19 files changed, 5982 insertions, 0 deletions

diff --git a/fs/ocfs2/cluster/Makefile b/fs/ocfs2/cluster/Makefile
new file mode 100644
index 00000000000000..cdd162f13650b1
--- /dev/null
+++ b/fs/ocfs2/cluster/Makefile
@@ -0,0 +1,4 @@
+obj-$(CONFIG_OCFS2_FS) += ocfs2_nodemanager.o
+
+ocfs2_nodemanager-objs := heartbeat.o masklog.o sys.o nodemanager.o \
+	quorum.o tcp.o ver.o
diff --git a/fs/ocfs2/cluster/endian.h b/fs/ocfs2/cluster/endian.h
new file mode 100644
index 00000000000000..2df9082f4e3599
--- /dev/null
+++ b/fs/ocfs2/cluster/endian.h
@@ -0,0 +1,30 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_CLUSTER_ENDIAN_H
+#define OCFS2_CLUSTER_ENDIAN_H
+
+static inline void be32_add_cpu(__be32 *var, u32 val)
+{
+	*var = cpu_to_be32(be32_to_cpu(*var) + val);
+}
+
+#endif /* OCFS2_CLUSTER_ENDIAN_H */
diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
new file mode 100644
index 00000000000000..7307ba528913e6
--- /dev/null
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -0,0 +1,1797 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/jiffies.h>
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/bio.h>
+#include <linux/blkdev.h>
+#include <linux/delay.h>
+#include <linux/file.h>
+#include <linux/kthread.h>
+#include <linux/configfs.h>
+#include <linux/random.h>
+#include <linux/crc32.h>
+#include <linux/time.h>
+
+#include "heartbeat.h"
+#include "tcp.h"
+#include "nodemanager.h"
+#include "quorum.h"
+
+#include "masklog.h"
+
+
+/*
+ * The first heartbeat pass had one global thread that would serialize all hb
+ * callback calls.  This global serializing sem should only be removed once
+ * we've made sure that all callees can deal with being called concurrently
+ * from multiple hb region threads.
+ */
+static DECLARE_RWSEM(o2hb_callback_sem);
+
+/*
+ * multiple hb threads are watching multiple regions.  A node is live
+ * whenever any of the threads sees activity from the node in its region.
+ */
+static spinlock_t o2hb_live_lock = SPIN_LOCK_UNLOCKED;
+static struct list_head o2hb_live_slots[O2NM_MAX_NODES];
+static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
+static LIST_HEAD(o2hb_node_events);
+static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue);
+
+static LIST_HEAD(o2hb_all_regions);
+
+static struct o2hb_callback {
+	struct list_head list;
+} o2hb_callbacks[O2HB_NUM_CB];
+
+static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
+
+#define O2HB_DEFAULT_BLOCK_BITS       9
+
+unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
+
+/* Only sets a new threshold if there are no active regions. 
+ *
+ * No locking or otherwise interesting code is required for reading
+ * o2hb_dead_threshold as it can't change once regions are active and
+ * it's not interesting to anyone until then anyway. */
+static void o2hb_dead_threshold_set(unsigned int threshold)
+{
+	if (threshold > O2HB_MIN_DEAD_THRESHOLD) {
+		spin_lock(&o2hb_live_lock);
+		if (list_empty(&o2hb_all_regions))
+			o2hb_dead_threshold = threshold;
+		spin_unlock(&o2hb_live_lock);
+	}
+}
+
+struct o2hb_node_event {
+	struct list_head        hn_item;
+	enum o2hb_callback_type hn_event_type;
+	struct o2nm_node        *hn_node;
+	int                     hn_node_num;
+};
+
+struct o2hb_disk_slot {
+	struct o2hb_disk_heartbeat_block *ds_raw_block;
+	u8			ds_node_num;
+	u64			ds_last_time;
+	u64			ds_last_generation;
+	u16			ds_equal_samples;
+	u16			ds_changed_samples;
+	struct list_head	ds_live_item;
+};
+
+/* each thread owns a region.. when we're asked to tear down the region
+ * we ask the thread to stop, who cleans up the region */
+struct o2hb_region {
+	struct config_item	hr_item;
+
+	struct list_head	hr_all_item;
+	unsigned		hr_unclean_stop:1;
+
+	/* protected by the hr_callback_sem */
+	struct task_struct 	*hr_task;
+
+	unsigned int		hr_blocks;
+	unsigned long long	hr_start_block;
+
+	unsigned int		hr_block_bits;
+	unsigned int		hr_block_bytes;
+
+	unsigned int		hr_slots_per_page;
+	unsigned int		hr_num_pages;
+
+	struct page             **hr_slot_data;
+	struct block_device	*hr_bdev;
+	struct o2hb_disk_slot	*hr_slots;
+
+	/* let the person setting up hb wait for it to return until it
+	 * has reached a 'steady' state.  This will be fixed when we have
+	 * a more complete api that doesn't lead to this sort of fragility. */
+	atomic_t		hr_steady_iterations;
+
+	char			hr_dev_name[BDEVNAME_SIZE];
+
+	unsigned int		hr_timeout_ms;
+
+	/* randomized as the region goes up and down so that a node
+	 * recognizes a node going up and down in one iteration */
+	u64			hr_generation;
+
+	struct work_struct	hr_write_timeout_work;
+	unsigned long		hr_last_timeout_start;
+
+	/* Used during o2hb_check_slot to hold a copy of the block
+	 * being checked because we temporarily have to zero out the
+	 * crc field. */
+	struct o2hb_disk_heartbeat_block *hr_tmp_block;
+};
+
+struct o2hb_bio_wait_ctxt {
+	atomic_t          wc_num_reqs;
+	struct completion wc_io_complete;
+};
+
+static void o2hb_write_timeout(void *arg)
+{
+	struct o2hb_region *reg = arg;
+
+	mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
+	     "milliseconds\n", reg->hr_dev_name,
+	     jiffies_to_msecs(jiffies - reg->hr_last_timeout_start)); 
+	o2quo_disk_timeout();
+}
+
+static void o2hb_arm_write_timeout(struct o2hb_region *reg)
+{
+	mlog(0, "Queue write timeout for %u ms\n", O2HB_MAX_WRITE_TIMEOUT_MS);
+
+	cancel_delayed_work(&reg->hr_write_timeout_work);
+	reg->hr_last_timeout_start = jiffies;
+	schedule_delayed_work(&reg->hr_write_timeout_work,
+			      msecs_to_jiffies(O2HB_MAX_WRITE_TIMEOUT_MS));
+}
+
+static void o2hb_disarm_write_timeout(struct o2hb_region *reg)
+{
+	cancel_delayed_work(&reg->hr_write_timeout_work);
+	flush_scheduled_work();
+}
+
+static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc,
+				      unsigned int num_ios)
+{
+	atomic_set(&wc->wc_num_reqs, num_ios);
+	init_completion(&wc->wc_io_complete);
+}
+
+/* Used in error paths too */
+static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc,
+				     unsigned int num)
+{
+	/* sadly atomic_sub_and_test() isn't available on all platforms.  The
+	 * good news is that the fast path only completes one at a time */
+	while(num--) {
+		if (atomic_dec_and_test(&wc->wc_num_reqs)) {
+			BUG_ON(num > 0);
+			complete(&wc->wc_io_complete);
+		}
+	}
+}
+
+static void o2hb_wait_on_io(struct o2hb_region *reg,
+			    struct o2hb_bio_wait_ctxt *wc)
+{
+	struct address_space *mapping = reg->hr_bdev->bd_inode->i_mapping;
+
+	blk_run_address_space(mapping);
+
+	wait_for_completion(&wc->wc_io_complete);
+}
+
+static int o2hb_bio_end_io(struct bio *bio,
+			   unsigned int bytes_done,
+			   int error)
+{
+	struct o2hb_bio_wait_ctxt *wc = bio->bi_private;
+
+	if (error)
+		mlog(ML_ERROR, "IO Error %d\n", error);
+
+	if (bio->bi_size)
+		return 1;
+
+	o2hb_bio_wait_dec(wc, 1);
+	return 0;
+}
+
+/* Setup a Bio to cover I/O against num_slots slots starting at
+ * start_slot. */
+static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg,
+				      struct o2hb_bio_wait_ctxt *wc,
+				      unsigned int start_slot,
+				      unsigned int num_slots)
+{
+	int i, nr_vecs, len, first_page, last_page;
+	unsigned int vec_len, vec_start;
+	unsigned int bits = reg->hr_block_bits;
+	unsigned int spp = reg->hr_slots_per_page;
+	struct bio *bio;
+	struct page *page;
+
+	nr_vecs = (num_slots + spp - 1) / spp;
+
+	/* Testing has shown this allocation to take long enough under
+	 * GFP_KERNEL that the local node can get fenced. It would be
+	 * nicest if we could pre-allocate these bios and avoid this
+	 * all together. */
+	bio = bio_alloc(GFP_ATOMIC, nr_vecs);
+	if (!bio) {
+		mlog(ML_ERROR, "Could not alloc slots BIO!\n");
+		bio = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	/* Must put everything in 512 byte sectors for the bio... */
+	bio->bi_sector = (reg->hr_start_block + start_slot) << (bits - 9);
+	bio->bi_bdev = reg->hr_bdev;
+	bio->bi_private = wc;
+	bio->bi_end_io = o2hb_bio_end_io;
+
+	first_page = start_slot / spp;
+	last_page = first_page + nr_vecs;
+	vec_start = (start_slot << bits) % PAGE_CACHE_SIZE;
+	for(i = first_page; i < last_page; i++) {
+		page = reg->hr_slot_data[i];
+
+		vec_len = PAGE_CACHE_SIZE;
+		/* last page might be short */
+		if (((i + 1) * spp) > (start_slot + num_slots))
+			vec_len = ((num_slots + start_slot) % spp) << bits;
+		vec_len -=  vec_start;
+
+		mlog(ML_HB_BIO, "page %d, vec_len = %u, vec_start = %u\n",
+		     i, vec_len, vec_start);
+
+		len = bio_add_page(bio, page, vec_len, vec_start);
+		if (len != vec_len) {
+			bio_put(bio);
+			bio = ERR_PTR(-EIO);
+
+			mlog(ML_ERROR, "Error adding page to bio i = %d, "
+			     "vec_len = %u, len = %d\n, start = %u\n",
+			     i, vec_len, len, vec_start);
+			goto bail;
+		}
+
+		vec_start = 0;
+	}
+
+bail:
+	return bio;
+}
+
+/*
+ * Compute the maximum number of sectors the bdev can handle in one bio,
+ * as a power of two.
+ *
+ * Stolen from oracleasm, thanks Joel!
+ */
+static int compute_max_sectors(struct block_device *bdev)
+{
+	int max_pages, max_sectors, pow_two_sectors;
+
+	struct request_queue *q;
+
+	q = bdev_get_queue(bdev);
+	max_pages = q->max_sectors >> (PAGE_SHIFT - 9);
+	if (max_pages > BIO_MAX_PAGES)
+		max_pages = BIO_MAX_PAGES;
+	if (max_pages > q->max_phys_segments)
+		max_pages = q->max_phys_segments;
+	if (max_pages > q->max_hw_segments)
+		max_pages = q->max_hw_segments;
+	max_pages--; /* Handle I/Os that straddle a page */
+
+	max_sectors = max_pages << (PAGE_SHIFT - 9);
+
+	/* Why is fls() 1-based???? */
+	pow_two_sectors = 1 << (fls(max_sectors) - 1);
+
+	return pow_two_sectors;
+}
+
+static inline void o2hb_compute_request_limits(struct o2hb_region *reg,
+					       unsigned int num_slots,
+					       unsigned int *num_bios,
+					       unsigned int *slots_per_bio)
+{
+	unsigned int max_sectors, io_sectors;
+
+	max_sectors = compute_max_sectors(reg->hr_bdev);
+
+	io_sectors = num_slots << (reg->hr_block_bits - 9);
+
+	*num_bios = (io_sectors + max_sectors - 1) / max_sectors;
+	*slots_per_bio = max_sectors >> (reg->hr_block_bits - 9);
+
+	mlog(ML_HB_BIO, "My io size is %u sectors for %u slots. This "
+	     "device can handle %u sectors of I/O\n", io_sectors, num_slots,
+	     max_sectors);
+	mlog(ML_HB_BIO, "Will need %u bios holding %u slots each\n",
+	     *num_bios, *slots_per_bio);
+}
+
+static int o2hb_read_slots(struct o2hb_region *reg,
+			   unsigned int max_slots)
+{
+	unsigned int num_bios, slots_per_bio, start_slot, num_slots;
+	int i, status;
+	struct o2hb_bio_wait_ctxt wc;
+	struct bio **bios;
+	struct bio *bio;
+
+	o2hb_compute_request_limits(reg, max_slots, &num_bios, &slots_per_bio);
+
+	bios = kcalloc(num_bios, sizeof(struct bio *), GFP_KERNEL);
+	if (!bios) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		return status;
+	}
+
+	o2hb_bio_wait_init(&wc, num_bios);
+
+	num_slots = slots_per_bio;
+	for(i = 0; i < num_bios; i++) {
+		start_slot = i * slots_per_bio;
+
+		/* adjust num_slots at last bio */
+		if (max_slots < (start_slot + num_slots))
+			num_slots = max_slots - start_slot;
+
+		bio = o2hb_setup_one_bio(reg, &wc, start_slot, num_slots);
+		if (IS_ERR(bio)) {
+			o2hb_bio_wait_dec(&wc, num_bios - i);
+
+			status = PTR_ERR(bio);
+			mlog_errno(status);
+			goto bail_and_wait;
+		}
+		bios[i] = bio;
+
+		submit_bio(READ, bio);
+	}
+
+	status = 0;
+
+bail_and_wait:
+	o2hb_wait_on_io(reg, &wc);
+
+	if (bios) {
+		for(i = 0; i < num_bios; i++)
+			if (bios[i])
+				bio_put(bios[i]);
+		kfree(bios);
+	}
+
+	return status;
+}
+
+static int o2hb_issue_node_write(struct o2hb_region *reg,
+				 struct bio **write_bio,
+				 struct o2hb_bio_wait_ctxt *write_wc)
+{
+	int status;
+	unsigned int slot;
+	struct bio *bio;
+
+	o2hb_bio_wait_init(write_wc, 1);
+
+	slot = o2nm_this_node();
+
+	bio = o2hb_setup_one_bio(reg, write_wc, slot, 1);
+	if (IS_ERR(bio)) {
+		status = PTR_ERR(bio);
+		mlog_errno(status);
+		goto bail;
+	}
+
+	submit_bio(WRITE, bio);
+
+	*write_bio = bio;
+	status = 0;
+bail:
+	return status;
+}
+
+static u32 o2hb_compute_block_crc_le(struct o2hb_region *reg,
+				     struct o2hb_disk_heartbeat_block *hb_block)
+{
+	__le32 old_cksum;
+	u32 ret;
+
+	/* We want to compute the block crc with a 0 value in the
+	 * hb_cksum field. Save it off here and replace after the
+	 * crc. */
+	old_cksum = hb_block->hb_cksum;
+	hb_block->hb_cksum = 0;
+
+	ret = crc32_le(0, (unsigned char *) hb_block, reg->hr_block_bytes);
+
+	hb_block->hb_cksum = old_cksum;
+
+	return ret;
+}
+
+static void o2hb_dump_slot(struct o2hb_disk_heartbeat_block *hb_block)
+{
+	mlog(ML_ERROR, "Dump slot information: seq = 0x%"MLFx64", node = %u, "
+	     "cksum = 0x%x, generation 0x%"MLFx64"\n",
+	     le64_to_cpu(hb_block->hb_seq), hb_block->hb_node,
+	     le32_to_cpu(hb_block->hb_cksum),
+	     le64_to_cpu(hb_block->hb_generation));
+}
+
+static int o2hb_verify_crc(struct o2hb_region *reg,
+			   struct o2hb_disk_heartbeat_block *hb_block)
+{
+	u32 read, computed;
+
+	read = le32_to_cpu(hb_block->hb_cksum);
+	computed = o2hb_compute_block_crc_le(reg, hb_block);
+
+	return read == computed;
+}
+
+/* We want to make sure that nobody is heartbeating on top of us --
+ * this will help detect an invalid configuration. */
+static int o2hb_check_last_timestamp(struct o2hb_region *reg)
+{
+	int node_num, ret;
+	struct o2hb_disk_slot *slot;
+	struct o2hb_disk_heartbeat_block *hb_block;
+
+	node_num = o2nm_this_node();
+
+	ret = 1;
+	slot = &reg->hr_slots[node_num];
+	/* Don't check on our 1st timestamp */
+	if (slot->ds_last_time) {
+		hb_block = slot->ds_raw_block;
+
+		if (le64_to_cpu(hb_block->hb_seq) != slot->ds_last_time)
+			ret = 0;
+	}
+
+	return ret;
+}
+
+static inline void o2hb_prepare_block(struct o2hb_region *reg,
+				      u64 generation)
+{
+	int node_num;
+	u64 cputime;
+	struct o2hb_disk_slot *slot;
+	struct o2hb_disk_heartbeat_block *hb_block;
+
+	node_num = o2nm_this_node();
+	slot = &reg->hr_slots[node_num];
+
+	hb_block = (struct o2hb_disk_heartbeat_block *)slot->ds_raw_block;
+	memset(hb_block, 0, reg->hr_block_bytes);
+	/* TODO: time stuff */
+	cputime = CURRENT_TIME.tv_sec;
+	if (!cputime)
+		cputime = 1;
+
+	hb_block->hb_seq = cpu_to_le64(cputime);
+	hb_block->hb_node = node_num;
+	hb_block->hb_generation = cpu_to_le64(generation);
+
+	/* This step must always happen last! */
+	hb_block->hb_cksum = cpu_to_le32(o2hb_compute_block_crc_le(reg,
+								   hb_block));
+
+	mlog(ML_HB_BIO, "our node generation = 0x%"MLFx64", cksum = 0x%x\n",
+	     cpu_to_le64(generation), le32_to_cpu(hb_block->hb_cksum));
+}
+
+static void o2hb_fire_callbacks(struct o2hb_callback *hbcall,
+				struct o2nm_node *node,
+				int idx)
+{
+	struct list_head *iter;
+	struct o2hb_callback_func *f;
+
+	list_for_each(iter, &hbcall->list) {
+		f = list_entry(iter, struct o2hb_callback_func, hc_item);
+		mlog(ML_HEARTBEAT, "calling funcs %p\n", f);
+		(f->hc_func)(node, idx, f->hc_data);
+	}
+}
+
+/* Will run the list in order until we process the passed event */
+static void o2hb_run_event_list(struct o2hb_node_event *queued_event)
+{
+	int empty;
+	struct o2hb_callback *hbcall;
+	struct o2hb_node_event *event;
+
+	spin_lock(&o2hb_live_lock);
+	empty = list_empty(&queued_event->hn_item);
+	spin_unlock(&o2hb_live_lock);
+	if (empty)
+		return;
+
+	/* Holding callback sem assures we don't alter the callback
+	 * lists when doing this, and serializes ourselves with other
+	 * processes wanting callbacks. */
+	down_write(&o2hb_callback_sem);
+
+	spin_lock(&o2hb_live_lock);
+	while (!list_empty(&o2hb_node_events)
+	       && !list_empty(&queued_event->hn_item)) {
+		event = list_entry(o2hb_node_events.next,
+				   struct o2hb_node_event,
+				   hn_item);
+		list_del_init(&event->hn_item);
+		spin_unlock(&o2hb_live_lock);
+
+		mlog(ML_HEARTBEAT, "Node %s event for %d\n",
+		     event->hn_event_type == O2HB_NODE_UP_CB ? "UP" : "DOWN",
+		     event->hn_node_num);
+
+		hbcall = hbcall_from_type(event->hn_event_type);
+
+		/* We should *never* have gotten on to the list with a
+		 * bad type... This isn't something that we should try
+		 * to recover from. */
+		BUG_ON(IS_ERR(hbcall));
+
+		o2hb_fire_callbacks(hbcall, event->hn_node, event->hn_node_num);
+
+		spin_lock(&o2hb_live_lock);
+	}
+	spin_unlock(&o2hb_live_lock);
+
+	up_write(&o2hb_callback_sem);
+}
+
+static void o2hb_queue_node_event(struct o2hb_node_event *event,
+				  enum o2hb_callback_type type,
+				  struct o2nm_node *node,
+				  int node_num)
+{
+	assert_spin_locked(&o2hb_live_lock);
+
+	event->hn_event_type = type;
+	event->hn_node = node;
+	event->hn_node_num = node_num;
+
+	mlog(ML_HEARTBEAT, "Queue node %s event for node %d\n",
+	     type == O2HB_NODE_UP_CB ? "UP" : "DOWN", node_num);
+
+	list_add_tail(&event->hn_item, &o2hb_node_events);
+}
+
+static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot)
+{
+	struct o2hb_node_event event =
+		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
+	struct o2nm_node *node;
+
+	node = o2nm_get_node_by_num(slot->ds_node_num);
+	if (!node)
+		return;
+
+	spin_lock(&o2hb_live_lock);
+	if (!list_empty(&slot->ds_live_item)) {
+		mlog(ML_HEARTBEAT, "Shutdown, node %d leaves region\n",
+		     slot->ds_node_num);
+
+		list_del_init(&slot->ds_live_item);
+
+		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
+			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
+
+			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
+					      slot->ds_node_num);
+		}
+	}
+	spin_unlock(&o2hb_live_lock);
+
+	o2hb_run_event_list(&event);
+
+	o2nm_node_put(node);
+}
+
+static int o2hb_check_slot(struct o2hb_region *reg,
+			   struct o2hb_disk_slot *slot)
+{
+	int changed = 0, gen_changed = 0;
+	struct o2hb_node_event event =
+		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
+	struct o2nm_node *node;
+	struct o2hb_disk_heartbeat_block *hb_block = reg->hr_tmp_block;
+	u64 cputime;
+
+	memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes);
+
+	/* Is this correct? Do we assume that the node doesn't exist
+	 * if we're not configured for him? */
+	node = o2nm_get_node_by_num(slot->ds_node_num);
+	if (!node)
+		return 0;
+
+	if (!o2hb_verify_crc(reg, hb_block)) {
+		/* all paths from here will drop o2hb_live_lock for
+		 * us. */
+		spin_lock(&o2hb_live_lock);
+
+		/* Don't print an error on the console in this case -
+		 * a freshly formatted heartbeat area will not have a
+		 * crc set on it. */
+		if (list_empty(&slot->ds_live_item))
+			goto out;
+
+		/* The node is live but pushed out a bad crc. We
+		 * consider it a transient miss but don't populate any
+		 * other values as they may be junk. */
+		mlog(ML_ERROR, "Node %d has written a bad crc to %s\n",
+		     slot->ds_node_num, reg->hr_dev_name);
+		o2hb_dump_slot(hb_block);
+
+		slot->ds_equal_samples++;
+		goto fire_callbacks;
+	}
+
+	/* we don't care if these wrap.. the state transitions below
+	 * clear at the right places */
+	cputime = le64_to_cpu(hb_block->hb_seq);
+	if (slot->ds_last_time != cputime)
+		slot->ds_changed_samples++;
+	else
+		slot->ds_equal_samples++;
+	slot->ds_last_time = cputime;
+
+	/* The node changed heartbeat generations. We assume this to
+	 * mean it dropped off but came back before we timed out. We
+	 * want to consider it down for the time being but don't want
+	 * to lose any changed_samples state we might build up to
+	 * considering it live again. */
+	if (slot->ds_last_generation != le64_to_cpu(hb_block->hb_generation)) {
+		gen_changed = 1;
+		slot->ds_equal_samples = 0;
+		mlog(ML_HEARTBEAT, "Node %d changed generation (0x%"MLFx64" "
+		     "to 0x%"MLFx64")\n", slot->ds_node_num,
+		     slot->ds_last_generation,
+		     le64_to_cpu(hb_block->hb_generation));
+	}
+
+	slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
+
+	mlog(ML_HEARTBEAT, "Slot %d gen 0x%"MLFx64" cksum 0x%x "
+	     "seq %"MLFu64" last %"MLFu64" changed %u equal %u\n",
+	     slot->ds_node_num, slot->ds_last_generation,
+	     le32_to_cpu(hb_block->hb_cksum), le64_to_cpu(hb_block->hb_seq), 
+	     slot->ds_last_time, slot->ds_changed_samples,
+	     slot->ds_equal_samples);
+
+	spin_lock(&o2hb_live_lock);
+
+fire_callbacks:
+	/* dead nodes only come to life after some number of
+	 * changes at any time during their dead time */
+	if (list_empty(&slot->ds_live_item) &&
+	    slot->ds_changed_samples >= O2HB_LIVE_THRESHOLD) {
+		mlog(ML_HEARTBEAT, "Node %d (id 0x%"MLFx64") joined my "
+		     "region\n", slot->ds_node_num, slot->ds_last_generation);
+
+		/* first on the list generates a callback */
+		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
+			set_bit(slot->ds_node_num, o2hb_live_node_bitmap);
+
+			o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node,
+					      slot->ds_node_num);
+
+			changed = 1;
+		}
+
+		list_add_tail(&slot->ds_live_item,
+			      &o2hb_live_slots[slot->ds_node_num]);
+
+		slot->ds_equal_samples = 0;
+		goto out;
+	}
+
+	/* if the list is dead, we're done.. */
+	if (list_empty(&slot->ds_live_item))
+		goto out;
+
+	/* live nodes only go dead after enough consequtive missed
+	 * samples..  reset the missed counter whenever we see
+	 * activity */
+	if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) {
+		mlog(ML_HEARTBEAT, "Node %d left my region\n",
+		     slot->ds_node_num);
+
+		/* last off the live_slot generates a callback */
+		list_del_init(&slot->ds_live_item);
+		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
+			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
+
+			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
+					      slot->ds_node_num);
+
+			changed = 1;
+		}
+
+		/* We don't clear this because the node is still
+		 * actually writing new blocks. */
+		if (!gen_changed)
+			slot->ds_changed_samples = 0;
+		goto out;
+	}
+	if (slot->ds_changed_samples) {
+		slot->ds_changed_samples = 0;
+		slot->ds_equal_samples = 0;
+	}
+out:
+	spin_unlock(&o2hb_live_lock);
+
+	o2hb_run_event_list(&event);
+
+	o2nm_node_put(node);
+	return changed;
+}
+
+/* This could be faster if we just implmented a find_last_bit, but I
+ * don't think the circumstances warrant it. */
+static int o2hb_highest_node(unsigned long *nodes,
+			     int numbits)
+{
+	int highest, node;
+
+	highest = numbits;
+	node = -1;
+	while ((node = find_next_bit(nodes, numbits, node + 1)) != -1) {
+		if (node >= numbits)
+			break;
+
+		highest = node;
+	}
+
+	return highest;
+}
+
+static void o2hb_do_disk_heartbeat(struct o2hb_region *reg)
+{
+	int i, ret, highest_node, change = 0;
+	unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)];
+	struct bio *write_bio;
+	struct o2hb_bio_wait_ctxt write_wc;
+
+	if (o2nm_configured_node_map(configured_nodes, sizeof(configured_nodes)))
+		return;
+
+	highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES);
+	if (highest_node >= O2NM_MAX_NODES) {
+		mlog(ML_NOTICE, "ocfs2_heartbeat: no configured nodes found!\n");
+		return;
+	}
+
+	/* No sense in reading the slots of nodes that don't exist
+	 * yet. Of course, if the node definitions have holes in them
+	 * then we're reading an empty slot anyway... Consider this
+	 * best-effort. */
+	ret = o2hb_read_slots(reg, highest_node + 1);
+	if (ret < 0) {
+		mlog_errno(ret);
+		return;
+	}
+
+	/* With an up to date view of the slots, we can check that no
+	 * other node has been improperly configured to heartbeat in
+	 * our slot. */
+	if (!o2hb_check_last_timestamp(reg))
+		mlog(ML_ERROR, "Device \"%s\": another node is heartbeating "
+		     "in our slot!\n", reg->hr_dev_name);
+
+	/* fill in the proper info for our next heartbeat */
+	o2hb_prepare_block(reg, reg->hr_generation);
+
+	/* And fire off the write. Note that we don't wait on this I/O
+	 * until later. */
+	ret = o2hb_issue_node_write(reg, &write_bio, &write_wc);
+	if (ret < 0) {
+		mlog_errno(ret);
+		return;
+	}
+
+	i = -1;
+	while((i = find_next_bit(configured_nodes, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
+
+		change |= o2hb_check_slot(reg, &reg->hr_slots[i]);
+	}
+
+	/*
+	 * We have to be sure we've advertised ourselves on disk
+	 * before we can go to steady state.  This ensures that
+	 * people we find in our steady state have seen us.
+	 */
+	o2hb_wait_on_io(reg, &write_wc);
+	bio_put(write_bio);
+	o2hb_arm_write_timeout(reg);
+
+	/* let the person who launched us know when things are steady */
+	if (!change && (atomic_read(&reg->hr_steady_iterations) != 0)) {
+		if (atomic_dec_and_test(&reg->hr_steady_iterations))
+			wake_up(&o2hb_steady_queue);
+	}
+}
+
+/* Subtract b from a, storing the result in a. a *must* have a larger
+ * value than b. */
+static void o2hb_tv_subtract(struct timeval *a,
+			     struct timeval *b)
+{
+	/* just return 0 when a is after b */
+	if (a->tv_sec < b->tv_sec ||
+	    (a->tv_sec == b->tv_sec && a->tv_usec < b->tv_usec)) {
+		a->tv_sec = 0;
+		a->tv_usec = 0;
+		return;
+	}
+
+	a->tv_sec -= b->tv_sec;
+	a->tv_usec -= b->tv_usec;
+	while ( a->tv_usec < 0 ) {
+		a->tv_sec--;
+		a->tv_usec += 1000000;
+	}
+}
+
+static unsigned int o2hb_elapsed_msecs(struct timeval *start,
+				       struct timeval *end)
+{
+	struct timeval res = *end;
+
+	o2hb_tv_subtract(&res, start);
+
+	return res.tv_sec * 1000 + res.tv_usec / 1000;
+}
+
+/*
+ * we ride the region ref that the region dir holds.  before the region
+ * dir is removed and drops it ref it will wait to tear down this
+ * thread.
+ */
+static int o2hb_thread(void *data)
+{
+	int i, ret;
+	struct o2hb_region *reg = data;
+	struct bio *write_bio;
+	struct o2hb_bio_wait_ctxt write_wc;
+	struct timeval before_hb, after_hb;
+	unsigned int elapsed_msec;
+
+	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n");
+
+	set_user_nice(current, -20);
+
+	while (!kthread_should_stop() && !reg->hr_unclean_stop) {
+		/* We track the time spent inside
+		 * o2hb_do_disk_heartbeat so that we avoid more then
+		 * hr_timeout_ms between disk writes. On busy systems
+		 * this should result in a heartbeat which is less
+		 * likely to time itself out. */
+		do_gettimeofday(&before_hb);
+
+		o2hb_do_disk_heartbeat(reg);
+
+		do_gettimeofday(&after_hb);
+		elapsed_msec = o2hb_elapsed_msecs(&before_hb, &after_hb);
+
+		mlog(0, "start = %lu.%lu, end = %lu.%lu, msec = %u\n",
+		     before_hb.tv_sec, before_hb.tv_usec,
+		     after_hb.tv_sec, after_hb.tv_usec, elapsed_msec);
+
+		if (elapsed_msec < reg->hr_timeout_ms) {
+			/* the kthread api has blocked signals for us so no
+			 * need to record the return value. */
+			msleep_interruptible(reg->hr_timeout_ms - elapsed_msec);
+		}
+	}
+
+	o2hb_disarm_write_timeout(reg);
+
+	/* unclean stop is only used in very bad situation */
+	for(i = 0; !reg->hr_unclean_stop && i < reg->hr_blocks; i++)
+		o2hb_shutdown_slot(&reg->hr_slots[i]);
+
+	/* Explicit down notification - avoid forcing the other nodes
+	 * to timeout on this region when we could just as easily
+	 * write a clear generation - thus indicating to them that
+	 * this node has left this region.
+	 *
+	 * XXX: Should we skip this on unclean_stop? */
+	o2hb_prepare_block(reg, 0);
+	ret = o2hb_issue_node_write(reg, &write_bio, &write_wc);
+	if (ret == 0) {
+		o2hb_wait_on_io(reg, &write_wc);
+		bio_put(write_bio);
+	} else {
+		mlog_errno(ret);
+	}
+
+	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread exiting\n");
+
+	return 0;
+}
+
+void o2hb_init(void)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(o2hb_callbacks); i++)
+		INIT_LIST_HEAD(&o2hb_callbacks[i].list);
+
+	for (i = 0; i < ARRAY_SIZE(o2hb_live_slots); i++)
+		INIT_LIST_HEAD(&o2hb_live_slots[i]);
+
+	INIT_LIST_HEAD(&o2hb_node_events);
+
+	memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap));
+}
+
+/* if we're already in a callback then we're already serialized by the sem */
+static void o2hb_fill_node_map_from_callback(unsigned long *map,
+					     unsigned bytes)
+{
+	BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long)));
+
+	memcpy(map, &o2hb_live_node_bitmap, bytes);
+}
+
+/*
+ * get a map of all nodes that are heartbeating in any regions
+ */
+void o2hb_fill_node_map(unsigned long *map, unsigned bytes)
+{
+	/* callers want to serialize this map and callbacks so that they
+	 * can trust that they don't miss nodes coming to the party */
+	down_read(&o2hb_callback_sem);
+	spin_lock(&o2hb_live_lock);
+	o2hb_fill_node_map_from_callback(map, bytes);
+	spin_unlock(&o2hb_live_lock);
+	up_read(&o2hb_callback_sem);
+}
+EXPORT_SYMBOL_GPL(o2hb_fill_node_map);
+
+/*
+ * heartbeat configfs bits.  The heartbeat set is a default set under
+ * the cluster set in nodemanager.c.
+ */
+
+static struct o2hb_region *to_o2hb_region(struct config_item *item)
+{
+	return item ? container_of(item, struct o2hb_region, hr_item) : NULL;
+}
+
+/* drop_item only drops its ref after killing the thread, nothing should
+ * be using the region anymore.  this has to clean up any state that
+ * attributes might have built up. */
+static void o2hb_region_release(struct config_item *item)
+{
+	int i;
+	struct page *page;
+	struct o2hb_region *reg = to_o2hb_region(item);
+
+	if (reg->hr_tmp_block)
+		kfree(reg->hr_tmp_block);
+
+	if (reg->hr_slot_data) {
+		for (i = 0; i < reg->hr_num_pages; i++) {
+			page = reg->hr_slot_data[i];
+			if (page)
+				__free_page(page);
+		}
+		kfree(reg->hr_slot_data);
+	}
+
+	if (reg->hr_bdev)
+		blkdev_put(reg->hr_bdev);
+
+	if (reg->hr_slots)
+		kfree(reg->hr_slots);
+
+	spin_lock(&o2hb_live_lock);
+	list_del(&reg->hr_all_item);
+	spin_unlock(&o2hb_live_lock);
+
+	kfree(reg);
+}
+
+static int o2hb_read_block_input(struct o2hb_region *reg,
+				 const char *page,
+				 size_t count,
+				 unsigned long *ret_bytes,
+				 unsigned int *ret_bits)
+{
+	unsigned long bytes;
+	char *p = (char *)page;
+
+	bytes = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	/* Heartbeat and fs min / max block sizes are the same. */
+	if (bytes > 4096 || bytes < 512)
+		return -ERANGE;
+	if (hweight16(bytes) != 1)
+		return -EINVAL;
+
+	if (ret_bytes)
+		*ret_bytes = bytes;
+	if (ret_bits)
+		*ret_bits = ffs(bytes) - 1;
+
+	return 0;
+}
+
+static ssize_t o2hb_region_block_bytes_read(struct o2hb_region *reg,
+					    char *page)
+{
+	return sprintf(page, "%u\n", reg->hr_block_bytes);
+}
+
+static ssize_t o2hb_region_block_bytes_write(struct o2hb_region *reg,
+					     const char *page,
+					     size_t count)
+{
+	int status;
+	unsigned long block_bytes;
+	unsigned int block_bits;
+
+	if (reg->hr_bdev)
+		return -EINVAL;
+
+	status = o2hb_read_block_input(reg, page, count,
+				       &block_bytes, &block_bits);
+	if (status)
+		return status;
+
+	reg->hr_block_bytes = (unsigned int)block_bytes;
+	reg->hr_block_bits = block_bits;
+
+	return count;
+}
+
+static ssize_t o2hb_region_start_block_read(struct o2hb_region *reg,
+					    char *page)
+{
+	return sprintf(page, "%llu\n", reg->hr_start_block);
+}
+
+static ssize_t o2hb_region_start_block_write(struct o2hb_region *reg,
+					     const char *page,
+					     size_t count)
+{
+	unsigned long long tmp;
+	char *p = (char *)page;
+
+	if (reg->hr_bdev)
+		return -EINVAL;
+
+	tmp = simple_strtoull(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	reg->hr_start_block = tmp;
+
+	return count;
+}
+
+static ssize_t o2hb_region_blocks_read(struct o2hb_region *reg,
+				       char *page)
+{
+	return sprintf(page, "%d\n", reg->hr_blocks);
+}
+
+static ssize_t o2hb_region_blocks_write(struct o2hb_region *reg,
+					const char *page,
+					size_t count)
+{
+	unsigned long tmp;
+	char *p = (char *)page;
+
+	if (reg->hr_bdev)
+		return -EINVAL;
+
+	tmp = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	if (tmp > O2NM_MAX_NODES || tmp == 0)
+		return -ERANGE;
+
+	reg->hr_blocks = (unsigned int)tmp;
+
+	return count;
+}
+
+static ssize_t o2hb_region_dev_read(struct o2hb_region *reg,
+				    char *page)
+{
+	unsigned int ret = 0;
+
+	if (reg->hr_bdev)
+		ret = sprintf(page, "%s\n", reg->hr_dev_name);
+
+	return ret;
+}
+
+static void o2hb_init_region_params(struct o2hb_region *reg)
+{
+	reg->hr_slots_per_page = PAGE_CACHE_SIZE >> reg->hr_block_bits;
+	reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS;
+
+	mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n",
+	     reg->hr_start_block, reg->hr_blocks);
+	mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n",
+	     reg->hr_block_bytes, reg->hr_block_bits);
+	mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms);
+	mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold);
+}
+
+static int o2hb_map_slot_data(struct o2hb_region *reg)
+{
+	int i, j;
+	unsigned int last_slot;
+	unsigned int spp = reg->hr_slots_per_page;
+	struct page *page;
+	char *raw;
+	struct o2hb_disk_slot *slot;
+
+	reg->hr_tmp_block = kmalloc(reg->hr_block_bytes, GFP_KERNEL);
+	if (reg->hr_tmp_block == NULL) {
+		mlog_errno(-ENOMEM);
+		return -ENOMEM;
+	}
+
+	reg->hr_slots = kcalloc(reg->hr_blocks,
+				sizeof(struct o2hb_disk_slot), GFP_KERNEL);
+	if (reg->hr_slots == NULL) {
+		mlog_errno(-ENOMEM);
+		return -ENOMEM;
+	}
+
+	for(i = 0; i < reg->hr_blocks; i++) {
+		slot = &reg->hr_slots[i];
+		slot->ds_node_num = i;
+		INIT_LIST_HEAD(&slot->ds_live_item);
+		slot->ds_raw_block = NULL;
+	}
+
+	reg->hr_num_pages = (reg->hr_blocks + spp - 1) / spp;
+	mlog(ML_HEARTBEAT, "Going to require %u pages to cover %u blocks "
+			   "at %u blocks per page\n",
+	     reg->hr_num_pages, reg->hr_blocks, spp);
+
+	reg->hr_slot_data = kcalloc(reg->hr_num_pages, sizeof(struct page *),
+				    GFP_KERNEL);
+	if (!reg->hr_slot_data) {
+		mlog_errno(-ENOMEM);
+		return -ENOMEM;
+	}
+
+	for(i = 0; i < reg->hr_num_pages; i++) {
+		page = alloc_page(GFP_KERNEL);
+		if (!page) {
+			mlog_errno(-ENOMEM);
+			return -ENOMEM;
+		}
+
+		reg->hr_slot_data[i] = page;
+
+		last_slot = i * spp;
+		raw = page_address(page);
+		for (j = 0;
+		     (j < spp) && ((j + last_slot) < reg->hr_blocks);
+		     j++) {
+			BUG_ON((j + last_slot) >= reg->hr_blocks);
+
+			slot = &reg->hr_slots[j + last_slot];
+			slot->ds_raw_block =
+				(struct o2hb_disk_heartbeat_block *) raw;
+
+			raw += reg->hr_block_bytes;
+		}
+	}
+
+	return 0;
+}
+
+/* Read in all the slots available and populate the tracking
+ * structures so that we can start with a baseline idea of what's
+ * there. */
+static int o2hb_populate_slot_data(struct o2hb_region *reg)
+{
+	int ret, i;
+	struct o2hb_disk_slot *slot;
+	struct o2hb_disk_heartbeat_block *hb_block;
+
+	mlog_entry_void();
+
+	ret = o2hb_read_slots(reg, reg->hr_blocks);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	/* We only want to get an idea of the values initially in each
+	 * slot, so we do no verification - o2hb_check_slot will
+	 * actually determine if each configured slot is valid and
+	 * whether any values have changed. */
+	for(i = 0; i < reg->hr_blocks; i++) {
+		slot = &reg->hr_slots[i];
+		hb_block = (struct o2hb_disk_heartbeat_block *) slot->ds_raw_block;
+
+		/* Only fill the values that o2hb_check_slot uses to
+		 * determine changing slots */
+		slot->ds_last_time = le64_to_cpu(hb_block->hb_seq);
+		slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
+	}
+
+out:
+	mlog_exit(ret);
+	return ret;
+}
+
+/* this is acting as commit; we set up all of hr_bdev and hr_task or nothing */
+static ssize_t o2hb_region_dev_write(struct o2hb_region *reg,
+				     const char *page,
+				     size_t count)
+{
+	long fd;
+	int sectsize;
+	char *p = (char *)page;
+	struct file *filp = NULL;
+	struct inode *inode = NULL;
+	ssize_t ret = -EINVAL;
+
+	if (reg->hr_bdev)
+		goto out;
+
+	/* We can't heartbeat without having had our node number
+	 * configured yet. */
+	if (o2nm_this_node() == O2NM_MAX_NODES)
+		goto out;
+
+	fd = simple_strtol(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		goto out;
+
+	if (fd < 0 || fd >= INT_MAX)
+		goto out;
+
+	filp = fget(fd);
+	if (filp == NULL)
+		goto out;
+
+	if (reg->hr_blocks == 0 || reg->hr_start_block == 0 ||
+	    reg->hr_block_bytes == 0)
+		goto out;
+
+	inode = igrab(filp->f_mapping->host);
+	if (inode == NULL)
+		goto out;
+
+	if (!S_ISBLK(inode->i_mode))
+		goto out;
+
+	reg->hr_bdev = I_BDEV(filp->f_mapping->host);
+	ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, 0);
+	if (ret) {
+		reg->hr_bdev = NULL;
+		goto out;
+	}
+	inode = NULL;
+
+	bdevname(reg->hr_bdev, reg->hr_dev_name);
+
+	sectsize = bdev_hardsect_size(reg->hr_bdev);
+	if (sectsize != reg->hr_block_bytes) {
+		mlog(ML_ERROR,
+		     "blocksize %u incorrect for device, expected %d",
+		     reg->hr_block_bytes, sectsize);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	o2hb_init_region_params(reg);
+
+	/* Generation of zero is invalid */
+	do {
+		get_random_bytes(&reg->hr_generation,
+				 sizeof(reg->hr_generation));
+	} while (reg->hr_generation == 0);
+
+	ret = o2hb_map_slot_data(reg);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = o2hb_populate_slot_data(reg);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	INIT_WORK(&reg->hr_write_timeout_work, o2hb_write_timeout, reg);
+
+	/*
+	 * A node is considered live after it has beat LIVE_THRESHOLD
+	 * times.  We're not steady until we've given them a chance
+	 * _after_ our first read.
+	 */
+	atomic_set(&reg->hr_steady_iterations, O2HB_LIVE_THRESHOLD + 1);
+
+	reg->hr_task = kthread_run(o2hb_thread, reg, "o2hb-%s",
+				   reg->hr_item.ci_name);
+	if (IS_ERR(reg->hr_task)) {
+		ret = PTR_ERR(reg->hr_task);
+		mlog_errno(ret);
+		reg->hr_task = NULL;
+		goto out;
+	}
+
+	ret = wait_event_interruptible(o2hb_steady_queue,
+				atomic_read(&reg->hr_steady_iterations) == 0);
+	if (ret) {
+		kthread_stop(reg->hr_task);
+		reg->hr_task = NULL;
+		goto out;
+	}
+
+	ret = count;
+out:
+	if (filp)
+		fput(filp);
+	if (inode)
+		iput(inode);
+	if (ret < 0) {
+		if (reg->hr_bdev) {
+			blkdev_put(reg->hr_bdev);
+			reg->hr_bdev = NULL;
+		}
+	}
+	return ret;
+}
+
+struct o2hb_region_attribute {
+	struct configfs_attribute attr;
+	ssize_t (*show)(struct o2hb_region *, char *);
+	ssize_t (*store)(struct o2hb_region *, const char *, size_t);
+};
+
+static struct o2hb_region_attribute o2hb_region_attr_block_bytes = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "block_bytes",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2hb_region_block_bytes_read,
+	.store	= o2hb_region_block_bytes_write,
+};
+
+static struct o2hb_region_attribute o2hb_region_attr_start_block = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "start_block",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2hb_region_start_block_read,
+	.store	= o2hb_region_start_block_write,
+};
+
+static struct o2hb_region_attribute o2hb_region_attr_blocks = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "blocks",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2hb_region_blocks_read,
+	.store	= o2hb_region_blocks_write,
+};
+
+static struct o2hb_region_attribute o2hb_region_attr_dev = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "dev",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2hb_region_dev_read,
+	.store	= o2hb_region_dev_write,
+};
+
+static struct configfs_attribute *o2hb_region_attrs[] = {
+	&o2hb_region_attr_block_bytes.attr,
+	&o2hb_region_attr_start_block.attr,
+	&o2hb_region_attr_blocks.attr,
+	&o2hb_region_attr_dev.attr,
+	NULL,
+};
+
+static ssize_t o2hb_region_show(struct config_item *item,
+				struct configfs_attribute *attr,
+				char *page)
+{
+	struct o2hb_region *reg = to_o2hb_region(item);
+	struct o2hb_region_attribute *o2hb_region_attr =
+		container_of(attr, struct o2hb_region_attribute, attr);
+	ssize_t ret = 0;
+
+	if (o2hb_region_attr->show)
+		ret = o2hb_region_attr->show(reg, page);
+	return ret;
+}
+
+static ssize_t o2hb_region_store(struct config_item *item,
+				 struct configfs_attribute *attr,
+				 const char *page, size_t count)
+{
+	struct o2hb_region *reg = to_o2hb_region(item);
+	struct o2hb_region_attribute *o2hb_region_attr =
+		container_of(attr, struct o2hb_region_attribute, attr);
+	ssize_t ret = -EINVAL;
+
+	if (o2hb_region_attr->store)
+		ret = o2hb_region_attr->store(reg, page, count);
+	return ret;
+}
+
+static struct configfs_item_operations o2hb_region_item_ops = {
+	.release		= o2hb_region_release,
+	.show_attribute		= o2hb_region_show,
+	.store_attribute	= o2hb_region_store,
+};
+
+static struct config_item_type o2hb_region_type = {
+	.ct_item_ops	= &o2hb_region_item_ops,
+	.ct_attrs	= o2hb_region_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* heartbeat set */
+
+struct o2hb_heartbeat_group {
+	struct config_group hs_group;
+	/* some stuff? */
+};
+
+static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group *group)
+{
+	return group ?
+		container_of(group, struct o2hb_heartbeat_group, hs_group)
+		: NULL;
+}
+
+static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group,
+							  const char *name)
+{
+	struct o2hb_region *reg = NULL;
+	struct config_item *ret = NULL;
+
+	reg = kcalloc(1, sizeof(struct o2hb_region), GFP_KERNEL);
+	if (reg == NULL)
+		goto out; /* ENOMEM */
+
+	config_item_init_type_name(&reg->hr_item, name, &o2hb_region_type);
+
+	ret = &reg->hr_item;
+
+	spin_lock(&o2hb_live_lock);
+	list_add_tail(&reg->hr_all_item, &o2hb_all_regions);
+	spin_unlock(&o2hb_live_lock);
+out:
+	if (ret == NULL)
+		kfree(reg);
+
+	return ret;
+}
+
+static void o2hb_heartbeat_group_drop_item(struct config_group *group,
+					   struct config_item *item)
+{
+	struct o2hb_region *reg = to_o2hb_region(item);
+
+	/* stop the thread when the user removes the region dir */
+	if (reg->hr_task) {
+		kthread_stop(reg->hr_task);
+		reg->hr_task = NULL;
+	}
+
+	config_item_put(item);
+}
+
+struct o2hb_heartbeat_group_attribute {
+	struct configfs_attribute attr;
+	ssize_t (*show)(struct o2hb_heartbeat_group *, char *);
+	ssize_t (*store)(struct o2hb_heartbeat_group *, const char *, size_t);
+};
+
+static ssize_t o2hb_heartbeat_group_show(struct config_item *item,
+					 struct configfs_attribute *attr,
+					 char *page)
+{
+	struct o2hb_heartbeat_group *reg = to_o2hb_heartbeat_group(to_config_group(item));
+	struct o2hb_heartbeat_group_attribute *o2hb_heartbeat_group_attr =
+		container_of(attr, struct o2hb_heartbeat_group_attribute, attr);
+	ssize_t ret = 0;
+
+	if (o2hb_heartbeat_group_attr->show)
+		ret = o2hb_heartbeat_group_attr->show(reg, page);
+	return ret;
+}
+
+static ssize_t o2hb_heartbeat_group_store(struct config_item *item,
+					  struct configfs_attribute *attr,
+					  const char *page, size_t count)
+{
+	struct o2hb_heartbeat_group *reg = to_o2hb_heartbeat_group(to_config_group(item));
+	struct o2hb_heartbeat_group_attribute *o2hb_heartbeat_group_attr =
+		container_of(attr, struct o2hb_heartbeat_group_attribute, attr);
+	ssize_t ret = -EINVAL;
+
+	if (o2hb_heartbeat_group_attr->store)
+		ret = o2hb_heartbeat_group_attr->store(reg, page, count);
+	return ret;
+}
+
+static ssize_t o2hb_heartbeat_group_threshold_show(struct o2hb_heartbeat_group *group,
+						     char *page)
+{
+	return sprintf(page, "%u\n", o2hb_dead_threshold);
+}
+
+static ssize_t o2hb_heartbeat_group_threshold_store(struct o2hb_heartbeat_group *group,
+						    const char *page,
+						    size_t count)
+{
+	unsigned long tmp;
+	char *p = (char *)page;
+
+	tmp = simple_strtoul(p, &p, 10);
+	if (!p || (*p && (*p != '\n')))
+                return -EINVAL;
+
+	/* this will validate ranges for us. */
+	o2hb_dead_threshold_set((unsigned int) tmp);
+
+	return count;
+}
+
+static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "dead_threshold",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2hb_heartbeat_group_threshold_show,
+	.store	= o2hb_heartbeat_group_threshold_store,
+};
+
+static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
+	&o2hb_heartbeat_group_attr_threshold.attr,
+	NULL,
+};
+
+static struct configfs_item_operations o2hb_hearbeat_group_item_ops = {
+	.show_attribute		= o2hb_heartbeat_group_show,
+	.store_attribute	= o2hb_heartbeat_group_store,
+};
+
+static struct configfs_group_operations o2hb_heartbeat_group_group_ops = {
+	.make_item	= o2hb_heartbeat_group_make_item,
+	.drop_item	= o2hb_heartbeat_group_drop_item,
+};
+
+static struct config_item_type o2hb_heartbeat_group_type = {
+	.ct_group_ops	= &o2hb_heartbeat_group_group_ops,
+	.ct_item_ops	= &o2hb_hearbeat_group_item_ops,
+	.ct_attrs	= o2hb_heartbeat_group_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* this is just here to avoid touching group in heartbeat.h which the
+ * entire damn world #includes */
+struct config_group *o2hb_alloc_hb_set(void)
+{
+	struct o2hb_heartbeat_group *hs = NULL;
+	struct config_group *ret = NULL;
+
+	hs = kcalloc(1, sizeof(struct o2hb_heartbeat_group), GFP_KERNEL);
+	if (hs == NULL)
+		goto out;
+
+	config_group_init_type_name(&hs->hs_group, "heartbeat",
+				    &o2hb_heartbeat_group_type);
+
+	ret = &hs->hs_group;
+out:
+	if (ret == NULL)
+		kfree(hs);
+	return ret;
+}
+
+void o2hb_free_hb_set(struct config_group *group)
+{
+	struct o2hb_heartbeat_group *hs = to_o2hb_heartbeat_group(group);
+	kfree(hs);
+}
+
+/* hb callback registration and issueing */
+
+static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type)
+{
+	if (type == O2HB_NUM_CB)
+		return ERR_PTR(-EINVAL);
+
+	return &o2hb_callbacks[type];
+}
+
+void o2hb_setup_callback(struct o2hb_callback_func *hc,
+			 enum o2hb_callback_type type,
+			 o2hb_cb_func *func,
+			 void *data,
+			 int priority)
+{
+	INIT_LIST_HEAD(&hc->hc_item);
+	hc->hc_func = func;
+	hc->hc_data = data;
+	hc->hc_priority = priority;
+	hc->hc_type = type;
+	hc->hc_magic = O2HB_CB_MAGIC;
+}
+EXPORT_SYMBOL_GPL(o2hb_setup_callback);
+
+int o2hb_register_callback(struct o2hb_callback_func *hc)
+{
+	struct o2hb_callback_func *tmp;
+	struct list_head *iter;
+	struct o2hb_callback *hbcall;
+	int ret;
+
+	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
+	BUG_ON(!list_empty(&hc->hc_item));
+
+	hbcall = hbcall_from_type(hc->hc_type);
+	if (IS_ERR(hbcall)) {
+		ret = PTR_ERR(hbcall);
+		goto out;
+	}
+
+	down_write(&o2hb_callback_sem);
+
+	list_for_each(iter, &hbcall->list) {
+		tmp = list_entry(iter, struct o2hb_callback_func, hc_item);
+		if (hc->hc_priority < tmp->hc_priority) {
+			list_add_tail(&hc->hc_item, iter);
+			break;
+		}
+	}
+	if (list_empty(&hc->hc_item))
+		list_add_tail(&hc->hc_item, &hbcall->list);
+
+	up_write(&o2hb_callback_sem);
+	ret = 0;
+out:
+	mlog(ML_HEARTBEAT, "returning %d on behalf of %p for funcs %p\n",
+	     ret, __builtin_return_address(0), hc);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(o2hb_register_callback);
+
+int o2hb_unregister_callback(struct o2hb_callback_func *hc)
+{
+	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
+
+	mlog(ML_HEARTBEAT, "on behalf of %p for funcs %p\n",
+	     __builtin_return_address(0), hc);
+
+	if (list_empty(&hc->hc_item))
+		return 0;
+
+	down_write(&o2hb_callback_sem);
+
+	list_del_init(&hc->hc_item);
+
+	up_write(&o2hb_callback_sem);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(o2hb_unregister_callback);
+
+int o2hb_check_node_heartbeating(u8 node_num)
+{
+	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
+
+	o2hb_fill_node_map(testing_map, sizeof(testing_map));
+	if (!test_bit(node_num, testing_map)) {
+		mlog(ML_HEARTBEAT,
+		     "node (%u) does not have heartbeating enabled.\n",
+		     node_num);
+		return 0;
+	}
+
+	return 1;
+}
+EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating);
+
+int o2hb_check_node_heartbeating_from_callback(u8 node_num)
+{
+	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
+
+	o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
+	if (!test_bit(node_num, testing_map)) {
+		mlog(ML_HEARTBEAT,
+		     "node (%u) does not have heartbeating enabled.\n",
+		     node_num);
+		return 0;
+	}
+
+	return 1;
+}
+EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_from_callback);
+
+/* Makes sure our local node is configured with a node number, and is
+ * heartbeating. */
+int o2hb_check_local_node_heartbeating(void)
+{
+	u8 node_num;
+
+	/* if this node was set then we have networking */
+	node_num = o2nm_this_node();
+	if (node_num == O2NM_MAX_NODES) {
+		mlog(ML_HEARTBEAT, "this node has not been configured.\n");
+		return 0;
+	}
+
+	return o2hb_check_node_heartbeating(node_num);
+}
+EXPORT_SYMBOL_GPL(o2hb_check_local_node_heartbeating);
+
+/*
+ * this is just a hack until we get the plumbing which flips file systems
+ * read only and drops the hb ref instead of killing the node dead.
+ */
+void o2hb_stop_all_regions(void)
+{
+	struct o2hb_region *reg;
+
+	mlog(ML_ERROR, "stopping heartbeat on all active regions.\n");
+
+	spin_lock(&o2hb_live_lock);
+
+	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item)
+		reg->hr_unclean_stop = 1;
+
+	spin_unlock(&o2hb_live_lock);
+}
+EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
diff --git a/fs/ocfs2/cluster/heartbeat.h b/fs/ocfs2/cluster/heartbeat.h
new file mode 100644
index 00000000000000..cac6223206a9bf
--- /dev/null
+++ b/fs/ocfs2/cluster/heartbeat.h
@@ -0,0 +1,82 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * heartbeat.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_HEARTBEAT_H
+#define O2CLUSTER_HEARTBEAT_H
+
+#include "ocfs2_heartbeat.h"
+
+#define O2HB_REGION_TIMEOUT_MS		2000
+
+/* number of changes to be seen as live */
+#define O2HB_LIVE_THRESHOLD	   2
+/* number of equal samples to be seen as dead */
+extern unsigned int o2hb_dead_threshold;
+#define O2HB_DEFAULT_DEAD_THRESHOLD	   7
+/* Otherwise MAX_WRITE_TIMEOUT will be zero... */
+#define O2HB_MIN_DEAD_THRESHOLD	  2
+#define O2HB_MAX_WRITE_TIMEOUT_MS (O2HB_REGION_TIMEOUT_MS * (o2hb_dead_threshold - 1))
+
+#define O2HB_CB_MAGIC		0x51d1e4ec
+
+/* callback stuff */
+enum o2hb_callback_type {
+	O2HB_NODE_DOWN_CB = 0,
+	O2HB_NODE_UP_CB,
+	O2HB_NUM_CB
+};
+
+struct o2nm_node;
+typedef void (o2hb_cb_func)(struct o2nm_node *, int, void *);
+
+struct o2hb_callback_func {
+	u32			hc_magic;
+	struct list_head	hc_item;
+	o2hb_cb_func		*hc_func;
+	void			*hc_data;
+	int			hc_priority;
+	enum o2hb_callback_type hc_type;
+};
+
+struct config_group *o2hb_alloc_hb_set(void);
+void o2hb_free_hb_set(struct config_group *group);
+
+void o2hb_setup_callback(struct o2hb_callback_func *hc,
+			 enum o2hb_callback_type type,
+			 o2hb_cb_func *func,
+			 void *data,
+			 int priority);
+int o2hb_register_callback(struct o2hb_callback_func *hc);
+int o2hb_unregister_callback(struct o2hb_callback_func *hc);
+void o2hb_fill_node_map(unsigned long *map,
+			unsigned bytes);
+void o2hb_init(void);
+int o2hb_check_node_heartbeating(u8 node_num);
+int o2hb_check_node_heartbeating_from_callback(u8 node_num);
+int o2hb_check_local_node_heartbeating(void);
+void o2hb_stop_all_regions(void);
+
+#endif /* O2CLUSTER_HEARTBEAT_H */
diff --git a/fs/ocfs2/cluster/masklog.c b/fs/ocfs2/cluster/masklog.c
new file mode 100644
index 00000000000000..fd741cea5705e0
--- /dev/null
+++ b/fs/ocfs2/cluster/masklog.c
@@ -0,0 +1,166 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#include <linux/string.h>
+#include <asm/uaccess.h>
+
+#include "masklog.h"
+
+struct mlog_bits mlog_and_bits = MLOG_BITS_RHS(MLOG_INITIAL_AND_MASK);
+EXPORT_SYMBOL_GPL(mlog_and_bits);
+struct mlog_bits mlog_not_bits = MLOG_BITS_RHS(MLOG_INITIAL_NOT_MASK);
+EXPORT_SYMBOL_GPL(mlog_not_bits);
+
+static ssize_t mlog_mask_show(u64 mask, char *buf)
+{
+	char *state;
+
+	if (__mlog_test_u64(mask, mlog_and_bits))
+		state = "allow";
+	else if (__mlog_test_u64(mask, mlog_not_bits))
+		state = "deny";
+	else
+		state = "off";
+
+	return snprintf(buf, PAGE_SIZE, "%s\n", state);
+}
+
+static ssize_t mlog_mask_store(u64 mask, const char *buf, size_t count)
+{
+	if (!strnicmp(buf, "allow", 5)) {
+		__mlog_set_u64(mask, mlog_and_bits);
+		__mlog_clear_u64(mask, mlog_not_bits);
+	} else if (!strnicmp(buf, "deny", 4)) {
+		__mlog_set_u64(mask, mlog_not_bits);
+		__mlog_clear_u64(mask, mlog_and_bits);
+	} else if (!strnicmp(buf, "off", 3)) {
+		__mlog_clear_u64(mask, mlog_not_bits);
+		__mlog_clear_u64(mask, mlog_and_bits);
+	} else
+		return -EINVAL;
+
+	return count;
+}
+
+struct mlog_attribute {
+	struct attribute attr;
+	u64 mask;
+};
+
+#define to_mlog_attr(_attr) container_of(_attr, struct mlog_attribute, attr)
+
+#define define_mask(_name) {			\
+	.attr = {				\
+		.name = #_name,			\
+		.mode = S_IRUGO | S_IWUSR,	\
+	},					\
+	.mask = ML_##_name,			\
+}
+
+static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = {
+	define_mask(ENTRY),
+	define_mask(EXIT),
+	define_mask(TCP),
+	define_mask(MSG),
+	define_mask(SOCKET),
+	define_mask(HEARTBEAT),
+	define_mask(HB_BIO),
+	define_mask(DLMFS),
+	define_mask(DLM),
+	define_mask(DLM_DOMAIN),
+	define_mask(DLM_THREAD),
+	define_mask(DLM_MASTER),
+	define_mask(DLM_RECOVERY),
+	define_mask(AIO),
+	define_mask(JOURNAL),
+	define_mask(DISK_ALLOC),
+	define_mask(SUPER),
+	define_mask(FILE_IO),
+	define_mask(EXTENT_MAP),
+	define_mask(DLM_GLUE),
+	define_mask(BH_IO),
+	define_mask(UPTODATE),
+	define_mask(NAMEI),
+	define_mask(INODE),
+	define_mask(VOTE),
+	define_mask(DCACHE),
+	define_mask(CONN),
+	define_mask(QUORUM),
+	define_mask(EXPORT),
+	define_mask(ERROR),
+	define_mask(NOTICE),
+	define_mask(KTHREAD),
+};
+
+static struct attribute *mlog_attr_ptrs[MLOG_MAX_BITS] = {NULL, };
+
+static ssize_t mlog_show(struct kobject *obj, struct attribute *attr,
+			 char *buf)
+{
+	struct mlog_attribute *mlog_attr = to_mlog_attr(attr);
+
+	return mlog_mask_show(mlog_attr->mask, buf);
+}
+
+static ssize_t mlog_store(struct kobject *obj, struct attribute *attr,
+			  const char *buf, size_t count)
+{
+	struct mlog_attribute *mlog_attr = to_mlog_attr(attr);
+
+	return mlog_mask_store(mlog_attr->mask, buf, count);
+}
+
+static struct sysfs_ops mlog_attr_ops = {
+	.show  = mlog_show,
+	.store = mlog_store,
+};
+
+static struct kobj_type mlog_ktype = {
+	.default_attrs = mlog_attr_ptrs,
+	.sysfs_ops     = &mlog_attr_ops,
+};
+
+static struct kset mlog_kset = {
+	.kobj   = {.name = "logmask", .ktype = &mlog_ktype},
+};
+
+int mlog_sys_init(struct subsystem *o2cb_subsys)
+{
+	int i = 0;
+
+	while (mlog_attrs[i].attr.mode) {
+		mlog_attr_ptrs[i] = &mlog_attrs[i].attr;
+		i++;
+	}
+	mlog_attr_ptrs[i] = NULL;
+
+	mlog_kset.subsys = o2cb_subsys;
+	return kset_register(&mlog_kset);
+}
+
+void mlog_sys_shutdown(void)
+{
+	kset_unregister(&mlog_kset);
+}
diff --git a/fs/ocfs2/cluster/masklog.h b/fs/ocfs2/cluster/masklog.h
new file mode 100644
index 00000000000000..f5ef5ea61a05af
--- /dev/null
+++ b/fs/ocfs2/cluster/masklog.h
@@ -0,0 +1,275 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef O2CLUSTER_MASKLOG_H
+#define O2CLUSTER_MASKLOG_H
+
+/*
+ * For now this is a trivial wrapper around printk() that gives the critical
+ * ability to enable sets of debugging output at run-time.  In the future this
+ * will almost certainly be redirected to relayfs so that it can pay a
+ * substantially lower heisenberg tax.
+ *
+ * Callers associate the message with a bitmask and a global bitmask is
+ * maintained with help from /proc.  If any of the bits match the message is
+ * output.
+ *
+ * We must have efficient bit tests on i386 and it seems gcc still emits crazy
+ * code for the 64bit compare.  It emits very good code for the dual unsigned
+ * long tests, though, completely avoiding tests that can never pass if the
+ * caller gives a constant bitmask that fills one of the longs with all 0s.  So
+ * the desire is to have almost all of the calls decided on by comparing just
+ * one of the longs.  This leads to having infrequently given bits that are
+ * frequently matched in the high bits.
+ *
+ * _ERROR and _NOTICE are used for messages that always go to the console and
+ * have appropriate KERN_ prefixes.  We wrap these in our function instead of
+ * just calling printk() so that this can eventually make its way through
+ * relayfs along with the debugging messages.  Everything else gets KERN_DEBUG.
+ * The inline tests and macro dance give GCC the opportunity to quite cleverly
+ * only emit the appropriage printk() when the caller passes in a constant
+ * mask, as is almost always the case.
+ *
+ * All this bitmask nonsense is hidden from the /proc interface so that Joel
+ * doesn't have an aneurism.  Reading the file gives a straight forward
+ * indication of which bits are on or off:
+ * 	ENTRY off
+ * 	EXIT off
+ * 	TCP off
+ * 	MSG off
+ * 	SOCKET off
+ * 	ERROR off
+ * 	NOTICE on
+ *
+ * Writing changes the state of a given bit and requires a strictly formatted
+ * single write() call:
+ *
+ * 	write(fd, "ENTRY on", 8);
+ *
+ * would turn the entry bit on.  "1" is also accepted in the place of "on", and
+ * "off" and "0" behave as expected.
+ *
+ * Some trivial shell can flip all the bits on or off:
+ *
+ * log_mask="/proc/fs/ocfs2_nodemanager/log_mask"
+ * cat $log_mask | (
+ * 	while read bit status; do
+ * 		# $1 is "on" or "off", say
+ * 		echo "$bit $1" > $log_mask
+ * 	done
+ * )
+ */
+
+/* for task_struct */
+#include <linux/sched.h>
+
+/* bits that are frequently given and infrequently matched in the low word */
+/* NOTE: If you add a flag, you need to also update mlog.c! */
+#define ML_ENTRY	0x0000000000000001ULL /* func call entry */
+#define ML_EXIT		0x0000000000000002ULL /* func call exit */
+#define ML_TCP		0x0000000000000004ULL /* net cluster/tcp.c */
+#define ML_MSG		0x0000000000000008ULL /* net network messages */
+#define ML_SOCKET	0x0000000000000010ULL /* net socket lifetime */
+#define ML_HEARTBEAT	0x0000000000000020ULL /* hb all heartbeat tracking */
+#define ML_HB_BIO	0x0000000000000040ULL /* hb io tracing */
+#define ML_DLMFS	0x0000000000000080ULL /* dlm user dlmfs */
+#define ML_DLM		0x0000000000000100ULL /* dlm general debugging */
+#define ML_DLM_DOMAIN	0x0000000000000200ULL /* dlm domain debugging */
+#define ML_DLM_THREAD	0x0000000000000400ULL /* dlm domain thread */
+#define ML_DLM_MASTER	0x0000000000000800ULL /* dlm master functions */
+#define ML_DLM_RECOVERY	0x0000000000001000ULL /* dlm master functions */
+#define ML_AIO		0x0000000000002000ULL /* ocfs2 aio read and write */
+#define ML_JOURNAL	0x0000000000004000ULL /* ocfs2 journalling functions */
+#define ML_DISK_ALLOC	0x0000000000008000ULL /* ocfs2 disk allocation */
+#define ML_SUPER	0x0000000000010000ULL /* ocfs2 mount / umount */
+#define ML_FILE_IO	0x0000000000020000ULL /* ocfs2 file I/O */
+#define ML_EXTENT_MAP	0x0000000000040000ULL /* ocfs2 extent map caching */
+#define ML_DLM_GLUE	0x0000000000080000ULL /* ocfs2 dlm glue layer */
+#define ML_BH_IO	0x0000000000100000ULL /* ocfs2 buffer I/O */
+#define ML_UPTODATE	0x0000000000200000ULL /* ocfs2 caching sequence #'s */
+#define ML_NAMEI	0x0000000000400000ULL /* ocfs2 directory / namespace */
+#define ML_INODE	0x0000000000800000ULL /* ocfs2 inode manipulation */
+#define ML_VOTE		0x0000000001000000ULL /* ocfs2 node messaging  */
+#define ML_DCACHE	0x0000000002000000ULL /* ocfs2 dcache operations */
+#define ML_CONN		0x0000000004000000ULL /* net connection management */
+#define ML_QUORUM	0x0000000008000000ULL /* net connection quorum */
+#define ML_EXPORT	0x0000000010000000ULL /* ocfs2 export operations */
+/* bits that are infrequently given and frequently matched in the high word */
+#define ML_ERROR	0x0000000100000000ULL /* sent to KERN_ERR */
+#define ML_NOTICE	0x0000000200000000ULL /* setn to KERN_NOTICE */
+#define ML_KTHREAD	0x0000000400000000ULL /* kernel thread activity */
+
+#define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE)
+#define MLOG_INITIAL_NOT_MASK (ML_ENTRY|ML_EXIT)
+#ifndef MLOG_MASK_PREFIX
+#define MLOG_MASK_PREFIX 0
+#endif
+
+#define MLOG_MAX_BITS 64
+
+struct mlog_bits {
+	unsigned long words[MLOG_MAX_BITS / BITS_PER_LONG];
+};
+
+extern struct mlog_bits mlog_and_bits, mlog_not_bits;
+
+#if BITS_PER_LONG == 32
+
+#define __mlog_test_u64(mask, bits)			\
+	( (u32)(mask & 0xffffffff) & bits.words[0] || 	\
+	  ((u64)(mask) >> 32) & bits.words[1] )
+#define __mlog_set_u64(mask, bits) do {			\
+	bits.words[0] |= (u32)(mask & 0xffffffff);	\
+       	bits.words[1] |= (u64)(mask) >> 32;		\
+} while (0)
+#define __mlog_clear_u64(mask, bits) do {		\
+	bits.words[0] &= ~((u32)(mask & 0xffffffff));	\
+       	bits.words[1] &= ~((u64)(mask) >> 32);		\
+} while (0)
+#define MLOG_BITS_RHS(mask) {				\
+	{						\
+		[0] = (u32)(mask & 0xffffffff),		\
+		[1] = (u64)(mask) >> 32,		\
+	}						\
+}
+
+#else /* 32bit long above, 64bit long below */
+
+#define __mlog_test_u64(mask, bits)	((mask) & bits.words[0])
+#define __mlog_set_u64(mask, bits) do {		\
+	bits.words[0] |= (mask);		\
+} while (0)
+#define __mlog_clear_u64(mask, bits) do {	\
+	bits.words[0] &= ~(mask);		\
+} while (0)
+#define MLOG_BITS_RHS(mask) { { (mask) } }
+
+#endif
+
+/*
+ * smp_processor_id() "helpfully" screams when called outside preemptible
+ * regions in current kernels.  sles doesn't have the variants that don't
+ * scream.  just do this instead of trying to guess which we're building
+ * against.. *sigh*.
+ */
+#define __mlog_cpu_guess ({		\
+	unsigned long _cpu = get_cpu();	\
+	put_cpu();			\
+	_cpu;				\
+})
+
+/* In the following two macros, the whitespace after the ',' just
+ * before ##args is intentional. Otherwise, gcc 2.95 will eat the
+ * previous token if args expands to nothing.
+ */
+#define __mlog_printk(level, fmt, args...)				\
+	printk(level "(%u,%lu):%s:%d " fmt, current->pid,		\
+	       __mlog_cpu_guess, __PRETTY_FUNCTION__, __LINE__ ,	\
+	       ##args)
+
+#define mlog(mask, fmt, args...) do {					\
+	u64 __m = MLOG_MASK_PREFIX | (mask);				\
+	if (__mlog_test_u64(__m, mlog_and_bits) &&			\
+	    !__mlog_test_u64(__m, mlog_not_bits)) {			\
+		if (__m & ML_ERROR)					\
+			__mlog_printk(KERN_ERR, "ERROR: "fmt , ##args);	\
+		else if (__m & ML_NOTICE)				\
+			__mlog_printk(KERN_NOTICE, fmt , ##args);	\
+		else __mlog_printk(KERN_INFO, fmt , ##args);		\
+	}								\
+} while (0)
+
+#define mlog_errno(st) do {						\
+	int _st = (st);							\
+	if (_st != -ERESTARTSYS && _st != -EINTR &&			\
+	    _st != AOP_TRUNCATED_PAGE)					\
+		mlog(ML_ERROR, "status = %lld\n", (long long)_st);	\
+} while (0)
+
+#define mlog_entry(fmt, args...) do {					\
+	mlog(ML_ENTRY, "ENTRY:" fmt , ##args);				\
+} while (0)
+
+#define mlog_entry_void() do {						\
+	mlog(ML_ENTRY, "ENTRY:\n");					\
+} while (0)
+
+/* We disable this for old compilers since they don't have support for
+ * __builtin_types_compatible_p.
+ */
+#if (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 1)) && \
+    !defined(__CHECKER__)
+#define mlog_exit(st) do {						     \
+	if (__builtin_types_compatible_p(typeof(st), unsigned long))	     \
+		mlog(ML_EXIT, "EXIT: %lu\n", (unsigned long) (st));	     \
+	else if (__builtin_types_compatible_p(typeof(st), signed long))      \
+		mlog(ML_EXIT, "EXIT: %ld\n", (signed long) (st));	     \
+	else if (__builtin_types_compatible_p(typeof(st), unsigned int)	     \
+		 || __builtin_types_compatible_p(typeof(st), unsigned short) \
+		 || __builtin_types_compatible_p(typeof(st), unsigned char)) \
+		mlog(ML_EXIT, "EXIT: %u\n", (unsigned int) (st));	     \
+	else if (__builtin_types_compatible_p(typeof(st), signed int)	     \
+		 || __builtin_types_compatible_p(typeof(st), signed short)   \
+		 || __builtin_types_compatible_p(typeof(st), signed char))   \
+		mlog(ML_EXIT, "EXIT: %d\n", (signed int) (st));		     \
+	else if (__builtin_types_compatible_p(typeof(st), long long))	     \
+		mlog(ML_EXIT, "EXIT: %lld\n", (long long) (st));	     \
+	else								     \
+		mlog(ML_EXIT, "EXIT: %llu\n", (unsigned long long) (st));    \
+} while (0)
+#else
+#define mlog_exit(st) do {						     \
+	mlog(ML_EXIT, "EXIT: %lld\n", (long long) (st));		     \
+} while (0)
+#endif
+
+#define mlog_exit_ptr(ptr) do {						\
+	mlog(ML_EXIT, "EXIT: %p\n", ptr);				\
+} while (0)
+
+#define mlog_exit_void() do {						\
+	mlog(ML_EXIT, "EXIT\n");					\
+} while (0)
+
+#define mlog_bug_on_msg(cond, fmt, args...) do {			\
+	if (cond) {							\
+		mlog(ML_ERROR, "bug expression: " #cond "\n");		\
+		mlog(ML_ERROR, fmt, ##args);				\
+		BUG();							\
+	}								\
+} while (0)
+
+#if (BITS_PER_LONG == 32) || defined(CONFIG_X86_64)
+#define MLFi64 "lld"
+#define MLFu64 "llu"
+#define MLFx64 "llx"
+#else
+#define MLFi64 "ld"
+#define MLFu64 "lu"
+#define MLFx64 "lx"
+#endif
+
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+int mlog_sys_init(struct subsystem *o2cb_subsys);
+void mlog_sys_shutdown(void);
+
+#endif /* O2CLUSTER_MASKLOG_H */
diff --git a/fs/ocfs2/cluster/nodemanager.c b/fs/ocfs2/cluster/nodemanager.c
new file mode 100644
index 00000000000000..5fd60c1059135e
--- /dev/null
+++ b/fs/ocfs2/cluster/nodemanager.c
@@ -0,0 +1,791 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/sysctl.h>
+#include <linux/configfs.h>
+
+#include "endian.h"
+#include "tcp.h"
+#include "nodemanager.h"
+#include "heartbeat.h"
+#include "masklog.h"
+#include "sys.h"
+#include "ver.h"
+
+/* for now we operate under the assertion that there can be only one
+ * cluster active at a time.  Changing this will require trickling
+ * cluster references throughout where nodes are looked up */
+static struct o2nm_cluster *o2nm_single_cluster = NULL;
+
+#define OCFS2_MAX_HB_CTL_PATH 256
+static char ocfs2_hb_ctl_path[OCFS2_MAX_HB_CTL_PATH] = "/sbin/ocfs2_hb_ctl";
+
+static ctl_table ocfs2_nm_table[] = {
+	{
+		.ctl_name	= 1,
+		.procname	= "hb_ctl_path",
+		.data		= ocfs2_hb_ctl_path,
+		.maxlen		= OCFS2_MAX_HB_CTL_PATH,
+		.mode		= 0644,
+		.proc_handler	= &proc_dostring,
+		.strategy	= &sysctl_string,
+	},
+	{ .ctl_name = 0 }
+};
+
+static ctl_table ocfs2_mod_table[] = {
+	{
+		.ctl_name	= KERN_OCFS2_NM,
+		.procname	= "nm",
+		.data		= NULL,
+		.maxlen		= 0,
+		.mode		= 0555,
+		.child		= ocfs2_nm_table
+	},
+	{ .ctl_name = 0}
+};
+
+static ctl_table ocfs2_kern_table[] = {
+	{
+		.ctl_name	= KERN_OCFS2,
+		.procname	= "ocfs2",
+		.data		= NULL,
+		.maxlen		= 0,
+		.mode		= 0555,
+		.child		= ocfs2_mod_table
+	},
+	{ .ctl_name = 0}
+};
+
+static ctl_table ocfs2_root_table[] = {
+	{
+		.ctl_name	= CTL_FS,
+		.procname	= "fs",
+		.data		= NULL,
+		.maxlen		= 0,
+		.mode		= 0555,
+		.child		= ocfs2_kern_table
+	},
+	{ .ctl_name = 0 }
+};
+
+static struct ctl_table_header *ocfs2_table_header = NULL;
+
+const char *o2nm_get_hb_ctl_path(void)
+{
+	return ocfs2_hb_ctl_path;
+}
+EXPORT_SYMBOL_GPL(o2nm_get_hb_ctl_path);
+
+struct o2nm_cluster {
+	struct config_group	cl_group;
+	unsigned		cl_has_local:1;
+	u8			cl_local_node;
+	rwlock_t		cl_nodes_lock;
+	struct o2nm_node  	*cl_nodes[O2NM_MAX_NODES];
+	struct rb_root		cl_node_ip_tree;
+	/* this bitmap is part of a hack for disk bitmap.. will go eventually. - zab */
+	unsigned long	cl_nodes_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
+};
+
+struct o2nm_node *o2nm_get_node_by_num(u8 node_num)
+{
+	struct o2nm_node *node = NULL;
+
+	if (node_num >= O2NM_MAX_NODES || o2nm_single_cluster == NULL)
+		goto out;
+
+	read_lock(&o2nm_single_cluster->cl_nodes_lock);
+	node = o2nm_single_cluster->cl_nodes[node_num];
+	if (node)
+		config_item_get(&node->nd_item);
+	read_unlock(&o2nm_single_cluster->cl_nodes_lock);
+out:
+	return node;
+}
+EXPORT_SYMBOL_GPL(o2nm_get_node_by_num);
+
+int o2nm_configured_node_map(unsigned long *map, unsigned bytes)
+{
+	struct o2nm_cluster *cluster = o2nm_single_cluster;
+
+	BUG_ON(bytes < (sizeof(cluster->cl_nodes_bitmap)));
+
+	if (cluster == NULL)
+		return -EINVAL;
+
+	read_lock(&cluster->cl_nodes_lock);
+	memcpy(map, cluster->cl_nodes_bitmap, sizeof(cluster->cl_nodes_bitmap));
+	read_unlock(&cluster->cl_nodes_lock);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(o2nm_configured_node_map);
+
+static struct o2nm_node *o2nm_node_ip_tree_lookup(struct o2nm_cluster *cluster,
+						  __be32 ip_needle,
+						  struct rb_node ***ret_p,
+						  struct rb_node **ret_parent)
+{
+	struct rb_node **p = &cluster->cl_node_ip_tree.rb_node;
+	struct rb_node *parent = NULL;
+	struct o2nm_node *node, *ret = NULL;
+
+	while (*p) {
+		parent = *p;
+		node = rb_entry(parent, struct o2nm_node, nd_ip_node);
+
+		if (memcmp(&ip_needle, &node->nd_ipv4_address,
+		           sizeof(ip_needle)) < 0)
+			p = &(*p)->rb_left;
+		else if (memcmp(&ip_needle, &node->nd_ipv4_address,
+			        sizeof(ip_needle)) > 0)
+			p = &(*p)->rb_right;
+		else {
+			ret = node;
+			break;
+		}
+	}
+
+	if (ret_p != NULL)
+		*ret_p = p;
+	if (ret_parent != NULL)
+		*ret_parent = parent;
+
+	return ret;
+}
+
+struct o2nm_node *o2nm_get_node_by_ip(__be32 addr)
+{
+	struct o2nm_node *node = NULL;
+	struct o2nm_cluster *cluster = o2nm_single_cluster;
+
+	if (cluster == NULL)
+		goto out;
+
+	read_lock(&cluster->cl_nodes_lock);
+	node = o2nm_node_ip_tree_lookup(cluster, addr, NULL, NULL);
+	if (node)
+		config_item_get(&node->nd_item);
+	read_unlock(&cluster->cl_nodes_lock);
+
+out:
+	return node;
+}
+EXPORT_SYMBOL_GPL(o2nm_get_node_by_ip);
+
+void o2nm_node_put(struct o2nm_node *node)
+{
+	config_item_put(&node->nd_item);
+}
+EXPORT_SYMBOL_GPL(o2nm_node_put);
+
+void o2nm_node_get(struct o2nm_node *node)
+{
+	config_item_get(&node->nd_item);
+}
+EXPORT_SYMBOL_GPL(o2nm_node_get);
+
+u8 o2nm_this_node(void)
+{
+	u8 node_num = O2NM_MAX_NODES;
+
+	if (o2nm_single_cluster && o2nm_single_cluster->cl_has_local)
+		node_num = o2nm_single_cluster->cl_local_node;
+
+	return node_num;
+}
+EXPORT_SYMBOL_GPL(o2nm_this_node);
+
+/* node configfs bits */
+
+static struct o2nm_cluster *to_o2nm_cluster(struct config_item *item)
+{
+	return item ?
+		container_of(to_config_group(item), struct o2nm_cluster,
+			     cl_group)
+		: NULL;
+}
+
+static struct o2nm_node *to_o2nm_node(struct config_item *item)
+{
+	return item ? container_of(item, struct o2nm_node, nd_item) : NULL;
+}
+
+static void o2nm_node_release(struct config_item *item)
+{
+	struct o2nm_node *node = to_o2nm_node(item);
+	kfree(node);
+}
+
+static ssize_t o2nm_node_num_read(struct o2nm_node *node, char *page)
+{
+	return sprintf(page, "%d\n", node->nd_num);
+}
+
+static struct o2nm_cluster *to_o2nm_cluster_from_node(struct o2nm_node *node)
+{
+	/* through the first node_set .parent
+	 * mycluster/nodes/mynode == o2nm_cluster->o2nm_node_group->o2nm_node */
+	return to_o2nm_cluster(node->nd_item.ci_parent->ci_parent);
+}
+
+enum {
+	O2NM_NODE_ATTR_NUM = 0,
+	O2NM_NODE_ATTR_PORT,
+	O2NM_NODE_ATTR_ADDRESS,
+	O2NM_NODE_ATTR_LOCAL,
+};
+
+static ssize_t o2nm_node_num_write(struct o2nm_node *node, const char *page,
+				   size_t count)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
+	unsigned long tmp;
+	char *p = (char *)page;
+
+	tmp = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	if (tmp >= O2NM_MAX_NODES)
+		return -ERANGE;
+
+	/* once we're in the cl_nodes tree networking can look us up by
+	 * node number and try to use our address and port attributes
+	 * to connect to this node.. make sure that they've been set
+	 * before writing the node attribute? */
+	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
+	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
+		return -EINVAL; /* XXX */
+
+	write_lock(&cluster->cl_nodes_lock);
+	if (cluster->cl_nodes[tmp])
+		p = NULL;
+	else  {
+		cluster->cl_nodes[tmp] = node;
+		node->nd_num = tmp;
+		set_bit(tmp, cluster->cl_nodes_bitmap);
+	}
+	write_unlock(&cluster->cl_nodes_lock);
+	if (p == NULL)
+		return -EEXIST;
+
+	return count;
+}
+static ssize_t o2nm_node_ipv4_port_read(struct o2nm_node *node, char *page)
+{
+	return sprintf(page, "%u\n", ntohs(node->nd_ipv4_port));
+}
+
+static ssize_t o2nm_node_ipv4_port_write(struct o2nm_node *node,
+					 const char *page, size_t count)
+{
+	unsigned long tmp;
+	char *p = (char *)page;
+
+	tmp = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	if (tmp == 0)
+		return -EINVAL;
+	if (tmp >= (u16)-1)
+		return -ERANGE;
+
+	node->nd_ipv4_port = htons(tmp);
+
+	return count;
+}
+
+static ssize_t o2nm_node_ipv4_address_read(struct o2nm_node *node, char *page)
+{
+	return sprintf(page, "%u.%u.%u.%u\n", NIPQUAD(node->nd_ipv4_address));
+}
+
+static ssize_t o2nm_node_ipv4_address_write(struct o2nm_node *node,
+					    const char *page,
+					    size_t count)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
+	int ret, i;
+	struct rb_node **p, *parent;
+	unsigned int octets[4];
+	__be32 ipv4_addr = 0;
+
+	ret = sscanf(page, "%3u.%3u.%3u.%3u", &octets[3], &octets[2],
+		     &octets[1], &octets[0]);
+	if (ret != 4)
+		return -EINVAL;
+
+	for (i = 0; i < ARRAY_SIZE(octets); i++) {
+		if (octets[i] > 255)
+			return -ERANGE;
+		be32_add_cpu(&ipv4_addr, octets[i] << (i * 8));
+	}
+
+	ret = 0;
+	write_lock(&cluster->cl_nodes_lock);
+	if (o2nm_node_ip_tree_lookup(cluster, ipv4_addr, &p, &parent))
+		ret = -EEXIST;
+	else {
+		rb_link_node(&node->nd_ip_node, parent, p);
+		rb_insert_color(&node->nd_ip_node, &cluster->cl_node_ip_tree);
+	}
+	write_unlock(&cluster->cl_nodes_lock);
+	if (ret)
+		return ret;
+
+	memcpy(&node->nd_ipv4_address, &ipv4_addr, sizeof(ipv4_addr));
+
+	return count;
+}
+
+static ssize_t o2nm_node_local_read(struct o2nm_node *node, char *page)
+{
+	return sprintf(page, "%d\n", node->nd_local);
+}
+
+static ssize_t o2nm_node_local_write(struct o2nm_node *node, const char *page,
+				     size_t count)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
+	unsigned long tmp;
+	char *p = (char *)page;
+	ssize_t ret;
+
+	tmp = simple_strtoul(p, &p, 0);
+	if (!p || (*p && (*p != '\n')))
+		return -EINVAL;
+
+	tmp = !!tmp; /* boolean of whether this node wants to be local */
+
+	/* setting local turns on networking rx for now so we require having
+	 * set everything else first */
+	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
+	    !test_bit(O2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
+	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
+		return -EINVAL; /* XXX */
+
+	/* the only failure case is trying to set a new local node
+	 * when a different one is already set */
+	if (tmp && tmp == cluster->cl_has_local &&
+	    cluster->cl_local_node != node->nd_num)
+		return -EBUSY;
+
+	/* bring up the rx thread if we're setting the new local node. */
+	if (tmp && !cluster->cl_has_local) {
+		ret = o2net_start_listening(node);
+		if (ret)
+			return ret;
+	}
+
+	if (!tmp && cluster->cl_has_local &&
+	    cluster->cl_local_node == node->nd_num) {
+		o2net_stop_listening(node);
+		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
+	}
+
+	node->nd_local = tmp;
+	if (node->nd_local) {
+		cluster->cl_has_local = tmp;
+		cluster->cl_local_node = node->nd_num;
+	}
+
+	return count;
+}
+
+struct o2nm_node_attribute {
+	struct configfs_attribute attr;
+	ssize_t (*show)(struct o2nm_node *, char *);
+	ssize_t (*store)(struct o2nm_node *, const char *, size_t);
+};
+
+static struct o2nm_node_attribute o2nm_node_attr_num = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "num",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_node_num_read,
+	.store	= o2nm_node_num_write,
+};
+
+static struct o2nm_node_attribute o2nm_node_attr_ipv4_port = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "ipv4_port",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_node_ipv4_port_read,
+	.store	= o2nm_node_ipv4_port_write,
+};
+
+static struct o2nm_node_attribute o2nm_node_attr_ipv4_address = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "ipv4_address",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_node_ipv4_address_read,
+	.store	= o2nm_node_ipv4_address_write,
+};
+
+static struct o2nm_node_attribute o2nm_node_attr_local = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "local",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= o2nm_node_local_read,
+	.store	= o2nm_node_local_write,
+};
+
+static struct configfs_attribute *o2nm_node_attrs[] = {
+	[O2NM_NODE_ATTR_NUM] = &o2nm_node_attr_num.attr,
+	[O2NM_NODE_ATTR_PORT] = &o2nm_node_attr_ipv4_port.attr,
+	[O2NM_NODE_ATTR_ADDRESS] = &o2nm_node_attr_ipv4_address.attr,
+	[O2NM_NODE_ATTR_LOCAL] = &o2nm_node_attr_local.attr,
+	NULL,
+};
+
+static int o2nm_attr_index(struct configfs_attribute *attr)
+{
+	int i;
+	for (i = 0; i < ARRAY_SIZE(o2nm_node_attrs); i++) {
+		if (attr == o2nm_node_attrs[i])
+			return i;
+	}
+	BUG();
+	return 0;
+}
+
+static ssize_t o2nm_node_show(struct config_item *item,
+			      struct configfs_attribute *attr,
+			      char *page)
+{
+	struct o2nm_node *node = to_o2nm_node(item);
+	struct o2nm_node_attribute *o2nm_node_attr =
+		container_of(attr, struct o2nm_node_attribute, attr);
+	ssize_t ret = 0;
+
+	if (o2nm_node_attr->show)
+		ret = o2nm_node_attr->show(node, page);
+	return ret;
+}
+
+static ssize_t o2nm_node_store(struct config_item *item,
+			       struct configfs_attribute *attr,
+			       const char *page, size_t count)
+{
+	struct o2nm_node *node = to_o2nm_node(item);
+	struct o2nm_node_attribute *o2nm_node_attr =
+		container_of(attr, struct o2nm_node_attribute, attr);
+	ssize_t ret;
+	int attr_index = o2nm_attr_index(attr);
+
+	if (o2nm_node_attr->store == NULL) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (test_bit(attr_index, &node->nd_set_attributes))
+		return -EBUSY;
+
+	ret = o2nm_node_attr->store(node, page, count);
+	if (ret < count)
+		goto out;
+
+	set_bit(attr_index, &node->nd_set_attributes);
+out:
+	return ret;
+}
+
+static struct configfs_item_operations o2nm_node_item_ops = {
+	.release		= o2nm_node_release,
+	.show_attribute		= o2nm_node_show,
+	.store_attribute	= o2nm_node_store,
+};
+
+static struct config_item_type o2nm_node_type = {
+	.ct_item_ops	= &o2nm_node_item_ops,
+	.ct_attrs	= o2nm_node_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* node set */
+
+struct o2nm_node_group {
+	struct config_group ns_group;
+	/* some stuff? */
+};
+
+#if 0
+static struct o2nm_node_group *to_o2nm_node_group(struct config_group *group)
+{
+	return group ?
+		container_of(group, struct o2nm_node_group, ns_group)
+		: NULL;
+}
+#endif
+
+static struct config_item *o2nm_node_group_make_item(struct config_group *group,
+						     const char *name)
+{
+	struct o2nm_node *node = NULL;
+	struct config_item *ret = NULL;
+
+	if (strlen(name) > O2NM_MAX_NAME_LEN)
+		goto out; /* ENAMETOOLONG */
+
+	node = kcalloc(1, sizeof(struct o2nm_node), GFP_KERNEL);
+	if (node == NULL)
+		goto out; /* ENOMEM */
+
+	strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
+	config_item_init_type_name(&node->nd_item, name, &o2nm_node_type);
+	spin_lock_init(&node->nd_lock);
+
+	ret = &node->nd_item;
+
+out:
+	if (ret == NULL)
+		kfree(node);
+
+	return ret;
+}
+
+static void o2nm_node_group_drop_item(struct config_group *group,
+				      struct config_item *item)
+{
+	struct o2nm_node *node = to_o2nm_node(item);
+	struct o2nm_cluster *cluster = to_o2nm_cluster(group->cg_item.ci_parent);
+
+	o2net_disconnect_node(node);
+
+	if (cluster->cl_has_local &&
+	    (cluster->cl_local_node == node->nd_num)) {
+		cluster->cl_has_local = 0;
+		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
+		o2net_stop_listening(node);
+	}
+
+	/* XXX call into net to stop this node from trading messages */
+
+	write_lock(&cluster->cl_nodes_lock);
+
+	/* XXX sloppy */
+	if (node->nd_ipv4_address)
+		rb_erase(&node->nd_ip_node, &cluster->cl_node_ip_tree);
+
+	/* nd_num might be 0 if the node number hasn't been set.. */
+	if (cluster->cl_nodes[node->nd_num] == node) {
+		cluster->cl_nodes[node->nd_num] = NULL;
+		clear_bit(node->nd_num, cluster->cl_nodes_bitmap);
+	}
+	write_unlock(&cluster->cl_nodes_lock);
+
+	config_item_put(item);
+}
+
+static struct configfs_group_operations o2nm_node_group_group_ops = {
+	.make_item	= o2nm_node_group_make_item,
+	.drop_item	= o2nm_node_group_drop_item,
+};
+
+static struct config_item_type o2nm_node_group_type = {
+	.ct_group_ops	= &o2nm_node_group_group_ops,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* cluster */
+
+static void o2nm_cluster_release(struct config_item *item)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
+
+	kfree(cluster->cl_group.default_groups);
+	kfree(cluster);
+}
+
+static struct configfs_item_operations o2nm_cluster_item_ops = {
+	.release	= o2nm_cluster_release,
+};
+
+static struct config_item_type o2nm_cluster_type = {
+	.ct_item_ops	= &o2nm_cluster_item_ops,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* cluster set */
+
+struct o2nm_cluster_group {
+	struct configfs_subsystem cs_subsys;
+	/* some stuff? */
+};
+
+#if 0
+static struct o2nm_cluster_group *to_o2nm_cluster_group(struct config_group *group)
+{
+	return group ?
+		container_of(to_configfs_subsystem(group), struct o2nm_cluster_group, cs_subsys)
+	       : NULL;
+}
+#endif
+
+static struct config_group *o2nm_cluster_group_make_group(struct config_group *group,
+							  const char *name)
+{
+	struct o2nm_cluster *cluster = NULL;
+	struct o2nm_node_group *ns = NULL;
+	struct config_group *o2hb_group = NULL, *ret = NULL;
+	void *defs = NULL;
+
+	/* this runs under the parent dir's i_sem; there can be only
+	 * one caller in here at a time */
+	if (o2nm_single_cluster)
+		goto out; /* ENOSPC */
+
+	cluster = kcalloc(1, sizeof(struct o2nm_cluster), GFP_KERNEL);
+	ns = kcalloc(1, sizeof(struct o2nm_node_group), GFP_KERNEL);
+	defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL);
+	o2hb_group = o2hb_alloc_hb_set();
+	if (cluster == NULL || ns == NULL || o2hb_group == NULL || defs == NULL)
+		goto out;
+
+	config_group_init_type_name(&cluster->cl_group, name,
+				    &o2nm_cluster_type);
+	config_group_init_type_name(&ns->ns_group, "node",
+				    &o2nm_node_group_type);
+
+	cluster->cl_group.default_groups = defs;
+	cluster->cl_group.default_groups[0] = &ns->ns_group;
+	cluster->cl_group.default_groups[1] = o2hb_group;
+	cluster->cl_group.default_groups[2] = NULL;
+	rwlock_init(&cluster->cl_nodes_lock);
+	cluster->cl_node_ip_tree = RB_ROOT;
+
+	ret = &cluster->cl_group;
+	o2nm_single_cluster = cluster;
+
+out:
+	if (ret == NULL) {
+		kfree(cluster);
+		kfree(ns);
+		o2hb_free_hb_set(o2hb_group);
+		kfree(defs);
+	}
+
+	return ret;
+}
+
+static void o2nm_cluster_group_drop_item(struct config_group *group, struct config_item *item)
+{
+	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
+	int i;
+	struct config_item *killme;
+
+	BUG_ON(o2nm_single_cluster != cluster);
+	o2nm_single_cluster = NULL;
+
+	for (i = 0; cluster->cl_group.default_groups[i]; i++) {
+		killme = &cluster->cl_group.default_groups[i]->cg_item;
+		cluster->cl_group.default_groups[i] = NULL;
+		config_item_put(killme);
+	}
+
+	config_item_put(item);
+}
+
+static struct configfs_group_operations o2nm_cluster_group_group_ops = {
+	.make_group	= o2nm_cluster_group_make_group,
+	.drop_item	= o2nm_cluster_group_drop_item,
+};
+
+static struct config_item_type o2nm_cluster_group_type = {
+	.ct_group_ops	= &o2nm_cluster_group_group_ops,
+	.ct_owner	= THIS_MODULE,
+};
+
+static struct o2nm_cluster_group o2nm_cluster_group = {
+	.cs_subsys = {
+		.su_group = {
+			.cg_item = {
+				.ci_namebuf = "cluster",
+				.ci_type = &o2nm_cluster_group_type,
+			},
+		},
+	},
+};
+
+static void __exit exit_o2nm(void)
+{
+	if (ocfs2_table_header)
+		unregister_sysctl_table(ocfs2_table_header);
+
+	/* XXX sync with hb callbacks and shut down hb? */
+	o2net_unregister_hb_callbacks();
+	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
+	o2cb_sys_shutdown();
+
+	o2net_exit();
+}
+
+static int __init init_o2nm(void)
+{
+	int ret = -1;
+
+	cluster_print_version();
+
+	o2hb_init();
+	o2net_init();
+
+	ocfs2_table_header = register_sysctl_table(ocfs2_root_table, 0);
+	if (!ocfs2_table_header) {
+		printk(KERN_ERR "nodemanager: unable to register sysctl\n");
+		ret = -ENOMEM; /* or something. */
+		goto out;
+	}
+
+	ret = o2net_register_hb_callbacks();
+	if (ret)
+		goto out_sysctl;
+
+	config_group_init(&o2nm_cluster_group.cs_subsys.su_group);
+	init_MUTEX(&o2nm_cluster_group.cs_subsys.su_sem);
+	ret = configfs_register_subsystem(&o2nm_cluster_group.cs_subsys);
+	if (ret) {
+		printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
+		goto out_callbacks;
+	}
+
+	ret = o2cb_sys_init();
+	if (!ret)
+		goto out;
+
+	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
+out_callbacks:
+	o2net_unregister_hb_callbacks();
+out_sysctl:
+	unregister_sysctl_table(ocfs2_table_header);
+out:
+	return ret;
+}
+
+MODULE_AUTHOR("Oracle");
+MODULE_LICENSE("GPL");
+
+module_init(init_o2nm)
+module_exit(exit_o2nm)
diff --git a/fs/ocfs2/cluster/nodemanager.h b/fs/ocfs2/cluster/nodemanager.h
new file mode 100644
index 00000000000000..fce8033c310fb0
--- /dev/null
+++ b/fs/ocfs2/cluster/nodemanager.h
@@ -0,0 +1,64 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * nodemanager.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_NODEMANAGER_H
+#define O2CLUSTER_NODEMANAGER_H
+
+#include "ocfs2_nodemanager.h"
+
+/* This totally doesn't belong here. */
+#include <linux/configfs.h>
+#include <linux/rbtree.h>
+
+#define KERN_OCFS2		988
+#define KERN_OCFS2_NM		1
+
+const char *o2nm_get_hb_ctl_path(void);
+
+struct o2nm_node {
+	spinlock_t		nd_lock;
+	struct config_item	nd_item;
+	char			nd_name[O2NM_MAX_NAME_LEN+1]; /* replace? */
+	__u8			nd_num;
+	/* only one address per node, as attributes, for now. */
+	__be32			nd_ipv4_address;
+	__be16			nd_ipv4_port;
+	struct rb_node		nd_ip_node;
+	/* there can be only one local node for now */
+	int			nd_local;
+
+	unsigned long		nd_set_attributes;
+};
+
+u8 o2nm_this_node(void);
+
+int o2nm_configured_node_map(unsigned long *map, unsigned bytes);
+struct o2nm_node *o2nm_get_node_by_num(u8 node_num);
+struct o2nm_node *o2nm_get_node_by_ip(__be32 addr);
+void o2nm_node_get(struct o2nm_node *node);
+void o2nm_node_put(struct o2nm_node *node);
+
+#endif /* O2CLUSTER_NODEMANAGER_H */
diff --git a/fs/ocfs2/cluster/ocfs2_heartbeat.h b/fs/ocfs2/cluster/ocfs2_heartbeat.h
new file mode 100644
index 00000000000000..94096069cb43a0
--- /dev/null
+++ b/fs/ocfs2/cluster/ocfs2_heartbeat.h
@@ -0,0 +1,37 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_heartbeat.h
+ *
+ * On-disk structures for ocfs2_heartbeat
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef _OCFS2_HEARTBEAT_H
+#define _OCFS2_HEARTBEAT_H
+
+struct o2hb_disk_heartbeat_block {
+	__le64 hb_seq;
+	__u8  hb_node;
+	__u8  hb_pad1[3];
+	__le32 hb_cksum;
+	__le64 hb_generation;
+};
+
+#endif /* _OCFS2_HEARTBEAT_H */
diff --git a/fs/ocfs2/cluster/ocfs2_nodemanager.h b/fs/ocfs2/cluster/ocfs2_nodemanager.h
new file mode 100644
index 00000000000000..5b9854bad571c2
--- /dev/null
+++ b/fs/ocfs2/cluster/ocfs2_nodemanager.h
@@ -0,0 +1,39 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_nodemanager.h
+ *
+ * Header describing the interface between userspace and the kernel
+ * for the ocfs2_nodemanager module.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef _OCFS2_NODEMANAGER_H
+#define _OCFS2_NODEMANAGER_H
+
+#define O2NM_API_VERSION	5
+
+#define O2NM_MAX_NODES		255
+#define O2NM_INVALID_NODE_NUM	255
+
+/* host name, group name, cluster name all 64 bytes */
+#define O2NM_MAX_NAME_LEN        64    // __NEW_UTS_LEN
+
+#endif /* _OCFS2_NODEMANAGER_H */
diff --git a/fs/ocfs2/cluster/quorum.c b/fs/ocfs2/cluster/quorum.c
new file mode 100644
index 00000000000000..7bba98fbfc1532
--- /dev/null
+++ b/fs/ocfs2/cluster/quorum.c
@@ -0,0 +1,315 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ *
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+/* This quorum hack is only here until we transition to some more rational
+ * approach that is driven from userspace.  Honest.  No foolin'.
+ *
+ * Imagine two nodes lose network connectivity to each other but they're still
+ * up and operating in every other way.  Presumably a network timeout indicates
+ * that a node is broken and should be recovered.  They can't both recover each
+ * other and both carry on without serialising their access to the file system.
+ * They need to decide who is authoritative.  Now extend that problem to
+ * arbitrary groups of nodes losing connectivity between each other.
+ *
+ * So we declare that a node which has given up on connecting to a majority
+ * of nodes who are still heartbeating will fence itself.
+ *
+ * There are huge opportunities for races here.  After we give up on a node's
+ * connection we need to wait long enough to give heartbeat an opportunity
+ * to declare the node as truly dead.  We also need to be careful with the
+ * race between when we see a node start heartbeating and when we connect
+ * to it.
+ *
+ * So nodes that are in this transtion put a hold on the quorum decision
+ * with a counter.  As they fall out of this transition they drop the count
+ * and if they're the last, they fire off the decision.
+ */
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/workqueue.h>
+
+#include "heartbeat.h"
+#include "nodemanager.h"
+#define MLOG_MASK_PREFIX ML_QUORUM
+#include "masklog.h"
+#include "quorum.h"
+
+static struct o2quo_state {
+	spinlock_t		qs_lock;
+	struct work_struct	qs_work;
+	int			qs_pending;
+	int			qs_heartbeating;
+	unsigned long		qs_hb_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
+	int			qs_connected;
+	unsigned long		qs_conn_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
+	int			qs_holds;
+	unsigned long		qs_hold_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
+} o2quo_state;
+
+/* this is horribly heavy-handed.  It should instead flip the file
+ * system RO and call some userspace script. */
+static void o2quo_fence_self(void)
+{
+	/* panic spins with interrupts enabled.  with preempt
+	 * threads can still schedule, etc, etc */
+	o2hb_stop_all_regions();
+	panic("ocfs2 is very sorry to be fencing this system by panicing\n");
+}
+
+/* Indicate that a timeout occured on a hearbeat region write. The
+ * other nodes in the cluster may consider us dead at that time so we
+ * want to "fence" ourselves so that we don't scribble on the disk
+ * after they think they've recovered us. This can't solve all
+ * problems related to writeout after recovery but this hack can at
+ * least close some of those gaps. When we have real fencing, this can
+ * go away as our node would be fenced externally before other nodes
+ * begin recovery. */
+void o2quo_disk_timeout(void)
+{
+	o2quo_fence_self();
+}
+
+static void o2quo_make_decision(void *arg)
+{
+	int quorum;
+	int lowest_hb, lowest_reachable = 0, fence = 0;
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	lowest_hb = find_first_bit(qs->qs_hb_bm, O2NM_MAX_NODES);
+	if (lowest_hb != O2NM_MAX_NODES)
+		lowest_reachable = test_bit(lowest_hb, qs->qs_conn_bm);
+
+	mlog(0, "heartbeating: %d, connected: %d, "
+	     "lowest: %d (%sreachable)\n", qs->qs_heartbeating,
+	     qs->qs_connected, lowest_hb, lowest_reachable ? "" : "un");
+
+	if (!test_bit(o2nm_this_node(), qs->qs_hb_bm) ||
+	    qs->qs_heartbeating == 1)
+		goto out;
+
+	if (qs->qs_heartbeating & 1) {
+		/* the odd numbered cluster case is straight forward --
+		 * if we can't talk to the majority we're hosed */
+		quorum = (qs->qs_heartbeating + 1)/2;
+		if (qs->qs_connected < quorum) {
+			mlog(ML_ERROR, "fencing this node because it is "
+			     "only connected to %u nodes and %u is needed "
+			     "to make a quorum out of %u heartbeating nodes\n",
+			     qs->qs_connected, quorum,
+			     qs->qs_heartbeating);
+			fence = 1;
+		}
+	} else {
+		/* the even numbered cluster adds the possibility of each half
+		 * of the cluster being able to talk amongst themselves.. in
+		 * that case we're hosed if we can't talk to the group that has
+		 * the lowest numbered node */
+		quorum = qs->qs_heartbeating / 2;
+		if (qs->qs_connected < quorum) {
+			mlog(ML_ERROR, "fencing this node because it is "
+			     "only connected to %u nodes and %u is needed "
+			     "to make a quorum out of %u heartbeating nodes\n",
+			     qs->qs_connected, quorum,
+			     qs->qs_heartbeating);
+			fence = 1;
+		}
+		else if ((qs->qs_connected == quorum) &&
+			 !lowest_reachable) {
+			mlog(ML_ERROR, "fencing this node because it is "
+			     "connected to a half-quorum of %u out of %u "
+			     "nodes which doesn't include the lowest active "
+			     "node %u\n", quorum, qs->qs_heartbeating,
+			     lowest_hb);
+			fence = 1;
+		}
+	}
+
+out:
+	spin_unlock(&qs->qs_lock);
+	if (fence)
+		o2quo_fence_self();
+}
+
+static void o2quo_set_hold(struct o2quo_state *qs, u8 node)
+{
+	assert_spin_locked(&qs->qs_lock);
+
+	if (!test_and_set_bit(node, qs->qs_hold_bm)) {
+		qs->qs_holds++;
+		mlog_bug_on_msg(qs->qs_holds == O2NM_MAX_NODES,
+			        "node %u\n", node);
+		mlog(0, "node %u, %d total\n", node, qs->qs_holds);
+	}
+}
+
+static void o2quo_clear_hold(struct o2quo_state *qs, u8 node)
+{
+	assert_spin_locked(&qs->qs_lock);
+
+	if (test_and_clear_bit(node, qs->qs_hold_bm)) {
+		mlog(0, "node %u, %d total\n", node, qs->qs_holds - 1);
+		if (--qs->qs_holds == 0) {
+			if (qs->qs_pending) {
+				qs->qs_pending = 0;
+				schedule_work(&qs->qs_work);
+			}
+		}
+		mlog_bug_on_msg(qs->qs_holds < 0, "node %u, holds %d\n",
+				node, qs->qs_holds);
+	}
+}
+
+/* as a node comes up we delay the quorum decision until we know the fate of
+ * the connection.  the hold will be droped in conn_up or hb_down.  it might be
+ * perpetuated by con_err until hb_down.  if we already have a conn, we might
+ * be dropping a hold that conn_up got. */
+void o2quo_hb_up(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	qs->qs_heartbeating++;
+	mlog_bug_on_msg(qs->qs_heartbeating == O2NM_MAX_NODES,
+		        "node %u\n", node);
+	mlog_bug_on_msg(test_bit(node, qs->qs_hb_bm), "node %u\n", node);
+	set_bit(node, qs->qs_hb_bm);
+
+	mlog(0, "node %u, %d total\n", node, qs->qs_heartbeating);
+
+	if (!test_bit(node, qs->qs_conn_bm))
+		o2quo_set_hold(qs, node);
+	else
+		o2quo_clear_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+/* hb going down releases any holds we might have had due to this node from
+ * conn_up, conn_err, or hb_up */
+void o2quo_hb_down(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	qs->qs_heartbeating--;
+	mlog_bug_on_msg(qs->qs_heartbeating < 0,
+			"node %u, %d heartbeating\n",
+			node, qs->qs_heartbeating);
+	mlog_bug_on_msg(!test_bit(node, qs->qs_hb_bm), "node %u\n", node);
+	clear_bit(node, qs->qs_hb_bm);
+
+	mlog(0, "node %u, %d total\n", node, qs->qs_heartbeating);
+
+	o2quo_clear_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+/* this tells us that we've decided that the node is still heartbeating
+ * even though we've lost it's conn.  it must only be called after conn_err
+ * and indicates that we must now make a quorum decision in the future,
+ * though we might be doing so after waiting for holds to drain.  Here
+ * we'll be dropping the hold from conn_err. */
+void o2quo_hb_still_up(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	mlog(0, "node %u\n", node);
+
+	qs->qs_pending = 1;
+	o2quo_clear_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+/* This is analagous to hb_up.  as a node's connection comes up we delay the
+ * quorum decision until we see it heartbeating.  the hold will be droped in
+ * hb_up or hb_down.  it might be perpetuated by con_err until hb_down.  if
+ * it's already heartbeating we we might be dropping a hold that conn_up got.
+ * */
+void o2quo_conn_up(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	qs->qs_connected++;
+	mlog_bug_on_msg(qs->qs_connected == O2NM_MAX_NODES,
+		        "node %u\n", node);
+	mlog_bug_on_msg(test_bit(node, qs->qs_conn_bm), "node %u\n", node);
+	set_bit(node, qs->qs_conn_bm);
+
+	mlog(0, "node %u, %d total\n", node, qs->qs_connected);
+
+	if (!test_bit(node, qs->qs_hb_bm))
+		o2quo_set_hold(qs, node);
+	else
+		o2quo_clear_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+/* we've decided that we won't ever be connecting to the node again.  if it's
+ * still heartbeating we grab a hold that will delay decisions until either the
+ * node stops heartbeating from hb_down or the caller decides that the node is
+ * still up and calls still_up */
+void o2quo_conn_err(u8 node)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock(&qs->qs_lock);
+
+	if (test_bit(node, qs->qs_conn_bm)) {
+		qs->qs_connected--;
+		mlog_bug_on_msg(qs->qs_connected < 0,
+				"node %u, connected %d\n",
+				node, qs->qs_connected);
+
+		clear_bit(node, qs->qs_conn_bm);
+	}
+
+	mlog(0, "node %u, %d total\n", node, qs->qs_connected);
+
+	if (test_bit(node, qs->qs_hb_bm))
+		o2quo_set_hold(qs, node);
+
+	spin_unlock(&qs->qs_lock);
+}
+
+void o2quo_init(void)
+{
+	struct o2quo_state *qs = &o2quo_state;
+
+	spin_lock_init(&qs->qs_lock);
+	INIT_WORK(&qs->qs_work, o2quo_make_decision, NULL);
+}
+
+void o2quo_exit(void)
+{
+	flush_scheduled_work();
+}
diff --git a/fs/ocfs2/cluster/quorum.h b/fs/ocfs2/cluster/quorum.h
new file mode 100644
index 00000000000000..6649cc6f67c9c7
--- /dev/null
+++ b/fs/ocfs2/cluster/quorum.h
@@ -0,0 +1,36 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_QUORUM_H
+#define O2CLUSTER_QUORUM_H
+
+void o2quo_init(void);
+void o2quo_exit(void);
+
+void o2quo_hb_up(u8 node);
+void o2quo_hb_down(u8 node);
+void o2quo_hb_still_up(u8 node);
+void o2quo_conn_up(u8 node);
+void o2quo_conn_err(u8 node);
+void o2quo_disk_timeout(void);
+
+#endif /* O2CLUSTER_QUORUM_H */
diff --git a/fs/ocfs2/cluster/sys.c b/fs/ocfs2/cluster/sys.c
new file mode 100644
index 00000000000000..1d9f6acafa2e50
--- /dev/null
+++ b/fs/ocfs2/cluster/sys.c
@@ -0,0 +1,124 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * sys.c
+ *
+ * OCFS2 cluster sysfs interface
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation,
+ * version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+
+#include "ocfs2_nodemanager.h"
+#include "masklog.h"
+#include "sys.h"
+
+struct o2cb_attribute {
+	struct attribute	attr;
+	ssize_t (*show)(char *buf);
+	ssize_t (*store)(const char *buf, size_t count);
+};
+
+#define O2CB_ATTR(_name, _mode, _show, _store)	\
+struct o2cb_attribute o2cb_attr_##_name = __ATTR(_name, _mode, _show, _store)
+
+#define to_o2cb_subsys(k) container_of(to_kset(k), struct subsystem, kset)
+#define to_o2cb_attr(_attr) container_of(_attr, struct o2cb_attribute, attr)
+
+static ssize_t o2cb_interface_revision_show(char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%u\n", O2NM_API_VERSION);
+}
+
+static O2CB_ATTR(interface_revision, S_IFREG | S_IRUGO, o2cb_interface_revision_show, NULL);
+
+static struct attribute *o2cb_attrs[] = {
+	&o2cb_attr_interface_revision.attr,
+	NULL,
+};
+
+static ssize_t
+o2cb_show(struct kobject * kobj, struct attribute * attr, char * buffer);
+static ssize_t
+o2cb_store(struct kobject * kobj, struct attribute * attr,
+	   const char * buffer, size_t count);
+static struct sysfs_ops o2cb_sysfs_ops = {
+	.show	= o2cb_show,
+	.store	= o2cb_store,
+};
+
+static struct kobj_type o2cb_subsys_type = {
+	.default_attrs	= o2cb_attrs,
+	.sysfs_ops	= &o2cb_sysfs_ops,
+};
+
+/* gives us o2cb_subsys */
+static decl_subsys(o2cb, NULL, NULL);
+
+static ssize_t
+o2cb_show(struct kobject * kobj, struct attribute * attr, char * buffer)
+{
+	struct o2cb_attribute *o2cb_attr = to_o2cb_attr(attr);
+	struct subsystem *sbs = to_o2cb_subsys(kobj);
+
+	BUG_ON(sbs != &o2cb_subsys);
+
+	if (o2cb_attr->show)
+		return o2cb_attr->show(buffer);
+	return -EIO;
+}
+
+static ssize_t
+o2cb_store(struct kobject * kobj, struct attribute * attr,
+	     const char * buffer, size_t count)
+{
+	struct o2cb_attribute *o2cb_attr = to_o2cb_attr(attr);
+	struct subsystem *sbs = to_o2cb_subsys(kobj);
+
+	BUG_ON(sbs != &o2cb_subsys);
+
+	if (o2cb_attr->store)
+		return o2cb_attr->store(buffer, count);
+	return -EIO;
+}
+
+void o2cb_sys_shutdown(void)
+{
+	mlog_sys_shutdown();
+	subsystem_unregister(&o2cb_subsys);
+}
+
+int o2cb_sys_init(void)
+{
+	int ret;
+
+	o2cb_subsys.kset.kobj.ktype = &o2cb_subsys_type;
+	ret = subsystem_register(&o2cb_subsys);
+	if (ret)
+		return ret;
+
+	ret = mlog_sys_init(&o2cb_subsys);
+	if (ret)
+		subsystem_unregister(&o2cb_subsys);
+	return ret;
+}
diff --git a/fs/ocfs2/cluster/sys.h b/fs/ocfs2/cluster/sys.h
new file mode 100644
index 00000000000000..d66b8ab0045ecf
--- /dev/null
+++ b/fs/ocfs2/cluster/sys.h
@@ -0,0 +1,33 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * sys.h
+ *
+ * Function prototypes for o2cb sysfs interface
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation,
+ * version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_SYS_H
+#define O2CLUSTER_SYS_H
+
+void o2cb_sys_shutdown(void);
+int o2cb_sys_init(void);
+
+#endif /* O2CLUSTER_SYS_H */
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
new file mode 100644
index 00000000000000..35d92c01a97241
--- /dev/null
+++ b/fs/ocfs2/cluster/tcp.c
@@ -0,0 +1,1829 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ *
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ * ----
+ *
+ * Callers for this were originally written against a very simple synchronus
+ * API.  This implementation reflects those simple callers.  Some day I'm sure
+ * we'll need to move to a more robust posting/callback mechanism.
+ *
+ * Transmit calls pass in kernel virtual addresses and block copying this into
+ * the socket's tx buffers via a usual blocking sendmsg.  They'll block waiting
+ * for a failed socket to timeout.  TX callers can also pass in a poniter to an
+ * 'int' which gets filled with an errno off the wire in response to the
+ * message they send.
+ *
+ * Handlers for unsolicited messages are registered.  Each socket has a page
+ * that incoming data is copied into.  First the header, then the data.
+ * Handlers are called from only one thread with a reference to this per-socket
+ * page.  This page is destroyed after the handler call, so it can't be
+ * referenced beyond the call.  Handlers may block but are discouraged from
+ * doing so.
+ *
+ * Any framing errors (bad magic, large payload lengths) close a connection.
+ *
+ * Our sock_container holds the state we associate with a socket.  It's current
+ * framing state is held there as well as the refcounting we do around when it
+ * is safe to tear down the socket.  The socket is only finally torn down from
+ * the container when the container loses all of its references -- so as long
+ * as you hold a ref on the container you can trust that the socket is valid
+ * for use with kernel socket APIs.
+ *
+ * Connections are initiated between a pair of nodes when the node with the
+ * higher node number gets a heartbeat callback which indicates that the lower
+ * numbered node has started heartbeating.  The lower numbered node is passive
+ * and only accepts the connection if the higher numbered node is heartbeating.
+ */
+
+#include <linux/kernel.h>
+#include <linux/jiffies.h>
+#include <linux/slab.h>
+#include <linux/idr.h>
+#include <linux/kref.h>
+#include <net/tcp.h>
+
+#include <asm/uaccess.h>
+
+#include "heartbeat.h"
+#include "tcp.h"
+#include "nodemanager.h"
+#define MLOG_MASK_PREFIX ML_TCP
+#include "masklog.h"
+#include "quorum.h"
+
+#include "tcp_internal.h"
+
+/* 
+ * The linux network stack isn't sparse endian clean.. It has macros like
+ * ntohs() which perform the endian checks and structs like sockaddr_in
+ * which aren't annotated.  So __force is found here to get the build
+ * clean.  When they emerge from the dark ages and annotate the code
+ * we can remove these.
+ */
+
+#define SC_NODEF_FMT "node %s (num %u) at %u.%u.%u.%u:%u"
+#define SC_NODEF_ARGS(sc) sc->sc_node->nd_name, sc->sc_node->nd_num,	\
+			  NIPQUAD(sc->sc_node->nd_ipv4_address),	\
+			  ntohs(sc->sc_node->nd_ipv4_port)
+
+/*
+ * In the following two log macros, the whitespace after the ',' just
+ * before ##args is intentional. Otherwise, gcc 2.95 will eat the
+ * previous token if args expands to nothing.
+ */
+#define msglog(hdr, fmt, args...) do {					\
+	typeof(hdr) __hdr = (hdr);					\
+	mlog(ML_MSG, "[mag %u len %u typ %u stat %d sys_stat %d "	\
+	     "key %08x num %u] " fmt,					\
+	     be16_to_cpu(__hdr->magic), be16_to_cpu(__hdr->data_len), 	\
+	     be16_to_cpu(__hdr->msg_type), be32_to_cpu(__hdr->status),	\
+	     be32_to_cpu(__hdr->sys_status), be32_to_cpu(__hdr->key),	\
+	     be32_to_cpu(__hdr->msg_num) ,  ##args);			\
+} while (0)
+
+#define sclog(sc, fmt, args...) do {					\
+	typeof(sc) __sc = (sc);						\
+	mlog(ML_SOCKET, "[sc %p refs %d sock %p node %u page %p "	\
+	     "pg_off %zu] " fmt, __sc,					\
+	     atomic_read(&__sc->sc_kref.refcount), __sc->sc_sock,	\
+	    __sc->sc_node->nd_num, __sc->sc_page, __sc->sc_page_off ,	\
+	    ##args);							\
+} while (0)
+
+static rwlock_t o2net_handler_lock = RW_LOCK_UNLOCKED;
+static struct rb_root o2net_handler_tree = RB_ROOT;
+
+static struct o2net_node o2net_nodes[O2NM_MAX_NODES];
+
+/* XXX someday we'll need better accounting */
+static struct socket *o2net_listen_sock = NULL;
+
+/*
+ * listen work is only queued by the listening socket callbacks on the
+ * o2net_wq.  teardown detaches the callbacks before destroying the workqueue.
+ * quorum work is queued as sock containers are shutdown.. stop_listening
+ * tears down all the node's sock containers, preventing future shutdowns
+ * and queued quroum work, before canceling delayed quorum work and
+ * destroying the work queue.
+ */
+static struct workqueue_struct *o2net_wq;
+static struct work_struct o2net_listen_work;
+
+static struct o2hb_callback_func o2net_hb_up, o2net_hb_down;
+#define O2NET_HB_PRI 0x1
+
+static struct o2net_handshake *o2net_hand;
+static struct o2net_msg *o2net_keep_req, *o2net_keep_resp;
+
+static int o2net_sys_err_translations[O2NET_ERR_MAX] =
+		{[O2NET_ERR_NONE]	= 0,
+		 [O2NET_ERR_NO_HNDLR]	= -ENOPROTOOPT,
+		 [O2NET_ERR_OVERFLOW]	= -EOVERFLOW,
+		 [O2NET_ERR_DIED]	= -EHOSTDOWN,};
+
+/* can't quite avoid *all* internal declarations :/ */
+static void o2net_sc_connect_completed(void *arg);
+static void o2net_rx_until_empty(void *arg);
+static void o2net_shutdown_sc(void *arg);
+static void o2net_listen_data_ready(struct sock *sk, int bytes);
+static void o2net_sc_send_keep_req(void *arg);
+static void o2net_idle_timer(unsigned long data);
+static void o2net_sc_postpone_idle(struct o2net_sock_container *sc);
+
+static inline int o2net_sys_err_to_errno(enum o2net_system_error err)
+{
+	int trans;
+	BUG_ON(err >= O2NET_ERR_MAX);
+	trans = o2net_sys_err_translations[err];
+
+	/* Just in case we mess up the translation table above */
+	BUG_ON(err != O2NET_ERR_NONE && trans == 0);
+	return trans;
+}
+
+static struct o2net_node * o2net_nn_from_num(u8 node_num)
+{
+	BUG_ON(node_num >= ARRAY_SIZE(o2net_nodes));
+	return &o2net_nodes[node_num];
+}
+
+static u8 o2net_num_from_nn(struct o2net_node *nn)
+{
+	BUG_ON(nn == NULL);
+	return nn - o2net_nodes;
+}
+
+/* ------------------------------------------------------------ */
+
+static int o2net_prep_nsw(struct o2net_node *nn, struct o2net_status_wait *nsw)
+{
+	int ret = 0;
+
+	do {
+		if (!idr_pre_get(&nn->nn_status_idr, GFP_ATOMIC)) {
+			ret = -EAGAIN;
+			break;
+		}
+		spin_lock(&nn->nn_lock);
+		ret = idr_get_new(&nn->nn_status_idr, nsw, &nsw->ns_id);
+		if (ret == 0)
+			list_add_tail(&nsw->ns_node_item,
+				      &nn->nn_status_list);
+		spin_unlock(&nn->nn_lock);
+	} while (ret == -EAGAIN);
+
+	if (ret == 0)  {
+		init_waitqueue_head(&nsw->ns_wq);
+		nsw->ns_sys_status = O2NET_ERR_NONE;
+		nsw->ns_status = 0;
+	}
+
+	return ret;
+}
+
+static void o2net_complete_nsw_locked(struct o2net_node *nn,
+				      struct o2net_status_wait *nsw,
+				      enum o2net_system_error sys_status,
+				      s32 status)
+{
+	assert_spin_locked(&nn->nn_lock);
+
+	if (!list_empty(&nsw->ns_node_item)) {
+		list_del_init(&nsw->ns_node_item);
+		nsw->ns_sys_status = sys_status;
+		nsw->ns_status = status;
+		idr_remove(&nn->nn_status_idr, nsw->ns_id);
+		wake_up(&nsw->ns_wq);
+	}
+}
+
+static void o2net_complete_nsw(struct o2net_node *nn,
+			       struct o2net_status_wait *nsw,
+			       u64 id, enum o2net_system_error sys_status,
+			       s32 status)
+{
+	spin_lock(&nn->nn_lock);
+	if (nsw == NULL) {
+		if (id > INT_MAX)
+			goto out;
+
+		nsw = idr_find(&nn->nn_status_idr, id);
+		if (nsw == NULL)
+			goto out;
+	}
+
+	o2net_complete_nsw_locked(nn, nsw, sys_status, status);
+
+out:
+	spin_unlock(&nn->nn_lock);
+	return;
+}
+
+static void o2net_complete_nodes_nsw(struct o2net_node *nn)
+{
+	struct list_head *iter, *tmp;
+	unsigned int num_kills = 0;
+	struct o2net_status_wait *nsw;
+
+	assert_spin_locked(&nn->nn_lock);
+
+	list_for_each_safe(iter, tmp, &nn->nn_status_list) {
+		nsw = list_entry(iter, struct o2net_status_wait, ns_node_item);
+		o2net_complete_nsw_locked(nn, nsw, O2NET_ERR_DIED, 0);
+		num_kills++;
+	}
+
+	mlog(0, "completed %d messages for node %u\n", num_kills,
+	     o2net_num_from_nn(nn));
+}
+
+static int o2net_nsw_completed(struct o2net_node *nn,
+			       struct o2net_status_wait *nsw)
+{
+	int completed;
+	spin_lock(&nn->nn_lock);
+	completed = list_empty(&nsw->ns_node_item);
+	spin_unlock(&nn->nn_lock);
+	return completed;
+}
+
+/* ------------------------------------------------------------ */
+
+static void sc_kref_release(struct kref *kref)
+{
+	struct o2net_sock_container *sc = container_of(kref,
+					struct o2net_sock_container, sc_kref);
+	sclog(sc, "releasing\n");
+
+	if (sc->sc_sock) {
+		sock_release(sc->sc_sock);
+		sc->sc_sock = NULL;
+	}
+
+	o2nm_node_put(sc->sc_node);
+	sc->sc_node = NULL;
+
+	kfree(sc);
+}
+
+static void sc_put(struct o2net_sock_container *sc)
+{
+	sclog(sc, "put\n");
+	kref_put(&sc->sc_kref, sc_kref_release);
+}
+static void sc_get(struct o2net_sock_container *sc)
+{
+	sclog(sc, "get\n");
+	kref_get(&sc->sc_kref);
+}
+static struct o2net_sock_container *sc_alloc(struct o2nm_node *node)
+{
+	struct o2net_sock_container *sc, *ret = NULL;
+	struct page *page = NULL;
+
+	page = alloc_page(GFP_NOFS);
+	sc = kcalloc(1, sizeof(*sc), GFP_NOFS);
+	if (sc == NULL || page == NULL)
+		goto out;
+
+	kref_init(&sc->sc_kref);
+	o2nm_node_get(node);
+	sc->sc_node = node;
+
+	INIT_WORK(&sc->sc_connect_work, o2net_sc_connect_completed, sc);
+	INIT_WORK(&sc->sc_rx_work, o2net_rx_until_empty, sc);
+	INIT_WORK(&sc->sc_shutdown_work, o2net_shutdown_sc, sc);
+	INIT_WORK(&sc->sc_keepalive_work, o2net_sc_send_keep_req, sc);
+
+	init_timer(&sc->sc_idle_timeout);
+	sc->sc_idle_timeout.function = o2net_idle_timer;
+	sc->sc_idle_timeout.data = (unsigned long)sc;
+
+	sclog(sc, "alloced\n");
+
+	ret = sc;
+	sc->sc_page = page;
+	sc = NULL;
+	page = NULL;
+
+out:
+	if (page)
+		__free_page(page);
+	kfree(sc);
+
+	return ret;
+}
+
+/* ------------------------------------------------------------ */
+
+static void o2net_sc_queue_work(struct o2net_sock_container *sc,
+				struct work_struct *work)
+{
+	sc_get(sc);
+	if (!queue_work(o2net_wq, work))
+		sc_put(sc);
+}
+static void o2net_sc_queue_delayed_work(struct o2net_sock_container *sc,
+					struct work_struct *work,
+					int delay)
+{
+	sc_get(sc);
+	if (!queue_delayed_work(o2net_wq, work, delay))
+		sc_put(sc);
+}
+static void o2net_sc_cancel_delayed_work(struct o2net_sock_container *sc,
+					 struct work_struct *work)
+{
+	if (cancel_delayed_work(work))
+		sc_put(sc);
+}
+
+static void o2net_set_nn_state(struct o2net_node *nn,
+			       struct o2net_sock_container *sc,
+			       unsigned valid, int err)
+{
+	int was_valid = nn->nn_sc_valid;
+	int was_err = nn->nn_persistent_error;
+	struct o2net_sock_container *old_sc = nn->nn_sc;
+
+	assert_spin_locked(&nn->nn_lock);
+
+	/* the node num comparison and single connect/accept path should stop
+	 * an non-null sc from being overwritten with another */
+	BUG_ON(sc && nn->nn_sc && nn->nn_sc != sc);
+	mlog_bug_on_msg(err && valid, "err %d valid %u\n", err, valid);
+	mlog_bug_on_msg(valid && !sc, "valid %u sc %p\n", valid, sc);
+
+	/* we won't reconnect after our valid conn goes away for
+	 * this hb iteration.. here so it shows up in the logs */
+	if (was_valid && !valid && err == 0)
+		err = -ENOTCONN;
+
+	mlog(ML_CONN, "node %u sc: %p -> %p, valid %u -> %u, err %d -> %d\n",
+	     o2net_num_from_nn(nn), nn->nn_sc, sc, nn->nn_sc_valid, valid,
+	     nn->nn_persistent_error, err);
+
+	nn->nn_sc = sc;
+	nn->nn_sc_valid = valid ? 1 : 0;
+	nn->nn_persistent_error = err;
+
+	/* mirrors o2net_tx_can_proceed() */
+	if (nn->nn_persistent_error || nn->nn_sc_valid)
+		wake_up(&nn->nn_sc_wq);
+
+	if (!was_err && nn->nn_persistent_error) {
+		o2quo_conn_err(o2net_num_from_nn(nn));
+		queue_delayed_work(o2net_wq, &nn->nn_still_up,
+				   msecs_to_jiffies(O2NET_QUORUM_DELAY_MS));
+	}
+
+	if (was_valid && !valid) {
+		mlog(ML_NOTICE, "no longer connected to " SC_NODEF_FMT "\n",
+		     SC_NODEF_ARGS(old_sc));
+		o2net_complete_nodes_nsw(nn);
+	}
+
+	if (!was_valid && valid) {
+		o2quo_conn_up(o2net_num_from_nn(nn));
+		/* this is a bit of a hack.  we only try reconnecting
+		 * when heartbeating starts until we get a connection.
+		 * if that connection then dies we don't try reconnecting.
+		 * the only way to start connecting again is to down
+		 * heartbeat and bring it back up. */
+		cancel_delayed_work(&nn->nn_connect_expired);
+		mlog(ML_NOTICE, "%s " SC_NODEF_FMT "\n", 
+		     o2nm_this_node() > sc->sc_node->nd_num ?
+		     	"connected to" : "accepted connection from",
+		     SC_NODEF_ARGS(sc));
+	}
+
+	/* trigger the connecting worker func as long as we're not valid,
+	 * it will back off if it shouldn't connect.  This can be called
+	 * from node config teardown and so needs to be careful about
+	 * the work queue actually being up. */
+	if (!valid && o2net_wq) {
+		unsigned long delay;
+		/* delay if we're withing a RECONNECT_DELAY of the
+		 * last attempt */
+		delay = (nn->nn_last_connect_attempt +
+			 msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS))
+			- jiffies;
+		if (delay > msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS))
+			delay = 0;
+		mlog(ML_CONN, "queueing conn attempt in %lu jiffies\n", delay);
+		queue_delayed_work(o2net_wq, &nn->nn_connect_work, delay);
+	}
+
+	/* keep track of the nn's sc ref for the caller */
+	if ((old_sc == NULL) && sc)
+		sc_get(sc);
+	if (old_sc && (old_sc != sc)) {
+		o2net_sc_queue_work(old_sc, &old_sc->sc_shutdown_work);
+		sc_put(old_sc);
+	}
+}
+
+/* see o2net_register_callbacks() */
+static void o2net_data_ready(struct sock *sk, int bytes)
+{
+	void (*ready)(struct sock *sk, int bytes);
+
+	read_lock(&sk->sk_callback_lock);
+	if (sk->sk_user_data) {
+		struct o2net_sock_container *sc = sk->sk_user_data;
+		sclog(sc, "data_ready hit\n");
+		do_gettimeofday(&sc->sc_tv_data_ready);
+		o2net_sc_queue_work(sc, &sc->sc_rx_work);
+		ready = sc->sc_data_ready;
+	} else {
+		ready = sk->sk_data_ready;
+	}
+	read_unlock(&sk->sk_callback_lock);
+
+	ready(sk, bytes);
+}
+
+/* see o2net_register_callbacks() */
+static void o2net_state_change(struct sock *sk)
+{
+	void (*state_change)(struct sock *sk);
+	struct o2net_sock_container *sc;
+
+	read_lock(&sk->sk_callback_lock);
+	sc = sk->sk_user_data;
+	if (sc == NULL) {
+		state_change = sk->sk_state_change;
+		goto out;
+	}
+
+	sclog(sc, "state_change to %d\n", sk->sk_state);
+
+	state_change = sc->sc_state_change;
+
+	switch(sk->sk_state) {
+		/* ignore connecting sockets as they make progress */
+		case TCP_SYN_SENT:
+		case TCP_SYN_RECV:
+			break;
+		case TCP_ESTABLISHED:
+			o2net_sc_queue_work(sc, &sc->sc_connect_work);
+			break;
+		default:
+			o2net_sc_queue_work(sc, &sc->sc_shutdown_work);
+			break;
+	}
+out:
+	read_unlock(&sk->sk_callback_lock);
+	state_change(sk);
+}
+
+/*
+ * we register callbacks so we can queue work on events before calling
+ * the original callbacks.  our callbacks our careful to test user_data
+ * to discover when they've reaced with o2net_unregister_callbacks().
+ */
+static void o2net_register_callbacks(struct sock *sk,
+				     struct o2net_sock_container *sc)
+{
+	write_lock_bh(&sk->sk_callback_lock);
+
+	/* accepted sockets inherit the old listen socket data ready */
+	if (sk->sk_data_ready == o2net_listen_data_ready) {
+		sk->sk_data_ready = sk->sk_user_data;
+		sk->sk_user_data = NULL;
+	}
+
+	BUG_ON(sk->sk_user_data != NULL);
+	sk->sk_user_data = sc;
+	sc_get(sc);
+
+	sc->sc_data_ready = sk->sk_data_ready;
+	sc->sc_state_change = sk->sk_state_change;
+	sk->sk_data_ready = o2net_data_ready;
+	sk->sk_state_change = o2net_state_change;
+
+	write_unlock_bh(&sk->sk_callback_lock);
+}
+
+static int o2net_unregister_callbacks(struct sock *sk,
+			           struct o2net_sock_container *sc)
+{
+	int ret = 0;
+
+	write_lock_bh(&sk->sk_callback_lock);
+	if (sk->sk_user_data == sc) {
+		ret = 1;
+		sk->sk_user_data = NULL;
+		sk->sk_data_ready = sc->sc_data_ready;
+		sk->sk_state_change = sc->sc_state_change;
+	}
+	write_unlock_bh(&sk->sk_callback_lock);
+
+	return ret;
+}
+
+/*
+ * this is a little helper that is called by callers who have seen a problem
+ * with an sc and want to detach it from the nn if someone already hasn't beat
+ * them to it.  if an error is given then the shutdown will be persistent
+ * and pending transmits will be canceled.
+ */
+static void o2net_ensure_shutdown(struct o2net_node *nn,
+			           struct o2net_sock_container *sc,
+				   int err)
+{
+	spin_lock(&nn->nn_lock);
+	if (nn->nn_sc == sc)
+		o2net_set_nn_state(nn, NULL, 0, err);
+	spin_unlock(&nn->nn_lock);
+}
+
+/*
+ * This work queue function performs the blocking parts of socket shutdown.  A
+ * few paths lead here.  set_nn_state will trigger this callback if it sees an
+ * sc detached from the nn.  state_change will also trigger this callback
+ * directly when it sees errors.  In that case we need to call set_nn_state
+ * ourselves as state_change couldn't get the nn_lock and call set_nn_state
+ * itself.
+ */
+static void o2net_shutdown_sc(void *arg)
+{
+	struct o2net_sock_container *sc = arg;
+	struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
+
+	sclog(sc, "shutting down\n");
+
+	/* drop the callbacks ref and call shutdown only once */
+	if (o2net_unregister_callbacks(sc->sc_sock->sk, sc)) {
+		/* we shouldn't flush as we're in the thread, the
+		 * races with pending sc work structs are harmless */
+		del_timer_sync(&sc->sc_idle_timeout);
+		o2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work);
+		sc_put(sc);
+		sc->sc_sock->ops->shutdown(sc->sc_sock,
+					   RCV_SHUTDOWN|SEND_SHUTDOWN);
+	}
+
+	/* not fatal so failed connects before the other guy has our
+	 * heartbeat can be retried */
+	o2net_ensure_shutdown(nn, sc, 0);
+	sc_put(sc);
+}
+
+/* ------------------------------------------------------------ */
+
+static int o2net_handler_cmp(struct o2net_msg_handler *nmh, u32 msg_type,
+			     u32 key)
+{
+	int ret = memcmp(&nmh->nh_key, &key, sizeof(key));
+
+	if (ret == 0)
+		ret = memcmp(&nmh->nh_msg_type, &msg_type, sizeof(msg_type));
+
+	return ret;
+}
+
+static struct o2net_msg_handler *
+o2net_handler_tree_lookup(u32 msg_type, u32 key, struct rb_node ***ret_p,
+			  struct rb_node **ret_parent)
+{
+        struct rb_node **p = &o2net_handler_tree.rb_node;
+        struct rb_node *parent = NULL;
+	struct o2net_msg_handler *nmh, *ret = NULL;
+	int cmp;
+
+        while (*p) {
+                parent = *p;
+                nmh = rb_entry(parent, struct o2net_msg_handler, nh_node);
+		cmp = o2net_handler_cmp(nmh, msg_type, key);
+
+                if (cmp < 0)
+                        p = &(*p)->rb_left;
+                else if (cmp > 0)
+                        p = &(*p)->rb_right;
+                else {
+			ret = nmh;
+                        break;
+		}
+        }
+
+        if (ret_p != NULL)
+                *ret_p = p;
+        if (ret_parent != NULL)
+                *ret_parent = parent;
+
+        return ret;
+}
+
+static void o2net_handler_kref_release(struct kref *kref)
+{
+	struct o2net_msg_handler *nmh;
+	nmh = container_of(kref, struct o2net_msg_handler, nh_kref);
+
+	kfree(nmh);
+}
+
+static void o2net_handler_put(struct o2net_msg_handler *nmh)
+{
+	kref_put(&nmh->nh_kref, o2net_handler_kref_release);
+}
+
+/* max_len is protection for the handler func.  incoming messages won't
+ * be given to the handler if their payload is longer than the max. */
+int o2net_register_handler(u32 msg_type, u32 key, u32 max_len,
+			   o2net_msg_handler_func *func, void *data,
+			   struct list_head *unreg_list)
+{
+	struct o2net_msg_handler *nmh = NULL;
+	struct rb_node **p, *parent;
+	int ret = 0;
+
+	if (max_len > O2NET_MAX_PAYLOAD_BYTES) {
+		mlog(0, "max_len for message handler out of range: %u\n",
+			max_len);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (!msg_type) {
+		mlog(0, "no message type provided: %u, %p\n", msg_type, func);
+		ret = -EINVAL;
+		goto out;
+
+	}
+	if (!func) {
+		mlog(0, "no message handler provided: %u, %p\n",
+		       msg_type, func);
+		ret = -EINVAL;
+		goto out;
+	}
+
+       	nmh = kcalloc(1, sizeof(struct o2net_msg_handler), GFP_NOFS);
+	if (nmh == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	nmh->nh_func = func;
+	nmh->nh_func_data = data;
+	nmh->nh_msg_type = msg_type;
+	nmh->nh_max_len = max_len;
+	nmh->nh_key = key;
+	/* the tree and list get this ref.. they're both removed in
+	 * unregister when this ref is dropped */
+	kref_init(&nmh->nh_kref);
+	INIT_LIST_HEAD(&nmh->nh_unregister_item);
+
+	write_lock(&o2net_handler_lock);
+	if (o2net_handler_tree_lookup(msg_type, key, &p, &parent))
+		ret = -EEXIST;
+	else {
+	        rb_link_node(&nmh->nh_node, parent, p);
+		rb_insert_color(&nmh->nh_node, &o2net_handler_tree);
+		list_add_tail(&nmh->nh_unregister_item, unreg_list);
+
+		mlog(ML_TCP, "registered handler func %p type %u key %08x\n",
+		     func, msg_type, key);
+		/* we've had some trouble with handlers seemingly vanishing. */
+		mlog_bug_on_msg(o2net_handler_tree_lookup(msg_type, key, &p,
+							  &parent) == NULL,
+			        "couldn't find handler we *just* registerd "
+				"for type %u key %08x\n", msg_type, key);
+	}
+	write_unlock(&o2net_handler_lock);
+	if (ret)
+		goto out;
+
+out:
+	if (ret)
+		kfree(nmh);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(o2net_register_handler);
+
+void o2net_unregister_handler_list(struct list_head *list)
+{
+	struct list_head *pos, *n;
+	struct o2net_msg_handler *nmh;
+
+	write_lock(&o2net_handler_lock);
+	list_for_each_safe(pos, n, list) {
+		nmh = list_entry(pos, struct o2net_msg_handler,
+				 nh_unregister_item);
+		mlog(ML_TCP, "unregistering handler func %p type %u key %08x\n",
+		     nmh->nh_func, nmh->nh_msg_type, nmh->nh_key);
+		rb_erase(&nmh->nh_node, &o2net_handler_tree);
+		list_del_init(&nmh->nh_unregister_item);
+		kref_put(&nmh->nh_kref, o2net_handler_kref_release);
+	}
+	write_unlock(&o2net_handler_lock);
+}
+EXPORT_SYMBOL_GPL(o2net_unregister_handler_list);
+
+static struct o2net_msg_handler *o2net_handler_get(u32 msg_type, u32 key)
+{
+	struct o2net_msg_handler *nmh;
+
+	read_lock(&o2net_handler_lock);
+	nmh = o2net_handler_tree_lookup(msg_type, key, NULL, NULL);
+	if (nmh)
+		kref_get(&nmh->nh_kref);
+	read_unlock(&o2net_handler_lock);
+
+	return nmh;
+}
+
+/* ------------------------------------------------------------ */
+
+static int o2net_recv_tcp_msg(struct socket *sock, void *data, size_t len)
+{
+	int ret;
+	mm_segment_t oldfs;
+	struct kvec vec = {
+		.iov_len = len,
+		.iov_base = data,
+	};
+	struct msghdr msg = {
+		.msg_iovlen = 1,
+		.msg_iov = (struct iovec *)&vec,
+       		.msg_flags = MSG_DONTWAIT,
+	};
+
+	oldfs = get_fs();
+	set_fs(get_ds());
+	ret = sock_recvmsg(sock, &msg, len, msg.msg_flags);
+	set_fs(oldfs);
+
+	return ret;
+}
+
+static int o2net_send_tcp_msg(struct socket *sock, struct kvec *vec,
+			      size_t veclen, size_t total)
+{
+	int ret;
+	mm_segment_t oldfs;
+	struct msghdr msg = {
+		.msg_iov = (struct iovec *)vec,
+		.msg_iovlen = veclen,
+	};
+
+	if (sock == NULL) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	oldfs = get_fs();
+	set_fs(get_ds());
+	ret = sock_sendmsg(sock, &msg, total);
+	set_fs(oldfs);
+	if (ret != total) {
+		mlog(ML_ERROR, "sendmsg returned %d instead of %zu\n", ret,
+		     total);
+		if (ret >= 0)
+			ret = -EPIPE; /* should be smarter, I bet */
+		goto out;
+	}
+
+	ret = 0;
+out:
+	if (ret < 0)
+		mlog(0, "returning error: %d\n", ret);
+	return ret;
+}
+
+static void o2net_sendpage(struct o2net_sock_container *sc,
+			   void *kmalloced_virt,
+			   size_t size)
+{
+	struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
+	ssize_t ret;
+
+
+	ret = sc->sc_sock->ops->sendpage(sc->sc_sock,
+					 virt_to_page(kmalloced_virt),
+					 (long)kmalloced_virt & ~PAGE_MASK,
+					 size, MSG_DONTWAIT);
+	if (ret != size) {
+		mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT 
+		     " failed with %zd\n", size, SC_NODEF_ARGS(sc), ret);
+		o2net_ensure_shutdown(nn, sc, 0);
+	}
+}
+
+static void o2net_init_msg(struct o2net_msg *msg, u16 data_len, u16 msg_type, u32 key)
+{
+	memset(msg, 0, sizeof(struct o2net_msg));
+	msg->magic = cpu_to_be16(O2NET_MSG_MAGIC);
+	msg->data_len = cpu_to_be16(data_len);
+	msg->msg_type = cpu_to_be16(msg_type);
+	msg->sys_status = cpu_to_be32(O2NET_ERR_NONE);
+	msg->status = 0;
+	msg->key = cpu_to_be32(key);
+}
+
+static int o2net_tx_can_proceed(struct o2net_node *nn,
+			        struct o2net_sock_container **sc_ret,
+				int *error)
+{
+	int ret = 0;
+
+	spin_lock(&nn->nn_lock);
+	if (nn->nn_persistent_error) {
+		ret = 1;
+		*sc_ret = NULL;
+		*error = nn->nn_persistent_error;
+	} else if (nn->nn_sc_valid) {
+		kref_get(&nn->nn_sc->sc_kref);
+
+		ret = 1;
+		*sc_ret = nn->nn_sc;
+		*error = 0;
+	}
+	spin_unlock(&nn->nn_lock);
+
+	return ret;
+}
+
+int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec,
+			   size_t caller_veclen, u8 target_node, int *status)
+{
+	int ret, error = 0;
+	struct o2net_msg *msg = NULL;
+	size_t veclen, caller_bytes = 0;
+	struct kvec *vec = NULL;
+	struct o2net_sock_container *sc = NULL;
+	struct o2net_node *nn = o2net_nn_from_num(target_node);
+	struct o2net_status_wait nsw = {
+		.ns_node_item = LIST_HEAD_INIT(nsw.ns_node_item),
+	};
+
+	if (o2net_wq == NULL) {
+		mlog(0, "attempt to tx without o2netd running\n");
+		ret = -ESRCH;
+		goto out;
+	}
+
+	if (caller_veclen == 0) {
+		mlog(0, "bad kvec array length\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	caller_bytes = iov_length((struct iovec *)caller_vec, caller_veclen);
+	if (caller_bytes > O2NET_MAX_PAYLOAD_BYTES) {
+		mlog(0, "total payload len %zu too large\n", caller_bytes);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (target_node == o2nm_this_node()) {
+		ret = -ELOOP;
+		goto out;
+	}
+
+	ret = wait_event_interruptible(nn->nn_sc_wq,
+				       o2net_tx_can_proceed(nn, &sc, &error));
+	if (!ret && error)
+		ret = error;
+	if (ret)
+		goto out;
+
+	veclen = caller_veclen + 1;
+	vec = kmalloc(sizeof(struct kvec) * veclen, GFP_ATOMIC);
+	if (vec == NULL) {
+		mlog(0, "failed to %zu element kvec!\n", veclen);
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	msg = kmalloc(sizeof(struct o2net_msg), GFP_ATOMIC);
+	if (!msg) {
+		mlog(0, "failed to allocate a o2net_msg!\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	o2net_init_msg(msg, caller_bytes, msg_type, key);
+
+	vec[0].iov_len = sizeof(struct o2net_msg);
+	vec[0].iov_base = msg;
+	memcpy(&vec[1], caller_vec, caller_veclen * sizeof(struct kvec));
+
+	ret = o2net_prep_nsw(nn, &nsw);
+	if (ret)
+		goto out;
+
+	msg->msg_num = cpu_to_be32(nsw.ns_id);
+
+	/* finally, convert the message header to network byte-order
+	 * and send */
+	ret = o2net_send_tcp_msg(sc->sc_sock, vec, veclen,
+				 sizeof(struct o2net_msg) + caller_bytes);
+	msglog(msg, "sending returned %d\n", ret);
+	if (ret < 0) {
+		mlog(0, "error returned from o2net_send_tcp_msg=%d\n", ret);
+		goto out;
+	}
+
+	/* wait on other node's handler */
+	wait_event(nsw.ns_wq, o2net_nsw_completed(nn, &nsw));
+
+	/* Note that we avoid overwriting the callers status return
+	 * variable if a system error was reported on the other
+	 * side. Callers beware. */
+	ret = o2net_sys_err_to_errno(nsw.ns_sys_status);
+	if (status && !ret)
+		*status = nsw.ns_status;
+
+	mlog(0, "woken, returning system status %d, user status %d\n",
+	     ret, nsw.ns_status);
+out:
+	if (sc)
+		sc_put(sc);
+	if (vec)
+		kfree(vec);
+	if (msg)
+		kfree(msg);
+	o2net_complete_nsw(nn, &nsw, 0, 0, 0);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(o2net_send_message_vec);
+
+int o2net_send_message(u32 msg_type, u32 key, void *data, u32 len,
+		       u8 target_node, int *status)
+{
+	struct kvec vec = {
+		.iov_base = data,
+		.iov_len = len,
+	};
+	return o2net_send_message_vec(msg_type, key, &vec, 1,
+				      target_node, status);
+}
+EXPORT_SYMBOL_GPL(o2net_send_message);
+
+static int o2net_send_status_magic(struct socket *sock, struct o2net_msg *hdr,
+				   enum o2net_system_error syserr, int err)
+{
+	struct kvec vec = {
+		.iov_base = hdr,
+		.iov_len = sizeof(struct o2net_msg),
+	};
+
+	BUG_ON(syserr >= O2NET_ERR_MAX);
+
+	/* leave other fields intact from the incoming message, msg_num
+	 * in particular */
+	hdr->sys_status = cpu_to_be32(syserr);
+	hdr->status = cpu_to_be32(err);
+	hdr->magic = cpu_to_be16(O2NET_MSG_STATUS_MAGIC);  // twiddle the magic
+	hdr->data_len = 0;
+
+	msglog(hdr, "about to send status magic %d\n", err);
+	/* hdr has been in host byteorder this whole time */
+	return o2net_send_tcp_msg(sock, &vec, 1, sizeof(struct o2net_msg));
+}
+
+/* this returns -errno if the header was unknown or too large, etc.
+ * after this is called the buffer us reused for the next message */
+static int o2net_process_message(struct o2net_sock_container *sc,
+				 struct o2net_msg *hdr)
+{
+	struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
+	int ret = 0, handler_status;
+	enum  o2net_system_error syserr;
+	struct o2net_msg_handler *nmh = NULL;
+
+	msglog(hdr, "processing message\n");
+
+	o2net_sc_postpone_idle(sc);
+
+	switch(be16_to_cpu(hdr->magic)) {
+		case O2NET_MSG_STATUS_MAGIC:
+			/* special type for returning message status */
+			o2net_complete_nsw(nn, NULL,
+					   be32_to_cpu(hdr->msg_num),
+					   be32_to_cpu(hdr->sys_status),
+					   be32_to_cpu(hdr->status));
+			goto out;
+		case O2NET_MSG_KEEP_REQ_MAGIC:
+			o2net_sendpage(sc, o2net_keep_resp,
+				       sizeof(*o2net_keep_resp));
+			goto out;
+		case O2NET_MSG_KEEP_RESP_MAGIC:
+			goto out;
+		case O2NET_MSG_MAGIC:
+			break;
+		default:
+			msglog(hdr, "bad magic\n");
+			ret = -EINVAL;
+			goto out;
+			break;
+	}
+
+	/* find a handler for it */
+	handler_status = 0;
+	nmh = o2net_handler_get(be16_to_cpu(hdr->msg_type),
+				be32_to_cpu(hdr->key));
+	if (!nmh) {
+		mlog(ML_TCP, "couldn't find handler for type %u key %08x\n",
+		     be16_to_cpu(hdr->msg_type), be32_to_cpu(hdr->key));
+		syserr = O2NET_ERR_NO_HNDLR;
+		goto out_respond;
+	}
+
+	syserr = O2NET_ERR_NONE;
+
+	if (be16_to_cpu(hdr->data_len) > nmh->nh_max_len)
+		syserr = O2NET_ERR_OVERFLOW;
+
+	if (syserr != O2NET_ERR_NONE)
+		goto out_respond;
+
+	do_gettimeofday(&sc->sc_tv_func_start);
+	sc->sc_msg_key = be32_to_cpu(hdr->key);
+	sc->sc_msg_type = be16_to_cpu(hdr->msg_type);
+	handler_status = (nmh->nh_func)(hdr, sizeof(struct o2net_msg) +
+					     be16_to_cpu(hdr->data_len),
+					nmh->nh_func_data);
+	do_gettimeofday(&sc->sc_tv_func_stop);
+
+out_respond:
+	/* this destroys the hdr, so don't use it after this */
+	ret = o2net_send_status_magic(sc->sc_sock, hdr, syserr,
+				      handler_status);
+	hdr = NULL;
+	mlog(0, "sending handler status %d, syserr %d returned %d\n",
+	     handler_status, syserr, ret);
+
+out:
+	if (nmh)
+		o2net_handler_put(nmh);
+	return ret;
+}
+
+static int o2net_check_handshake(struct o2net_sock_container *sc)
+{
+	struct o2net_handshake *hand = page_address(sc->sc_page);
+	struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
+
+	if (hand->protocol_version != cpu_to_be64(O2NET_PROTOCOL_VERSION)) {
+		mlog(ML_NOTICE, SC_NODEF_FMT " advertised net protocol "
+		     "version %llu but %llu is required, disconnecting\n",
+		     SC_NODEF_ARGS(sc),
+		     (unsigned long long)be64_to_cpu(hand->protocol_version),
+		     O2NET_PROTOCOL_VERSION);
+
+		/* don't bother reconnecting if its the wrong version. */
+		o2net_ensure_shutdown(nn, sc, -ENOTCONN);
+		return -1;
+	}
+
+	sc->sc_handshake_ok = 1;
+
+	spin_lock(&nn->nn_lock);
+	/* set valid and queue the idle timers only if it hasn't been
+	 * shut down already */
+	if (nn->nn_sc == sc) {
+		o2net_sc_postpone_idle(sc);
+		o2net_set_nn_state(nn, sc, 1, 0);
+	}
+	spin_unlock(&nn->nn_lock);
+
+	/* shift everything up as though it wasn't there */
+	sc->sc_page_off -= sizeof(struct o2net_handshake);
+	if (sc->sc_page_off)
+		memmove(hand, hand + 1, sc->sc_page_off);
+
+	return 0;
+}
+
+/* this demuxes the queued rx bytes into header or payload bits and calls
+ * handlers as each full message is read off the socket.  it returns -error,
+ * == 0 eof, or > 0 for progress made.*/
+static int o2net_advance_rx(struct o2net_sock_container *sc)
+{
+	struct o2net_msg *hdr;
+	int ret = 0;
+	void *data;
+	size_t datalen;
+
+	sclog(sc, "receiving\n");
+	do_gettimeofday(&sc->sc_tv_advance_start);
+
+	/* do we need more header? */
+	if (sc->sc_page_off < sizeof(struct o2net_msg)) {
+		data = page_address(sc->sc_page) + sc->sc_page_off;
+		datalen = sizeof(struct o2net_msg) - sc->sc_page_off;
+		ret = o2net_recv_tcp_msg(sc->sc_sock, data, datalen);
+		if (ret > 0) {
+			sc->sc_page_off += ret;
+
+			/* this working relies on the handshake being
+			 * smaller than the normal message header */
+			if (sc->sc_page_off >= sizeof(struct o2net_handshake)&&
+			    !sc->sc_handshake_ok && o2net_check_handshake(sc)) {
+				ret = -EPROTO;
+				goto out;
+			}
+
+			/* only swab incoming here.. we can
+			 * only get here once as we cross from
+			 * being under to over */
+			if (sc->sc_page_off == sizeof(struct o2net_msg)) {
+				hdr = page_address(sc->sc_page);
+				if (be16_to_cpu(hdr->data_len) >
+				    O2NET_MAX_PAYLOAD_BYTES)
+					ret = -EOVERFLOW;
+			}
+		}
+		if (ret <= 0)
+			goto out;
+	}
+
+	if (sc->sc_page_off < sizeof(struct o2net_msg)) {
+		/* oof, still don't have a header */
+		goto out;
+	}
+
+	/* this was swabbed above when we first read it */
+	hdr = page_address(sc->sc_page);
+
+	msglog(hdr, "at page_off %zu\n", sc->sc_page_off);
+
+	/* do we need more payload? */
+	if (sc->sc_page_off - sizeof(struct o2net_msg) < be16_to_cpu(hdr->data_len)) {
+		/* need more payload */
+		data = page_address(sc->sc_page) + sc->sc_page_off;
+		datalen = (sizeof(struct o2net_msg) + be16_to_cpu(hdr->data_len)) -
+			  sc->sc_page_off;
+		ret = o2net_recv_tcp_msg(sc->sc_sock, data, datalen);
+		if (ret > 0)
+			sc->sc_page_off += ret;
+		if (ret <= 0)
+			goto out;
+	}
+
+	if (sc->sc_page_off - sizeof(struct o2net_msg) == be16_to_cpu(hdr->data_len)) {
+		/* we can only get here once, the first time we read
+		 * the payload.. so set ret to progress if the handler
+		 * works out. after calling this the message is toast */
+		ret = o2net_process_message(sc, hdr);
+		if (ret == 0)
+			ret = 1;
+		sc->sc_page_off = 0;
+	}
+
+out:
+	sclog(sc, "ret = %d\n", ret);
+	do_gettimeofday(&sc->sc_tv_advance_stop);
+	return ret;
+}
+
+/* this work func is triggerd by data ready.  it reads until it can read no
+ * more.  it interprets 0, eof, as fatal.  if data_ready hits while we're doing
+ * our work the work struct will be marked and we'll be called again. */
+static void o2net_rx_until_empty(void *arg)
+{
+	struct o2net_sock_container *sc = arg;
+	int ret;
+
+	do {
+		ret = o2net_advance_rx(sc);
+	} while (ret > 0);
+
+	if (ret <= 0 && ret != -EAGAIN) {
+		struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
+		sclog(sc, "saw error %d, closing\n", ret);
+		/* not permanent so read failed handshake can retry */
+		o2net_ensure_shutdown(nn, sc, 0);
+	}
+
+	sc_put(sc);
+}
+
+static int o2net_set_nodelay(struct socket *sock)
+{
+	int ret, val = 1;
+	mm_segment_t oldfs;
+
+	oldfs = get_fs();
+	set_fs(KERNEL_DS);
+
+	/*
+	 * Dear unsuspecting programmer,
+	 *
+	 * Don't use sock_setsockopt() for SOL_TCP.  It doesn't check its level
+	 * argument and assumes SOL_SOCKET so, say, your TCP_NODELAY will
+	 * silently turn into SO_DEBUG.
+	 *
+	 * Yours,
+	 * Keeper of hilariously fragile interfaces.
+	 */
+	ret = sock->ops->setsockopt(sock, SOL_TCP, TCP_NODELAY,
+				    (char __user *)&val, sizeof(val));
+
+	set_fs(oldfs);
+	return ret;
+}
+
+/* ------------------------------------------------------------ */
+
+/* called when a connect completes and after a sock is accepted.  the
+ * rx path will see the response and mark the sc valid */
+static void o2net_sc_connect_completed(void *arg)
+{
+	struct o2net_sock_container *sc = arg;
+
+	mlog(ML_MSG, "sc sending handshake with ver %llu id %llx\n",
+              (unsigned long long)O2NET_PROTOCOL_VERSION,
+	      (unsigned long long)be64_to_cpu(o2net_hand->connector_id));
+
+	o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand));
+	sc_put(sc);
+}
+
+/* this is called as a work_struct func. */
+static void o2net_sc_send_keep_req(void *arg)
+{
+	struct o2net_sock_container *sc = arg;
+
+	o2net_sendpage(sc, o2net_keep_req, sizeof(*o2net_keep_req));
+	sc_put(sc);
+}
+
+/* socket shutdown does a del_timer_sync against this as it tears down.
+ * we can't start this timer until we've got to the point in sc buildup
+ * where shutdown is going to be involved */
+static void o2net_idle_timer(unsigned long data)
+{
+	struct o2net_sock_container *sc = (struct o2net_sock_container *)data;
+	struct timeval now;
+
+	do_gettimeofday(&now);
+
+	mlog(ML_NOTICE, "connection to " SC_NODEF_FMT " has been idle for 10 "
+	     "seconds, shutting it down.\n", SC_NODEF_ARGS(sc));
+	mlog(ML_NOTICE, "here are some times that might help debug the "
+	     "situation: (tmr %ld.%ld now %ld.%ld dr %ld.%ld adv "
+	     "%ld.%ld:%ld.%ld func (%08x:%u) %ld.%ld:%ld.%ld)\n",
+	     sc->sc_tv_timer.tv_sec, sc->sc_tv_timer.tv_usec, 
+	     now.tv_sec, now.tv_usec,
+	     sc->sc_tv_data_ready.tv_sec, sc->sc_tv_data_ready.tv_usec, 
+	     sc->sc_tv_advance_start.tv_sec, sc->sc_tv_advance_start.tv_usec, 
+	     sc->sc_tv_advance_stop.tv_sec, sc->sc_tv_advance_stop.tv_usec, 
+	     sc->sc_msg_key, sc->sc_msg_type,
+	     sc->sc_tv_func_start.tv_sec, sc->sc_tv_func_start.tv_usec,
+	     sc->sc_tv_func_stop.tv_sec, sc->sc_tv_func_stop.tv_usec);
+
+	o2net_sc_queue_work(sc, &sc->sc_shutdown_work);
+}
+
+static void o2net_sc_postpone_idle(struct o2net_sock_container *sc)
+{
+	o2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work);
+	o2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work,
+				    O2NET_KEEPALIVE_DELAY_SECS * HZ);
+	do_gettimeofday(&sc->sc_tv_timer);
+	mod_timer(&sc->sc_idle_timeout,
+		  jiffies + (O2NET_IDLE_TIMEOUT_SECS * HZ));
+}
+
+/* this work func is kicked whenever a path sets the nn state which doesn't
+ * have valid set.  This includes seeing hb come up, losing a connection,
+ * having a connect attempt fail, etc. This centralizes the logic which decides
+ * if a connect attempt should be made or if we should give up and all future
+ * transmit attempts should fail */
+static void o2net_start_connect(void *arg)
+{
+	struct o2net_node *nn = arg;
+	struct o2net_sock_container *sc = NULL;
+	struct o2nm_node *node = NULL;
+	struct socket *sock = NULL;
+	struct sockaddr_in myaddr = {0, }, remoteaddr = {0, };
+	int ret = 0;
+
+	/* if we're greater we initiate tx, otherwise we accept */
+	if (o2nm_this_node() <= o2net_num_from_nn(nn))
+		goto out;
+
+	/* watch for racing with tearing a node down */
+	node = o2nm_get_node_by_num(o2net_num_from_nn(nn));
+	if (node == NULL) {
+		ret = 0;
+		goto out;
+	}
+
+	spin_lock(&nn->nn_lock);
+	/* see if we already have one pending or have given up */
+	if (nn->nn_sc || nn->nn_persistent_error)
+		arg = NULL;
+	spin_unlock(&nn->nn_lock);
+	if (arg == NULL) /* *shrug*, needed some indicator */
+		goto out;
+
+	nn->nn_last_connect_attempt = jiffies;
+
+	sc = sc_alloc(node);
+	if (sc == NULL) {
+		mlog(0, "couldn't allocate sc\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ret = sock_create(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
+	if (ret < 0) {
+		mlog(0, "can't create socket: %d\n", ret);
+		goto out;
+	}
+	sc->sc_sock = sock; /* freed by sc_kref_release */
+
+	sock->sk->sk_allocation = GFP_ATOMIC;
+
+	myaddr.sin_family = AF_INET;
+	myaddr.sin_port = (__force u16)htons(0); /* any port */
+
+	ret = sock->ops->bind(sock, (struct sockaddr *)&myaddr,
+			      sizeof(myaddr));
+	if (ret) {
+		mlog(0, "bind failed: %d\n", ret);
+		goto out;
+	}
+
+	ret = o2net_set_nodelay(sc->sc_sock);
+	if (ret) {
+		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
+		goto out;
+	}
+
+	o2net_register_callbacks(sc->sc_sock->sk, sc);
+
+	spin_lock(&nn->nn_lock);
+	/* handshake completion will set nn->nn_sc_valid */
+	o2net_set_nn_state(nn, sc, 0, 0);
+	spin_unlock(&nn->nn_lock);
+
+	remoteaddr.sin_family = AF_INET;
+	remoteaddr.sin_addr.s_addr = (__force u32)node->nd_ipv4_address;
+	remoteaddr.sin_port = (__force u16)node->nd_ipv4_port;
+
+	ret = sc->sc_sock->ops->connect(sc->sc_sock,
+					(struct sockaddr *)&remoteaddr,
+					sizeof(remoteaddr),
+					O_NONBLOCK);
+	if (ret == -EINPROGRESS)
+		ret = 0;
+
+out:
+	if (ret) {
+		mlog(ML_NOTICE, "connect attempt to " SC_NODEF_FMT " failed "
+		     "with errno %d\n", SC_NODEF_ARGS(sc), ret);
+		/* 0 err so that another will be queued and attempted
+		 * from set_nn_state */
+		if (sc)
+			o2net_ensure_shutdown(nn, sc, 0);
+	}
+	if (sc)
+		sc_put(sc);
+	if (node)
+		o2nm_node_put(node);
+
+	return;
+}
+
+static void o2net_connect_expired(void *arg)
+{
+	struct o2net_node *nn = arg;
+
+	spin_lock(&nn->nn_lock);
+	if (!nn->nn_sc_valid) {
+		mlog(ML_ERROR, "no connection established with node %u after "
+		     "%u seconds, giving up and returning errors.\n",
+		     o2net_num_from_nn(nn), O2NET_IDLE_TIMEOUT_SECS);
+
+		o2net_set_nn_state(nn, NULL, 0, -ENOTCONN);
+	}
+	spin_unlock(&nn->nn_lock);
+}
+
+static void o2net_still_up(void *arg)
+{
+	struct o2net_node *nn = arg;
+
+	o2quo_hb_still_up(o2net_num_from_nn(nn));
+}
+
+/* ------------------------------------------------------------ */
+
+void o2net_disconnect_node(struct o2nm_node *node)
+{
+	struct o2net_node *nn = o2net_nn_from_num(node->nd_num);
+
+	/* don't reconnect until it's heartbeating again */
+	spin_lock(&nn->nn_lock);
+	o2net_set_nn_state(nn, NULL, 0, -ENOTCONN);
+	spin_unlock(&nn->nn_lock);
+
+	if (o2net_wq) {
+		cancel_delayed_work(&nn->nn_connect_expired);
+		cancel_delayed_work(&nn->nn_connect_work);
+		cancel_delayed_work(&nn->nn_still_up);
+		flush_workqueue(o2net_wq);
+	}
+}
+
+static void o2net_hb_node_down_cb(struct o2nm_node *node, int node_num,
+				  void *data)
+{
+	o2quo_hb_down(node_num);
+
+	if (node_num != o2nm_this_node())
+		o2net_disconnect_node(node);
+}
+
+static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num,
+				void *data)
+{
+	struct o2net_node *nn = o2net_nn_from_num(node_num);
+
+	o2quo_hb_up(node_num);
+
+	/* ensure an immediate connect attempt */
+	nn->nn_last_connect_attempt = jiffies -
+		(msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS) + 1);
+
+	if (node_num != o2nm_this_node()) {
+		/* heartbeat doesn't work unless a local node number is
+		 * configured and doing so brings up the o2net_wq, so we can
+		 * use it.. */
+		queue_delayed_work(o2net_wq, &nn->nn_connect_expired,
+				   O2NET_IDLE_TIMEOUT_SECS * HZ);
+
+		/* believe it or not, accept and node hearbeating testing
+		 * can succeed for this node before we got here.. so
+		 * only use set_nn_state to clear the persistent error
+		 * if that hasn't already happened */
+		spin_lock(&nn->nn_lock);
+		if (nn->nn_persistent_error)
+			o2net_set_nn_state(nn, NULL, 0, 0);
+		spin_unlock(&nn->nn_lock);
+	}
+}
+
+void o2net_unregister_hb_callbacks(void)
+{
+	int ret;
+
+	ret = o2hb_unregister_callback(&o2net_hb_up);
+	if (ret < 0)
+		mlog(ML_ERROR, "Status return %d unregistering heartbeat up "
+		     "callback!\n", ret);
+
+	ret = o2hb_unregister_callback(&o2net_hb_down);
+	if (ret < 0)
+		mlog(ML_ERROR, "Status return %d unregistering heartbeat down "
+		     "callback!\n", ret);
+}
+
+int o2net_register_hb_callbacks(void)
+{
+	int ret;
+
+	o2hb_setup_callback(&o2net_hb_down, O2HB_NODE_DOWN_CB,
+			    o2net_hb_node_down_cb, NULL, O2NET_HB_PRI);
+	o2hb_setup_callback(&o2net_hb_up, O2HB_NODE_UP_CB,
+			    o2net_hb_node_up_cb, NULL, O2NET_HB_PRI);
+
+	ret = o2hb_register_callback(&o2net_hb_up);
+	if (ret == 0)
+		ret = o2hb_register_callback(&o2net_hb_down);
+
+	if (ret)
+		o2net_unregister_hb_callbacks();
+
+	return ret;
+}
+
+/* ------------------------------------------------------------ */
+
+static int o2net_accept_one(struct socket *sock)
+{
+	int ret, slen;
+	struct sockaddr_in sin;
+	struct socket *new_sock = NULL;
+	struct o2nm_node *node = NULL;
+	struct o2net_sock_container *sc = NULL;
+	struct o2net_node *nn;
+
+	BUG_ON(sock == NULL);
+	ret = sock_create_lite(sock->sk->sk_family, sock->sk->sk_type,
+			       sock->sk->sk_protocol, &new_sock);
+	if (ret)
+		goto out;
+
+	new_sock->type = sock->type;
+	new_sock->ops = sock->ops;
+	ret = sock->ops->accept(sock, new_sock, O_NONBLOCK);
+	if (ret < 0)
+		goto out;
+
+	new_sock->sk->sk_allocation = GFP_ATOMIC;
+
+	ret = o2net_set_nodelay(new_sock);
+	if (ret) {
+		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
+		goto out;
+	}
+
+	slen = sizeof(sin);
+	ret = new_sock->ops->getname(new_sock, (struct sockaddr *) &sin,
+				       &slen, 1);
+	if (ret < 0)
+		goto out;
+
+	node = o2nm_get_node_by_ip((__force __be32)sin.sin_addr.s_addr);
+	if (node == NULL) {
+		mlog(ML_NOTICE, "attempt to connect from unknown node at "
+		     "%u.%u.%u.%u:%d\n", NIPQUAD(sin.sin_addr.s_addr),
+		     ntohs((__force __be16)sin.sin_port));
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (o2nm_this_node() > node->nd_num) {
+		mlog(ML_NOTICE, "unexpected connect attempted from a lower "
+		     "numbered node '%s' at " "%u.%u.%u.%u:%d with num %u\n",
+		     node->nd_name, NIPQUAD(sin.sin_addr.s_addr),
+		     ntohs((__force __be16)sin.sin_port), node->nd_num);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* this happens all the time when the other node sees our heartbeat
+	 * and tries to connect before we see their heartbeat */
+	if (!o2hb_check_node_heartbeating_from_callback(node->nd_num)) {
+		mlog(ML_CONN, "attempt to connect from node '%s' at "
+		     "%u.%u.%u.%u:%d but it isn't heartbeating\n",
+		     node->nd_name, NIPQUAD(sin.sin_addr.s_addr),
+		     ntohs((__force __be16)sin.sin_port));
+		ret = -EINVAL;
+		goto out;
+	}
+
+	nn = o2net_nn_from_num(node->nd_num);
+
+	spin_lock(&nn->nn_lock);
+	if (nn->nn_sc)
+		ret = -EBUSY;
+	else
+		ret = 0;
+	spin_unlock(&nn->nn_lock);
+	if (ret) {
+		mlog(ML_NOTICE, "attempt to connect from node '%s' at "
+		     "%u.%u.%u.%u:%d but it already has an open connection\n",
+		     node->nd_name, NIPQUAD(sin.sin_addr.s_addr),
+		     ntohs((__force __be16)sin.sin_port));
+		goto out;
+	}
+
+	sc = sc_alloc(node);
+	if (sc == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	sc->sc_sock = new_sock;
+	new_sock = NULL;
+
+	spin_lock(&nn->nn_lock);
+	o2net_set_nn_state(nn, sc, 0, 0);
+	spin_unlock(&nn->nn_lock);
+
+	o2net_register_callbacks(sc->sc_sock->sk, sc);
+	o2net_sc_queue_work(sc, &sc->sc_rx_work);
+
+	o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand));
+
+out:
+	if (new_sock)
+		sock_release(new_sock);
+	if (node)
+		o2nm_node_put(node);
+	if (sc)
+		sc_put(sc);
+	return ret;
+}
+
+static void o2net_accept_many(void *arg)
+{
+	struct socket *sock = arg;
+	while (o2net_accept_one(sock) == 0)
+		cond_resched();
+}
+
+static void o2net_listen_data_ready(struct sock *sk, int bytes)
+{
+	void (*ready)(struct sock *sk, int bytes);
+
+	read_lock(&sk->sk_callback_lock);
+	ready = sk->sk_user_data;
+	if (ready == NULL) { /* check for teardown race */
+		ready = sk->sk_data_ready;
+		goto out;
+	}
+
+	/* ->sk_data_ready is also called for a newly established child socket
+	 * before it has been accepted and the acceptor has set up their
+	 * data_ready.. we only want to queue listen work for our listening
+	 * socket */
+	if (sk->sk_state == TCP_LISTEN) {
+		mlog(ML_TCP, "bytes: %d\n", bytes);
+		queue_work(o2net_wq, &o2net_listen_work);
+	}
+
+out:
+	read_unlock(&sk->sk_callback_lock);
+	ready(sk, bytes);
+}
+
+static int o2net_open_listening_sock(__be16 port)
+{
+	struct socket *sock = NULL;
+	int ret;
+	struct sockaddr_in sin = {
+		.sin_family = PF_INET,
+		.sin_addr = { .s_addr = (__force u32)htonl(INADDR_ANY) },
+		.sin_port = (__force u16)port,
+	};
+
+	ret = sock_create(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
+	if (ret < 0) {
+		mlog(ML_ERROR, "unable to create socket, ret=%d\n", ret);
+		goto out;
+	}
+
+	sock->sk->sk_allocation = GFP_ATOMIC;
+
+	write_lock_bh(&sock->sk->sk_callback_lock);
+	sock->sk->sk_user_data = sock->sk->sk_data_ready;
+	sock->sk->sk_data_ready = o2net_listen_data_ready;
+	write_unlock_bh(&sock->sk->sk_callback_lock);
+
+	o2net_listen_sock = sock;
+	INIT_WORK(&o2net_listen_work, o2net_accept_many, sock);
+
+	sock->sk->sk_reuse = 1;
+	ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
+	if (ret < 0) {
+		mlog(ML_ERROR, "unable to bind socket to port %d, ret=%d\n",
+		     ntohs(port), ret);
+		goto out;
+	}
+
+	ret = sock->ops->listen(sock, 64);
+	if (ret < 0) {
+		mlog(ML_ERROR, "unable to listen on port %d, ret=%d\n",
+		     ntohs(port), ret);
+	}
+
+out:
+	if (ret) {
+		o2net_listen_sock = NULL;
+		if (sock)
+			sock_release(sock);
+	}
+	return ret;
+}
+
+/*
+ * called from node manager when we should bring up our network listening
+ * socket.  node manager handles all the serialization to only call this
+ * once and to match it with o2net_stop_listening().  note,
+ * o2nm_this_node() doesn't work yet as we're being called while it
+ * is being set up.
+ */
+int o2net_start_listening(struct o2nm_node *node)
+{
+	int ret = 0;
+
+	BUG_ON(o2net_wq != NULL);
+	BUG_ON(o2net_listen_sock != NULL);
+
+	mlog(ML_KTHREAD, "starting o2net thread...\n");
+	o2net_wq = create_singlethread_workqueue("o2net");
+	if (o2net_wq == NULL) {
+		mlog(ML_ERROR, "unable to launch o2net thread\n");
+		return -ENOMEM; /* ? */
+	}
+
+	ret = o2net_open_listening_sock(node->nd_ipv4_port);
+	if (ret) {
+		destroy_workqueue(o2net_wq);
+		o2net_wq = NULL;
+	} else
+		o2quo_conn_up(node->nd_num);
+
+	return ret;
+}
+
+/* again, o2nm_this_node() doesn't work here as we're involved in
+ * tearing it down */
+void o2net_stop_listening(struct o2nm_node *node)
+{
+	struct socket *sock = o2net_listen_sock;
+	size_t i;
+
+	BUG_ON(o2net_wq == NULL);
+	BUG_ON(o2net_listen_sock == NULL);
+
+	/* stop the listening socket from generating work */
+	write_lock_bh(&sock->sk->sk_callback_lock);
+	sock->sk->sk_data_ready = sock->sk->sk_user_data;
+	sock->sk->sk_user_data = NULL;
+	write_unlock_bh(&sock->sk->sk_callback_lock);
+
+	for (i = 0; i < ARRAY_SIZE(o2net_nodes); i++) {
+		struct o2nm_node *node = o2nm_get_node_by_num(i);
+		if (node) {
+			o2net_disconnect_node(node);
+			o2nm_node_put(node);
+		}
+	}
+
+	/* finish all work and tear down the work queue */
+	mlog(ML_KTHREAD, "waiting for o2net thread to exit....\n");
+	destroy_workqueue(o2net_wq);
+	o2net_wq = NULL;
+
+	sock_release(o2net_listen_sock);
+	o2net_listen_sock = NULL;
+
+	o2quo_conn_err(node->nd_num);
+}
+
+/* ------------------------------------------------------------ */
+
+int o2net_init(void)
+{
+	unsigned long i;
+
+	o2quo_init();
+
+	o2net_hand = kcalloc(1, sizeof(struct o2net_handshake), GFP_KERNEL);
+	o2net_keep_req = kcalloc(1, sizeof(struct o2net_msg), GFP_KERNEL);
+	o2net_keep_resp = kcalloc(1, sizeof(struct o2net_msg), GFP_KERNEL);
+	if (!o2net_hand || !o2net_keep_req || !o2net_keep_resp) {
+		kfree(o2net_hand);
+		kfree(o2net_keep_req);
+		kfree(o2net_keep_resp);
+		return -ENOMEM;
+	}
+
+	o2net_hand->protocol_version = cpu_to_be64(O2NET_PROTOCOL_VERSION);
+	o2net_hand->connector_id = cpu_to_be64(1);
+
+	o2net_keep_req->magic = cpu_to_be16(O2NET_MSG_KEEP_REQ_MAGIC);
+	o2net_keep_resp->magic = cpu_to_be16(O2NET_MSG_KEEP_RESP_MAGIC);
+
+	for (i = 0; i < ARRAY_SIZE(o2net_nodes); i++) {
+		struct o2net_node *nn = o2net_nn_from_num(i);
+
+		spin_lock_init(&nn->nn_lock);
+		INIT_WORK(&nn->nn_connect_work, o2net_start_connect, nn);
+		INIT_WORK(&nn->nn_connect_expired, o2net_connect_expired, nn);
+		INIT_WORK(&nn->nn_still_up, o2net_still_up, nn);
+		/* until we see hb from a node we'll return einval */
+		nn->nn_persistent_error = -ENOTCONN;
+		init_waitqueue_head(&nn->nn_sc_wq);
+		idr_init(&nn->nn_status_idr);
+		INIT_LIST_HEAD(&nn->nn_status_list);
+	}
+
+	return 0;
+}
+
+void o2net_exit(void)
+{
+	o2quo_exit();
+	kfree(o2net_hand);
+	kfree(o2net_keep_req);
+	kfree(o2net_keep_resp);
+}
diff --git a/fs/ocfs2/cluster/tcp.h b/fs/ocfs2/cluster/tcp.h
new file mode 100644
index 00000000000000..a6f4585501c82a
--- /dev/null
+++ b/fs/ocfs2/cluster/tcp.h
@@ -0,0 +1,113 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * tcp.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ */
+
+#ifndef O2CLUSTER_TCP_H
+#define O2CLUSTER_TCP_H
+
+#include <linux/socket.h>
+#ifdef __KERNEL__
+#include <net/sock.h>
+#include <linux/tcp.h>
+#else
+#include <sys/socket.h>
+#endif
+#include <linux/inet.h>
+#include <linux/in.h>
+
+struct o2net_msg
+{
+	__be16 magic;
+	__be16 data_len;
+	__be16 msg_type;
+	__be16 pad1;
+	__be32 sys_status;
+	__be32 status;
+	__be32 key;
+	__be32 msg_num;
+	__u8  buf[0];
+};
+
+typedef int (o2net_msg_handler_func)(struct o2net_msg *msg, u32 len, void *data);
+
+#define O2NET_MAX_PAYLOAD_BYTES  (4096 - sizeof(struct o2net_msg))
+
+/* TODO: figure this out.... */
+static inline int o2net_link_down(int err, struct socket *sock)
+{
+	if (sock) {
+		if (sock->sk->sk_state != TCP_ESTABLISHED &&
+	    	    sock->sk->sk_state != TCP_CLOSE_WAIT)
+			return 1;
+	}
+
+	if (err >= 0)
+		return 0;
+	switch (err) {
+		/* ????????????????????????? */
+		case -ERESTARTSYS:
+		case -EBADF:
+		/* When the server has died, an ICMP port unreachable
+		 * message prompts ECONNREFUSED. */
+		case -ECONNREFUSED:
+		case -ENOTCONN:
+		case -ECONNRESET:
+		case -EPIPE:
+			return 1;
+	}
+	return 0;
+}
+
+enum {
+	O2NET_DRIVER_UNINITED,
+	O2NET_DRIVER_READY,
+};
+
+int o2net_init_tcp_sock(struct inode *inode);
+int o2net_send_message(u32 msg_type, u32 key, void *data, u32 len,
+		       u8 target_node, int *status);
+int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *vec,
+			   size_t veclen, u8 target_node, int *status);
+int o2net_broadcast_message(u32 msg_type, u32 key, void *data, u32 len,
+			    struct inode *group);
+
+int o2net_register_handler(u32 msg_type, u32 key, u32 max_len,
+			   o2net_msg_handler_func *func, void *data,
+			   struct list_head *unreg_list);
+void o2net_unregister_handler_list(struct list_head *list);
+
+struct o2nm_node;
+int o2net_register_hb_callbacks(void);
+void o2net_unregister_hb_callbacks(void);
+int o2net_start_listening(struct o2nm_node *node);
+void o2net_stop_listening(struct o2nm_node *node);
+void o2net_disconnect_node(struct o2nm_node *node);
+
+int o2net_init(void);
+void o2net_exit(void);
+int o2net_proc_init(struct proc_dir_entry *parent);
+void o2net_proc_exit(struct proc_dir_entry *parent);
+
+#endif /* O2CLUSTER_TCP_H */
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
new file mode 100644
index 00000000000000..ff9e2e2104c245
--- /dev/null
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -0,0 +1,174 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef O2CLUSTER_TCP_INTERNAL_H
+#define O2CLUSTER_TCP_INTERNAL_H
+
+#define O2NET_MSG_MAGIC           ((u16)0xfa55)
+#define O2NET_MSG_STATUS_MAGIC    ((u16)0xfa56)
+#define O2NET_MSG_KEEP_REQ_MAGIC  ((u16)0xfa57)
+#define O2NET_MSG_KEEP_RESP_MAGIC ((u16)0xfa58)
+
+/* same as hb delay, we're waiting for another node to recognize our hb */
+#define O2NET_RECONNECT_DELAY_MS	O2HB_REGION_TIMEOUT_MS
+
+/* we're delaying our quorum decision so that heartbeat will have timed
+ * out truly dead nodes by the time we come around to making decisions
+ * on their number */
+#define O2NET_QUORUM_DELAY_MS	((o2hb_dead_threshold + 2) * O2HB_REGION_TIMEOUT_MS)
+
+#define O2NET_KEEPALIVE_DELAY_SECS	5
+#define O2NET_IDLE_TIMEOUT_SECS		10
+
+/* 
+ * This version number represents quite a lot, unfortunately.  It not
+ * only represents the raw network message protocol on the wire but also
+ * locking semantics of the file system using the protocol.  It should 
+ * be somewhere else, I'm sure, but right now it isn't.
+ *
+ * New in version 2:
+ * 	- full 64 bit i_size in the metadata lock lvbs
+ * 	- introduction of "rw" lock and pushing meta/data locking down
+ */
+#define O2NET_PROTOCOL_VERSION 2ULL
+struct o2net_handshake {
+	__be64	protocol_version;
+	__be64	connector_id;
+};
+
+struct o2net_node {
+	/* this is never called from int/bh */
+	spinlock_t			nn_lock;
+
+	/* set the moment an sc is allocated and a connect is started */
+	struct o2net_sock_container	*nn_sc;
+	/* _valid is only set after the handshake passes and tx can happen */
+	unsigned			nn_sc_valid:1;
+	/* if this is set tx just returns it */
+	int				nn_persistent_error;
+
+	/* threads waiting for an sc to arrive wait on the wq for generation
+	 * to increase.  it is increased when a connecting socket succeeds
+	 * or fails or when an accepted socket is attached. */
+	wait_queue_head_t		nn_sc_wq;
+
+	struct idr			nn_status_idr;
+	struct list_head		nn_status_list;
+
+	/* connects are attempted from when heartbeat comes up until either hb
+	 * goes down, the node is unconfigured, no connect attempts succeed
+	 * before O2NET_CONN_IDLE_DELAY, or a connect succeeds.  connect_work
+	 * is queued from set_nn_state both from hb up and from itself if a
+	 * connect attempt fails and so can be self-arming.  shutdown is
+	 * careful to first mark the nn such that no connects will be attempted
+	 * before canceling delayed connect work and flushing the queue. */
+	struct work_struct		nn_connect_work;
+	unsigned long			nn_last_connect_attempt;
+
+	/* this is queued as nodes come up and is canceled when a connection is
+	 * established.  this expiring gives up on the node and errors out
+	 * transmits */
+	struct work_struct		nn_connect_expired;
+
+	/* after we give up on a socket we wait a while before deciding
+	 * that it is still heartbeating and that we should do some
+	 * quorum work */
+	struct work_struct		nn_still_up;
+};
+
+struct o2net_sock_container {
+	struct kref		sc_kref;
+	/* the next two are vaild for the life time of the sc */
+	struct socket		*sc_sock;
+	struct o2nm_node	*sc_node;
+
+	/* all of these sc work structs hold refs on the sc while they are
+	 * queued.  they should not be able to ref a freed sc.  the teardown
+	 * race is with o2net_wq destruction in o2net_stop_listening() */
+
+	/* rx and connect work are generated from socket callbacks.  sc
+	 * shutdown removes the callbacks and then flushes the work queue */
+	struct work_struct	sc_rx_work;
+	struct work_struct	sc_connect_work;
+	/* shutdown work is triggered in two ways.  the simple way is
+	 * for a code path calls ensure_shutdown which gets a lock, removes
+	 * the sc from the nn, and queues the work.  in this case the
+	 * work is single-shot.  the work is also queued from a sock
+	 * callback, though, and in this case the work will find the sc
+	 * still on the nn and will call ensure_shutdown itself.. this
+	 * ends up triggering the shutdown work again, though nothing
+	 * will be done in that second iteration.  so work queue teardown
+	 * has to be careful to remove the sc from the nn before waiting
+	 * on the work queue so that the shutdown work doesn't remove the
+	 * sc and rearm itself.
+	 */
+	struct work_struct	sc_shutdown_work;
+
+	struct timer_list	sc_idle_timeout;
+	struct work_struct	sc_keepalive_work;
+
+	unsigned		sc_handshake_ok:1;
+
+	struct page 		*sc_page;
+	size_t			sc_page_off;
+
+	/* original handlers for the sockets */
+	void			(*sc_state_change)(struct sock *sk);
+	void			(*sc_data_ready)(struct sock *sk, int bytes);
+
+	struct timeval 		sc_tv_timer;
+	struct timeval 		sc_tv_data_ready;
+	struct timeval 		sc_tv_advance_start;
+	struct timeval 		sc_tv_advance_stop;
+	struct timeval 		sc_tv_func_start;
+	struct timeval 		sc_tv_func_stop;
+	u32			sc_msg_key;
+	u16			sc_msg_type;
+};
+
+struct o2net_msg_handler {
+	struct rb_node		nh_node;
+	u32			nh_max_len;
+	u32			nh_msg_type;
+	u32			nh_key;
+	o2net_msg_handler_func	*nh_func;
+	o2net_msg_handler_func	*nh_func_data;
+	struct kref		nh_kref;
+	struct list_head	nh_unregister_item;
+};
+
+enum o2net_system_error {
+	O2NET_ERR_NONE = 0,
+	O2NET_ERR_NO_HNDLR,
+	O2NET_ERR_OVERFLOW,
+	O2NET_ERR_DIED,
+	O2NET_ERR_MAX
+};
+
+struct o2net_status_wait {
+	enum o2net_system_error	ns_sys_status;
+	s32			ns_status;
+	int			ns_id;
+	wait_queue_head_t	ns_wq;
+	struct list_head	ns_node_item;
+};
+
+#endif /* O2CLUSTER_TCP_INTERNAL_H */
diff --git a/fs/ocfs2/cluster/ver.c b/fs/ocfs2/cluster/ver.c
new file mode 100644
index 00000000000000..7286c48bb30d45
--- /dev/null
+++ b/fs/ocfs2/cluster/ver.c
@@ -0,0 +1,42 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ver.c
+ *
+ * version string
+ *
+ * Copyright (C) 2002, 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+
+#include "ver.h"
+
+#define CLUSTER_BUILD_VERSION "1.3.3"
+
+#define VERSION_STR "OCFS2 Node Manager " CLUSTER_BUILD_VERSION
+
+void cluster_print_version(void)
+{
+	printk(KERN_INFO "%s\n", VERSION_STR);
+}
+
+MODULE_DESCRIPTION(VERSION_STR);
+
+MODULE_VERSION(CLUSTER_BUILD_VERSION);
diff --git a/fs/ocfs2/cluster/ver.h b/fs/ocfs2/cluster/ver.h
new file mode 100644
index 00000000000000..32554c3382c225
--- /dev/null
+++ b/fs/ocfs2/cluster/ver.h
@@ -0,0 +1,31 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ver.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef O2CLUSTER_VER_H
+#define O2CLUSTER_VER_H
+
+void cluster_print_version(void);
+
+#endif /* O2CLUSTER_VER_H */