ChangeSet@1.1570, 2004-10-22 11:03:01-02:00, paulkf@microgate.com [PATCH] serial send_break duration fix Fix tty_io.c send_break() to assert break for proper duration. If driver break_ctl() changes task state, then break may end prematurely. USB serial driver break_ctl() sends a URB, changing task state to TASK_RUNNING. Signed-off-by: Paul Fulghum -- Paul Fulghum paulkf@microgate.com ChangeSet@1.1569, 2004-10-21 14:21:15-02:00, marcelo@logos.cnet Changed EXTRAVERSION to -rc1 ChangeSet@1.1568, 2004-10-21 08:17:44-02:00, axboe@suse.de [PATCH] scsi io completion bug Hi Marcelo, There's a bug in scsi_end_request() in that it uses 'req' after having called complete() on its ->waiting member. If 'req' is on the stack, it could be invalid if the process exits right after getting the wakeup. Move the wakeup last, and also correct bogus comment on this being used for swapping. Signed-off-by: Jens Axboe ===== drivers/scsi/scsi_lib.c 1.18 vs edited ===== ChangeSet@1.1567, 2004-10-20 15:28:34-02:00, chaus@rz.uni-potsdam.de [PATCH] Fix bug in PIIX code where DMA could be turned on without proper hw configuration (bugzilla bug #3473) The function piix_config_drive_xfer_rate() consisted mainly of some nested if-then-statements. Some missing else-branches allowed for turning on DMA without proper hardware configuration. This may lead to malfunctioning harddisks transfers. However these if-then-statements were redundant as the same checks are done by a subsequent call of piix_config_drive_for_dma() which in turn calls ide_dma_speed(), where the real work is done. So the fix was to drop these statements plus some comments describing this behavior. ChangeSet@1.1566, 2004-10-18 15:34:46-02:00, benh@kernel.crashing.org [PATCH] Mikael Pettersson: PowerPC 745x coherency fix Two months ago, on August 15, I reported to the old linuxppc-dev list that the MPC745x coherency fix was needed also for the 2.4 kernels, and I included a patch for doing just that. This hasn't been included in 2.4.28-pre yet by the PPC maintainers, so I'm wondering what the holdup is? I can send it to Marcelo myself, but arch updates of this magnitude ought to come from, or at least be signed-off-by, the arch maintainers. ChangeSet@1.1565, 2004-10-17 20:13:09-02:00, moilanen@austin.ibm.com [PATCH] PPC64 build break Build break using the defconfig. Signed-off-by: Jake Moilanen ChangeSet@1.1560.1.3, 2004-10-15 02:11:17-04:00, bzolnier@gmail.com [libata piix] Fix PATA UDMA masks piix_pata_cbl_detect() handles cable detection just fine. Signed-off-by: Bartlomiej Zolnierkiewicz ChangeSet@1.1560.1.2, 2004-10-15 00:57:40-04:00, bzolnier@gmail.com [libata] do not memset() SCSI request buf in a get-reference style function fixes INQUIRY command handling for ATAPI. Signed-off-by: Bartlomiej Zolnierkiewicz ChangeSet@1.1542.11.5, 2004-10-14 23:35:56-04:00, bzolnier@elka.pw.edu.pl [PATCH] libata: PCI IDE legacy mode fix In PCI IDE legacy mode ap->port_no is incorrectly set to zero for the second port. Fix it by adding ->hard_port_no to struct ata_probe_ent and struct ata_port (per Jeff's suggestion) and teaching ata_piix.c to use it instead of ->port_no. Signed-off-by: Bartlomiej Zolnierkiewicz ChangeSet@1.1563, 2004-10-13 15:51:20-03:00, zaitcev@redhat.com [PATCH] Crash with cat /proc/bus/usb/devices and disconnect Here's a patch, I'd like to be in -pre. It is not the best fix. The 2.6 took a more fundamental approach, but I do not wish to rock the boat too much. Also, I'm not sure if 2.6 even gets it right at all, considering Fedora Core 3 bug 135171. At least this patch fixes the problem for me! :-) so I suppose better this than nothing, because getting oops otherwise is just too easy. I would like this to be in -pre. Here's the 2.6 bug (unfixed yet): https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=135171 The 2.4 bug (fixed by this patch - admittedly a contrived scenario): https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=129265 ChangeSet@1.1562, 2004-10-13 08:35:53-03:00, herbert@gondor.apana.org.au [PATCH] Fix hiddev devfs oops There is a long-standing devfs_unregister oops in hid/hiddev. It's caused by hid calling hiddev_exit before unregistering itself which in turn calls hiddev_disconnect. hiddev_exit removes the directory which contains the hiddev devices. Therefore it needs to be called after the hiddev devices have been disconnected. This patch fixes that. ===== drivers/usb/hid-core.c 1.30 vs edited ===== ChangeSet@1.1561, 2004-10-11 13:19:05-03:00, davej@redhat.com [PATCH] davej CREDITS update Wow, this has been out of date for over a year. ChangeSet@1.1560, 2004-10-11 08:05:05-03:00, sezeroz@ttnet.net.tr [PATCH] e1000 driver, gcc-3.4 inlining fix ChangeSet@1.1559, 2004-10-08 09:16:45-03:00, hugh@veritas.com [PATCH] tmpfs: fix shmem_file_write return value During some testing, it was noted that writing more than 2GB using fwrite() to a file on tmpfs would continue to retry the write() syscall until the entire filesystem was full. This patch makes the return from the write() syscall actually an ssize_t instead of an int. Signed-off-by: Robin Holt Signed-off-by: Hugh Dickins ChangeSet@1.1558, 2004-10-08 09:16:30-03:00, hugh@veritas.com [PATCH] tmpfs: stop negative dentries A tmpfs user reported increasingly slow directory reads when repeatedly creating and unlinking in a mkstemp-like way. The negative dentries accumulate alarmingly (until memory pressure finally frees them), and are just a hindrance to any in-memory filesystem. Two more users have reported this since. shmem_lookup set d_op to arrange for negative dentries to be deleted immediately. (But I failed to discover how it is that on-disk filesystems seem to keep their negative dentries within manageable bounds: this effect was gross with tmpfs or ramfs, but no problem at all with extN or reiser.) Signed-off-by: Hugh Dickins ChangeSet@1.1557, 2004-10-07 08:17:45-03:00, janitor@sternwelten.at [PATCH] menuconfig fix crash due to infinite recursion gawk(1) tells that getline "returns 0 on end of file and -1 on an error." in current script for menuconfig if getline has an error, it is still treated as true, fix _both_ of its invocations. 2.6 don't use that script anymore. fix suggestion from Aharon Robbins debian bts has 2 bugs open concerning that issue, this is the one containing belows fix: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=147469 [TM] tested. menuconfig still works for me. ;) TAG: v2.4.28-pre4