the correct
+	 * task
+	 */
+	pmu_owners[smp_processor_id()] = ta;
 
 	/*
 	 * XXX: this will need to be extended beyong just counters 
 	 */
-	for (i=0; i< IA64_NUM_PM_REGS ; i++) {
-		ia64_set_pmd(4+i, t->pmd[i]);
-		pmds[smp_processor_id()][i] = t->pmod[i];
-		ia64_set_pmc(4+i, t->pmc[i]);
+	for (i=0,j=4; i< IA64_NUM_PMD_COUNTERS; i++,j++) {
+		ia64_set_pmd(j, t->pmd[i]);
+		ia64_set_pmc(j, t->pmc[i]);
 	}
+	/*
+	 * unfreeze PMU
+	 */
 	ia64_set_pmc(0, 0);
 	ia64_srlz_d();
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/process.c linux/arch/ia64/kernel/process.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/process.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/process.c	Thu Jan  4 12:50:17 2001
@@ -137,23 +137,6 @@
 		check_pgt_cache();
 		if (pm_idle)
 			(*pm_idle)();
-#ifdef CONFIG_ITANIUM_ASTEP_SPECIFIC
-		local_irq_disable();
-		{
-			u64 itc, itm;
-
-			itc = ia64_get_itc();
-			itm = ia64_get_itm();
-			if (time_after(itc, itm + 1000)) {
-				extern void ia64_reset_itm (void);
-
-				printk("cpu_idle: ITM in past (itc=%lx,itm=%lx:%lums)\n",
-				       itc, itm, (itc - itm)/500000);
-				ia64_reset_itm();
-			}
-		}
-		local_irq_enable();
-#endif
 	}
 }
 
@@ -164,7 +147,7 @@
 		ia64_save_debug_regs(&task->thread.dbr[0]);
 #ifdef CONFIG_PERFMON
 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
-		ia64_save_pm_regs(&task->thread);
+		ia64_save_pm_regs(task);
 #endif
 	if (IS_IA32_PROCESS(ia64_task_regs(task)))
 		ia32_save_state(&task->thread);
@@ -177,7 +160,7 @@
 		ia64_load_debug_regs(&task->thread.dbr[0]);
 #ifdef CONFIG_PERFMON
 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
-		ia64_load_pm_regs(&task->thread);
+		ia64_load_pm_regs(task);
 #endif
 	if (IS_IA32_PROCESS(ia64_task_regs(task)))
 		ia32_load_state(&task->thread);
@@ -299,6 +282,14 @@
 #	define THREAD_FLAGS_TO_SET	0
 	p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
 			   | THREAD_FLAGS_TO_SET);
+#ifdef CONFIG_IA32_SUPPORT
+	/*
+	 * If we're cloning an IA32 task then save the IA32 extra
+	 * state from the current task to the new task
+	 */
+	if (IS_IA32_PROCESS(ia64_task_regs(current)))
+		ia32_save_state(&p->thread);
+#endif
 	return 0;
 }
 
@@ -554,7 +545,7 @@
 		 * we garantee no race.  this call we also stop
 		 * monitoring
 		 */
-		ia64_save_pm_regs(&current->thread);
+		ia64_save_pm_regs(current);
 		/*
 		 * make sure that switch_to() will not save context again
 		 */
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/ptrace.c linux/arch/ia64/kernel/ptrace.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/ptrace.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/ptrace.c	Thu Jan  4 12:50:17 2001
@@ -617,7 +617,6 @@
 	struct switch_stack *sw;
 	struct unw_frame_info info;
 	struct pt_regs *pt;
-	unsigned long pmd_tmp;
 
 	pt = ia64_task_regs(child);
 	sw = (struct switch_stack *) (child->thread.ksp + 16);
@@ -794,11 +793,7 @@
 				addr);
 			return -1;
 		}
-	} else 
-#ifdef CONFIG_PERFMON
-		if (addr < PT_PMD) 
-#endif
-		{
+	} else {
 		/* access debug registers */
 
 		if (!(child->thread.flags & IA64_THREAD_DBG_VALID)) {
@@ -820,33 +815,14 @@
 		}
 
 		ptr += regnum;
-	}
-#ifdef CONFIG_PERFMON
-	else {
-		/*
-		 * XXX: will eventually move back to perfmonctl()
-		 */
-		unsigned long pmd = (addr - PT_PMD) >> 3;
-		extern unsigned long perf_ovfl_val;
-
-		/* we just use ptrace to read */
-		if (write_access) return -1;
-
-		if (pmd > 3) {
-			printk("ptrace: rejecting access to PMD[%ld] address 0x%lx\n", pmd, addr);
-			return -1;
-		}
 
-		/* 
-		 * We always need to mask upper 32bits of pmd because value is random
-		 */
-		pmd_tmp = child->thread.pmod[pmd]+(child->thread.pmd[pmd]& perf_ovfl_val);
-
-		/*printk(__FUNCTION__" child=%d reading pmd[%ld]=%lx\n", child->pid, pmd, pmd_tmp);*/
-
-		ptr = &pmd_tmp;
+		if (write_access)
+			/* don't let the user set kernel-level breakpoints... */
+			*ptr = *data & ~(7UL << 56);
+		else
+			*data = *ptr;
+		return 0;
 	}
-#endif
 	if (write_access)
 		*ptr = *data;
 	else
@@ -861,7 +837,6 @@
 {
 	unsigned long *ptr = NULL, *rbs, *bspstore, ndirty, regnum;
 	struct switch_stack *sw;
-	unsigned long pmd_tmp;
 	struct pt_regs *pt;
 
 	if ((addr & 0x7) != 0)
@@ -977,11 +952,7 @@
 			/* disallow accessing anything else... */
 			return -1;
 		}
-	} else 
-#ifdef CONFIG_PERFMON
-		if (addr < PT_PMD) 
-#endif
-		{
+	} else {
 
 		/* access debug registers */
 
@@ -1002,34 +973,14 @@
 			return -1;
 
 		ptr += regnum;
-	}
-#ifdef CONFIG_PERFMON
-	else {
-		/*
-		 * XXX: will eventually move back to perfmonctl()
-		 */
-		unsigned long pmd = (addr - PT_PMD) >> 3;
-		extern unsigned long perf_ovfl_val;
 
-		/* we just use ptrace to read */
-		if (write_access) return -1;
-
-		if (pmd > 3) {
-			printk("ptrace: rejecting access to PMD[%ld] address 0x%lx\n", pmd, addr);
-			return -1;
-		}
-
-		/* 
-		 * We always need to mask upper 32bits of pmd because value is random
-		 */
-		pmd_tmp = child->thread.pmod[pmd]+(child->thread.pmd[pmd]& perf_ovfl_val);
-
-		/*printk(__FUNCTION__" child=%d reading pmd[%ld]=%lx\n", child->pid, pmd, pmd_tmp);*/
-
-		ptr = &pmd_tmp;
+		if (write_access)
+			/* don't let the user set kernel-level breakpoints... */
+			*ptr = *data & ~(7UL << 56);
+		else
+			*data = *ptr;
+		return 0;
 	}
-#endif
-
 	if (write_access)
 		*ptr = *data;
 	else
@@ -1107,7 +1058,7 @@
 		goto out_tsk;
 
 	if (child->state != TASK_STOPPED) {
-		if (request != PTRACE_KILL && request != PTRACE_PEEKUSR)
+		if (request != PTRACE_KILL)
 			goto out_tsk;
 	}
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/sal.c linux/arch/ia64/kernel/sal.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/sal.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/sal.c	Thu Jan  4 12:50:17 2001
@@ -104,9 +104,11 @@
 	if (strncmp(systab->signature, "SST_", 4) != 0)
 		printk("bad signature in system table!");
 
-	printk("SAL v%u.%02u: ia32bios=%s, oem=%.32s, product=%.32s\n",
+	/* 
+	 * revisions are coded in BCD, so %x does the job for us
+	 */
+	printk("SAL v%x.%02x: oem=%.32s, product=%.32s\n",
 	       systab->sal_rev_major, systab->sal_rev_minor,
-	       systab->ia32_bios_present ? "present" : "absent",
 	       systab->oem_id, systab->product_id);
 
 	min = ~0UL;
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/setup.c linux/arch/ia64/kernel/setup.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/setup.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/setup.c	Thu Jan  4 12:50:17 2001
@@ -235,6 +235,12 @@
 	machvec_init(acpi_get_sysname());
 #endif
 
+#ifdef	CONFIG_ACPI20
+	if (efi.acpi20) {
+		/* Parse the ACPI 2.0 tables */
+		acpi20_parse(efi.acpi20);
+	} else 
+#endif
 	if (efi.acpi) {
 		/* Parse the ACPI tables */
 		acpi_parse(efi.acpi);
@@ -255,13 +261,6 @@
 
 	paging_init();
 	platform_setup(cmdline_p);
-
-#ifdef CONFIG_SWIOTLB
-	{
-		extern void setup_swiotlb (void);
-		setup_swiotlb();
-	}
-#endif
 }
 
 /*
@@ -271,9 +270,9 @@
 get_cpuinfo (char *buffer)
 {
 #ifdef CONFIG_SMP
-#	define lps	c->loops_per_sec
+#	define lpj	c->loops_per_jiffy
 #else
-#	define lps	loops_per_sec
+#	define lpj	loops_per_jiffy
 #endif
 	char family[32], model[32], features[128], *cp, *p = buffer;
 	struct cpuinfo_ia64 *c;
@@ -325,7 +324,7 @@
 			     features,
 			     c->ppn, c->number, c->proc_freq / 1000000, c->proc_freq % 1000000,
 			     c->itc_freq / 1000000, c->itc_freq % 1000000,
-			     lps / 500000, (lps / 5000) % 100);
+			     lpj*HZ/500000, (lpj*HZ/5000) % 100);
         }
 	return p - buffer;
 }
@@ -376,15 +375,7 @@
 
 	status = ia64_pal_vm_summary(&vm1, &vm2);
 	if (status == PAL_STATUS_SUCCESS) {
-#if 1
-		/*
-		 * XXX the current PAL code returns IMPL_VA_MSB==60, which is dead-wrong.
-		 * --davidm 00/05/26
-		 s*/
-		impl_va_msb = 50;
-#else
 		impl_va_msb = vm2.pal_vm_info_2_s.impl_va_msb;
-#endif
 		phys_addr_size = vm1.pal_vm_info_1_s.phys_add_size;
 	}
 	printk("CPU %d: %lu virtual and %lu physical address bits\n",
@@ -408,6 +399,8 @@
 {
 	extern void __init ia64_rid_init (void);
 	extern void __init ia64_tlb_init (void);
+	pal_vm_info_2_u_t vmi;
+	unsigned int max_ctx;
 
 	identify_cpu(&my_cpu_data);
 
@@ -415,15 +408,12 @@
 	memset(ia64_task_regs(current), 0, sizeof(struct pt_regs));
 
 	/*
-	 * Initialize default control register to defer speculative
-	 * faults.  On a speculative load, we want to defer access
-	 * right, key miss, and key permission faults.  We currently
-	 * do NOT defer TLB misses, page-not-present, access bit, or
-	 * debug faults but kernel code should not rely on any
-	 * particular setting of these bits.
-	ia64_set_dcr(IA64_DCR_DR | IA64_DCR_DK | IA64_DCR_DX | IA64_DCR_PP);
+	 * Initialize default control register to defer all speculative faults.  The
+	 * kernel MUST NOT depend on a particular setting of these bits (in other words,
+	 * the kernel must have recovery code for all speculative accesses).
 	 */
-	ia64_set_dcr(IA64_DCR_DR | IA64_DCR_DK | IA64_DCR_DX );
+	ia64_set_dcr(  IA64_DCR_DM | IA64_DCR_DP | IA64_DCR_DK | IA64_DCR_DX | IA64_DCR_DR
+		     | IA64_DCR_DA | IA64_DCR_DD);
 #ifndef CONFIG_SMP
 	ia64_set_fpu_owner(0);		/* initialize ar.k5 */
 #endif
@@ -444,4 +434,17 @@
 #ifdef CONFIG_SMP
 	normal_xtp();
 #endif
+
+	/* set ia64_ctx.max_rid to the maximum RID that is supported by all CPUs: */
+	if (ia64_pal_vm_summary(NULL, &vmi) == 0)
+		max_ctx = (1U << (vmi.pal_vm_info_2_s.rid_size - 3)) - 1;
+	else {
+		printk("ia64_rid_init: PAL VM summary failed, assuming 18 RID bits\n");
+		max_ctx = (1U << 15) - 1;	/* use architected minimum */
+	}
+	while (max_ctx < ia64_ctx.max_ctx) {
+		unsigned int old = ia64_ctx.max_ctx;
+		if (cmpxchg(&ia64_ctx.max_ctx, old, max_ctx) == old)
+			break;
+	}
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/signal.c linux/arch/ia64/kernel/signal.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/signal.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/signal.c	Thu Jan  4 12:50:17 2001
@@ -91,7 +91,7 @@
 		scr->pt.r10 = -1;
 	}
 	while (1) {
-		set_current_state(TASK_INTERRUPTIBLE);
+		current->state = TASK_INTERRUPTIBLE;
 		schedule();
 		if (ia64_do_signal(&oldset, scr, 1))
 			return -EINTR;
@@ -499,9 +499,10 @@
 			/* Let the debugger run.  */
 			current->exit_code = signr;
 			current->thread.siginfo = &info;
-			set_current_state(TASK_STOPPED);
+			current->state = TASK_STOPPED;
 			notify_parent(current, SIGCHLD);
 			schedule();
+
 			signr = current->exit_code;
 			current->thread.siginfo = 0;
 
@@ -557,7 +558,7 @@
 				/* FALLTHRU */
 
 			      case SIGSTOP:
-				set_current_state(TASK_STOPPED);
+				current->state = TASK_STOPPED;
 				current->exit_code = signr;
 				if (!(current->p_pptr->sig->action[SIGCHLD-1].sa.sa_flags
 				      & SA_NOCLDSTOP))
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/smp.c linux/arch/ia64/kernel/smp.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/smp.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/smp.c	Thu Jan  4 12:50:17 2001
@@ -6,11 +6,13 @@
  * 
  * Lots of stuff stolen from arch/alpha/kernel/smp.c
  *
- *  00/09/11 David Mosberger <davidm@hpl.hp.com> Do loops_per_sec calibration on each CPU.
+ *  00/09/11 David Mosberger <davidm@hpl.hp.com> Do loops_per_jiffy calibration on each CPU.
  *  00/08/23 Asit Mallick <asit.k.mallick@intel.com> fixed logical processor id
  *  00/03/31 Rohit Seth <rohit.seth@intel.com>	Fixes for Bootstrap Processor & cpu_online_map
  *			now gets done here (instead of setup.c)
  *  99/10/05 davidm	Update to bring it in sync with new command-line processing scheme.
+ *  10/13/00 Goutham Rao <goutham.rao@intel.com> Updated smp_call_function and
+ *		smp_call_function_single to resend IPI on timeouts
  */
 #define __KERNEL_SYSCALLS__
 
@@ -30,6 +32,7 @@
 #include <asm/current.h>
 #include <asm/delay.h>
 #include <asm/efi.h>
+#include <asm/machvec.h>
 
 #include <asm/io.h>
 #include <asm/irq.h>
@@ -78,10 +81,6 @@
 };
 static volatile struct smp_call_struct *smp_call_function_data;
 
-#ifdef	CONFIG_ITANIUM_A1_SPECIFIC
-extern spinlock_t ivr_read_lock;
-#endif
-
 #define IPI_RESCHEDULE	        0
 #define IPI_CALL_FUNC	        1
 #define IPI_CPU_STOP	        2
@@ -269,14 +268,14 @@
 }
 
 static inline void
-send_IPI_single(int dest_cpu, int op) 
+send_IPI_single (int dest_cpu, int op) 
 {
 	
 	if (dest_cpu == -1) 
                 return;
         
 	set_bit(op, &ipi_op[dest_cpu]);
-	ipi_send(dest_cpu, IPI_IRQ, IA64_IPI_DM_INT, 0);
+	platform_send_ipi(dest_cpu, IPI_IRQ, IA64_IPI_DM_INT, 0);
 }
 
 static inline void
@@ -358,6 +357,7 @@
 	if (pointer_lock(&smp_call_function_data, &data, retry))
 		return -EBUSY;
 
+resend:
 	/*  Send a message to all other CPUs and wait for them to respond  */
 	send_IPI_single(cpuid, IPI_CALL_FUNC);
 
@@ -366,8 +366,12 @@
 	while ((atomic_read(&data.unstarted_count) > 0) && time_before(jiffies, timeout))
 		barrier();
 	if (atomic_read(&data.unstarted_count) > 0) {
+#if (defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_BSTEP_SPECIFIC))
+		goto resend;
+#else
 		smp_call_function_data = NULL;
 		return -ETIMEDOUT;
+#endif
 	}
 	if (wait)
 		while (atomic_read(&data.unfinished_count) > 0)
@@ -411,13 +415,23 @@
 	/*  Send a message to all other CPUs and wait for them to respond  */
 	send_IPI_allbutself(IPI_CALL_FUNC);
 
+retry:
 	/*  Wait for response  */
 	timeout = jiffies + HZ;
 	while ((atomic_read(&data.unstarted_count) > 0) && time_before(jiffies, timeout))
 		barrier();
 	if (atomic_read(&data.unstarted_count) > 0) {
+#if (defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_BSTEP_SPECIFIC))
+		int i;
+		for (i = 0; i < smp_num_cpus; i++) {
+			if (i != smp_processor_id())
+				platform_send_ipi(i, IPI_IRQ, IA64_IPI_DM_INT, 0);
+		}
+		goto retry;
+#else
 		smp_call_function_data = NULL;
 		return -ETIMEDOUT;
+#endif
 	}
 	if (wait)
 		while (atomic_read(&data.unfinished_count) > 0)
@@ -430,8 +444,6 @@
 /*
  * Flush all other CPU's tlb and then mine.  Do this with smp_call_function() as we
  * want to ensure all TLB's flushed before proceeding.
- *
- * XXX: Is it OK to use the same ptc.e info on all cpus?
  */
 void
 smp_flush_tlb_all(void)
@@ -502,7 +514,7 @@
 	local_irq_enable();		/* Interrupts have been off until now */
 
 	calibrate_delay();
-	my_cpu_data.loops_per_sec = loops_per_sec;
+	my_cpu_data.loops_per_jiffy = loops_per_jiffy;
 
 	/* allow the master to continue */
 	set_bit(cpu, &cpu_callin_map);
@@ -569,7 +581,7 @@
 	cpu_now_booting = cpu;
 
 	/* Kick the AP in the butt */
-	ipi_send(cpu, ap_wakeup_vector, IA64_IPI_DM_INT, 0);
+	platform_send_ipi(cpu, ap_wakeup_vector, IA64_IPI_DM_INT, 0);
 
 	/* wait up to 10s for the AP to start  */
 	for (timeout = 0; timeout < 100000; timeout++) {
@@ -603,7 +615,7 @@
 	__cpu_physical_id[0] = hard_smp_processor_id();
 
 	/* on the BP, the kernel already called calibrate_delay_loop() in init/main.c */
-	my_cpu_data.loops_per_sec = loops_per_sec;
+	my_cpu_data.loops_per_jiffy = loops_per_jiffy;
 #if 0
 	smp_tune_scheduling();
 #endif
@@ -653,13 +665,11 @@
 	bogosum = 0;
         for (i = 0; i < NR_CPUS; i++) {
 		if (cpu_online_map & (1L << i))
-			bogosum += cpu_data[i].loops_per_sec;
+			bogosum += cpu_data[i].loops_per_jiffy;
         }
 
-	printk(KERN_INFO "SMP: Total of %d processors activated "
-	       "(%lu.%02lu BogoMIPS).\n",
-	       cpu_count, (bogosum + 2500) / 500000,
-	       ((bogosum + 2500) / 5000) % 100);
+	printk(KERN_INFO "SMP: Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
+	       cpu_count, bogosum*HZ/500000, (bogosum*HZ/5000) % 100);
 
 	smp_num_cpus = cpu_count;
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/smpboot.c linux/arch/ia64/kernel/smpboot.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/smpboot.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/smpboot.c	Thu Jan  4 12:50:17 2001
@@ -4,6 +4,8 @@
  * Application processor startup code, moved from smp.c to better support kernel profile
  */
 
+#include <linux/config.h>
+
 #include <linux/kernel.h>
 #include <linux/sched.h>
 #include <linux/init.h>
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/sys_ia64.c linux/arch/ia64/kernel/sys_ia64.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/sys_ia64.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/sys_ia64.c	Thu Jan  4 12:50:17 2001
@@ -16,8 +16,38 @@
 #include <linux/smp_lock.h>
 #include <linux/highuid.h>
 
+#include <asm/shmparam.h>
 #include <asm/uaccess.h>
 
+#define COLOR_ALIGN(addr)	(((addr) + SHMLBA - 1) & ~(SHMLBA - 1))
+
+unsigned long
+get_unmapped_area (unsigned long addr, unsigned long len)
+{
+	struct vm_area_struct * vmm;
+
+	if (len > RGN_MAP_LIMIT)
+		return 0;
+	if (!addr)
+		addr = TASK_UNMAPPED_BASE;
+
+	if (current->thread.flags & IA64_THREAD_MAP_SHARED)
+		addr = COLOR_ALIGN(addr);
+	else
+		addr = PAGE_ALIGN(addr);
+
+	for (vmm = find_vma(current->mm, addr); ; vmm = vmm->vm_next) {
+		/* At this point:  (!vmm || addr < vmm->vm_end). */
+		if (TASK_SIZE - len < addr)
+			return 0;
+		if (rgn_offset(addr) + len > RGN_MAP_LIMIT)	/* no risk of overflow here... */
+			return 0;
+		if (!vmm || addr + len <= vmm->vm_start)
+			return addr;
+		addr = vmm->vm_end;
+	}
+}
+
 asmlinkage long
 ia64_getpriority (int which, int who, long arg2, long arg3, long arg4, long arg5, long arg6, 
 		  long arg7, long stack)
@@ -34,6 +64,7 @@
 	return prio;
 }
 
+/* XXX obsolete, but leave it here until the old libc is gone... */
 asmlinkage unsigned long
 sys_getpagesize (void)
 {
@@ -58,16 +89,61 @@
 }
 
 asmlinkage unsigned long
-ia64_brk (long brk, long arg1, long arg2, long arg3,
+ia64_brk (unsigned long brk, long arg1, long arg2, long arg3,
 	  long arg4, long arg5, long arg6, long arg7, long stack)
 {
-	extern unsigned long sys_brk (unsigned long brk);
+	extern int vm_enough_memory (long pages);
 	struct pt_regs *regs = (struct pt_regs *) &stack;
-	unsigned long retval;
+	unsigned long rlim, retval, newbrk, oldbrk;
+	struct mm_struct *mm = current->mm;
+
+	/*
+	 * Most of this replicates the code in sys_brk() except for an additional safety
+	 * check and the clearing of r8.  However, we can't call sys_brk() because we need
+	 * to acquire the mmap_sem before we can do the test...
+	 */
+	down(&mm->mmap_sem);
+
+	if (brk < mm->end_code)
+		goto out;
+	newbrk = PAGE_ALIGN(brk);
+	oldbrk = PAGE_ALIGN(mm->brk);
+	if (oldbrk == newbrk)
+		goto set_brk;
+
+	/* Always allow shrinking brk. */
+	if (brk <= mm->brk) {
+		if (!do_munmap(mm, newbrk, oldbrk-newbrk))
+			goto set_brk;
+		goto out;
+	}
+
+	/* Check against unimplemented/unmapped addresses: */
+	if ((newbrk - oldbrk) > RGN_MAP_LIMIT || rgn_offset(newbrk) > RGN_MAP_LIMIT)
+		goto out;
+
+	/* Check against rlimit.. */
+	rlim = current->rlim[RLIMIT_DATA].rlim_cur;
+	if (rlim < RLIM_INFINITY && brk - mm->start_data > rlim)
+		goto out;
+
+	/* Check against existing mmap mappings. */
+	if (find_vma_intersection(mm, oldbrk, newbrk+PAGE_SIZE))
+		goto out;
 
-	retval = sys_brk(brk);
+	/* Check if we have enough memory.. */
+	if (!vm_enough_memory((newbrk-oldbrk) >> PAGE_SHIFT))
+		goto out;
 
-	regs->r8 = 0;	/* ensure large retval isn't mistaken as error code */
+	/* Ok, looks good - let it rip. */
+	if (do_brk(oldbrk, newbrk-oldbrk) != oldbrk)
+		goto out;
+set_brk:
+	mm->brk = brk;
+out:
+	retval = mm->brk;
+	up(&mm->mmap_sem);
+	regs->r8 = 0;		/* ensure large retval isn't mistaken as error code */
 	return retval;
 }
 
@@ -95,10 +171,8 @@
 static inline unsigned long
 do_mmap2 (unsigned long addr, unsigned long len, int prot, int flags, int fd, unsigned long pgoff)
 {
-	unsigned long loff, hoff;
+	unsigned long roff;
 	struct file *file = 0;
-	/* the virtual address space that is mappable in each region: */
-#	define OCTANT_SIZE	((PTRS_PER_PGD<<PGDIR_SHIFT)/8)
 
 	/*
 	 * A zero mmap always succeeds in Linux, independent of
@@ -107,15 +181,12 @@
 	if (PAGE_ALIGN(len) == 0)
 		return addr;
 
-	/* Don't permit mappings into or across the address hole in a region: */
-	loff = rgn_offset(addr);
-	hoff = loff - (RGN_SIZE - OCTANT_SIZE/2);
-	if ((len | loff | (loff + len)) >= OCTANT_SIZE/2
-	    && (len | hoff | (hoff + len)) >= OCTANT_SIZE/2)
+	/* don't permit mappings into unmapped space or the virtual page table of a region: */
+	roff = rgn_offset(addr);
+	if ((len | roff | (roff + len)) >= RGN_MAP_LIMIT)
 		return -EINVAL;
 
-	/* Don't permit mappings that would cross a region boundary: */
-
+	/* don't permit mappings that would cross a region boundary: */
 	if (rgn_index(addr) != rgn_index(addr + len))
 		return -EINVAL;
 
@@ -126,9 +197,14 @@
 			return -EBADF;
 	}
 
+	if (flags & MAP_SHARED)
+		current->thread.flags |= IA64_THREAD_MAP_SHARED;
+
 	down(&current->mm->mmap_sem);
 	addr = do_mmap_pgoff(file, addr, len, prot, flags, pgoff);
 	up(&current->mm->mmap_sem);
+
+	current->thread.flags &= ~IA64_THREAD_MAP_SHARED;
 
 	if (file)
 		fput(file);
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/time.c linux/arch/ia64/kernel/time.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/time.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/time.c	Thu Jan  4 12:50:17 2001
@@ -152,19 +152,7 @@
 {
 	int cpu = smp_processor_id();
 	unsigned long new_itm;
-#if 0
-	static unsigned long last_time;
-	static unsigned char count;
-	int printed = 0;
-#endif
 
-	/*
-	 * Here we are in the timer irq handler. We have irqs locally
-	 * disabled, but we don't know if the timer_bh is running on
-	 * another CPU. We need to avoid to SMP race by acquiring the
-	 * xtime_lock.
-	 */
-	write_lock(&xtime_lock);
 	new_itm = itm.next[cpu].count;
 
 	if (!time_after(ia64_get_itc(), new_itm))
@@ -173,48 +161,33 @@
 
 	while (1) {
 		/*
-		 * Do kernel PC profiling here.  We multiply the
-		 * instruction number by four so that we can use a
-		 * prof_shift of 2 to get instruction-level instead of
-		 * just bundle-level accuracy.
+		 * Do kernel PC profiling here.  We multiply the instruction number by
+		 * four so that we can use a prof_shift of 2 to get instruction-level
+		 * instead of just bundle-level accuracy.
 		 */
 		if (!user_mode(regs)) 
 			do_profile(regs->cr_iip + 4*ia64_psr(regs)->ri);
 
 #ifdef CONFIG_SMP
 		smp_do_timer(regs);
-		if (smp_processor_id() == 0)
-			do_timer(regs);
-#else
-		do_timer(regs);
 #endif
+		if (smp_processor_id() == 0) {
+			/*
+			 * Here we are in the timer irq handler. We have irqs locally
+			 * disabled, but we don't know if the timer_bh is running on
+			 * another CPU. We need to avoid to SMP race by acquiring the
+			 * xtime_lock.
+			 */
+			write_lock(&xtime_lock);
+			do_timer(regs);
+			write_unlock(&xtime_lock);
+		}
 
 		new_itm += itm.delta;
 		itm.next[cpu].count = new_itm;
 		if (time_after(new_itm, ia64_get_itc()))
 			break;
-
-#if 0
-		/*
-		 * SoftSDV in SMP mode is _slow_, so we do "lose" ticks, 
-		 * but it's really OK...
-		 */
-		if (count > 0 && jiffies - last_time > 5*HZ)
-			count = 0;
-		if (count++ == 0) {
-			last_time = jiffies;
-			if (!printed) {
-				printk("Lost clock tick on CPU %d (now=%lx, next=%lx)!!\n",
-				       cpu, ia64_get_itc(), itm.next[cpu].count);
-				printed = 1;
-# ifdef CONFIG_IA64_DEBUG_IRQ
-				printk("last_cli_ip=%lx\n", last_cli_ip);
-# endif
-			}
-		}
-#endif
 	}
-	write_unlock(&xtime_lock);
 
 	/*
 	 * If we're too close to the next clock tick for comfort, we
@@ -229,7 +202,7 @@
 	ia64_set_itm(new_itm);
 }
 
-#if defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_IA64_SOFTSDV_HACKS)
+#ifdef CONFIG_IA64_SOFTSDV_HACKS
 
 /*
  * Interrupts must be disabled before calling this routine.
@@ -240,7 +213,7 @@
 	timer_interrupt(0, 0, ia64_task_regs(current));
 }
 
-#endif /* CONFIG_ITANIUM_ASTEP_SPECIFIC */
+#endif
 
 /*
  * Encapsulate access to the itm structure for SMP.
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/traps.c linux/arch/ia64/kernel/traps.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/traps.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/traps.c	Thu Jan  4 12:50:17 2001
@@ -78,7 +78,7 @@
 die_if_kernel (char *str, struct pt_regs *regs, long err)
 {
 	if (user_mode(regs)) {
-#if 1
+#if 0
 		/* XXX for debugging only */
 		printk ("!!die_if_kernel: %s(%d): %s %ld\n",
 			current->comm, current->pid, str, err);
@@ -484,6 +484,20 @@
 		sprintf(buf, "Disabled FPL fault---not supposed to happen!");
 		break;
 
+	      case 26: /* NaT Consumption */
+	      case 31: /* Unsupported Data Reference */
+		if (user_mode(regs)) {
+			siginfo.si_signo = SIGILL;
+			siginfo.si_code = ILL_ILLOPN;
+			siginfo.si_errno = 0;
+			siginfo.si_addr = (void *) (regs->cr_iip + ia64_psr(regs)->ri);
+			siginfo.si_imm = vector;
+			force_sig_info(SIGILL, &siginfo, current);
+			return;
+		}
+		sprintf(buf, (vector == 26) ? "NaT consumption" : "Unsupported data reference");
+		break;
+
 	      case 29: /* Debug */
 	      case 35: /* Taken Branch Trap */
 	      case 36: /* Single Step Trap */
@@ -522,10 +536,10 @@
 
 	      case 34:		/* Unimplemented Instruction Address Trap */
 		if (user_mode(regs)) {
-			printk("Woah! Unimplemented Instruction Address Trap!\n");
-			siginfo.si_code = ILL_BADIADDR;
 			siginfo.si_signo = SIGILL;
+			siginfo.si_code = ILL_BADIADDR;
 			siginfo.si_errno = 0;
+			siginfo.si_addr = (void *) (regs->cr_iip + ia64_psr(regs)->ri);
 			force_sig_info(SIGILL, &siginfo, current);
 			return;
 		}
@@ -544,7 +558,8 @@
 
 	      case 46:
 		printk("Unexpected IA-32 intercept trap (Trap 46)\n");
-		printk("  iip - 0x%lx, ifa - 0x%lx, isr - 0x%lx\n", regs->cr_iip, ifa, isr);
+		printk("  iip - 0x%lx, ifa - 0x%lx, isr - 0x%lx, iim - 0x%lx\n",
+		       regs->cr_iip, ifa, isr, iim);
 		force_sig(SIGSEGV, current);
 		return;
 
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/unaligned.c linux/arch/ia64/kernel/unaligned.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/unaligned.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/unaligned.c	Thu Jan  4 12:50:17 2001
@@ -572,7 +572,8 @@
 	 */
 	if (regnum == 0) {
 		*val = 0;
-		*nat = 0;
+		if (nat)
+			*nat = 0;
 		return;
 	}
 
@@ -1563,9 +1564,13 @@
 
 	DPRINT(("ret=%d\n", ret));
 	if (ret) {
-		lock_kernel();
-	        force_sig(SIGSEGV, current);
-	        unlock_kernel();
+		struct siginfo si;
+
+		si.si_signo = SIGBUS;
+		si.si_errno = 0;
+		si.si_code = BUS_ADRALN;
+		si.si_addr = (void *) ifa;
+	        force_sig_info(SIGBUS, &si, current);
 	} else {
 		/*
 	 	 * given today's architecture this case is not likely to happen
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/kernel/unwind.c linux/arch/ia64/kernel/unwind.c
--- v2.4.0-prerelease/linux/arch/ia64/kernel/unwind.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/kernel/unwind.c	Thu Jan  4 12:50:17 2001
@@ -46,16 +46,6 @@
 #define MIN(a,b)	((a) < (b) ? (a) : (b))
 #define p5		5
 
-/*
- * The unwind tables are supposed to be sorted, but the GNU toolchain
- * currently fails to produce a sorted table in the presence of
- * functions that go into sections other than .text.  For example, the
- * kernel likes to put initialization code into .text.init, which
- * messes up the sort order.  Hopefully, this will get fixed sometime
- * soon.  --davidm 00/05/23
- */
-#define UNWIND_TABLE_SORT_BUG
-
 #define UNW_LOG_CACHE_SIZE	7	/* each unw_script is ~256 bytes in size */
 #define UNW_CACHE_SIZE		(1 << UNW_LOG_CACHE_SIZE)
 
@@ -531,6 +521,10 @@
 	struct unw_reg_state *rs;
 
 	rs = alloc_reg_state();
+	if (!rs) {
+		printk("unwind: cannot stack reg state!\n");
+		return;
+	}
 	memcpy(rs, &sr->curr, sizeof(*rs));
 	rs->next = sr->stack;
 	sr->stack = rs;
@@ -1964,23 +1958,6 @@
 {
 	struct unw_table_entry *start = table_start, *end = table_end;
 
-#ifdef UNWIND_TABLE_SORT_BUG
-	{
-		struct unw_table_entry *e1, *e2, tmp;
-
-		/* stupid bubble sort... */
-
-		for (e1 = start; e1 < end; ++e1) {
-			for (e2 = e1 + 1; e2 < end; ++e2) {
-				if (e2->start_offset < e1->start_offset) {
-					tmp = *e1;
-					*e1 = *e2;
-					*e2 = tmp;
-				}
-			}
-		}
-	}
-#endif
 	table->name = name;
 	table->segment_base = segment_base;
 	table->gp = gp;
@@ -2023,8 +2000,8 @@
 void
 unw_remove_unwind_table (void *handle)
 {
-	struct unw_table *table, *prevt;
-	struct unw_script *tmp, *prev;
+	struct unw_table *table, *prev;
+	struct unw_script *tmp;
 	unsigned long flags;
 	long index;
 
@@ -2043,41 +2020,35 @@
 	{
 		/* first, delete the table: */
 
-		for (prevt = (struct unw_table *) &unw.tables; prevt; prevt = prevt->next)
-			if (prevt->next == table)
+		for (prev = (struct unw_table *) &unw.tables; prev; prev = prev->next)
+			if (prev->next == table)
 				break;
-		if (!prevt) {
+		if (!prev) {
 			dprintk("unwind: failed to find unwind table %p\n", (void *) table);
 			spin_unlock_irqrestore(&unw.lock, flags);
 			return;
 		}
-		prevt->next = table->next;
+		prev->next = table->next;
+	}
+	spin_unlock_irqrestore(&unw.lock, flags);
 
-		/* next, remove hash table entries for this table */
+	/* next, remove hash table entries for this table */
 
-		for (index = 0; index <= UNW_HASH_SIZE; ++index) {
-			if (unw.hash[index] >= UNW_CACHE_SIZE)
-				continue;
+	for (index = 0; index <= UNW_HASH_SIZE; ++index) {
+		tmp = unw.cache + unw.hash[index];
+		if (unw.hash[index] >= UNW_CACHE_SIZE
+		    || tmp->ip < table->start || tmp->ip >= table->end)
+			continue;
 
-			tmp = unw.cache + unw.hash[index];
-			prev = 0;
-			while (1) {
-				write_lock(&tmp->lock);
-				{
-					if (tmp->ip >= table->start && tmp->ip < table->end) {
-						if (prev)
-							prev->coll_chain = tmp->coll_chain;
-						else
-							unw.hash[index] = -1;
-						tmp->ip = 0;
-					} else
-						prev = tmp;
-				}
-				write_unlock(&tmp->lock);
+		write_lock(&tmp->lock);
+		{
+			if (tmp->ip >= table->start && tmp->ip < table->end) {
+				unw.hash[index] = tmp->coll_chain;
+				tmp->ip = 0;
 			}
 		}
+		write_unlock(&tmp->lock);
 	}
-	spin_unlock_irqrestore(&unw.lock, flags);
 
 	kfree(table);
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/lib/Makefile linux/arch/ia64/lib/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/lib/Makefile	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/lib/Makefile	Thu Jan  4 12:50:17 2001
@@ -7,22 +7,23 @@
 
 L_TARGET = lib.a
 
-L_OBJS  = __divsi3.o __udivsi3.o __modsi3.o __umodsi3.o					\
+obj-y := __divsi3.o __udivsi3.o __modsi3.o __umodsi3.o					\
 	__divdi3.o __udivdi3.o __moddi3.o __umoddi3.o					\
 	checksum.o clear_page.o csum_partial_copy.o copy_page.o				\
 	copy_user.o clear_user.o strncpy_from_user.o strlen_user.o strnlen_user.o	\
-	flush.o do_csum.o
+	flush.o do_csum.o								\
+	swiotlb.o
 
 ifneq ($(CONFIG_ITANIUM_ASTEP_SPECIFIC),y)
-  L_OBJS += memcpy.o memset.o strlen.o
+  obj-y += memcpy.o memset.o strlen.o
 endif
 
-LX_OBJS = io.o
+export-objs += io.o
 
 IGNORE_FLAGS_OBJS =	__divsi3.o __udivsi3.o __modsi3.o __umodsi3.o \
 			__divdi3.o __udivdi3.o __moddi3.o __umoddi3.o
 
-$(L_TARGET):
+$(L_TARGET): $(obj-y) $(export-objs)
 
 __divdi3.o: idiv64.S
 	$(CC) $(AFLAGS) $(AFLAGS_KERNEL) -c -o $@ $<
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/lib/copy_user.S linux/arch/ia64/lib/copy_user.S
--- v2.4.0-prerelease/linux/arch/ia64/lib/copy_user.S	Fri Jul 14 16:08:12 2000
+++ linux/arch/ia64/lib/copy_user.S	Thu Jan  4 12:50:17 2001
@@ -65,6 +65,12 @@
 //
 // local registers
 //
+#define t1		r2	// rshift in bytes
+#define t2		r3	// lshift in bytes
+#define rshift		r14	// right shift in bits
+#define lshift		r15	// left shift in bits
+#define word1		r16
+#define word2		r17
 #define cnt		r18
 #define len2		r19
 #define saved_lc	r20
@@ -134,6 +140,190 @@
 	br.ret.sptk.few rp	// end of short memcpy
 
 	//
+	// Not 8-byte aligned
+	//
+diff_align_copy_user:
+	// At this point we know we have more than 16 bytes to copy
+	// and also that src and dest do _not_ have the same alignment.
+	and src2=0x7,src1				// src offset
+	and dst2=0x7,dst1				// dst offset
+	;;
+	// The basic idea is that we copy byte-by-byte at the head so 
+	// that we can reach 8-byte alignment for both src1 and dst1. 
+	// Then copy the body using software pipelined 8-byte copy, 
+	// shifting the two back-to-back words right and left, then copy 
+	// the tail by copying byte-by-byte.
+	//
+	// Fault handling. If the byte-by-byte at the head fails on the
+	// load, then restart and finish the pipleline by copying zeros
+	// to the dst1. Then copy zeros for the rest of dst1.
+	// If 8-byte software pipeline fails on the load, do the same as
+	// failure_in3 does. If the byte-by-byte at the tail fails, it is
+	// handled simply by failure_in_pipe1.
+	//
+	// The case p14 represents the source has more bytes in the
+	// the first word (by the shifted part), whereas the p15 needs to 
+	// copy some bytes from the 2nd word of the source that has the 
+	// tail of the 1st of the destination.
+	//
+
+	//
+	// Optimization. If dst1 is 8-byte aligned (not rarely), we don't need 
+	// to copy the head to dst1, to start 8-byte copy software pipleline. 
+	// We know src1 is not 8-byte aligned in this case.
+	//
+	cmp.eq p14,p15=r0,dst2
+(p15)	br.cond.spnt.few 1f				
+	;;
+	sub t1=8,src2
+	mov t2=src2
+	;;
+	shl rshift=t2,3
+	sub len1=len,t1					// set len1
+	;;
+	sub lshift=64,rshift
+	;; 
+	br.cond.spnt.few word_copy_user
+	;; 
+1:			
+	cmp.leu	p14,p15=src2,dst2
+	sub t1=dst2,src2
+	;;
+	.pred.rel "mutex", p14, p15
+(p14)	sub word1=8,src2				// (8 - src offset)
+(p15)	sub t1=r0,t1					// absolute value
+(p15)	sub word1=8,dst2				// (8 - dst offset)
+	;;
+	// For the case p14, we don't need to copy the shifted part to
+	// the 1st word of destination.
+	sub t2=8,t1	
+(p14)	sub word1=word1,t1
+	;;
+	sub len1=len,word1				// resulting len
+(p15)	shl rshift=t1,3					// in bits
+(p14)	shl rshift=t2,3
+	;; 
+(p14)	sub len1=len1,t1
+	adds cnt=-1,word1
+	;; 
+	sub lshift=64,rshift
+	mov ar.ec=PIPE_DEPTH
+	mov pr.rot=1<<16	// p16=true all others are false
+	mov ar.lc=cnt
+	;; 
+2:	
+	EX(failure_in_pipe2,(p16) ld1 val1[0]=[src1],1)
+	;; 
+	EX(failure_out,(EPI) st1 [dst1]=val1[PIPE_DEPTH-1],1)
+	br.ctop.dptk.few 2b
+	;;
+	clrrrb	
+	;; 
+word_copy_user:		
+	cmp.gtu p9,p0=16,len1
+(p9)	br.cond.spnt.few 4f		// if (16 > len1) skip 8-byte copy
+	;;
+	shr.u cnt=len1,3		// number of 64-bit words
+	;;
+	adds cnt=-1,cnt
+	;;
+	.pred.rel "mutex", p14, p15	
+(p14)	sub src1=src1,t2
+(p15)	sub src1=src1,t1
+	//
+	// Now both src1 and dst1 point to an 8-byte aligned address. And
+	// we have more than 8 bytes to copy.
+	//
+	mov ar.lc=cnt
+	mov ar.ec=PIPE_DEPTH
+	mov pr.rot=1<<16	// p16=true all others are false
+	;; 
+3:
+	//
+	// The pipleline consists of 3 stages:	
+	// 1 (p16):	Load a word from src1
+	// 2 (EPI_1):	Shift right pair, saving to tmp
+	// 3 (EPI):	Store tmp to dst1
+	//
+	// To make it simple, use at least 2 (p16) loops to set up val1[n] 
+	// because we need 2 back-to-back val1[] to get tmp.
+	// Note that this implies EPI_2 must be p18 or greater.
+	// 
+
+#define EPI_1		p[PIPE_DEPTH-2]
+#define SWITCH(pred, shift)	cmp.eq pred,p0=shift,rshift
+#define CASE(pred, shift)	\
+	(pred)	br.cond.spnt.few copy_user_bit##shift	
+#define BODY(rshift)							\
+copy_user_bit##rshift:							\
+1:									\
+	EX(failure_out,(EPI) st8 [dst1]=tmp,8);				\
+(EPI_1) shrp tmp=val1[PIPE_DEPTH-3],val1[PIPE_DEPTH-2],rshift;		\
+	EX(failure_in2,(p16) ld8 val1[0]=[src1],8);			\
+	br.ctop.dptk.few 1b;						\
+	;;								\
+	br.cond.spnt.few .diff_align_do_tail
+
+	//
+	// Since the instruction 'shrp' requires a fixed 128-bit value
+	// specifying the bits to shift, we need to provide 7 cases
+	// below. 
+	//
+	SWITCH(p6, 8)
+	SWITCH(p7, 16)
+	SWITCH(p8, 24)	
+	SWITCH(p9, 32)
+	SWITCH(p10, 40)
+	SWITCH(p11, 48)
+	SWITCH(p12, 56)
+	;;
+	CASE(p6, 8)
+	CASE(p7, 16)
+	CASE(p8, 24)
+	CASE(p9, 32)
+	CASE(p10, 40)
+	CASE(p11, 48)
+	CASE(p12, 56)
+	;;
+	BODY(8)
+	BODY(16)
+	BODY(24)
+	BODY(32)
+	BODY(40)		
+	BODY(48)
+	BODY(56)
+	;; 
+.diff_align_do_tail:	
+	.pred.rel "mutex", p14, p15		
+(p14)	sub src1=src1,t1
+(p14)	adds dst1=-8,dst1			
+(p15)	sub dst1=dst1,t1
+	;; 
+4:	
+	// Tail correction.
+	//
+	// The problem with this piplelined loop is that the last word is not
+	// loaded and thus parf of the last word written is not correct. 
+	// To fix that, we simply copy the tail byte by byte.
+	
+	sub len1=endsrc,src1,1
+	clrrrb
+	;; 
+	mov ar.ec=PIPE_DEPTH
+	mov pr.rot=1<<16	// p16=true all others are false
+	mov ar.lc=len1
+	;;
+5:		
+	EX(failure_in_pipe1,(p16) ld1 val1[0]=[src1],1)
+	
+	EX(failure_out,(EPI) st1 [dst1]=val1[PIPE_DEPTH-1],1)
+	br.ctop.dptk.few 5b
+	;;
+	mov pr=saved_pr,0xffffffffffff0000
+	mov ar.pfs=saved_pfs
+	br.ret.dptk.few rp
+	
+	//
 	// Beginning of long mempcy (i.e. > 16 bytes)
 	//
 long_copy_user:
@@ -142,7 +332,7 @@
 	;;
 	cmp.eq p10,p8=r0,tmp
 	mov len1=len		// copy because of rotation
-(p8)	br.cond.dpnt.few 1b	// XXX Fixme. memcpy_diff_align 
+(p8)	br.cond.dpnt.few diff_align_copy_user
 	;;
 	// At this point we know we have more than 16 bytes to copy
 	// and also that both src and dest have the same alignment
@@ -267,6 +457,21 @@
 	mov ar.pfs=saved_pfs
 	br.ret.dptk.few rp
 
+	//
+	// This is the case where the byte by byte copy fails on the load
+	// when we copy the head. We need to finish the pipeline and copy 
+	// zeros for the rest of the destination. Since this happens
+	// at the top we still need to fill the body and tail.
+failure_in_pipe2:
+	sub ret0=endsrc,src1	// number of bytes to zero, i.e. not copied
+2:
+(p16)	mov val1[0]=r0
+(EPI)	st1 [dst1]=val1[PIPE_DEPTH-1],1
+	br.ctop.dptk.few 2b
+	;;
+	sub len=enddst,dst1,1		// precompute len
+	br.cond.dptk.few failure_in1bis
+	;; 
 
 	//
 	// Here we handle the head & tail part when we check for alignment.
@@ -395,6 +600,23 @@
 	mov ar.pfs=saved_pfs
 	br.ret.dptk.few rp
 
+failure_in2:
+	sub ret0=endsrc,src1	// number of bytes to zero, i.e. not copied
+	;;
+3:
+(p16)	mov val1[0]=r0
+(EPI)	st8 [dst1]=val1[PIPE_DEPTH-1],8
+	br.ctop.dptk.few 3b
+	;;
+	cmp.ne p6,p0=dst1,enddst	// Do we need to finish the tail ?
+	sub len=enddst,dst1,1		// precompute len
+(p6)	br.cond.dptk.few failure_in1bis	
+	;;
+	mov pr=saved_pr,0xffffffffffff0000
+	mov ar.lc=saved_lc
+	mov ar.pfs=saved_pfs
+	br.ret.dptk.few rp
+	
 	//
 	// handling of failures on stores: that's the easy part
 	//
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/lib/flush.S linux/arch/ia64/lib/flush.S
--- v2.4.0-prerelease/linux/arch/ia64/lib/flush.S	Thu Jun 22 07:09:44 2000
+++ linux/arch/ia64/lib/flush.S	Thu Jan  4 12:50:17 2001
@@ -12,29 +12,33 @@
 	.psr lsb
 	.lsb
 
-GLOBAL_ENTRY(ia64_flush_icache_page)
+	/*
+	 * flush_icache_range(start,end)
+	 *	Must flush range from start to end-1 but nothing else (need to
+	 *	be careful not to touch addresses that may be unmapped).
+	 */
+GLOBAL_ENTRY(flush_icache_range)
 	UNW(.prologue)
-	alloc r2=ar.pfs,1,0,0,0
+	alloc r2=ar.pfs,2,0,0,0
+	sub r8=in1,in0,1
+	;;
+	shr.u r8=r8,5			// we flush 32 bytes per iteration
 	UNW(.save ar.lc, r3)
 	mov r3=ar.lc			// save ar.lc	
+	;;
 
 	.body
 
-	mov r8=PAGE_SIZE/64-1		// repeat/until loop
-	;;
 	mov ar.lc=r8
-	add r8=32,in0
 	;;
-.Loop1:	fc in0				// issuable on M0 only
-	add in0=64,in0
-	fc r8
-	add r8=64,r8
-	br.cloop.sptk.few .Loop1
+.Loop:	fc in0				// issuable on M0 only
+	add in0=32,in0
+	br.cloop.sptk.few .Loop
 	;;
 	sync.i
 	;;
 	srlz.i
 	;;	
 	mov ar.lc=r3			// restore ar.lc
-	br.ret.sptk.few rp
-END(ia64_flush_icache_page)
+	br.ret.sptk.many rp
+END(flush_icache_range)
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/lib/io.c linux/arch/ia64/lib/io.c
--- v2.4.0-prerelease/linux/arch/ia64/lib/io.c	Tue Oct 31 12:42:26 2000
+++ linux/arch/ia64/lib/io.c	Thu Jan  4 12:50:17 2001
@@ -1,3 +1,4 @@
+#include <linux/config.h>
 #include <linux/types.h>
 
 #include <asm/io.h>
@@ -48,3 +49,54 @@
 	}
 }
 
+#ifdef CONFIG_IA64_GENERIC
+
+unsigned int
+ia64_inb (unsigned long port)
+{
+	return __ia64_inb(port);
+}
+
+unsigned int
+ia64_inw (unsigned long port)
+{
+	return __ia64_inw(port);
+}
+
+unsigned int
+ia64_inl (unsigned long port)
+{
+	return __ia64_inl(port);
+}
+
+void
+ia64_outb (unsigned char val, unsigned long port)
+{
+	__ia64_outb(val, port);
+}
+
+void
+ia64_outw (unsigned short val, unsigned long port)
+{
+	__ia64_outw(val, port);
+}
+
+void
+ia64_outl (unsigned int val, unsigned long port)
+{
+	__ia64_outl(val, port);
+}
+
+/* define aliases: */
+
+asm (".global __ia64_inb, __ia64_inw, __ia64_inl");
+asm ("__ia64_inb = ia64_inb");
+asm ("__ia64_inw = ia64_inw");
+asm ("__ia64_inl = ia64_inl");
+
+asm (".global __ia64_outb, __ia64_outw, __ia64_outl");
+asm ("__ia64_outb = ia64_outb");
+asm ("__ia64_outw = ia64_outw");
+asm ("__ia64_outl = ia64_outl");
+
+#endif /* CONFIG_IA64_GENERIC */
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/lib/memcpy.S linux/arch/ia64/lib/memcpy.S
--- v2.4.0-prerelease/linux/arch/ia64/lib/memcpy.S	Fri Aug 11 19:09:06 2000
+++ linux/arch/ia64/lib/memcpy.S	Thu Jan  4 12:50:17 2001
@@ -17,17 +17,31 @@
 
 #include <asm/asmmacro.h>
 
+#if defined(CONFIG_ITANIUM_B0_SPECIFIC) || defined(CONFIG_ITANIUM_B1_SPECIFIC)
+# define BRP(args...)	nop.b 0
+#else
+# define BRP(args...)	brp.loop.imp args
+#endif
+
 GLOBAL_ENTRY(bcopy)
 	.regstk 3,0,0,0
 	mov r8=in0
 	mov in0=in1
 	;;
 	mov in1=r8
+	// gas doesn't handle control flow across procedures, so it doesn't
+	// realize that a stop bit is needed before the "alloc" instruction
+	// below
+{
+	nop.m 0
+	nop.f 0
+	nop.i 0
+}	;;
 END(bcopy)
 	// FALL THROUGH
 GLOBAL_ENTRY(memcpy)
 
-#	define MEM_LAT	2		/* latency to L1 cache */
+#	define MEM_LAT	21		/* latency to memory */
 
 #	define dst	r2
 #	define src	r3
@@ -57,20 +71,17 @@
 	UNW(.prologue)
 	UNW(.save ar.pfs, saved_pfs)
 	alloc saved_pfs=ar.pfs,3,Nrot,0,Nrot
-#if !(defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_BSTEP_SPECIFIC))
-	lfetch [in1]
-#else
-	nop.m 0
-#endif
+	UNW(.save ar.lc, saved_lc)
+	mov saved_lc=ar.lc
 	or t0=in0,in1
 	;;
 
 	or t0=t0,in2
-	UNW(.save ar.lc, saved_lc)
-	mov saved_lc=ar.lc
 	UNW(.save pr, saved_pr)
 	mov saved_pr=pr
 
+	UNW(.body)
+
 	cmp.eq p6,p0=in2,r0	// zero length?
 	mov retval=in0		// return dst
 (p6)	br.ret.spnt.many rp	// zero length, return immediately
@@ -83,7 +94,6 @@
 
 	adds cnt=-1,cnt		// br.ctop is repeat/until
 	cmp.gtu p7,p0=16,in2	// copying less than 16 bytes?
-	UNW(.body)
 	mov ar.ec=N
 	;;
 
@@ -96,12 +106,26 @@
 (p7)	br.cond.spnt.few memcpy_short
 (p6)	br.cond.spnt.few memcpy_long
 	;;
+	nop.m	0
+	;;
+	nop.m	0
+	nop.i	0
+	;;
+	nop.m	0
+	;;
 	.rotr val[N]
 	.rotp p[N]
-1:
+	.align 32
+1: { .mib
 (p[0])	ld8 val[0]=[src],8
+	nop.i 0
+	BRP(1b, 2f)
+}
+2: { .mfb
 (p[N-1])st8 [dst]=val[N-1],8
+	nop.f 0
 	br.ctop.dptk.few 1b
+}
 	;;
 	mov ar.lc=saved_lc
 	mov pr=saved_pr,-1
@@ -118,19 +142,34 @@
 memcpy_short:
 	adds cnt=-1,in2		// br.ctop is repeat/until
 	mov ar.ec=MEM_LAT
+	BRP(1f, 2f)
 	;;
 	mov ar.lc=cnt
 	;;
+	nop.m	0			
+	;;
+	nop.m	0
+	nop.i	0
+	;;
+	nop.m	0
+	;;
+	nop.m	0
+	;;
 	/*
 	 * It is faster to put a stop bit in the loop here because it makes
 	 * the pipeline shorter (and latency is what matters on short copies).
 	 */
-1:
+	.align 32
+1: { .mib
 (p[0])	ld1 val[0]=[src],1
-	;;
+	nop.i 0
+	BRP(1b, 2f)
+} ;;
+2: { .mfb
 (p[MEM_LAT-1])st1 [dst]=val[MEM_LAT-1],1
+	nop.f 0
 	br.ctop.dptk.few 1b
-	;;
+} ;;
 	mov ar.lc=saved_lc
 	mov pr=saved_pr,-1
 	mov ar.pfs=saved_pfs
@@ -227,6 +266,13 @@
 	mov pr=cnt,0x38			// set (p5,p4,p3) to # of bytes last-word bytes to copy
 	mov ar.lc=t2
 	;;
+	nop.m	0			
+	;;
+	nop.m	0
+	nop.i	0
+	;;
+	nop.m	0
+	;;
 (p6)	ld8 val[1]=[src2],8		// prime the pump...
 	mov b6=t4
 	br.sptk.few b6
@@ -251,17 +297,16 @@
 	.align 64
 
 #define COPY(shift,index)									\
- 1:												\
-  { .mfi											\
+ 1: { .mib											\
 	(p[0])		ld8 val[0]=[src2],8;							\
-			nop.f 0;								\
 	(p[MEM_LAT+3])	shrp w[0]=val[MEM_LAT+3],val[MEM_LAT+4-index],shift;			\
-  };												\
-  { .mbb											\
+			BRP(1b, 2f)								\
+    };												\
+ 2: { .mfb											\
 	(p[MEM_LAT+4])	st8 [dst]=w[1],8;							\
-			nop.b 0;								\
+			nop.f 0;								\
 			br.ctop.dptk.few 1b;							\
-  };												\
+    };												\
 			;;									\
 			ld8 val[N-1]=[src_end];	/* load last word (may be same as val[N]) */	\
 			;;									\
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/lib/swiotlb.c linux/arch/ia64/lib/swiotlb.c
--- v2.4.0-prerelease/linux/arch/ia64/lib/swiotlb.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/lib/swiotlb.c	Thu Jan  4 12:50:17 2001
@@ -0,0 +1,458 @@
+/*
+ * Dynamic DMA mapping support.
+ *
+ * This implementation is for IA-64 platforms that do not support
+ * I/O TLBs (aka DMA address translation hardware).
+ * Copyright (C) 2000 Asit Mallick <Asit.K.Mallick@intel.com>
+ * Copyright (C) 2000 Goutham Rao <goutham.rao@intel.com>
+ *
+ * 00/12/13 davidm	Rename to swiotlb.c and add mark_clean() to avoid
+ *			unnecessary i-cache flushing.
+ */
+
+#include <linux/config.h>
+
+#include <linux/mm.h>
+#include <linux/pci.h>
+#include <linux/spinlock.h>
+#include <linux/string.h>
+#include <linux/types.h>
+
+#include <asm/io.h>
+#include <asm/pci.h>
+#include <asm/dma.h>
+
+#include <linux/init.h>
+#include <linux/bootmem.h>
+
+#define ALIGN(val, align) ((unsigned long)	\
+	(((unsigned long) (val) + ((align) - 1)) & ~((align) - 1)))
+
+/*
+ * log of the size of each IO TLB slab.  The number of slabs is command line controllable.
+ */
+#define IO_TLB_SHIFT 11
+
+/*
+ * Used to do a quick range check in swiotlb_unmap_single and swiotlb_sync_single, to see
+ * if the memory was in fact allocated by this API.
+ */
+static char *io_tlb_start, *io_tlb_end;
+
+/*
+ * The number of IO TLB blocks (in groups of 64) betweeen io_tlb_start and io_tlb_end.
+ * This is command line adjustable via setup_io_tlb_npages.
+ */
+static unsigned long io_tlb_nslabs = 1024;
+
+/*
+ * This is a free list describing the number of free entries available from each index
+ */
+static unsigned int *io_tlb_list;
+static unsigned int io_tlb_index;
+
+/*
+ * We need to save away the original address corresponding to a mapped entry for the sync 
+ * operations.
+ */
+static unsigned char **io_tlb_orig_addr;
+
+/*
+ * Protect the above data structures in the map and unmap calls
+ */ 
+static spinlock_t io_tlb_lock = SPIN_LOCK_UNLOCKED;
+
+static int __init
+setup_io_tlb_npages (char *str)
+{
+	io_tlb_nslabs = simple_strtoul(str, NULL, 0) << (PAGE_SHIFT - IO_TLB_SHIFT);
+	return 1;
+}
+__setup("swiotlb=", setup_io_tlb_npages);
+
+/*
+ * Statically reserve bounce buffer space and initialize bounce buffer data structures for
+ * the software IO TLB used to implement the PCI DMA API.
+ */
+void
+swiotlb_init (void)
+{
+	int i;
+
+	/*
+	 * Get IO TLB memory from the low pages
+	 */
+	io_tlb_start = alloc_bootmem_low_pages(io_tlb_nslabs * (1 << IO_TLB_SHIFT));
+	if (!io_tlb_start)
+		BUG();
+	io_tlb_end = io_tlb_start + io_tlb_nslabs * (1 << IO_TLB_SHIFT);
+
+	/*
+	 * Allocate and initialize the free list array.  This array is used
+	 * to find contiguous free memory regions of size 2^IO_TLB_SHIFT between
+	 * io_tlb_start and io_tlb_end.
+	 */
+	io_tlb_list = alloc_bootmem(io_tlb_nslabs * sizeof(int));
+	for (i = 0; i < io_tlb_nslabs; i++)
+		io_tlb_list[i] = io_tlb_nslabs - i;
+	io_tlb_index = 0;
+	io_tlb_orig_addr = alloc_bootmem(io_tlb_nslabs * sizeof(char *));
+
+	printk("Placing software IO TLB between 0x%p - 0x%p\n",
+	       (void *) io_tlb_start, (void *) io_tlb_end);
+}
+
+/*
+ * Allocates bounce buffer and returns its kernel virtual address.
+ */
+static void *
+map_single (struct pci_dev *hwdev, char *buffer, size_t size, int direction)
+{
+	unsigned long flags;
+	char *dma_addr;
+	unsigned int nslots, stride, index, wrap;
+	int i;
+
+	/*
+	 * For mappings greater than a page size, we limit the stride (and hence alignment)
+	 * to a page size.
+	 */
+	nslots = ALIGN(size, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT;
+	if (size > (1 << PAGE_SHIFT))
+		stride = (1 << (PAGE_SHIFT - IO_TLB_SHIFT));
+	else
+		stride = nslots;
+
+	if (!nslots)
+		BUG();
+
+	/*
+	 * Find suitable number of IO TLB entries size that will fit this request and
+	 * allocate a buffer from that IO TLB pool.
+	 */
+	spin_lock_irqsave(&io_tlb_lock, flags);
+	{
+		wrap = index = ALIGN(io_tlb_index, stride);
+
+		if (index >= io_tlb_nslabs) 
+			wrap = index = 0;
+
+		do {
+			/*
+			 * If we find a slot that indicates we have 'nslots' number of
+			 * contiguous buffers, we allocate the buffers from that slot and
+			 * mark the entries as '0' indicating unavailable.
+			 */
+			if (io_tlb_list[index] >= nslots) {
+				int count = 0;
+
+				for (i = index; i < index + nslots; i++)
+					io_tlb_list[i] = 0;
+				for (i = index - 1; (i >= 0) && io_tlb_list[i]; i--)
+					io_tlb_list[i] = ++count;
+				dma_addr = io_tlb_start + (index << IO_TLB_SHIFT);
+
+				/*
+				 * Update the indices to avoid searching in the next round.
+				 */
+				io_tlb_index = ((index + nslots) < io_tlb_nslabs
+						? (index + nslots) : 0);
+
+				goto found;
+			}
+			index += stride;
+			if (index >= io_tlb_nslabs)
+				index = 0;
+		} while (index != wrap);
+
+		/*
+		 * XXX What is a suitable recovery mechanism here?  We cannot 
+		 * sleep because we are called from with in interrupts!
+		 */
+		panic("map_single: could not allocate software IO TLB (%ld bytes)", size);
+found:
+	}
+	spin_unlock_irqrestore(&io_tlb_lock, flags);
+
+	/*
+	 * Save away the mapping from the original address to the DMA address.  This is
+	 * needed when we sync the memory.  Then we sync the buffer if needed.
+	 */
+	io_tlb_orig_addr[index] = buffer;
+	if (direction == PCI_DMA_TODEVICE || direction == PCI_DMA_BIDIRECTIONAL)
+		memcpy(dma_addr, buffer, size);
+
+	return dma_addr;
+}
+
+/*
+ * dma_addr is the kernel virtual address of the bounce buffer to unmap.
+ */
+static void
+unmap_single (struct pci_dev *hwdev, char *dma_addr, size_t size, int direction)
+{
+	unsigned long flags;
+	int i, nslots = ALIGN(size, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT;
+	int index = (dma_addr - io_tlb_start) >> IO_TLB_SHIFT;
+	char *buffer = io_tlb_orig_addr[index];
+
+	/*
+	 * First, sync the memory before unmapping the entry
+	 */
+	if ((direction == PCI_DMA_FROMDEVICE) || (direction == PCI_DMA_BIDIRECTIONAL))
+		/*
+ 	 	 * bounce... copy the data back into the original buffer * and delete the
+ 	 	 * bounce buffer.
+ 	 	 */
+		memcpy(buffer, dma_addr, size);
+
+	/*
+	 * Return the buffer to the free list by setting the corresponding entries to
+	 * indicate the number of contigous entries available.  While returning the
+	 * entries to the free list, we merge the entries with slots below and above the
+	 * pool being returned.
+	 */
+	spin_lock_irqsave(&io_tlb_lock, flags);
+	{
+		int count = ((index + nslots) < io_tlb_nslabs ? io_tlb_list[index + nslots] : 0);
+		/*
+		 * Step 1: return the slots to the free list, merging the slots with
+		 * superceeding slots
+		 */
+		for (i = index + nslots - 1; i >= index; i--)
+			io_tlb_list[i] = ++count;
+		/*
+		 * Step 2: merge the returned slots with the preceeding slots, if
+		 * available (non zero)
+		 */
+		for (i = index - 1; (i >= 0) && io_tlb_list[i]; i--)
+			io_tlb_list[i] = ++count;
+	}
+	spin_unlock_irqrestore(&io_tlb_lock, flags);
+}
+
+static void
+sync_single (struct pci_dev *hwdev, char *dma_addr, size_t size, int direction)
+{
+	int index = (dma_addr - io_tlb_start) >> IO_TLB_SHIFT;
+	char *buffer = io_tlb_orig_addr[index];
+
+	/*
+  	 * bounce... copy the data back into/from the original buffer
+	 * XXX How do you handle PCI_DMA_BIDIRECTIONAL here ?
+ 	 */
+	if (direction == PCI_DMA_FROMDEVICE)
+		memcpy(buffer, dma_addr, size);
+	else if (direction == PCI_DMA_TODEVICE)
+		memcpy(dma_addr, buffer, size);
+	else
+		BUG();
+}
+
+void *
+swiotlb_alloc_consistent (struct pci_dev *hwdev, size_t size, dma_addr_t *dma_handle)
+{
+	unsigned long pci_addr;
+	int gfp = GFP_ATOMIC;
+	void *ret;
+
+	if (!hwdev || hwdev->dma_mask <= 0xffffffff)
+		gfp |= GFP_DMA; /* XXX fix me: should change this to GFP_32BIT or ZONE_32BIT */
+	ret = (void *)__get_free_pages(gfp, get_order(size));
+	if (!ret)
+		return NULL;
+
+	memset(ret, 0, size);
+	pci_addr = virt_to_phys(ret);
+	if ((pci_addr & ~hwdev->dma_mask) != 0)
+		panic("swiotlb_alloc_consistent: allocated memory is out of range for PCI device");
+	*dma_handle = pci_addr;
+	return ret;
+}
+
+void
+swiotlb_free_consistent (struct pci_dev *hwdev, size_t size, void *vaddr, dma_addr_t dma_handle)
+{
+	free_pages((unsigned long) vaddr, get_order(size));
+}
+
+/*
+ * Map a single buffer of the indicated size for DMA in streaming mode.  The PCI address
+ * to use is returned.
+ *
+ * Once the device is given the dma address, the device owns this memory until either
+ * swiotlb_unmap_single or swiotlb_dma_sync_single is performed.
+ */
+dma_addr_t
+swiotlb_map_single (struct pci_dev *hwdev, void *ptr, size_t size, int direction)
+{
+	unsigned long pci_addr = virt_to_phys(ptr);
+
+	if (direction == PCI_DMA_NONE)
+		BUG();
+	/*
+	 * Check if the PCI device can DMA to ptr... if so, just return ptr
+	 */
+	if ((pci_addr & ~hwdev->dma_mask) == 0)
+		/*
+		 * Device is bit capable of DMA'ing to the buffer... just return the PCI
+		 * address of ptr
+		 */
+		return pci_addr;
+
+	/* 
+	 * get a bounce buffer: 
+	 */
+	pci_addr = virt_to_phys(map_single(hwdev, ptr, size, direction));
+
+	/*
+	 * Ensure that the address returned is DMA'ble:
+	 */
+	if ((pci_addr & ~hwdev->dma_mask) != 0)
+		panic("map_single: bounce buffer is not DMA'ble");
+
+	return pci_addr;
+}
+
+/*
+ * Since DMA is i-cache coherent, any (complete) pages that were written via
+ * DMA can be marked as "clean" so that update_mmu_cache() doesn't have to
+ * flush them when they get mapped into an executable vm-area.
+ */
+static void
+mark_clean (void *addr, size_t size)
+{
+	unsigned long pg_addr, end;
+
+	pg_addr = PAGE_ALIGN((unsigned long) addr);
+	end = (unsigned long) addr + size;
+	while (pg_addr + PAGE_SIZE <= end) {
+#if 0
+		set_bit(PG_arch_1, virt_to_page(pg_addr));
+#else
+		if (!VALID_PAGE(virt_to_page(pg_addr)))
+			printk("Invalid addr %lx!!!\n", pg_addr);
+#endif
+		pg_addr += PAGE_SIZE;
+	}
+}
+
+/*
+ * Unmap a single streaming mode DMA translation.  The dma_addr and size must match what
+ * was provided for in a previous swiotlb_map_single call.  All other usages are
+ * undefined.
+ *
+ * After this call, reads by the cpu to the buffer are guarenteed to see whatever the
+ * device wrote there.
+ */
+void
+swiotlb_unmap_single (struct pci_dev *hwdev, dma_addr_t pci_addr, size_t size, int direction)
+{
+	char *dma_addr = phys_to_virt(pci_addr);
+
+	if (direction == PCI_DMA_NONE)
+		BUG();
+	if (dma_addr >= io_tlb_start && dma_addr < io_tlb_end)
+		unmap_single(hwdev, dma_addr, size, direction);
+	else if (direction == PCI_DMA_FROMDEVICE)
+		mark_clean(dma_addr, size);
+}
+
+/*
+ * Make physical memory consistent for a single streaming mode DMA translation after a
+ * transfer.
+ *
+ * If you perform a swiotlb_map_single() but wish to interrogate the buffer using the cpu,
+ * yet do not wish to teardown the PCI dma mapping, you must call this function before
+ * doing so.  At the next point you give the PCI dma address back to the card, the device
+ * again owns the buffer.
+ */
+void
+swiotlb_sync_single (struct pci_dev *hwdev, dma_addr_t pci_addr, size_t size, int direction)
+{
+	char *dma_addr = phys_to_virt(pci_addr);
+
+	if (direction == PCI_DMA_NONE)
+		BUG();
+	if (dma_addr >= io_tlb_start && dma_addr < io_tlb_end)
+		sync_single(hwdev, dma_addr, size, direction);
+	else if (direction == PCI_DMA_FROMDEVICE)
+		mark_clean(dma_addr, size);
+}
+
+/*
+ * Map a set of buffers described by scatterlist in streaming mode for DMA.  This is the
+ * scather-gather version of the above swiotlb_map_single interface.  Here the scatter
+ * gather list elements are each tagged with the appropriate dma address and length.  They
+ * are obtained via sg_dma_{address,length}(SG).
+ *
+ * NOTE: An implementation may be able to use a smaller number of
+ *       DMA address/length pairs than there are SG table elements.
+ *       (for example via virtual mapping capabilities)
+ *       The routine returns the number of addr/length pairs actually
+ *       used, at most nents.
+ *
+ * Device ownership issues as mentioned above for swiotlb_map_single are the same here.
+ */
+int
+swiotlb_map_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction)
+{
+	int i;
+
+	if (direction == PCI_DMA_NONE)
+		BUG();
+
+	for (i = 0; i < nelems; i++, sg++) {
+		sg->orig_address = sg->address;
+		if ((virt_to_phys(sg->address) & ~hwdev->dma_mask) != 0) {
+			sg->address = map_single(hwdev, sg->address, sg->length, direction);
+		}
+	}
+	return nelems;
+}
+
+/*
+ * Unmap a set of streaming mode DMA translations.  Again, cpu read rules concerning calls
+ * here are the same as for swiotlb_unmap_single() above.
+ */
+void
+swiotlb_unmap_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction)
+{
+	int i;
+
+	if (direction == PCI_DMA_NONE)
+		BUG();
+
+	for (i = 0; i < nelems; i++, sg++)
+		if (sg->orig_address != sg->address) {
+			unmap_single(hwdev, sg->address, sg->length, direction);
+			sg->address = sg->orig_address;
+		} else if (direction == PCI_DMA_FROMDEVICE)
+			mark_clean(sg->address, sg->length);
+}
+
+/*
+ * Make physical memory consistent for a set of streaming mode DMA translations after a
+ * transfer.
+ *
+ * The same as swiotlb_dma_sync_single but for a scatter-gather list, same rules and
+ * usage.
+ */
+void
+swiotlb_sync_sg (struct pci_dev *hwdev, struct scatterlist *sg, int nelems, int direction)
+{
+	int i;
+
+	if (direction == PCI_DMA_NONE)
+		BUG();
+
+	for (i = 0; i < nelems; i++, sg++)
+		if (sg->orig_address != sg->address)
+			sync_single(hwdev, sg->address, sg->length, direction);
+}
+
+unsigned long
+swiotlb_dma_address (struct scatterlist *sg)
+{
+	return virt_to_phys(sg->address);
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/mm/Makefile linux/arch/ia64/mm/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/mm/Makefile	Sun Feb  6 18:42:40 2000
+++ linux/arch/ia64/mm/Makefile	Thu Jan  4 12:50:17 2001
@@ -8,7 +8,7 @@
 # Note 2! The CFLAGS definition is now in the main makefile...
 
 O_TARGET := mm.o
-#O_OBJS	 := ioremap.o
-O_OBJS	 := init.o fault.o tlb.o extable.o
+
+obj-y	 := init.o fault.o tlb.o extable.o
 
 include $(TOPDIR)/Rules.make
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/mm/fault.c linux/arch/ia64/mm/fault.c
--- v2.4.0-prerelease/linux/arch/ia64/mm/fault.c	Thu Jun 22 07:09:45 2000
+++ linux/arch/ia64/mm/fault.c	Thu Jan  4 12:50:17 2001
@@ -94,7 +94,7 @@
 	 * sure we exit gracefully rather than endlessly redo the
 	 * fault.
 	 */
-	switch (handle_mm_fault(mm, vma, address, (mask & VM_WRITE) != 0)) {
+	switch (handle_mm_fault(mm, vma, address, mask) != 0) {
 	      case 1:
 		++current->min_flt;
 		break;
@@ -119,19 +119,27 @@
 	if (!(prev_vma && (prev_vma->vm_flags & VM_GROWSUP) && (address == prev_vma->vm_end))) {
 		if (!(vma->vm_flags & VM_GROWSDOWN))
 			goto bad_area;
+		if (rgn_index(address) != rgn_index(vma->vm_start)
+		    || rgn_offset(address) >= RGN_MAP_LIMIT)
+			goto bad_area;
 		if (expand_stack(vma, address))
 			goto bad_area;
-	} else if (expand_backing_store(prev_vma, address))
-		goto bad_area;
+	} else {
+		vma = prev_vma;
+		if (rgn_index(address) != rgn_index(vma->vm_start)
+		    || rgn_offset(address) >= RGN_MAP_LIMIT)
+			goto bad_area;
+		if (expand_backing_store(vma, address))
+			goto bad_area;
+	}
 	goto good_area;
 
   bad_area:
 	up(&mm->mmap_sem);
 	if (isr & IA64_ISR_SP) {
 		/*
-		 * This fault was due to a speculative load set the
-		 * "ed" bit in the psr to ensure forward progress
-		 * (target register will get a NaT).
+		 * This fault was due to a speculative load set the "ed" bit in the psr to
+		 * ensure forward progress (target register will get a NaT).
 		 */
 		ia64_psr(regs)->ed = 1;
 		return;
@@ -146,6 +154,15 @@
 	}
 
   no_context:
+	if (isr & IA64_ISR_SP) {
+		/*
+		 * This fault was due to a speculative load set the "ed" bit in the psr to
+		 * ensure forward progress (target register will get a NaT).
+		 */
+		ia64_psr(regs)->ed = 1;
+		return;
+	}
+
 	fix = search_exception_table(regs->cr_iip);
 	if (fix) {
 		regs->r8 = -EFAULT;
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/mm/init.c linux/arch/ia64/mm/init.c
--- v2.4.0-prerelease/linux/arch/ia64/mm/init.c	Mon Dec 11 17:59:43 2000
+++ linux/arch/ia64/mm/init.c	Thu Jan  4 12:50:17 2001
@@ -1,8 +1,8 @@
 /*
  * Initialize MMU support.
  *
- * Copyright (C) 1998, 1999 Hewlett-Packard Co
- * Copyright (C) 1998, 1999 David Mosberger-Tang <davidm@hpl.hp.com>
+ * Copyright (C) 1998-2000 Hewlett-Packard Co
+ * Copyright (C) 1998-2000 David Mosberger-Tang <davidm@hpl.hp.com>
  */
 #include <linux/config.h>
 #include <linux/kernel.h>
@@ -19,6 +19,7 @@
 #include <asm/efi.h>
 #include <asm/ia32.h>
 #include <asm/io.h>
+#include <asm/machvec.h>
 #include <asm/pgalloc.h>
 #include <asm/sal.h>
 #include <asm/system.h>
@@ -303,7 +304,7 @@
 		return 0;
 	}
 	flush_page_to_ram(page);
-	set_pte(pte, page_pte_prot(page, PAGE_GATE));
+	set_pte(pte, mk_pte(page, PAGE_GATE));
 	/* no need for flush_tlb */
 	return page;
 }
@@ -311,7 +312,12 @@
 void __init
 ia64_rid_init (void)
 {
-	unsigned long flags, rid, pta, impl_va_msb;
+	unsigned long flags, rid, pta, impl_va_bits;
+#ifdef CONFIG_DISABLE_VHPT
+#	define VHPT_ENABLE_BIT	0
+#else
+#	define VHPT_ENABLE_BIT	1
+#endif
 
 	/* Set up the kernel identity mappings (regions 6 & 7) and the vmalloc area (region 5): */
 	ia64_clear_ic(flags);
@@ -328,44 +334,46 @@
 	__restore_flags(flags);
 
 	/*
-	 * Check if the virtually mapped linear page table (VMLPT)
-	 * overlaps with a mapped address space.  The IA-64
-	 * architecture guarantees that at least 50 bits of virtual
-	 * address space are implemented but if we pick a large enough
-	 * page size (e.g., 64KB), the VMLPT is big enough that it
-	 * will overlap with the upper half of the kernel mapped
-	 * region.  I assume that once we run on machines big enough
-	 * to warrant 64KB pages, IMPL_VA_MSB will be significantly
-	 * bigger, so we can just adjust the number below to get
-	 * things going.  Alternatively, we could truncate the upper
-	 * half of each regions address space to not permit mappings
-	 * that would overlap with the VMLPT.  --davidm 99/11/13
+	 * Check if the virtually mapped linear page table (VMLPT) overlaps with a mapped
+	 * address space.  The IA-64 architecture guarantees that at least 50 bits of
+	 * virtual address space are implemented but if we pick a large enough page size
+	 * (e.g., 64KB), the mapped address space is big enough that it will overlap with
+	 * VMLPT.  I assume that once we run on machines big enough to warrant 64KB pages,
+	 * IMPL_VA_MSB will be significantly bigger, so this is unlikely to become a
+	 * problem in practice.  Alternatively, we could truncate the top of the mapped
+	 * address space to not permit mappings that would overlap with the VMLPT.
+	 * --davidm 00/12/06
+	 */
+#	define pte_bits			3
+#	define mapped_space_bits	(3*(PAGE_SHIFT - pte_bits) + PAGE_SHIFT)
+	/*
+	 * The virtual page table has to cover the entire implemented address space within
+	 * a region even though not all of this space may be mappable.  The reason for
+	 * this is that the Access bit and Dirty bit fault handlers perform
+	 * non-speculative accesses to the virtual page table, so the address range of the
+	 * virtual page table itself needs to be covered by virtual page table.
 	 */
-#	define ld_pte_size		3
-#	define ld_max_addr_space_pages	3*(PAGE_SHIFT - ld_pte_size) /* max # of mappable pages */
-#	define ld_max_addr_space_size	(ld_max_addr_space_pages + PAGE_SHIFT)
-#	define ld_max_vpt_size		(ld_max_addr_space_pages + ld_pte_size)
+#	define vmlpt_bits		(impl_va_bits - PAGE_SHIFT + pte_bits)
 #	define POW2(n)			(1ULL << (n))
-	impl_va_msb = ffz(~my_cpu_data.unimpl_va_mask) - 1;
 
-	if (impl_va_msb < 50 || impl_va_msb > 60)
-		panic("Bogus impl_va_msb value of %lu!\n", impl_va_msb);
+	impl_va_bits = ffz(~my_cpu_data.unimpl_va_mask);
+
+	if (impl_va_bits < 51 || impl_va_bits > 61)
+		panic("CPU has bogus IMPL_VA_MSB value of %lu!\n", impl_va_bits - 1);
+
+	/* place the VMLPT at the end of each page-table mapped region: */
+	pta = POW2(61) - POW2(vmlpt_bits);
 
-	if (POW2(ld_max_addr_space_size - 1) + POW2(ld_max_vpt_size) > POW2(impl_va_msb))
+	if (POW2(mapped_space_bits) >= pta)
 		panic("mm/init: overlap between virtually mapped linear page table and "
 		      "mapped kernel space!");
-	pta = POW2(61) - POW2(impl_va_msb);
-#ifndef CONFIG_DISABLE_VHPT
 	/*
 	 * Set the (virtually mapped linear) page table address.  Bit
 	 * 8 selects between the short and long format, bits 2-7 the
 	 * size of the table, and bit 0 whether the VHPT walker is
 	 * enabled.
 	 */
-	ia64_set_pta(pta | (0<<8) | ((3*(PAGE_SHIFT-3)+3)<<2) | 1);
-#else
-	ia64_set_pta(pta | (0<<8) | ((3*(PAGE_SHIFT-3)+3)<<2) | 0);
-#endif
+	ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | VHPT_ENABLE_BIT);
 }
 
 /*
@@ -420,6 +428,15 @@
 {
 	extern char __start_gate_section[];
 	long reserved_pages, codesize, datasize, initsize;
+
+#ifdef CONFIG_PCI
+	/*
+	 * This needs to be called _after_ the command line has been parsed but _before_
+	 * any drivers that may need the PCI DMA interface are initialized or bootmem has
+	 * been freed.
+	 */
+	platform_pci_dma_init();
+#endif
 
 	if (!mem_map)
 		BUG();
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/mm/tlb.c linux/arch/ia64/mm/tlb.c
--- v2.4.0-prerelease/linux/arch/ia64/mm/tlb.c	Fri Aug 11 19:09:06 2000
+++ linux/arch/ia64/mm/tlb.c	Thu Jan  4 12:50:17 2001
@@ -6,6 +6,8 @@
  *
  * 08/02/00 A. Mallick <asit.k.mallick@intel.com>	
  *		Modified RID allocation for SMP 
+ *          Goutham Rao <goutham.rao@intel.com>
+ *              IPI based ptc implementation and A-step IPI implementation.
  */
 #include <linux/config.h>
 #include <linux/init.h>
@@ -17,6 +19,7 @@
 #include <asm/mmu_context.h>
 #include <asm/pgalloc.h>
 #include <asm/pal.h>
+#include <asm/delay.h>
 
 #define SUPPORTED_PGBITS (			\
 		1 << _PAGE_SIZE_256M |		\
@@ -33,15 +36,10 @@
 struct ia64_ctx ia64_ctx = {
 	lock:	SPIN_LOCK_UNLOCKED,
 	next:	1,
-	limit:	(1UL << IA64_HW_CONTEXT_BITS)
+	limit:	(1 << 15) - 1,		/* start out with the safe (architected) limit */
+	max_ctx: ~0U
 };
 
- /*
-  * Put everything in a struct so we avoid the global offset table whenever
-  * possible.
-  */
-ia64_ptce_info_t ia64_ptce_info;
-
 /*
  * Seralize usage of ptc.g 
  */
@@ -99,9 +97,22 @@
 	/*
 	 * Wait for other CPUs to finish purging entries.
 	 */
+#if (defined(CONFIG_ITANIUM_ASTEP_SPECIFIC) || defined(CONFIG_ITANIUM_BSTEP_SPECIFIC))
+	{
+		unsigned long start = ia64_get_itc();
+		while (atomic_read(&flush_cpu_count) > 0) {
+			if ((ia64_get_itc() - start) > 40000UL) {
+				atomic_set(&flush_cpu_count, smp_num_cpus - 1);
+				smp_send_flush_tlb();
+				start = ia64_get_itc();
+			}
+		}
+	}
+#else
 	while (atomic_read(&flush_cpu_count)) {
 		/* Nothing */
 	}
+#endif
 	if (!(flags & IA64_PSR_I)) {
 		local_irq_disable();
 		ia64_set_tpr(saved_tpr);
@@ -117,12 +128,12 @@
 void
 wrap_mmu_context (struct mm_struct *mm)
 {
+	unsigned long tsk_context, max_ctx = ia64_ctx.max_ctx;
 	struct task_struct *tsk;
-	unsigned long tsk_context;
 
-	if (ia64_ctx.next >= (1UL << IA64_HW_CONTEXT_BITS)) 
+	if (ia64_ctx.next > max_ctx)
 		ia64_ctx.next = 300;	/* skip daemons */
-	ia64_ctx.limit = (1UL << IA64_HW_CONTEXT_BITS);
+	ia64_ctx.limit = max_ctx + 1;
 
 	/*
 	 * Scan all the task's mm->context and set proper safe range
@@ -137,9 +148,9 @@
 		if (tsk_context == ia64_ctx.next) {
 			if (++ia64_ctx.next >= ia64_ctx.limit) {
 				/* empty range: reset the range limit and start over */
-				if (ia64_ctx.next >= (1UL << IA64_HW_CONTEXT_BITS)) 
+				if (ia64_ctx.next > max_ctx) 
 					ia64_ctx.next = 300;
-				ia64_ctx.limit = (1UL << IA64_HW_CONTEXT_BITS);
+				ia64_ctx.limit = max_ctx + 1;
 				goto repeat;
 			}
 		}
@@ -153,12 +164,13 @@
 void
 __flush_tlb_all (void)
 {
-	unsigned long i, j, flags, count0, count1, stride0, stride1, addr = ia64_ptce_info.base;
+	unsigned long i, j, flags, count0, count1, stride0, stride1, addr;
 
-	count0  = ia64_ptce_info.count[0];
-	count1  = ia64_ptce_info.count[1];
-	stride0 = ia64_ptce_info.stride[0];
-	stride1 = ia64_ptce_info.stride[1];
+	addr    = my_cpu_data.ptce_base;
+	count0  = my_cpu_data.ptce_count[0];
+	count1  = my_cpu_data.ptce_count[1];
+	stride0 = my_cpu_data.ptce_stride[0];
+	stride1 = my_cpu_data.ptce_stride[1];
 
 	local_irq_save(flags);
 	for (i = 0; i < count0; ++i) {
@@ -182,7 +194,11 @@
 
 	if (mm != current->active_mm) {
 		/* this does happen, but perhaps it's not worth optimizing for? */
+#ifdef CONFIG_SMP
+		flush_tlb_all();
+#else
 		mm->context = 0;
+#endif
 		return;
 	}
 
@@ -230,6 +246,14 @@
 void __init
 ia64_tlb_init (void)
 {
-	ia64_get_ptce(&ia64_ptce_info);
+	ia64_ptce_info_t ptce_info;
+
+	ia64_get_ptce(&ptce_info);
+	my_cpu_data.ptce_base = ptce_info.base;
+	my_cpu_data.ptce_count[0] = ptce_info.count[0];
+	my_cpu_data.ptce_count[1] = ptce_info.count[1];
+	my_cpu_data.ptce_stride[0] = ptce_info.stride[0];
+	my_cpu_data.ptce_stride[1] = ptce_info.stride[1];
+
 	__flush_tlb_all();		/* nuke left overs from bootstrapping... */
 }
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/Makefile linux/arch/ia64/sn/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/sn/Makefile	Thu Mar 30 16:56:04 2000
+++ linux/arch/ia64/sn/Makefile	Thu Jan  4 13:00:15 2001
@@ -5,15 +5,10 @@
 # Copyright (C) Srinivasa Thirumalachar (sprasad@engr.sgi.com)
 #
 
-CFLAGS          :=     $(CFLAGS) -DCONFIG_SGI_SN1 -DSN1 -DSN -DSOFTSDV \
-			-DLANGUAGE_C=1 -D_LANGUAGE_C=1
-AFLAGS          :=      $(AFLAGS) -DCONFIG_SGI_SN1 -DSN1 -DSOFTSDV
-
-.S.s:
-	$(CPP) $(AFLAGS) -o $*.s $<
-.S.o:
-	$(CC) $(AFLAGS) -c -o $*.o $<
-
+EXTRA_CFLAGS	:= -DSN -DLANGUAGE_C=1 -D_LANGUAGE_C=1 -I. -DBRINGUP \
+		   -DDIRECT_L1_CONSOLE -DNUMA_BASE -DSIMULATED_KLGRAPH \
+		   -DNUMA_MIGR_CONTROL -DLITTLE_ENDIAN -DREAL_HARDWARE \
+		   -DNEW_INTERRUPTS -DCONFIG_IA64_SGI_IO
 all: sn.a
 
 O_TARGET        = sn.a
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/Makefile linux/arch/ia64/sn/fprom/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/Makefile	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/Makefile	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,30 @@
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.  See the file "COPYING" in the main directory of this archive
+# for more details.
+#
+# Copyright (C) 2000 Silicon Graphics, Inc.
+# Copyright (C) Jack Steiner (steiner@sgi.com)
+#
+
+TOPDIR=../../../..
+HPATH           = $(TOPDIR)/include
+
+LIB     = ../../lib/lib.a
+
+OBJ=fpromasm.o main.o  fw-emu.o fpmem.o
+
+fprom: $(OBJ)
+	$(LD) -static -Tfprom.lds -o fprom $(OBJ) $(LIB)
+
+.S.o:
+	$(CC)  -D__ASSEMBLY__ $(AFLAGS) $(AFLAGS_KERNEL) -c -o $*.o $<
+.c.o:
+	$(CC)  $(CFLAGS) $(CFLAGS_KERNEL) -c -o $*.o $<
+
+clean:
+	rm -f *.o fprom
+
+
+include $(TOPDIR)/Rules.make
+
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/README linux/arch/ia64/sn/fprom/README
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/README	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/README	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,85 @@
+This directory contains the files required to build
+the fake PROM image that is currently being used to
+boot IA64 kernels running under the SGI Medusa kernel.
+
+The FPROM currently provides the following functions:
+
+	- PAL emulation for all PAL calls we've made so far.
+	- SAL emulation for all SAL calls we've made so far.
+	- EFI emulation for all EFI calls we've made so far.
+	- builds the "ia64_bootparam" structure that is
+	  passed to the kernel from SAL. This structure 
+	  shows the cpu & memory configurations.
+	- supports medusa boottime options for changing
+	  the number of cpus present
+	- supports medusa boottime options for changing
+	  the memory configuration.
+
+
+
+At some point, this fake PROM will be replaced by the
+real PROM.
+
+
+
+
+To build a fake PROM, cd to this directory & type:
+
+	make
+
+This will (or should) build a fake PROM named "fprom".
+
+
+
+
+Use this fprom image when booting the Medusa simulator. The
+control file used to boot Medusa should include the 
+following lines:
+
+	load fprom
+	load vmlinux
+	sr pc 0x100000
+	sr g 9 <address of kernel _start function> #(currently 0xe000000000520000)
+
+NOTE: There is a script "runsim" in this directory that can be used to
+simplify setting up an environment for running under Medusa.
+
+
+
+
+The following parameters may be passed to the fake PROM to
+control the PAL/SAL/EFI parameters passed to the kernel:
+
+	GR[8] = # of cpus
+	GR[9] = address of primary entry point into the kernel
+	GR[20] = memory configuration for node 0
+	GR[21] = memory configuration for node 1
+	GR[22] = memory configuration for node 2
+	GR[23] = memory configuration for node 3
+
+
+Registers GR[20] - GR[23] contain information to specify the
+amount of memory present on nodes 0-3.
+
+  - if nothing is specified (all registers are 0), the configuration
+    defaults to 8 MB on node 0.
+
+  - a mem config entry for node N is passed in GR[20+N]
+
+  - a mem config entry consists of 8 hex digits. Each digit gives the
+    amount of physical memory available on the node starting at
+    1GB*<dn>, where dn is the digit number. The amount of memory
+    is 8MB*2**<d>. (If <d> = 0, the memory size is 0).
+
+    SN1 doesnt support dimms this small but small memory systems 
+    boot faster on Medusa.
+
+
+
+An example helps a lot. The following specifies that node 0 has
+physical memory 0 to 8MB and 1GB to 1GB+32MB, and that node 1 has
+64MB starting at address 0 of the node which is 8GB.
+
+      gr[20] = 0x21           # 0 to 8MB, 1GB to 1GB+32MB
+      gr[21] = 0x4            # 8GB to 8GB+64MB
+
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fpmem.c linux/arch/ia64/sn/fprom/fpmem.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fpmem.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/fpmem.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,200 @@
+/* 
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Jack Steiner (steiner@sgi.com)
+ */
+
+
+/*
+ * FPROM EFI memory descriptor build routines
+ *
+ * 	- Routines to build the EFI memory descriptor map
+ * 	- Should also be usable by the SGI SN1 prom to convert
+ * 	  klconfig to efi_memmap
+ */
+
+#include <asm/efi.h>
+#include "fpmem.h"
+
+/*
+ * args points to a layout in memory like this
+ *
+ *		32 bit		32 bit
+ *
+ * 		numnodes	numcpus
+ *
+ *		16 bit   16 bit		   32 bit
+ *		nasid0	cpuconf		membankdesc0
+ *		nasid1	cpuconf		membankdesc1
+ *			   .
+ *			   .
+ *			   .
+ *			   .
+ *			   .
+ */
+
+sn_memmap_t	*sn_memmap ;
+sn_config_t	*sn_config ;
+
+/*
+ * There is a hole in the node 0 address space. Dont put it
+ * in the memory map
+ */
+#define NODE0_HOLE_SIZE         (20*MB)
+#define NODE0_HOLE_END          (4UL*GB)
+
+#define	MB			(1024*1024)
+#define GB			(1024*MB)
+#define KERNEL_SIZE		(4*MB)
+#define PROMRESERVED_SIZE	(1*MB)
+#define MD_BANK_SHFT 30
+
+#define TO_NODE(_n, _x)		(((long)_n<<33L) | (long)_x)
+
+/*
+ * For SN, this may not take an arg and gets the numnodes from 
+ * the prom variable or by traversing klcfg or promcfg
+ */
+int
+GetNumNodes(void)
+{
+	return sn_config->nodes;
+}
+
+int
+GetNumCpus(void)
+{
+	return sn_config->cpus;
+}
+
+/* For SN1, get the index th nasid */
+
+int
+GetNasid(int index)
+{
+	return sn_memmap[index].nasid ;
+}
+
+node_memmap_t
+GetMemBankInfo(int index)
+{
+	return sn_memmap[index].node_memmap ;
+}
+
+int
+IsCpuPresent(int cnode, int cpu)
+{
+	return  sn_memmap[cnode].cpuconfig & (1<<cpu);
+}
+
+
+/*
+ * Made this into an explicit case statement so that
+ * we can assign specific properties to banks like bank0
+ * actually disabled etc.
+ */
+
+int
+IsBankPresent(int index, node_memmap_t nmemmap)
+{
+	switch (index) {
+		case 0:return nmemmap.b0;
+		case 1:return nmemmap.b1;
+		case 2:return nmemmap.b2;
+		case 3:return nmemmap.b3;
+		case 4:return nmemmap.b4;
+		case 5:return nmemmap.b5;
+		case 6:return nmemmap.b6;
+		case 7:return nmemmap.b7;
+		default:return -1 ;
+	}
+}
+
+int
+GetBankSize(int index, node_memmap_t nmemmap)
+{
+        switch (index) {
+                case 0:
+                case 1:return nmemmap.b01size;
+                case 2:
+                case 3:return nmemmap.b23size;
+                case 4:
+                case 5:return nmemmap.b45size;
+                case 6:
+                case 7:return nmemmap.b67size;
+                default:return -1 ;
+        }
+}
+
+void
+build_mem_desc(efi_memory_desc_t *md, int type, long paddr, long numbytes)
+{
+        md->type = type;
+        md->phys_addr = paddr;
+        md->virt_addr = 0;
+        md->num_pages = numbytes >> 12;
+        md->attribute = EFI_MEMORY_WB;
+}
+
+int
+build_efi_memmap(void *md, int mdsize)
+{
+	int		numnodes = GetNumNodes() ;
+	int		cnode,bank ;
+	int		nasid ;
+	node_memmap_t	membank_info ;
+	int		bsize;
+	int		count = 0 ;
+	long		paddr, hole, numbytes;
+
+
+	for (cnode=0;cnode<numnodes;cnode++) {
+		nasid = GetNasid(cnode) ;
+		membank_info = GetMemBankInfo(cnode) ;
+		for (bank=0;bank<SN1_MAX_BANK_PER_NODE;bank++) {
+			if (IsBankPresent(bank, membank_info)) {
+				bsize = GetBankSize(bank, membank_info) ;
+                                paddr = TO_NODE(nasid, (long)bank<<MD_BANK_SHFT);
+                                numbytes = BankSizeBytes(bsize);
+
+                                /*
+                                 * Check for the node 0 hole. Since banks cant
+                                 * span the hole, we only need to check if the end of
+                                 * the range is the end of the hole.
+                                 */
+                                if (paddr+numbytes == NODE0_HOLE_END)
+                                        numbytes -= NODE0_HOLE_SIZE;
+                                /*
+                                 * UGLY hack - we must skip overr the kernel and
+                                 * PROM runtime services but we dont exactly where it is.
+                                 * So lets just reserve 0-12MB.
+                                 */
+                                if (bank == 0) {
+					hole = (cnode == 0) ? KERNEL_SIZE : PROMRESERVED_SIZE;
+					numbytes -= hole;
+                                        build_mem_desc(md, EFI_RUNTIME_SERVICES_DATA, paddr, hole);
+                                        paddr += hole;
+			        	count++ ;
+                                        md += mdsize;
+                                }
+                                build_mem_desc(md, EFI_CONVENTIONAL_MEMORY, paddr, numbytes);
+
+			        md += mdsize ;
+			        count++ ;
+			}
+		}
+	}
+	return count ;
+}
+
+void
+build_init(unsigned long args)
+{
+	sn_config = (sn_config_t *) (args);	
+	sn_memmap = (sn_memmap_t *)(args + 8) ; /* SN equiv for this is */
+						/* init to klconfig start */
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fpmem.h linux/arch/ia64/sn/fprom/fpmem.h
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fpmem.h	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/fpmem.h	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,35 @@
+/* 
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Jack Steiner (steiner@sgi.com)
+ */
+
+
+
+#include <asm/sn/mmzone_sn1.h>
+
+typedef struct sn_memmap_s
+{
+	short		nasid ;
+	short		cpuconfig;
+	node_memmap_t 	node_memmap ;
+} sn_memmap_t ;
+
+typedef struct sn_config_s
+{
+	int		cpus;
+	int		nodes;
+	sn_memmap_t	memmap[1];		/* start of array */
+} sn_config_t;
+
+
+extern void build_init(unsigned long);
+extern int build_efi_memmap(void *, int);
+extern int GetNumNodes(void);
+extern int GetNumCpus(void);
+extern int IsCpuPresent(int, int);
+extern int GetNasid(int);
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fprom.lds linux/arch/ia64/sn/fprom/fprom.lds
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fprom.lds	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/fprom.lds	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,96 @@
+
+OUTPUT_FORMAT("elf64-ia64-little")
+OUTPUT_ARCH(ia64)
+ENTRY(_start)
+SECTIONS
+{
+  v = 0x0000000000000000 ;	/* this symbol is here to make debugging with kdb easier... */
+
+  . = (0x000000000000000  + 0x100000) ;
+
+  _text = .;
+  .text : AT(ADDR(.text) - 0x0000000000000000 )
+    {
+	*(__ivt_section)
+	/* these are not really text pages, but the zero page needs to be in a fixed location: */
+	*(__special_page_section)
+	__start_gate_section = .;
+	*(__gate_section)
+	__stop_gate_section = .;
+	*(.text)
+    }
+
+  /* Global data */
+  _data = .;
+
+  .rodata : AT(ADDR(.rodata) - 0x0000000000000000 )
+	{ *(.rodata) }
+  .opd : AT(ADDR(.opd) - 0x0000000000000000 )
+	{ *(.opd) }
+  .data : AT(ADDR(.data) - 0x0000000000000000 )
+	{ *(.data) *(.gnu.linkonce.d*) CONSTRUCTORS }
+
+  __gp = ALIGN (8) + 0x200000;
+
+  .got : AT(ADDR(.got) - 0x0000000000000000 )
+	{ *(.got.plt) *(.got) }
+  /* We want the small data sections together, so single-instruction offsets
+     can access them all, and initialized data all before uninitialized, so
+     we can shorten the on-disk segment size.  */
+  .sdata : AT(ADDR(.sdata) - 0x0000000000000000 )
+	{ *(.sdata) }
+  _edata  =  .;
+  _bss = .;
+  .sbss : AT(ADDR(.sbss) - 0x0000000000000000 )
+	{ *(.sbss) *(.scommon) }
+  .bss : AT(ADDR(.bss) - 0x0000000000000000 )
+	{ *(.bss) *(COMMON) }
+  . = ALIGN(64 / 8);
+  _end = .;
+
+  /* Sections to be discarded */
+  /DISCARD/ : {
+	*(.text.exit)
+	*(.data.exit)
+	}
+
+  /* Stabs debugging sections.  */
+  .stab 0 : { *(.stab) }
+  .stabstr 0 : { *(.stabstr) }
+  .stab.excl 0 : { *(.stab.excl) }
+  .stab.exclstr 0 : { *(.stab.exclstr) }
+  .stab.index 0 : { *(.stab.index) }
+  .stab.indexstr 0 : { *(.stab.indexstr) }
+  /* DWARF debug sections.
+     Symbols in the DWARF debugging sections are relative to the beginning
+     of the section so we begin them at 0.  */
+  /* DWARF 1 */
+  .debug          0 : { *(.debug) }
+  .line           0 : { *(.line) }
+  /* GNU DWARF 1 extensions */
+  .debug_srcinfo  0 : { *(.debug_srcinfo) }
+  .debug_sfnames  0 : { *(.debug_sfnames) }
+  /* DWARF 1.1 and DWARF 2 */
+  .debug_aranges  0 : { *(.debug_aranges) }
+  .debug_pubnames 0 : { *(.debug_pubnames) }
+  /* DWARF 2 */
+  .debug_info     0 : { *(.debug_info) }
+  .debug_abbrev   0 : { *(.debug_abbrev) }
+  .debug_line     0 : { *(.debug_line) }
+  .debug_frame    0 : { *(.debug_frame) }
+  .debug_str      0 : { *(.debug_str) }
+  .debug_loc      0 : { *(.debug_loc) }
+  .debug_macinfo  0 : { *(.debug_macinfo) }
+  /* SGI/MIPS DWARF 2 extensions */
+  .debug_weaknames 0 : { *(.debug_weaknames) }
+  .debug_funcnames 0 : { *(.debug_funcnames) }
+  .debug_typenames 0 : { *(.debug_typenames) }
+  .debug_varnames  0 : { *(.debug_varnames) }
+  /* These must appear regardless of  .  */
+  /* Discard them for now since Intel SoftSDV cannot handle them.
+  .comment 0 : { *(.comment) }
+  .note 0 : { *(.note) }
+  */
+  /DISCARD/ : { *(.comment) }
+  /DISCARD/ : { *(.note) }
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fpromasm.S linux/arch/ia64/sn/fprom/fpromasm.S
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fpromasm.S	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/fpromasm.S	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,314 @@
+/* 
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ *   (Code copied from or=ther files)
+ * Copyright (C) 1998-2000 Hewlett-Packard Co
+ * Copyright (C) 1998-2000 David Mosberger-Tang <davidm@hpl.hp.com>
+ *
+ * Copyright (C) 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Jack Steiner (steiner@sgi.com)
+ */
+
+
+
+#define __ASSEMBLY__ 1
+#include "asm/processor.h"
+
+/*
+ * This file contains additional set up code that is needed to get going on
+ * Medusa.  This code should disappear once real hw is available.
+ *
+ * On entry to this routine, the following register values are assumed:
+ *
+ *	gr[8]	- BSP cpu
+ *	pr[9]	- kernel entry address
+ *
+ * NOTE:
+ *   This FPROM may be loaded/executed at an address different from the
+ *   address that it was linked at. The FPROM is linked to run on node 0
+ *   at address 0x100000. If the code in loaded into another node, it
+ *   must be loaded at offset 0x100000 of the node. In addition, the
+ *   FPROM does the following things:
+ *		- determine the base address of the node it is loaded on
+ *		- add the node base to _gp.
+ *		- add the node base to all addresses derived from "movl" 
+ *		  instructions. (I couldnt get GPREL addressing to work)
+ *		  (maybe newer versions of the tools will support this)
+ *		- scan the .got section and add the node base to all
+ *		  pointers in this section.
+ *		- add the node base to all physical addresses in the
+ *		  SAL/PAL/EFI table built by the C code. (This is done
+ *		  in the C code - not here)
+ *		- add the node base to the TLB entries for vmlinux
+ */
+
+#define KERNEL_BASE	0xe000000000000000
+#define PAGESIZE_256M	28
+
+/* 
+ * ar.k0 gets set to IOPB_PA value, on 460gx chipset it should 
+ * be 0x00000ffffc000000, but on snia we use the (inverse swizzled)
+ * IOSPEC_BASE value
+ */
+#define IOPB_PA		0x00000a0000000000 /* inv swizzle IOSPEC_BASE */
+
+#define RR_RID		8
+
+
+
+// ====================================================================================	
+        .text
+        .align 16
+	.global _start
+	.proc _start
+_start:
+
+// Setup psr and rse for system init
+	mov		psr.l = r0;;
+	srlz.d;;
+	invala
+	mov		ar.rsc = r0;;
+	loadrs
+	;;
+
+// Set CALIAS size to zero. We dont use it.
+	movl		r24=0x80000a0001000028;;	// BR_PI_CALIAS_SIZE
+	st8 		[r24]=r0
+
+// Isolate node number we are running on.
+	mov		r6 = ip;;
+	shr		r5 = r6,33;;			// r5 = node number
+	shl		r6 = r5,33			// r6 = base memory address of node
+
+// Set & relocate gp.
+	movl		r1= __gp;;			// Add base memory address
+	add 		r1 = r1,r6			// Relocate to boot node
+
+// Lets figure out who we are & put it in the LID register.
+// The BR_PI_SELF_CPU_NUM register gives us a value of 0-3.
+// This identifies the cpu on the node. 
+// Merge the cpu number with the NASID to generate the LID.
+	movl		r24=0x80000a0001000020;;	// BR_PI_SELF_CPU_NUM
+	ld8 		r25=[r24]			// Fetch PI_SELF
+	movl		r27=0x80000a0001600000;;	// Fetch REVID to get local NASID
+	ld8 		r27=[r27];;
+	extr.u		r27=r27,32,8
+	shl 		r26=r25,16;;			// Align local cpu# to lid.eid
+	shl 		r27=r27,24;;			// Align NASID to lid.id
+	or  		r26=r26,r27;;			// build the LID
+	mov 		cr.lid=r26			// Now put in in the LID register
+	
+	movl		r2=FPSR_DEFAULT;;
+	mov 		ar.fpsr=r2
+	movl		sp = bootstacke-16;;
+	add 		sp = sp,r6			// Relocate to boot node			
+
+// Save the NASID that we are loaded on.
+	movl		r2=base_nasid;;			// Save base_nasid for C code
+	add 		r2 = r2,r6;;			// Relocate to boot node
+  	st8 		[r2]=r5				// Uncond st8 - same on all cpus
+
+// Save the kernel entry address. It is passed in r9 on one of
+// the cpus.
+	movl		r2=bsp_entry_pc
+	cmp.ne		p6,p0=r9,r0;;
+	add 		r2 = r2,r6;;			// Relocate to boot node
+(p6)  	st8 		[r2]=r9				// Uncond st8 - same on all cpus
+
+
+// The following can ONLY be done by 1 cpu. Lets set a lock - the
+// cpu that gets it does the initilization. The rest just spin waiting
+// til initilization is complete.
+	movl		r22 = initlock;;
+	add		r22 = r22,r6			// Relocate to boot node
+	mov		r23 = 1;;
+	xchg8		r23 = [r22],r23;;
+	cmp.eq 		p6,p0 = 0,r23
+(p6)	br.cond.spnt.few init
+1:	ld4		r23 = [r22];;
+	cmp.eq		p6,p0 = 1,r23
+(p6)	br.cond.sptk	1b
+	br		initx
+
+// Add base address of node memory to each pointer in the .got section.
+init:	movl		r16 = _GLOBAL_OFFSET_TABLE_;;
+	add		r16 = r16,r6;;			// Relocate to boot node
+1: 	ld8		r17 = [r16];;
+	cmp.eq		p6,p7=0,r17
+(p6)	br.cond.sptk.few.clr 2f;;
+	add		r17 = r17,r6;;			// Relocate to boot node
+	st8		[r16] = r17,8
+	br		1b
+2:
+	mov		r23 = 2;;			// All done, release the spinning cpus
+	st4		[r22] = r23
+initx:
+
+//
+//	I/O-port space base address:
+//
+	movl		r2 = IOPB_PA;;
+	mov		ar.k0 = r2
+
+
+// Now call main & pass it the current LID value.
+	alloc 		r0=ar.pfs,0,0,2,0
+	mov    		r32=r26
+	mov   		r33=r8;;
+	br.call.sptk.few rp=fmain
+	
+// Initialize Region Registers
+//
+	mov		r10 = r0
+	mov		r2 = (13<<2) 
+	mov		r3 = r0;;
+1:	cmp4.gtu	p6,p7 = 7, r3
+	dep		r10 = r3, r10, 61, 3
+	dep		r2 = r3, r2, RR_RID, 4;;
+(p7)	dep		r2 = 0, r2, 0, 1;;
+(p6)	dep		r2 = -1, r2, 0, 1;;
+	mov		rr[r10] = r2
+	add		r3 = 1, r3;;
+	srlz.d;;
+	cmp4.gtu	p6,p0 = 8, r3
+(p6)	br.cond.sptk.few.clr 1b
+
+//
+// Return value indicates if we are the BSP or AP.
+// 	   1 = BSP, 0 = AP
+	mov             cr.tpr=r0;;
+	cmp.eq		p6,p0=r8,r0
+(p6)	br.cond.spnt	slave
+
+//
+// Initialize the protection key registers with only pkr[0] = valid.
+//
+// Should be initialized in accordance with the OS.
+//
+	mov		r2 = 1
+	mov		r3 = r0;;
+	mov		pkr[r3] = r2;;
+	srlz.d;;
+	mov		r2 = r0
+
+1:	add		r3 = r3, r0, 1;;		// increment PKR
+	cmp.gtu		p6, p0 = 16, r3;;
+(p6)	mov		pkr[r3] = r2
+(p6)	br.cond.sptk.few.clr 1b
+
+	mov		ar.rnat = r0			// clear RNAT register
+
+//
+// Setup system address translation for kernel
+//
+// Note: The setup of Kernel Virtual address space can be done by the
+// C code of the boot loader.
+//
+//
+
+#define LINUX_PAGE_OFFSET       0xe000000000000000
+#define ITIR(key, ps)           ((key<<8) | (ps<<2))
+#define ITRGR(ed,ar,ma)         ((ed<<52) | (ar<<9) | (ma<<2) | 0x61)
+
+#define AR_RX                   1                       // RX permission
+#define AR_RW                   4                       // RW permission
+#define MA_WB                   0                       // WRITEBACK memory attribute
+
+#define TLB_PAGESIZE		28			// Use 256MB pages for now.
+	mov		r16=r5
+
+//
+//     text section
+//
+        movl            r2 = LINUX_PAGE_OFFSET;;        // Set up IFA with VPN of linux
+        mov             cr.ifa = r2
+        movl            r3 = ITIR(0,TLB_PAGESIZE);;     // Set ITIR to default pagesize
+        mov             cr.itir = r3
+
+        shl             r4 = r16,33;;                   // physical addr of start of node
+        movl            r5 = ITRGR(1,AR_RX,MA_WB);;     // TLB attributes
+        or              r10=r4,r5;;
+
+        itr.i           itr[r0] = r10;;                   // Dropin ITR entry
+	srlz.i;;
+
+//
+//     data section
+//
+        movl            r2 = LINUX_PAGE_OFFSET;;        // Set up IFA with VPN of linux
+        mov             cr.ifa = r2
+        movl            r3 = ITIR(0,TLB_PAGESIZE);;     // Set ITIR to default pagesize
+        mov             cr.itir = r3
+
+        shl             r4 = r16,33;;                   // physical addr of start of node
+        movl            r5 = ITRGR(1,AR_RW,MA_WB);;     // TLB attributes
+        or              r10=r4,r5;;
+
+        itr.d           dtr[r0] = r10;;                 // Dropin DTR entry
+	srlz.d;;
+
+
+
+
+//
+// Turn on address translation, interrupt collection, psr.ed, protection key.
+// Interrupts (PSR.i) are still off here.
+//
+
+	movl		r3 = (	IA64_PSR_BN | \
+				IA64_PSR_AC | \
+				IA64_PSR_IT | \
+				IA64_PSR_DB | \
+				IA64_PSR_DA | \
+				IA64_PSR_RT | \
+				IA64_PSR_DT | \
+				IA64_PSR_IC   \
+			     )
+	;;
+	mov		cr.ipsr = r3
+
+//
+// Go to kernel C startup routines
+//	Need to do a "rfi" in order set "it" and "ed" bits in the PSR.
+//	This is the only way to set them.
+
+	movl		r2=bsp_entry_pc;;
+	add 		r2 = r2,r6;;			// Relocate to boot node
+	ld8		r2=[r2];;
+	mov		cr.iip = r2
+	srlz.d;;
+	rfi;;
+	.endp		_start
+
+// Slave processors come here to spin til they get an interrupt. Then they launch themselves to
+// the place ap_entry points. No initialization is necessary - the kernel makes no
+// assumptions about state on this entry.
+//	Note: should verify that the interrupt we got was really the ap_wakeup
+//	      interrupt but this should not be an issue on medusa
+slave:
+	nop.i		0x8beef				// Medusa - put cpu to sleep til interrupt occurs
+	mov		r8=cr.irr0;;			// Check for interrupt pending.
+	cmp.eq		p6,p0=r8,r0
+(p6)	br.cond.sptk	slave;;
+
+	mov		r8=cr.ivr;;			// Got one. Must read ivr to accept it
+	srlz.d;;
+	mov		cr.eoi=r0;;			// must write eoi to clear
+	movl		r8=ap_entry;;			// now jump to kernel entry
+	add 		r8 = r8,r6;;			// Relocate to boot node
+	ld8		r9=[r8],8;;
+	ld8		r1=[r8]
+	mov		b0=r9;;
+	br		b0
+
+// Here is the kernel stack used for the fake PROM
+	.bss
+	.align		16384
+bootstack:
+	.skip		16384
+bootstacke:
+initlock:
+	data4
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fw-emu.c linux/arch/ia64/sn/fprom/fw-emu.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/fw-emu.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/fw-emu.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,494 @@
+/*
+ * PAL & SAL emulation.
+ *
+ * Copyright (C) 1998-2000 Hewlett-Packard Co
+ * Copyright (C) 1998-2000 David Mosberger-Tang <davidm@hpl.hp.com>
+ *
+ *
+ * Copyright (C) 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Jack Steiner (steiner@sgi.com)
+ */
+#include <linux/config.h>
+
+#include <asm/efi.h>
+#include <asm/pal.h>
+#include <asm/sal.h>
+#include <asm/processor.h>
+#include <asm/acpi-ext.h>
+#include "fpmem.h"
+
+#define MB	(1024*1024UL)
+#define GB	(MB*1024UL)
+
+#define FPROM_BUG()		do {while (1);} while (0)
+#define MAX_NODES		128
+#define MAX_LSAPICS		512
+#define MAX_CPUS		512
+#define MAX_CPUS_NODE		4
+#define CPUS_PER_NODE		4
+#define CPUS_PER_FSB		2
+#define CPUS_PER_FSB_MASK	(CPUS_PER_FSB-1)
+
+#define NUM_EFI_DESCS		2
+
+typedef union ia64_nasid_va {
+        struct {
+                unsigned long off   : 33;       /* intra-region offset */
+		unsigned long nasid :  7;	/* NASID */
+		unsigned long off2  : 21;	/* fill */
+                unsigned long reg   :  3;       /* region number */
+        } f;
+        unsigned long l;
+        void *p;
+} ia64_nasid_va;
+
+typedef struct {
+	unsigned long	pc;
+	unsigned long	gp;
+} func_ptr_t;
+ 
+#define IS_VIRTUAL_MODE() 	 ({struct ia64_psr psr; asm("mov %0=psr" : "=r"(psr)); psr.dt;})
+#define ADDR_OF(p)		(IS_VIRTUAL_MODE() ? ((void*)((long)(p)+PAGE_OFFSET)) : ((void*) (p)))
+#define __fwtab_pa(n,x)		({ia64_nasid_va _v; _v.l = (long) (x); _v.f.nasid = (x) ? (n) : 0; _v.f.reg = 0; _v.l;})
+
+/*
+ * The following variables are passed thru registersfrom the configuration file and
+ * are set via the _start function.
+ */
+long		base_nasid;
+long		num_cpus;
+long		bsp_entry_pc=0;
+long		num_nodes;
+long		app_entry_pc;
+int		bsp_lid;
+func_ptr_t	ap_entry;
+
+
+static char fw_mem[(  sizeof(efi_system_table_t)
+		    + sizeof(efi_runtime_services_t)
+		    + NUM_EFI_DESCS*sizeof(efi_config_table_t)
+		    + sizeof(struct ia64_sal_systab)
+		    + sizeof(struct ia64_sal_desc_entry_point)
+		    + sizeof(struct ia64_sal_desc_ap_wakeup)
+		    + sizeof(acpi_rsdp_t)
+		    + sizeof(acpi_rsdt_t)
+		    + sizeof(acpi_sapic_t)
+		    + MAX_LSAPICS*(sizeof(acpi_entry_lsapic_t))
+		    + (1+8*MAX_NODES)*(sizeof(efi_memory_desc_t))
+		    + sizeof(ia64_sal_desc_ptc_t) +
+		    + MAX_NODES*sizeof(ia64_sal_ptc_domain_info_t) +
+		    + MAX_CPUS*sizeof(ia64_sal_ptc_domain_proc_entry_t) +
+		    + 1024)] __attribute__ ((aligned (8)));
+
+/*
+ * Very ugly, but we need this in the simulator only.  Once we run on
+ * real hw, this can all go away.
+ */
+extern void pal_emulator_static (void);
+
+asm ("
+	.text
+	.proc pal_emulator_static
+pal_emulator_static:
+	mov r8=-1
+	cmp.eq p6,p7=6,r28		/* PAL_PTCE_INFO */
+(p7)	br.cond.sptk.few 1f
+	;;
+	mov r8=0			/* status = 0 */
+	movl r9=0x500000000		/* tc.base */
+	movl r10=0x0000000200000003	/* count[0], count[1] */
+	movl r11=0x1000000000002000	/* stride[0], stride[1] */
+	br.cond.sptk.few rp
+
+1:	cmp.eq p6,p7=14,r28		/* PAL_FREQ_RATIOS */
+(p7)	br.cond.sptk.few 1f
+	mov r8=0			/* status = 0 */
+	movl r9 =0x100000064		/* proc_ratio (1/100) */
+	movl r10=0x100000100		/* bus_ratio<<32 (1/256) */
+	movl r11=0x10000000a		/* itc_ratio<<32 (1/100) */
+
+1:	cmp.eq p6,p7=22,r28		/* PAL_MC_DRAIN */
+(p7)	br.cond.sptk.few 1f
+	mov r8=0
+	br.cond.sptk.few rp
+
+1:	cmp.eq p6,p7=23,r28		/* PAL_MC_EXPECTED */
+(p7)	br.cond.sptk.few 1f
+	mov r8=0
+	br.cond.sptk.few rp
+
+1:	br.cond.sptk.few rp
+	.endp pal_emulator_static\n");
+
+
+static efi_status_t
+efi_get_time (efi_time_t *tm, efi_time_cap_t *tc)
+{
+	if (tm) {
+		memset(tm, 0, sizeof(*tm));
+		tm->year = 2000;
+		tm->month = 2;
+		tm->day = 13;
+		tm->hour = 10;
+		tm->minute = 11;
+		tm->second = 12;
+	}
+
+	if (tc) {
+		tc->resolution = 10;
+		tc->accuracy = 12;
+		tc->sets_to_zero = 1;
+	}
+
+	return EFI_SUCCESS;
+}
+
+static void
+efi_reset_system (int reset_type, efi_status_t status, unsigned long data_size, efi_char16_t *data)
+{
+	while(1);	/* Is there a pseudo-op to stop medusa */
+}
+
+static efi_status_t
+efi_success (void)
+{
+	return EFI_SUCCESS;
+}
+
+static efi_status_t
+efi_unimplemented (void)
+{
+	return EFI_UNSUPPORTED;
+}
+
+static long
+sal_emulator (long index, unsigned long in1, unsigned long in2,
+	      unsigned long in3, unsigned long in4, unsigned long in5,
+	      unsigned long in6, unsigned long in7)
+{
+	register long r9 asm ("r9") = 0;
+	register long r10 asm ("r10") = 0;
+	register long r11 asm ("r11") = 0;
+	long status;
+
+	/*
+	 * Don't do a "switch" here since that gives us code that
+	 * isn't self-relocatable.
+	 */
+	status = 0;
+	if (index == SAL_FREQ_BASE) {
+		switch (in1) {
+		      case SAL_FREQ_BASE_PLATFORM:
+			r9 = 500000000;
+			break;
+
+		      case SAL_FREQ_BASE_INTERVAL_TIMER:
+			/*
+			 * Is this supposed to be the cr.itc frequency
+			 * or something platform specific?  The SAL
+			 * doc ain't exactly clear on this...
+			 */
+			r9 = 700000000;
+			break;
+
+		      case SAL_FREQ_BASE_REALTIME_CLOCK:
+			r9 = 1;
+			break;
+
+		      default:
+			status = -1;
+			break;
+		}
+	} else if (index == SAL_SET_VECTORS) {
+		if (in1 == SAL_VECTOR_OS_BOOT_RENDEZ) {
+			func_ptr_t	*fp;
+			fp = ADDR_OF(&ap_entry);
+			fp->pc = in2;
+			fp->gp = in3;
+		} else {
+			status = -1;
+		}
+		;
+	} else if (index == SAL_GET_STATE_INFO) {
+		;
+	} else if (index == SAL_GET_STATE_INFO_SIZE) {
+		;
+	} else if (index == SAL_CLEAR_STATE_INFO) {
+		;
+	} else if (index == SAL_MC_RENDEZ) {
+		;
+	} else if (index == SAL_MC_SET_PARAMS) {
+		;
+	} else if (index == SAL_CACHE_FLUSH) {
+		;
+	} else if (index == SAL_CACHE_INIT) {
+		;
+	} else if (index == SAL_UPDATE_PAL) {
+		;
+	} else {
+		status = -1;
+	}
+	asm volatile ("" :: "r"(r9), "r"(r10), "r"(r11));
+	return status;
+}
+
+
+/*
+ * This is here to work around a bug in egcs-1.1.1b that causes the
+ * compiler to crash (seems like a bug in the new alias analysis code.
+ */
+void *
+id (long addr)
+{
+	return (void *) addr;
+}
+
+
+/*
+ * Fix the addresses in a function pointer by adding base node address
+ * to pc & gp.
+ */
+void
+fix_function_pointer(void *fp)
+{
+	func_ptr_t	*_fp;
+
+	_fp = fp;
+	_fp->pc = __fwtab_pa(base_nasid, _fp->pc);
+	_fp->gp = __fwtab_pa(base_nasid, _fp->gp);
+}
+
+
+void
+sys_fw_init (const char *args, int arglen, int bsp)
+{
+	/*
+	 * Use static variables to keep from overflowing the RSE stack
+	 */
+	static efi_system_table_t *efi_systab;
+	static efi_runtime_services_t *efi_runtime;
+	static efi_config_table_t *efi_tables;
+	static ia64_sal_desc_ptc_t *sal_ptc;
+	static ia64_sal_ptc_domain_info_t *sal_ptcdi;
+	static ia64_sal_ptc_domain_proc_entry_t *sal_ptclid;
+	static acpi_rsdp_t *acpi_systab;
+	static acpi_rsdt_t *acpi_rsdt;
+	static acpi_sapic_t *acpi_sapic;
+	static acpi_entry_lsapic_t *acpi_lsapic;
+	static struct ia64_sal_systab *sal_systab;
+	static efi_memory_desc_t *efi_memmap, *md;
+	static unsigned long *pal_desc, *sal_desc;
+	static struct ia64_sal_desc_entry_point *sal_ed;
+	static struct ia64_boot_param *bp;
+	static struct ia64_sal_desc_ap_wakeup *sal_apwake;
+	static unsigned char checksum = 0;
+	static char *cp, *cmd_line, *vendor;
+	static int mdsize, domain, last_domain ;
+	static int cnode, nasid, cpu, num_memmd, cpus_found;
+
+	/*
+	 * Pass the parameter base address to the build_efi_xxx routines.
+	 */
+	build_init(8LL*GB*base_nasid);
+
+	num_nodes = GetNumNodes();
+	num_cpus = GetNumCpus();
+
+
+	memset(fw_mem, 0, sizeof(fw_mem));
+
+	pal_desc = (unsigned long *) &pal_emulator_static;
+	sal_desc = (unsigned long *) &sal_emulator;
+	fix_function_pointer(&pal_emulator_static);
+	fix_function_pointer(&sal_emulator);
+
+	/* Align this to 16 bytes, probably EFI does this  */
+	mdsize = (sizeof(efi_memory_desc_t) + 15) & ~15 ;
+
+	cp = fw_mem;
+	efi_systab  = (void *) cp; cp += sizeof(*efi_systab);
+	efi_runtime = (void *) cp; cp += sizeof(*efi_runtime);
+	efi_tables  = (void *) cp; cp += NUM_EFI_DESCS*sizeof(*efi_tables);
+	sal_systab  = (void *) cp; cp += sizeof(*sal_systab);
+	sal_ed      = (void *) cp; cp += sizeof(*sal_ed);
+	sal_ptc     = (void *) cp; cp += sizeof(*sal_ptc);
+	sal_apwake  = (void *) cp; cp += sizeof(*sal_apwake);
+	acpi_systab = (void *) cp; cp += sizeof(*acpi_systab);
+	acpi_rsdt   = (void *) cp; cp += sizeof(*acpi_rsdt);
+	acpi_sapic  = (void *) cp; cp += sizeof(*acpi_sapic);
+	acpi_lsapic = (void *) cp; cp += num_cpus*sizeof(*acpi_lsapic);
+	vendor 	    = (char *) cp; cp += 32;
+	efi_memmap  = (void *) cp; cp += 8*32*sizeof(*efi_memmap);
+	sal_ptcdi   = (void *) cp; cp += CPUS_PER_FSB*(1+num_nodes)*sizeof(*sal_ptcdi);
+	sal_ptclid  = (void *) cp; cp += ((3+num_cpus)*sizeof(*sal_ptclid)+7)/8*8;
+	cmd_line    = (void *) cp;
+
+	if (args) {
+		if (arglen >= 1024)
+			arglen = 1023;
+		memcpy(cmd_line, args, arglen);
+	} else {
+		arglen = 0;
+	}
+	cmd_line[arglen] = '\0';
+#ifdef BRINGUP
+	/* for now, just bring up bash */
+	strcpy(cmd_line, "init=/bin/bash");
+#else
+	strcpy(cmd_line, "");
+#endif
+
+	memset(efi_systab, 0, sizeof(efi_systab));
+	efi_systab->hdr.signature = EFI_SYSTEM_TABLE_SIGNATURE;
+	efi_systab->hdr.revision  = EFI_SYSTEM_TABLE_REVISION;
+	efi_systab->hdr.headersize = sizeof(efi_systab->hdr);
+	efi_systab->fw_vendor = __fwtab_pa(base_nasid, vendor);
+	efi_systab->fw_revision = 1;
+	efi_systab->runtime = __fwtab_pa(base_nasid, efi_runtime);
+	efi_systab->nr_tables = 2;
+	efi_systab->tables = __fwtab_pa(base_nasid, efi_tables);
+	memcpy(vendor, "S\0i\0l\0i\0c\0o\0n\0-\0G\0r\0a\0p\0h\0i\0c\0s\0\0", 32);
+
+	efi_runtime->hdr.signature = EFI_RUNTIME_SERVICES_SIGNATURE;
+	efi_runtime->hdr.revision = EFI_RUNTIME_SERVICES_REVISION;
+	efi_runtime->hdr.headersize = sizeof(efi_runtime->hdr);
+	efi_runtime->get_time = __fwtab_pa(base_nasid, &efi_get_time);
+	efi_runtime->set_time = __fwtab_pa(base_nasid, &efi_unimplemented);
+	efi_runtime->get_wakeup_time = __fwtab_pa(base_nasid, &efi_unimplemented);
+	efi_runtime->set_wakeup_time = __fwtab_pa(base_nasid, &efi_unimplemented);
+	efi_runtime->set_virtual_address_map = __fwtab_pa(base_nasid, &efi_success);
+	efi_runtime->get_variable = __fwtab_pa(base_nasid, &efi_unimplemented);
+	efi_runtime->get_next_variable = __fwtab_pa(base_nasid, &efi_unimplemented);
+	efi_runtime->set_variable = __fwtab_pa(base_nasid, &efi_unimplemented);
+	efi_runtime->get_next_high_mono_count = __fwtab_pa(base_nasid, &efi_unimplemented);
+	efi_runtime->reset_system = __fwtab_pa(base_nasid, &efi_reset_system);
+
+	efi_tables->guid = SAL_SYSTEM_TABLE_GUID;
+	efi_tables->table = __fwtab_pa(base_nasid, sal_systab);
+	efi_tables++;
+	efi_tables->guid = ACPI_TABLE_GUID;
+	efi_tables->table = __fwtab_pa(base_nasid, acpi_systab);
+	fix_function_pointer(&efi_unimplemented);
+	fix_function_pointer(&efi_get_time);
+	fix_function_pointer(&efi_success);
+	fix_function_pointer(&efi_reset_system);
+
+	/* fill in the ACPI system table: */
+	memcpy(acpi_systab->signature, "RSD PTR ", 8);
+	acpi_systab->rsdt = (acpi_rsdt_t*)__fwtab_pa(base_nasid, acpi_rsdt);
+
+	memcpy(acpi_rsdt->header.signature, "RSDT",4);
+	acpi_rsdt->header.length = sizeof(acpi_rsdt_t);
+	memcpy(acpi_rsdt->header.oem_id, "SGI", 3);
+	memcpy(acpi_rsdt->header.oem_table_id, "SN1", 3);
+	acpi_rsdt->header.oem_revision = 0x00010001;
+	acpi_rsdt->entry_ptrs[0] = __fwtab_pa(base_nasid, acpi_sapic);
+
+	memcpy(acpi_sapic->header.signature, "SPIC ", 4);
+	acpi_sapic->header.length = sizeof(acpi_sapic_t)+num_cpus*sizeof(acpi_entry_lsapic_t);
+	for (cnode=0; cnode<num_nodes; cnode++) {
+		nasid = GetNasid(cnode);
+		for(cpu=0; cpu<CPUS_PER_NODE; cpu++) {
+			if (!IsCpuPresent(cnode, cpu))
+				continue;
+			acpi_lsapic->type = ACPI_ENTRY_LOCAL_SAPIC;
+			acpi_lsapic->length = sizeof(acpi_entry_lsapic_t);
+			acpi_lsapic->acpi_processor_id = cnode*4+cpu;
+			acpi_lsapic->flags = LSAPIC_ENABLED|LSAPIC_PRESENT;
+			acpi_lsapic->eid = cpu;
+			acpi_lsapic->id = nasid;
+			acpi_lsapic++;
+		}
+	}
+
+
+	/* fill in the SAL system table: */
+	memcpy(sal_systab->signature, "SST_", 4);
+	sal_systab->size = sizeof(*sal_systab);
+	sal_systab->sal_rev_minor = 1;
+	sal_systab->sal_rev_major = 0;
+	sal_systab->entry_count = 3;
+
+	strcpy(sal_systab->oem_id, "SGI");
+	strcpy(sal_systab->product_id, "SN1");
+
+	/* fill in an entry point: */	
+	sal_ed->type = SAL_DESC_ENTRY_POINT;
+	sal_ed->pal_proc = __fwtab_pa(base_nasid, pal_desc[0]);
+	sal_ed->sal_proc = __fwtab_pa(base_nasid, sal_desc[0]);
+	sal_ed->gp = __fwtab_pa(base_nasid, sal_desc[1]);
+
+	/* kludge the PTC domain info */
+	sal_ptc->type = SAL_DESC_PTC;
+	sal_ptc->num_domains = 0;
+	sal_ptc->domain_info = __fwtab_pa(base_nasid, sal_ptcdi);
+	cpus_found = 0;
+	last_domain = -1;
+	sal_ptcdi--;
+	for (cnode=0; cnode<num_nodes; cnode++) {
+		nasid = GetNasid(cnode);
+		for(cpu=0; cpu<CPUS_PER_NODE; cpu++) {
+			if (IsCpuPresent(cnode, cpu)) {
+				domain = cnode*CPUS_PER_NODE + cpu/CPUS_PER_FSB;
+				if (domain != last_domain) {
+					sal_ptc->num_domains++;
+					sal_ptcdi++;
+					sal_ptcdi->proc_count = 0;
+					sal_ptcdi->proc_list = __fwtab_pa(base_nasid, sal_ptclid);
+					last_domain = domain;
+				}
+				sal_ptcdi->proc_count++;
+				sal_ptclid->id = nasid;
+				sal_ptclid->eid = cpu;
+				sal_ptclid++;
+				cpus_found++;
+			}
+		}
+	}
+
+	if (cpus_found != num_cpus)
+		FPROM_BUG();
+
+	/* Make the AP WAKEUP entry */
+	sal_apwake->type = SAL_DESC_AP_WAKEUP;
+	sal_apwake->mechanism = IA64_SAL_AP_EXTERNAL_INT;
+	sal_apwake->vector = 18;
+
+	for (cp = (char *) sal_systab; cp < (char *) efi_memmap; ++cp)
+		checksum += *cp;
+
+	sal_systab->checksum = -checksum;
+
+	md = &efi_memmap[0];
+	num_memmd = build_efi_memmap((void *)md, mdsize) ;
+
+	bp = id(ZERO_PAGE_ADDR + (((long)base_nasid)<<33));
+	bp->efi_systab = __fwtab_pa(base_nasid, &fw_mem);
+	bp->efi_memmap = __fwtab_pa(base_nasid, efi_memmap);
+	bp->efi_memmap_size = num_memmd*mdsize;
+	bp->efi_memdesc_size = mdsize;
+	bp->efi_memdesc_version = 0x101;
+	bp->command_line = __fwtab_pa(base_nasid, cmd_line);
+	bp->console_info.num_cols = 80;
+	bp->console_info.num_rows = 25;
+	bp->console_info.orig_x = 0;
+	bp->console_info.orig_y = 24;
+	bp->num_pci_vectors = 0;
+	bp->fpswa = 0;
+
+	/*
+	 * Now pick the BSP & store it LID value in
+	 * a global variable. Note if BSP is greater than last cpu,
+	 * pick the last cpu.
+	 */
+	for (cnode=0; cnode<num_nodes; cnode++) {
+		for(cpu=0; cpu<CPUS_PER_NODE; cpu++) {
+			if (!IsCpuPresent(cnode, cpu))
+				continue;
+			bsp_lid = (GetNasid(cnode)<<24) | (cpu<<16);
+			if (bsp-- > 0)
+				continue;
+			return;
+		}
+	}
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/main.c linux/arch/ia64/sn/fprom/main.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/main.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/main.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,110 @@
+/* 
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Jack Steiner (steiner@sgi.com)
+ */
+
+
+
+#include <linux/types.h>
+#include <asm/bitops.h>
+
+void bedrock_init(int);
+void synergy_init(int, int);
+void sys_fw_init (const char *args, int arglen, int bsp);
+
+volatile int	bootmaster=0;		/* Used to pick bootmaster */
+volatile int	nasidmaster[128]={0};	/* Used to pick node/synergy masters */
+int		init_done=0;
+extern int	bsp_lid;
+
+#define get_bit(b,p)	(((*p)>>(b))&1)
+
+int
+fmain(int lid, int bsp) {
+	int	syn, nasid, cpu;
+
+	/*
+	 * First lets figure out who we are. This is done from the
+	 * LID passed to us.
+	 */
+	nasid = (lid>>24);
+	syn = (lid>>17)&1;
+	cpu = (lid>>16)&1;
+
+	/*
+	 * Now pick a synergy master to initialize synergy registers.
+	 */
+	if (test_and_set_bit(syn, &nasidmaster[nasid]) == 0) {
+		synergy_init(nasid, syn);
+		test_and_set_bit(syn+2, &nasidmaster[nasid]);
+	} else
+		while (get_bit(syn+2, &nasidmaster[nasid]) == 0);
+	
+	/*
+	 * Now pick a nasid master to initialize Bedrock registers.
+	 */
+	if (test_and_set_bit(8, &nasidmaster[nasid]) == 0) {
+		bedrock_init(nasid);
+		test_and_set_bit(9, &nasidmaster[nasid]);
+	} else
+		while (get_bit(9, &nasidmaster[nasid]) == 0);
+	
+
+	/*
+	 * Now pick a BSP & finish init.
+	 */
+	if (test_and_set_bit(0, &bootmaster) == 0) {
+		sys_fw_init(0, 0, bsp);
+		test_and_set_bit(1, &bootmaster);
+	} else
+		while (get_bit(1, &bootmaster) == 0);
+
+	return (lid == bsp_lid);
+}
+
+
+void
+bedrock_init(int nasid)
+{
+	nasid = nasid;		/* to quiet gcc */
+}
+
+
+void
+synergy_init(int nasid, int syn)
+{
+	long	*base;
+	long	off;
+
+	/*
+	 * Enable all FSB flashed interrupts.
+	 * ZZZ - I'd really like defines for this......
+	 */
+	base = (long*)0x80000e0000000000LL;		/* base of synergy regs */
+	for (off = 0x2a0; off < 0x2e0; off+=8)		/* offset for VEC_MASK_{0-3}_A/B */
+		*(base+off/8) = -1LL;
+
+	/*
+	 * Set the NASID in the FSB_CONFIG register.
+	 */
+	base = (long*)0x80000e0000000450LL;
+	*base = (long)((nasid<<16)|(syn<<9));
+}
+
+
+/* Why isnt there a bcopy/memcpy in lib64.a */
+
+void* 
+memcpy(void * dest, const void *src, size_t count)
+{
+	char *s, *se, *d;
+
+	for(d=dest, s=(char*)src, se=s+count; s<se; s++, d++)
+		*d = *s;
+	return dest;
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/fprom/runsim linux/arch/ia64/sn/fprom/runsim
--- v2.4.0-prerelease/linux/arch/ia64/sn/fprom/runsim	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/fprom/runsim	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,372 @@
+#!/bin/sh
+
+# Script for running PROMs and LINUX kernwls on medusa. 
+# Type "sim -H" for instructions.
+
+MEDUSA=${MEDUSA:-/home/rickc/official_medusa/medusa}
+
+# ------------------ err -----------------------
+err() {
+	echo "ERROR - $1"
+	exit 1
+}
+
+# ----------------  help ----------------------
+help() {
+cat <<END
+Script for running a PROM or LINUX kernel under medusa.
+This script creates a control file, creates links to the appropriate
+linux/prom files, and/or calls medusa to make simulation runs.
+
+Usage:  
+   Initial setup:
+   	sim [-c <config_file>] <-p> | <-k>  [<work_dir>]
+		-p	Create PROM control file & links
+		-k	Create LINUX control file & links
+		-c<cf>	Control file name				[Default: cf]
+		<work_dir> Path to directory that contains the linux or PROM files.
+		    The directory can be any of the following:
+		       (linux simulations)
+		       		worktree
+				worktree/linux
+				any directory with vmlinux, vmlinux.sym & fprom files
+			(prom simulations)
+				worktree
+				worktree/stand/arcs/IP37prom/dev
+				any directory with fw.bin & fw.sim files
+
+    Simulations:
+	sim  [-X <n>] [-o <output>] [-M] [<config_file>]
+		-c<cf>	Control file name				[Default: cf]
+		-M	Pipe output thru fmtmedusa
+		-o	Output filename (copy of all commands/output)	[Default: simout]
+		-X	Specifies number of instructions to execute	[Default: 0]
+			(Used only in auto test mode - not described here)
+
+Examples:
+	sim -p <promtree>	# create control file (cf) & links for prom simulations
+	sim -k <linuxtree>	# create control file (cf) & links for linux simulations
+	sim -p -c cfprom	# create a prom control file (cfprom) only. No links are made.
+
+	sim			# run medusa using previously created links &
+				#   control file (cf).
+END
+exit 1
+}
+
+# ----------------------- create control file header --------------------
+create_cf_header() {
+cat <<END >>$CF
+#
+# Template for a control file for running linux kernels under medusa. 
+# You probably want to make mods here but this is a good starting point.
+#
+
+# Preferences
+setenv cpu_stepping A
+setenv exceptionPrint off
+setenv interrupt_messages off
+setenv lastPCsize 100000
+setenv low_power_mode on
+setenv partialIntelChipSet on
+setenv printIntelMessages off
+setenv prom_write_action halt
+setenv prom_write_messages on
+setenv step_quantum 100
+setenv swizzling on
+setenv tsconsole on
+setenv uart_echo on
+symbols on
+
+# IDE disk params
+setenv diskCylinders 611
+setenv bootDrive C
+setenv diskHeads 16
+setenv diskPath idedisk
+setenv diskPresent 1
+setenv diskSpt 63
+
+# Hardware config
+setenv coherency_type nasid
+setenv cpu_cache_type default
+setenv synergy_cache_type syn_cac_64m_8w
+
+# Numalink config
+setenv route_enable on
+setenv network_type xbar		# Select [xbar|router]
+setenv network_warning 0xff
+
+END
+}
+
+
+# ------------------ create control file entries for linux simulations -------------
+create_cf_linux() {
+cat <<END >>$CF
+# Kernel specific options
+setenv mca_on_memory_failure off
+setenv LOADPC 0x00100000		# FPROM load address/entry point (8 digits!)
+sr g 9 0xe000000000520000		# Kernel entry point
+setenv symbol_table vmlinux.sym
+load fprom
+load vmlinux
+
+# Useful breakpoints to always have set. Add more if desired.
+break 0xe000000000505e00	all	# dispatch_to_fault_handler
+break panic			all	# stop on panic
+break die_if_kernel		all	# may as well stop
+
+END
+}
+
+# ------------------ create control file entries for prom simulations ---------------
+create_cf_prom() {
+	SYM2=""
+	ADDR="0x80000000ff800000"
+	[ "$EMBEDDED_LINUX" != "0" ] || SYM2="setenv symbol_table2 vmlinux.sym"
+	[ "$SIZE" = "8MB" ] || ADDR="0x80000000ffc00000"
+	cat <<END >>$CF
+# PROM specific options
+setenv mca_on_memory_failure on
+setenv LOADPC 0x80000000ffffffb0
+setenv promFile fw.bin
+setenv promAddr $ADDR
+setenv symbol_table fw.sym
+$SYM2
+
+# Useful breakpoints to always have set. Add more if desired.
+break Pr_ivt_gexx 		all
+break Pr_ivt_brk		all
+break Pr_PROM_Panic_Spin	all
+break Pr_PROM_Panic		all
+break Pr_PROM_C_Panic		all
+break Pr_fled_die		all
+break Pr_ResetNow		all
+break Pr_zzzbkpt		all
+
+END
+}
+
+
+# ------------------ create control file entries for memory configuration -------------
+create_cf_memory() {
+cat <<END >>$CF
+# CPU/Memory map format:
+#	setenv nodeN_memory_config 0xBSBSBSBS
+#		B=banksize (0=unused, 1=64M, 2=128M, .., 5-1G, c=8M, d=16M, e=32M)
+#		S=bank enable (0=both disable, 3=both enable, 2=bank1 enable, 1=bank0 enable)
+#		  rightmost digits are for bank 0, the lowest address.
+#	setenv nodeN_nasid <nasid>
+#		specifies the NASID for the node. This is used ONLY if booting the kernel.
+#		On PROM configurations, set to 0 - PROM will change it later.
+#	setenv nodeN_cpu_config <cpu_mask>
+#		Set bit number N to 1 to enable cpu N. Ex., a value of 5 enables cpu 0 & 2.
+#
+# Repeat the above 3 commands for each node.
+#
+# For kernel, default to 32MB. Although this is not a valid hardware configuration,
+# it runs faster on medusa. For PROM, 64MB is smallest allowed value.
+
+setenv node0_cpu_config		0x1	# Enable only cpu 0 on the node
+END
+
+if [ $LINUX -eq 1 ] ; then
+cat <<END >>$CF
+setenv node0_nasid		0	# cnode 0 has NASID 0
+setenv node0_memory_config 	0xe1	# 32MB
+END
+else
+cat <<END >>$CF
+setenv node0_memory_config 	0x11	# 64MB
+END
+fi
+}
+
+# -------------------- set links to linux files -------------------------
+set_linux_links() {
+	if [ -d $D/linux/arch ] ; then
+		D=$D/linux
+	elif [ -d $D/arch -o -e vmlinux.sym ] ; then
+		D=$D
+	else
+		err "cant determine directory for linux binaries"
+	fi
+	rm -rf vmlinux vmlinux.sym fprom
+	ln -s $D/vmlinux vmlinux
+	ln -s $D/vmlinux.sym vmlinux.sym
+	if [ -d $D/arch ] ; then
+		ln -s $D/arch/ia64/sn/fprom/fprom fprom
+	else
+		ln -s $D/fprom fprom
+	fi
+	echo "  .. Created links to linux files"	
+}
+
+# -------------------- set links to prom files -------------------------
+set_prom_links() {
+	if [ -d $D/stand ] ; then
+		D=$D/stand/arcs/IP37prom/dev
+	elif [ -d $D/sal ] ; then
+		D=$D
+	else
+		err "cant determine directory for PROM binaries"
+	fi
+	SETUP="$D/../../../../.setup"
+	grep -q '^ *setenv *PROMSIZE *8MB' $SETUP
+	if [ $? -eq 0 ] ; then
+		SIZE="8MB"
+	else
+		SIZE="4MB"
+	fi
+	grep -q '^ *setenv *LAUNCH_VMLINUX' $SETUP
+	EMBEDDED_LINUX=$?
+	rm -f fw.bin fw.map fw.sym vmlinux vmlinux.sym fprom
+	SDIR="SN1IA${SIZE}.O"
+	BIN="SN1IAip37prom${SIZE}"
+	ln -s $D/$SDIR/$BIN.bin fw.bin
+	ln -s $D/$SDIR/$BIN.map fw.map
+	ln -s $D/$SDIR/$BIN.sym fw.sym
+	echo "  .. Created links to $SIZE prom files"
+	if [ $EMBEDDED_LINUX -eq 0 ] ; then
+		ln -s $D/linux/vmlinux vmlinux
+		ln -s $D/linux/vmlinux.sym vmlinux.sym
+		if [ -d linux/arch ] ; then
+			ln -s $D/linux/arch/ia64/sn/fprom/fprom fprom
+		else
+			ln -s $D/linux/fprom fprom
+		fi
+		echo "  .. Created links to embedded linux files in prom tree"
+	fi
+}
+
+# --------------- start of shell script --------------------------------
+OUT="simout"
+FMTMED=0
+STEPCNT=0
+PROM=0
+LINUX=0
+NCF="cf"
+while getopts "HMX:c:o:pk" c ; do
+        case ${c} in
+                H) help;;
+		M) FMTMED=1;;
+		X) STEPCNT=${OPTARG};;
+		c) NCF=${OPTARG};;
+		k) PROM=0;LINUX=1;;
+		p) PROM=1;LINUX=0;;
+		o) OUT=${OPTARG};;
+                \?) exit 1;;
+        esac
+done
+shift `expr ${OPTIND} - 1`
+
+# Check if command is for creating control file and/or links to images.
+if [ $PROM -eq 1 -o $LINUX -eq 1 ] ; then
+	CF=$NCF
+	[ ! -f $CF ] || err "wont overwrite an existing control file ($CF)"
+	if [ $# -gt 0 ] ; then
+		D=$1
+		[ -d $D ] || err "cannot find directory $D"
+		[ $PROM -eq 0 ]  || set_prom_links
+		[ $LINUX -eq 0 ] || set_linux_links
+	fi
+	create_cf_header
+	[ $PROM -eq 0 ]  || create_cf_prom
+	[ $LINUX -eq 0 ] || create_cf_linux
+	create_cf_memory
+	echo "  .. Basic control file created (in $CF). You might want to edit"
+	echo "     this file (at least, look at it)."
+	exit 0
+fi
+
+# Verify that the control file exists
+CF=${1:-$NCF}
+[ -f $CF ] || err "No control file exists. For help, type: $0 -H"
+
+# Build the .cf files from the user control file. The .cf file is
+# identical except that the actual start & load addresses are inserted
+# into the file. In addition, the FPROM commands for configuring memory
+# and LIDs are generated. 
+
+rm -f .cf .cf1 .cf2
+awk '
+function strtonum(n) {
+	 if (substr(n,1,2) != "0x")
+	 	return int(n)
+	 n = substr(n,3)
+	 r=0
+	 while (length(n) > 0) {
+	 	r = r*16+(index("0123456789abcdef", substr(n,1,1))-1)
+		n = substr(n,2)
+	 }
+	 return r
+	}
+/^#/   	{next}
+/^$/	{next}
+/^setenv *LOADPC/               {loadpc = $3; next}
+/^setenv *node._cpu_config/	{n=int(substr($2,5,1)); cpuconf[n] = strtonum($3); print; next}
+/^setenv *node._memory_config/	{n=int(substr($2,5,1)); memconf[n] = strtonum($3); print; next}
+/^setenv *node._nasid/		{n=int(substr($2,5,1)); nasid[n] = strtonum($3); print; next}
+		{print}
+END	{
+	 # Generate the memmap info that starts at the beginning of
+	 # the node the kernel was loaded on.
+	 loadnasid = nasid[0]
+	 cnode = 0
+	 for (i=0; i<128; i++) {
+		if (memconf[i] != "") {
+			printf "sm 0x%x%08x 0x%x%04x%04x\n", 
+				2*loadnasid, 8*cnodes+8, memconf[i], cpuconf[i], nasid[i]
+			cnodes++
+			cpus += substr("0112122312232334", cpuconf[i]+1,1)
+		}
+	 }
+	 printf "sm 0x%x00000000 0x%x%08x\n", 2*loadnasid, cnodes, cpus
+	 printf "setenv number_of_nodes %d\n", cnodes
+
+	 # Now set the starting PC for each cpu.
+	 cnode = 0
+	 lowcpu=-1
+	 for (i=0; i<128; i++) {
+		if (memconf[i] != "") {
+			printf "setnode %d\n", cnode
+			conf = cpuconf[i]
+			for (j=0; j<4; j++) {
+				if (conf != int(conf/2)*2) {
+	 				printf "setcpu %d\n", j
+					if (length(loadpc) == 18)
+						printf "sr pc %s\n", loadpc
+					else
+						printf "sr pc 0x%x%s\n", 2*loadnasid, substr(loadpc,3)
+					if (lowcpu == -1)
+						lowcpu = j
+				}
+				conf = int(conf/2)
+			}
+			cnode++
+		}
+	 }
+	 printf "setnode 0\n"
+	 printf "setcpu %d\n", lowcpu
+	}
+' <$CF >.cf
+
+# Now build the .cf1 & .cf2 control files.
+CF2_LINES="^sm |^break |^run |^si |^quit |^symbols "
+egrep  "$CF2_LINES" .cf >.cf2
+egrep -v "$CF2_LINES" .cf >.cf1
+if [ $STEPCNT -ne 0 ] ; then
+	echo "s $STEPCNT" >>.cf2
+	echo "lastpc 1000" >>.cf2
+	echo "q" >>.cf2
+fi
+echo "script-on $OUT" >>.cf2
+
+# Now start medusa....
+if [ $FMTMED -ne 0 ] ; then
+	$MEDUSA -system mpsn1 -c .cf1 -i .cf2 |  fmtmedusa
+elif [ $STEPCNT -eq 0 ] ; then
+	$MEDUSA -system mpsn1 -c .cf1 -i .cf2 
+else
+	$MEDUSA -system mpsn1 -c .cf1 -i .cf2 2>&1 
+fi
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/Makefile linux/arch/ia64/sn/io/Makefile
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/Makefile	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/Makefile	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,32 @@
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.  See the file "COPYING" in the main directory of this archive
+# for more details.
+#
+# Copyright (C) 2000 Silicon Graphics, Inc.
+# Copyright (C) Jack Steiner (steiner@sgi.com)
+#
+#
+# Makefile for the linux kernel.
+#
+# Note! Dependencies are done automagically by 'make dep', which also
+# removes any old dependencies. DON'T put your own dependencies here
+# unless it's something special (ie not a .c file).
+#
+# Note 2! The CFLAGS definitions are now in the main makefile...
+
+EXTRA_CFLAGS	:= -DSN -DLANGUAGE_C=1 -D_LANGUAGE_C=1 -I. -DBRINGUP \
+		   -DDIRECT_L1_CONSOLE -DNUMA_BASE -DSIMULATED_KLGRAPH \
+		   -DNUMA_MIGR_CONTROL -DLITTLE_ENDIAN -DREAL_HARDWARE \
+		   -DNEW_INTERRUPTS -DCONFIG_IA64_SGI_IO
+O_TARGET := sgiio.o
+O_OBJS   := stubs.o sgi_if.o pciio.o pcibr.o xtalk.o xbow.o xswitch.o hubspc.o \
+		klgraph_hack.o io.o hubdev.o \
+		hcl.o labelcl.o invent.o klgraph.o klconflib.o sgi_io_sim.o \
+		module.o sgi_io_init.o klgraph_hack.o ml_SN_init.o \
+		ml_SN_intr.o ip37.o \
+		ml_iograph.o hcl_util.o cdl.o \
+		mem_refcnt.o devsupport.o alenlist.o pci_bus_cvlink.o \
+		eeprom.o pci.o pci_dma.o l1.o l1_command.o
+
+include $(TOPDIR)/Rules.make
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/alenlist.c linux/arch/ia64/sn/io/alenlist.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/alenlist.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/alenlist.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,901 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+/* Implementation of Address/Length Lists. */
+
+
+#include <linux/types.h>
+#include <linux/config.h>
+#include <linux/slab.h>
+#include <asm/sn/sgi.h>
+#include <asm/sn/alenlist.h>
+#include <asm/sn/mmzone_sn1.h>
+
+/*
+ * Logically, an Address/Length List is a list of Pairs, where each pair
+ * holds an Address and a Length, all in some Address Space.  In this
+ * context, "Address Space" is a particular Crosstalk Widget address
+ * space, a PCI device address space, a VME bus address space, a
+ * physical memory address space, etc.
+ *
+ * The main use for these Lists is to provide a single mechanism that
+ * describes where in an address space a DMA occurs.  This allows the
+ * various I/O Bus support layers to provide a single interface for
+ * DMA mapping and DMA translation without regard to how the DMA target
+ * was specified by upper layers.  The upper layers commonly specify a 
+ * DMA target via a buf structure page list, a kernel virtual address,
+ * a user virtual address, a vector of addresses (a la uio and iov), 
+ * or possibly a pfn list.
+ *
+ * Address/Length Lists also enable drivers to take advantage of their
+ * inate scatter/gather capabilities in systems where some address
+ * translation may be required between bus adapters.  The driver forms
+ * a List that represents physical memory targets.  This list is passed
+ * to the various adapters, which apply various translations.  The final
+ * list that's returned to the driver is in terms of its local address
+ * address space -- addresses which can be passed off to a scatter/gather
+ * capable DMA controller.
+ *
+ * The current implementation is intended to be useful both in kernels
+ * that support interrupt threads (INTR_KTHREAD) and in systems that do
+ * not support interrupt threads.  Of course, in the latter case, some
+ * interfaces can be called only within a suspendable context.
+ *
+ * Basic operations on Address/Length Lists include:
+ *	alenlist_create		Create a list
+ *	alenlist_clear		Clear a list
+ *	alenlist_destroy	Destroy a list
+ *	alenlist_append		Append a Pair to the end of a list
+ *	alenlist_replace	Replace a Pair in the middle of a list
+ *	alenlist_get		Get an Address/Length Pair from a list
+ *	alenlist_size		Return the number of Pairs in a list
+ *	alenlist_concat		Append one list to the end of another
+ *	alenlist_clone		Create a new copy of a list
+ *
+ * Operations that convert from upper-level specifications to Address/
+ * Length Lists currently include:
+ *	kvaddr_to_alenlist	Convert from a kernel virtual address
+ *	uvaddr_to_alenlist	Convert from a user virtual address
+ *	buf_to_alenlist		Convert from a buf structure
+ *	alenlist_done		Tell system that we're done with an alenlist
+ *				obtained from a conversion.
+ * Additional convenience operations:
+ *	alenpair_init		Create a list and initialize it with a Pair
+ *	alenpair_get		Peek at the first pair on a List
+ *
+ * A supporting type for Address/Length Lists is an alenlist_cursor_t.  A
+ * cursor marks a position in a List, and determines which Pair is fetched
+ * by alenlist_get.
+ *	alenlist_cursor_create	Allocate and initialize a cursor
+ *	alenlist_cursor_destroy	Free space consumed by a cursor
+ *	alenlist_cursor_init	(Re-)Initialize a cursor to point 
+ *				to the start of a list
+ *	alenlist_cursor_clone	Clone a cursor (at the current offset)
+ *	alenlist_cursor_offset	Return the number of bytes into
+ *				a list that this cursor marks
+ * Multiple cursors can point at various points into a List.  Also, each
+ * list maintains one "internal cursor" which may be updated by alenlist_clear
+ * and alenlist_get.  If calling code simply wishes to scan sequentially
+ * through a list starting at the beginning, and if it is the only user of
+ * a list, it can rely on this internal cursor rather than managing a 
+ * separate explicit cursor.
+ *
+ * The current implementation allows callers to allocate both cursors and
+ * the lists as local stack (structure) variables.  This allows for some
+ * extra efficiency at the expense of forward binary compatibility.  It 
+ * is recommended that customer drivers refrain from local allocation.
+ * In fact, we likely will choose to move the structures out of the public 
+ * header file into a private place in order to discourage this usage.
+ *
+ * Currently, no locking is provided by the alenlist implementation.
+ *
+ * Implementation notes:
+ * For efficiency, Pairs are grouped into "chunks" of, say, 32 Pairs
+ * and a List consists of some number of these chunks.  Chunks are completely
+ * invisible to calling code.  Chunks should be large enough to hold most
+ * standard-sized DMA's, but not so large that they consume excessive space.
+ *
+ * It is generally expected that Lists will be constructed at one time and
+ * scanned at a later time.  It is NOT expected that drivers will scan
+ * a List while the List is simultaneously extended, although this is
+ * theoretically possible with sufficient upper-level locking.
+ *
+ * In order to support demands of Real-Time drivers and in order to support
+ * swapping under low-memory conditions, we support the concept of a
+ * "pre-allocated fixed-sized List".  After creating a List with 
+ * alenlist_create, a driver may explicitly grow the list (via "alenlist_grow")
+ * to a specific number of Address/Length pairs.  It is guaranteed that future 
+ * operations involving this list will never automatically grow the list 
+ * (i.e. if growth is ever required, the operation will fail).  Additionally, 
+ * operations that use alenlist's (e.g. DMA operations) accept a flag which 
+ * causes processing to take place "in-situ"; that is, the input alenlist 
+ * entries are replaced with output alenlist entries.  The combination of 
+ * pre-allocated Lists and in-situ processing allows us to avoid the 
+ * potential deadlock scenario where we sleep (waiting for memory) in the 
+ * swap out path.
+ *
+ * For debugging, we track the number of allocated Lists in alenlist_count
+ * the number of allocated chunks in alenlist_chunk_count, and the number
+ * of allocate cursors in alenlist_cursor_count.  We also provide a debug 
+ * routine, alenlist_show, which dumps the contents of an Address/Length List.
+ *
+ * Currently, Lists are formed by drivers on-demand.  Eventually, we may
+ * associate an alenlist with a buf structure and keep it up to date as
+ * we go along.  In that case, buf_to_alenlist simply returns a pointer
+ * to the existing List, and increments the Lists's reference count.
+ * alenlist_done would decrement the reference count and destroys the List
+ * if it was the last reference.
+ *
+ * Eventually alenlist's may allow better support for user-level scatter/
+ * gather operations (e.g. via readv/writev):  With proper support, we
+ * could potentially handle a vector of reads with a single scatter/gather
+ * DMA operation.  This could be especially useful on NUMA systems where
+ * there's more of a reason for users to use vector I/O operations.
+ *
+ * Eventually, alenlist's may replace kaio lists, vhand page lists,
+ * buffer cache pfdat lists, DMA page lists, etc.
+ */
+
+/* Opaque data types */
+
+/* An Address/Length pair.  */
+typedef struct alen_s {
+	alenaddr_t	al_addr;
+	size_t		al_length;
+} alen_t;
+
+/* 
+ * Number of elements in one chunk of an Address/Length List.
+ *
+ * This size should be sufficient to hold at least an "average" size
+ * DMA request.  Must be at least 1, and should be a power of 2,
+ * for efficiency.
+ */
+#define ALEN_CHUNK_SZ ((512*1024)/NBPP)
+
+/*
+ * A fixed-size set of Address/Length Pairs.  Chunks of Pairs are strung together 
+ * to form a complete Address/Length List.  Chunking is entirely hidden within the 
+ * alenlist implementation, and it simply makes allocation and growth of lists more 
+ * efficient.
+ */
+typedef struct alenlist_chunk_s {
+	alen_t			alc_pair[ALEN_CHUNK_SZ];/* list of addr/len pairs */
+	struct alenlist_chunk_s *alc_next;		/* point to next chunk of pairs */
+} *alenlist_chunk_t;
+
+/* 
+ * An Address/Length List.  An Address/Length List is allocated with alenlist_create.  
+ * Alternatively, a list can be allocated on the stack (local variable of type 
+ * alenlist_t) and initialized with alenpair_init or with a combination of 
+ * alenlist_clear and alenlist_append, etc.  Code which statically allocates these
+ * structures loses forward binary compatibility!
+ *
+ * A statically allocated List is sufficiently large to hold ALEN_CHUNK_SZ pairs.
+ */
+struct alenlist_s {
+	unsigned short		al_flags;
+	unsigned short		al_logical_size;	/* logical size of list, in pairs */
+	unsigned short		al_actual_size;		/* actual size of list, in pairs */
+	struct alenlist_chunk_s	*al_last_chunk;		/* pointer to last logical chunk */
+	struct alenlist_cursor_s al_cursor;		/* internal cursor */
+	struct alenlist_chunk_s	al_chunk;		/* initial set of pairs */
+	alenaddr_t		al_compaction_address;	/* used to compact pairs */
+};
+
+/* al_flags field */
+#define AL_FIXED_SIZE	0x1	/* List is pre-allocated, and of fixed size */
+
+
+zone_t *alenlist_zone = NULL;
+zone_t *alenlist_chunk_zone = NULL;
+zone_t *alenlist_cursor_zone = NULL;
+
+#if DEBUG
+int alenlist_count=0;		/* Currently allocated Lists */
+int alenlist_chunk_count = 0;	/* Currently allocated chunks */
+int alenlist_cursor_count = 0;	/* Currently allocate cursors */
+#define INCR_COUNT(ptr) atomicAddInt((ptr), 1);
+#define DECR_COUNT(ptr) atomicAddInt((ptr), -1);
+#else
+#define INCR_COUNT(ptr)
+#define DECR_COUNT(ptr)
+#endif /* DEBUG */
+
+#if DEBUG
+static void alenlist_show(alenlist_t);
+#endif /* DEBUG */
+
+/*
+ * Initialize Address/Length List management.  One time initialization.
+ */
+void
+alenlist_init(void)
+{
+	alenlist_zone = kmem_zone_init(sizeof(struct alenlist_s), "alenlist");
+	alenlist_chunk_zone = kmem_zone_init(sizeof(struct alenlist_chunk_s), "alchunk");
+	alenlist_cursor_zone = kmem_zone_init(sizeof(struct alenlist_cursor_s), "alcursor");
+#if DEBUG
+	idbg_addfunc("alenshow", alenlist_show);
+#endif /* DEBUG */
+}
+
+
+/*
+ * Initialize an Address/Length List cursor.
+ */
+static void
+do_cursor_init(alenlist_t alenlist, alenlist_cursor_t cursorp)
+{
+	cursorp->al_alenlist = alenlist;
+	cursorp->al_offset = 0;
+	cursorp->al_chunk = &alenlist->al_chunk;
+	cursorp->al_index = 0;
+	cursorp->al_bcount = 0;
+}
+
+
+/*
+ * Create an Address/Length List, and clear it.
+ * Set the cursor to the beginning.
+ */
+alenlist_t 
+alenlist_create(unsigned flags)
+{
+	alenlist_t alenlist;
+
+	alenlist = kmem_zone_alloc(alenlist_zone, flags & AL_NOSLEEP ? VM_NOSLEEP : 0);
+	if (alenlist) {
+		INCR_COUNT(&alenlist_count);
+
+		alenlist->al_flags = 0;
+		alenlist->al_logical_size = 0;
+		alenlist->al_actual_size = ALEN_CHUNK_SZ;
+		alenlist->al_last_chunk = &alenlist->al_chunk;
+		alenlist->al_chunk.alc_next = NULL;
+		do_cursor_init(alenlist, &alenlist->al_cursor);
+	}
+
+	return(alenlist);
+}
+
+
+/*
+ * Grow an Address/Length List so that all resources needed to contain
+ * the specified number of Pairs are pre-allocated.  An Address/Length
+ * List that has been explicitly "grown" will never *automatically*
+ * grow, shrink, or be destroyed.
+ *
+ * Pre-allocation is useful for Real-Time drivers and for drivers that
+ * may be used along the swap-out path and therefore cannot afford to 
+ * sleep until memory is freed.
+ * 
+ * The cursor is set to the beginning of the list.
+ */
+int
+alenlist_grow(alenlist_t alenlist, size_t npairs)
+{
+	/* 
+	 * This interface should be used relatively rarely, so
+	 * the implementation is kept simple: We clear the List,
+	 * then append npairs bogus entries.  Finally, we mark
+	 * the list as FIXED_SIZE and re-initialize the internal
+	 * cursor.
+	 */
+
+	/* 
+	 * Temporarily mark as non-fixed size, since we're about
+	 * to shrink and expand it.
+	 */
+	alenlist->al_flags &= ~AL_FIXED_SIZE;
+
+	/* Free whatever was in the alenlist. */
+	alenlist_clear(alenlist);
+
+	/* Allocate everything that we need via automatic expansion. */
+	while (npairs--)
+		if (alenlist_append(alenlist, 0, 0, AL_NOCOMPACT) == ALENLIST_FAILURE)
+			return(ALENLIST_FAILURE);
+
+	/* Now, mark as FIXED_SIZE */
+	alenlist->al_flags |= AL_FIXED_SIZE;
+
+	/* Clear out bogus entries */
+	alenlist_clear(alenlist);
+
+	/* Initialize internal cursor to the beginning */
+	do_cursor_init(alenlist, &alenlist->al_cursor);
+
+	return(ALENLIST_SUCCESS);
+}
+
+
+/*
+ * Clear an Address/Length List so that it holds no pairs.
+ */
+void
+alenlist_clear(alenlist_t alenlist)
+{
+	alenlist_chunk_t chunk, freechunk;
+
+	/*
+	 * If this List is not FIXED_SIZE, free all the
+	 * extra chunks.
+	 */
+	if (!(alenlist->al_flags & AL_FIXED_SIZE)) {
+		/* First, free any extension alenlist chunks */
+		chunk = alenlist->al_chunk.alc_next;
+		while (chunk) {
+			freechunk = chunk;
+			chunk = chunk->alc_next;
+			kmem_zone_free(alenlist_chunk_zone, freechunk);
+			DECR_COUNT(&alenlist_chunk_count);
+		}
+		alenlist->al_actual_size = ALEN_CHUNK_SZ;
+		alenlist->al_chunk.alc_next = NULL;
+	}
+
+	alenlist->al_logical_size = 0;
+	alenlist->al_last_chunk = &alenlist->al_chunk;
+	do_cursor_init(alenlist, &alenlist->al_cursor);
+}
+
+
+/*
+ * Create and initialize an Address/Length Pair.
+ * This is intended for degenerate lists, consisting of a single 
+ * address/length pair.
+ */
+alenlist_t
+alenpair_init(	alenaddr_t address, 
+		size_t length)
+{
+	alenlist_t alenlist;
+
+	alenlist = alenlist_create(0);
+
+	alenlist->al_logical_size = 1;
+	ASSERT(alenlist->al_last_chunk == &alenlist->al_chunk);
+	alenlist->al_chunk.alc_pair[0].al_length = length;
+	alenlist->al_chunk.alc_pair[0].al_addr = address;
+
+	return(alenlist);
+}
+
+/*
+ * Return address/length from a degenerate (1-pair) List, or
+ * first pair from a larger list.  Does NOT update the internal cursor,
+ * so this is an easy way to peek at a start address.
+ */
+int
+alenpair_get(	alenlist_t alenlist,
+		alenaddr_t *address,
+		size_t *length)
+{
+	if (alenlist->al_logical_size == 0)
+		return(ALENLIST_FAILURE);
+
+	*length = alenlist->al_chunk.alc_pair[0].al_length;
+	*address = alenlist->al_chunk.alc_pair[0].al_addr;
+	return(ALENLIST_SUCCESS);
+}
+
+
+/*
+ * Destroy an Address/Length List.
+ */
+void 
+alenlist_destroy(alenlist_t alenlist)
+{
+	if (alenlist == NULL)
+		return;
+
+	/* 
+	 * Turn off FIXED_SIZE so this List can be 
+	 * automatically shrunk.
+	 */
+	alenlist->al_flags &= ~AL_FIXED_SIZE;
+
+	/* Free extension chunks first */
+	if (alenlist->al_chunk.alc_next)
+		alenlist_clear(alenlist);
+
+	/* Now, free the alenlist itself */
+	kmem_zone_free(alenlist_zone, alenlist);
+	DECR_COUNT(&alenlist_count);
+}
+
+/*
+ * Release an Address/Length List.
+ * This is in preparation for a day when alenlist's may be longer-lived, and
+ * perhaps associated with a buf structure.  We'd add a reference count, and
+ * this routine would decrement the count.  For now, we create alenlist's on
+ * on demand and free them when done.  If the driver is not explicitly managing
+ * a List for its own use, it should call alenlist_done rather than alenlist_destroy.
+ */
+void
+alenlist_done(alenlist_t alenlist)
+{
+	alenlist_destroy(alenlist);
+}
+
+
+/*
+ * Append another address/length to the end of an Address/Length List,
+ * growing the list if permitted and necessary.
+ *
+ * Returns: SUCCESS/FAILURE
+ */
+int 
+alenlist_append(	alenlist_t alenlist, 		/* append to this list */
+			alenaddr_t address, 		/* address to append */
+			size_t length,			/* length to append */
+			unsigned flags)
+{
+	alen_t *alenp;
+	int index, last_index;
+
+	index = alenlist->al_logical_size % ALEN_CHUNK_SZ;
+
+	if ((alenlist->al_logical_size > 0)) {
+		/*
+		 * See if we can compact this new pair in with the previous entry.
+		 * al_compaction_address holds that value that we'd need to see
+		 * in order to compact.
+		 */
+		if (!(flags & AL_NOCOMPACT) &&
+		    (alenlist->al_compaction_address == address)) {
+			last_index = (alenlist->al_logical_size-1) % ALEN_CHUNK_SZ;
+			alenp = &(alenlist->al_last_chunk->alc_pair[last_index]);
+			alenp->al_length += length;
+			alenlist->al_compaction_address += length;
+			return(ALENLIST_SUCCESS);
+		}
+
+		/*
+		 * If we're out of room in this chunk, move to a new chunk.
+	 	 */
+		if (index == 0) {
+			if (alenlist->al_flags & AL_FIXED_SIZE) {
+				alenlist->al_last_chunk = alenlist->al_last_chunk->alc_next;
+
+				/* If we're out of space in a FIXED_SIZE List, quit. */
+				if (alenlist->al_last_chunk == NULL) {
+					ASSERT(alenlist->al_logical_size == alenlist->al_actual_size);
+					return(ALENLIST_FAILURE);
+				}
+			} else {
+				alenlist_chunk_t new_chunk;
+
+				new_chunk = kmem_zone_alloc(alenlist_chunk_zone, 
+							flags & AL_NOSLEEP ? VM_NOSLEEP : 0);
+
+				if (new_chunk == NULL)
+					return(ALENLIST_FAILURE);
+
+				alenlist->al_last_chunk->alc_next = new_chunk;
+				new_chunk->alc_next = NULL;
+				alenlist->al_last_chunk = new_chunk;
+				alenlist->al_actual_size += ALEN_CHUNK_SZ;
+				INCR_COUNT(&alenlist_chunk_count);
+			}
+		}
+	}
+
+	alenp = &(alenlist->al_last_chunk->alc_pair[index]);
+	alenp->al_addr = address;
+	alenp->al_length = length;
+	
+	alenlist->al_logical_size++;
+	alenlist->al_compaction_address = address + length;
+
+	return(ALENLIST_SUCCESS);
+}
+
+
+/*
+ * Replace an item in an Address/Length List.  Cursor is updated so
+ * that alenlist_get will get the next item in the list.  This interface 
+ * is not very useful for drivers; but it is useful to bus providers 
+ * that need to translate between address spaced in situ.  The old Address
+ * and Length are returned.
+ */
+/* ARGSUSED */
+int
+alenlist_replace(	alenlist_t alenlist, 		/* in: replace in this list */
+			alenlist_cursor_t cursorp, 	/* inout: which item to replace */
+			alenaddr_t *addrp, 		/* inout: address */
+			size_t *lengthp,		/* inout: length */
+			unsigned flags)
+{
+	alen_t *alenp;
+	alenlist_chunk_t chunk;
+	unsigned int index;
+	size_t length;
+	alenaddr_t addr;
+
+	if ((addrp == NULL) || (lengthp == NULL))
+		return(ALENLIST_FAILURE);
+
+	if (alenlist->al_logical_size == 0)
+		return(ALENLIST_FAILURE);
+
+	addr = *addrp;
+	length = *lengthp;
+
+	/* 
+	 * If no cursor explicitly specified, use the Address/Length List's 
+	 * internal cursor.
+	 */
+	if (cursorp == NULL)
+		cursorp = &alenlist->al_cursor;
+
+	chunk = cursorp->al_chunk;
+	index = cursorp->al_index;
+
+	ASSERT(cursorp->al_alenlist == alenlist);
+	if (cursorp->al_alenlist != alenlist)
+		return(ALENLIST_FAILURE);
+
+	alenp = &chunk->alc_pair[index];
+
+	/* Return old values */
+	*addrp = alenp->al_length;
+	*lengthp = alenp->al_addr;
+
+	/* Set up new values */
+	alenp->al_length = length;
+	alenp->al_addr = addr;
+
+	/* Update cursor to point to next item */
+	cursorp->al_bcount = length;
+
+	return(ALENLIST_SUCCESS);
+}
+
+
+/*
+ * Initialize a cursor in order to walk an alenlist.
+ * An alenlist_cursor always points to the last thing that was obtained
+ * from the list.  If al_chunk is NULL, then nothing has yet been obtained.
+ *
+ * Note: There is an "internal cursor" associated with every Address/Length List.
+ * For users that scan sequentially through a List, it is more efficient to
+ * simply use the internal cursor.  The caller must insure that no other users
+ * will simultaneously scan the List.  The caller can reposition the internal
+ * cursor by calling alenlist_cursor_init with a NULL cursorp.
+ */
+int
+alenlist_cursor_init(alenlist_t alenlist, size_t offset, alenlist_cursor_t cursorp)
+{
+	size_t byte_count;
+
+	if (cursorp == NULL)
+		cursorp = &alenlist->al_cursor;
+
+	/* Get internal cursor's byte count for use as a hint.
+	 *
+	 * If the internal cursor points passed the point that we're interested in,
+	 * we need to seek forward from the beginning.  Otherwise, we can seek forward
+	 * from the internal cursor.
+	 */
+	if ((offset > 0) &&
+	   ((byte_count = alenlist_cursor_offset(alenlist, (alenlist_cursor_t)NULL)) <= offset)) {
+		offset -= byte_count;
+		alenlist_cursor_clone(alenlist, NULL, cursorp);
+	} else
+		do_cursor_init(alenlist, cursorp);
+
+	/* We could easily speed this up, but it shouldn't be used very often. */
+	while (offset != 0) {
+		alenaddr_t addr;
+		size_t length;
+
+		if (alenlist_get(alenlist, cursorp, offset, &addr, &length, 0) != ALENLIST_SUCCESS)
+			return(ALENLIST_FAILURE);
+		offset -= length;
+	}
+	return(ALENLIST_SUCCESS);
+}
+
+
+/*
+ * Copy a cursor.  The source cursor is either an internal alenlist cursor
+ * or an explicit cursor.
+ */
+int
+alenlist_cursor_clone(	alenlist_t alenlist, 
+			alenlist_cursor_t cursorp_in, 
+			alenlist_cursor_t cursorp_out)
+{
+	ASSERT(cursorp_out);
+
+	if (alenlist && cursorp_in)
+		if (alenlist != cursorp_in->al_alenlist)
+			return(ALENLIST_FAILURE);
+
+	if (alenlist)
+		*cursorp_out = alenlist->al_cursor; /* small structure copy */
+	else if (cursorp_in)
+		*cursorp_out = *cursorp_in; /* small structure copy */
+	else
+		return(ALENLIST_FAILURE); /* no source */
+
+	return(ALENLIST_SUCCESS);
+}
+
+/*
+ * Return the number of bytes passed so far according to the specified cursor.
+ * If cursorp is NULL, use the alenlist's internal cursor.
+ */
+size_t
+alenlist_cursor_offset(alenlist_t alenlist, alenlist_cursor_t cursorp)
+{
+	ASSERT(!alenlist || !cursorp || (alenlist == cursorp->al_alenlist));
+
+	if (cursorp == NULL) {
+		ASSERT(alenlist);
+		cursorp = &alenlist->al_cursor;
+	}
+
+	return(cursorp->al_offset);
+}
+
+/*
+ * Allocate and initialize an Address/Length List cursor.
+ */
+alenlist_cursor_t
+alenlist_cursor_create(alenlist_t alenlist, unsigned flags)
+{
+	alenlist_cursor_t cursorp;
+
+	ASSERT(alenlist != NULL);
+	cursorp = kmem_zone_alloc(alenlist_cursor_zone, flags & AL_NOSLEEP ? VM_NOSLEEP : 0);
+	if (cursorp) {
+		INCR_COUNT(&alenlist_cursor_count);
+		alenlist_cursor_init(alenlist, 0, cursorp);
+	}
+	return(cursorp);
+}
+
+/*
+ * Free an Address/Length List cursor.
+ */
+void
+alenlist_cursor_destroy(alenlist_cursor_t cursorp)
+{
+	DECR_COUNT(&alenlist_cursor_count);
+	kmem_zone_free(alenlist_cursor_zone, cursorp);
+}
+
+
+/*
+ * Fetch an address/length pair from an Address/Length List.  Update
+ * the "cursor" so that next time this routine is called, we'll get
+ * the next address range.  Never return a length that exceeds maxlength
+ * (if non-zero).  If maxlength is a power of 2, never return a length 
+ * that crosses a maxlength boundary.  [This may seem strange at first,
+ * but it's what many drivers want.]
+ *
+ * Returns: SUCCESS/FAILURE
+ */
+int
+alenlist_get(	alenlist_t alenlist, 		/* in: get from this list */
+		alenlist_cursor_t cursorp, 	/* inout: which item to get */
+		size_t	maxlength,		/* in: at most this length */
+		alenaddr_t *addrp, 		/* out: address */
+		size_t *lengthp,		/* out: length */
+		unsigned flags)
+{
+	alen_t *alenp;
+	alenlist_chunk_t chunk;
+	unsigned int index;
+	size_t bcount;
+	size_t length;
+
+	/* 
+	 * If no cursor explicitly specified, use the Address/Length List's 
+	 * internal cursor.
+	 */
+	if (cursorp == NULL) {
+		if (alenlist->al_logical_size == 0)
+			return(ALENLIST_FAILURE);
+		cursorp = &alenlist->al_cursor;
+	}
+
+	chunk = cursorp->al_chunk;
+	index = cursorp->al_index;
+	bcount = cursorp->al_bcount;
+
+	ASSERT(cursorp->al_alenlist == alenlist);
+	if (cursorp->al_alenlist != alenlist)
+		return(ALENLIST_FAILURE);
+
+	alenp = &chunk->alc_pair[index];
+	length = alenp->al_length - bcount;
+
+	/* Bump up to next pair, if we're done with this pair. */
+	if (length == 0) {
+		cursorp->al_bcount = bcount = 0;
+		cursorp->al_index = index = (index + 1) % ALEN_CHUNK_SZ;
+
+		/* Bump up to next chunk, if we're done with this chunk. */
+		if (index == 0) {
+			if (cursorp->al_chunk == alenlist->al_last_chunk)
+				return(ALENLIST_FAILURE);
+			chunk = chunk->alc_next;
+			ASSERT(chunk != NULL);
+		} else {
+			/* If in last chunk, don't go beyond end. */
+			if (cursorp->al_chunk == alenlist->al_last_chunk) {
+				int last_size = alenlist->al_logical_size % ALEN_CHUNK_SZ;
+				if (last_size && (index >= last_size))
+					return(ALENLIST_FAILURE);
+			}
+		}
+
+		alenp = &chunk->alc_pair[index];
+		length = alenp->al_length;
+	}
+
+	/* Constrain what we return according to maxlength */
+	if (maxlength) {
+		size_t maxlen1 = maxlength - 1;
+
+		if ((maxlength & maxlen1) == 0) /* power of 2 */
+			maxlength -= 
+			   ((alenp->al_addr + cursorp->al_bcount) & maxlen1);
+
+		length = MIN(maxlength, length);
+	}
+
+	/* Update the cursor, if desired. */
+	if (!(flags & AL_LEAVE_CURSOR)) {
+		cursorp->al_bcount += length;
+		cursorp->al_chunk = chunk;
+	}
+
+	*lengthp = length;
+	*addrp = alenp->al_addr + bcount;
+
+	return(ALENLIST_SUCCESS);
+}
+
+
+/*
+ * Return the number of pairs in the specified Address/Length List.
+ * (For FIXED_SIZE Lists, this returns the logical size of the List, 
+ * not the actual capacity of the List.)
+ */
+int
+alenlist_size(alenlist_t alenlist)
+{
+	return(alenlist->al_logical_size);
+}
+
+
+/*
+ * Concatenate two Address/Length Lists.
+ */
+void
+alenlist_concat(alenlist_t from,
+		alenlist_t to)
+{
+	struct alenlist_cursor_s cursor;
+	alenaddr_t addr;
+	size_t length;
+
+	alenlist_cursor_init(from, 0, &cursor);
+
+	while(alenlist_get(from, &cursor, (size_t)0, &addr, &length, 0) == ALENLIST_SUCCESS)
+		alenlist_append(to, addr, length, 0);
+}
+
+/*
+ * Create a copy of a list.
+ * (Not all attributes of the old list are cloned.  For instance, if
+ * a FIXED_SIZE list is cloned, the resulting list is NOT FIXED_SIZE.)
+ */
+alenlist_t
+alenlist_clone(alenlist_t old_list, unsigned flags)
+{
+	alenlist_t new_list;
+
+	new_list = alenlist_create(flags);
+	if (new_list != NULL)
+		alenlist_concat(old_list, new_list);
+
+	return(new_list);
+}
+
+
+/* 
+ * Convert a kernel virtual address to a Physical Address/Length List.
+ */
+alenlist_t
+kvaddr_to_alenlist(alenlist_t alenlist, caddr_t kvaddr, size_t length, unsigned flags)
+{
+	alenaddr_t paddr;
+	long offset;
+	size_t piece_length;
+	int created_alenlist;
+
+	if (length <=0)
+		return(NULL);
+
+	/* If caller supplied a List, use it.  Otherwise, allocate one. */
+	if (alenlist == NULL) {
+		alenlist = alenlist_create(0);
+		created_alenlist = 1;
+	} else {
+		alenlist_clear(alenlist);
+		created_alenlist = 0;
+	}
+
+	paddr = kvtophys(kvaddr);
+	offset = poff(kvaddr);
+
+	/* Handle first page */
+	piece_length = MIN(NBPP - offset, length);
+	if (alenlist_append(alenlist, paddr, piece_length, flags) == ALENLIST_FAILURE)
+		goto failure;
+	length -= piece_length;
+	kvaddr += piece_length;
+
+	/* Handle middle pages */
+	while (length >= NBPP) {
+		paddr = kvtophys(kvaddr);
+		if (alenlist_append(alenlist, paddr, NBPP, flags) == ALENLIST_FAILURE)
+			goto failure;
+		length -= NBPP;
+		kvaddr += NBPP;
+	}
+
+	/* Handle last page */
+	if (length) {
+		ASSERT(length < NBPP);
+		paddr = kvtophys(kvaddr);
+		if (alenlist_append(alenlist, paddr, length, flags) == ALENLIST_FAILURE)
+			goto failure;
+	}
+
+	alenlist_cursor_init(alenlist, 0, NULL);
+	return(alenlist);
+
+failure:
+	if (created_alenlist)
+		alenlist_destroy(alenlist);
+	return(NULL);
+}
+
+
+#if DEBUG
+static void
+alenlist_show(alenlist_t alenlist)
+{
+	struct alenlist_cursor_s cursor;
+	alenaddr_t addr;
+	size_t length;
+	int i = 0;
+
+	alenlist_cursor_init(alenlist, 0, &cursor);
+
+	qprintf("Address/Length List@0x%x:\n", alenlist);
+	qprintf("logical size=0x%x actual size=0x%x last_chunk at 0x%x\n", 
+		alenlist->al_logical_size, alenlist->al_actual_size, 
+		alenlist->al_last_chunk);
+	qprintf("cursor: chunk=0x%x index=%d offset=0x%x\n",
+		alenlist->al_cursor.al_chunk, 
+		alenlist->al_cursor.al_index,
+		alenlist->al_cursor.al_bcount);
+	while(alenlist_get(alenlist, &cursor, (size_t)0, &addr, &length, 0) == ALENLIST_SUCCESS)
+		qprintf("%d:\t0x%lx 0x%lx\n", ++i, addr, length);
+}
+#endif /* DEBUG */
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/cdl.c linux/arch/ia64/sn/io/cdl.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/cdl.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/cdl.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,231 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+#include <linux/types.h>
+#include <linux/config.h>
+#include <asm/sn/sgi.h>
+#include <asm/io.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/pci/bridge.h>
+#include "asm/sn/ioerror_handling.h"
+#include <asm/sn/xtalk/xbow.h>
+
+#ifdef BRINGUP
+/* these get called directly in cdl_add_connpt in fops bypass hack */
+extern int pcibr_attach(devfs_handle_t);
+extern int xbow_attach(devfs_handle_t);
+#endif /* BRINGUP */
+
+/*
+ *    cdl: Connection and Driver List
+ *
+ *	We are not porting this to Linux.  Devices are registered via 
+ *	the normal Linux PCI layer.  This is a very simplified version 
+ *	of cdl that will allow us to register and call our very own 
+ *	IO Infrastructure Drivers e.g. pcibr.
+ */
+
+struct cdl {
+    int		part_num;
+    int		mfg_num;
+    int (*attach) (devfs_handle_t);
+} dummy_reg;
+
+typedef struct cdl     *cdl_p;
+
+#define MAX_SGI_IO_INFRA_DRVR 4
+struct cdl sgi_infrastructure_drivers[MAX_SGI_IO_INFRA_DRVR] =
+{
+	{ XBRIDGE_WIDGET_PART_NUM, XBRIDGE_WIDGET_MFGR_NUM, pcibr_attach /* &pcibr_fops  */},
+	{ BRIDGE_WIDGET_PART_NUM,  BRIDGE_WIDGET_MFGR_NUM,  pcibr_attach /* &pcibr_fops */},
+	{ XXBOW_WIDGET_PART_NUM,   XXBOW_WIDGET_MFGR_NUM,   xbow_attach /* &xbow_fops */},
+	{ XBOW_WIDGET_PART_NUM,    XBOW_WIDGET_MFGR_NUM,    xbow_attach /* &xbow_fops */},
+};
+
+/*
+ * cdl_new:  Called by pciio and xtalk.
+ */
+cdl_p
+cdl_new(char *name, char *k1str, char *k2str)
+{
+    /*
+     * Just return a dummy pointer.
+     */
+    return((cdl_p)&dummy_reg);
+}
+
+/*
+ * cdl_del: Do nothing.
+ */
+void
+cdl_del(cdl_p reg)
+{
+	printk("SGI IO INFRASTRUCTURE - cdl_del not supported.\n");
+}
+
+/*
+ * cdl_add_driver: The driver part number and manufacturers number 
+ * are statically initialized above.
+ * 
+  Do nothing.
+ */
+int
+cdl_add_driver(cdl_p reg, int key1, int key2, char *prefix, int flags)
+{
+    return 0;
+}
+
+/*
+ * cdl_del_driver: Not supported.
+ */
+void
+cdl_del_driver(cdl_p reg,
+	       char *prefix)
+{
+
+	printk("SGI IO INFRASTRUCTURE - cdl_del_driver not supported.\n");
+}
+
+/*
+ * cdl_add_connpt: We found a device and it's connect point.  Call the 
+ * attach routine of that driver.
+ *
+ * May need support for pciba registration here ...
+ *
+ * This routine use to create /hw/.id/pci/.../.. that links to 
+ * /hw/module/006c06/Pbrick/xtalk/15/pci/<slotnum> .. do we still need 
+ * it?  The specified driver attach routine does not reference these 
+ * vertices.
+ */
+int
+cdl_add_connpt(cdl_p reg, int part_num, int mfg_num, 
+	       devfs_handle_t connpt)
+{
+	int i;
+	
+	/*
+	 * Find the driver entry point and call the attach routine.
+	 */
+	for (i = 0; i < MAX_SGI_IO_INFRA_DRVR; i++) {
+
+		if ( (part_num == sgi_infrastructure_drivers[i].part_num) &&
+		   ( mfg_num == sgi_infrastructure_drivers[i].mfg_num) ) {
+			/*
+			 * Call the device attach routines.
+			 */
+			if (sgi_infrastructure_drivers[i].attach) {
+			    return(sgi_infrastructure_drivers[i].attach(connpt));
+			}
+#ifdef BRINGUP
+			/*
+			 * XXX HACK ALERT bypassing fops for now..
+			 */
+			else {
+			    printk("cdl_add_connpt: NEED FOPS FOR OUR DRIVERS!!\n");
+			    printk("cdl_add_connpt: part_num= 0x%x  mfg_num= 0x%x\n",
+				part_num, mfg_num);
+			    return(-1);
+			}
+#endif /* BRINGUP */
+		} else {
+			continue;
+		}
+
+		printk("**** cdl_add_connpt: driver not found for part_num %d mfg_num %d ****\n", part_num, mfg_num);
+
+		return(-1);
+	}	
+	if ( (i == MAX_SGI_IO_INFRA_DRVR) ) 
+		printk("**** cdl_add_connpt: Driver not found for part_num 0x%x mfg_num 0x%x ****\n", part_num, mfg_num);
+
+	return (0);
+}
+
+/*
+ * cdl_del_connpt: Not implemented.
+ */
+void
+cdl_del_connpt(cdl_p reg, int key1, int key2, devfs_handle_t connpt)
+{
+
+	printk("SGI IO INFRASTRUCTURE - cdl_del_cdl_del_connpt not supported.\n");
+}
+
+/*
+ *    cdl_iterate: Not Implemented.
+ */
+void
+cdl_iterate(cdl_p reg,
+	    char *prefix,
+	    cdl_iter_f * func)
+{
+
+	printk("SGI IO INFRASTRUCTURE - cdl_iterate not supported.\n");
+}
+
+async_attach_t 
+async_attach_new(void)
+{
+
+	printk("SGI IO INFRASTRUCTURE - async_attach_new not supported.\n");
+	return(0);
+}
+
+void 
+async_attach_free(async_attach_t aa)
+{
+	printk("SGI IO INFRASTRUCTURE - async_attach_free not supported.\n");
+}
+
+async_attach_t 
+async_attach_get_info(devfs_handle_t vhdl)
+{
+
+	printk("SGI IO INFRASTRUCTURE - async_attach_get_info not supported.\n");
+	return(0);
+}
+
+void            
+async_attach_add_info(devfs_handle_t vhdl, async_attach_t aa)
+{
+	printk("SGI IO INFRASTRUCTURE - async_attach_add_info not supported.\n");
+
+}
+
+void            
+async_attach_del_info(devfs_handle_t vhdl)
+{
+
+	printk("SGI IO INFRASTRUCTURE - async_attach_del_info not supported.\n");
+
+}
+
+void async_attach_signal_start(async_attach_t aa)
+{
+
+	printk("SGI IO INFRASTRUCTURE - async_attach_signal_start not supported.\n");
+
+}
+
+void async_attach_signal_done(async_attach_t aa)
+{
+
+	printk("SGI IO INFRASTRUCTURE - async_attach_signal_done not supported.\n");
+
+}
+
+void async_attach_waitall(async_attach_t aa)
+{
+
+	printk("SGI IO INFRASTRUCTURE - async_attach_waitall not supported.\n");
+
+}
+
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/devsupport.c linux/arch/ia64/sn/io/devsupport.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/devsupport.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/devsupport.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,1292 @@
+#define ilvt_t int
+
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/slab.h>
+#include <linux/config.h>
+#include <asm/sn/sgi.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/iobus.h>
+#include <asm/sn/iograph.h>
+
+/* 
+ * Interfaces in this file are all platform-independent AND IObus-independent.
+ * Be aware that there may be macro equivalents to each of these hiding in
+ * header files which supercede these functions.
+ */
+
+/* =====Generic iobus support===== */
+
+/* String table to hold names of interrupts. */
+#ifdef notyet
+static struct string_table device_desc_string_table;
+#endif
+
+/* One time initialization for device descriptor support. */
+static void
+device_desc_init(void)
+{
+#ifdef notyet
+	string_table_init(&device_desc_string_table);
+#endif
+	FIXME("device_desc_init");
+}
+
+
+/* Drivers use these interfaces to manage device descriptors */
+static device_desc_t
+device_desc_alloc(void)
+{
+#ifdef notyet
+	device_desc_t device_desc;
+
+	device_desc = (device_desc_t)kmem_zalloc(sizeof(struct device_desc_s), 0);
+	device_desc->intr_target = GRAPH_VERTEX_NONE;
+
+	ASSERT(device_desc->intr_policy == 0);
+	device_desc->intr_swlevel = -1;
+	ASSERT(device_desc->intr_name == NULL);
+	ASSERT(device_desc->flags == 0);
+
+	ASSERT(!(device_desc->flags & D_IS_ASSOC));
+	return(device_desc);
+#else
+	FIXME("device_desc_alloc");
+	return((device_desc_t)0);
+#endif
+}
+
+void
+device_desc_free(device_desc_t device_desc)
+{
+#ifdef notyet
+	if (!(device_desc->flags & D_IS_ASSOC)) /* sanity */
+		kfree(device_desc);
+#endif
+	FIXME("device_desc_free");
+}
+
+device_desc_t
+device_desc_dup(devfs_handle_t dev)
+{
+#ifdef notyet
+	device_desc_t orig_device_desc, new_device_desc;
+
+
+	new_device_desc = device_desc_alloc();
+	orig_device_desc = device_desc_default_get(dev);
+	if (orig_device_desc)
+		*new_device_desc = *orig_device_desc;/* small structure copy */
+	else {
+		device_driver_t		driver;
+		ilvl_t			pri;		
+		/* 
+		 * Use the driver's thread priority in 
+		 * case the device thread priority has not
+		 * been given.
+		 */
+		if (driver = device_driver_getbydev(dev)) {
+			pri = device_driver_thread_pri_get(driver);
+			device_desc_intr_swlevel_set(new_device_desc,pri);
+		}
+	}		
+	new_device_desc->flags &= ~D_IS_ASSOC;
+	return(new_device_desc);
+#else
+	FIXME("device_desc_dup");
+	return((device_desc_t)0);
+#endif
+}
+
+device_desc_t	
+device_desc_default_get(devfs_handle_t dev)
+{
+#ifdef notyet
+	graph_error_t rc;
+	device_desc_t device_desc;
+
+	rc = hwgraph_info_get_LBL(dev, INFO_LBL_DEVICE_DESC, (arbitrary_info_t *)&device_desc);
+
+	if (rc == GRAPH_SUCCESS)
+		return(device_desc);
+	else
+		return(NULL);
+#else
+	FIXME("device_desc_default_get");
+	return((device_desc_t)0);
+#endif
+}
+
+void		
+device_desc_default_set(devfs_handle_t dev, device_desc_t new_device_desc)
+{
+#ifdef notyet
+	graph_error_t rc;
+	device_desc_t old_device_desc = NULL;
+
+	if (new_device_desc) {
+		new_device_desc->flags |= D_IS_ASSOC;
+		rc = hwgraph_info_add_LBL(dev, INFO_LBL_DEVICE_DESC, 
+						(arbitrary_info_t)new_device_desc);
+		if (rc == GRAPH_DUP) {
+			rc = hwgraph_info_replace_LBL(dev, INFO_LBL_DEVICE_DESC, 
+				(arbitrary_info_t)new_device_desc, 
+				(arbitrary_info_t *)&old_device_desc);
+
+			ASSERT(rc == GRAPH_SUCCESS);
+		}
+		hwgraph_info_export_LBL(dev, INFO_LBL_DEVICE_DESC,
+					sizeof(struct device_desc_s));
+	} else {
+		rc = hwgraph_info_remove_LBL(dev, INFO_LBL_DEVICE_DESC,
+					(arbitrary_info_t *)&old_device_desc);
+	}
+
+	if (old_device_desc) {
+		ASSERT(old_device_desc->flags & D_IS_ASSOC);
+		old_device_desc->flags &= ~D_IS_ASSOC;
+		device_desc_free(old_device_desc);
+	}
+#endif
+	FIXME("device_desc_default_set");
+}
+
+devfs_handle_t
+device_desc_intr_target_get(device_desc_t device_desc)
+{
+#ifdef notyet
+	return(device_desc->intr_target);
+#else
+	FIXME("device_desc_intr_target_get");
+	return((devfs_handle_t)0);
+#endif
+}
+
+int
+device_desc_intr_policy_get(device_desc_t device_desc)
+{
+#ifdef notyet
+	return(device_desc->intr_policy);
+#else
+	FIXME("device_desc_intr_policy_get");
+	return(0);
+#endif
+}
+
+ilvl_t
+device_desc_intr_swlevel_get(device_desc_t device_desc)
+{
+#ifdef notyet
+	return(device_desc->intr_swlevel);
+#else
+	FIXME("device_desc_intr_swlevel_get");
+	return((ilvl_t)0);
+#endif
+}
+
+char *
+device_desc_intr_name_get(device_desc_t device_desc)
+{
+#ifdef notyet
+	return(device_desc->intr_name);
+#else
+	FIXME("device_desc_intr_name_get");
+	return(NULL);
+#endif
+}
+
+int
+device_desc_flags_get(device_desc_t device_desc)
+{
+#ifdef notyet
+	return(device_desc->flags);
+#else
+	FIXME("device_desc_flags_get");
+	return(0);
+#endif
+}
+
+void
+device_desc_intr_target_set(device_desc_t device_desc, devfs_handle_t target)
+{
+	if ( device_desc != (device_desc_t)0 )
+		device_desc->intr_target = target;
+}
+
+void
+device_desc_intr_policy_set(device_desc_t device_desc, int policy)
+{
+	if ( device_desc != (device_desc_t)0 )
+		device_desc->intr_policy = policy;
+}
+
+void
+device_desc_intr_swlevel_set(device_desc_t device_desc, ilvl_t swlevel)
+{
+	if ( device_desc != (device_desc_t)0 )
+		device_desc->intr_swlevel = swlevel;
+}
+
+void
+device_desc_intr_name_set(device_desc_t device_desc, char *name)
+{
+#ifdef notyet
+	if ( device_desc != (device_desc_t)0 )
+		device_desc->intr_name = string_table_insert(&device_desc_string_table, name);
+#else
+	FIXME("device_desc_intr_name_set");
+#endif
+}
+
+void
+device_desc_flags_set(device_desc_t device_desc, int flags)
+{
+	if ( device_desc != (device_desc_t)0 )
+		device_desc->flags = flags;
+}
+
+
+
+/*============= device admin registry routines ===================== */
+
+/* Linked list of <admin-name,admin-val> pairs */
+typedef struct dev_admin_list_s {
+	struct dev_admin_list_s		*admin_next; 	/* next entry in the
+							 * list 
+							 */
+	char				*admin_name;	/* info label */
+	char				*admin_val;	/* actual info */
+} dev_admin_list_t;
+
+/* Device/Driver administration registry */
+typedef struct dev_admin_registry_s {
+	mrlock_t			reg_lock;	/* To allow
+							 * exclusive
+							 * access
+							 */
+	dev_admin_list_t		*reg_first;	/* first entry in 
+							 * the list
+							 */
+	dev_admin_list_t		**reg_last;	/* pointer to the
+							 * next to last entry
+							 * in the last which 
+							 * is also the place
+							 * where the new
+							 * entry gets
+							 * inserted
+							 */
+} dev_admin_registry_t;
+
+/*
+** device_driver_s associates a device driver prefix with device switch entries.
+*/
+struct device_driver_s {
+	struct device_driver_s	*dd_next;	/* next element on hash chain */
+	struct device_driver_s	*dd_prev;	/* previous element on hash chain */
+	char			*dd_prefix;	/* driver prefix string */
+	struct bdevsw		*dd_bdevsw;	/* driver's bdevsw */
+	struct cdevsw		*dd_cdevsw;	/* driver's cdevsw */
+	
+	/* driver administration specific data structures need to
+	 * maintain the list of <driver-paramater,value> pairs
+	 */
+	dev_admin_registry_t	dd_dev_admin_registry;
+	ilvl_t			dd_thread_pri;	/* default thread priority for
+						 *  all this driver's
+						 * threads.
+						 */
+
+};
+
+#define	NEW(_p)		(_p = kmalloc(sizeof(*_p), GFP_KERNEL))
+#define FREE(_p)	(kmem_free(_p))
+	
+/*
+ * helpful lock macros
+ */
+
+#define DEV_ADMIN_REGISTRY_INITLOCK(lockp,name)	mrinit(lockp,name)
+#define DEV_ADMIN_REGISTRY_RDLOCK(lockp)	mraccess(lockp)	       
+#define DEV_ADMIN_REGISTRY_WRLOCK(lockp)	mrupdate(lockp)	       
+#define DEV_ADMIN_REGISTRY_UNLOCK(lockp)	mrunlock(lockp)
+
+/* Initialize the registry 
+ */
+static void
+dev_admin_registry_init(dev_admin_registry_t *registry)
+{
+#ifdef notyet
+	if ( registry != (dev_admin_registry_t *)0 )
+		DEV_ADMIN_REGISTRY_INITLOCK(&registry->reg_lock,
+				    "dev_admin_registry_lock");
+		registry->reg_first = NULL;
+		registry->reg_last = &registry->reg_first;
+	}
+#else
+	FIXME("dev_admin_registry_init");
+#endif
+}
+
+/*
+ * add an <name , value > entry to the dev admin registry.
+ * if the name already exists in the registry then change the
+ * value iff the new value differs from the old value.
+ * if the name doesn't exist a new list entry is created and put
+ * at the end.
+ */
+static void
+dev_admin_registry_add(dev_admin_registry_t	*registry,
+		       char			*name,
+		       char			*val)
+{
+#ifdef notyet
+	dev_admin_list_t	*reg_entry;
+	dev_admin_list_t	*scan = 0;
+
+	DEV_ADMIN_REGISTRY_WRLOCK(&registry->reg_lock);
+
+	/* check if the name already exists in the registry */
+	scan = registry->reg_first;
+
+	while (scan) {
+		if (strcmp(scan->admin_name,name) == 0) {
+			/* name is there in the registry */
+			if (strcmp(scan->admin_val,val)) {
+				/* old value != new value 
+				 * reallocate  memory and copy the new value
+				 */
+				FREE(scan->admin_val);
+				scan->admin_val = 
+					(char *)kern_calloc(1,strlen(val)+1);
+				strcpy(scan->admin_val,val);
+				goto out;
+			}
+			goto out;	/* old value == new value */
+		}
+		scan = scan->admin_next;
+	}
+
+	/* name is not there in the registry.
+	 * allocate memory for the new registry entry 
+	 */
+	NEW(reg_entry);
+	
+	reg_entry->admin_next   	= 0;
+	reg_entry->admin_name	= (char *)kern_calloc(1,strlen(name)+1);
+	strcpy(reg_entry->admin_name,name);
+	reg_entry->admin_val	= (char *)kern_calloc(1,strlen(val)+1);
+	strcpy(reg_entry->admin_val,val);
+
+	/* add the entry at the end of the registry */
+
+	*(registry->reg_last)	= reg_entry;
+	registry->reg_last	= &reg_entry->admin_next;
+
+out:	DEV_ADMIN_REGISTRY_UNLOCK(&registry->reg_lock);
+#endif
+	FIXME("dev_admin_registry_add");
+}
+/*
+ * check if there is an info corr. to a particular
+ * name starting from the cursor position in the 
+ * registry
+ */
+static char *
+dev_admin_registry_find(dev_admin_registry_t *registry,char *name)
+{
+#ifdef notyet
+	dev_admin_list_t	*scan = 0;
+	
+	DEV_ADMIN_REGISTRY_RDLOCK(&registry->reg_lock);
+	scan = registry->reg_first;
+
+	while (scan) {
+		if (strcmp(scan->admin_name,name) == 0) {
+			DEV_ADMIN_REGISTRY_UNLOCK(&registry->reg_lock);
+			return scan->admin_val;
+		}
+		scan = scan->admin_next;
+	}
+	DEV_ADMIN_REGISTRY_UNLOCK(&registry->reg_lock);
+	return 0;
+#else
+	FIXME("dev_admin_registry_find");
+	return(NULL);
+#endif
+}
+/*============= MAIN DEVICE/ DRIVER ADMINISTRATION INTERFACE================ */
+/*
+ * return any labelled info associated with a device.
+ * called by any kernel code including device drivers.
+ */
+char *
+device_admin_info_get(devfs_handle_t	dev_vhdl,
+		      char		*info_lbl)
+{
+#ifdef notyet
+	char		*info = 0;
+
+	/* return value need not be GRAPH_SUCCESS as the labelled
+	 * info may not be present
+	 */
+	(void)hwgraph_info_get_LBL(dev_vhdl,info_lbl,
+				   (arbitrary_info_t *)&info);
+
+	
+	return info;
+#else
+	FIXME("device_admin_info_get");
+	return(NULL);
+#endif
+}
+
+/*
+ * set labelled info associated with a device.
+ * called by hwgraph infrastructure . may also be called
+ * by device drivers etc.
+ */
+int
+device_admin_info_set(devfs_handle_t	dev_vhdl,
+		      char		*dev_info_lbl,
+		      char		*dev_info_val)
+{
+#ifdef notyet
+	graph_error_t		rv;
+	arbitrary_info_t	old_info;
+
+	/* Handle the labelled info
+	 *		intr_target
+	 *		sw_level 
+	 * in a special way. These are part of device_desc_t
+	 * Right now this is the only case where we have 
+	 * a set of related device_admin attributes which 
+	 * are grouped together.
+	 * In case there is a need for another set we need to
+	 * take a more generic approach to solving this.
+	 * Basically a registry should be implemented. This
+	 * registry is initialized with the callbacks for the
+	 * attributes which need to handled in a special way
+	 * For example:
+	 * Consider
+	 * 		device_desc
+	 *			intr_target
+	 *			intr_swlevel
+	 * register "do_intr_target" for intr_target
+	 * register "do_intr_swlevel" for intr_swlevel.
+	 * When the device_admin interface layer gets an <attr,val> pair
+	 * it looks in the registry to see if there is a function registered to
+	 * handle "attr. If not follow the default path of setting the <attr,val>
+	 * as labelled information hanging off the vertex.
+	 * In the above example:
+	 * "do_intr_target" does what is being done below for the ADMIN_LBL_INTR_TARGET
+	 * case
+	 */		
+	if (!strcmp(dev_info_lbl,ADMIN_LBL_INTR_TARGET) ||
+	    !strcmp(dev_info_lbl,ADMIN_LBL_INTR_SWLEVEL)) {
+
+		device_desc_t	device_desc;
+		
+		/* Check if there is a default device descriptor
+		 * information for this vertex. If not dup one .
+		 */
+		if (!(device_desc = device_desc_default_get(dev_vhdl))) {
+			device_desc = device_desc_dup(dev_vhdl);
+			device_desc_default_set(dev_vhdl,device_desc);
+
+		}
+		if (!strcmp(dev_info_lbl,ADMIN_LBL_INTR_TARGET)) {
+			/* Check if a target cpu has been specified
+			 * for this device by a device administration
+			 * directive
+			 */
+#ifdef DEBUG	
+			printf(ADMIN_LBL_INTR_TARGET
+			       " dev = 0x%x "
+			       "dev_admin_info = %s"
+			       " target = 0x%x\n",
+			       dev_vhdl,
+			       dev_info_lbl,
+			       hwgraph_path_to_vertex(dev_info_val));
+#endif	
+
+			device_desc->intr_target = 
+				hwgraph_path_to_vertex(dev_info_val);
+		} else if (!strcmp(dev_info_lbl,ADMIN_LBL_INTR_SWLEVEL)) {
+			/* Check if the ithread priority level  has been 
+			 * specified for this device by a device administration
+			 * directive
+			 */
+#ifdef DEBUG	
+			printf(ADMIN_LBL_INTR_SWLEVEL
+			       " dev = 0x%x "
+			       "dev_admin_info = %s"
+			       " sw level = 0x%x\n",
+			       dev_vhdl,
+			       dev_info_lbl,
+			       atoi(dev_info_val));
+#endif	
+			device_desc->intr_swlevel = atoi(dev_info_val);
+		}
+
+	}
+	if (!dev_info_val)
+		rv = hwgraph_info_remove_LBL(dev_vhdl,
+					     dev_info_lbl,
+					     &old_info);
+	else {
+
+		rv = hwgraph_info_add_LBL(dev_vhdl,
+					  dev_info_lbl,
+					  (arbitrary_info_t)dev_info_val);
+	
+		if (rv == GRAPH_DUP)  {
+			rv = hwgraph_info_replace_LBL(dev_vhdl,
+					      dev_info_lbl,
+					      (arbitrary_info_t)dev_info_val,
+					      &old_info);
+		}
+	}
+	ASSERT(rv == GRAPH_SUCCESS);
+#endif
+	FIXME("device_admin_info_set");
+	return 0;
+}
+
+/*
+ * return labelled info associated with a device driver
+ * called by kernel code including device drivers
+ */
+char *
+device_driver_admin_info_get(char		*driver_prefix,
+			     char		*driver_info_lbl)
+{
+#ifdef notyet
+	device_driver_t driver;
+
+	driver = device_driver_get(driver_prefix);
+	return (dev_admin_registry_find(&driver->dd_dev_admin_registry,
+					driver_info_lbl));
+#else
+	FIXME("device_driver_admin_info_get");
+	return(NULL);
+#endif
+}
+
+/*
+ * set labelled info associated with a device driver.
+ * called by hwgraph infrastructure . may also be called
+ * from drivers etc.
+ */
+int
+device_driver_admin_info_set(char		*driver_prefix,
+			     char		*driver_info_lbl,
+			     char		*driver_info_val)
+{
+#ifdef notyet
+	device_driver_t driver;
+
+	driver = device_driver_get(driver_prefix);
+	dev_admin_registry_add(&driver->dd_dev_admin_registry,	
+			       driver_info_lbl,
+			       driver_info_val);
+#endif
+	FIXME("device_driver_admin_info_set");
+	return 0;
+}
+/*================== device / driver  admin support routines================*/
+
+/* static tables created by lboot */
+extern dev_admin_info_t	dev_admin_table[];
+extern dev_admin_info_t	drv_admin_table[];
+extern int		dev_admin_table_size;
+extern int		drv_admin_table_size;
+
+/* Extend the device admin table to allow the kernel startup code to 
+ * provide some device specific administrative hints
+ */
+#define ADMIN_TABLE_CHUNK	100
+static dev_admin_info_t extended_dev_admin_table[ADMIN_TABLE_CHUNK];	
+static int		extended_dev_admin_table_size = 0;
+static mrlock_t		extended_dev_admin_table_lock;
+
+/* Initialize the extended device admin table */
+void
+device_admin_table_init(void)
+{
+#ifdef notyet
+	extended_dev_admin_table_size = 0;
+	mrinit(&extended_dev_admin_table_lock,
+	       "extended_dev_admin_table_lock");
+#endif
+	FIXME("device_admin_table_init");
+}
+/* Add <device-name , parameter-name , parameter-value> triple to
+ * the extended device administration info table. This is helpful
+ * for kernel startup code to put some hints before the hwgraph
+ * is setup 
+ */
+void
+device_admin_table_update(char *name,char *label,char *value)
+{
+#ifdef notyet
+	dev_admin_info_t	*p;
+
+	mrupdate(&extended_dev_admin_table_lock);
+
+	/* Safety check that we haven't exceeded array limits */
+	ASSERT(extended_dev_admin_table_size < ADMIN_TABLE_CHUNK);
+
+	if (extended_dev_admin_table_size == ADMIN_TABLE_CHUNK)
+		goto out;
+	
+	/* Get the pointer to the entry in the table where we are
+	 * going to put the new information 
+	 */
+	p = &extended_dev_admin_table[extended_dev_admin_table_size++];
+
+	/* Allocate memory for the strings and copy them in */
+	p->dai_name = (char *)kern_calloc(1,strlen(name)+1);
+	strcpy(p->dai_name,name);
+	p->dai_param_name = (char *)kern_calloc(1,strlen(label)+1);
+	strcpy(p->dai_param_name,label);
+	p->dai_param_val = (char *)kern_calloc(1,strlen(value)+1);
+	strcpy(p->dai_param_val,value);
+
+out:	mrunlock(&extended_dev_admin_table_lock);
+#endif
+	FIXME("device_admin_table_update");
+}
+/* Extend the device driver  admin table to allow the kernel startup code to 
+ * provide some device driver specific administrative hints
+ */
+
+static dev_admin_info_t extended_drv_admin_table[ADMIN_TABLE_CHUNK];	
+static int		extended_drv_admin_table_size = 0;
+mrlock_t		extended_drv_admin_table_lock;
+
+/* Initialize the extended device driver admin table */
+void
+device_driver_admin_table_init(void)
+{
+#ifdef notyet
+	extended_drv_admin_table_size = 0;
+	mrinit(&extended_drv_admin_table_lock,
+	       "extended_drv_admin_table_lock");
+#endif
+	FIXME("device_driver_admin_table_init");
+}
+/* Add <device-driver prefix , parameter-name , parameter-value> triple to
+ * the extended device administration info table. This is helpful
+ * for kernel startup code to put some hints before the hwgraph
+ * is setup 
+ */
+void
+device_driver_admin_table_update(char *name,char *label,char *value)
+{
+#ifdef notyet
+	dev_admin_info_t	*p;
+
+	mrupdate(&extended_dev_admin_table_lock);
+
+	/* Safety check that we haven't exceeded array limits */
+	ASSERT(extended_drv_admin_table_size < ADMIN_TABLE_CHUNK);
+
+	if (extended_drv_admin_table_size == ADMIN_TABLE_CHUNK)
+		goto out;
+	
+	/* Get the pointer to the entry in the table where we are
+	 * going to put the new information 
+	 */
+	p = &extended_drv_admin_table[extended_drv_admin_table_size++];
+
+	/* Allocate memory for the strings and copy them in */
+	p->dai_name = (char *)kern_calloc(1,strlen(name)+1);
+	strcpy(p->dai_name,name);
+	p->dai_param_name = (char *)kern_calloc(1,strlen(label)+1);
+	strcpy(p->dai_param_name,label);
+	p->dai_param_val = (char *)kern_calloc(1,strlen(value)+1);
+	strcpy(p->dai_param_val,value);
+
+out:	mrunlock(&extended_drv_admin_table_lock);
+#endif
+	FIXME("device_driver_admin_table_update");
+}
+/*
+ * keeps on adding the labelled info for each new (lbl,value) pair
+ * that it finds in the static dev admin table (  created by lboot)
+ * and the extended dev admin table ( created if at all by the kernel startup
+ *  code) corresponding to a device in the hardware graph.
+ */
+void
+device_admin_info_update(devfs_handle_t	dev_vhdl)
+{
+#ifdef notyet
+	int			i = 0;
+	dev_admin_info_t	*scan;
+	devfs_handle_t		scan_vhdl;
+	
+	/* Check the static device administration info table */
+	scan = dev_admin_table;
+	while (i < dev_admin_table_size) {
+		
+		scan_vhdl = hwgraph_path_to_dev(scan->dai_name);
+		if (scan_vhdl == dev_vhdl) {
+			device_admin_info_set(dev_vhdl,
+					      scan->dai_param_name,
+					      scan->dai_param_val);
+		}
+		if (scan_vhdl != NODEV)
+			hwgraph_vertex_unref(scan_vhdl);
+		scan++;i++;
+
+	}
+	i = 0;
+	/* Check the extended device administration info table */
+	scan = extended_dev_admin_table;
+	while (i < extended_dev_admin_table_size) {
+		scan_vhdl = hwgraph_path_to_dev(scan->dai_name);
+		if (scan_vhdl == dev_vhdl) {
+			device_admin_info_set(dev_vhdl,
+					      scan->dai_param_name,
+					      scan->dai_param_val);
+		}
+		if (scan_vhdl != NODEV)
+			hwgraph_vertex_unref(scan_vhdl);
+		scan++;i++;
+
+	}
+
+
+#endif
+	FIXME("device_admin_info_update");
+}
+
+/* looks up the static drv admin table ( created by the lboot) and the extended
+ * drv admin table (created if at all by the kernel startup code) 
+ * for this driver specific administration info and adds it to the admin info 
+ * associated with this device driver's object
+ */
+void
+device_driver_admin_info_update(device_driver_t	driver)
+{
+#ifdef notyet
+	int			i = 0;
+	dev_admin_info_t	*scan;
+
+	/* Check the static device driver administration info table */
+	scan = drv_admin_table;
+	while (i < drv_admin_table_size) {
+
+		if (strcmp(scan->dai_name,driver->dd_prefix) == 0) {
+			dev_admin_registry_add(&driver->dd_dev_admin_registry,
+						scan->dai_param_name,
+					 	scan->dai_param_val);
+		}
+		scan++;i++;
+	}
+	i = 0;
+	/* Check the extended device driver administration info table */
+	scan = extended_drv_admin_table;
+	while (i < extended_drv_admin_table_size) {
+
+		if (strcmp(scan->dai_name,driver->dd_prefix) == 0) {
+			dev_admin_registry_add(&driver->dd_dev_admin_registry,
+						scan->dai_param_name,
+					 	scan->dai_param_val);
+		}
+		scan++;i++;
+	}
+#endif
+	FIXME("device_driver_admin_info_update");
+}
+
+/* =====Device Driver Support===== */
+
+
+
+/*
+** Generic device driver support routines for use by kernel modules that
+** deal with device drivers (but NOT for use by the drivers themselves).
+** EVERY registered driver currently in the system -- static or loadable --
+** has an entry in the device_driver_hash table.  A pointer to such an entry
+** serves as a generic device driver handle.
+*/
+
+#define DEVICE_DRIVER_HASH_SIZE 32
+#ifdef notyet
+lock_t device_driver_lock[DEVICE_DRIVER_HASH_SIZE];
+device_driver_t device_driver_hash[DEVICE_DRIVER_HASH_SIZE];
+static struct string_table driver_prefix_string_table;
+#endif
+
+/*
+** Initialize device driver infrastructure.
+*/
+void
+device_driver_init(void)
+{
+#ifdef notyet
+	int i;
+	extern void alenlist_init(void);
+	extern void hwgraph_init(void);
+	extern void device_desc_init(void);
+
+	ASSERT(DEVICE_DRIVER_NONE == NULL);
+	alenlist_init();
+	hwgraph_init();
+	device_desc_init();
+
+	string_table_init(&driver_prefix_string_table);
+
+	for (i=0; i<DEVICE_DRIVER_HASH_SIZE; i++) {
+		spinlock_init(&device_driver_lock[i], "devdrv");
+		device_driver_hash[i] = NULL;
+	}
+
+	/* Initialize static drivers from master.c table */
+	for (i=0; i<static_devsw_count; i++) {
+		device_driver_t driver;
+		static_device_driver_desc_t desc;
+		int pri;
+
+		desc = &static_device_driver_table[i];
+		driver = device_driver_get(desc->sdd_prefix);
+		if (!driver)
+			driver = device_driver_alloc(desc->sdd_prefix);
+		pri = device_driver_sysgen_thread_pri_get(desc->sdd_prefix);
+		device_driver_thread_pri_set(driver, pri);
+		device_driver_devsw_put(driver, desc->sdd_bdevsw, desc->sdd_cdevsw);
+	}
+#endif
+	FIXME("device_driver_init");
+}
+
+/*
+** Hash a prefix string into a hash table chain.
+*/
+static int
+driver_prefix_hash(char *prefix)
+{
+#ifdef notyet
+	int accum = 0;
+	char nextchar;
+
+	while (nextchar = *prefix++)
+		accum = accum ^ nextchar;
+
+	return(accum % DEVICE_DRIVER_HASH_SIZE);
+#else
+	FIXME("driver_prefix_hash");
+	return(0);
+#endif
+}
+
+
+/*
+** Allocate a driver handle.
+** Returns the driver handle, or NULL if the driver prefix 
+** already has a handle.
+** 
+** Upper layers prevent races among device_driver_alloc,
+** device_driver_free, and device_driver_get*.
+*/
+device_driver_t
+device_driver_alloc(char *prefix)
+{
+#ifdef notyet
+	int which_hash;
+	device_driver_t new_driver;
+	int s;
+		
+	which_hash = driver_prefix_hash(prefix);
+
+	new_driver = kern_calloc(1, sizeof(*new_driver));
+	ASSERT(new_driver != NULL);
+	new_driver->dd_prev = NULL;
+	new_driver->dd_prefix = string_table_insert(&driver_prefix_string_table, prefix);
+	new_driver->dd_bdevsw = NULL;
+	new_driver->dd_cdevsw = NULL;
+
+	dev_admin_registry_init(&new_driver->dd_dev_admin_registry);
+	device_driver_admin_info_update(new_driver);
+
+	s = mutex_spinlock(&device_driver_lock[which_hash]);
+
+#if DEBUG
+	{
+		device_driver_t drvscan;
+
+		/* Make sure we haven't already added a driver with this prefix */
+		drvscan = device_driver_hash[which_hash];
+		while (drvscan && 
+		        strcmp(drvscan->dd_prefix, prefix)) {
+			drvscan = drvscan->dd_next;
+		}
+
+		ASSERT(!drvscan);
+	}
+#endif /* DEBUG */
+
+
+	/* Add new_driver to front of hash chain. */
+	new_driver->dd_next = device_driver_hash[which_hash];
+	if (new_driver->dd_next)
+		new_driver->dd_next->dd_prev = new_driver;
+	device_driver_hash[which_hash] = new_driver;
+
+	mutex_spinunlock(&device_driver_lock[which_hash], s);
+
+	return(new_driver);
+#else
+	FIXME("device_driver_alloc");
+	return((device_driver_t)0);
+#endif
+}
+
+/*
+** Free a driver handle.
+**
+** Statically loaded drivers should never device_driver_free.
+** Dynamically loaded drivers device_driver_free when either an
+** unloaded driver is unregistered, or when an unregistered driver
+** is unloaded.
+*/
+void
+device_driver_free(device_driver_t driver)
+{
+#ifdef notyet
+	int which_hash;
+	int s;
+
+	if (!driver)
+		return;
+
+	which_hash = driver_prefix_hash(driver->dd_prefix);
+
+	s = mutex_spinlock(&device_driver_lock[which_hash]);
+
+#if DEBUG
+	{
+		device_driver_t drvscan;
+
+		/* Make sure we're dealing with the right list */
+		drvscan = device_driver_hash[which_hash];
+		while (drvscan && (drvscan != driver))
+			drvscan = drvscan->dd_next;
+
+		ASSERT(drvscan);
+	}
+#endif /* DEBUG */
+
+	if (driver->dd_next)
+		driver->dd_next->dd_prev = driver->dd_prev;
+
+	if (driver->dd_prev)
+		driver->dd_prev->dd_next = driver->dd_next;
+	else
+		device_driver_hash[which_hash] = driver->dd_next;
+
+	mutex_spinunlock(&device_driver_lock[which_hash], s);
+
+	driver->dd_next = NULL;		/* sanity */
+	driver->dd_prev = NULL;		/* sanity */
+	driver->dd_prefix = NULL;	/* sanity */
+
+	if (driver->dd_bdevsw) {
+		driver->dd_bdevsw->d_driver = NULL;
+		driver->dd_bdevsw = NULL;
+	}
+
+	if (driver->dd_cdevsw) {
+		if (driver->dd_cdevsw->d_str) {
+			str_free_mux_node(driver);
+		}
+		driver->dd_cdevsw->d_driver = NULL;
+		driver->dd_cdevsw = NULL;
+	}
+
+	kern_free(driver);
+#endif
+	FIXME("device_driver_free");
+}
+
+
+/*
+** Given a device driver prefix, return a handle to the caller.
+*/
+device_driver_t
+device_driver_get(char *prefix)
+{
+#ifdef notyet
+	int which_hash;
+	device_driver_t drvscan;
+	int s;
+
+	if (prefix == NULL)
+		return(NULL);
+		
+	which_hash = driver_prefix_hash(prefix);
+
+	s = mutex_spinlock(&device_driver_lock[which_hash]);
+
+	drvscan = device_driver_hash[which_hash];
+	while (drvscan && strcmp(drvscan->dd_prefix, prefix))
+		drvscan = drvscan->dd_next;
+
+	mutex_spinunlock(&device_driver_lock[which_hash], s);
+
+	return(drvscan);
+#else
+	FIXME("device_driver_get");
+	return((device_driver_t)0);
+#endif
+}
+
+
+/*
+** Given a block or char special file devfs_handle_t, find the 
+** device driver that controls it.
+*/
+device_driver_t
+device_driver_getbydev(devfs_handle_t device)
+{
+#ifdef notyet
+	struct bdevsw *my_bdevsw;
+	struct cdevsw *my_cdevsw;
+
+	my_cdevsw = get_cdevsw(device);
+	if (my_cdevsw != NULL)
+		return(my_cdevsw->d_driver);
+
+	my_bdevsw = get_bdevsw(device);
+	if (my_bdevsw != NULL)
+		return(my_bdevsw->d_driver);
+
+#endif
+	FIXME("device_driver_getbydev");
+	return((device_driver_t)0);
+}
+
+
+/*
+** Associate a driver with bdevsw/cdevsw pointers.
+**
+** Statically loaded drivers are permanently and automatically associated
+** with the proper bdevsw/cdevsw.  Dynamically loaded drivers associate
+** themselves when the driver is registered, and disassociate when the
+** driver unregisters.
+**
+** Returns 0 on success, -1 on failure (devsw already associated with driver)
+*/
+int
+device_driver_devsw_put(device_driver_t driver,
+			struct bdevsw *my_bdevsw,
+			struct cdevsw *my_cdevsw)
+{
+#ifdef notyet
+	int i;
+
+	if (!driver)
+		return(-1);
+
+	/* Trying to re-register data?  */
+	if (((my_bdevsw != NULL) && (driver->dd_bdevsw != NULL)) ||
+	    ((my_cdevsw != NULL) && (driver->dd_cdevsw != NULL)))
+		return(-1);
+
+	if (my_bdevsw != NULL) {
+		driver->dd_bdevsw = my_bdevsw;
+		my_bdevsw->d_driver = driver;
+		for (i = 0; i < bdevmax; i++) {
+			if (driver->dd_bdevsw->d_flags == bdevsw[i].d_flags) {
+				bdevsw[i].d_driver = driver;
+				break;
+			}
+		}
+	}
+
+	if (my_cdevsw != NULL) {
+		driver->dd_cdevsw = my_cdevsw;
+		my_cdevsw->d_driver = driver;
+		for (i = 0; i < cdevmax; i++) {
+			if (driver->dd_cdevsw->d_flags == cdevsw[i].d_flags) {
+				cdevsw[i].d_driver = driver;
+				break;
+			}
+		}
+	}
+#endif
+	FIXME("device_driver_devsw_put");
+	return(0);
+}
+
+
+/*
+** Given a driver, return the corresponding bdevsw and cdevsw pointers.
+*/
+void
+device_driver_devsw_get(	device_driver_t driver, 
+				struct bdevsw **bdevswp,
+				struct cdevsw **cdevswp)
+{
+	if (!driver) {
+		*bdevswp = NULL;
+		*cdevswp = NULL;
+	} else {
+		*bdevswp = driver->dd_bdevsw;
+		*cdevswp = driver->dd_cdevsw;
+	}
+}
+
+/*
+ * device_driver_thread_pri_set
+ *	Given a driver try to set its thread priority.
+ *	Returns 0 on success , -1 on failure.
+ */ 
+int
+device_driver_thread_pri_set(device_driver_t driver,ilvl_t pri)
+{
+	if (!driver)
+		return(-1);
+	driver->dd_thread_pri = pri;
+	return(0);
+}
+/*
+ * device_driver_thread_pri_get
+ * 	Given a driver return the driver thread priority.
+ * 	If the driver is NULL return invalid driver thread
+ * 	priority.
+ */
+ilvl_t
+device_driver_thread_pri_get(device_driver_t driver)
+{
+	if (driver)
+		return(driver->dd_thread_pri);
+	else
+		return(DRIVER_THREAD_PRI_INVALID);
+}
+/*
+** Given a device driver, return it's handle (prefix).
+*/
+void
+device_driver_name_get(device_driver_t driver, char *buffer, int length)
+{
+	if (driver == NULL)
+		return;
+
+	strncpy(buffer, driver->dd_prefix, length);
+}
+
+
+/*
+** Associate a pointer-sized piece of information with a device.
+*/
+void 
+device_info_set(devfs_handle_t device, void *info)
+{
+#ifdef notyet
+	hwgraph_fastinfo_set(device, (arbitrary_info_t)info);
+#endif
+	FIXME("device_info_set");
+}
+
+
+/*
+** Retrieve a pointer-sized piece of information associated with a device.
+*/
+void *
+device_info_get(devfs_handle_t device)
+{
+#ifdef notyet
+	return((void *)hwgraph_fastinfo_get(device));
+#else
+	FIXME("device_info_get");
+	return(NULL);
+#endif
+}
+
+/*
+ * Find the thread priority for a device, from the various
+ * sysgen files.
+ */
+int
+device_driver_sysgen_thread_pri_get(char *dev_prefix)
+{
+#ifdef notyet
+	int pri;
+	char *pri_s;
+	char *class;
+
+	extern default_intr_pri;
+	extern disk_intr_pri;
+	extern serial_intr_pri;
+	extern parallel_intr_pri;
+	extern tape_intr_pri;
+	extern graphics_intr_pri;
+	extern network_intr_pri;
+	extern scsi_intr_pri;
+	extern audio_intr_pri;
+	extern video_intr_pri;
+	extern external_intr_pri;
+	extern tserialio_intr_pri;
+
+	/* Check if there is a thread priority specified for
+	 * this driver's thread thru admin hints. If so 
+	 * use that value. Otherwise set it to its default
+	 * class value, otherwise set it to the default
+	 * value.
+	 */
+
+	if (pri_s = device_driver_admin_info_get(dev_prefix,
+						ADMIN_LBL_THREAD_PRI)) {
+		pri = atoi(pri_s);
+	} else if (class = device_driver_admin_info_get(dev_prefix,
+						ADMIN_LBL_THREAD_CLASS)) {
+		if (strcmp(class, "disk") == 0)
+			pri = disk_intr_pri;
+		else if (strcmp(class, "serial") == 0)
+			pri = serial_intr_pri;
+		else if (strcmp(class, "parallel") == 0)
+			pri = parallel_intr_pri;
+		else if (strcmp(class, "tape") == 0)
+			pri = tape_intr_pri;
+		else if (strcmp(class, "graphics") == 0)
+			pri = graphics_intr_pri;
+		else if (strcmp(class, "network") == 0)
+			pri = network_intr_pri;
+		else if (strcmp(class, "scsi") == 0)
+			pri = scsi_intr_pri;
+		else if (strcmp(class, "audio") == 0)
+			pri = audio_intr_pri;
+		else if (strcmp(class, "video") == 0)
+			pri = video_intr_pri;
+		else if (strcmp(class, "external") == 0)
+			pri = external_intr_pri;
+		else if (strcmp(class, "tserialio") == 0)
+			pri = tserialio_intr_pri;
+		else
+			pri = default_intr_pri;
+	} else
+		pri = default_intr_pri;
+
+	if (pri > 255)
+		pri = 255;
+	else if (pri < 0)
+		pri = 0;
+	return pri;
+#else
+	FIXME("device_driver_sysgen_thread_pri_get");
+	return(-1);
+#endif
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/eeprom.c linux/arch/ia64/sn/io/eeprom.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/eeprom.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/eeprom.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,1457 @@
+/*
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Jack Steiner (steiner@sgi.com)
+ */
+
+
+/*
+ * WARNING:     There is more than one copy of this file in different isms.
+ *              All copies must be kept exactly in sync.
+ *              Do not modify this file without also updating the following:
+ *
+ *              irix/kern/io/eeprom.c
+ *              stand/arcs/lib/libsk/ml/eeprom.c
+ *		stand/arcs/lib/libkl/io/eeprom.c
+ *
+ *      (from time to time they might not be in sync but that's due to bringup
+ *       activity - this comment is to remind us that they eventually have to
+ *       get back together)
+ *
+ * eeprom.c
+ *
+ * access to board-mounted EEPROMs via the L1 system controllers
+ *
+ */
+
+/**************************************************************************
+ *                                                                        *
+ *  Copyright (C) 1999 Silicon Graphics, Inc.                             *
+ *                                                                        *
+ *  These coded instructions, statements, and computer programs  contain  *
+ *  unpublished  proprietary  information of Silicon Graphics, Inc., and  *
+ *  are protected by Federal copyright law.  They  may  not be disclosed  *
+ *  to  third  parties  or copied or duplicated in any form, in whole or  *
+ *  in part, without the prior written consent of Silicon Graphics, Inc.  *
+ *                                                                        *
+ **************************************************************************
+ */
+
+
+#include <linux/types.h>
+#include <linux/config.h>
+#include <linux/slab.h>
+#include <asm/sn/sgi.h>
+#include <asm/sn/iograph.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/hcl_util.h>
+#include <asm/sn/labelcl.h>
+#include <asm/sn/eeprom.h>
+#include <asm/sn/ksys/i2c.h>
+#include <asm/sn/cmn_err.h>
+/* #include <sys/SN/SN1/ip27log.h> */
+#include <asm/sn/router.h>
+#include <asm/sn/module.h>
+#include <asm/sn/ksys/l1.h>
+#include <asm/sn/nodepda.h>
+#include <asm/sn/clksupport.h>
+
+#if defined(EEPROM_DEBUG)
+#define db_printf(x) printk x
+#else
+#define db_printf(x) printk x
+#endif
+
+#define BCOPY(x,y,z)	memcpy(y,x,z)
+
+#define UNDERSCORE	0	/* don't convert underscores to hyphens */
+#define HYPHEN		1	/* convert underscores to hyphens */
+
+void		copy_ascii_field( char *to, char *from, int length,
+			          int change_underscore );
+uint64_t	generate_unique_id( char *sn, int sn_len );
+uchar_t		char_to_base36( char c );
+int		nicify( char *dst, eeprom_brd_record_t *src );
+static void	int64_to_hex_string( char *out, uint64_t val );
+
+// extern int router_lock( net_vec_t, int, int );
+// extern int router_unlock( net_vec_t );
+#define ROUTER_LOCK(p) 	// router_lock(p, 10000, 3000000)
+#define ROUTER_UNLOCK(p) 	// router_unlock(p)
+
+#define IP27LOG_OVNIC           "OverrideNIC"
+
+
+/* the following function converts an EEPROM record to a close facsimile
+ * of the string returned by reading a Dallas Semiconductor NIC (see
+ * one of the many incarnations of nic.c for details on that driver)
+ */
+int nicify( char *dst, eeprom_brd_record_t *src )
+{
+    int field_len;
+    uint64_t unique_id;
+    char *cur_dst = dst;
+    eeprom_board_ia_t   *board;
+
+    board   = src->board_ia;
+    ASSERT( board );  /* there should always be a board info area */
+
+    /* copy part number */
+    strcpy( cur_dst, "Part:" );
+    cur_dst += strlen( cur_dst );
+    ASSERT( (board->part_num_tl & FIELD_FORMAT_MASK)
+	    == FIELD_FORMAT_ASCII );
+    field_len = board->part_num_tl & FIELD_LENGTH_MASK;
+    copy_ascii_field( cur_dst, board->part_num, field_len, HYPHEN );
+    cur_dst += field_len;
+
+    /* copy product name */
+    strcpy( cur_dst, ";Name:" );
+    cur_dst += strlen( cur_dst );
+    ASSERT( (board->product_tl & FIELD_FORMAT_MASK) == FIELD_FORMAT_ASCII );
+    field_len = board->product_tl & FIELD_LENGTH_MASK;
+    copy_ascii_field( cur_dst, board->product, field_len, UNDERSCORE );
+    cur_dst += field_len;
+
+    /* copy serial number */
+    strcpy( cur_dst, ";Serial:" );
+    cur_dst += strlen( cur_dst );
+    ASSERT( (board->serial_num_tl & FIELD_FORMAT_MASK)
+	    == FIELD_FORMAT_ASCII );
+    field_len = board->serial_num_tl & FIELD_LENGTH_MASK;
+    copy_ascii_field( cur_dst, board->serial_num, field_len,
+		      HYPHEN);
+
+    cur_dst += field_len;
+
+    /* copy revision */
+    strcpy( cur_dst, ";Revision:");
+    cur_dst += strlen( cur_dst );
+    ASSERT( (board->board_rev_tl & FIELD_FORMAT_MASK)
+	    == FIELD_FORMAT_ASCII );
+    field_len = board->board_rev_tl & FIELD_LENGTH_MASK;
+    copy_ascii_field( cur_dst, board->board_rev, field_len, HYPHEN );
+    cur_dst += field_len;
+
+    /* EEPROMs don't have equivalents for the Group, Capability and
+     * Variety fields, so we pad these with 0's
+     */
+    strcpy( cur_dst, ";Group:ff;Capability:ffffffff;Variety:ff" );
+    cur_dst += strlen( cur_dst );
+
+    /* use the board serial number to "fake" a laser id */
+    strcpy( cur_dst, ";Laser:" );
+    cur_dst += strlen( cur_dst );
+    unique_id = generate_unique_id( board->serial_num,
+				    board->serial_num_tl & FIELD_LENGTH_MASK );
+    int64_to_hex_string( cur_dst, unique_id );
+    strcat( dst, ";" );
+
+    return 1;
+}
+
+
+/* These functions borrow heavily from chars2* in nic.c
+ */
+void copy_ascii_field( char *to, char *from, int length,
+		       int change_underscore )
+{
+    int i;
+    for( i = 0; i < length; i++ ) {
+
+	/* change underscores to hyphens if requested */
+	if( from[i] == '_' && change_underscore == HYPHEN )
+	    to[i] = '-';
+
+	/* ; and ; are separators, so mustn't appear within
+	 * a field */
+	else if( from[i] == ':' || from[i] == ';' )
+	    to[i] = '?';
+
+	/* I'm not sure why or if ASCII character 0xff would
+	 * show up in an EEPROM field, but the NIC parsing
+	 * routines wouldn't like it if it did... so we
+	 * get rid of it, just in case. */
+	else if( (unsigned char)from[i] == (unsigned char)0xff )
+	    to[i] = ' ';
+	
+	/* unprintable characters are replaced with . */
+	else if( from[i] < ' ' || from[i] >= 0x7f )
+	    to[i] = '.';
+
+	/* otherwise, just copy the character */
+	else
+	    to[i] = from[i];
+    }
+
+    if( i == 0 ) {
+	to[i] = ' '; /* return at least a space... */
+	i++;
+    }
+    to[i] = 0;	     /* terminating null */
+}
+
+/* Note that int64_to_hex_string currently only has a big-endian
+ * implementation.
+ */
+#ifdef _MIPSEB
+static void int64_to_hex_string( char *out, uint64_t val )
+{
+    int i;
+    uchar_t table[] = "0123456789abcdef";
+    uchar_t *byte_ptr = (uchar_t *)&val;
+    for( i = 0; i < sizeof(uint64_t); i++ ) {
+	out[i*2] = table[ ((*byte_ptr) >> 4) & 0x0f ];
+	out[i*2+1] = table[ (*byte_ptr) & 0x0f ];
+	byte_ptr++;
+    }
+    out[i*2] = '\0';
+}
+
+#else /* little endian */
+
+static void int64_to_hex_string( char *out, uint64_t val )
+{
+
+
+	printk("int64_to_hex_string needs a little-endian implementation.\n");
+}
+#endif /* _MIPSEB */
+
+/* Convert a standard ASCII serial number to a unique integer
+ * id number by treating the serial number string as though
+ * it were a base 36 number
+ */
+uint64_t generate_unique_id( char *sn, int sn_len )
+{
+    int uid = 0;
+    int i;
+
+    #define VALID_BASE36(c)	((c >= '0' && c <='9') \
+			    ||   (c >= 'A' && c <='Z') \
+			    ||   (c >= 'a' && c <='z'))
+
+    for( i = 0; i < sn_len; i++ ) {
+	if( !VALID_BASE36(sn[i]) )
+	    continue;
+	uid *= 36;
+	uid += char_to_base36( sn[i] );
+    }
+
+    if( uid == 0 )
+	return rtc_time();
+
+    return uid;
+}
+
+uchar_t char_to_base36( char c )
+{
+    uchar_t val;
+
+    if( c >= '0' && c <= '9' )
+	val = (c - '0');
+
+    else if( c >= 'A' && c <= 'Z' )
+	val = (c - 'A' + 10);
+
+    else if( c >= 'a' && c <= 'z' )
+	val = (c - 'a' + 10);
+
+    else val = 0;
+
+    return val;
+}
+
+
+/* given a pointer to the three-byte little-endian EEPROM representation
+ * of date-of-manufacture, this function translates to a big-endian
+ * integer format
+ */
+int eeprom_xlate_board_mfr_date( uchar_t *src )
+{
+    int rval = 0;
+    rval += *src; src++;
+    rval += ((int)(*src) << 8); src ++;
+    rval += ((int)(*src) << 16);
+    return rval;
+}
+
+
+int eeprom_str( char *nic_str, nasid_t nasid, int component )
+{
+    eeprom_brd_record_t eep;
+    eeprom_board_ia_t board;
+    eeprom_chassis_ia_t chassis;
+    int r;
+
+    if( (component & C_DIMM) == C_DIMM ) {
+	/* this function isn't applicable to DIMMs */
+	return EEP_PARAM;
+    }
+    else {
+	eep.board_ia = &board;
+	eep.spd = NULL;
+	if( !(component & SUBORD_MASK) )
+	    eep.chassis_ia = &chassis;  /* only main boards have a chassis
+					 * info area */
+	else
+	    eep.chassis_ia = NULL;
+    }
+    
+    switch( component & BRICK_MASK ) {
+      case C_BRICK:
+	r = cbrick_eeprom_read( &eep, nasid, component );
+	break;
+      case IO_BRICK:
+	r = iobrick_eeprom_read( &eep, nasid, component );
+	break;
+      default:
+	return EEP_PARAM;  /* must be an invalid component */
+    }
+    if( r )
+	return r;
+    if( !nicify( nic_str, &eep ) )
+	return EEP_NICIFY;
+
+    return EEP_OK;
+}
+
+int vector_eeprom_str( char *nic_str, nasid_t nasid,
+		       int component, net_vec_t path )
+{
+    eeprom_brd_record_t eep;
+    eeprom_board_ia_t board;
+    eeprom_chassis_ia_t chassis;
+    int r;
+
+    eep.board_ia = &board;
+    if( !(component & SUBORD_MASK) )
+        eep.chassis_ia = &chassis;  /* only main boards have a chassis
+                                     * info area */
+    else
+        eep.chassis_ia = NULL;
+
+    if( !(component & VECTOR) )
+	return EEP_PARAM;
+
+    if( (r = vector_eeprom_read( &eep, nasid, path, component )) )
+	return r;
+
+    if( !nicify( nic_str, &eep ) )
+        return EEP_NICIFY;
+
+    return EEP_OK;
+}
+
+
+int is_iobrick( int nasid, int widget_num )
+{
+    uint32_t wid_reg;
+    int part_num, mfg_num;
+
+    /* Read the widget's WIDGET_ID register to get
+     * its part number and mfg number
+     */
+    wid_reg = *(volatile int32_t *)
+        (NODE_SWIN_BASE( nasid, widget_num ) + WIDGET_ID);
+
+    part_num = (wid_reg & WIDGET_PART_NUM) >> WIDGET_PART_NUM_SHFT;
+    mfg_num = (wid_reg & WIDGET_MFG_NUM) >> WIDGET_MFG_NUM_SHFT;
+
+    /* Is this the "xbow part" of an XBridge?  If so, this
+     * widget is definitely part of an I/O brick.
+     */
+    if( part_num == XXBOW_WIDGET_PART_NUM &&
+	mfg_num == XXBOW_WIDGET_MFGR_NUM )
+
+	return 1;
+
+    /* Is this a "bridge part" of an XBridge?  If so, once
+     * again, we know this widget is part of an I/O brick.
+     */
+    if( part_num == XBRIDGE_WIDGET_PART_NUM &&
+	mfg_num == XBRIDGE_WIDGET_MFGR_NUM )
+
+	return 1;
+
+    return 0;
+}
+
+
+int cbrick_uid_get( nasid_t nasid, uint64_t *uid )
+{
+#if !defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+    return EEP_L1;
+#else
+    char uid_str[32];
+    char msg[BRL1_QSIZE];
+    int subch, len;
+    l1sc_t sc;
+    l1sc_t *scp;
+    int local = (nasid == get_nasid());
+
+    if ( IS_RUNNING_ON_SIMULATOR() )
+	return EEP_L1;
+
+    /* If the promlog variable pointed to by IP27LOG_OVNIC is set,
+     * use that value for the cbrick UID rather than the EEPROM
+     * serial number.
+     */
+#ifdef LOG_GETENV
+    if( ip27log_getenv( nasid, IP27LOG_OVNIC, uid_str, NULL, 0 ) >= 0 )
+    {
+	/* We successfully read IP27LOG_OVNIC, so return it as the UID. */
+	db_printf(( "cbrick_uid_get:"
+		    "Overriding UID with environment variable %s\n", 
+		    IP27LOG_OVNIC ));
+	*uid = strtoull( uid_str, NULL, 0 );
+	return EEP_OK;
+    }
+#endif
+
+    /* If this brick is retrieving its own uid, use the local l1sc_t to
+     * arbitrate access to the l1; otherwise, set up a new one.
+     */
+    if( local ) {
+	scp = get_l1sc();
+    }
+    else {
+	scp = &sc;
+	sc_init( &sc, nasid, BRL1_LOCALUART );
+    }
+
+    /* fill in msg with the opcode & params */
+    BZERO( msg, BRL1_QSIZE );
+    if( (subch = sc_open( scp, L1_ADDR_LOCAL )) < 0 )
+	return EEP_L1;
+
+    if( (len = sc_construct_msg( scp, subch, msg, BRL1_QSIZE,
+				 L1_ADDR_TASK_GENERAL,
+				 L1_REQ_SER_NUM, 0 )) < 0 )
+    {
+	sc_close( scp, subch );
+	return( EEP_L1 );
+    }
+
+    /* send the request to the L1 */
+    if( sc_command( scp, subch, msg, msg, &len ) ) {
+	sc_close( scp, subch );
+	return( EEP_L1 );
+    }
+
+    /* free up subchannel */
+    sc_close(scp, subch);
+
+    /* check response */
+    if( sc_interpret_resp( msg, 2, L1_ARG_ASCII, uid_str ) < 0 )
+    {
+	return( EEP_L1 );
+    }
+
+    *uid = generate_unique_id( uid_str, strlen( uid_str ) );
+
+    return EEP_OK;
+#endif /* CONFIG_SERIAL_SGI_L1_PROTOCOL */
+}
+
+
+int rbrick_uid_get( nasid_t nasid, net_vec_t path, uint64_t *uid )
+{
+#if !defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+    return EEP_L1;
+#else
+    char uid_str[32];
+    char msg[BRL1_QSIZE];
+    int subch, len;
+    l1sc_t sc;
+
+    if ( IS_RUNNING_ON_SIMULATOR() )
+	return EEP_L1;
+
+#ifdef BRINGUP
+#define FAIL								\
+    {									\
+	*uid = rtc_time();						\
+	printk( "rbrick_uid_get failed; using current time as uid\n" );	\
+	return EEP_OK;							\
+    }
+#endif /* BRINGUP */
+
+    ROUTER_LOCK(path);
+    sc_init( &sc, nasid, path );
+
+    /* fill in msg with the opcode & params */
+    BZERO( msg, BRL1_QSIZE );
+    if( (subch = sc_open( &sc, L1_ADDR_LOCAL )) < 0 ) {
+	ROUTER_UNLOCK(path);
+	FAIL;
+    }
+
+    if( (len = sc_construct_msg( &sc, subch, msg, BRL1_QSIZE,
+				 L1_ADDR_TASK_GENERAL,
+				 L1_REQ_SER_NUM, 0 )) < 0 )
+    {
+	ROUTER_UNLOCK(path);
+	sc_close( &sc, subch );
+	FAIL;
+    }
+
+    /* send the request to the L1 */
+    if( sc_command( &sc, subch, msg, msg, &len ) ) {
+	ROUTER_UNLOCK(path);
+	sc_close( &sc, subch );
+	FAIL;
+    }
+
+    /* free up subchannel */
+    ROUTER_UNLOCK(path);
+    sc_close(&sc, subch);
+
+    /* check response */
+    if( sc_interpret_resp( msg, 2, L1_ARG_ASCII, uid_str ) < 0 )
+    {
+	FAIL;
+    }
+
+    *uid = generate_unique_id( uid_str, strlen( uid_str ) );
+
+    return EEP_OK;
+#endif /* CONFIG_SERIAL_SGI_L1_PROTOCOL */
+}
+
+int iobrick_uid_get( nasid_t nasid, uint64_t *uid )
+{
+    eeprom_brd_record_t eep;
+    eeprom_board_ia_t board;
+    eeprom_chassis_ia_t chassis;
+    int r;
+
+    eep.board_ia = &board;
+    eep.chassis_ia = &chassis;
+    eep.spd = NULL;
+
+    r = iobrick_eeprom_read( &eep, nasid, IO_BRICK );
+    if( r != EEP_OK ) {
+        *uid = rtc_time();
+        return r;
+    }
+
+    *uid = generate_unique_id( board.serial_num,
+                               board.serial_num_tl & FIELD_LENGTH_MASK );
+
+    return EEP_OK;
+}
+
+
+int ibrick_mac_addr_get( nasid_t nasid, char *eaddr )
+{
+    eeprom_brd_record_t eep;
+    eeprom_board_ia_t board;
+    eeprom_chassis_ia_t chassis;
+    int r;
+    char *tmp;
+
+    eep.board_ia = &board;
+    eep.chassis_ia = &chassis;
+    eep.spd = NULL;
+
+    r = iobrick_eeprom_read( &eep, nasid, IO_BRICK );
+    if( (r != EEP_OK) || (board.mac_addr[0] == '\0') ) {
+	db_printf(( "ibrick_mac_addr_get: "
+		    "Couldn't read MAC address from EEPROM\n" ));
+	return EEP_L1;
+    }
+    else {
+	/* successfully read info area */
+	int ix;
+	tmp = board.mac_addr;
+	for( ix = 0; ix < (board.mac_addr_tl & FIELD_LENGTH_MASK); ix++ )
+	{
+	    *eaddr++ = *tmp++;
+	}
+	*eaddr = '\0';
+    }
+
+    return EEP_OK;
+}
+
+
+/* 
+ * eeprom_vertex_info_set
+ *
+ * Given a vertex handle, a component designation, a starting nasid
+ * and (in the case of a router) a vector path to the component, this
+ * function will read the EEPROM and attach the resulting information
+ * to the vertex in the same string format as that provided by the
+ * Dallas Semiconductor NIC drivers.  If the vertex already has the
+ * string, this function just returns the string.
+ */
+
+extern char *nic_vertex_info_get( devfs_handle_t );
+extern void nic_vmc_check( devfs_handle_t, char * );
+#ifdef BRINGUP
+/* the following were lifted from nic.c - change later? */
+#define MAX_INFO 2048
+#define NEWSZ(ptr,sz)   ((ptr) = kern_malloc((sz)))
+#define DEL(ptr) (kern_free((ptr)))
+#endif /* BRINGUP */
+
+char *eeprom_vertex_info_set( int component, int nasid, devfs_handle_t v,
+                              net_vec_t path )
+{
+        char *info_tmp;
+        int info_len;
+        char *info;
+
+        /* see if this vertex is already marked */
+        info_tmp = nic_vertex_info_get(v);
+        if (info_tmp) return info_tmp;
+
+        /* get a temporary place for the data */
+        NEWSZ(info_tmp, MAX_INFO);
+        if (!info_tmp) return NULL;
+
+        /* read the EEPROM */
+	if( component & R_BRICK ) {
+	    if( RBRICK_EEPROM_STR( info_tmp, nasid, path ) != EEP_OK )
+		return NULL;
+	}
+	else {
+            if( eeprom_str( info_tmp, nasid, component ) != EEP_OK )
+	        return NULL;
+	}
+
+        /* allocate a smaller final place */
+        info_len = strlen(info_tmp)+1;
+        NEWSZ(info, info_len);
+        if (info) {
+                strcpy(info, info_tmp);
+                DEL(info_tmp);
+        } else {
+                info = info_tmp;
+        }
+
+        /* add info to the vertex */
+        hwgraph_info_add_LBL(v, INFO_LBL_NIC,
+                             (arbitrary_info_t) info);
+
+        /* see if someone else got there first */
+        info_tmp = nic_vertex_info_get(v);
+        if (info != info_tmp) {
+            DEL(info);
+            return info_tmp;
+        }
+
+        /* export the data */
+        hwgraph_info_export_LBL(v, INFO_LBL_NIC, info_len);
+
+        /* trigger all matching callbacks */
+        nic_vmc_check(v, info);
+
+        return info;
+}
+
+
+/*********************************************************************
+ *
+ * stubs for use until the Bedrock/L1 link is available
+ *
+ */
+
+#include <asm/sn/nic.h>
+
+/* #define EEPROM_TEST */
+
+/* fake eeprom reading functions (replace when the BR/L1 communication
+ * channel is in working order)
+ */
+
+
+/* generate a charater in [0-9A-Z]; if an "extra" character is
+ * specified (such as '_'), include it as one of the possibilities.
+ */
+char random_eeprom_ch( char extra ) 
+{
+    char ch;
+    int modval = 36;
+    if( extra )
+	modval++;
+    
+    ch = rtc_time() % modval;
+
+    if( ch < 10 )
+        ch += '0';
+    else if( ch >= 10 && ch < 36 )
+	ch += ('A' - 10);
+    else
+	ch = extra;
+
+    return ch;
+}
+
+/* create a part number of the form xxx-xxxx-xxx.
+ * It may be important later to generate different
+ * part numbers depending on the component we're
+ * supposed to be "reading" from, so the component
+ * paramter is provided.
+ */
+void fake_a_part_number( char *buf, int component )
+{
+    int i;
+    switch( component ) {
+
+    /* insert component-specific routines here */
+
+    case C_BRICK:
+	strcpy( buf, "030-1266-001" );
+	break;
+    default:
+        for( i = 0; i < 12; i++ ) {
+	    if( i == 3 || i == 8 )
+	        buf[i] = '-';
+	    else
+	        buf[i] = random_eeprom_ch(0);
+        }
+    }
+}
+
+
+/* create a six-character serial number */
+void fake_a_serial_number( char *buf, uint64_t ser )
+{
+    int i;
+    static const char hexchars[] = "0123456789ABCDEF";
+
+    if (ser) {
+	for( i = 5; i >=0; i-- ) {
+	    buf[i] = hexchars[ser & 0xf];
+	    ser >>= 4;
+	}
+    }
+    else {
+	for( i = 0; i < 6; i++ )
+	    buf[i] = random_eeprom_ch(0);
+    }
+}
+
+
+void fake_a_product_name( uchar_t *format, char* buf, int component )
+{
+    switch( component & BRICK_MASK ) {
+
+    case C_BRICK:
+	if( component & SUBORD_MASK ) {
+	    strcpy( buf, "C_BRICK_SUB" );
+	    *format = 0xCB;
+	}
+	else {
+	    strcpy( buf, "IP35" );
+	    *format = 0xC4;
+	}
+	break;
+
+    case R_BRICK:
+        if( component & SUBORD_MASK ) {
+            strcpy( buf, "R_BRICK_SUB" );
+            *format = 0xCB;
+        }
+        else {
+            strcpy( buf, "R_BRICK" );
+            *format = 0xC7;
+        }
+        break;
+
+    case IO_BRICK:
+        if( component & SUBORD_MASK ) {
+            strcpy( buf, "IO_BRICK_SUB" );
+            *format = 0xCC;
+        }
+        else {
+            strcpy( buf, "IO_BRICK" );
+            *format = 0xC8;
+        }
+        break;
+
+    default:
+	strcpy( buf, "UNK_DEVICE" );
+	*format = 0xCA;
+    }
+}
+
+
+
+int fake_an_eeprom_record( eeprom_brd_record_t *buf, int component, 
+			   uint64_t ser )
+{
+    eeprom_board_ia_t *board;
+    eeprom_chassis_ia_t *chassis;
+    int i, cs;
+
+    board = buf->board_ia;
+    chassis = buf->chassis_ia;
+
+    if( !(component & SUBORD_MASK) ) {
+	if( !chassis )
+	    return EEP_PARAM;
+	chassis->format = 0;
+	chassis->length = 5;
+	chassis->type = 0x17;
+
+	chassis->part_num_tl = 0xCC;
+	fake_a_part_number( chassis->part_num, component );
+	chassis->serial_num_tl = 0xC6;
+	fake_a_serial_number( chassis->serial_num, ser );
+
+	cs = chassis->format + chassis->length + chassis->type
+	    + chassis->part_num_tl + chassis->serial_num_tl;
+	for( i = 0; i < (chassis->part_num_tl & FIELD_LENGTH_MASK); i++ )
+	    cs += chassis->part_num[i];
+	for( i = 0; i < (chassis->serial_num_tl & FIELD_LENGTH_MASK); i++ )
+	    cs += chassis->serial_num[i];
+	chassis->checksum = 256 - (cs % 256);
+    }
+
+    if( !board )
+	return EEP_PARAM;
+    board->format = 0;
+    board->length = 10;
+    board->language = 0;
+    board->mfg_date = 1789200; /* noon, 5/26/99 */
+    board->manuf_tl = 0xC3;
+    strcpy( board->manuf, "SGI" );
+
+    fake_a_product_name( &(board->product_tl), board->product, component );
+
+    board->serial_num_tl = 0xC6;
+    fake_a_serial_number( board->serial_num, ser );
+
+    board->part_num_tl = 0xCC;
+    fake_a_part_number( board->part_num, component );
+
+    board->board_rev_tl = 0xC2;
+    board->board_rev[0] = '0';
+    board->board_rev[1] = '1';
+
+    board->eeprom_size_tl = 0x01;
+    board->eeprom_size = 1;
+
+    board->temp_waiver_tl = 0xC2;
+    board->temp_waiver[0] = '0';
+    board->temp_waiver[1] = '1';
+
+    cs = board->format + board->length + board->language
+	+ (board->mfg_date & 0xFF)
+	+ (board->mfg_date & 0xFF00)
+	+ (board->mfg_date & 0xFF0000)
+	+ board->manuf_tl + board->product_tl + board->serial_num_tl
+	+ board->part_num_tl + board->board_rev_tl
+	+ board->board_rev[0] + board->board_rev[1]
+	+ board->eeprom_size_tl + board->eeprom_size + board->temp_waiver_tl
+	+ board->temp_waiver[0] + board->temp_waiver[1];
+    for( i = 0; i < (board->manuf_tl & FIELD_LENGTH_MASK); i++ )
+	cs += board->manuf[i];
+    for( i = 0; i < (board->product_tl & FIELD_LENGTH_MASK); i++ )
+	cs += board->product[i];
+    for( i = 0; i < (board->serial_num_tl & FIELD_LENGTH_MASK); i++ )
+	cs += board->serial_num[i];
+    for( i = 0; i < (board->part_num_tl & FIELD_LENGTH_MASK); i++ )
+	cs += board->part_num[i];
+    
+    board->checksum = 256 - (cs % 256);
+
+    return EEP_OK;
+}
+
+#define EEPROM_CHUNKSIZE	64
+
+#if defined(EEPROM_DEBUG)
+#define RETURN_ERROR							\
+{									\
+    printk( "read_ia error return, component 0x%x, line %d"		\
+	    ", address 0x%x, ia code 0x%x\n",				\
+	    l1_compt, __LINE__, sc->subch[subch].target, ia_code );	\
+    return EEP_L1;							\
+}
+
+#else
+#define RETURN_ERROR	return(EEP_L1)
+#endif
+
+int read_ia( l1sc_t *sc, int subch, int l1_compt, 
+	     int ia_code, char *eep_record )
+{
+#if !defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+    return EEP_L1;
+#else
+    char msg[BRL1_QSIZE]; 	   /* message buffer */
+    int len;              	   /* number of bytes used in message buffer */
+    int ia_len = EEPROM_CHUNKSIZE; /* remaining bytes in info area */
+    int offset = 0;                /* current offset into info area */
+
+    if ( IS_RUNNING_ON_SIMULATOR() )
+	return EEP_L1;
+
+    BZERO( msg, BRL1_QSIZE );
+
+    /* retrieve EEPROM data in 64-byte chunks
+     */
+
+    while( ia_len )
+    {
+	/* fill in msg with opcode & params */
+	if( (len = sc_construct_msg( sc, subch, msg, BRL1_QSIZE,
+				     L1_ADDR_TASK_GENERAL,
+				     L1_REQ_EEPROM, 8,
+				     L1_ARG_INT, l1_compt,
+				     L1_ARG_INT, ia_code,
+				     L1_ARG_INT, offset,
+				     L1_ARG_INT, ia_len )) < 0 )
+	{
+	    RETURN_ERROR;
+	}
+
+	/* send the request to the L1 */
+
+	if( sc_command( sc, subch, msg, msg, &len ) ) {
+	    RETURN_ERROR;
+	}
+
+	/* check response */
+	if( sc_interpret_resp( msg, 5, 
+			       L1_ARG_INT, &ia_len,
+			       L1_ARG_UNKNOWN, &len, eep_record ) < 0 )
+	{
+	    RETURN_ERROR;
+	}
+
+	if( ia_len > EEPROM_CHUNKSIZE )
+	    ia_len = EEPROM_CHUNKSIZE;
+
+	eep_record += EEPROM_CHUNKSIZE;
+	offset += EEPROM_CHUNKSIZE;
+    }
+
+    return EEP_OK;
+#endif /* CONFIG_SERIAL_SGI_L1_PROTOCOL */
+}
+
+
+int read_spd( l1sc_t *sc, int subch, int l1_compt,
+	      eeprom_spd_u *spd )
+{
+#if !defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+    return EEP_L1;
+#else
+    char msg[BRL1_QSIZE]; 	    /* message buffer */
+    int len;              	    /* number of bytes used in message buffer */
+    int spd_len = EEPROM_CHUNKSIZE; /* remaining bytes in spd record */
+    int offset = 0;		    /* current offset into spd record */
+    char *spd_p = spd->bytes;	    /* "thumb" for writing to spd */
+
+    if ( IS_RUNNING_ON_SIMULATOR() )
+	return EEP_L1;
+
+    BZERO( msg, BRL1_QSIZE );
+
+    /* retrieve EEPROM data in 64-byte chunks
+     */
+
+    while( spd_len )
+    {
+	/* fill in msg with opcode & params */
+	if( (len = sc_construct_msg( sc, subch, msg, BRL1_QSIZE,
+				     L1_ADDR_TASK_GENERAL,
+				     L1_REQ_EEPROM, 8,
+				     L1_ARG_INT, l1_compt,
+				     L1_ARG_INT, L1_EEP_SPD,
+				     L1_ARG_INT, offset,
+				     L1_ARG_INT, spd_len )) < 0 )
+	{
+	    return( EEP_L1 );
+	}
+
+	/* send the request to the L1 */
+	if( sc_command( sc, subch, msg, msg, &len ) ) {
+	    return( EEP_L1 );
+	}
+
+	/* check response */
+	if( sc_interpret_resp( msg, 5, 
+			       L1_ARG_INT, &spd_len,
+			       L1_ARG_UNKNOWN, &len, spd_p ) < 0 )
+	{
+	    return( EEP_L1 );
+	}
+
+	if( spd_len > EEPROM_CHUNKSIZE )
+	    spd_len = EEPROM_CHUNKSIZE;
+
+	spd_p += EEPROM_CHUNKSIZE;
+	offset += EEPROM_CHUNKSIZE;
+    }
+    return EEP_OK;
+#endif /* CONFIG_SERIAL_SGI_L1_PROTOCOL */
+}
+
+
+int read_chassis_ia( l1sc_t *sc, int subch, int l1_compt,
+		     eeprom_chassis_ia_t *ia )
+{
+    char eep_record[512];          /* scratch area for building up info area */
+    char *eep_rec_p = eep_record;  /* thumb for moving through eep_record */
+    int checksum = 0;              /* use to verify eeprom record checksum */
+    int i;
+
+    /* Read in info area record from the L1.
+     */
+    if( read_ia( sc, subch, l1_compt, L1_EEP_CHASSIS, eep_record )
+	!= EEP_OK )
+    {
+	return EEP_L1;
+    }
+
+    /* Now we've got the whole info area.  Transfer it to the data structure.
+     */
+
+    eep_rec_p = eep_record;
+    ia->format = *eep_rec_p++;
+    ia->length = *eep_rec_p++;
+    if( ia->length == 0 ) {
+	/* since we're using 8*ia->length-1 as an array index later, make
+	 * sure it's sane.
+	 */
+	db_printf(( "read_chassis_ia: eeprom length byte of ZERO\n" ));
+	return EEP_L1;
+    }
+    ia->type = *eep_rec_p++;
+   
+    ia->part_num_tl = *eep_rec_p++;
+
+    (void)BCOPY( eep_rec_p, ia->part_num, (ia->part_num_tl & FIELD_LENGTH_MASK) );
+    eep_rec_p += (ia->part_num_tl & FIELD_LENGTH_MASK);
+
+    ia->serial_num_tl = *eep_rec_p++;
+
+    BCOPY( eep_rec_p, ia->serial_num, 
+	   (ia->serial_num_tl & FIELD_LENGTH_MASK) );
+    eep_rec_p += (ia->serial_num_tl & FIELD_LENGTH_MASK);
+
+    ia->checksum = eep_record[(8 * ia->length) - 1];
+
+    /* verify checksum */
+    eep_rec_p = eep_record;
+    checksum = 0;
+    for( i = 0; i < (8 * ia->length); i++ ) {
+	checksum += *eep_rec_p++;
+    }
+
+    if( (checksum & 0xff) != 0 )
+    {
+	db_printf(( "read_chassis_ia: bad checksum\n" ));
+	db_printf(( "read_chassis_ia: target 0x%x  uart 0x%x\n",
+			   sc->subch[subch].target, sc->uart ));
+	return EEP_BAD_CHECKSUM;
+    }
+
+    return EEP_OK;
+}
+
+
+int read_board_ia( l1sc_t *sc, int subch, int l1_compt,
+		   eeprom_board_ia_t *ia )
+{
+    char eep_record[512];          /* scratch area for building up info area */
+    char *eep_rec_p = eep_record;  /* thumb for moving through eep_record */
+    int checksum = 0;              /* running checksum total */
+    int i;
+
+    BZERO( ia, sizeof( eeprom_board_ia_t ) );
+
+    /* Read in info area record from the L1.
+     */
+    if( read_ia( sc, subch, l1_compt, L1_EEP_BOARD, eep_record )
+	!= EEP_OK )
+    {
+	db_printf(( "read_board_ia: error reading info area from L1\n" ));
+	return EEP_L1;
+    }
+
+     /* Now we've got the whole info area.  Transfer it to the data structure.
+      */
+
+    eep_rec_p = eep_record;
+    ia->format = *eep_rec_p++;
+    ia->length = *eep_rec_p++;
+    if( ia->length == 0 ) {
+	/* since we're using 8*ia->length-1 as an array index later, make
+	 * sure it's sane.
+	 */
+	db_printf(( "read_board_ia: eeprom length byte of ZERO\n" ));
+	return EEP_L1;
+    }
+    ia->language = *eep_rec_p++;
+    
+    ia->mfg_date = eeprom_xlate_board_mfr_date( (uchar_t *)eep_rec_p );
+    eep_rec_p += 3;
+
+    ia->manuf_tl = *eep_rec_p++;
+    
+    BCOPY( eep_rec_p, ia->manuf, (ia->manuf_tl & FIELD_LENGTH_MASK) );
+    eep_rec_p += (ia->manuf_tl & FIELD_LENGTH_MASK);
+
+    ia->product_tl = *eep_rec_p++;
+    
+    BCOPY( eep_rec_p, ia->product, (ia->product_tl & FIELD_LENGTH_MASK) );
+    eep_rec_p += (ia->product_tl & FIELD_LENGTH_MASK);
+
+    ia->serial_num_tl = *eep_rec_p++;
+    
+    BCOPY(eep_rec_p, ia->serial_num, (ia->serial_num_tl & FIELD_LENGTH_MASK));
+    eep_rec_p += (ia->serial_num_tl & FIELD_LENGTH_MASK);
+
+    ia->part_num_tl = *eep_rec_p++;
+
+    BCOPY( eep_rec_p, ia->part_num, (ia->part_num_tl & FIELD_LENGTH_MASK) );
+    eep_rec_p += (ia->part_num_tl & FIELD_LENGTH_MASK);
+
+    eep_rec_p++; /* we do not use the FRU file id */
+    
+    ia->board_rev_tl = *eep_rec_p++;
+    
+    BCOPY( eep_rec_p, ia->board_rev, (ia->board_rev_tl & FIELD_LENGTH_MASK) );
+    eep_rec_p += (ia->board_rev_tl & FIELD_LENGTH_MASK);
+
+    ia->eeprom_size_tl = *eep_rec_p++;
+    ia->eeprom_size = *eep_rec_p++;
+
+    ia->temp_waiver_tl = *eep_rec_p++;
+    
+    BCOPY( eep_rec_p, ia->temp_waiver, 
+	   (ia->temp_waiver_tl & FIELD_LENGTH_MASK) );
+    eep_rec_p += (ia->temp_waiver_tl & FIELD_LENGTH_MASK);
+
+    /* if there's more, we must be reading a main board; get
+     * additional fields
+     */
+    if( ((unsigned char)*eep_rec_p != (unsigned char)EEPROM_EOF) ) {
+
+	ia->ekey_G_tl = *eep_rec_p++;
+	BCOPY( eep_rec_p, (char *)&ia->ekey_G, 
+	       ia->ekey_G_tl & FIELD_LENGTH_MASK );
+	eep_rec_p += (ia->ekey_G_tl & FIELD_LENGTH_MASK);
+	
+	ia->ekey_P_tl = *eep_rec_p++;
+	BCOPY( eep_rec_p, (char *)&ia->ekey_P, 
+	       ia->ekey_P_tl & FIELD_LENGTH_MASK );
+	eep_rec_p += (ia->ekey_P_tl & FIELD_LENGTH_MASK);
+	
+	ia->ekey_Y_tl = *eep_rec_p++;
+	BCOPY( eep_rec_p, (char *)&ia->ekey_Y, 
+	       ia->ekey_Y_tl & FIELD_LENGTH_MASK );
+	eep_rec_p += (ia->ekey_Y_tl & FIELD_LENGTH_MASK);
+	
+	/* 
+	 * need to get a couple more fields if this is an I brick 
+	 */
+	if( ((unsigned char)*eep_rec_p != (unsigned char)EEPROM_EOF) ) {
+
+	    ia->mac_addr_tl = *eep_rec_p++;
+	    BCOPY( eep_rec_p, ia->mac_addr, 
+		   ia->mac_addr_tl & FIELD_LENGTH_MASK );
+	    eep_rec_p += (ia->mac_addr_tl & FIELD_LENGTH_MASK);
+	    
+	    ia->ieee1394_cfg_tl = *eep_rec_p++;
+	    BCOPY( eep_rec_p, ia->ieee1394_cfg,
+		   ia->ieee1394_cfg_tl & FIELD_LENGTH_MASK );
+	    
+	}
+    }
+
+    ia->checksum = eep_record[(ia->length * 8) - 1];
+
+    /* verify checksum */
+    eep_rec_p = eep_record;
+    checksum = 0;
+    for( i = 0; i < (8 * ia->length); i++ ) {
+	checksum += *eep_rec_p++;
+    }
+
+    if( (checksum & 0xff) != 0 )
+    {
+	db_printf(( "read_board_ia: bad checksum\n" ));
+	db_printf(( "read_board_ia: target 0x%x  uart 0x%x\n",
+		    sc->subch[subch].target, sc->uart ));
+	return EEP_BAD_CHECKSUM;
+    }
+
+    return EEP_OK;
+}
+
+
+int _cbrick_eeprom_read( eeprom_brd_record_t *buf, l1sc_t *scp,
+			 int component )
+{
+#if !defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+    return EEP_L1;
+#else
+    int r;
+    uint64_t uid = 0;
+    char uid_str[32];
+    int l1_compt, subch;
+
+    if ( IS_RUNNING_ON_SIMULATOR() )
+	return EEP_L1;
+
+    /* make sure we're targeting a cbrick */
+    if( !(component & C_BRICK) )
+	return EEP_PARAM;
+
+    /* If the promlog variable pointed to by IP27LOG_OVNIC is set,
+     * use that value for the cbrick UID rather than the EEPROM
+     * serial number.
+     */
+#ifdef LOG_GETENV
+    if( ip27log_getenv( scp->nasid, IP27LOG_OVNIC, uid_str, "0", 0 ) >= 0 )
+    {
+	db_printf(( "_cbrick_eeprom_read: "
+		    "Overriding UID with environment variable %s\n", 
+		    IP27LOG_OVNIC ));
+	uid = strtoull( uid_str, NULL, 0 );
+    }
+#endif
+
+    if( (subch = sc_open( scp, L1_ADDR_LOCAL )) < 0 )
+	return EEP_L1;
+
+    switch( component )
+    {
+      case C_BRICK:
+	/* c-brick motherboard */
+	l1_compt = L1_EEP_NODE;
+	r = read_chassis_ia( scp, subch, l1_compt, buf->chassis_ia );
+	if( r != EEP_OK ) {
+	    sc_close( scp, subch );
+	    db_printf(( "_cbrick_eeprom_read: using a fake eeprom record\n" ));
+	    return fake_an_eeprom_record( buf, component, uid );
+	}
+	if( uid ) {
+	    /* If IP27LOG_OVNIC is set, we want to put that value
+	     * in as our UID. */
+	    fake_a_serial_number( buf->chassis_ia->serial_num, uid );
+	    buf->chassis_ia->serial_num_tl = 6;
+	}
+	break;
+
+      case C_PIMM:
+	/* one of the PIMM boards */
+	l1_compt = L1_EEP_PIMM( component & COMPT_MASK );
+	break;
+
+      case C_DIMM:
+	/* one of the DIMMs */
+	l1_compt = L1_EEP_DIMM( component & COMPT_MASK );
+	r = read_spd( scp, subch, l1_compt, buf->spd );
+	sc_close( scp, subch );
+	return r;
+
+      default:
+	/* unsupported board type */
+	sc_close( scp, subch );
+	return EEP_PARAM;
+    }
+	      
+    r = read_board_ia( scp, subch, l1_compt, buf->board_ia );
+    sc_close( scp, subch );
+    if( r != EEP_OK ) 
+    {
+	db_printf(( "_cbrick_eeprom_read: using a fake eeprom record\n" ));
+	return fake_an_eeprom_record( buf, component, uid );
+    }
+    return EEP_OK;
+#endif /* CONFIG_SERIAL_SGI_L1_PROTOCOL */
+}
+
+
+int cbrick_eeprom_read( eeprom_brd_record_t *buf, nasid_t nasid,
+    		        int component )
+{
+#if !defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+    return EEP_L1;
+#else
+    l1sc_t *scp;
+    int local = (nasid == get_nasid());
+
+    if ( IS_RUNNING_ON_SIMULATOR() )
+	return EEP_L1;
+
+    /* If this brick is retrieving its own uid, use the local l1sc_t to
+     * arbitrate access to the l1; otherwise, set up a new one (prom) or
+     * use an existing remote l1sc_t (kernel)
+     */
+    if( local ) {
+	scp = get_l1sc();
+    }
+    else {
+	elsc_t *get_elsc(void);
+	scp = get_elsc();
+    }
+
+    return _cbrick_eeprom_read( buf, scp, component );
+#endif /* CONFIG_SERIAL_SGI_L1_PROTOCOL */
+}
+
+
+int iobrick_eeprom_read( eeprom_brd_record_t *buf, nasid_t nasid,
+			 int component )
+{
+#if !defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+    return EEP_L1;
+#else
+    int r;
+    int l1_compt, subch;
+    l1sc_t *scp;
+    int local = (nasid == get_nasid());
+
+    if ( IS_RUNNING_ON_SIMULATOR() )
+	return EEP_L1;
+
+    /* make sure we're talking to an applicable brick */
+    if( !(component & IO_BRICK) ) {
+	return EEP_PARAM;
+    }
+
+    /* If we're talking to this c-brick's attached io brick, use
+     * the local l1sc_t; otherwise, set up a new one (prom) or
+     * use an existing remote l1sc_t (kernel)
+     */
+    if( local ) {
+	scp = get_l1sc();
+    }
+    else {
+	elsc_t *get_elsc(void);
+	scp = get_elsc();
+    }
+
+    if( (subch = sc_open( scp, L1_ADDR_LOCALIO )) < 0 )
+	return EEP_L1;
+
+
+    switch( component )
+    {
+      case IO_BRICK:
+	/* IO brick motherboard */
+	l1_compt = L1_EEP_LOGIC;
+	r = read_chassis_ia( scp, subch, l1_compt, buf->chassis_ia );
+
+	if( r != EEP_OK ) {
+	    sc_close( scp, subch );
+#ifdef BRINGUP /* Once EEPROMs are universally available, remove this */
+	    r = fake_an_eeprom_record( buf, component, rtc_time() );
+#endif /* BRINGUP */
+	    return r;
+	}
+	break;
+
+      case IO_POWER:
+	/* IO brick power board */
+	l1_compt = L1_EEP_POWER;
+	break;
+
+      default:
+	/* unsupported board type */
+	sc_close( scp, subch );
+	return EEP_PARAM;
+    }
+
+    r = read_board_ia( scp, subch, l1_compt, buf->board_ia );
+    sc_close( scp, subch );
+    if( r != EEP_OK ) {
+	return r;
+    }
+    return EEP_OK;
+#endif /* CONFIG_SERIAL_SGI_L1_PROTOCOL */    
+}
+
+
+int vector_eeprom_read( eeprom_brd_record_t *buf, nasid_t nasid,
+			net_vec_t path, int component )
+{
+#if !defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+    return EEP_L1;
+#else
+    int r;
+    uint64_t uid = 0;
+    int l1_compt, subch;
+    l1sc_t sc;
+
+    if ( IS_RUNNING_ON_SIMULATOR() )
+	return EEP_L1;
+
+    /* make sure we're targeting an applicable brick */
+    if( !(component & VECTOR) )
+	return EEP_PARAM;
+
+    switch( component & BRICK_MASK )
+    {
+      case R_BRICK:
+	ROUTER_LOCK( path );
+	sc_init( &sc, nasid, path );
+
+	if( (subch = sc_open( &sc, L1_ADDR_LOCAL )) < 0 )
+	{
+	    db_printf(( "vector_eeprom_read: couldn't open subch\n" ));
+	    ROUTER_UNLOCK(path);
+	    return EEP_L1;
+	}
+	switch( component )
+	{
+	  case R_BRICK:
+	    /* r-brick motherboard */
+	    l1_compt = L1_EEP_LOGIC;
+    	    r = read_chassis_ia( &sc, subch, l1_compt, buf->chassis_ia );
+	    if( r != EEP_OK ) {
+		sc_close( &sc, subch );
+		ROUTER_UNLOCK( path );
+		printk( "vector_eeprom_read: couldn't get rbrick eeprom info;"
+			" using current time as uid\n" );
+		uid = rtc_time();
+		db_printf(("vector_eeprom_read: using a fake eeprom record\n"));
+		return fake_an_eeprom_record( buf, component, uid );
+	    }
+	    break;
+
+	  case R_POWER:
+	    /* r-brick power board */
+	    l1_compt = L1_EEP_POWER;
+	    break;
+
+	  default:
+	    /* unsupported board type */
+	    sc_close( &sc, subch );
+	    ROUTER_UNLOCK( path );
+	    return EEP_PARAM;
+	}
+	r = read_board_ia( &sc, subch, l1_compt, buf->board_ia );
+	sc_close( &sc, subch );
+	ROUTER_UNLOCK( path );
+	if( r != EEP_OK ) {
+	    db_printf(( "vector_eeprom_read: using a fake eeprom record\n" ));
+	    return fake_an_eeprom_record( buf, component, uid );
+	}
+	return EEP_OK;
+
+      case C_BRICK:
+	sc_init( &sc, nasid, path );
+	return _cbrick_eeprom_read( buf, &sc, component );
+
+      default:
+	/* unsupported brick type */
+	return EEP_PARAM;
+    }
+#endif /* CONFIG_SERIAL_SGI_L1_PROTOCOL */
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/hcl.c linux/arch/ia64/sn/io/hcl.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/hcl.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/hcl.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,1506 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ *  hcl - SGI's Hardware Graph compatibility layer.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+#include <linux/types.h>
+#include <linux/config.h>
+#include <linux/slab.h>
+#include <linux/ctype.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <asm/sn/sgi.h>
+#include <linux/devfs_fs.h>
+#include <linux/devfs_fs_kernel.h>
+#include <asm/io.h>
+#include <asm/sn/iograph.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/labelcl.h>
+
+#define HCL_NAME "SGI-HWGRAPH COMPATIBILITY DRIVER"
+#define HCL_TEMP_NAME "HCL_TEMP_NAME_USED_FOR_HWGRAPH_VERTEX_CREATE"
+#define HCL_TEMP_NAME_LEN 44 
+#define HCL_VERSION "1.0"
+devfs_handle_t hwgraph_root = NULL;
+
+/*
+ * Debug flag definition.
+ */
+#define OPTION_NONE             0x00
+#define HCL_DEBUG_NONE 0x00000
+#define HCL_DEBUG_ALL  0x0ffff
+#if defined(CONFIG_HCL_DEBUG)
+static unsigned int hcl_debug_init __initdata = HCL_DEBUG_NONE;
+#endif
+static unsigned int hcl_debug = HCL_DEBUG_NONE;
+static unsigned int boot_options = OPTION_NONE;
+
+/*
+ * Some Global definitions.
+ */
+spinlock_t hcl_spinlock;
+devfs_handle_t hcl_handle = NULL;
+
+/*
+ * HCL device driver.
+ * The purpose of this device driver is to provide a facility 
+ * for User Level Apps e.g. hinv, ioconfig etc. an ioctl path 
+ * to manipulate label entries without having to implement
+ * system call interfaces.  This methodology will enable us to 
+ * make this feature module loadable.
+ */
+static int hcl_open(struct inode * inode, struct file * filp)
+{
+	if (hcl_debug) {
+        	printk("HCL: hcl_open called.\n");
+	}
+
+        return(0);
+
+}
+
+static int hcl_close(struct inode * inode, struct file * filp)
+{
+
+	if (hcl_debug) {
+        	printk("HCL: hcl_close called.\n");
+	}
+
+        return(0);
+
+}
+
+static int hcl_ioctl(struct inode * inode, struct file * file,
+        unsigned int cmd, unsigned long arg)
+{
+
+	if (hcl_debug) {
+		printk("HCL: hcl_ioctl called.\n");
+	}
+
+	switch (cmd) {
+		default:
+			if (hcl_debug) {
+				printk("HCL: hcl_ioctl cmd = 0x%x\n", cmd);
+			}
+	}
+
+	return(0);
+
+}
+
+struct file_operations hcl_fops = {
+	NULL,		/* lseek - default */
+	NULL,		/* read - general block-dev read */
+	NULL,		/* write - general block-dev write */
+	NULL,		/* readdir - bad */
+	NULL,		/* poll */
+	hcl_ioctl,      /* ioctl */
+	NULL,		/* mmap */
+	hcl_open,	/* open */
+	NULL,		/* flush */
+	hcl_close,	/* release */
+	NULL,		/* fsync */
+	NULL,		/* fasync */
+	NULL,		/* check_media_change */
+	NULL,		/* revalidate */
+	NULL		/* lock */
+};
+
+
+/*
+ * init_hcl() - Boot time initialization.  Ensure that it is called 
+ *	after devfs has been initialized.
+ *
+ * For now this routine is being called out of devfs/base.c.  Actually 
+ * Not a bad place to be ..
+ *
+ */
+#ifdef MODULE
+int init_module (void)
+#else
+int __init init_hcl(void)
+#endif
+{
+	extern void string_table_init(struct string_table *);
+	extern struct string_table label_string_table;
+	int rv = 0;
+
+	printk ("\n%s: v%s Colin Ngam (cngam@sgi.com)\n",
+		HCL_NAME, HCL_VERSION);
+#if defined(CONFIG_HCL_DEBUG) && !defined(MODULE)
+	hcl_debug = hcl_debug_init;
+	printk ("%s: hcl_debug: 0x%0x\n", HCL_NAME, hcl_debug);
+#endif
+	printk ("\n%s: boot_options: 0x%0x\n", HCL_NAME, boot_options);
+	spin_lock_init(&hcl_spinlock);
+
+	/*
+	 * Create the hwgraph_root on devfs.
+	 */
+	rv = hwgraph_path_add(NULL, "hw", &hwgraph_root);
+	if (rv)
+		printk ("init_hcl: Failed to create hwgraph_root. Error = %d.\n", rv);
+
+	/*
+	 * Create the hcl driver to support inventory entry manipulations.
+	 * By default, it is expected that devfs is mounted on /dev.
+	 *
+	 */
+	hcl_handle = hwgraph_register(hwgraph_root, ".hcl",
+			0, DEVFS_FL_AUTO_DEVNUM,
+			0, 0,
+			S_IFCHR | S_IRUSR | S_IWUSR | S_IRGRP, 0, 0,
+			&hcl_fops, NULL);
+
+	if (hcl_handle == NULL) {
+		panic("HCL: Unable to create HCL Driver in init_hcl().\n");
+		return(0);
+	}
+
+	/*
+	 * Initialize the HCL string table.
+	 */
+	string_table_init(&label_string_table);
+
+	return(0);
+
+}
+
+
+/*
+ * hcl_setup() - Process boot time parameters if given.
+ *	"hcl="
+ *	This routine gets called only if "hcl=" is given in the 
+ *	boot line and before init_hcl().
+ *
+ *	We currently do not have any boot options .. when we do, 
+ *	functionalities can be added here.
+ *
+ */
+static int __init hcl_setup(char *str)
+{
+    while ( (*str != '\0') && !isspace (*str) )
+    {
+	printk("HCL: Boot time parameter %s\n", str);
+#ifdef CONFIG_HCL_DEBUG
+        if (strncmp (str, "all", 3) == 0) {
+            hcl_debug_init |= HCL_DEBUG_ALL;
+            str += 3;
+        } else 
+        	return 0;
+#endif
+        if (*str != ',') return 0;
+        ++str;
+    }
+
+    return 1;
+
+}
+
+__setup("hcl=", hcl_setup);
+
+
+/*
+ * Set device specific "fast information".
+ *
+ */
+void
+hwgraph_fastinfo_set(devfs_handle_t de, arbitrary_info_t fastinfo)
+{
+
+	if (hcl_debug) {
+		printk("HCL: hwgraph_fastinfo_set handle 0x%p fastinfo %ld\n",
+			de, fastinfo);
+	}
+		
+	labelcl_info_replace_IDX(de, HWGRAPH_FASTINFO, fastinfo, NULL);
+
+}
+
+
+/*
+ * Get device specific "fast information".
+ *
+ */
+arbitrary_info_t
+hwgraph_fastinfo_get(devfs_handle_t de)
+{
+	arbitrary_info_t fastinfo;
+	int rv;
+
+	if (!de) {
+		printk(KERN_WARNING "HCL: hwgraph_fastinfo_get handle given is NULL.\n");
+		return(-1);
+	}
+
+	rv = labelcl_info_get_IDX(de, HWGRAPH_FASTINFO, &fastinfo);
+	if (rv == 0)
+		return(fastinfo);
+
+	return(0);
+}
+
+
+/*
+ * hwgraph_connectpt_set - Sets the connect point handle in de to the 
+ *	given connect_de handle.  By default, the connect point of the 
+ *	devfs node is the parent.  This effectively changes this assumption.
+ */
+int
+hwgraph_connectpt_set(devfs_handle_t de, devfs_handle_t connect_de)
+{
+	int rv;
+
+	if (!de)
+		return(-1);
+
+	rv = labelcl_info_connectpt_set(de, connect_de);
+
+	return(rv);
+}
+
+
+/*
+ * hwgraph_connectpt_get: Returns the entry's connect point  in the devfs 
+ *	tree.
+ */
+devfs_handle_t
+hwgraph_connectpt_get(devfs_handle_t de)
+{
+	int rv;
+	arbitrary_info_t info;
+	devfs_handle_t connect;
+
+	rv = labelcl_info_get_IDX(de, HWGRAPH_CONNECTPT, &info);
+	if (rv != 0) {
+		return(NULL);
+	}
+
+	connect = (devfs_handle_t)info;
+	return(connect);
+
+}
+
+
+/*
+ * hwgraph_mk_dir - Creates a directory entry with devfs.
+ *	Note that a directory entry in devfs can have children 
+ *	but it cannot be a char|block special file.
+ */
+devfs_handle_t
+hwgraph_mk_dir(devfs_handle_t de, const char *name,
+                unsigned int namelen, void *info)
+{
+
+	int rv;
+	labelcl_info_t *labelcl_info = NULL;
+	devfs_handle_t new_devfs_handle = NULL;
+	devfs_handle_t parent = NULL;
+
+	/*
+	 * Create the device info structure for hwgraph compatiblity support.
+	 */
+	labelcl_info = labelcl_info_create();
+	if (!labelcl_info)
+		return(NULL);
+
+	/*
+	 * Create a devfs entry.
+	 */
+	new_devfs_handle = devfs_mk_dir(de, name, (void *)labelcl_info);
+	if (!new_devfs_handle) {
+		labelcl_info_destroy(labelcl_info);
+		return(NULL);
+	}
+
+	/*
+	 * Get the parent handle.
+	 */
+	parent = devfs_get_parent (new_devfs_handle);
+
+	/*
+	 * To provide the same semantics as the hwgraph, set the connect point.
+	 */
+	rv = hwgraph_connectpt_set(new_devfs_handle, parent);
+	if (!rv) {
+		/*
+		 * We need to clean up!
+		 */
+	}
+
+	/*
+	 * If the caller provides a private data pointer, save it in the 
+	 * labelcl info structure(fastinfo).  This can be retrieved via
+	 * hwgraph_fastinfo_get()
+	 */
+	if (info)
+		hwgraph_fastinfo_set(new_devfs_handle, (arbitrary_info_t)info);
+		
+	return(new_devfs_handle);
+
+}
+
+/*
+ * hwgraph_vertex_create - Create a vertex by giving it a temp name.
+ */
+
+/*
+ * hwgraph_path_add - Create a directory node with the given path starting 
+ * from the given devfs_handle_t.
+ */
+extern char * dev_to_name(devfs_handle_t, char *, uint);
+int
+hwgraph_path_add(devfs_handle_t  fromv,
+		 char *path,
+		 devfs_handle_t *new_de)
+{
+
+	unsigned int	namelen = strlen(path);
+	int		rv;
+
+	/*
+	 * We need to handle the case when fromv is NULL ..
+	 * in this case we need to create the path from the 
+	 * hwgraph root!
+	 */
+	if (fromv == NULL)
+		fromv = hwgraph_root;
+
+	/*
+	 * check the entry doesn't already exist, if it does
+	 * then we simply want new_de to point to it (otherwise
+	 * we'll overwrite the existing labelcl_info struct)
+	 */
+	rv = hwgraph_edge_get(fromv, path, new_de);
+	if (rv)	{	/* couldn't find entry so we create it */
+		*new_de = hwgraph_mk_dir(fromv, path, namelen, NULL);
+		if (new_de == NULL)
+			return(-1);
+		else
+			return(0);
+	}
+	else 
+ 		return(0);
+
+}
+
+/*
+ * hwgraph_register  - Creates a file entry with devfs.
+ *	Note that a file entry cannot have children .. it is like a 
+ *	char|block special vertex in hwgraph.
+ */
+devfs_handle_t
+hwgraph_register(devfs_handle_t de, const char *name,
+                unsigned int namelen, unsigned int flags, 
+		unsigned int major, unsigned int minor,
+                umode_t mode, uid_t uid, gid_t gid, 
+		struct file_operations *fops,
+                void *info)
+{
+
+	int rv;
+        void *labelcl_info = NULL;
+        devfs_handle_t new_devfs_handle = NULL;
+	devfs_handle_t parent = NULL;
+
+        /*
+         * Create the labelcl info structure for hwgraph compatiblity support.
+         */
+        labelcl_info = labelcl_info_create();
+        if (!labelcl_info)
+                return(NULL);
+
+        /*
+         * Create a devfs entry.
+         */
+        new_devfs_handle = devfs_register(de, name, flags, major,
+				minor, mode, fops, labelcl_info);
+        if (!new_devfs_handle) {
+                labelcl_info_destroy((labelcl_info_t *)labelcl_info);
+                return(NULL);
+        }
+
+	/*
+	 * Get the parent handle.
+	 */
+	if (de == NULL)
+		parent = devfs_get_parent (new_devfs_handle);
+	else
+		parent = de;
+		
+	/*
+	 * To provide the same semantics as the hwgraph, set the connect point.
+	 */
+	rv = hwgraph_connectpt_set(new_devfs_handle, parent);
+	if (rv) {
+		/*
+		 * We need to clean up!
+		 */
+		printk("HCL: Unable to set the connect point to it's parent 0x%p\n",
+			new_devfs_handle);
+	}
+
+        /*
+         * If the caller provides a private data pointer, save it in the 
+         * labelcl info structure(fastinfo).  This can be retrieved via
+         * hwgraph_fastinfo_get()
+         */
+        if (info)
+                hwgraph_fastinfo_set(new_devfs_handle, (arbitrary_info_t)info);
+
+        return(new_devfs_handle);
+
+}
+
+
+/*
+ * hwgraph_mk_symlink - Create a symbolic link.
+ */
+int
+hwgraph_mk_symlink(devfs_handle_t de, const char *name, unsigned int namelen,
+                unsigned int flags, const char *link, unsigned int linklen, 
+		devfs_handle_t *handle, void *info)
+{
+
+	void *labelcl_info = NULL;
+	int status = 0;
+	devfs_handle_t new_devfs_handle = NULL;
+
+	/*
+	 * Create the labelcl info structure for hwgraph compatiblity support.
+	 */
+	labelcl_info = labelcl_info_create();
+	if (!labelcl_info)
+		return(-1);
+
+	/*
+	 * Create a symbolic link devfs entry.
+	 */
+	status = devfs_mk_symlink(de, name, flags, link,
+				&new_devfs_handle, labelcl_info);
+	if ( (!new_devfs_handle) || (!status) ){
+		labelcl_info_destroy((labelcl_info_t *)labelcl_info);
+		return(-1);
+	}
+
+	/*
+	 * If the caller provides a private data pointer, save it in the 
+	 * labelcl info structure(fastinfo).  This can be retrieved via
+	 * hwgraph_fastinfo_get()
+	 */
+	if (info)
+		hwgraph_fastinfo_set(new_devfs_handle, (arbitrary_info_t)info);
+
+	*handle = new_devfs_handle;
+	return(0);
+
+}
+
+/*
+ * hwgraph_vertex_get_next - this routine returns the next sibbling for the 
+ *	device entry given in de.  If there are no more sibbling, NULL 
+ * 	is returned in next_sibbling.
+ *
+ *	Currently we do not have any protection against de being deleted 
+ *	while it's handle is being held.
+ */
+int
+hwgraph_vertex_get_next(devfs_handle_t *next_sibbling, devfs_handle_t *de)
+{
+	*next_sibbling = devfs_get_next_sibling (*de);
+
+	if (*next_sibbling != NULL)
+		*de = *next_sibbling;
+	return (0);
+}
+
+
+/*
+ * hwgraph_vertex_destroy - Destroy the devfs entry
+ */
+int
+hwgraph_vertex_destroy(devfs_handle_t de)
+{
+
+	void *labelcl_info = NULL;
+
+	labelcl_info = devfs_get_info(de);
+	devfs_unregister(de);
+
+	if (labelcl_info)
+		labelcl_info_destroy((labelcl_info_t *)labelcl_info);
+
+	return(0);
+}
+
+/*
+** See if a vertex has an outgoing edge with a specified name.
+** Vertices in the hwgraph *implicitly* contain these edges:
+**	"." 	refers to "current vertex"
+**	".." 	refers to "connect point vertex"
+**	"char"	refers to current vertex (character device access)
+**	"block"	refers to current vertex (block device access)
+*/
+
+/*
+ * hwgraph_edge_add - This routines has changed from the original conext.
+ * All it does now is to create a symbolic link from "from" to "to".
+ */
+/* ARGSUSED */
+int
+hwgraph_edge_add(devfs_handle_t from, devfs_handle_t to, char *name)
+{
+
+	char *path;
+	int name_start;
+	devfs_handle_t handle = NULL;
+	int rv;
+
+	path = kmalloc(1024, GFP_KERNEL);
+	name_start = devfs_generate_path (to, path, 1024);
+
+	/*
+	 * Otherwise, just create a symlink to the vertex.
+	 * In this case the vertex was previous created with a REAL pathname.
+	 */
+	rv = devfs_mk_symlink (from, (const char *)name, 
+			       DEVFS_FL_DEFAULT, (const char *)&path[name_start],
+			       &handle, NULL);
+
+	name_start = devfs_generate_path (handle, path, 1024);
+	return(rv);
+
+	
+}
+/* ARGSUSED */
+int
+hwgraph_edge_get(devfs_handle_t from, char *name, devfs_handle_t *toptr)
+{
+
+	int namelen = 0;
+	devfs_handle_t target_handle = NULL;
+
+	if (name == NULL)
+		return(-1);
+
+	if (toptr == NULL)
+		return(-1);
+
+	/*
+	 * If the name is "." just return the current devfs entry handle.
+	 */
+	if (!strcmp(name, HWGRAPH_EDGELBL_DOT)) {
+		if (toptr) {
+			*toptr = from;
+		}
+	} else if (!strcmp(name, HWGRAPH_EDGELBL_DOTDOT)) {
+		/*
+		 * Hmmm .. should we return the connect point or parent ..
+		 * see in hwgraph, the concept of parent is the connectpt!
+		 *
+		 * Maybe we should see whether the connectpt is set .. if 
+		 * not just return the parent!
+		 */
+		target_handle = hwgraph_connectpt_get(from);
+		if (target_handle) {
+			/*
+			 * Just return the connect point.
+			 */
+			*toptr = target_handle;
+			return(0);
+		}
+		target_handle = devfs_get_parent(from);
+		*toptr = target_handle;
+
+	} else {
+		/*
+		 * Call devfs to get the devfs entry.
+		 */
+		namelen = (int) strlen(name);
+		target_handle = devfs_find_handle (from, name, 0, 0,
+					0, 1); /* Yes traverse symbolic links */
+		if (target_handle == NULL)
+			return(-1);
+		else
+		*toptr = target_handle;
+	}
+
+	return(0);
+}
+
+
+/*
+ * hwgraph_edge_get_next - Retrieves the next sibbling given the current
+ *	entry number "placeptr".
+ *
+ * 	Allow the caller to retrieve walk through the sibblings of "source" 
+ * 	devfs_handle_t.  The implicit edges "." and ".." is returned first 
+ * 	followed by each of the real children.
+ *
+ *	We may end up returning garbage if another thread perform any deletion 
+ *	in this directory before "placeptr".
+ *
+ */
+/* ARGSUSED */
+int
+hwgraph_edge_get_next(devfs_handle_t source, char *name, devfs_handle_t *target,
+                              uint *placeptr)
+
+{
+
+        uint which_place;
+	unsigned int namelen = 0;
+	const char *tempname = NULL;
+
+        if (placeptr == NULL)
+                return(-1);
+
+        which_place = *placeptr;
+
+again:
+        if (which_place <= HWGRAPH_RESERVED_PLACES) {
+                if (which_place == EDGE_PLACE_WANT_CURRENT) {
+			/*
+			 * Looking for "."
+			 * Return the current devfs handle.
+			 */
+                        if (name != NULL)
+                                strcpy(name, HWGRAPH_EDGELBL_DOT);
+
+                        if (target != NULL) {
+                                *target = source; 
+				/* XXX should incr "source" ref count here if we
+				 * ever implement ref counts */
+                        }
+
+                } else if (which_place == EDGE_PLACE_WANT_CONNECTPT) {
+			/*
+			 * Looking for the connect point or parent.
+			 * If the connect point is set .. it returns the connect point.
+			 * Otherwise, it returns the parent .. will we support 
+			 * connect point?
+			 */
+                        devfs_handle_t connect_point = hwgraph_connectpt_get(source);
+
+                        if (connect_point == NULL) {
+				/*
+				 * No connectpoint set .. either the User
+				 * explicitly NULL it or this node was not 
+				 * created via hcl.
+				 */
+                                which_place++;
+                                goto again;
+                        }
+
+                        if (name != NULL)
+                                strcpy(name, HWGRAPH_EDGELBL_DOTDOT);
+
+                        if (target != NULL)
+                                *target = connect_point;
+
+                } else if (which_place == EDGE_PLACE_WANT_REAL_EDGES) {
+			/* 
+			 * return first "real" entry in directory, and increment
+			 * placeptr.  Next time around we should have 
+			 * which_place > HWGRAPH_RESERVED_EDGES so we'll fall through
+			 * this nested if block.
+			 */
+			*target = devfs_get_first_child(source);
+			if (*target && name) {
+				tempname = devfs_get_name(*target, &namelen);
+				if (tempname && namelen)
+					strcpy(name, tempname);
+			}
+					
+			*placeptr = which_place + 1;
+			return (0);
+                }
+
+                *placeptr = which_place+1;
+                return(0);
+        }
+
+	/*
+	 * walk linked list, (which_place - HWGRAPH_RESERVED_PLACES) times
+	 */
+	{
+		devfs_handle_t	curr;
+		int		i = 0;
+
+		for (curr=devfs_get_first_child(source), i= i+HWGRAPH_RESERVED_PLACES; 
+			curr!=NULL && i<which_place; 
+			curr=devfs_get_next_sibling(curr), i++)
+			;
+		*target = curr;
+		*placeptr = which_place + 1;
+		if (curr && name) {
+			tempname = devfs_get_name(*target, &namelen);
+			printk("hwgraph_edge_get_next: Component name = %s, length = %d\n", tempname, namelen);
+			if (tempname && namelen)
+				strcpy(name, tempname);
+		}
+	}
+	if (target == NULL)
+		return(-1);
+	else
+        	return(0);
+}
+
+/*
+ * hwgraph_info_add_LBL - Adds a new label for the device.  Mark the info_desc
+ *	of the label as INFO_DESC_PRIVATE and store the info in the label.
+ */
+/* ARGSUSED */
+int
+hwgraph_info_add_LBL(	devfs_handle_t de,
+			char *name,
+			arbitrary_info_t info)
+{
+	return(labelcl_info_add_LBL(de, name, INFO_DESC_PRIVATE, info));
+}
+
+/*
+ * hwgraph_info_remove_LBL - Remove the label entry for the device.
+ */
+/* ARGSUSED */
+int
+hwgraph_info_remove_LBL(	devfs_handle_t de,
+				char *name,
+				arbitrary_info_t *old_info)
+{
+	return(labelcl_info_remove_LBL(de, name, NULL, old_info));
+}
+
+/*
+ * hwgraph_info_replace_LBL - replaces an existing label with 
+ *	a new label info value.
+ */
+/* ARGSUSED */
+int
+hwgraph_info_replace_LBL(	devfs_handle_t de,
+				char *name,
+				arbitrary_info_t info,
+				arbitrary_info_t *old_info)
+{
+	return(labelcl_info_replace_LBL(de, name,
+			INFO_DESC_PRIVATE, info,
+			NULL, old_info));
+}
+/*
+ * hwgraph_info_get_LBL - Get and return the info value in the label of the 
+ * 	device.
+ */
+/* ARGSUSED */
+int
+hwgraph_info_get_LBL(	devfs_handle_t de,
+			char *name,
+			arbitrary_info_t *infop)
+{
+	return(labelcl_info_get_LBL(de, name, NULL, infop));
+}
+
+/*
+ * hwgraph_info_get_exported_LBL - Retrieve the info_desc and info pointer 
+ *	of the given label for the device.  The weird thing is that the label 
+ *	that matches the name is return irrespective of the info_desc value!
+ *	Do not understand why the word "exported" is used!
+ */
+/* ARGSUSED */
+int
+hwgraph_info_get_exported_LBL(	devfs_handle_t de,
+				char *name,
+				int *export_info,
+				arbitrary_info_t *infop)
+{
+	int rc;
+	arb_info_desc_t info_desc;
+
+	rc = labelcl_info_get_LBL(de, name, &info_desc, infop);
+	if (rc == 0)
+		*export_info = (int)info_desc;
+
+	return(rc);
+}
+
+/*
+ * hwgraph_info_get_next_LBL - Returns the next label info given the 
+ *	current label entry in place.
+ *
+ *	Once again this has no locking or reference count for protection.
+ *
+ */
+/* ARGSUSED */
+int
+hwgraph_info_get_next_LBL(	devfs_handle_t de,
+				char *buf,
+				arbitrary_info_t *infop,
+				labelcl_info_place_t *place)
+{
+	return(labelcl_info_get_next_LBL(de, buf, NULL, infop, place));
+}
+
+/*
+ * hwgraph_info_export_LBL - Retrieve the specified label entry and modify 
+ *	the info_desc field with the given value in nbytes.
+ */
+/* ARGSUSED */
+int
+hwgraph_info_export_LBL(devfs_handle_t de, char *name, int nbytes)
+{
+	arbitrary_info_t info;
+	int rc;
+
+	if (nbytes == 0)
+		nbytes = INFO_DESC_EXPORT;
+
+	if (nbytes < 0)
+		return(-1);
+
+	rc = labelcl_info_get_LBL(de, name, NULL, &info);
+	if (rc != 0)
+		return(rc);
+
+	rc = labelcl_info_replace_LBL(de, name,
+				nbytes, info, NULL, NULL);
+
+	return(rc);
+}
+
+/*
+ * hwgraph_info_unexport_LBL - Retrieve the given label entry and change the 
+ * label info_descr filed to INFO_DESC_PRIVATE.
+ */
+/* ARGSUSED */
+int
+hwgraph_info_unexport_LBL(devfs_handle_t de, char *name)
+{
+	arbitrary_info_t info;
+	int rc;
+
+	rc = labelcl_info_get_LBL(de, name, NULL, &info);
+	if (rc != 0)
+		return(rc);
+
+	rc = labelcl_info_replace_LBL(de, name,
+				INFO_DESC_PRIVATE, info, NULL, NULL);
+
+	return(rc);
+}
+
+/*
+ * hwgraph_path_lookup - return the handle for the given path.
+ *
+ */
+int
+hwgraph_path_lookup(	devfs_handle_t start_vertex_handle,
+			char *lookup_path,
+			devfs_handle_t *vertex_handle_ptr,
+			char **remainder)
+{
+	*vertex_handle_ptr = devfs_find_handle(start_vertex_handle,	/* start dir */
+					lookup_path,		/* path */
+					0,			/* major */
+					0,			/* minor */
+					0,			/* char | block */
+					1);			/* traverse symlinks */
+	if (*vertex_handle_ptr == NULL)
+		return(-1);
+	else
+		return(0);
+}
+
+/*
+ * hwgraph_traverse - Find and return the devfs handle starting from de.
+ *
+ */
+graph_error_t
+hwgraph_traverse(devfs_handle_t de, char *path, devfs_handle_t *found)
+{
+	/* 
+	 * get the directory entry (path should end in a directory)
+	 */
+
+	*found = devfs_find_handle(de,	/* start dir */
+			    path,	/* path */
+			    0,		/* major */
+			    0,		/* minor */
+			    0,		/* char | block */
+			    1);		/* traverse symlinks */
+	if (*found == NULL)
+		return(GRAPH_NOT_FOUND);
+	else
+		return(GRAPH_SUCCESS);
+}
+
+/*
+ * hwgraph_path_to_vertex - Return the devfs entry handle for the given 
+ *	pathname .. assume traverse symlinks too!.
+ */
+devfs_handle_t
+hwgraph_path_to_vertex(char *path)
+{
+	return(devfs_find_handle(NULL,	/* start dir */
+			path,		/* path */
+		    	0,		/* major */
+		    	0,		/* minor */
+		    	0,		/* char | block */
+		    	1));		/* traverse symlinks */
+}
+
+/*
+ * hwgraph_path_to_dev - Returns the devfs_handle_t of the given path ..
+ *	We only deal with devfs handle and not devfs_handle_t.
+*/
+devfs_handle_t
+hwgraph_path_to_dev(char *path)
+{
+	devfs_handle_t  de;
+
+	de = hwgraph_path_to_vertex(path);
+	return(de);
+}
+
+/*
+ * hwgraph_block_device_get - return the handle of the block device file.
+ *	The assumption here is that de is a directory.
+*/
+devfs_handle_t
+hwgraph_block_device_get(devfs_handle_t de)
+{
+	return(devfs_find_handle(de,		/* start dir */
+			"block",		/* path */
+		    	0,			/* major */
+		    	0,			/* minor */
+		    	DEVFS_SPECIAL_BLK,	/* char | block */
+		    	1));			/* traverse symlinks */
+}
+
+/*
+ * hwgraph_char_device_get - return the handle of the char device file.
+ *      The assumption here is that de is a directory.
+*/
+devfs_handle_t
+hwgraph_char_device_get(devfs_handle_t de)
+{
+	return(devfs_find_handle(de,		/* start dir */
+			"char",			/* path */
+		    	0,			/* major */
+		    	0,			/* minor */
+		    	DEVFS_SPECIAL_CHR,	/* char | block */
+		    	1));			/* traverse symlinks */
+}
+
+/*
+ * hwgraph_cdevsw_get - returns the fops of the given devfs entry.
+ */
+struct file_operations *
+hwgraph_cdevsw_get(devfs_handle_t de)
+{
+	return(devfs_get_ops(de));
+}
+
+/*
+ * hwgraph_bdevsw_get - returns the fops of the given devfs entry.
+*/
+struct file_operations *
+hwgraph_bdevsw_get(devfs_handle_t de)
+{
+	return(devfs_get_ops(de));
+}
+
+/*
+** Inventory is now associated with a vertex in the graph.  For items that
+** belong in the inventory but have no vertex 
+** (e.g. old non-graph-aware drivers), we create a bogus vertex under the 
+** INFO_LBL_INVENT name.
+**
+** For historical reasons, we prevent exact duplicate entries from being added
+** to a single vertex.
+*/
+
+/*
+ * hwgraph_inventory_add - Adds an inventory entry into de.
+ */
+int
+hwgraph_inventory_add(	devfs_handle_t de,
+			int class,
+			int type,
+			major_t controller,
+			minor_t unit,
+			int state)
+{
+	inventory_t *pinv = NULL, *old_pinv = NULL, *last_pinv = NULL;
+	int rv;
+
+	/*
+	 * Add our inventory data to the list of inventory data
+	 * associated with this vertex.
+	 */
+again:
+	/* GRAPH_LOCK_UPDATE(&invent_lock); */
+	rv = labelcl_info_get_LBL(de,
+			INFO_LBL_INVENT,
+			NULL, (arbitrary_info_t *)&old_pinv);
+	if ((rv != LABELCL_SUCCESS) && (rv != LABELCL_NOT_FOUND))
+		goto failure;
+
+	/*
+	 * Seek to end of inventory items associated with this
+	 * vertex.  Along the way, make sure we're not duplicating
+	 * an inventory item (for compatibility with old add_to_inventory)
+	 */
+	for (;old_pinv; last_pinv = old_pinv, old_pinv = old_pinv->inv_next) {
+		if ((int)class != -1 && old_pinv->inv_class != class)
+			continue;
+		if ((int)type != -1 && old_pinv->inv_type != type)
+			continue;
+		if ((int)state != -1 && old_pinv->inv_state != state)
+			continue;
+		if ((int)controller != -1
+		    && old_pinv->inv_controller != controller)
+			continue;
+		if ((int)unit != -1 && old_pinv->inv_unit != unit)
+			continue;
+
+		/* exact duplicate of previously-added inventory item */
+		rv = LABELCL_DUP;
+		goto failure;
+	}
+
+	/* Not a duplicate, so we know that we need to add something. */
+	if (pinv == NULL) {
+		/* Release lock while we wait for memory. */
+		/* GRAPH_LOCK_DONE_UPDATE(&invent_lock); */
+		pinv = (inventory_t *)kmalloc(sizeof(inventory_t), GFP_KERNEL);
+		replace_in_inventory(pinv, class, type, controller, unit, state);
+		goto again;
+	}
+
+	pinv->inv_next = NULL;
+	if (last_pinv) {
+		last_pinv->inv_next = pinv;
+	} else {
+		rv = labelcl_info_add_LBL(de, INFO_LBL_INVENT, 
+			sizeof(inventory_t), (arbitrary_info_t)pinv);
+
+		if (!rv)
+			goto failure;
+	}
+
+	/* GRAPH_LOCK_DONE_UPDATE(&invent_lock); */
+	return(0);
+
+failure:
+	/* GRAPH_LOCK_DONE_UPDATE(&invent_lock); */
+	if (pinv)
+		kfree(pinv);
+	return(rv);
+}
+
+
+/*
+ * hwgraph_inventory_remove - Removes an inventory entry.
+ *
+ *	Remove an inventory item associated with a vertex.   It is the caller's
+ *	responsibility to make sure that there are no races between removing
+ *	inventory from a vertex and simultaneously removing that vertex.
+*/
+int
+hwgraph_inventory_remove(	devfs_handle_t de,
+				int class,
+				int type,
+				major_t controller,
+				minor_t unit,
+				int state)
+{
+	inventory_t *pinv = NULL, *last_pinv = NULL, *next_pinv = NULL;
+	labelcl_error_t rv;
+
+	/*
+	 * We never remove stuff from ".invent" ..
+	 */
+	if (!de)
+		return (-1);
+
+	/*
+	 * Remove our inventory data to the list of inventory data
+	 * associated with this vertex.
+	 */
+	/* GRAPH_LOCK_UPDATE(&invent_lock); */
+	rv = labelcl_info_get_LBL(de,
+			INFO_LBL_INVENT,
+			NULL, (arbitrary_info_t *)&pinv);
+	if (rv != LABELCL_SUCCESS)
+		goto failure;
+
+	/*
+	 * Search through inventory items associated with this
+	 * vertex, looking for a match.
+	 */
+	for (;pinv; pinv = next_pinv) {
+		next_pinv = pinv->inv_next;
+
+		if(((int)class == -1 || pinv->inv_class == class) &&
+		   ((int)type == -1 || pinv->inv_type == type) &&
+		   ((int)state == -1 || pinv->inv_state == state) &&
+		   ((int)controller == -1 || pinv->inv_controller == controller) &&
+		   ((int)unit == -1 || pinv->inv_unit == unit)) {
+
+			/* Found a matching inventory item. Remove it. */
+			if (last_pinv) {
+				last_pinv->inv_next = pinv->inv_next;
+			} else {
+				rv = hwgraph_info_replace_LBL(de, INFO_LBL_INVENT, (arbitrary_info_t)pinv->inv_next, NULL);
+				if (rv != LABELCL_SUCCESS)
+					goto failure;
+			}
+
+			pinv->inv_next = NULL; /* sanity */
+			kfree(pinv);
+		} else
+			last_pinv = pinv;
+	}
+
+	if (last_pinv == NULL) {
+		rv = hwgraph_info_remove_LBL(de, INFO_LBL_INVENT, NULL);
+		if (rv != LABELCL_SUCCESS)
+			goto failure;
+	}
+
+	rv = LABELCL_SUCCESS;
+
+failure:
+	/* GRAPH_LOCK_DONE_UPDATE(&invent_lock); */
+	return(rv);
+}
+
+/*
+ * hwgraph_inventory_get_next - Get next inventory item associated with the 
+ *	specified vertex.
+ *
+ *	No locking is really needed.  We don't yet have the ability
+ *	to remove inventory items, and new items are always added to
+ *	the end of a vertex' inventory list.
+ *
+ * 	However, a devfs entry can be removed!
+*/
+int
+hwgraph_inventory_get_next(devfs_handle_t de, invplace_t *place, inventory_t **ppinv)
+{
+	inventory_t *pinv;
+	labelcl_error_t rv;
+
+	if (de == NULL)
+		return(LABELCL_BAD_PARAM);
+
+	if (place->invplace_vhdl == NULL) {
+		place->invplace_vhdl = de;
+		place->invplace_inv = NULL;
+	}
+
+	if (de != place->invplace_vhdl)
+		return(LABELCL_BAD_PARAM);
+
+	if (place->invplace_inv == NULL) {
+		/* Just starting on this vertex */
+		rv = labelcl_info_get_LBL(de, INFO_LBL_INVENT,
+						NULL, (arbitrary_info_t *)&pinv);
+		if (rv != LABELCL_SUCCESS)
+			return(LABELCL_NOT_FOUND);
+
+	} else {
+		/* Advance to next item on this vertex */
+		pinv = place->invplace_inv->inv_next;
+	}
+	place->invplace_inv = pinv;
+	*ppinv = pinv;
+
+	return(LABELCL_SUCCESS);
+}
+
+/*
+ * hwgraph_controller_num_get - Returns the controller number in the inventory 
+ *	entry.
+ */
+int
+hwgraph_controller_num_get(devfs_handle_t device)
+{
+	inventory_t *pinv;
+	invplace_t invplace = { NULL, NULL, NULL };
+	int val = -1;
+	if ((pinv = device_inventory_get_next(device, &invplace)) != NULL) {
+		val = (pinv->inv_class == INV_NETWORK)? pinv->inv_unit: pinv->inv_controller;
+	}
+#ifdef DEBUG
+	/*
+	 * It does not make any sense to call this on vertexes with multiple
+	 * inventory structs chained together
+	 */
+	if ( device_inventory_get_next(device, &invplace) != NULL ) {
+		printk("Should panic here ... !\n");
+#endif
+	return (val);	
+}
+
+/*
+ * hwgraph_controller_num_set - Sets the controller number in the inventory 
+ *	entry.
+ */
+void
+hwgraph_controller_num_set(devfs_handle_t device, int contr_num)
+{
+	inventory_t *pinv;
+	invplace_t invplace = { NULL, NULL, NULL };
+	if ((pinv = device_inventory_get_next(device, &invplace)) != NULL) {
+		if (pinv->inv_class == INV_NETWORK)
+			pinv->inv_unit = contr_num;
+		else {
+			if (pinv->inv_class == INV_FCNODE)
+				pinv = device_inventory_get_next(device, &invplace);
+			if (pinv != NULL)
+				pinv->inv_controller = contr_num;
+		}
+	}
+#ifdef DEBUG
+	/*
+	 * It does not make any sense to call this on vertexes with multiple
+	 * inventory structs chained together
+	 */
+	if(pinv != NULL)
+		ASSERT(device_inventory_get_next(device, &invplace) == NULL);
+#endif
+}
+
+/*
+ * Find the canonical name for a given vertex by walking back through
+ * connectpt's until we hit the hwgraph root vertex (or until we run
+ * out of buffer space or until something goes wrong).
+ *
+ *	COMPATIBILITY FUNCTIONALITY
+ * Walks back through 'parents', not necessarily the same as connectpts.
+ *
+ * Need to resolve the fact that devfs does not return the path from 
+ * "/" but rather it just stops right before /dev ..
+ */
+int
+hwgraph_vertex_name_get(devfs_handle_t vhdl, char *buf, uint buflen)
+{
+	char *locbuf;
+	int   pos;
+
+	if (buflen < 1)
+		return(-1);	/* XXX should be GRAPH_BAD_PARAM ? */
+
+	locbuf = kmalloc(buflen, GFP_KERNEL);
+
+	pos = devfs_generate_path(vhdl, locbuf, buflen);
+	if (pos < 0) {
+		kfree(locbuf);
+		return pos;
+	}
+
+	strcpy(buf, &locbuf[pos]);
+	kfree(locbuf);
+	return 0;
+}
+
+/*
+** vertex_to_name converts a vertex into a canonical name by walking
+** back through connect points until we hit the hwgraph root (or until
+** we run out of buffer space).
+**
+** Usually returns a pointer to the original buffer, filled in as
+** appropriate.  If the buffer is too small to hold the entire name,
+** or if anything goes wrong while determining the name, vertex_to_name
+** returns "UnknownDevice".
+*/
+
+#define DEVNAME_UNKNOWN "UnknownDevice"
+
+char *
+vertex_to_name(devfs_handle_t vhdl, char *buf, uint buflen)
+{
+	if (hwgraph_vertex_name_get(vhdl, buf, buflen) == GRAPH_SUCCESS)
+		return(buf);
+	else
+		return(DEVNAME_UNKNOWN);
+}
+
+#ifdef IRIX
+/*
+** Return the compact node id of the node that ultimately "owns" the specified
+** vertex.  In order to do this, we walk back through masters and connect points
+** until we reach a vertex that represents a node.
+*/
+cnodeid_t
+master_node_get(devfs_handle_t vhdl)
+{
+	cnodeid_t cnodeid;
+	devfs_handle_t master;
+
+	for (;;) {
+		cnodeid = nodevertex_to_cnodeid(vhdl);
+		if (cnodeid != CNODEID_NONE)
+			return(cnodeid);
+
+		master = device_master_get(vhdl);
+
+		/* Check for exceptional cases */
+		if (master == vhdl) {
+			/* Since we got a reference to the "master" thru
+			 * device_master_get() we should decrement
+			 * its reference count by 1
+			 */
+			hwgraph_vertex_unref(master);
+			return(CNODEID_NONE);
+		}
+
+		if (master == GRAPH_VERTEX_NONE) {
+			master = hwgraph_connectpt_get(vhdl);
+			if ((master == GRAPH_VERTEX_NONE) ||
+			    (master == vhdl)) {
+				if (master == vhdl)
+					/* Since we got a reference to the
+					 * "master" thru
+					 * hwgraph_connectpt_get() we should
+					 * decrement its reference count by 1
+					 */
+					hwgraph_vertex_unref(master);
+				return(CNODEID_NONE);
+			}
+		}
+		
+		vhdl = master;
+		/* Decrement the reference to "master" which was got
+		 * either thru device_master_get() or hwgraph_connectpt_get()
+		 * above.
+		 */
+		hwgraph_vertex_unref(master);
+	}
+}
+
+/*
+ * Using the canonical path name to get hold of the desired vertex handle will
+ * not work on multi-hub sn0 nodes. Hence, we use the following (slightly
+ * convoluted) algorithm.
+ *
+ * - Start at the vertex corresponding to the driver (provided as input parameter)
+ * - Loop till you reach a vertex which has EDGE_LBL_MEMORY
+ *    - If EDGE_LBL_CONN exists, follow that up.
+ *      else if EDGE_LBL_MASTER exists, follow that up.
+ *      else follow EDGE_LBL_DOTDOT up.
+ *
+ * * We should be at desired hub/heart vertex now *
+ * - Follow EDGE_LBL_CONN to the widget vertex.
+ *
+ * - return vertex handle of this widget.
+ */
+devfs_handle_t
+mem_vhdl_get(devfs_handle_t drv_vhdl)
+{
+devfs_handle_t cur_vhdl, cur_upper_vhdl;
+devfs_handle_t tmp_mem_vhdl, mem_vhdl;
+graph_error_t loop_rv;
+
+  /* Initializations */
+  cur_vhdl = drv_vhdl;
+  loop_rv = ~GRAPH_SUCCESS;
+
+  /* Loop till current vertex has EDGE_LBL_MEMORY */
+  while (loop_rv != GRAPH_SUCCESS) {
+
+    if ((hwgraph_edge_get(cur_vhdl, EDGE_LBL_CONN, &cur_upper_vhdl)) == GRAPH_SUCCESS) {
+
+    } else if ((hwgraph_edge_get(cur_vhdl, EDGE_LBL_MASTER, &cur_upper_vhdl)) == GRAPH_SUCCESS) {
+      } else { /* Follow HWGRAPH_EDGELBL_DOTDOT up */
+           (void) hwgraph_edge_get(cur_vhdl, HWGRAPH_EDGELBL_DOTDOT, &cur_upper_vhdl);
+        }
+
+    cur_vhdl = cur_upper_vhdl;
+
+#if DEBUG && HWG_DEBUG
+    printf("Current vhdl %d \n", cur_vhdl);
+#endif /* DEBUG */
+
+    loop_rv = hwgraph_edge_get(cur_vhdl, EDGE_LBL_MEMORY, &tmp_mem_vhdl);
+  }
+
+  /* We should be at desired hub/heart vertex now */
+  if ((hwgraph_edge_get(cur_vhdl, EDGE_LBL_CONN, &mem_vhdl)) != GRAPH_SUCCESS)
+    return (GRAPH_VERTEX_NONE);
+
+  return (mem_vhdl);
+}
+#endif /* IRIX */
+
+
+/*
+** Add a char device -- if the driver supports it -- at a specified vertex.
+*/
+graph_error_t
+hwgraph_char_device_add(        devfs_handle_t from,
+                                char *path,
+                                char *prefix,
+                                devfs_handle_t *devhdl)
+{
+	devfs_handle_t xx = NULL;
+
+	printk("FIXME: hwgraph_char_device_add() called. Use hwgraph_register.\n");
+	*devhdl = xx;	// Must set devhdl
+	return(GRAPH_SUCCESS);
+}
+
+graph_error_t
+hwgraph_edge_remove(devfs_handle_t from, char *name, devfs_handle_t *toptr)
+{
+	printk("FIXME: hwgraph_edge_remove\n");
+	return(GRAPH_ILLEGAL_REQUEST);
+}
+
+graph_error_t
+hwgraph_vertex_unref(devfs_handle_t vhdl)
+{
+	printk("FIXME: hwgraph_vertex_unref\n");
+	return(GRAPH_ILLEGAL_REQUEST);
+}
+
+
+EXPORT_SYMBOL(hwgraph_mk_dir);
+EXPORT_SYMBOL(hwgraph_path_add);
+EXPORT_SYMBOL(hwgraph_char_device_add);
+EXPORT_SYMBOL(hwgraph_register);
+EXPORT_SYMBOL(hwgraph_vertex_destroy);
+
+EXPORT_SYMBOL(hwgraph_fastinfo_get);
+EXPORT_SYMBOL(hwgraph_edge_get);
+
+EXPORT_SYMBOL(hwgraph_fastinfo_set);
+EXPORT_SYMBOL(hwgraph_connectpt_set);
+EXPORT_SYMBOL(hwgraph_connectpt_get);
+EXPORT_SYMBOL(hwgraph_edge_get_next);
+EXPORT_SYMBOL(hwgraph_info_add_LBL);
+EXPORT_SYMBOL(hwgraph_info_remove_LBL);
+EXPORT_SYMBOL(hwgraph_info_replace_LBL);
+EXPORT_SYMBOL(hwgraph_info_get_LBL);
+EXPORT_SYMBOL(hwgraph_info_get_exported_LBL);
+EXPORT_SYMBOL(hwgraph_info_get_next_LBL);
+EXPORT_SYMBOL(hwgraph_info_export_LBL);
+EXPORT_SYMBOL(hwgraph_info_unexport_LBL);
+EXPORT_SYMBOL(hwgraph_path_lookup);
+EXPORT_SYMBOL(hwgraph_traverse);
+EXPORT_SYMBOL(hwgraph_path_to_vertex);
+EXPORT_SYMBOL(hwgraph_path_to_dev);
+EXPORT_SYMBOL(hwgraph_block_device_get);
+EXPORT_SYMBOL(hwgraph_char_device_get);
+EXPORT_SYMBOL(hwgraph_cdevsw_get);
+EXPORT_SYMBOL(hwgraph_bdevsw_get);
+EXPORT_SYMBOL(hwgraph_vertex_name_get);
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/hcl_util.c linux/arch/ia64/sn/io/hcl_util.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/hcl_util.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/hcl_util.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,160 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+#include <linux/types.h>
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/devfs_fs.h>
+#include <linux/devfs_fs_kernel.h>
+#include <asm/sn/sgi.h>
+#include <asm/io.h>
+#include <asm/sn/iograph.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/labelcl.h>
+#include <asm/sn/hcl_util.h>
+#include <asm/sn/nodepda.h>
+
+static devfs_handle_t hwgraph_all_cnodes = GRAPH_VERTEX_NONE;
+extern devfs_handle_t hwgraph_root;
+
+
+/*
+** Return the "master" for a given vertex.  A master vertex is a
+** controller or adapter or other piece of hardware that the given
+** vertex passes through on the way to the rest of the system.
+*/
+devfs_handle_t
+device_master_get(devfs_handle_t vhdl)
+{
+	graph_error_t rc;
+	devfs_handle_t master;
+
+	rc = hwgraph_edge_get(vhdl, EDGE_LBL_MASTER, &master);
+	if (rc == GRAPH_SUCCESS)
+		return(master);
+	else
+		return(GRAPH_VERTEX_NONE);
+}
+
+/*
+** Set the master for a given vertex.
+** Returns 0 on success, non-0 indicates failure
+*/
+int
+device_master_set(devfs_handle_t vhdl, devfs_handle_t master)
+{
+	graph_error_t rc;
+
+	rc = hwgraph_edge_add(vhdl, master, EDGE_LBL_MASTER);
+	return(rc != GRAPH_SUCCESS);
+}
+
+
+/*
+** Return the compact node id of the node that ultimately "owns" the specified
+** vertex.  In order to do this, we walk back through masters and connect points
+** until we reach a vertex that represents a node.
+*/
+cnodeid_t
+master_node_get(devfs_handle_t vhdl)
+{
+	cnodeid_t cnodeid;
+	devfs_handle_t master;
+
+	for (;;) {
+		cnodeid = nodevertex_to_cnodeid(vhdl);
+		if (cnodeid != CNODEID_NONE)
+			return(cnodeid);
+
+		master = device_master_get(vhdl);
+
+		/* Check for exceptional cases */
+		if (master == vhdl) {
+			/* Since we got a reference to the "master" thru
+			 * device_master_get() we should decrement
+			 * its reference count by 1
+			 */
+			return(CNODEID_NONE);
+		}
+
+		if (master == GRAPH_VERTEX_NONE) {
+			master = hwgraph_connectpt_get(vhdl);
+			if ((master == GRAPH_VERTEX_NONE) ||
+			    (master == vhdl)) {
+				return(CNODEID_NONE);
+			}
+		}
+
+		vhdl = master;
+	}
+}
+
+/*
+** If the specified device represents a node, return its
+** compact node ID; otherwise, return CNODEID_NONE.
+*/
+cnodeid_t
+nodevertex_to_cnodeid(devfs_handle_t vhdl)
+{
+	int rv = 0;
+	arbitrary_info_t cnodeid = CNODEID_NONE;
+
+	rv = labelcl_info_get_LBL(vhdl, INFO_LBL_CNODEID, NULL, &cnodeid);
+
+	return((cnodeid_t)cnodeid);
+}
+
+void
+mark_nodevertex_as_node(devfs_handle_t vhdl, cnodeid_t cnodeid)
+{
+	if (cnodeid == CNODEID_NONE)
+		return;
+
+	cnodeid_to_vertex(cnodeid) = vhdl;
+	labelcl_info_add_LBL(vhdl, INFO_LBL_CNODEID, INFO_DESC_EXPORT, 
+		(arbitrary_info_t)cnodeid);
+
+	{
+		char cnodeid_buffer[10];
+
+		if (hwgraph_all_cnodes == GRAPH_VERTEX_NONE) {
+			(void)hwgraph_path_add( hwgraph_root,
+						EDGE_LBL_NODENUM,
+						&hwgraph_all_cnodes);
+		}
+
+		sprintf(cnodeid_buffer, "%d", cnodeid);
+		(void)hwgraph_edge_add( hwgraph_all_cnodes,
+					vhdl,
+					cnodeid_buffer);
+	}
+}
+
+
+/*
+** dev_to_name converts a devfs_handle_t into a canonical name.  If the devfs_handle_t
+** represents a vertex in the hardware graph, it is converted in the
+** normal way for vertices.  If the devfs_handle_t is an old devfs_handle_t (one which
+** does not represent a hwgraph vertex), we synthesize a name based
+** on major/minor number.
+**
+** Usually returns a pointer to the original buffer, filled in as
+** appropriate.  If the buffer is too small to hold the entire name,
+** or if anything goes wrong while determining the name, dev_to_name
+** returns "UnknownDevice".
+*/
+char *
+dev_to_name(devfs_handle_t dev, char *buf, uint buflen)
+{
+        return(vertex_to_name(dev, buf, buflen));
+}
+
+
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/hubdev.c linux/arch/ia64/sn/io/hubdev.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/hubdev.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/hubdev.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,127 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+#include <linux/types.h>
+#include <linux/config.h>
+#include <linux/slab.h>
+#include <asm/sn/sgi.h>
+#include <asm/sn/iograph.h>
+#include <asm/sn/sn1/hubdev.h>
+#include <asm/sn/sn_private.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/labelcl.h>
+
+struct hubdev_callout {
+        int (*attach_method)(devfs_handle_t);
+        struct hubdev_callout *fp;
+};
+
+typedef struct hubdev_callout hubdev_callout_t;
+
+mutex_t hubdev_callout_mutex;
+hubdev_callout_t *hubdev_callout_list = NULL;
+
+void
+hubdev_init(void)
+{
+	mutex_init(&hubdev_callout_mutex, MUTEX_DEFAULT, "hubdev");
+        hubdev_callout_list = NULL;
+}
+        
+void
+hubdev_register(int (*attach_method)(devfs_handle_t))
+{
+        hubdev_callout_t *callout;
+        
+        ASSERT(attach_method);
+
+        callout =  (hubdev_callout_t *)kmem_zalloc(sizeof(hubdev_callout_t), KM_SLEEP);
+        ASSERT(callout);
+        
+	mutex_lock(&hubdev_callout_mutex, PZERO);
+        /*
+         * Insert at the front of the list
+         */
+        callout->fp = hubdev_callout_list;
+        hubdev_callout_list = callout;
+        callout->attach_method = attach_method;
+	mutex_unlock(&hubdev_callout_mutex);
+}
+
+int
+hubdev_unregister(int (*attach_method)(devfs_handle_t))
+{
+        hubdev_callout_t **p;
+        
+        ASSERT(attach_method);
+   
+	mutex_lock(&hubdev_callout_mutex, PZERO);
+        /*
+         * Remove registry element containing attach_method
+         */
+        for (p = &hubdev_callout_list; *p != NULL; p = &(*p)->fp) {
+                if ((*p)->attach_method == attach_method) {
+                        hubdev_callout_t* victim = *p;
+                        *p = (*p)->fp;
+                        kfree(victim);
+                        mutex_unlock(&hubdev_callout_mutex);
+                        return (0);
+                }
+        }
+        mutex_unlock(&hubdev_callout_mutex);
+        return (ENOENT);
+}
+
+
+int
+hubdev_docallouts(devfs_handle_t hub)
+{
+        hubdev_callout_t *p;
+        int errcode;
+
+	mutex_lock(&hubdev_callout_mutex, PZERO);
+        
+        for (p = hubdev_callout_list; p != NULL; p = p->fp) {
+                ASSERT(p->attach_method);
+                errcode = (*p->attach_method)(hub);
+                if (errcode != 0) {
+			mutex_unlock(&hubdev_callout_mutex);
+                        return (errcode);
+                }
+        }
+        mutex_unlock(&hubdev_callout_mutex);
+        return (0);
+}
+
+/*
+ * Given a hub vertex, return the base address of the Hspec space
+ * for that hub.
+ */
+caddr_t
+hubdev_prombase_get(devfs_handle_t hub)
+{
+	hubinfo_t	hinfo = NULL;
+
+	hubinfo_get(hub, &hinfo);
+	ASSERT(hinfo);
+
+	return ((caddr_t)NODE_RBOOT_BASE(hinfo->h_nasid));
+}
+
+cnodeid_t
+hubdev_cnodeid_get(devfs_handle_t hub)
+{
+	hubinfo_t	hinfo = NULL;
+	hubinfo_get(hub, &hinfo);
+	ASSERT(hinfo);
+
+	return hinfo->h_cnodeid;
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/hubspc.c linux/arch/ia64/sn/io/hubspc.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/hubspc.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/hubspc.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,447 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+/*
+ * hubspc.c - Hub Memory Space Management Driver
+ * This driver implements the managers for the following
+ * memory resources:
+ * 1) reference counters
+ */
+
+#include <linux/types.h>
+#include <linux/config.h>
+#include <linux/slab.h>
+#include <asm/sn/sgi.h>
+#include <linux/devfs_fs.h>
+#include <linux/devfs_fs_kernel.h>
+#include <asm/io.h>
+#include <asm/sn/iograph.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/labelcl.h>
+#include <asm/sn/mem_refcnt.h>
+#include <asm/sn/agent.h>
+#include <asm/sn/addrs.h>
+
+
+#if defined(CONFIG_SGI_IP35) || defined(CONFIG_IA64_SGI_SN1) || defined(CONFIG_IA64_GENERIC)
+#include <asm/sn/sn1/ip27config.h>
+#include <asm/sn/sn1/hubdev.h>
+#include <asm/sn/ksys/elsc.h>
+#endif
+
+#include <asm/sn/hubspc.h>
+
+
+/* Uncomment the following line for tracing */
+/* #define HUBSPC_DEBUG 1 */
+
+int hubspc_devflag = D_MP;
+
+extern void *device_info_get(devfs_handle_t device);
+extern void device_info_set(devfs_handle_t device, void *info);
+
+
+
+/***********************************************************************/
+/* CPU Prom Space 						       */
+/***********************************************************************/
+
+typedef struct cpuprom_info {
+	devfs_handle_t	prom_dev;
+	devfs_handle_t	nodevrtx;
+	struct	cpuprom_info *next;
+}cpuprom_info_t;
+
+static cpuprom_info_t	*cpuprom_head;
+lock_t	cpuprom_spinlock;
+#define	PROM_LOCK()	mutex_spinlock(&cpuprom_spinlock)
+#define	PROM_UNLOCK(s)	mutex_spinunlock(&cpuprom_spinlock, (s))
+
+/*
+ * Add prominfo to the linked list maintained.
+ */
+void
+prominfo_add(devfs_handle_t hub, devfs_handle_t prom)
+{
+	cpuprom_info_t	*info;
+	int	s;
+
+	info = kmalloc(sizeof(cpuprom_info_t), GFP_KERNEL);
+	ASSERT(info);
+	info->prom_dev = prom;
+	info->nodevrtx = hub;
+
+
+	s = PROM_LOCK();
+	info->next = cpuprom_head;
+	cpuprom_head = info;
+	PROM_UNLOCK(s);
+}
+
+void
+prominfo_del(devfs_handle_t prom)
+{
+	int	s;
+	cpuprom_info_t	*info;
+	cpuprom_info_t	**prev;
+
+	s = PROM_LOCK();
+	prev = &cpuprom_head;
+	while ( (info = *prev) ) {
+		if (info->prom_dev == prom) {
+			*prev = info->next;
+			PROM_UNLOCK(s);
+			return;
+		}
+		
+		prev = &info->next;
+	}
+	PROM_UNLOCK(s);
+	ASSERT(0);
+}
+
+devfs_handle_t
+prominfo_nodeget(devfs_handle_t prom)
+{
+	int	s;
+	cpuprom_info_t	*info;
+
+	s = PROM_LOCK();
+	info = cpuprom_head;
+	while (info) {
+		if(info->prom_dev == prom) {
+			PROM_UNLOCK(s);
+			return info->nodevrtx;
+		}
+		info = info->next;
+	}
+	PROM_UNLOCK(s);
+	return 0;
+}
+
+#if defined(CONFIG_SGI_IP35) || defined(CONFIG_IA64_SGI_SN1) || defined(CONFIG_IA64_GENERIC)
+#define	SN_PROMVERSION		INV_IP35PROM
+#endif
+
+/* Add "detailed" labelled inventory information to the
+ * prom vertex 
+ */
+void
+cpuprom_detailed_inventory_info_add(devfs_handle_t prom_dev,devfs_handle_t node)
+{
+	invent_miscinfo_t 	*cpuprom_inventory_info;
+	extern invent_generic_t *klhwg_invent_alloc(cnodeid_t cnode, 
+						     int class, int size);
+	cnodeid_t		cnode = hubdev_cnodeid_get(node);
+
+	/* Allocate memory for the extra inventory information
+	 * for the  prom
+	 */
+	cpuprom_inventory_info = (invent_miscinfo_t *) 
+		klhwg_invent_alloc(cnode, INV_PROM, sizeof(invent_miscinfo_t));
+
+	ASSERT(cpuprom_inventory_info);
+
+	/* Set the enabled flag so that the hinv interprets this
+	 * information
+	 */
+	cpuprom_inventory_info->im_gen.ig_flag = INVENT_ENABLED;
+	cpuprom_inventory_info->im_type = SN_PROMVERSION;
+	/* Store prom revision into inventory information */
+	cpuprom_inventory_info->im_rev = IP27CONFIG.pvers_rev;
+	cpuprom_inventory_info->im_version = IP27CONFIG.pvers_vers;
+
+
+	/* Store this info as labelled information hanging off the
+	 * prom device vertex
+	 */
+	hwgraph_info_add_LBL(prom_dev, INFO_LBL_DETAIL_INVENT, 
+			     (arbitrary_info_t) cpuprom_inventory_info);
+	/* Export this information so that user programs can get to
+	 * this by using attr_get()
+	 */
+        hwgraph_info_export_LBL(prom_dev, INFO_LBL_DETAIL_INVENT,
+				sizeof(invent_miscinfo_t));
+}
+
+int
+cpuprom_attach(devfs_handle_t node)
+{
+        devfs_handle_t prom_dev;
+
+        hwgraph_char_device_add(node, EDGE_LBL_PROM, "hubspc_", &prom_dev);
+#ifdef	HUBSPC_DEBUG
+	printf("hubspc: prom_attach hub: 0x%x prom: 0x%x\n", node, prom_dev);
+#endif	/* HUBSPC_DEBUG */
+	device_inventory_add(prom_dev, INV_PROM, SN_PROMVERSION,
+				(major_t)0, (minor_t)0, 0);
+
+	/* Add additional inventory info about the cpu prom like
+	 * revision & version numbers etc.
+	 */
+	cpuprom_detailed_inventory_info_add(prom_dev,node);
+        device_info_set(prom_dev, (void*)(ulong)HUBSPC_PROM);
+	prominfo_add(node, prom_dev);
+
+        return (0);
+}
+
+#if defined(CONFIG_SGI_IP35) || defined(CONFIG_IA64_SGI_SN1) || defined(CONFIG_IA64_GENERIC)
+#define FPROM_CONFIG_ADDR	MD_JUNK_BUS_TIMING
+#define FPROM_ENABLE_MASK	MJT_FPROM_ENABLE_MASK
+#define FPROM_ENABLE_SHFT	MJT_FPROM_ENABLE_SHFT
+#define FPROM_SETUP_MASK	MJT_FPROM_SETUP_MASK
+#define FPROM_SETUP_SHFT	MJT_FPROM_SETUP_SHFT
+#endif
+
+/*ARGSUSED*/
+int
+cpuprom_map(devfs_handle_t dev, vhandl_t *vt, off_t addr, size_t len)
+{
+        int 		errcode;
+	caddr_t 	kvaddr;
+	devfs_handle_t		node;
+	cnodeid_t 	cnode;
+
+	node = prominfo_nodeget(dev);
+
+	if (!node)
+		return EIO;
+        
+
+	kvaddr = hubdev_prombase_get(node);
+	cnode  = hubdev_cnodeid_get(node);
+#ifdef	HUBSPC_DEBUG
+	printf("cpuprom_map: hubnode %d kvaddr 0x%x\n", node, kvaddr);
+#endif
+
+	if (len > RBOOT_SIZE)
+		len = RBOOT_SIZE;
+        /*
+         * Map in the prom space
+         */
+	errcode = v_mapphys(vt, kvaddr, len);
+
+	if (errcode == 0 ){
+		/*
+		 * Set the MD configuration registers suitably.
+		 */
+		nasid_t		nasid;
+		uint64_t	value;
+		volatile hubreg_t	*regaddr;
+
+		nasid = COMPACT_TO_NASID_NODEID(cnode);
+		regaddr = REMOTE_HUB_ADDR(nasid, FPROM_CONFIG_ADDR);
+		value = HUB_L(regaddr);
+		value &= ~(FPROM_SETUP_MASK | FPROM_ENABLE_MASK);
+		{
+			value |= (((long)CONFIG_FPROM_SETUP << FPROM_SETUP_SHFT) | 
+				  ((long)CONFIG_FPROM_ENABLE << FPROM_ENABLE_SHFT));
+		}
+		HUB_S(regaddr, value);
+
+	}
+        return (errcode);
+}
+
+/*ARGSUSED*/
+int
+cpuprom_unmap(devfs_handle_t dev, vhandl_t *vt)
+{
+        return 0;
+}
+
+/***********************************************************************/
+/* Base Hub Space Driver                                               */
+/***********************************************************************/
+
+// extern int l1_attach( devfs_handle_t );
+
+/*
+ * hubspc_init
+ * Registration of the hubspc devices with the hub manager
+ */
+void
+hubspc_init(void)
+{
+        /*
+         * Register with the hub manager
+         */
+
+        /* The reference counters */
+        hubdev_register(mem_refcnt_attach);
+
+	/* Prom space */
+	hubdev_register(cpuprom_attach);
+
+#if defined(CONFIG_SERIAL_SGI_L1_PROTOCOL)
+	/* L1 system controller link */
+	if ( !IS_RUNNING_ON_SIMULATOR() ) {
+		/* initialize the L1 link */
+		void l1_cons_init( l1sc_t *sc );
+		elsc_t *get_elsc(void);
+
+		l1_cons_init((l1sc_t *)get_elsc());
+	}
+#endif
+
+#ifdef	HUBSPC_DEBUG
+	printf("hubspc_init: Completed\n");
+#endif	/* HUBSPC_DEBUG */
+	/* Initialize spinlocks */
+	spinlock_init(&cpuprom_spinlock, "promlist");
+}
+
+/* ARGSUSED */
+int
+hubspc_open(devfs_handle_t *devp, mode_t oflag, int otyp, cred_t *crp)
+{
+        int errcode = 0;
+        
+        switch ((hubspc_subdevice_t)(ulong)device_info_get(*devp)) {
+        case HUBSPC_REFCOUNTERS:
+                errcode = mem_refcnt_open(devp, oflag, otyp, crp);
+                break;
+
+        case HUBSPC_PROM:
+		/* Check if the user has proper access rights to 
+		 * read/write the prom space.
+		 */
+                if (!cap_able(CAP_DEVICE_MGT)) {
+                        errcode = EPERM;
+                }                
+                break;
+
+        default:
+                errcode = ENODEV;
+        }
+
+#ifdef	HUBSPC_DEBUG
+	printf("hubspc_open: Completed open for type %d\n",
+               (hubspc_subdevice_t)(ulong)device_info_get(*devp));
+#endif	/* HUBSPC_DEBUG */
+
+        return (errcode);
+}
+
+
+/* ARGSUSED */
+int
+hubspc_close(devfs_handle_t dev, int oflag, int otyp, cred_t *crp)
+{
+        int errcode = 0;
+        
+        switch ((hubspc_subdevice_t)(ulong)device_info_get(dev)) {
+        case HUBSPC_REFCOUNTERS:
+                errcode = mem_refcnt_close(dev, oflag, otyp, crp);
+                break;
+
+        case HUBSPC_PROM:
+                break;
+        default:
+                errcode = ENODEV;
+        }
+
+#ifdef	HUBSPC_DEBUG
+	printf("hubspc_close: Completed close for type %d\n",
+               (hubspc_subdevice_t)(ulong)device_info_get(dev));
+#endif	/* HUBSPC_DEBUG */
+
+        return (errcode);
+}
+
+/* ARGSUSED */
+int
+hubspc_map(devfs_handle_t dev, vhandl_t *vt, off_t off, size_t len, uint prot)
+{
+	/*REFERENCED*/
+        hubspc_subdevice_t subdevice;
+        int errcode = 0;
+
+	/* check validity of request */
+	if( len == 0 ) {
+		return ENXIO;
+        }
+
+        subdevice = (hubspc_subdevice_t)(ulong)device_info_get(dev);
+
+#ifdef	HUBSPC_DEBUG
+	printf("hubspc_map: subdevice: %d vaddr: 0x%x phyaddr: 0x%x len: 0x%x\n",
+	       subdevice, v_getaddr(vt), off, len);
+#endif /* HUBSPC_DEBUG */
+
+        switch ((hubspc_subdevice_t)(ulong)device_info_get(dev)) {
+        case HUBSPC_REFCOUNTERS:
+                errcode = mem_refcnt_mmap(dev, vt, off, len, prot);
+                break;
+
+        case HUBSPC_PROM:
+		errcode = cpuprom_map(dev, vt, off, len);
+                break;
+        default:
+                errcode = ENODEV;
+        }
+
+#ifdef	HUBSPC_DEBUG
+	printf("hubspc_map finished: spctype: %d vaddr: 0x%x len: 0x%x\n",
+	       (hubspc_subdevice_t)(ulong)device_info_get(dev), v_getaddr(vt), len);
+#endif /* HUBSPC_DEBUG */
+
+	return errcode;
+}
+
+/* ARGSUSED */
+int
+hubspc_unmap(devfs_handle_t dev, vhandl_t *vt)
+{
+        int errcode = 0;
+        
+        switch ((hubspc_subdevice_t)(ulong)device_info_get(dev)) {
+        case HUBSPC_REFCOUNTERS:
+                errcode = mem_refcnt_unmap(dev, vt);
+                break;
+
+        case HUBSPC_PROM:
+                errcode = cpuprom_unmap(dev, vt);
+                break;
+
+        default:
+                errcode = ENODEV;
+        }
+	return errcode;
+
+}
+
+/* ARGSUSED */
+int
+hubspc_ioctl(devfs_handle_t dev,
+             int cmd,
+             void *arg,
+             int mode,
+             cred_t *cred_p,
+             int *rvalp)
+{
+        int errcode = 0;
+        
+        switch ((hubspc_subdevice_t)(ulong)device_info_get(dev)) {
+        case HUBSPC_REFCOUNTERS:
+                errcode = mem_refcnt_ioctl(dev, cmd, arg, mode, cred_p, rvalp);
+                break;
+
+        case HUBSPC_PROM:
+                break;
+
+        default:
+                errcode = ENODEV;
+        }
+	return errcode;
+
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/invent.c linux/arch/ia64/sn/io/invent.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/invent.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/invent.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,198 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+/*
+ * Hardware Inventory
+ *
+ * See sys/sn/invent.h for an explanation of the hardware inventory contents.
+ *
+ */
+#include <linux/types.h>
+#include <linux/config.h>
+#include <asm/sn/sgi.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/labelcl.h>
+
+void
+inventinit(void)
+{
+}
+
+/*
+ * For initializing/updating an inventory entry.
+ */
+void
+replace_in_inventory(
+	inventory_t *pinv, int class, int type,
+	int controller, int unit, int state)
+{
+	pinv->inv_class = class;
+	pinv->inv_type = type;
+	pinv->inv_controller = controller;
+	pinv->inv_unit = unit;
+	pinv->inv_state = state;
+}
+
+/*
+ * Inventory addition 
+ *
+ * XXX NOTE: Currently must be called after dynamic memory allocator is
+ * initialized.
+ *
+ */
+void
+add_to_inventory(int class, int type, int controller, int unit, int state)
+{
+	(void)device_inventory_add((devfs_handle_t)GRAPH_VERTEX_NONE, class, type, 
+					controller, unit, state);
+}
+
+
+/*
+ * Inventory retrieval 
+ *
+ * These two routines are intended to prevent the caller from having to know
+ * the internal structure of the inventory table.
+ *
+ */
+inventory_t *
+get_next_inventory(invplace_t *place)
+{
+	inventory_t *pinv;
+	devfs_handle_t device = place->invplace_vhdl;
+	int rv;
+
+	while ((pinv = device_inventory_get_next(device, place)) == NULL) {
+		/*
+		 * We've exhausted inventory items on the last device.
+		 * Advance to next device.
+		 */
+		rv = hwgraph_vertex_get_next(&device, &place->invplace_vplace);
+		if (rv != LABELCL_SUCCESS)
+			return(NULL);
+		place->invplace_vhdl = device;
+		place->invplace_inv = NULL; /* Start from beginning invent on this device */
+	}
+
+	return(pinv);
+}
+
+/* ARGSUSED */
+int
+get_sizeof_inventory(int abi)
+{
+	return sizeof(inventory_t);
+}
+
+/*
+ * Hardware inventory scanner.
+ *
+ * Calls fun() for every entry in inventory list unless fun() returns something
+ * other than 0.
+ */
+int
+scaninvent(int (*fun)(inventory_t *, void *), void *arg)
+{
+	inventory_t *ie;
+	invplace_t iplace = { NULL,NULL, NULL };
+	int rc;
+
+	ie = 0;
+	rc = 0;
+	while ( (ie = (inventory_t *)get_next_inventory(&iplace)) ) {
+		rc = (*fun)(ie, arg);
+		if (rc)
+			break;
+	}
+	return rc;
+}
+
+/*
+ * Find a particular inventory object
+ *
+ * pinv can be a pointer to an inventory entry and the search will begin from
+ * there, or it can be 0 in which case the search starts at the beginning.
+ * A -1 for any of the other arguments is a wildcard (i.e. it always matches).
+ */
+inventory_t *
+find_inventory(inventory_t *pinv, int class, int type, int controller,
+	       int unit, int state)
+{
+	invplace_t iplace =  { NULL,NULL, NULL };
+
+	while ((pinv = (inventory_t *)get_next_inventory(&iplace)) != NULL) {
+		if (class != -1 && pinv->inv_class != class)
+			continue;
+		if (type != -1 && pinv->inv_type != type)
+			continue;
+
+		/* XXXX - perhaps the "state" entry should be ignored so an
+		 * an existing entry can be updated.  See vino_init() and
+		 * ml/IP22.c:add_ioboard() for an example.
+		 */
+		if (state != -1 && pinv->inv_state != state)
+			continue;
+		if (controller != -1
+		    && pinv->inv_controller != controller)
+			continue;
+		if (unit != -1 && pinv->inv_unit != unit)
+			continue;
+		break;
+	}
+
+	return(pinv);
+}
+
+
+/*
+** Retrieve inventory data associated with a device.
+*/
+inventory_t *
+device_inventory_get_next(	devfs_handle_t device,
+				invplace_t *invplace)
+{
+	inventory_t *pinv;
+	int rv;
+
+	rv = hwgraph_inventory_get_next(device, invplace, &pinv);
+	if (rv == LABELCL_SUCCESS)
+		return(pinv);
+	else
+		return(NULL);
+}
+
+
+/*
+** Associate canonical inventory information with a device (and
+** add it to the general inventory).
+*/
+void
+device_inventory_add(	devfs_handle_t device,
+			int class, 
+			int type, 
+			major_t controller, 
+			minor_t unit, 
+			int state)
+{
+	hwgraph_inventory_add(device, class, type, controller, unit, state);
+}
+
+int
+device_controller_num_get(devfs_handle_t device)
+{
+	return (hwgraph_controller_num_get(device));
+}
+
+void
+device_controller_num_set(devfs_handle_t device, int contr_num)
+{
+	hwgraph_controller_num_set(device, contr_num);
+}
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/io.c linux/arch/ia64/sn/io/io.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/io.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/io.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,1312 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+#include <linux/types.h>
+#include <linux/config.h>
+#include <linux/slab.h>
+#include <asm/sn/types.h>
+#include <asm/sn/sgi.h>
+#include <asm/sn/cmn_err.h>
+#include <asm/sn/iobus.h>
+#include <asm/sn/iograph.h>
+#include <asm/param.h>
+#include <asm/sn/pio.h>
+#include <asm/sn/xtalk/xwidget.h>
+#include <asm/sn/sn_private.h>
+#include <asm/sn/addrs.h>
+#include <asm/sn/invent.h>
+#include <asm/sn/hcl.h>
+#include <asm/sn/hcl_util.h>
+#include <asm/sn/agent.h>
+#include <asm/sn/intr.h>
+#include <asm/sn/xtalk/xtalkaddrs.h>
+#include <asm/sn/klconfig.h>
+#include <asm/sn/xtalk/xwidget.h>
+#include <asm/sn/io.h>
+#include <asm/sn/sn_cpuid.h>
+
+extern xtalk_provider_t hub_provider;
+
+#ifndef CONFIG_IA64_SGI_IO
+/* Global variables */
+extern pdaindr_t       pdaindr[MAXCPUS];
+#endif
+
+/*
+ * Perform any initializations needed to support hub-based I/O.
+ * Called once during startup.
+ */
+void
+hubio_init(void)
+{
+#if 0
+	/* This isn't needed unless we port the entire sio driver ... */
+        extern void early_brl1_port_init( void );
+	early_brl1_port_init();
+#endif
+}
+
+/* 
+ * Implementation of hub iobus operations.
+ *
+ * Hub provides a crosstalk "iobus" on IP27 systems.  These routines
+ * provide a platform-specific implementation of xtalk used by all xtalk 
+ * cards on IP27 systems.
+ *
+ * Called from corresponding xtalk_* routines.
+ */
+
+
+/* PIO MANAGEMENT */
+/* For mapping system virtual address space to xtalk space on a specified widget */
+
+/*
+ * Setup pio structures needed for a particular hub.
+ */
+static void
+hub_pio_init(devfs_handle_t hubv)
+{
+	xwidgetnum_t widget;
+	hubinfo_t hubinfo;
+	nasid_t nasid;
+	int bigwin;
+	hub_piomap_t hub_piomap;
+
+	hubinfo_get(hubv, &hubinfo);
+	nasid = hubinfo->h_nasid;
+
+	/* Initialize small window piomaps for this hub */
+	for (widget=0; widget <= HUB_WIDGET_ID_MAX; widget++) {
+		hub_piomap = hubinfo_swin_piomap_get(hubinfo, (int)widget);
+		hub_piomap->hpio_xtalk_info.xp_target = widget;
+		hub_piomap->hpio_xtalk_info.xp_xtalk_addr = 0;
+		hub_piomap->hpio_xtalk_info.xp_mapsz = SWIN_SIZE;
+		hub_piomap->hpio_xtalk_info.xp_kvaddr = (caddr_t)NODE_SWIN_BASE(nasid, widget);
+		hub_piomap->hpio_hub = hubv;
+		hub_piomap->hpio_flags = HUB_PIOMAP_IS_VALID;
+	}
+
+	/* Initialize big window piomaps for this hub */
+	for (bigwin=0; bigwin < HUB_NUM_BIG_WINDOW; bigwin++) {
+		hub_piomap = hubinfo_bwin_piomap_get(hubinfo, bigwin);
+		hub_piomap->hpio_xtalk_info.xp_mapsz = BWIN_SIZE;
+		hub_piomap->hpio_hub = hubv;
+		hub_piomap->hpio_holdcnt = 0;
+		hub_piomap->hpio_flags = HUB_PIOMAP_IS_BIGWINDOW;
+		IIO_ITTE_DISABLE(nasid, bigwin);
+	}
+#ifdef	BRINGUP
+	hub_set_piomode(nasid, HUB_PIO_CONVEYOR);
+#else
+	/* Set all the xwidgets in fire-and-forget mode
+	 * by default
+	 */
+	hub_set_piomode(nasid, HUB_PIO_FIRE_N_FORGET);
+#endif	/* BRINGUP */
+
+	sv_init(&hubinfo->h_bwwait, SV_FIFO, "bigwin");
+	spinlock_init(&hubinfo->h_bwlock, "bigwin");
+}
+
+/* 
+ * Create a caddr_t-to-xtalk_addr mapping.
+ *
+ * Use a small window if possible (that's the usual case), but
+ * manage big windows if needed.  Big window mappings can be
+ * either FIXED or UNFIXED -- we keep at least 1 big window available
+ * for UNFIXED mappings.
+ *
+ * Returns an opaque pointer-sized type which can be passed to
+ * other hub_pio_* routines on success, or NULL if the request
+ * cannot be satisfied.
+ */
+/* ARGSUSED */
+hub_piomap_t
+hub_piomap_alloc(devfs_handle_t dev,	/* set up mapping for this device */
+		device_desc_t dev_desc,	/* device descriptor */
+		iopaddr_t xtalk_addr,	/* map for this xtalk_addr range */
+		size_t byte_count,
+		size_t byte_count_max, 	/* maximum size of a mapping */
+		unsigned flags)		/* defined in sys/pio.h */
+{
+	xwidget_info_t widget_info = xwidget_info_get(dev);
+	xwidgetnum_t widget = xwidget_info_id_get(widget_info);
+	devfs_handle_t hubv = xwidget_info_master_get(widget_info);
+	hubinfo_t hubinfo;
+	hub_piomap_t bw_piomap;
+	int bigwin, free_bw_index;
+	nasid_t nasid;
+	volatile hubreg_t junk;
+	int s;
+
+	/* sanity check */
+	if (byte_count_max > byte_count)
+		return(NULL);
+
+	hubinfo_get(hubv, &hubinfo);
+
+	/* If xtalk_addr range is mapped by a small window, we don't have 
+	 * to do much 
+	 */
+	if (xtalk_addr + byte_count <= SWIN_SIZE)
+		return(hubinfo_swin_piomap_get(hubinfo, (int)widget));
+
+	/* We need to use a big window mapping.  */
+
+	/*
+	 * TBD: Allow requests that would consume multiple big windows --
+	 * split the request up and use multiple mapping entries.
+	 * For now, reject requests that span big windows.
+	 */
+	if ((xtalk_addr % BWIN_SIZE) + byte_count > BWIN_SIZE)
+		return(NULL);
+
+
+	/* Round xtalk address down for big window alignement */
+	xtalk_addr = xtalk_addr & ~(BWIN_SIZE-1);
+
+	/*
+	 * Check to see if an existing big window mapping will suffice.
+	 */
+tryagain:
+	free_bw_index = -1;
+	s = mutex_spinlock(&hubinfo->h_bwlock);
+	for (bigwin=0; bigwin < HUB_NUM_BIG_WINDOW; bigwin++) {
+		bw_piomap = hubinfo_bwin_piomap_get(hubinfo, bigwin);
+
+		/* If mapping is not valid, skip it */
+		if (!(bw_piomap->hpio_flags & HUB_PIOMAP_IS_VALID)) {
+			free_bw_index = bigwin;
+			continue;
+		}
+
+		/* 
+		 * If mapping is UNFIXED, skip it.  We don't allow sharing
+		 * of UNFIXED mappings, because this would allow starvation.
+		 */
+		if (!(bw_piomap->hpio_flags & HUB_PIOMAP_IS_FIXED))
+			continue;
+
+		if ( xtalk_addr == bw_piomap->hpio_xtalk_info.xp_xtalk_addr &&
+		     widget == bw_piomap->hpio_xtalk_info.xp_target) {
+			bw_piomap->hpio_holdcnt++;
+			mutex_spinunlock(&hubinfo->h_bwlock, s);
+			return(bw_piomap);
+		}
+	}
+
+	/*
+	 * None of the existing big window mappings will work for us --
+	 * we need to establish a new mapping.
+	 */
+
+	/* Insure that we don't consume all big windows with FIXED mappings */
+	if (flags & PIOMAP_FIXED) {
+		if (hubinfo->h_num_big_window_fixed < HUB_NUM_BIG_WINDOW-1) {
+			ASSERT(free_bw_index >= 0);
+			hubinfo->h_num_big_window_fixed++;
+		} else {
+			bw_piomap = NULL;
+			goto done;
+		}
+	} else /* PIOMAP_UNFIXED */ {
+		if (free_bw_index < 0) {
+			if (flags & PIOMAP_NOSLEEP) {
+				bw_piomap = NULL;
+				goto done;
+			}
+
+			sv_wait(&hubinfo->h_bwwait, PZERO, &hubinfo->h_bwlock, s);
+			goto tryagain;
+		}
+	}
+
+
+	/* OK!  Allocate big window free_bw_index for this mapping. */
+ 	/* 
+	 * The code below does a PIO write to setup an ITTE entry.
+	 * We need to prevent other CPUs from seeing our updated memory 
+	 * shadow of the ITTE (in the piomap) until the ITTE entry is 
+	 * actually set up; otherwise, another CPU might attempt a PIO 
+	 * prematurely.  
+	 *
+	 * Also, the only way we can know that an entry has been received 
+	 * by the hub and can be used by future PIO reads/writes is by 
+	 * reading back the ITTE entry after writing it.
+	 *
+	 * For these two reasons, we PIO read back the ITTE entry after
+	 * we write it.
+	 */
+
+	nasid = hubinfo->h_nasid;
+	IIO_ITTE_PUT(nasid, free_bw_index, HUB_PIO_MAP_TO_MEM, widget, xtalk_addr);	
+	junk = HUB_L(IIO_ITTE_GET(nasid, free_bw_index));
+
+	bw_piomap = hubinfo_bwin_piomap_get(hubinfo, free_bw_index);
+	bw_piomap->hpio_xtalk_info.xp_dev = dev;
+	bw_piomap->hpio_xtalk_info.xp_target = widget;
+	bw_piomap->hpio_xtalk_info.xp_xtalk_addr = xtalk_addr;
+	bw_piomap->hpio_xtalk_info.xp_kvaddr = (caddr_t)NODE_BWIN_BASE(nasid, free_bw_index);
+	bw_piomap->hpio_holdcnt++;
+	bw_piomap->hpio_bigwin_num = free_bw_index;
+
+	if (flags & PIOMAP_FIXED)
+		bw_piomap->hpio_flags |= HUB_PIOMAP_IS_VALID | HUB_PIOMAP_IS_FIXED;
+	else
+		bw_piomap->hpio_flags |= HUB_PIOMAP_IS_VALID;
+
+done:
+	mutex_spinunlock(&hubinfo->h_bwlock, s);
+	return(bw_piomap);
+}
+
+/*
+ * hub_piomap_free destroys a caddr_t-to-xtalk pio mapping and frees
+ * any associated mapping resources.  
+ *
+ * If this * piomap was handled with a small window, or if it was handled
+ * in a big window that's still in use by someone else, then there's 
+ * nothing to do.  On the other hand, if this mapping was handled 
+ * with a big window, AND if we were the final user of that mapping, 
+ * then destroy the mapping.
+ */
+void
+hub_piomap_free(hub_piomap_t hub_piomap)
+{
+	devfs_handle_t hubv;
+	hubinfo_t hubinfo;
+	nasid_t nasid;
+	int s;
+
+	/* 
+	 * Small windows are permanently mapped to corresponding widgets,
+	 * so there're no resources to free.
+	 */
+	if (!(hub_piomap->hpio_flags & HUB_PIOMAP_IS_BIGWINDOW))
+		return;
+
+	ASSERT(hub_piomap->hpio_flags & HUB_PIOMAP_IS_VALID);
+	ASSERT(hub_piomap->hpio_holdcnt > 0);
+
+	hubv = hub_piomap->hpio_hub;
+	hubinfo_get(hubv, &hubinfo);
+	nasid = hubinfo->h_nasid;
+
+	s = mutex_spinlock(&hubinfo->h_bwlock);
+
+	/*
+	 * If this is the last hold on this mapping, free it.
+	 */
+	if (--hub_piomap->hpio_holdcnt == 0) {
+		IIO_ITTE_DISABLE(nasid, hub_piomap->hpio_bigwin_num );
+
+		if (hub_piomap->hpio_flags & HUB_PIOMAP_IS_FIXED) {
+			hub_piomap->hpio_flags &= ~(HUB_PIOMAP_IS_VALID | HUB_PIOMAP_IS_FIXED);
+			hubinfo->h_num_big_window_fixed--;
+			ASSERT(hubinfo->h_num_big_window_fixed >= 0);
+		} else
+			hub_piomap->hpio_flags &= ~HUB_PIOMAP_IS_VALID;
+
+		(void)sv_signal(&hubinfo->h_bwwait);
+	}
+
+	mutex_spinunlock(&hubinfo->h_bwlock, s);
+}
+
+/*
+ * Establish a mapping to a given xtalk address range using the resources
+ * allocated earlier.
+ */
+caddr_t
+hub_piomap_addr(hub_piomap_t hub_piomap,	/* mapping resources */
+		iopaddr_t xtalk_addr,		/* map for this xtalk address */
+		size_t byte_count)		/* map this many bytes */
+{
+	/* Verify that range can be mapped using the specified piomap */
+	if (xtalk_addr < hub_piomap->hpio_xtalk_info.xp_xtalk_addr)
+		return(0);
+
+	if (xtalk_addr + byte_count > 
+		( hub_piomap->hpio_xtalk_info.xp_xtalk_addr + 
+			hub_piomap->hpio_xtalk_info.xp_mapsz))
+		return(0);
+
+	if (hub_piomap->hpio_flags & HUB_PIOMAP_IS_VALID)
+		return(hub_piomap->hpio_xtalk_info.xp_kvaddr + 
+			(xtalk_addr % hub_piomap->hpio_xtalk_info.xp_mapsz));
+	else
+		return(0);
+}
+
+
+/*
+ * Driver indicates that it's done with PIO's from an earlier piomap_addr.
+ */
+/* ARGSUSED */
+void
+hub_piomap_done(hub_piomap_t hub_piomap)	/* done with these mapping resources */
+{
+	/* Nothing to do */
+}
+
+
+/*
+ * For translations that require no mapping resources, supply a kernel virtual
+ * address that maps to the specified xtalk address range.
+ */
+/* ARGSUSED */
+caddr_t
+hub_piotrans_addr(	devfs_handle_t dev,	/* translate to this device */
+			device_desc_t dev_desc,	/* device descriptor */
+			iopaddr_t xtalk_addr,	/* Crosstalk address */
+			size_t byte_count,	/* map this many bytes */
+			unsigned flags)		/* (currently unused) */
+{
+	xwidget_info_t widget_info = xwidget_info_get(dev);
+	xwidgetnum_t widget = xwidget_info_id_get(widget_info);
+	devfs_handle_t hubv = xwidget_info_master_get(widget_info);
+	hub_piomap_t hub_piomap;
+	hubinfo_t hubinfo;
+
+	hubinfo_get(hubv, &hubinfo);
+
+	if (xtalk_addr + byte_count <= SWIN_SIZE) {
+		hub_piomap = hubinfo_swin_piomap_get(hubinfo, (int)widget);
+		return(hub_piomap_addr(hub_piomap, xtalk_addr, byte_count));
+	} else
+		return(0);
+}
+
+
+/* DMA MANAGEMENT */
+/* Mapping from crosstalk space to system physical space */
+
+/* 
+ * There's not really very much to do here, since crosstalk maps
+ * directly to system physical space.  It's quite possible that this
+ * DMA layer will be bypassed in performance kernels.
+ */
+
+
+/* ARGSUSED */
+static void
+hub_dma_init(devfs_handle_t hubv)
+{
+}
+
+
+/*
+ * Allocate resources needed to set up DMA mappings up to a specified size
+ * on a specified adapter.
+ * 
+ * We don't actually use the adapter ID for anything.  It's just the adapter
+ * that the lower level driver plans to use for DMA.
+ */
+/* ARGSUSED */
+hub_dmamap_t
+hub_dmamap_alloc(	devfs_handle_t dev,	/* set up mappings for this device */
+			device_desc_t dev_desc,	/* device descriptor */
+			size_t byte_count_max, 	/* max size of a mapping */
+			unsigned flags)		/* defined in dma.h */
+{
+	hub_dmamap_t dmamap;
+	xwidget_info_t widget_info = xwidget_info_get(dev);
+	xwidgetnum_t widget = xwidget_info_id_get(widget_info);
+	devfs_handle_t hubv = xwidget_info_master_get(widget_info);
+
+	dmamap = kern_malloc(sizeof(struct hub_dmamap_s));
+	dmamap->hdma_xtalk_info.xd_dev = dev;
+	dmamap->hdma_xtalk_info.xd_target = widget;
+	dmamap->hdma_hub = hubv;
+	dmamap->hdma_flags = HUB_DMAMAP_IS_VALID;
+ 	if (flags & XTALK_FIXED)
+		dmamap->hdma_flags |= HUB_DMAMAP_IS_FIXED;
+
+	return(dmamap);
+}
+
+/*
+ * Destroy a DMA mapping from crosstalk space to system address space.
+ * There is no actual mapping hardware to destroy, but we at least mark
+ * the dmamap INVALID and free the space that it took.
+ */
+void
+hub_dmamap_free(hub_dmamap_t hub_dmamap)
+{
+	hub_dmamap->hdma_flags &= ~HUB_DMAMAP_IS_VALID;
+	kern_free(hub_dmamap);
+}
+
+/*
+ * Establish a DMA mapping using the resources allocated in a previous dmamap_alloc.
+ * Return an appropriate crosstalk address range that maps to the specified physical 
+ * address range.
+ */
+/* ARGSUSED */
+extern iopaddr_t
+hub_dmamap_addr(	hub_dmamap_t dmamap,	/* use these mapping resources */
+			paddr_t paddr,		/* map for this address */
+			size_t byte_count)	/* map this many bytes */
+{
+	devfs_handle_t vhdl;
+
+	ASSERT(dmamap->hdma_flags & HUB_DMAMAP_IS_VALID);
+
+	if (dmamap->hdma_flags & HUB_DMAMAP_USED) {
+	    /* If the map is FIXED, re-use is OK. */
+	    if (!(dmamap->hdma_flags & HUB_DMAMAP_IS_FIXED)) {
+		vhdl = dmamap->hdma_xtalk_info.xd_dev;
+#if defined(SUPPORT_PRINTING_V_FORMAT)
+		cmn_err(CE_WARN, "%v: hub_dmamap_addr re-uses dmamap.\n",vhdl);
+#else
+		cmn_err(CE_WARN, "0x%p: hub_dmamap_addr re-uses dmamap.\n", &vhdl);
+#endif
+	    }
+	} else {
+		dmamap->hdma_flags |= HUB_DMAMAP_USED;
+	}
+
+	/* There isn't actually any DMA mapping hardware on the hub. */
+	return(paddr);
+}
+
+/*
+ * Establish a DMA mapping using the resources allocated in a previous dmamap_alloc.
+ * Return an appropriate crosstalk address list that maps to the specified physical 
+ * address list.
+ */
+/* ARGSUSED */
+alenlist_t
+hub_dmamap_list(hub_dmamap_t hub_dmamap,	/* use these mapping resources */
+		alenlist_t palenlist,		/* map this area of memory */
+		unsigned flags)
+{
+	devfs_handle_t vhdl;
+
+	ASSERT(hub_dmamap->hdma_flags & HUB_DMAMAP_IS_VALID);
+
+	if (hub_dmamap->hdma_flags & HUB_DMAMAP_USED) {
+	    /* If the map is FIXED, re-use is OK. */
+	    if (!(hub_dmamap->hdma_flags & HUB_DMAMAP_IS_FIXED)) {
+		vhdl = hub_dmamap->hdma_xtalk_info.xd_dev;
+#if defined(SUPPORT_PRINTING_V_FORMAT)
+		cmn_err(CE_WARN,"%v: hub_dmamap_list re-uses dmamap\n",vhdl);
+#else
+		cmn_err(CE_WARN,"0x%p: hub_dmamap_list re-uses dmamap\n", &vhdl);
+#endif
+	    }
+	} else {
+		hub_dmamap->hdma_flags |= HUB_DMAMAP_USED;
+	}
+
+	/* There isn't actually any DMA mapping hardware on the hub.  */
+	return(palenlist);
+}
+
+/*
+ * Driver indicates that it has completed whatever DMA it may have started
+ * after an earlier dmamap_addr or dmamap_list call.
+ */
+void
+hub_dmamap_done(hub_dmamap_t hub_dmamap)	/* done with these mapping resources */
+{
+	devfs_handle_t vhdl;
+
+	if (hub_dmamap->hdma_flags & HUB_DMAMAP_USED) {
+		hub_dmamap->hdma_flags &= ~HUB_DMAMAP_USED;
+	} else {
+	    /* If the map is FIXED, re-done is OK. */
+	    if (!(hub_dmamap->hdma_flags & HUB_DMAMAP_IS_FIXED)) {
+		vhdl = hub_dmamap->hdma_xtalk_info.xd_dev;
+#if defined(SUPPORT_PRINTING_V_FORMAT)
+		cmn_err(CE_WARN, "%v: hub_dmamap_done already done with dmamap\n",vhdl);
+#else
+		cmn_err(CE_WARN, "0x%p: hub_dmamap_done already done with dmamap\n", &vhdl);
+#endif
+	    }
+	}
+}
+
+/*
+ * Translate a single system physical address into a crosstalk address.
+ */
+/* ARGSUSED */
+iopaddr_t
+hub_dmatrans_addr(	devfs_handle_t dev,	/* translate for this device */
+			device_desc_t dev_desc,	/* device descriptor */
+			paddr_t paddr,		/* system physical address */
+			size_t byte_count,	/* length */
+			unsigned flags)		/* defined in dma.h */
+{
+	/* no translation needed */
+	return(paddr);
+}
+
+/*
+ * Translate a list of IP27 addresses and lengths into a list of crosstalk 
+ * addresses and lengths.  No actual hardware mapping takes place; the hub 
+ * has no DMA mapping registers -- crosstalk addresses map directly.
+ */
+/* ARGSUSED */
+alenlist_t
+hub_dmatrans_list(	devfs_handle_t dev,	/* translate for this device */
+			device_desc_t dev_desc,	/* device descriptor */
+			alenlist_t palenlist,	/* system address/length list */
+			unsigned flags)		/* defined in dma.h */
+{
+	/* no translation needed */
+	return(palenlist);
+}
+
+/*ARGSUSED*/
+void
+hub_dmamap_drain(	hub_dmamap_t map)
+{
+    /* XXX- flush caches, if cache coherency WAR is needed */
+}
+
+/*ARGSUSED*/
+void
+hub_dmaaddr_drain(	devfs_handle_t vhdl,
+			paddr_t addr,
+			size_t bytes)
+{
+    /* XXX- flush caches, if cache coherency WAR is needed */
+}
+
+/*ARGSUSED*/
+void
+hub_dmalist_drain(	devfs_handle_t vhdl,
+			alenlist_t list)
+{
+    /* XXX- flush caches, if cache coherency WAR is needed */
+}
+
+
+
+/* INTERRUPT MANAGEMENT */
+
+/* ARGSUSED */
+static void
+hub_intr_init(devfs_handle_t hubv)
+{
+}
+
+/*
+ * hub_device_desc_update
+ *	Update the passed in device descriptor with the actual the
+ * 	target cpu number and interrupt priority level.
+ *	NOTE : These might be the same as the ones passed in thru
+ *	the descriptor.
+ */
+static void
+hub_device_desc_update(device_desc_t 	dev_desc, 
+		       ilvl_t 		intr_swlevel,
+		       cpuid_t		cpu)
+{
+	char	cpuname[40];
+	
+	/* Store the interrupt priority level in the device descriptor */
+	device_desc_intr_swlevel_set(dev_desc, intr_swlevel);
+
+	/* Convert the cpuid to the vertex handle in the hwgraph and
+	 * save it in the device descriptor.
+	 */
+	sprintf(cpuname,"/hw/cpunum/%ld",cpu);
+	device_desc_intr_target_set(dev_desc, 
+				    hwgraph_path_to_dev(cpuname));
+}
+
+int allocate_my_bit = INTRCONNECT_ANYBIT;
+
+/*
+ * Allocate resources required for an interrupt as specified in dev_desc.
+ * Returns a hub interrupt handle on success, or 0 on failure.
+ */
+hub_intr_t
+hub_intr_alloc(	devfs_handle_t dev,		/* which crosstalk device */
+		device_desc_t dev_desc,		/* device descriptor */
+		devfs_handle_t owner_dev)		/* owner of this interrupt, if known */
+{
+	cpuid_t cpu;			/* cpu to receive interrupt */
+        int cpupicked = 0;
+	int bit;			/* interrupt vector */
+	/*REFERENCED*/
+	int intr_resflags;
+	hub_intr_t intr_hdl;
+	cnodeid_t nodeid;		/* node to receive interrupt */
+	/*REFERENCED*/
+	nasid_t nasid;			/* nasid to receive interrupt */
+	struct xtalk_intr_s *xtalk_info;
+	iopaddr_t xtalk_addr;		/* xtalk addr on hub to set intr */
+	xwidget_info_t xwidget_info;	/* standard crosstalk widget info handle */
+	char *intr_name = NULL;
+	ilvl_t intr_swlevel;
+	extern int default_intr_pri;
+#ifdef CONFIG_IA64_SGI_SN1 
+	extern void synergy_intr_alloc(int, int);
+#endif
+	
+	/*
+	 * If caller didn't explicily specify a device descriptor, see if there's
+	 * a default descriptor associated with the device.
+	 */
+	if (!dev_desc) 
+		dev_desc = device_desc_default_get(dev);
+
+	if (dev_desc) {
+		intr_name = device_desc_intr_name_get(dev_desc);
+		intr_swlevel = device_desc_intr_swlevel_get(dev_desc);
+		if (dev_desc->flags & D_INTR_ISERR) {
+			intr_resflags = II_ERRORINT;
+		} else if (!(dev_desc->flags & D_INTR_NOTHREAD)) {
+			intr_resflags = II_THREADED;
+		} else {
+			/* Neither an error nor a thread. */
+			intr_resflags = 0;
+		}
+	} else {
+		intr_swlevel = default_intr_pri;
+		intr_resflags = II_THREADED;
+	}
+
+	/* XXX - Need to determine if the interrupt should be threaded. */
+
+	/* If the cpu has not been picked already then choose a candidate 
+	 * interrupt target and reserve the interrupt bit 
+	 */
+#if defined(NEW_INTERRUPTS)
+	if (!cpupicked) {
+		cpu = intr_heuristic(dev,dev_desc,allocate_my_bit,
+				     intr_resflags,owner_dev,
+				     intr_name,&bit);
+	}
+#endif
+
+	/* At this point we SHOULD have a valid cpu */
+	if (cpu == CPU_NONE) {
+#if defined(SUPPORT_PRINTING_V_FORMAT)
+		cmn_err(CE_WARN, 
+			"%v hub_intr_alloc could not allocate interrupt\n",
+			owner_dev);
+#else
+		cmn_err(CE_WARN, 
+			"0x%p hub_intr_alloc could not allocate interrupt\n",
+			&owner_dev);
+#endif
+		return(0);
+
+	}
+
+	/* If the cpu has been picked already (due to the bridge data 
+	 * corruption bug) then try to reserve an interrupt bit .
+	 */
+#if defined(NEW_INTERRUPTS)
+	if (cpupicked) {
+		bit = intr_reserve_level(cpu, allocate_my_bit, 
+					 intr_resflags, 
+					 owner_dev, intr_name);
+		if (bit < 0) {
+#if defined(SUPPORT_PRINTING_V_FORMAT)
+			cmn_err(CE_WARN,
+				"Could not reserve an interrupt bit for cpu "
+				" %d and dev %v\n",
+				cpu,owner_dev);
+#else
+			cmn_err(CE_WARN,
+				"Could not reserve an interrupt bit for cpu "
+				" %d and dev 0x%x\n",
+				cpu, &owner_dev);
+#endif
+				
+			return(0);
+		}
+	}
+#endif	/* NEW_INTERRUPTS */
+
+	nodeid = cpuid_to_cnodeid(cpu);
+	nasid = cpuid_to_nasid(cpu);
+	xtalk_addr = HUBREG_AS_XTALKADDR(nasid, PIREG(PI_INT_PEND_MOD, cpuid_to_subnode(cpu)));
+
+	/*
+	 * Allocate an interrupt handle, and fill it in.  There are two
+	 * pieces to an interrupt handle: the piece needed by generic
+	 * xtalk code which is used by crosstalk device drivers, and
+	 * the piece needed by low-level IP27 hardware code.
+	 */
+	intr_hdl = kmem_alloc_node(sizeof(struct hub_intr_s), KM_NOSLEEP, nodeid);
+	ASSERT_ALWAYS(intr_hdl);
+
+	/* 
+	 * Fill in xtalk information for generic xtalk interfaces that
+	 * operate on xtalk_intr_hdl's.
+	 */
+	xtalk_info = &intr_hdl->i_xtalk_info;
+	xtalk_info->xi_dev = dev;
+	xtalk_info->xi_vector = bit;
+	xtalk_info->xi_addr = xtalk_addr;
+	xtalk_info->xi_flags =  (intr_resflags == II_THREADED) ?
+				0 : XTALK_INTR_NOTHREAD;
+
+	/*
+	 * Regardless of which CPU we ultimately interrupt, a given crosstalk
+	 * widget always handles interrupts (and PIO and DMA) through its 
+	 * designated "master" crosstalk provider.
+	 */
+	xwidget_info = xwidget_info_get(dev);
+	if (xwidget_info)
+		xtalk_info->xi_target = xwidget_info_masterid_get(xwidget_info);
+
+	/* Fill in low level hub information for hub_* interrupt interface */
+	intr_hdl->i_swlevel = intr_swlevel;
+	intr_hdl->i_cpuid = cpu;
+	intr_hdl->i_bit = bit;
+	intr_hdl->i_flags = HUB_INTR_IS_ALLOCED;
+
+	/* Store the actual interrupt priority level & interrupt target
+	 * cpu back in the device descriptor.
+	 */
+	hub_device_desc_update(dev_desc, intr_swlevel, cpu);
+#ifdef CONFIG_IA64_SGI_SN1
+	synergy_intr_alloc((int)bit, (int)cpu);
+#endif
+	return(intr_hdl);
+}
+
+
+/*
+ * Free resources consumed by intr_alloc.
+ */
+void
+hub_intr_free(hub_intr_t intr_hdl)
+{
+	cpuid_t cpu = intr_hdl->i_cpuid;
+	int bit = intr_hdl->i_bit;
+	xtalk_intr_t xtalk_info;
+
+	if (intr_hdl->i_flags & HUB_INTR_IS_CONNECTED) {
+		/* Setting the following fields in the xtalk interrupt info
+	 	 * clears the interrupt target register in the xtalk user
+	 	 */
+		xtalk_info = &intr_hdl->i_xtalk_info;
+		xtalk_info->xi_dev = NODEV;
+		xtalk_info->xi_vector = 0;
+		xtalk_info->xi_addr = 0;
+		hub_intr_disconnect(intr_hdl);
+	}
+
+	if (intr_hdl->i_flags & HUB_INTR_IS_ALLOCED)
+		kfree(intr_hdl);
+
+#if defined(NEW_INTERRUPTS)
+	intr_unreserve_level(cpu, bit);
+#endif
+}
+
+
+/*
+ * Associate resources allocated with a previous hub_intr_alloc call with the
+ * described handler, arg, name, etc.
+ */
+/*ARGSUSED*/
+int
+hub_intr_connect(	hub_intr_t intr_hdl,		/* xtalk intr resource handle */
+			intr_func_t intr_func,		/* xtalk intr handler */
+			void *intr_arg,			/* arg to intr handler */
+			xtalk_intr_setfunc_t setfunc,	/* func to set intr hw */
+			void *setfunc_arg,		/* arg to setfunc */
+			void *thread)			/* intr thread to use */
+{
+	int rv;
+	cpuid_t cpu = intr_hdl->i_cpuid;
+	int bit = intr_hdl->i_bit;
+#ifdef CONFIG_IA64_SGI_SN1
+	extern int synergy_intr_connect(int, int);
+#endif
+
+	ASSERT(intr_hdl->i_flags & HUB_INTR_IS_ALLOCED);
+
+#if defined(NEW_INTERRUPTS)
+	rv = intr_connect_level(cpu, bit, intr_hdl->i_swlevel, 
+					intr_func, intr_arg, NULL);
+	if (rv < 0)
+		return(rv);
+
+#endif
+	intr_hdl->i_xtalk_info.xi_setfunc = setfunc;
+	intr_hdl->i_xtalk_info.xi_sfarg = setfunc_arg;
+
+	if (setfunc) (*setfunc)((xtalk_intr_t)intr_hdl);
+
+	intr_hdl->i_flags |= HUB_INTR_IS_CONNECTED;
+#ifdef CONFIG_IA64_SGI_SN1
+	return(synergy_intr_connect((int)bit, (int)cpu));
+#endif
+}
+
+
+/*
+ * Disassociate handler with the specified interrupt.
+ */
+void
+hub_intr_disconnect(hub_intr_t intr_hdl)
+{
+	/*REFERENCED*/
+	int rv;
+	cpuid_t cpu = intr_hdl->i_cpuid;
+	int bit = intr_hdl->i_bit;
+	xtalk_intr_setfunc_t setfunc;
+
+	setfunc = intr_hdl->i_xtalk_info.xi_setfunc;
+
+	/* TBD: send disconnected interrupts somewhere harmless */
+	if (setfunc) (*setfunc)((xtalk_intr_t)intr_hdl);
+
+#if defined(NEW_INTERRUPTS)
+	rv = intr_disconnect_level(cpu, bit);
+	ASSERT(rv == 0);
+#endif
+
+	intr_hdl->i_flags &= ~HUB_INTR_IS_CONNECTED;
+}
+
+
+/*
+ * Return a hwgraph vertex that represents the CPU currently
+ * targeted by an interrupt.
+ */
+devfs_handle_t
+hub_intr_cpu_get(hub_intr_t intr_hdl)
+{
+	cpuid_t cpuid = intr_hdl->i_cpuid;
+	ASSERT(cpuid != CPU_NONE);
+
+	return(cpuid_to_vertex(cpuid));
+}
+
+
+
+/* CONFIGURATION MANAGEMENT */
+
+/*
+ * Perform initializations that allow this hub to start crosstalk support.
+ */
+void
+hub_provider_startup(devfs_handle_t hubv)
+{
+	hub_pio_init(hubv);
+	hub_dma_init(hubv);
+	hub_intr_init(hubv);
+}
+
+/*
+ * Shutdown crosstalk support from a hub.
+ */
+void
+hub_provider_shutdown(devfs_handle_t hub)
+{
+	/* TBD */
+	xtalk_provider_unregister(hub);
+}
+
+/*
+ * Check that an address is in teh real small window widget 0 space
+ * or else in the big window we're using to emulate small window 0
+ * in the kernel.
+ */
+int
+hub_check_is_widget0(void *addr)
+{
+	nasid_t nasid = NASID_GET(addr);
+
+	if (((__psunsigned_t)addr >= RAW_NODE_SWIN_BASE(nasid, 0)) &&
+	    ((__psunsigned_t)addr < RAW_NODE_SWIN_BASE(nasid, 1)))
+		return 1;
+	return 0;
+}
+
+
+/*
+ * Check that two addresses use the same widget
+ */
+int
+hub_check_window_equiv(void *addra, void *addrb)
+{
+	if (hub_check_is_widget0(addra) && hub_check_is_widget0(addrb))
+		return 1;
+
+	/* XXX - Assume this is really a small window address */
+	if (WIDGETID_GET((__psunsigned_t)addra) ==
+	    WIDGETID_GET((__psunsigned_t)addrb))
+		return 1;
+
+	return 0;
+}
+
+
+/*
+ * Determine whether two PCI addresses actually refer to the same device.
+ * This only works if both addresses are in small windows.  It's used to
+ * determine whether prom addresses refer to particular PCI devices.
+ */
+/*	
+ * XXX - This won't work as written if we ever have more than two nodes
+ * on a crossbow.  In that case, we'll need an array or partners.
+ */
+int
+hub_check_pci_equiv(void *addra, void *addrb)
+{
+	nasid_t nasida, nasidb;
+
+	/*
+	 * This is for a permanent workaround that causes us to use a
+	 * big window in place of small window 0.
+	 */
+	if (!hub_check_window_equiv(addra, addrb))
+		return 0;
+
+	/* If the offsets aren't the same, forget it. */
+	if (SWIN_WIDGETADDR((__psunsigned_t)addra) !=
+	    (SWIN_WIDGETADDR((__psunsigned_t)addrb)))
+		return 0;
+
+	/* Now, check the nasids */
+	nasida = NASID_GET(addra);
+	nasidb = NASID_GET(addrb);
+
+	ASSERT(NASID_TO_COMPACT_NODEID(nasida) != INVALID_NASID);
+	ASSERT(NASID_TO_COMPACT_NODEID(nasidb) != INVALID_NASID);
+
+	/*
+	 * Either the NASIDs must be the same or they must be crossbow
+	 * partners (on the same crossbow).
+	 */
+	return (check_nasid_equiv(nasida, nasidb));
+}
+
+/*
+ * hub_setup_prb(nasid, prbnum, credits, conveyor)
+ *
+ * 	Put a PRB into fire-and-forget mode if conveyor isn't set.  Otehrwise,
+ * 	put it into conveyor belt mode with the specified number of credits.
+ */
+void
+hub_setup_prb(nasid_t nasid, int prbnum, int credits, int conveyor)
+{
+	iprb_t prb;
+	int prb_offset;
+#ifdef IRIX
+	extern int force_fire_and_forget;
+	extern volatile int ignore_conveyor_override;
+
+	if (force_fire_and_forget && !ignore_conveyor_override)
+	    if (conveyor == HUB_PIO_CONVEYOR)
+		conveyor = HUB_PIO_FIRE_N_FORGET;
+#endif
+
+	/*
+	 * Get the current register value.
+	 */
+	prb_offset = IIO_IOPRB(prbnum);
+	prb.iprb_regval = REMOTE_HUB_L(nasid, prb_offset);
+
+	/*
+	 * Clear out some fields.
+	 */
+	prb.iprb_ovflow = 1;
+	prb.iprb_bnakctr = 0;
+	prb.iprb_anakctr = 0;
+
+	/*
+	 * Enable or disable fire-and-forget mode.
+	 */
+	prb.iprb_ff = ((conveyor == HUB_PIO_CONVEYOR) ? 0 : 1);
+
+	/*
+	 * Set the appropriate number of PIO cresits for the widget.
+	 */
+	prb.iprb_xtalkctr = credits;
+
+	/*
+	 * Store the new value to the register.
+	 */
+	REMOTE_HUB_S(nasid, prb_offset, prb.iprb_regval);
+}
+
+/*
+ * hub_set_piomode()
+ *
+ * 	Put the hub into either "PIO conveyor belt" mode or "fire-and-forget"
+ * 	mode.  To do this, we have to make absolutely sure that no PIOs
+ *	are in progress so we turn off access to all widgets for the duration
+ *	of the function.
+ * 
+ * XXX - This code should really check what kind of widget we're talking
+ * to.  Bridges can only handle three requests, but XG will do more.
+ * How many can crossbow handle to widget 0?  We're assuming 1.
+ *
+ * XXX - There is a bug in the crossbow that link reset PIOs do not
+ * return write responses.  The easiest solution to this problem is to
+ * leave widget 0 (xbow) in fire-and-forget mode at all times.  This
+ * only affects pio's to xbow registers, which should be rare.
+ */
+void
+hub_set_piomode(nasid_t nasid, int conveyor)
+{
+	hubreg_t ii_iowa;
+	int direct_connect;
+	hubii_wcr_t ii_wcr;
+	int prbnum;
+	int s, cons_lock = 0;
+
+	ASSERT(NASID_TO_COMPACT_NODEID(nasid) != INVALID_CNODEID);
+	if (nasid == get_console_nasid()) {
+		PUTBUF_LOCK(s);	
+		cons_lock = 1;
+	}
+
+	ii_iowa = REMOTE_HUB_L(nasid, IIO_OUTWIDGET_ACCESS);
+	REMOTE_HUB_S(nasid, IIO_OUTWIDGET_ACCESS, 0);
+
+	ii_wcr.wcr_reg_value = REMOTE_HUB_L(nasid, IIO_WCR);
+	direct_connect = ii_wcr.iwcr_dir_con;
+
+	if (direct_connect) {
+		/* 
+		 * Assume a bridge here.
+		 */
+		hub_setup_prb(nasid, 0, 3, conveyor);
+	} else {
+		/* 
+		 * Assume a crossbow here.
+		 */
+		hub_setup_prb(nasid, 0, 1, conveyor);
+	}
+
+	for (prbnum = HUB_WIDGET_ID_MIN; prbnum <= HUB_WIDGET_ID_MAX; prbnum++) {
+		/*
+		 * XXX - Here's where we should take the widget type into
+		 * when account assigning credits.
+		 */
+		/* Always set the PRBs in fire-and-forget mode */
+		hub_setup_prb(nasid, prbnum, 3, conveyor);
+	}
+
+#ifdef IRIX
+	/*
+	 * In direct connect mode, disable access to all widgets but 0.
+	 * Later, the prom will do this for us.
+	 */
+	if (direct_connect)
+		ii_iowa = 1;
+#endif
+
+	REMOTE_HUB_S(nasid, IIO_OUTWIDGET_ACCESS, ii_iowa);
+
+	if (cons_lock)
+	    PUTBUF_UNLOCK(s);
+}
+/* Interface to allow special drivers to set hub specific
+ * device flags.
+ * Return 0 on failure , 1 on success
+ */
+int
+hub_widget_flags_set(nasid_t		nasid,
+		     xwidgetnum_t	widget_num,
+		     hub_widget_flags_t	flags)
+{
+
+	ASSERT((flags & HUB_WIDGET_FLAGS) == flags);
+
+	if (flags & HUB_PIO_CONVEYOR) {
+		hub_setup_prb(nasid,widget_num,
+			      3,HUB_PIO_CONVEYOR); /* set the PRB in conveyor 
+						    * belt mode with 3 credits
+						    */
+	} else if (flags & HUB_PIO_FIRE_N_FORGET) {
+		hub_setup_prb(nasid,widget_num,
+			      3,HUB_PIO_FIRE_N_FORGET); /* set the PRB in fire
+							 *  and forget mode 
+							 */
+	}
+
+	return 1;
+}
+/* Interface to allow special drivers to set hub specific
+ * device flags.
+ * Return 0 on failure , 1 on success
+ */
+int
+hub_device_flags_set(devfs_handle_t	widget_vhdl,
+		     hub_widget_flags_t	flags)
+{
+	xwidget_info_t		widget_info = xwidget_info_get(widget_vhdl);
+	xwidgetnum_t		widget_num  = xwidget_info_id_get(widget_info);
+	devfs_handle_t		hub_vhdl    = xwidget_info_master_get(widget_info);
+	hubinfo_t		hub_info = 0;
+	nasid_t			nasid;
+	int			s,rv;
+
+	/* Use the nasid from the hub info hanging off the hub vertex
+	 * and widget number from the widget vertex
+	 */
+	hubinfo_get(hub_vhdl, &hub_info);
+	/* Being over cautious by grabbing a lock */
+	s 	= mutex_spinlock(&hub_info->h_bwlock);
+	nasid 	= hub_info->h_nasid;
+	rv 	= hub_widget_flags_set(nasid,widget_num,flags);
+	mutex_spinunlock(&hub_info->h_bwlock, s);
+
+	return rv;
+}
+
+#if ((defined(CONFIG_IA64_SGI_SN1) || defined(CONFIG_IA64_GENERIC)) && defined(BRINGUP))
+/* BRINGUP:  This ought to be useful for IP27 too but, for now,
+ * make it SN1 only because `ii_ixtt_u_t' is not in IP27/hubio.h
+ * (or anywhere else :-).
+ */
+int
+hubii_ixtt_set(devfs_handle_t widget_vhdl, ii_ixtt_u_t *ixtt)
+{
+	xwidget_info_t		widget_info = xwidget_info_get(widget_vhdl);
+	devfs_handle_t		hub_vhdl    = xwidget_info_master_get(widget_info);
+	hubinfo_t		hub_info = 0;
+	nasid_t			nasid;
+	int			s;
+
+	/* Use the nasid from the hub info hanging off the hub vertex
+	 * and widget number from the widget vertex
+	 */
+	hubinfo_get(hub_vhdl, &hub_info);
+	/* Being over cautious by grabbing a lock */
+	s 	= mutex_spinlock(&hub_info->h_bwlock);
+	nasid 	= hub_info->h_nasid;
+
+	REMOTE_HUB_S(nasid, IIO_IXTT, ixtt->ii_ixtt_regval);
+
+	mutex_spinunlock(&hub_info->h_bwlock, s);
+	return 0;
+}
+
+int
+hubii_ixtt_get(devfs_handle_t widget_vhdl, ii_ixtt_u_t *ixtt)
+{
+	xwidget_info_t		widget_info = xwidget_info_get(widget_vhdl);
+	devfs_handle_t		hub_vhdl    = xwidget_info_master_get(widget_info);
+	hubinfo_t		hub_info = 0;
+	nasid_t			nasid;
+	int			s;
+
+	/* Use the nasid from the hub info hanging off the hub vertex
+	 * and widget number from the widget vertex
+	 */
+	hubinfo_get(hub_vhdl, &hub_info);
+	/* Being over cautious by grabbing a lock */
+	s 	= mutex_spinlock(&hub_info->h_bwlock);
+	nasid 	= hub_info->h_nasid;
+
+	ixtt->ii_ixtt_regval = REMOTE_HUB_L(nasid, IIO_IXTT);
+
+	mutex_spinunlock(&hub_info->h_bwlock, s);
+	return 0;
+}
+#endif /* CONFIG_IA64_SGI_SN1 */
+
+/*
+ * hub_device_inquiry
+ *	Find out the xtalk widget related information stored in this 
+ *	hub's II.
+ */
+void
+hub_device_inquiry(devfs_handle_t	xbus_vhdl, xwidgetnum_t widget)
+{
+	devfs_handle_t	xconn, hub_vhdl;
+	char		widget_name[8];
+	hubreg_t	ii_iidem,ii_iiwa, ii_iowa;
+	hubinfo_t	hubinfo;
+	nasid_t		nasid;
+	int		d;
+
+	sprintf(widget_name, "%d", widget);
+	if (hwgraph_traverse(xbus_vhdl, widget_name, &xconn)
+	    != GRAPH_SUCCESS)
+		return;
+
+	hub_vhdl = device_master_get(xconn);
+	if (hub_vhdl == GRAPH_VERTEX_NONE)
+		return;
+
+	hubinfo_get(hub_vhdl, &hubinfo);
+	if (!hubinfo)
+		return;
+	
+	nasid = hubinfo->h_nasid;
+
+	ii_iidem	= REMOTE_HUB_L(nasid, IIO_IIDEM);
+	ii_iiwa 	= REMOTE_HUB_L(nasid, IIO_IIWA);
+	ii_iowa 	= REMOTE_HUB_L(nasid, IIO_IOWA);
+
+#if defined(SUPPORT_PRINTING_V_FORMAT)
+	cmn_err(CE_CONT, "Inquiry Info for %v\n", xconn);
+#else
+	cmn_err(CE_CONT, "Inquiry Info for 0x%p\n", &xconn);
+#endif
+
+	cmn_err(CE_CONT,"\tDevices shutdown [ ");
+
+	for (d = 0 ; d <= 7 ; d++)
+		if (!(ii_iidem & (IIO_IIDEM_WIDGETDEV_MASK(widget,d))))
+			cmn_err(CE_CONT, " %d", d);
+
+	cmn_err(CE_CONT,"]\n");
+
+	cmn_err(CE_CONT,
+		"\tInbound access ? %s\n",
+		ii_iiwa & IIO_IIWA_WIDGET(widget) ? "yes" : "no");
+
+	cmn_err(CE_CONT,
+		"\tOutbound access ? %s\n",
+		ii_iowa & IIO_IOWA_WIDGET(widget) ? "yes" : "no");
+
+}
+
+/*
+ * A pointer to this structure hangs off of every hub hwgraph vertex.
+ * The generic xtalk layer may indirect through it to get to this specific
+ * crosstalk bus provider.
+ */
+xtalk_provider_t hub_provider = {
+	(xtalk_piomap_alloc_f *)	hub_piomap_alloc,
+	(xtalk_piomap_free_f *)		hub_piomap_free,
+	(xtalk_piomap_addr_f *)		hub_piomap_addr,
+	(xtalk_piomap_done_f *)		hub_piomap_done,
+	(xtalk_piotrans_addr_f *)	hub_piotrans_addr,
+
+	(xtalk_dmamap_alloc_f *)	hub_dmamap_alloc,
+	(xtalk_dmamap_free_f *)		hub_dmamap_free,
+	(xtalk_dmamap_addr_f *)		hub_dmamap_addr,
+	(xtalk_dmamap_list_f *)		hub_dmamap_list,
+	(xtalk_dmamap_done_f *)		hub_dmamap_done,
+	(xtalk_dmatrans_addr_f *)	hub_dmatrans_addr,
+	(xtalk_dmatrans_list_f *)	hub_dmatrans_list,
+	(xtalk_dmamap_drain_f *)	hub_dmamap_drain,
+	(xtalk_dmaaddr_drain_f *)	hub_dmaaddr_drain,
+	(xtalk_dmalist_drain_f *)	hub_dmalist_drain,
+
+	(xtalk_intr_alloc_f *)		hub_intr_alloc,
+	(xtalk_intr_free_f *)		hub_intr_free,
+	(xtalk_intr_connect_f *)	hub_intr_connect,
+	(xtalk_intr_disconnect_f *)	hub_intr_disconnect,
+	(xtalk_intr_cpu_get_f *)	hub_intr_cpu_get,
+
+	(xtalk_provider_startup_f *)	hub_provider_startup,
+	(xtalk_provider_shutdown_f *)	hub_provider_shutdown,
+};
+
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/ip37.c linux/arch/ia64/sn/io/ip37.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/ip37.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/ip37.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,127 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyright (C) 2000 by Colin Ngam
+ */
+
+/*
+ * ip37.c
+ *	Support for IP35/IP37 machines
+ */
+
+#include <linux/types.h>
+#include <linux/config.h>
+
+#if defined(CONFIG_SGI_IP35) || defined(CONFIG_IA64_SGI_SN1) || defined(CONFIG_IA64_GENERIC)
+#include <asm/sn/sgi.h>
+#include <asm/sn/klconfig.h>
+#include <asm/sn/sn1/hubdev.h>
+#include <asm/sn/pci/bridge.h>     /* for bridge_t */
+
+
+xwidgetnum_t
+hub_widget_id(nasid_t nasid)
+{
+	hubii_wcr_t	ii_wcr;	/* the control status register */
+		
+	ii_wcr.wcr_reg_value = REMOTE_HUB_L(nasid,IIO_WCR);
+
+	printk("hub_widget_id: Found Hub Widget ID 0x%x from Register 0x%p\n", ii_wcr.wcr_fields_s.wcr_widget_id, REMOTE_HUB_ADDR(nasid, IIO_WCR));
+
+	printk("hub_widget_id: Found Hub Widget 0x%lx wcr_reg_value 0x%lx\n", REMOTE_HUB_L(nasid,IIO_WCR), ii_wcr.wcr_reg_value);
+
+	return ii_wcr.wcr_fields_s.wcr_widget_id;
+}
+
+/*
+ * get_nasid() returns the physical node id number of the caller.
+ */
+nasid_t
+get_nasid(void)
+{
+	return (nasid_t)((LOCAL_HUB_L(LB_REV_ID) & LRI_NODEID_MASK) >> LRI_NODEID_SHFT);
+}
+
+int
+get_slice(void)
+{
+	return LOCAL_HUB_L(PI_CPU_NUM);
+}
+
+int
+is_fine_dirmode(void)
+{
+	return (((LOCAL_HUB_L(LB_REV_ID) & LRI_SYSTEM_SIZE_MASK)
+		>> LRI_SYSTEM_SIZE_SHFT) == SYSTEM_SIZE_SMALL);
+
+}
+
+hubreg_t
+get_hub_chiprev(nasid_t nasid)
+{
+
+	printk("get_hub_chiprev: Hub Chip Rev 0x%lx\n",
+		(REMOTE_HUB_L(nasid, LB_REV_ID) & LRI_REV_MASK) >> LRI_REV_SHFT);
+	return ((REMOTE_HUB_L(nasid, LB_REV_ID) & LRI_REV_MASK)
+		                                         >> LRI_REV_SHFT);
+}
+
+int
+verify_snchip_rev(void)
+{
+	int hub_chip_rev;
+	int i;
+	static int min_hub_rev = 0;
+	nasid_t nasid;
+	static int first_time = 1;
+	extern int maxnodes;
+
+        
+	if (first_time) {
+	    for (i = 0; i < maxnodes; i++) {	
+		nasid = COMPACT_TO_NASID_NODEID(i);
+		hub_chip_rev = get_hub_chiprev(nasid);
+
+		if ((hub_chip_rev < min_hub_rev) || (i == 0))
+		    min_hub_rev = hub_chip_rev;
+	    }
+
+	
+	    first_time = 0;
+	}
+
+	return min_hub_rev;
+	
+}
+
+#ifdef SN1_USE_POISON_BITS
+int
+hub_bte_poison_ok(void)
+{
+	/*
+	 * For now, assume poisoning is ok. If it turns out there are chip
+	 * bugs that prevent its use in early revs, there is some neat code
+	 * to steal from the IP27 equivalent of this code.
+	 */
+
+#ifdef BRINGUP	/* temp disable BTE poisoning - might be sw bugs in this area */
+	return 0;
+#else
+	return 1;
+#endif
+}
+#endif /* SN1_USE_POISON_BITS */
+                
+
+void
+ni_reset_port(void)
+{
+	LOCAL_HUB_S(NI_RESET_ENABLE, NRE_RESETOK);
+	LOCAL_HUB_S(NI_PORT_RESET, NPR_PORTRESET | NPR_LOCALRESET);
+}
+
+#endif	/* CONFIG_SGI_IP35 || CONFIG_IA64_SGI_SN1 */
diff -u --recursive --new-file v2.4.0-prerelease/linux/arch/ia64/sn/io/klconflib.c linux/arch/ia64/sn/io/klconflib.c
--- v2.4.0-prerelease/linux/arch/ia64/sn/io/klconflib.c	Wed Dec 31 16:00:00 1969
+++ linux/arch/ia64/sn/io/klconflib.c	Thu Jan  4 13:00:15 2001
@@ -0,0 +1,1334 @@
+/* $Id$
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1992 - 1997, 2000 Silicon Graphics, Inc.
+ * Copyrig