linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH V5r1 0/6] staging: ramster: multi-machine memory capacity management
@ 2012-02-15 15:54 Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 1/6] staging: ramster: cluster/messaging foundation Dan Magenheimer
                   ` (5 more replies)
  0 siblings, 6 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-02-15 15:54 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, dan.magenheimer

[Note: identical to V5 but rebased to staging-next and patchset threaded,
 due to gregkh request... sorry for the noise and thanks for your patience!]

HIGH LEVEL OVERVIEW

RAMster implements peer-to-peer transcendent memory, allowing a "cluster" of
kernels to dynamically pool their RAM so that a RAM-hungry workload on one
machine can temporarily and transparently utilize RAM on another machine which
is presumably idle or running a non-RAM-hungry workload.  Other than the
already-merged cleancache patchset and the not-yet-merged frontswap patchset,
no core kernel changes are currently required.

(Note that, unlike previous public descriptions of RAMster, this implementation
does NOT require synchronous "gets" or core networking changes. As of V5,
it also co-exists with ocfs2.)

RAMster combines a clustering and messaging foundation based on the ocfs2
cluster layer with the in-kernel compression implementation of zcache, and
adds code to glue them together.  When a page is "put" to RAMster, it is
compressed and stored locally.  Periodically, a thread will "remotify" these
pages by sending them via messages to a remote machine.  When the page is
later needed as indicated by a page fault, a "get" is issued.  If the data
is local, it is uncompressed and the fault is resolved.  If the data is
remote, a message is sent to fetch the data and the faulting thread sleeps;
when the data arrives, the thread awakens, the data is decompressed and
the fault is resolved.

As of V5, clusters up to eight nodes are supported; each node can remotify
pages to one specified node, so clusters can be configured as clients to
a "memory server".  Some simple policy is in place that will need to be
refined over time.  Larger clusters and fault-resistant protocols can also
be added over time.

A git branch containing these patches built on linux-3.2 can be found at:
git://oss.oracle.com/git/djm/tmem.git #ramster-v5
Note that that tree also includes frontswap-v11 and "WasActive" patches

A HOW-TO is available at:
http://oss.oracle.com/projects/tmem/dist/files/RAMster/HOWTO-v5-120214

v4->v5: support multi-node clusters (up to 8 nodes)
v4->v5: add settable to choose remotify target node for memory server config
v4->v5: incorporate ocfs2 cluster layer directly to allow co-exist with ocfs2
v4->v5: incorporate xvmalloc directly to avoid upstream zsmalloc conflicts
v4->v5: support ramster-tools (instead of ocfs2-tools) in userland
v3->v4: rebase to 3.2 (including updates in zcache)
v3->v4: fix a couple of bad memory leaks to get cleancache fully working
v3->v4: fix preemption calls to remove dependency on CONFIG_PREEMPT_NONE
v3->v4: various cleanup
v2->v3: documentation and commit message changes required [gregkh]

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>


---

Diffstat:

 drivers/staging/Kconfig                            |    2 +
 drivers/staging/Makefile                           |    1 +
 drivers/staging/ramster/Kconfig                    |   13 +
 drivers/staging/ramster/Makefile                   |    1 +
 drivers/staging/ramster/TODO                       |   13 +
 drivers/staging/ramster/cluster/Makefile           |    3 +
 drivers/staging/ramster/cluster/heartbeat.c        |  464 +++
 drivers/staging/ramster/cluster/heartbeat.h        |   87 +
 drivers/staging/ramster/cluster/masklog.c          |  155 +
 drivers/staging/ramster/cluster/masklog.h          |  220 ++
 drivers/staging/ramster/cluster/nodemanager.c      |  992 ++++++
 drivers/staging/ramster/cluster/nodemanager.h      |   88 +
 .../staging/ramster/cluster/ramster_nodemanager.h  |   39 +
 drivers/staging/ramster/cluster/tcp.c              | 2256 +++++++++++++
 drivers/staging/ramster/cluster/tcp.h              |  159 +
 drivers/staging/ramster/cluster/tcp_internal.h     |  248 ++
 drivers/staging/ramster/r2net.c                    |  401 +++
 drivers/staging/ramster/ramster.h                  |  118 +
 drivers/staging/ramster/tmem.c                     |  851 +++++
 drivers/staging/ramster/tmem.h                     |  244 ++
 drivers/staging/ramster/xvmalloc.c                 |  510 +++
 drivers/staging/ramster/xvmalloc.h                 |   30 +
 drivers/staging/ramster/xvmalloc_int.h             |   95 +
 drivers/staging/ramster/zcache-main.c              | 3320 ++++++++++++++++++++
 drivers/staging/ramster/zcache.h                   |   22 +
 25 files changed, 10332 insertions(+), 0 deletions(-)

^ permalink raw reply	[flat|nested] 10+ messages in thread

* [PATCH 1/6] staging: ramster: cluster/messaging foundation
  2012-02-15 15:54 [PATCH V5r1 0/6] staging: ramster: multi-machine memory capacity management Dan Magenheimer
@ 2012-02-15 15:54 ` Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 2/6] staging: ramster: local compression + tmem Dan Magenheimer
                   ` (4 subsequent siblings)
  5 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-02-15 15:54 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, dan.magenheimer

RAMster implements peer-to-peer transcendent memory, allowing a "cluster"
of kernels to dynamically pool their RAM.

This patch provides the cluster and messaging foundation for RAMster,
implementing the basic cluster discovery, mapping, heartbeat / keepalive,
and messaging ("r2net") that RAMster requires for internode communication.
This code heavily leverages code from the ocfs2 cluster layer but
has been extended, interfaces to userland changed, and external functions
renamed so that RAMster and ocfs2 can co-exist in the kernel and userland.

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/ramster/cluster/Makefile           |    3 +
 drivers/staging/ramster/cluster/heartbeat.c        |  464 ++++
 drivers/staging/ramster/cluster/heartbeat.h        |   87 +
 drivers/staging/ramster/cluster/masklog.c          |  155 ++
 drivers/staging/ramster/cluster/masklog.h          |  220 ++
 drivers/staging/ramster/cluster/nodemanager.c      |  992 +++++++++
 drivers/staging/ramster/cluster/nodemanager.h      |   88 +
 .../staging/ramster/cluster/ramster_nodemanager.h  |   39 +
 drivers/staging/ramster/cluster/tcp.c              | 2256 ++++++++++++++++++++
 drivers/staging/ramster/cluster/tcp.h              |  159 ++
 drivers/staging/ramster/cluster/tcp_internal.h     |  248 +++
 11 files changed, 4711 insertions(+), 0 deletions(-)
 create mode 100644 drivers/staging/ramster/cluster/Makefile
 create mode 100644 drivers/staging/ramster/cluster/heartbeat.c
 create mode 100644 drivers/staging/ramster/cluster/heartbeat.h
 create mode 100644 drivers/staging/ramster/cluster/masklog.c
 create mode 100644 drivers/staging/ramster/cluster/masklog.h
 create mode 100644 drivers/staging/ramster/cluster/nodemanager.c
 create mode 100644 drivers/staging/ramster/cluster/nodemanager.h
 create mode 100644 drivers/staging/ramster/cluster/ramster_nodemanager.h
 create mode 100644 drivers/staging/ramster/cluster/tcp.c
 create mode 100644 drivers/staging/ramster/cluster/tcp.h
 create mode 100644 drivers/staging/ramster/cluster/tcp_internal.h

diff --git a/drivers/staging/ramster/cluster/Makefile b/drivers/staging/ramster/cluster/Makefile
new file mode 100644
index 0000000..9c69436
--- /dev/null
+++ b/drivers/staging/ramster/cluster/Makefile
@@ -0,0 +1,3 @@
+obj-$(CONFIG_RAMSTER) += ramster_nodemanager.o
+
+ramster_nodemanager-objs := heartbeat.o masklog.o nodemanager.o tcp.o
diff --git a/drivers/staging/ramster/cluster/heartbeat.c b/drivers/staging/ramster/cluster/heartbeat.c
new file mode 100644
index 0000000..0020949
--- /dev/null
+++ b/drivers/staging/ramster/cluster/heartbeat.c
@@ -0,0 +1,464 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004, 2005, 2012 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/configfs.h>
+
+#include "heartbeat.h"
+#include "tcp.h"
+#include "nodemanager.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(r2hb_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 DEFINE_SPINLOCK(r2hb_live_lock);
+static unsigned long r2hb_live_node_bitmap[BITS_TO_LONGS(R2NM_MAX_NODES)];
+
+static struct r2hb_callback {
+	struct list_head list;
+} r2hb_callbacks[R2HB_NUM_CB];
+
+enum r2hb_heartbeat_modes {
+	R2HB_HEARTBEAT_LOCAL		= 0,
+	R2HB_HEARTBEAT_GLOBAL,
+	R2HB_HEARTBEAT_NUM_MODES,
+};
+
+char *r2hb_heartbeat_mode_desc[R2HB_HEARTBEAT_NUM_MODES] = {
+		"local",	/* R2HB_HEARTBEAT_LOCAL */
+		"global",	/* R2HB_HEARTBEAT_GLOBAL */
+};
+
+unsigned int r2hb_dead_threshold = R2HB_DEFAULT_DEAD_THRESHOLD;
+unsigned int r2hb_heartbeat_mode = R2HB_HEARTBEAT_LOCAL;
+
+/* Only sets a new threshold if there are no active regions.
+ *
+ * No locking or otherwise interesting code is required for reading
+ * r2hb_dead_threshold as it can't change once regions are active and
+ * it's not interesting to anyone until then anyway. */
+static void r2hb_dead_threshold_set(unsigned int threshold)
+{
+	if (threshold > R2HB_MIN_DEAD_THRESHOLD) {
+		spin_lock(&r2hb_live_lock);
+		r2hb_dead_threshold = threshold;
+		spin_unlock(&r2hb_live_lock);
+	}
+}
+
+static int r2hb_global_hearbeat_mode_set(unsigned int hb_mode)
+{
+	int ret = -1;
+
+	if (hb_mode < R2HB_HEARTBEAT_NUM_MODES) {
+		spin_lock(&r2hb_live_lock);
+		r2hb_heartbeat_mode = hb_mode;
+		ret = 0;
+		spin_unlock(&r2hb_live_lock);
+	}
+
+	return ret;
+}
+
+void r2hb_exit(void)
+{
+}
+
+int r2hb_init(void)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(r2hb_callbacks); i++)
+		INIT_LIST_HEAD(&r2hb_callbacks[i].list);
+
+	memset(r2hb_live_node_bitmap, 0, sizeof(r2hb_live_node_bitmap));
+
+	return 0;
+}
+
+/* if we're already in a callback then we're already serialized by the sem */
+static void r2hb_fill_node_map_from_callback(unsigned long *map,
+					     unsigned bytes)
+{
+	BUG_ON(bytes < (BITS_TO_LONGS(R2NM_MAX_NODES) * sizeof(unsigned long)));
+
+	memcpy(map, &r2hb_live_node_bitmap, bytes);
+}
+
+/*
+ * get a map of all nodes that are heartbeating in any regions
+ */
+void r2hb_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(&r2hb_callback_sem);
+	spin_lock(&r2hb_live_lock);
+	r2hb_fill_node_map_from_callback(map, bytes);
+	spin_unlock(&r2hb_live_lock);
+	up_read(&r2hb_callback_sem);
+}
+EXPORT_SYMBOL_GPL(r2hb_fill_node_map);
+
+/*
+ * heartbeat configfs bits.  The heartbeat set is a default set under
+ * the cluster set in nodemanager.c.
+ */
+
+/* heartbeat set */
+
+struct r2hb_hb_group {
+	struct config_group hs_group;
+	/* some stuff? */
+};
+
+static struct r2hb_hb_group *to_r2hb_hb_group(struct config_group *group)
+{
+	return group ?
+		container_of(group, struct r2hb_hb_group, hs_group)
+		: NULL;
+}
+
+static struct config_item r2hb_config_item;
+
+static struct config_item *r2hb_hb_group_make_item(struct config_group *group,
+							  const char *name)
+{
+	int ret;
+
+	if (strlen(name) > R2HB_MAX_REGION_NAME_LEN) {
+		ret = -ENAMETOOLONG;
+		goto free;
+	}
+
+	config_item_put(&r2hb_config_item);
+
+	return &r2hb_config_item;
+free:
+	return ERR_PTR(ret);
+}
+
+static void r2hb_hb_group_drop_item(struct config_group *group,
+					   struct config_item *item)
+{
+	if (r2hb_global_heartbeat_active()) {
+		printk(KERN_NOTICE "ramster: Heartbeat %s "
+			"on region %s (%s)\n",
+			"stopped/aborted", config_item_name(item),
+			"no region");
+	}
+
+	config_item_put(item);
+}
+
+struct r2hb_hb_group_attribute {
+	struct configfs_attribute attr;
+	ssize_t (*show)(struct r2hb_hb_group *, char *);
+	ssize_t (*store)(struct r2hb_hb_group *, const char *, size_t);
+};
+
+static ssize_t r2hb_hb_group_show(struct config_item *item,
+					 struct configfs_attribute *attr,
+					 char *page)
+{
+	struct r2hb_hb_group *reg = to_r2hb_hb_group(to_config_group(item));
+	struct r2hb_hb_group_attribute *r2hb_hb_group_attr =
+		container_of(attr, struct r2hb_hb_group_attribute, attr);
+	ssize_t ret = 0;
+
+	if (r2hb_hb_group_attr->show)
+		ret = r2hb_hb_group_attr->show(reg, page);
+	return ret;
+}
+
+static ssize_t r2hb_hb_group_store(struct config_item *item,
+					  struct configfs_attribute *attr,
+					  const char *page, size_t count)
+{
+	struct r2hb_hb_group *reg = to_r2hb_hb_group(to_config_group(item));
+	struct r2hb_hb_group_attribute *r2hb_hb_group_attr =
+		container_of(attr, struct r2hb_hb_group_attribute, attr);
+	ssize_t ret = -EINVAL;
+
+	if (r2hb_hb_group_attr->store)
+		ret = r2hb_hb_group_attr->store(reg, page, count);
+	return ret;
+}
+
+static ssize_t r2hb_hb_group_threshold_show(struct r2hb_hb_group *group,
+						     char *page)
+{
+	return sprintf(page, "%u\n", r2hb_dead_threshold);
+}
+
+static ssize_t r2hb_hb_group_threshold_store(struct r2hb_hb_group *group,
+						    const char *page,
+						    size_t count)
+{
+	unsigned long tmp;
+	char *p = (char *)page;
+	int err;
+
+	err = kstrtoul(p, 10, &tmp);
+	if (err)
+		return err;
+
+	/* this will validate ranges for us. */
+	r2hb_dead_threshold_set((unsigned int) tmp);
+
+	return count;
+}
+
+static
+ssize_t r2hb_hb_group_mode_show(struct r2hb_hb_group *group,
+				       char *page)
+{
+	return sprintf(page, "%s\n",
+		       r2hb_heartbeat_mode_desc[r2hb_heartbeat_mode]);
+}
+
+static
+ssize_t r2hb_hb_group_mode_store(struct r2hb_hb_group *group,
+					const char *page, size_t count)
+{
+	unsigned int i;
+	int ret;
+	size_t len;
+
+	len = (page[count - 1] == '\n') ? count - 1 : count;
+	if (!len)
+		return -EINVAL;
+
+	for (i = 0; i < R2HB_HEARTBEAT_NUM_MODES; ++i) {
+		if (strnicmp(page, r2hb_heartbeat_mode_desc[i], len))
+			continue;
+
+		ret = r2hb_global_hearbeat_mode_set(i);
+		if (!ret)
+			printk(KERN_NOTICE "ramster: Heartbeat mode "
+						"set to %s\n",
+			       r2hb_heartbeat_mode_desc[i]);
+		return count;
+	}
+
+	return -EINVAL;
+
+}
+
+static struct r2hb_hb_group_attribute r2hb_hb_group_attr_threshold = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "dead_threshold",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2hb_hb_group_threshold_show,
+	.store	= r2hb_hb_group_threshold_store,
+};
+
+static struct r2hb_hb_group_attribute r2hb_hb_group_attr_mode = {
+	.attr   = { .ca_owner = THIS_MODULE,
+		.ca_name = "mode",
+		.ca_mode = S_IRUGO | S_IWUSR },
+	.show   = r2hb_hb_group_mode_show,
+	.store  = r2hb_hb_group_mode_store,
+};
+
+static struct configfs_attribute *r2hb_hb_group_attrs[] = {
+	&r2hb_hb_group_attr_threshold.attr,
+	&r2hb_hb_group_attr_mode.attr,
+	NULL,
+};
+
+static struct configfs_item_operations r2hb_hearbeat_group_item_ops = {
+	.show_attribute		= r2hb_hb_group_show,
+	.store_attribute	= r2hb_hb_group_store,
+};
+
+static struct configfs_group_operations r2hb_hb_group_group_ops = {
+	.make_item	= r2hb_hb_group_make_item,
+	.drop_item	= r2hb_hb_group_drop_item,
+};
+
+static struct config_item_type r2hb_hb_group_type = {
+	.ct_group_ops	= &r2hb_hb_group_group_ops,
+	.ct_item_ops	= &r2hb_hearbeat_group_item_ops,
+	.ct_attrs	= r2hb_hb_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 *r2hb_alloc_hb_set(void)
+{
+	struct r2hb_hb_group *hs = NULL;
+	struct config_group *ret = NULL;
+
+	hs = kzalloc(sizeof(struct r2hb_hb_group), GFP_KERNEL);
+	if (hs == NULL)
+		goto out;
+
+	config_group_init_type_name(&hs->hs_group, "heartbeat",
+				    &r2hb_hb_group_type);
+
+	ret = &hs->hs_group;
+out:
+	if (ret == NULL)
+		kfree(hs);
+	return ret;
+}
+
+void r2hb_free_hb_set(struct config_group *group)
+{
+	struct r2hb_hb_group *hs = to_r2hb_hb_group(group);
+	kfree(hs);
+}
+
+/* hb callback registration and issuing */
+
+static struct r2hb_callback *hbcall_from_type(enum r2hb_callback_type type)
+{
+	if (type == R2HB_NUM_CB)
+		return ERR_PTR(-EINVAL);
+
+	return &r2hb_callbacks[type];
+}
+
+void r2hb_setup_callback(struct r2hb_callback_func *hc,
+			 enum r2hb_callback_type type,
+			 r2hb_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 = R2HB_CB_MAGIC;
+}
+EXPORT_SYMBOL_GPL(r2hb_setup_callback);
+
+int r2hb_register_callback(const char *region_uuid,
+			   struct r2hb_callback_func *hc)
+{
+	struct r2hb_callback_func *tmp;
+	struct list_head *iter;
+	struct r2hb_callback *hbcall;
+	int ret;
+
+	BUG_ON(hc->hc_magic != R2HB_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(&r2hb_callback_sem);
+
+	list_for_each(iter, &hbcall->list) {
+		tmp = list_entry(iter, struct r2hb_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(&r2hb_callback_sem);
+	ret = 0;
+out:
+	mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n",
+	     ret, __builtin_return_address(0), hc);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(r2hb_register_callback);
+
+void r2hb_unregister_callback(const char *region_uuid,
+			      struct r2hb_callback_func *hc)
+{
+	BUG_ON(hc->hc_magic != R2HB_CB_MAGIC);
+
+	mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n",
+	     __builtin_return_address(0), hc);
+
+	/* XXX Can this happen _with_ a region reference? */
+	if (list_empty(&hc->hc_item))
+		return;
+
+	down_write(&r2hb_callback_sem);
+
+	list_del_init(&hc->hc_item);
+
+	up_write(&r2hb_callback_sem);
+}
+EXPORT_SYMBOL_GPL(r2hb_unregister_callback);
+
+int r2hb_check_node_heartbeating_from_callback(u8 node_num)
+{
+	unsigned long testing_map[BITS_TO_LONGS(R2NM_MAX_NODES)];
+
+	r2hb_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(r2hb_check_node_heartbeating_from_callback);
+
+void r2hb_stop_all_regions(void)
+{
+}
+EXPORT_SYMBOL_GPL(r2hb_stop_all_regions);
+
+/*
+ * 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.
+ */
+int r2hb_global_heartbeat_active(void)
+{
+	return (r2hb_heartbeat_mode == R2HB_HEARTBEAT_GLOBAL);
+}
+EXPORT_SYMBOL(r2hb_global_heartbeat_active);
+
+/* added for RAMster */
+void r2hb_manual_set_node_heartbeating(int node_num)
+{
+	if (node_num < R2NM_MAX_NODES)
+		set_bit(node_num, r2hb_live_node_bitmap);
+}
+EXPORT_SYMBOL(r2hb_manual_set_node_heartbeating);
diff --git a/drivers/staging/ramster/cluster/heartbeat.h b/drivers/staging/ramster/cluster/heartbeat.h
new file mode 100644
index 0000000..6cbc775
--- /dev/null
+++ b/drivers/staging/ramster/cluster/heartbeat.h
@@ -0,0 +1,87 @@
+/* -*- 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 R2CLUSTER_HEARTBEAT_H
+#define R2CLUSTER_HEARTBEAT_H
+
+#define R2HB_REGION_TIMEOUT_MS		2000
+
+#define R2HB_MAX_REGION_NAME_LEN	32
+
+/* number of changes to be seen as live */
+#define R2HB_LIVE_THRESHOLD	   2
+/* number of equal samples to be seen as dead */
+extern unsigned int r2hb_dead_threshold;
+#define R2HB_DEFAULT_DEAD_THRESHOLD	   31
+/* Otherwise MAX_WRITE_TIMEOUT will be zero... */
+#define R2HB_MIN_DEAD_THRESHOLD	  2
+#define R2HB_MAX_WRITE_TIMEOUT_MS \
+	(R2HB_REGION_TIMEOUT_MS * (r2hb_dead_threshold - 1))
+
+#define R2HB_CB_MAGIC		0x51d1e4ec
+
+/* callback stuff */
+enum r2hb_callback_type {
+	R2HB_NODE_DOWN_CB = 0,
+	R2HB_NODE_UP_CB,
+	R2HB_NUM_CB
+};
+
+struct r2nm_node;
+typedef void (r2hb_cb_func)(struct r2nm_node *, int, void *);
+
+struct r2hb_callback_func {
+	u32			hc_magic;
+	struct list_head	hc_item;
+	r2hb_cb_func		*hc_func;
+	void			*hc_data;
+	int			hc_priority;
+	enum r2hb_callback_type hc_type;
+};
+
+struct config_group *r2hb_alloc_hb_set(void);
+void r2hb_free_hb_set(struct config_group *group);
+
+void r2hb_setup_callback(struct r2hb_callback_func *hc,
+			 enum r2hb_callback_type type,
+			 r2hb_cb_func *func,
+			 void *data,
+			 int priority);
+int r2hb_register_callback(const char *region_uuid,
+			   struct r2hb_callback_func *hc);
+void r2hb_unregister_callback(const char *region_uuid,
+			      struct r2hb_callback_func *hc);
+void r2hb_fill_node_map(unsigned long *map,
+			unsigned bytes);
+void r2hb_exit(void);
+int r2hb_init(void);
+int r2hb_check_node_heartbeating_from_callback(u8 node_num);
+void r2hb_stop_all_regions(void);
+int r2hb_get_all_regions(char *region_uuids, u8 numregions);
+int r2hb_global_heartbeat_active(void);
+void r2hb_manual_set_node_heartbeating(int);
+
+#endif /* R2CLUSTER_HEARTBEAT_H */
diff --git a/drivers/staging/ramster/cluster/masklog.c b/drivers/staging/ramster/cluster/masklog.c
new file mode 100644
index 0000000..c2af3c7
--- /dev/null
+++ b/drivers/staging/ramster/cluster/masklog.c
@@ -0,0 +1,155 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004, 2005, 2012 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 <linux/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(0);
+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(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(DLM_GLUE),
+	define_mask(VOTE),
+	define_mask(CONN),
+	define_mask(QUORUM),
+	define_mask(BASTS),
+	define_mask(CLUSTER),
+	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 const 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   = {.ktype = &mlog_ktype},
+};
+
+int mlog_sys_init(struct kset *r2cb_kset)
+{
+	int i = 0;
+
+	while (mlog_attrs[i].attr.mode) {
+		mlog_attr_ptrs[i] = &mlog_attrs[i].attr;
+		i++;
+	}
+	mlog_attr_ptrs[i] = NULL;
+
+	kobject_set_name(&mlog_kset.kobj, "logmask");
+	mlog_kset.kobj.kset = r2cb_kset;
+	return kset_register(&mlog_kset);
+}
+
+void mlog_sys_shutdown(void)
+{
+	kset_unregister(&mlog_kset);
+}
diff --git a/drivers/staging/ramster/cluster/masklog.h b/drivers/staging/ramster/cluster/masklog.h
new file mode 100644
index 0000000..7609e66
--- /dev/null
+++ b/drivers/staging/ramster/cluster/masklog.h
@@ -0,0 +1,220 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005, 2012 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 R2CLUSTER_MASKLOG_H
+#define R2CLUSTER_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 managed from the files under
+ * /sys/fs/r2cb/logmask/.  Reading the files gives a straightforward
+ * indication of which bits are allowed (allow) or denied (off/deny).
+ *	ENTRY deny
+ *	EXIT deny
+ *	TCP off
+ *	MSG off
+ *	SOCKET off
+ *	ERROR allow
+ *	NOTICE allow
+ *
+ * Writing changes the state of a given bit and requires a strictly formatted
+ * single write() call:
+ *
+ *	write(fd, "allow", 5);
+ *
+ * Echoing allow/deny/off string into the logmask files can flip the bits
+ * on or off as expected; here is the bash script for example:
+ *
+ * log_mask="/sys/fs/r2cb/log_mask"
+ * for node in ENTRY EXIT TCP MSG SOCKET ERROR NOTICE; do
+ *	echo allow >"$log_mask"/"$node"
+ * done
+ *
+ * The debugfs.ramster tool can also flip the bits with the -l option:
+ *
+ * debugfs.ramster -l TCP allow
+ */
+
+/* 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 masklog.c! */
+#define ML_TCP		0x0000000000000001ULL /* net cluster/tcp.c */
+#define ML_MSG		0x0000000000000002ULL /* net network messages */
+#define ML_SOCKET	0x0000000000000004ULL /* net socket lifetime */
+#define ML_HEARTBEAT	0x0000000000000008ULL /* hb all heartbeat tracking */
+#define ML_HB_BIO	0x0000000000000010ULL /* hb io tracing */
+#define ML_DLMFS	0x0000000000000020ULL /* dlm user dlmfs */
+#define ML_DLM		0x0000000000000040ULL /* dlm general debugging */
+#define ML_DLM_DOMAIN	0x0000000000000080ULL /* dlm domain debugging */
+#define ML_DLM_THREAD	0x0000000000000100ULL /* dlm domain thread */
+#define ML_DLM_MASTER	0x0000000000000200ULL /* dlm master functions */
+#define ML_DLM_RECOVERY	0x0000000000000400ULL /* dlm master functions */
+#define ML_DLM_GLUE	0x0000000000000800ULL /* ramster dlm glue layer */
+#define ML_VOTE		0x0000000000001000ULL /* ramster node messaging  */
+#define ML_CONN		0x0000000000002000ULL /* net connection management */
+#define ML_QUORUM	0x0000000000004000ULL /* net connection quorum */
+#define ML_BASTS	0x0000000000008000ULL /* dlmglue asts and basts */
+#define ML_CLUSTER	0x0000000000010000ULL /* cluster stack */
+
+/* bits that are infrequently given and frequently matched in the high word */
+#define ML_ERROR	0x1000000000000000ULL /* sent to KERN_ERR */
+#define ML_NOTICE	0x2000000000000000ULL /* setn to KERN_NOTICE */
+#define ML_KTHREAD	0x4000000000000000ULL /* kernel thread activity */
+
+#define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE)
+#ifndef MLOG_MASK_PREFIX
+#define MLOG_MASK_PREFIX 0
+#endif
+
+/*
+ * When logging is disabled, force the bit test to 0 for anything other
+ * than errors and notices, allowing gcc to remove the code completely.
+ * When enabled, allow all masks.
+ */
+#if defined(CONFIG_RAMSTER_DEBUG_MASKLOG)
+#define ML_ALLOWED_BITS (~0)
+#else
+#define ML_ALLOWED_BITS (ML_ERROR|ML_NOTICE)
+#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 "(%s,%u,%lu):%s:%d " fmt, current->comm,		\
+	       task_pid_nr(current), __mlog_cpu_guess,			\
+	       __PRETTY_FUNCTION__, __LINE__ , ##args)
+
+#define mlog(mask, fmt, args...) do {					\
+	u64 __m = MLOG_MASK_PREFIX | (mask);				\
+	if ((__m & ML_ALLOWED_BITS) &&					\
+	    __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 && _st != -ENOSPC)		\
+		mlog(ML_ERROR, "status = %lld\n", (long long)_st);	\
+} 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)
+
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+int mlog_sys_init(struct kset *r2cb_subsys);
+void mlog_sys_shutdown(void);
+
+#endif /* R2CLUSTER_MASKLOG_H */
diff --git a/drivers/staging/ramster/cluster/nodemanager.c b/drivers/staging/ramster/cluster/nodemanager.c
new file mode 100644
index 0000000..de0e5c8
--- /dev/null
+++ b/drivers/staging/ramster/cluster/nodemanager.c
@@ -0,0 +1,992 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2004, 2005, 2012 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/slab.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/configfs.h>
+
+#include "tcp.h"
+#include "nodemanager.h"
+#include "heartbeat.h"
+#include "masklog.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 */
+struct r2nm_cluster *r2nm_single_cluster;
+
+char *r2nm_fence_method_desc[R2NM_FENCE_METHODS] = {
+		"reset",	/* R2NM_FENCE_RESET */
+		"panic",	/* R2NM_FENCE_PANIC */
+};
+
+struct r2nm_node *r2nm_get_node_by_num(u8 node_num)
+{
+	struct r2nm_node *node = NULL;
+
+	if (node_num >= R2NM_MAX_NODES || r2nm_single_cluster == NULL)
+		goto out;
+
+	read_lock(&r2nm_single_cluster->cl_nodes_lock);
+	node = r2nm_single_cluster->cl_nodes[node_num];
+	if (node)
+		config_item_get(&node->nd_item);
+	read_unlock(&r2nm_single_cluster->cl_nodes_lock);
+out:
+	return node;
+}
+EXPORT_SYMBOL_GPL(r2nm_get_node_by_num);
+
+int r2nm_configured_node_map(unsigned long *map, unsigned bytes)
+{
+	struct r2nm_cluster *cluster = r2nm_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(r2nm_configured_node_map);
+
+static struct r2nm_node *r2nm_node_ip_tree_lookup(struct r2nm_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 r2nm_node *node, *ret = NULL;
+
+	while (*p) {
+		int cmp;
+
+		parent = *p;
+		node = rb_entry(parent, struct r2nm_node, nd_ip_node);
+
+		cmp = memcmp(&ip_needle, &node->nd_ipv4_address,
+				sizeof(ip_needle));
+		if (cmp < 0)
+			p = &(*p)->rb_left;
+		else if (cmp > 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 r2nm_node *r2nm_get_node_by_ip(__be32 addr)
+{
+	struct r2nm_node *node = NULL;
+	struct r2nm_cluster *cluster = r2nm_single_cluster;
+
+	if (cluster == NULL)
+		goto out;
+
+	read_lock(&cluster->cl_nodes_lock);
+	node = r2nm_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(r2nm_get_node_by_ip);
+
+void r2nm_node_put(struct r2nm_node *node)
+{
+	config_item_put(&node->nd_item);
+}
+EXPORT_SYMBOL_GPL(r2nm_node_put);
+
+void r2nm_node_get(struct r2nm_node *node)
+{
+	config_item_get(&node->nd_item);
+}
+EXPORT_SYMBOL_GPL(r2nm_node_get);
+
+u8 r2nm_this_node(void)
+{
+	u8 node_num = R2NM_MAX_NODES;
+
+	if (r2nm_single_cluster && r2nm_single_cluster->cl_has_local)
+		node_num = r2nm_single_cluster->cl_local_node;
+
+	return node_num;
+}
+EXPORT_SYMBOL_GPL(r2nm_this_node);
+
+/* node configfs bits */
+
+static struct r2nm_cluster *to_r2nm_cluster(struct config_item *item)
+{
+	return item ?
+		container_of(to_config_group(item), struct r2nm_cluster,
+			     cl_group)
+		: NULL;
+}
+
+static struct r2nm_node *to_r2nm_node(struct config_item *item)
+{
+	return item ? container_of(item, struct r2nm_node, nd_item) : NULL;
+}
+
+static void r2nm_node_release(struct config_item *item)
+{
+	struct r2nm_node *node = to_r2nm_node(item);
+	kfree(node);
+}
+
+static ssize_t r2nm_node_num_read(struct r2nm_node *node, char *page)
+{
+	return sprintf(page, "%d\n", node->nd_num);
+}
+
+static struct r2nm_cluster *to_r2nm_cluster_from_node(struct r2nm_node *node)
+{
+	/* through the first node_set .parent
+	 * mycluster/nodes/mynode == r2nm_cluster->r2nm_node_group->r2nm_node */
+	return to_r2nm_cluster(node->nd_item.ci_parent->ci_parent);
+}
+
+enum {
+	R2NM_NODE_ATTR_NUM = 0,
+	R2NM_NODE_ATTR_PORT,
+	R2NM_NODE_ATTR_ADDRESS,
+	R2NM_NODE_ATTR_LOCAL,
+};
+
+static ssize_t r2nm_node_num_write(struct r2nm_node *node, const char *page,
+				   size_t count)
+{
+	struct r2nm_cluster *cluster = to_r2nm_cluster_from_node(node);
+	unsigned long tmp;
+	char *p = (char *)page;
+	int err;
+
+	err = kstrtoul(p, 10, &tmp);
+	if (err)
+		return err;
+
+	if (tmp >= R2NM_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(R2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
+	    !test_bit(R2NM_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 r2nm_node_ipv4_port_read(struct r2nm_node *node, char *page)
+{
+	return sprintf(page, "%u\n", ntohs(node->nd_ipv4_port));
+}
+
+static ssize_t r2nm_node_ipv4_port_write(struct r2nm_node *node,
+					 const char *page, size_t count)
+{
+	unsigned long tmp;
+	char *p = (char *)page;
+	int err;
+
+	err = kstrtoul(p, 10, &tmp);
+	if (err)
+		return err;
+
+	if (tmp == 0)
+		return -EINVAL;
+	if (tmp >= (u16)-1)
+		return -ERANGE;
+
+	node->nd_ipv4_port = htons(tmp);
+
+	return count;
+}
+
+static ssize_t r2nm_node_ipv4_address_read(struct r2nm_node *node, char *page)
+{
+	return sprintf(page, "%pI4\n", &node->nd_ipv4_address);
+}
+
+static ssize_t r2nm_node_ipv4_address_write(struct r2nm_node *node,
+					    const char *page,
+					    size_t count)
+{
+	struct r2nm_cluster *cluster = to_r2nm_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 (r2nm_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 r2nm_node_local_read(struct r2nm_node *node, char *page)
+{
+	return sprintf(page, "%d\n", node->nd_local);
+}
+
+static ssize_t r2nm_node_local_write(struct r2nm_node *node, const char *page,
+				     size_t count)
+{
+	struct r2nm_cluster *cluster = to_r2nm_cluster_from_node(node);
+	unsigned long tmp;
+	char *p = (char *)page;
+	ssize_t ret;
+	int err;
+
+	err = kstrtoul(p, 10, &tmp);
+	if (err)
+		return err;
+
+	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(R2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
+	    !test_bit(R2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
+	    !test_bit(R2NM_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 = r2net_start_listening(node);
+		if (ret)
+			return ret;
+	}
+
+	if (!tmp && cluster->cl_has_local &&
+	    cluster->cl_local_node == node->nd_num) {
+		r2net_stop_listening(node);
+		cluster->cl_local_node = R2NM_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 r2nm_node_attribute {
+	struct configfs_attribute attr;
+	ssize_t (*show)(struct r2nm_node *, char *);
+	ssize_t (*store)(struct r2nm_node *, const char *, size_t);
+};
+
+static struct r2nm_node_attribute r2nm_node_attr_num = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "num",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2nm_node_num_read,
+	.store	= r2nm_node_num_write,
+};
+
+static struct r2nm_node_attribute r2nm_node_attr_ipv4_port = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "ipv4_port",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2nm_node_ipv4_port_read,
+	.store	= r2nm_node_ipv4_port_write,
+};
+
+static struct r2nm_node_attribute r2nm_node_attr_ipv4_address = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "ipv4_address",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2nm_node_ipv4_address_read,
+	.store	= r2nm_node_ipv4_address_write,
+};
+
+static struct r2nm_node_attribute r2nm_node_attr_local = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "local",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2nm_node_local_read,
+	.store	= r2nm_node_local_write,
+};
+
+static struct configfs_attribute *r2nm_node_attrs[] = {
+	[R2NM_NODE_ATTR_NUM] = &r2nm_node_attr_num.attr,
+	[R2NM_NODE_ATTR_PORT] = &r2nm_node_attr_ipv4_port.attr,
+	[R2NM_NODE_ATTR_ADDRESS] = &r2nm_node_attr_ipv4_address.attr,
+	[R2NM_NODE_ATTR_LOCAL] = &r2nm_node_attr_local.attr,
+	NULL,
+};
+
+static int r2nm_attr_index(struct configfs_attribute *attr)
+{
+	int i;
+	for (i = 0; i < ARRAY_SIZE(r2nm_node_attrs); i++) {
+		if (attr == r2nm_node_attrs[i])
+			return i;
+	}
+	BUG();
+	return 0;
+}
+
+static ssize_t r2nm_node_show(struct config_item *item,
+			      struct configfs_attribute *attr,
+			      char *page)
+{
+	struct r2nm_node *node = to_r2nm_node(item);
+	struct r2nm_node_attribute *r2nm_node_attr =
+		container_of(attr, struct r2nm_node_attribute, attr);
+	ssize_t ret = 0;
+
+	if (r2nm_node_attr->show)
+		ret = r2nm_node_attr->show(node, page);
+	return ret;
+}
+
+static ssize_t r2nm_node_store(struct config_item *item,
+			       struct configfs_attribute *attr,
+			       const char *page, size_t count)
+{
+	struct r2nm_node *node = to_r2nm_node(item);
+	struct r2nm_node_attribute *r2nm_node_attr =
+		container_of(attr, struct r2nm_node_attribute, attr);
+	ssize_t ret;
+	int attr_index = r2nm_attr_index(attr);
+
+	if (r2nm_node_attr->store == NULL) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (test_bit(attr_index, &node->nd_set_attributes))
+		return -EBUSY;
+
+	ret = r2nm_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 r2nm_node_item_ops = {
+	.release		= r2nm_node_release,
+	.show_attribute		= r2nm_node_show,
+	.store_attribute	= r2nm_node_store,
+};
+
+static struct config_item_type r2nm_node_type = {
+	.ct_item_ops	= &r2nm_node_item_ops,
+	.ct_attrs	= r2nm_node_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* node set */
+
+struct r2nm_node_group {
+	struct config_group ns_group;
+	/* some stuff? */
+};
+
+#if 0
+static struct r2nm_node_group *to_r2nm_node_group(struct config_group *group)
+{
+	return group ?
+		container_of(group, struct r2nm_node_group, ns_group)
+		: NULL;
+}
+#endif
+
+struct r2nm_cluster_attribute {
+	struct configfs_attribute attr;
+	ssize_t (*show)(struct r2nm_cluster *, char *);
+	ssize_t (*store)(struct r2nm_cluster *, const char *, size_t);
+};
+
+static ssize_t r2nm_cluster_attr_write(const char *page, ssize_t count,
+					unsigned int *val)
+{
+	unsigned long tmp;
+	char *p = (char *)page;
+	int err;
+
+	err = kstrtoul(p, 10, &tmp);
+	if (err)
+		return err;
+
+	if (tmp == 0)
+		return -EINVAL;
+	if (tmp >= (u32)-1)
+		return -ERANGE;
+
+	*val = tmp;
+
+	return count;
+}
+
+static ssize_t r2nm_cluster_attr_idle_timeout_ms_read(
+	struct r2nm_cluster *cluster, char *page)
+{
+	return sprintf(page, "%u\n", cluster->cl_idle_timeout_ms);
+}
+
+static ssize_t r2nm_cluster_attr_idle_timeout_ms_write(
+	struct r2nm_cluster *cluster, const char *page, size_t count)
+{
+	ssize_t ret;
+	unsigned int val = 0;
+
+	ret =  r2nm_cluster_attr_write(page, count, &val);
+
+	if (ret > 0) {
+		if (cluster->cl_idle_timeout_ms != val
+			&& r2net_num_connected_peers()) {
+			mlog(ML_NOTICE,
+			     "r2net: cannot change idle timeout after "
+			     "the first peer has agreed to it."
+			     "  %d connected peers\n",
+			     r2net_num_connected_peers());
+			ret = -EINVAL;
+		} else if (val <= cluster->cl_keepalive_delay_ms) {
+			mlog(ML_NOTICE, "r2net: idle timeout must be larger "
+			     "than keepalive delay\n");
+			ret = -EINVAL;
+		} else {
+			cluster->cl_idle_timeout_ms = val;
+		}
+	}
+
+	return ret;
+}
+
+static ssize_t r2nm_cluster_attr_keepalive_delay_ms_read(
+	struct r2nm_cluster *cluster, char *page)
+{
+	return sprintf(page, "%u\n", cluster->cl_keepalive_delay_ms);
+}
+
+static ssize_t r2nm_cluster_attr_keepalive_delay_ms_write(
+	struct r2nm_cluster *cluster, const char *page, size_t count)
+{
+	ssize_t ret;
+	unsigned int val = 0;
+
+	ret =  r2nm_cluster_attr_write(page, count, &val);
+
+	if (ret > 0) {
+		if (cluster->cl_keepalive_delay_ms != val
+		    && r2net_num_connected_peers()) {
+			mlog(ML_NOTICE,
+			     "r2net: cannot change keepalive delay after"
+			     " the first peer has agreed to it."
+			     "  %d connected peers\n",
+			     r2net_num_connected_peers());
+			ret = -EINVAL;
+		} else if (val >= cluster->cl_idle_timeout_ms) {
+			mlog(ML_NOTICE, "r2net: keepalive delay must be "
+			     "smaller than idle timeout\n");
+			ret = -EINVAL;
+		} else {
+			cluster->cl_keepalive_delay_ms = val;
+		}
+	}
+
+	return ret;
+}
+
+static ssize_t r2nm_cluster_attr_reconnect_delay_ms_read(
+	struct r2nm_cluster *cluster, char *page)
+{
+	return sprintf(page, "%u\n", cluster->cl_reconnect_delay_ms);
+}
+
+static ssize_t r2nm_cluster_attr_reconnect_delay_ms_write(
+	struct r2nm_cluster *cluster, const char *page, size_t count)
+{
+	return r2nm_cluster_attr_write(page, count,
+					&cluster->cl_reconnect_delay_ms);
+}
+
+static ssize_t r2nm_cluster_attr_fence_method_read(
+	struct r2nm_cluster *cluster, char *page)
+{
+	ssize_t ret = 0;
+
+	if (cluster)
+		ret = sprintf(page, "%s\n",
+			      r2nm_fence_method_desc[cluster->cl_fence_method]);
+	return ret;
+}
+
+static ssize_t r2nm_cluster_attr_fence_method_write(
+	struct r2nm_cluster *cluster, const char *page, size_t count)
+{
+	unsigned int i;
+
+	if (page[count - 1] != '\n')
+		goto bail;
+
+	for (i = 0; i < R2NM_FENCE_METHODS; ++i) {
+		if (count != strlen(r2nm_fence_method_desc[i]) + 1)
+			continue;
+		if (strncasecmp(page, r2nm_fence_method_desc[i], count - 1))
+			continue;
+		if (cluster->cl_fence_method != i) {
+			printk(KERN_INFO "ramster: Changing fence method to %s\n",
+			       r2nm_fence_method_desc[i]);
+			cluster->cl_fence_method = i;
+		}
+		return count;
+	}
+
+bail:
+	return -EINVAL;
+}
+
+static struct r2nm_cluster_attribute r2nm_cluster_attr_idle_timeout_ms = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "idle_timeout_ms",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2nm_cluster_attr_idle_timeout_ms_read,
+	.store	= r2nm_cluster_attr_idle_timeout_ms_write,
+};
+
+static struct r2nm_cluster_attribute r2nm_cluster_attr_keepalive_delay_ms = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "keepalive_delay_ms",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2nm_cluster_attr_keepalive_delay_ms_read,
+	.store	= r2nm_cluster_attr_keepalive_delay_ms_write,
+};
+
+static struct r2nm_cluster_attribute r2nm_cluster_attr_reconnect_delay_ms = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "reconnect_delay_ms",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2nm_cluster_attr_reconnect_delay_ms_read,
+	.store	= r2nm_cluster_attr_reconnect_delay_ms_write,
+};
+
+static struct r2nm_cluster_attribute r2nm_cluster_attr_fence_method = {
+	.attr	= { .ca_owner = THIS_MODULE,
+		    .ca_name = "fence_method",
+		    .ca_mode = S_IRUGO | S_IWUSR },
+	.show	= r2nm_cluster_attr_fence_method_read,
+	.store	= r2nm_cluster_attr_fence_method_write,
+};
+
+static struct configfs_attribute *r2nm_cluster_attrs[] = {
+	&r2nm_cluster_attr_idle_timeout_ms.attr,
+	&r2nm_cluster_attr_keepalive_delay_ms.attr,
+	&r2nm_cluster_attr_reconnect_delay_ms.attr,
+	&r2nm_cluster_attr_fence_method.attr,
+	NULL,
+};
+static ssize_t r2nm_cluster_show(struct config_item *item,
+					struct configfs_attribute *attr,
+					char *page)
+{
+	struct r2nm_cluster *cluster = to_r2nm_cluster(item);
+	struct r2nm_cluster_attribute *r2nm_cluster_attr =
+		container_of(attr, struct r2nm_cluster_attribute, attr);
+	ssize_t ret = 0;
+
+	if (r2nm_cluster_attr->show)
+		ret = r2nm_cluster_attr->show(cluster, page);
+	return ret;
+}
+
+static ssize_t r2nm_cluster_store(struct config_item *item,
+					struct configfs_attribute *attr,
+					const char *page, size_t count)
+{
+	struct r2nm_cluster *cluster = to_r2nm_cluster(item);
+	struct r2nm_cluster_attribute *r2nm_cluster_attr =
+		container_of(attr, struct r2nm_cluster_attribute, attr);
+	ssize_t ret;
+
+	if (r2nm_cluster_attr->store == NULL) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ret = r2nm_cluster_attr->store(cluster, page, count);
+	if (ret < count)
+		goto out;
+out:
+	return ret;
+}
+
+static struct config_item *r2nm_node_group_make_item(struct config_group *group,
+						     const char *name)
+{
+	struct r2nm_node *node = NULL;
+
+	if (strlen(name) > R2NM_MAX_NAME_LEN)
+		return ERR_PTR(-ENAMETOOLONG);
+
+	node = kzalloc(sizeof(struct r2nm_node), GFP_KERNEL);
+	if (node == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
+	config_item_init_type_name(&node->nd_item, name, &r2nm_node_type);
+	spin_lock_init(&node->nd_lock);
+
+	mlog(ML_CLUSTER, "r2nm: Registering node %s\n", name);
+
+	return &node->nd_item;
+}
+
+static void r2nm_node_group_drop_item(struct config_group *group,
+				      struct config_item *item)
+{
+	struct r2nm_node *node = to_r2nm_node(item);
+	struct r2nm_cluster *cluster =
+				to_r2nm_cluster(group->cg_item.ci_parent);
+
+	r2net_disconnect_node(node);
+
+	if (cluster->cl_has_local &&
+	    (cluster->cl_local_node == node->nd_num)) {
+		cluster->cl_has_local = 0;
+		cluster->cl_local_node = R2NM_INVALID_NODE_NUM;
+		r2net_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);
+
+	mlog(ML_CLUSTER, "r2nm: Unregistered node %s\n",
+	     config_item_name(&node->nd_item));
+
+	config_item_put(item);
+}
+
+static struct configfs_group_operations r2nm_node_group_group_ops = {
+	.make_item	= r2nm_node_group_make_item,
+	.drop_item	= r2nm_node_group_drop_item,
+};
+
+static struct config_item_type r2nm_node_group_type = {
+	.ct_group_ops	= &r2nm_node_group_group_ops,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* cluster */
+
+static void r2nm_cluster_release(struct config_item *item)
+{
+	struct r2nm_cluster *cluster = to_r2nm_cluster(item);
+
+	kfree(cluster->cl_group.default_groups);
+	kfree(cluster);
+}
+
+static struct configfs_item_operations r2nm_cluster_item_ops = {
+	.release	= r2nm_cluster_release,
+	.show_attribute		= r2nm_cluster_show,
+	.store_attribute	= r2nm_cluster_store,
+};
+
+static struct config_item_type r2nm_cluster_type = {
+	.ct_item_ops	= &r2nm_cluster_item_ops,
+	.ct_attrs	= r2nm_cluster_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+/* cluster set */
+
+struct r2nm_cluster_group {
+	struct configfs_subsystem cs_subsys;
+	/* some stuff? */
+};
+
+#if 0
+static struct r2nm_cluster_group *
+to_r2nm_cluster_group(struct config_group *group)
+{
+	return group ?
+		container_of(to_configfs_subsystem(group),
+				struct r2nm_cluster_group, cs_subsys)
+	       : NULL;
+}
+#endif
+
+static struct config_group *
+r2nm_cluster_group_make_group(struct config_group *group,
+							  const char *name)
+{
+	struct r2nm_cluster *cluster = NULL;
+	struct r2nm_node_group *ns = NULL;
+	struct config_group *r2hb_group = NULL, *ret = NULL;
+	void *defs = NULL;
+
+	/* this runs under the parent dir's i_mutex; there can be only
+	 * one caller in here at a time */
+	if (r2nm_single_cluster)
+		return ERR_PTR(-ENOSPC);
+
+	cluster = kzalloc(sizeof(struct r2nm_cluster), GFP_KERNEL);
+	ns = kzalloc(sizeof(struct r2nm_node_group), GFP_KERNEL);
+	defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL);
+	r2hb_group = r2hb_alloc_hb_set();
+	if (cluster == NULL || ns == NULL || r2hb_group == NULL || defs == NULL)
+		goto out;
+
+	config_group_init_type_name(&cluster->cl_group, name,
+				    &r2nm_cluster_type);
+	config_group_init_type_name(&ns->ns_group, "node",
+				    &r2nm_node_group_type);
+
+	cluster->cl_group.default_groups = defs;
+	cluster->cl_group.default_groups[0] = &ns->ns_group;
+	cluster->cl_group.default_groups[1] = r2hb_group;
+	cluster->cl_group.default_groups[2] = NULL;
+	rwlock_init(&cluster->cl_nodes_lock);
+	cluster->cl_node_ip_tree = RB_ROOT;
+	cluster->cl_reconnect_delay_ms = R2NET_RECONNECT_DELAY_MS_DEFAULT;
+	cluster->cl_idle_timeout_ms    = R2NET_IDLE_TIMEOUT_MS_DEFAULT;
+	cluster->cl_keepalive_delay_ms = R2NET_KEEPALIVE_DELAY_MS_DEFAULT;
+	cluster->cl_fence_method       = R2NM_FENCE_RESET;
+
+	ret = &cluster->cl_group;
+	r2nm_single_cluster = cluster;
+
+out:
+	if (ret == NULL) {
+		kfree(cluster);
+		kfree(ns);
+		r2hb_free_hb_set(r2hb_group);
+		kfree(defs);
+		ret = ERR_PTR(-ENOMEM);
+	}
+
+	return ret;
+}
+
+static void r2nm_cluster_group_drop_item(struct config_group *group,
+						struct config_item *item)
+{
+	struct r2nm_cluster *cluster = to_r2nm_cluster(item);
+	int i;
+	struct config_item *killme;
+
+	BUG_ON(r2nm_single_cluster != cluster);
+	r2nm_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 r2nm_cluster_group_group_ops = {
+	.make_group	= r2nm_cluster_group_make_group,
+	.drop_item	= r2nm_cluster_group_drop_item,
+};
+
+static struct config_item_type r2nm_cluster_group_type = {
+	.ct_group_ops	= &r2nm_cluster_group_group_ops,
+	.ct_owner	= THIS_MODULE,
+};
+
+static struct r2nm_cluster_group r2nm_cluster_group = {
+	.cs_subsys = {
+		.su_group = {
+			.cg_item = {
+				.ci_namebuf = "cluster",
+				.ci_type = &r2nm_cluster_group_type,
+			},
+		},
+	},
+};
+
+int r2nm_depend_item(struct config_item *item)
+{
+	return configfs_depend_item(&r2nm_cluster_group.cs_subsys, item);
+}
+
+void r2nm_undepend_item(struct config_item *item)
+{
+	configfs_undepend_item(&r2nm_cluster_group.cs_subsys, item);
+}
+
+int r2nm_depend_this_node(void)
+{
+	int ret = 0;
+	struct r2nm_node *local_node;
+
+	local_node = r2nm_get_node_by_num(r2nm_this_node());
+	if (!local_node) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ret = r2nm_depend_item(&local_node->nd_item);
+	r2nm_node_put(local_node);
+
+out:
+	return ret;
+}
+
+void r2nm_undepend_this_node(void)
+{
+	struct r2nm_node *local_node;
+
+	local_node = r2nm_get_node_by_num(r2nm_this_node());
+	BUG_ON(!local_node);
+
+	r2nm_undepend_item(&local_node->nd_item);
+	r2nm_node_put(local_node);
+}
+
+
+static void __exit exit_r2nm(void)
+{
+	/* XXX sync with hb callbacks and shut down hb? */
+	r2net_unregister_hb_callbacks();
+	configfs_unregister_subsystem(&r2nm_cluster_group.cs_subsys);
+
+	r2net_exit();
+	r2hb_exit();
+}
+
+static int __init init_r2nm(void)
+{
+	int ret = -1;
+
+	ret = r2hb_init();
+	if (ret)
+		goto out;
+
+	ret = r2net_init();
+	if (ret)
+		goto out_r2hb;
+
+	ret = r2net_register_hb_callbacks();
+	if (ret)
+		goto out_r2net;
+
+	config_group_init(&r2nm_cluster_group.cs_subsys.su_group);
+	mutex_init(&r2nm_cluster_group.cs_subsys.su_mutex);
+	ret = configfs_register_subsystem(&r2nm_cluster_group.cs_subsys);
+	if (ret) {
+		printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
+		goto out_callbacks;
+	}
+
+	if (!ret)
+		goto out;
+
+	configfs_unregister_subsystem(&r2nm_cluster_group.cs_subsys);
+out_callbacks:
+	r2net_unregister_hb_callbacks();
+out_r2net:
+	r2net_exit();
+out_r2hb:
+	r2hb_exit();
+out:
+	return ret;
+}
+
+MODULE_AUTHOR("Oracle");
+MODULE_LICENSE("GPL");
+
+module_init(init_r2nm)
+module_exit(exit_r2nm)
diff --git a/drivers/staging/ramster/cluster/nodemanager.h b/drivers/staging/ramster/cluster/nodemanager.h
new file mode 100644
index 0000000..41a04df
--- /dev/null
+++ b/drivers/staging/ramster/cluster/nodemanager.h
@@ -0,0 +1,88 @@
+/* -*- 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 R2CLUSTER_NODEMANAGER_H
+#define R2CLUSTER_NODEMANAGER_H
+
+#include "ramster_nodemanager.h"
+
+/* This totally doesn't belong here. */
+#include <linux/configfs.h>
+#include <linux/rbtree.h>
+
+enum r2nm_fence_method {
+	R2NM_FENCE_RESET	= 0,
+	R2NM_FENCE_PANIC,
+	R2NM_FENCE_METHODS,	/* Number of fence methods */
+};
+
+struct r2nm_node {
+	spinlock_t		nd_lock;
+	struct config_item	nd_item;
+	char			nd_name[R2NM_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;
+};
+
+struct r2nm_cluster {
+	struct config_group	cl_group;
+	unsigned		cl_has_local:1;
+	u8			cl_local_node;
+	rwlock_t		cl_nodes_lock;
+	struct r2nm_node	*cl_nodes[R2NM_MAX_NODES];
+	struct rb_root		cl_node_ip_tree;
+	unsigned int		cl_idle_timeout_ms;
+	unsigned int		cl_keepalive_delay_ms;
+	unsigned int		cl_reconnect_delay_ms;
+	enum r2nm_fence_method	cl_fence_method;
+
+	/* part of a hack for disk bitmap.. will go eventually. - zab */
+	unsigned long	cl_nodes_bitmap[BITS_TO_LONGS(R2NM_MAX_NODES)];
+};
+
+extern struct r2nm_cluster *r2nm_single_cluster;
+
+u8 r2nm_this_node(void);
+
+int r2nm_configured_node_map(unsigned long *map, unsigned bytes);
+struct r2nm_node *r2nm_get_node_by_num(u8 node_num);
+struct r2nm_node *r2nm_get_node_by_ip(__be32 addr);
+void r2nm_node_get(struct r2nm_node *node);
+void r2nm_node_put(struct r2nm_node *node);
+
+int r2nm_depend_item(struct config_item *item);
+void r2nm_undepend_item(struct config_item *item);
+int r2nm_depend_this_node(void);
+void r2nm_undepend_this_node(void);
+
+#endif /* R2CLUSTER_NODEMANAGER_H */
diff --git a/drivers/staging/ramster/cluster/ramster_nodemanager.h b/drivers/staging/ramster/cluster/ramster_nodemanager.h
new file mode 100644
index 0000000..49f879d
--- /dev/null
+++ b/drivers/staging/ramster/cluster/ramster_nodemanager.h
@@ -0,0 +1,39 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ramster_nodemanager.h
+ *
+ * Header describing the interface between userspace and the kernel
+ * for the ramster_nodemanager module.
+ *
+ * Copyright (C) 2002, 2004, 2012 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 _RAMSTER_NODEMANAGER_H
+#define _RAMSTER_NODEMANAGER_H
+
+#define R2NM_API_VERSION	5
+
+#define R2NM_MAX_NODES		255
+#define R2NM_INVALID_NODE_NUM	255
+
+/* host name, group name, cluster name all 64 bytes */
+#define R2NM_MAX_NAME_LEN        64    /* __NEW_UTS_LEN */
+
+#endif /* _RAMSTER_NODEMANAGER_H */
diff --git a/drivers/staging/ramster/cluster/tcp.c b/drivers/staging/ramster/cluster/tcp.c
new file mode 100644
index 0000000..3af1b2c
--- /dev/null
+++ b/drivers/staging/ramster/cluster/tcp.c
@@ -0,0 +1,2256 @@
+/* -*- 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 <linux/net.h>
+#include <linux/export.h>
+#include <linux/uaccess.h>
+#include <net/tcp.h>
+
+
+#include "heartbeat.h"
+#include "tcp.h"
+#include "nodemanager.h"
+#define MLOG_MASK_PREFIX ML_TCP
+#include "masklog.h"
+
+#include "tcp_internal.h"
+
+#define SC_NODEF_FMT "node %s (num %u) at %pI4:%u"
+
+/*
+ * 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 DEFINE_RWLOCK(r2net_handler_lock);
+static struct rb_root r2net_handler_tree = RB_ROOT;
+
+static struct r2net_node r2net_nodes[R2NM_MAX_NODES];
+
+/* XXX someday we'll need better accounting */
+static struct socket *r2net_listen_sock;
+
+/*
+ * listen work is only queued by the listening socket callbacks on the
+ * r2net_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 *r2net_wq;
+static struct work_struct r2net_listen_work;
+
+static struct r2hb_callback_func r2net_hb_up, r2net_hb_down;
+#define R2NET_HB_PRI 0x1
+
+static struct r2net_handshake *r2net_hand;
+static struct r2net_msg *r2net_keep_req, *r2net_keep_resp;
+
+static int r2net_sys_err_translations[R2NET_ERR_MAX] = {
+		[R2NET_ERR_NONE]	= 0,
+		[R2NET_ERR_NO_HNDLR]	= -ENOPROTOOPT,
+		[R2NET_ERR_OVERFLOW]	= -EOVERFLOW,
+		[R2NET_ERR_DIED]	= -EHOSTDOWN,};
+
+/* can't quite avoid *all* internal declarations :/ */
+static void r2net_sc_connect_completed(struct work_struct *work);
+static void r2net_rx_until_empty(struct work_struct *work);
+static void r2net_shutdown_sc(struct work_struct *work);
+static void r2net_listen_data_ready(struct sock *sk, int bytes);
+static void r2net_sc_send_keep_req(struct work_struct *work);
+static void r2net_idle_timer(unsigned long data);
+static void r2net_sc_postpone_idle(struct r2net_sock_container *sc);
+static void r2net_sc_reset_idle_timer(struct r2net_sock_container *sc);
+
+#ifdef CONFIG_DEBUG_FS
+static void r2net_init_nst(struct r2net_send_tracking *nst, u32 msgtype,
+			   u32 msgkey, struct task_struct *task, u8 node)
+{
+	INIT_LIST_HEAD(&nst->st_net_debug_item);
+	nst->st_task = task;
+	nst->st_msg_type = msgtype;
+	nst->st_msg_key = msgkey;
+	nst->st_node = node;
+}
+
+static inline void r2net_set_nst_sock_time(struct r2net_send_tracking *nst)
+{
+	nst->st_sock_time = ktime_get();
+}
+
+static inline void r2net_set_nst_send_time(struct r2net_send_tracking *nst)
+{
+	nst->st_send_time = ktime_get();
+}
+
+static inline void r2net_set_nst_status_time(struct r2net_send_tracking *nst)
+{
+	nst->st_status_time = ktime_get();
+}
+
+static inline void r2net_set_nst_sock_container(struct r2net_send_tracking *nst,
+						struct r2net_sock_container *sc)
+{
+	nst->st_sc = sc;
+}
+
+static inline void r2net_set_nst_msg_id(struct r2net_send_tracking *nst,
+					u32 msg_id)
+{
+	nst->st_id = msg_id;
+}
+
+static inline void r2net_set_sock_timer(struct r2net_sock_container *sc)
+{
+	sc->sc_tv_timer = ktime_get();
+}
+
+static inline void r2net_set_data_ready_time(struct r2net_sock_container *sc)
+{
+	sc->sc_tv_data_ready = ktime_get();
+}
+
+static inline void r2net_set_advance_start_time(struct r2net_sock_container *sc)
+{
+	sc->sc_tv_advance_start = ktime_get();
+}
+
+static inline void r2net_set_advance_stop_time(struct r2net_sock_container *sc)
+{
+	sc->sc_tv_advance_stop = ktime_get();
+}
+
+static inline void r2net_set_func_start_time(struct r2net_sock_container *sc)
+{
+	sc->sc_tv_func_start = ktime_get();
+}
+
+static inline void r2net_set_func_stop_time(struct r2net_sock_container *sc)
+{
+	sc->sc_tv_func_stop = ktime_get();
+}
+
+#else  /* CONFIG_DEBUG_FS */
+# define r2net_init_nst(a, b, c, d, e)
+# define r2net_set_nst_sock_time(a)
+# define r2net_set_nst_send_time(a)
+# define r2net_set_nst_status_time(a)
+# define r2net_set_nst_sock_container(a, b)
+# define r2net_set_nst_msg_id(a, b)
+# define r2net_set_sock_timer(a)
+# define r2net_set_data_ready_time(a)
+# define r2net_set_advance_start_time(a)
+# define r2net_set_advance_stop_time(a)
+# define r2net_set_func_start_time(a)
+# define r2net_set_func_stop_time(a)
+#endif /* CONFIG_DEBUG_FS */
+
+#ifdef CONFIG_RAMSTER_FS_STATS
+static ktime_t r2net_get_func_run_time(struct r2net_sock_container *sc)
+{
+	return ktime_sub(sc->sc_tv_func_stop, sc->sc_tv_func_start);
+}
+
+static void r2net_update_send_stats(struct r2net_send_tracking *nst,
+				    struct r2net_sock_container *sc)
+{
+	sc->sc_tv_status_total = ktime_add(sc->sc_tv_status_total,
+					   ktime_sub(ktime_get(),
+						     nst->st_status_time));
+	sc->sc_tv_send_total = ktime_add(sc->sc_tv_send_total,
+					 ktime_sub(nst->st_status_time,
+						   nst->st_send_time));
+	sc->sc_tv_acquiry_total = ktime_add(sc->sc_tv_acquiry_total,
+					    ktime_sub(nst->st_send_time,
+						      nst->st_sock_time));
+	sc->sc_send_count++;
+}
+
+static void r2net_update_recv_stats(struct r2net_sock_container *sc)
+{
+	sc->sc_tv_process_total = ktime_add(sc->sc_tv_process_total,
+					    r2net_get_func_run_time(sc));
+	sc->sc_recv_count++;
+}
+
+#else
+
+# define r2net_update_send_stats(a, b)
+
+# define r2net_update_recv_stats(sc)
+
+#endif /* CONFIG_RAMSTER_FS_STATS */
+
+static inline int r2net_reconnect_delay(void)
+{
+	return r2nm_single_cluster->cl_reconnect_delay_ms;
+}
+
+static inline int r2net_keepalive_delay(void)
+{
+	return r2nm_single_cluster->cl_keepalive_delay_ms;
+}
+
+static inline int r2net_idle_timeout(void)
+{
+	return r2nm_single_cluster->cl_idle_timeout_ms;
+}
+
+static inline int r2net_sys_err_to_errno(enum r2net_system_error err)
+{
+	int trans;
+	BUG_ON(err >= R2NET_ERR_MAX);
+	trans = r2net_sys_err_translations[err];
+
+	/* Just in case we mess up the translation table above */
+	BUG_ON(err != R2NET_ERR_NONE && trans == 0);
+	return trans;
+}
+
+struct r2net_node *r2net_nn_from_num(u8 node_num)
+{
+	BUG_ON(node_num >= ARRAY_SIZE(r2net_nodes));
+	return &r2net_nodes[node_num];
+}
+
+static u8 r2net_num_from_nn(struct r2net_node *nn)
+{
+	BUG_ON(nn == NULL);
+	return nn - r2net_nodes;
+}
+
+/* ------------------------------------------------------------ */
+
+static int r2net_prep_nsw(struct r2net_node *nn, struct r2net_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 = R2NET_ERR_NONE;
+		nsw->ns_status = 0;
+	}
+
+	return ret;
+}
+
+static void r2net_complete_nsw_locked(struct r2net_node *nn,
+				      struct r2net_status_wait *nsw,
+				      enum r2net_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 r2net_complete_nsw(struct r2net_node *nn,
+			       struct r2net_status_wait *nsw,
+			       u64 id, enum r2net_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;
+	}
+
+	r2net_complete_nsw_locked(nn, nsw, sys_status, status);
+
+out:
+	spin_unlock(&nn->nn_lock);
+	return;
+}
+
+static void r2net_complete_nodes_nsw(struct r2net_node *nn)
+{
+	struct r2net_status_wait *nsw, *tmp;
+	unsigned int num_kills = 0;
+
+	assert_spin_locked(&nn->nn_lock);
+
+	list_for_each_entry_safe(nsw, tmp, &nn->nn_status_list, ns_node_item) {
+		r2net_complete_nsw_locked(nn, nsw, R2NET_ERR_DIED, 0);
+		num_kills++;
+	}
+
+	mlog(0, "completed %d messages for node %u\n", num_kills,
+	     r2net_num_from_nn(nn));
+}
+
+static int r2net_nsw_completed(struct r2net_node *nn,
+			       struct r2net_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 r2net_sock_container *sc = container_of(kref,
+					struct r2net_sock_container, sc_kref);
+	BUG_ON(timer_pending(&sc->sc_idle_timeout));
+
+	sclog(sc, "releasing\n");
+
+	if (sc->sc_sock) {
+		sock_release(sc->sc_sock);
+		sc->sc_sock = NULL;
+	}
+
+	r2nm_undepend_item(&sc->sc_node->nd_item);
+	r2nm_node_put(sc->sc_node);
+	sc->sc_node = NULL;
+
+	r2net_debug_del_sc(sc);
+	kfree(sc);
+}
+
+static void sc_put(struct r2net_sock_container *sc)
+{
+	sclog(sc, "put\n");
+	kref_put(&sc->sc_kref, sc_kref_release);
+}
+static void sc_get(struct r2net_sock_container *sc)
+{
+	sclog(sc, "get\n");
+	kref_get(&sc->sc_kref);
+}
+static struct r2net_sock_container *sc_alloc(struct r2nm_node *node)
+{
+	struct r2net_sock_container *sc, *ret = NULL;
+	struct page *page = NULL;
+	int status = 0;
+
+	page = alloc_page(GFP_NOFS);
+	sc = kzalloc(sizeof(*sc), GFP_NOFS);
+	if (sc == NULL || page == NULL)
+		goto out;
+
+	kref_init(&sc->sc_kref);
+	r2nm_node_get(node);
+	sc->sc_node = node;
+
+	/* pin the node item of the remote node */
+	status = r2nm_depend_item(&node->nd_item);
+	if (status) {
+		mlog_errno(status);
+		r2nm_node_put(node);
+		goto out;
+	}
+	INIT_WORK(&sc->sc_connect_work, r2net_sc_connect_completed);
+	INIT_WORK(&sc->sc_rx_work, r2net_rx_until_empty);
+	INIT_WORK(&sc->sc_shutdown_work, r2net_shutdown_sc);
+	INIT_DELAYED_WORK(&sc->sc_keepalive_work, r2net_sc_send_keep_req);
+
+	init_timer(&sc->sc_idle_timeout);
+	sc->sc_idle_timeout.function = r2net_idle_timer;
+	sc->sc_idle_timeout.data = (unsigned long)sc;
+
+	sclog(sc, "alloced\n");
+
+	ret = sc;
+	sc->sc_page = page;
+	r2net_debug_add_sc(sc);
+	sc = NULL;
+	page = NULL;
+
+out:
+	if (page)
+		__free_page(page);
+	kfree(sc);
+
+	return ret;
+}
+
+/* ------------------------------------------------------------ */
+
+static void r2net_sc_queue_work(struct r2net_sock_container *sc,
+				struct work_struct *work)
+{
+	sc_get(sc);
+	if (!queue_work(r2net_wq, work))
+		sc_put(sc);
+}
+static void r2net_sc_queue_delayed_work(struct r2net_sock_container *sc,
+					struct delayed_work *work,
+					int delay)
+{
+	sc_get(sc);
+	if (!queue_delayed_work(r2net_wq, work, delay))
+		sc_put(sc);
+}
+static void r2net_sc_cancel_delayed_work(struct r2net_sock_container *sc,
+					 struct delayed_work *work)
+{
+	if (cancel_delayed_work(work))
+		sc_put(sc);
+}
+
+static atomic_t r2net_connected_peers = ATOMIC_INIT(0);
+
+int r2net_num_connected_peers(void)
+{
+	return atomic_read(&r2net_connected_peers);
+}
+
+static void r2net_set_nn_state(struct r2net_node *nn,
+			       struct r2net_sock_container *sc,
+			       unsigned valid, int err)
+{
+	int was_valid = nn->nn_sc_valid;
+	int was_err = nn->nn_persistent_error;
+	struct r2net_sock_container *old_sc = nn->nn_sc;
+
+	assert_spin_locked(&nn->nn_lock);
+
+	if (old_sc && !sc)
+		atomic_dec(&r2net_connected_peers);
+	else if (!old_sc && sc)
+		atomic_inc(&r2net_connected_peers);
+
+	/* 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);
+
+	if (was_valid && !valid && err == 0)
+		err = -ENOTCONN;
+
+	mlog(ML_CONN, "node %u sc: %p -> %p, valid %u -> %u, err %d -> %d\n",
+	     r2net_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 r2net_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) {
+		queue_delayed_work(r2net_wq, &nn->nn_still_up,
+				   msecs_to_jiffies(R2NET_QUORUM_DELAY_MS));
+	}
+
+	if (was_valid && !valid) {
+		printk(KERN_NOTICE "ramster: No longer connected to "
+		       SC_NODEF_FMT "\n",
+			old_sc->sc_node->nd_name, old_sc->sc_node->nd_num,
+			&old_sc->sc_node->nd_ipv4_address,
+			ntohs(old_sc->sc_node->nd_ipv4_port));
+		r2net_complete_nodes_nsw(nn);
+	}
+
+	if (!was_valid && valid) {
+		cancel_delayed_work(&nn->nn_connect_expired);
+		printk(KERN_NOTICE "ramster: %s " SC_NODEF_FMT "\n",
+		       r2nm_this_node() > sc->sc_node->nd_num ?
+		       "Connected to" : "Accepted connection from",
+		       sc->sc_node->nd_name, sc->sc_node->nd_num,
+			&sc->sc_node->nd_ipv4_address,
+			ntohs(sc->sc_node->nd_ipv4_port));
+	}
+
+	/* 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 && r2net_wq) {
+		unsigned long delay;
+		/* delay if we're within a RECONNECT_DELAY of the
+		 * last attempt */
+		delay = (nn->nn_last_connect_attempt +
+			 msecs_to_jiffies(r2net_reconnect_delay()))
+			- jiffies;
+		if (delay > msecs_to_jiffies(r2net_reconnect_delay()))
+			delay = 0;
+		mlog(ML_CONN, "queueing conn attempt in %lu jiffies\n", delay);
+		queue_delayed_work(r2net_wq, &nn->nn_connect_work, delay);
+
+		/*
+		 * Delay the expired work after idle timeout.
+		 *
+		 * We might have lots of failed connection attempts that run
+		 * through here but we only cancel the connect_expired work when
+		 * a connection attempt succeeds.  So only the first enqueue of
+		 * the connect_expired work will do anything.  The rest will see
+		 * that it's already queued and do nothing.
+		 */
+		delay += msecs_to_jiffies(r2net_idle_timeout());
+		queue_delayed_work(r2net_wq, &nn->nn_connect_expired, 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)) {
+		r2net_sc_queue_work(old_sc, &old_sc->sc_shutdown_work);
+		sc_put(old_sc);
+	}
+}
+
+/* see r2net_register_callbacks() */
+static void r2net_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 r2net_sock_container *sc = sk->sk_user_data;
+		sclog(sc, "data_ready hit\n");
+		r2net_set_data_ready_time(sc);
+		r2net_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 r2net_register_callbacks() */
+static void r2net_state_change(struct sock *sk)
+{
+	void (*state_change)(struct sock *sk);
+	struct r2net_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:
+		r2net_sc_queue_work(sc, &sc->sc_connect_work);
+		break;
+	default:
+		printk(KERN_INFO "ramster: Connection to "
+			SC_NODEF_FMT " shutdown, state %d\n",
+			sc->sc_node->nd_name, sc->sc_node->nd_num,
+			&sc->sc_node->nd_ipv4_address,
+			ntohs(sc->sc_node->nd_ipv4_port), sk->sk_state);
+		r2net_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 r2net_unregister_callbacks().
+ */
+static void r2net_register_callbacks(struct sock *sk,
+				     struct r2net_sock_container *sc)
+{
+	write_lock_bh(&sk->sk_callback_lock);
+
+	/* accepted sockets inherit the old listen socket data ready */
+	if (sk->sk_data_ready == r2net_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 = r2net_data_ready;
+	sk->sk_state_change = r2net_state_change;
+
+	mutex_init(&sc->sc_send_lock);
+
+	write_unlock_bh(&sk->sk_callback_lock);
+}
+
+static int r2net_unregister_callbacks(struct sock *sk,
+					struct r2net_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 r2net_ensure_shutdown(struct r2net_node *nn,
+					struct r2net_sock_container *sc,
+				   int err)
+{
+	spin_lock(&nn->nn_lock);
+	if (nn->nn_sc == sc)
+		r2net_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 r2net_shutdown_sc(struct work_struct *work)
+{
+	struct r2net_sock_container *sc =
+		container_of(work, struct r2net_sock_container,
+			     sc_shutdown_work);
+	struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
+
+	sclog(sc, "shutting down\n");
+
+	/* drop the callbacks ref and call shutdown only once */
+	if (r2net_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);
+		r2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work);
+		sc_put(sc);
+		kernel_sock_shutdown(sc->sc_sock, SHUT_RDWR);
+	}
+
+	/* not fatal so failed connects before the other guy has our
+	 * heartbeat can be retried */
+	r2net_ensure_shutdown(nn, sc, 0);
+	sc_put(sc);
+}
+
+/* ------------------------------------------------------------ */
+
+static int r2net_handler_cmp(struct r2net_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 r2net_msg_handler *
+r2net_handler_tree_lookup(u32 msg_type, u32 key, struct rb_node ***ret_p,
+				struct rb_node **ret_parent)
+{
+	struct rb_node **p = &r2net_handler_tree.rb_node;
+	struct rb_node *parent = NULL;
+	struct r2net_msg_handler *nmh, *ret = NULL;
+	int cmp;
+
+	while (*p) {
+		parent = *p;
+		nmh = rb_entry(parent, struct r2net_msg_handler, nh_node);
+		cmp = r2net_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 r2net_handler_kref_release(struct kref *kref)
+{
+	struct r2net_msg_handler *nmh;
+	nmh = container_of(kref, struct r2net_msg_handler, nh_kref);
+
+	kfree(nmh);
+}
+
+static void r2net_handler_put(struct r2net_msg_handler *nmh)
+{
+	kref_put(&nmh->nh_kref, r2net_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 r2net_register_handler(u32 msg_type, u32 key, u32 max_len,
+			   r2net_msg_handler_func *func, void *data,
+			   r2net_post_msg_handler_func *post_func,
+			   struct list_head *unreg_list)
+{
+	struct r2net_msg_handler *nmh = NULL;
+	struct rb_node **p, *parent;
+	int ret = 0;
+
+	if (max_len > R2NET_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 = kzalloc(sizeof(struct r2net_msg_handler), GFP_NOFS);
+	if (nmh == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	nmh->nh_func = func;
+	nmh->nh_func_data = data;
+	nmh->nh_post_func = post_func;
+	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(&r2net_handler_lock);
+	if (r2net_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, &r2net_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(r2net_handler_tree_lookup(msg_type, key, &p,
+							  &parent) == NULL,
+				"couldn't find handler we *just* registered "
+				"for type %u key %08x\n", msg_type, key);
+	}
+	write_unlock(&r2net_handler_lock);
+	if (ret)
+		goto out;
+
+out:
+	if (ret)
+		kfree(nmh);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(r2net_register_handler);
+
+void r2net_unregister_handler_list(struct list_head *list)
+{
+	struct r2net_msg_handler *nmh, *n;
+
+	write_lock(&r2net_handler_lock);
+	list_for_each_entry_safe(nmh, n, list, 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, &r2net_handler_tree);
+		list_del_init(&nmh->nh_unregister_item);
+		kref_put(&nmh->nh_kref, r2net_handler_kref_release);
+	}
+	write_unlock(&r2net_handler_lock);
+}
+EXPORT_SYMBOL_GPL(r2net_unregister_handler_list);
+
+static struct r2net_msg_handler *r2net_handler_get(u32 msg_type, u32 key)
+{
+	struct r2net_msg_handler *nmh;
+
+	read_lock(&r2net_handler_lock);
+	nmh = r2net_handler_tree_lookup(msg_type, key, NULL, NULL);
+	if (nmh)
+		kref_get(&nmh->nh_kref);
+	read_unlock(&r2net_handler_lock);
+
+	return nmh;
+}
+
+/* ------------------------------------------------------------ */
+
+static int r2net_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 r2net_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 r2net_sendpage(struct r2net_sock_container *sc,
+			   void *kmalloced_virt,
+			   size_t size)
+{
+	struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
+	ssize_t ret;
+
+	while (1) {
+		mutex_lock(&sc->sc_send_lock);
+		ret = sc->sc_sock->ops->sendpage(sc->sc_sock,
+					virt_to_page(kmalloced_virt),
+					(long)kmalloced_virt & ~PAGE_MASK,
+					size, MSG_DONTWAIT);
+		mutex_unlock(&sc->sc_send_lock);
+		if (ret == size)
+			break;
+		if (ret == (ssize_t)-EAGAIN) {
+			mlog(0, "sendpage of size %zu to " SC_NODEF_FMT
+			     " returned EAGAIN\n", size, sc->sc_node->nd_name,
+				sc->sc_node->nd_num,
+				&sc->sc_node->nd_ipv4_address,
+				ntohs(sc->sc_node->nd_ipv4_port));
+			cond_resched();
+			continue;
+		}
+		mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT
+		     " failed with %zd\n", size, sc->sc_node->nd_name,
+			sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
+			ntohs(sc->sc_node->nd_ipv4_port), ret);
+		r2net_ensure_shutdown(nn, sc, 0);
+		break;
+	}
+}
+
+static void r2net_init_msg(struct r2net_msg *msg, u16 data_len,
+				u16 msg_type, u32 key)
+{
+	memset(msg, 0, sizeof(struct r2net_msg));
+	msg->magic = cpu_to_be16(R2NET_MSG_MAGIC);
+	msg->data_len = cpu_to_be16(data_len);
+	msg->msg_type = cpu_to_be16(msg_type);
+	msg->sys_status = cpu_to_be32(R2NET_ERR_NONE);
+	msg->status = 0;
+	msg->key = cpu_to_be32(key);
+}
+
+static int r2net_tx_can_proceed(struct r2net_node *nn,
+				struct r2net_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;
+}
+
+/* Get a map of all nodes to which this node is currently connected to */
+void r2net_fill_node_map(unsigned long *map, unsigned bytes)
+{
+	struct r2net_sock_container *sc;
+	int node, ret;
+
+	BUG_ON(bytes < (BITS_TO_LONGS(R2NM_MAX_NODES) * sizeof(unsigned long)));
+
+	memset(map, 0, bytes);
+	for (node = 0; node < R2NM_MAX_NODES; ++node) {
+		r2net_tx_can_proceed(r2net_nn_from_num(node), &sc, &ret);
+		if (!ret) {
+			set_bit(node, map);
+			sc_put(sc);
+		}
+	}
+}
+EXPORT_SYMBOL_GPL(r2net_fill_node_map);
+
+int r2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec,
+			   size_t caller_veclen, u8 target_node, int *status)
+{
+	int ret = 0;
+	struct r2net_msg *msg = NULL;
+	size_t veclen, caller_bytes = 0;
+	struct kvec *vec = NULL;
+	struct r2net_sock_container *sc = NULL;
+	struct r2net_node *nn = r2net_nn_from_num(target_node);
+	struct r2net_status_wait nsw = {
+		.ns_node_item = LIST_HEAD_INIT(nsw.ns_node_item),
+	};
+	struct r2net_send_tracking nst;
+
+	/* this may be a general bug fix */
+	init_waitqueue_head(&nsw.ns_wq);
+
+	r2net_init_nst(&nst, msg_type, key, current, target_node);
+
+	if (r2net_wq == NULL) {
+		mlog(0, "attempt to tx without r2netd 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 > R2NET_MAX_PAYLOAD_BYTES) {
+		mlog(0, "total payload len %zu too large\n", caller_bytes);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (target_node == r2nm_this_node()) {
+		ret = -ELOOP;
+		goto out;
+	}
+
+	r2net_debug_add_nst(&nst);
+
+	r2net_set_nst_sock_time(&nst);
+
+	wait_event(nn->nn_sc_wq, r2net_tx_can_proceed(nn, &sc, &ret));
+	if (ret)
+		goto out;
+
+	r2net_set_nst_sock_container(&nst, sc);
+
+	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 r2net_msg), GFP_ATOMIC);
+	if (!msg) {
+		mlog(0, "failed to allocate a r2net_msg!\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	r2net_init_msg(msg, caller_bytes, msg_type, key);
+
+	vec[0].iov_len = sizeof(struct r2net_msg);
+	vec[0].iov_base = msg;
+	memcpy(&vec[1], caller_vec, caller_veclen * sizeof(struct kvec));
+
+	ret = r2net_prep_nsw(nn, &nsw);
+	if (ret)
+		goto out;
+
+	msg->msg_num = cpu_to_be32(nsw.ns_id);
+	r2net_set_nst_msg_id(&nst, nsw.ns_id);
+
+	r2net_set_nst_send_time(&nst);
+
+	/* finally, convert the message header to network byte-order
+	 * and send */
+	mutex_lock(&sc->sc_send_lock);
+	ret = r2net_send_tcp_msg(sc->sc_sock, vec, veclen,
+				 sizeof(struct r2net_msg) + caller_bytes);
+	mutex_unlock(&sc->sc_send_lock);
+	msglog(msg, "sending returned %d\n", ret);
+	if (ret < 0) {
+		mlog(0, "error returned from r2net_send_tcp_msg=%d\n", ret);
+		goto out;
+	}
+
+	/* wait on other node's handler */
+	r2net_set_nst_status_time(&nst);
+	wait_event(nsw.ns_wq, r2net_nsw_completed(nn, &nsw));
+
+	r2net_update_send_stats(&nst, sc);
+
+	/* Note that we avoid overwriting the callers status return
+	 * variable if a system error was reported on the other
+	 * side. Callers beware. */
+	ret = r2net_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:
+	r2net_debug_del_nst(&nst); /* must be before dropping sc and node */
+	if (sc)
+		sc_put(sc);
+	kfree(vec);
+	kfree(msg);
+	r2net_complete_nsw(nn, &nsw, 0, 0, 0);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(r2net_send_message_vec);
+
+int r2net_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 r2net_send_message_vec(msg_type, key, &vec, 1,
+				      target_node, status);
+}
+EXPORT_SYMBOL_GPL(r2net_send_message);
+
+static int r2net_send_status_magic(struct socket *sock, struct r2net_msg *hdr,
+				   enum r2net_system_error syserr, int err)
+{
+	struct kvec vec = {
+		.iov_base = hdr,
+		.iov_len = sizeof(struct r2net_msg),
+	};
+
+	BUG_ON(syserr >= R2NET_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);
+	/* twiddle the magic */
+	hdr->magic = cpu_to_be16(R2NET_MSG_STATUS_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 r2net_send_tcp_msg(sock, &vec, 1, sizeof(struct r2net_msg));
+}
+
+/*
+ * "data magic" is a long version of "status magic" where the message
+ * payload actually contains data to be passed in reply to certain messages
+ */
+static int r2net_send_data_magic(struct r2net_sock_container *sc,
+			  struct r2net_msg *hdr,
+			  void *data, size_t data_len,
+			  enum r2net_system_error syserr, int err)
+{
+	struct kvec vec[2];
+	int ret;
+
+	vec[0].iov_base = hdr;
+	vec[0].iov_len = sizeof(struct r2net_msg);
+	vec[1].iov_base = data;
+	vec[1].iov_len = data_len;
+
+	BUG_ON(syserr >= R2NET_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(R2NET_MSG_DATA_MAGIC);  /* twiddle magic */
+	hdr->data_len = cpu_to_be16(data_len);
+
+	msglog(hdr, "about to send data magic %d\n", err);
+	/* hdr has been in host byteorder this whole time */
+	ret = r2net_send_tcp_msg(sc->sc_sock, vec, 2,
+			sizeof(struct r2net_msg) + data_len);
+	return ret;
+}
+
+/*
+ * called by a message handler to convert an otherwise normal reply
+ * message into a "data magic" message
+ */
+void r2net_force_data_magic(struct r2net_msg *hdr, u16 msgtype, u32 msgkey)
+{
+	hdr->magic = cpu_to_be16(R2NET_MSG_DATA_MAGIC);
+	hdr->msg_type = cpu_to_be16(msgtype);
+	hdr->key = cpu_to_be32(msgkey);
+}
+
+/* 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 r2net_process_message(struct r2net_sock_container *sc,
+				 struct r2net_msg *hdr)
+{
+	struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
+	int ret = 0, handler_status;
+	enum  r2net_system_error syserr;
+	struct r2net_msg_handler *nmh = NULL;
+	void *ret_data = NULL;
+	int data_magic = 0;
+
+	msglog(hdr, "processing message\n");
+
+	r2net_sc_postpone_idle(sc);
+
+	switch (be16_to_cpu(hdr->magic)) {
+
+	case R2NET_MSG_STATUS_MAGIC:
+		/* special type for returning message status */
+		r2net_complete_nsw(nn, NULL, be32_to_cpu(hdr->msg_num),
+						be32_to_cpu(hdr->sys_status),
+						be32_to_cpu(hdr->status));
+		goto out;
+	case R2NET_MSG_KEEP_REQ_MAGIC:
+		r2net_sendpage(sc, r2net_keep_resp, sizeof(*r2net_keep_resp));
+		goto out;
+	case R2NET_MSG_KEEP_RESP_MAGIC:
+		goto out;
+	case R2NET_MSG_MAGIC:
+		break;
+	case R2NET_MSG_DATA_MAGIC:
+		/*
+		 * unlike a normal status magic, a data magic DOES
+		 * (MUST) have a handler, so the control flow is
+		 * a little funky here as a result
+		 */
+		data_magic = 1;
+		break;
+	default:
+		msglog(hdr, "bad magic\n");
+		ret = -EINVAL;
+		goto out;
+		break;
+	}
+
+	/* find a handler for it */
+	handler_status = 0;
+	nmh = r2net_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 = R2NET_ERR_NO_HNDLR;
+		goto out_respond;
+	}
+
+	syserr = R2NET_ERR_NONE;
+
+	if (be16_to_cpu(hdr->data_len) > nmh->nh_max_len)
+		syserr = R2NET_ERR_OVERFLOW;
+
+	if (syserr != R2NET_ERR_NONE)
+		goto out_respond;
+
+	r2net_set_func_start_time(sc);
+	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 r2net_msg) +
+					     be16_to_cpu(hdr->data_len),
+					nmh->nh_func_data, &ret_data);
+	if (data_magic) {
+		/*
+		 * handler handled data sent in reply to request
+		 * so complete the transaction
+		 */
+		r2net_complete_nsw(nn, NULL, be32_to_cpu(hdr->msg_num),
+			be32_to_cpu(hdr->sys_status), handler_status);
+		goto out;
+	}
+	/*
+	 * handler changed magic to DATA_MAGIC to reply to request for data,
+	 * implies ret_data points to data to return and handler_status
+	 * is the number of bytes of data
+	 */
+	if (be16_to_cpu(hdr->magic) == R2NET_MSG_DATA_MAGIC) {
+		ret = r2net_send_data_magic(sc, hdr,
+						ret_data, handler_status,
+						syserr, 0);
+		hdr = NULL;
+		mlog(0, "sending data reply %d, syserr %d returned %d\n",
+			handler_status, syserr, ret);
+		r2net_set_func_stop_time(sc);
+
+		r2net_update_recv_stats(sc);
+		goto out;
+	}
+	r2net_set_func_stop_time(sc);
+
+	r2net_update_recv_stats(sc);
+
+out_respond:
+	/* this destroys the hdr, so don't use it after this */
+	mutex_lock(&sc->sc_send_lock);
+	ret = r2net_send_status_magic(sc->sc_sock, hdr, syserr,
+				      handler_status);
+	mutex_unlock(&sc->sc_send_lock);
+	hdr = NULL;
+	mlog(0, "sending handler status %d, syserr %d returned %d\n",
+	     handler_status, syserr, ret);
+
+	if (nmh) {
+		BUG_ON(ret_data != NULL && nmh->nh_post_func == NULL);
+		if (nmh->nh_post_func)
+			(nmh->nh_post_func)(handler_status, nmh->nh_func_data,
+					    ret_data);
+	}
+
+out:
+	if (nmh)
+		r2net_handler_put(nmh);
+	return ret;
+}
+
+static int r2net_check_handshake(struct r2net_sock_container *sc)
+{
+	struct r2net_handshake *hand = page_address(sc->sc_page);
+	struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
+
+	if (hand->protocol_version != cpu_to_be64(R2NET_PROTOCOL_VERSION)) {
+		printk(KERN_NOTICE "ramster: " SC_NODEF_FMT " Advertised net "
+		       "protocol version %llu but %llu is required. "
+		       "Disconnecting.\n", sc->sc_node->nd_name,
+			sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
+			ntohs(sc->sc_node->nd_ipv4_port),
+		       (unsigned long long)be64_to_cpu(hand->protocol_version),
+		       R2NET_PROTOCOL_VERSION);
+
+		/* don't bother reconnecting if its the wrong version. */
+		r2net_ensure_shutdown(nn, sc, -ENOTCONN);
+		return -1;
+	}
+
+	/*
+	 * Ensure timeouts are consistent with other nodes, otherwise
+	 * we can end up with one node thinking that the other must be down,
+	 * but isn't. This can ultimately cause corruption.
+	 */
+	if (be32_to_cpu(hand->r2net_idle_timeout_ms) !=
+				r2net_idle_timeout()) {
+		printk(KERN_NOTICE "ramster: " SC_NODEF_FMT " uses a network "
+		       "idle timeout of %u ms, but we use %u ms locally. "
+		       "Disconnecting.\n", sc->sc_node->nd_name,
+			sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
+			ntohs(sc->sc_node->nd_ipv4_port),
+		       be32_to_cpu(hand->r2net_idle_timeout_ms),
+		       r2net_idle_timeout());
+		r2net_ensure_shutdown(nn, sc, -ENOTCONN);
+		return -1;
+	}
+
+	if (be32_to_cpu(hand->r2net_keepalive_delay_ms) !=
+			r2net_keepalive_delay()) {
+		printk(KERN_NOTICE "ramster: " SC_NODEF_FMT " uses a keepalive "
+		       "delay of %u ms, but we use %u ms locally. "
+		       "Disconnecting.\n", sc->sc_node->nd_name,
+			sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
+			ntohs(sc->sc_node->nd_ipv4_port),
+		       be32_to_cpu(hand->r2net_keepalive_delay_ms),
+		       r2net_keepalive_delay());
+		r2net_ensure_shutdown(nn, sc, -ENOTCONN);
+		return -1;
+	}
+
+	if (be32_to_cpu(hand->r2hb_heartbeat_timeout_ms) !=
+			R2HB_MAX_WRITE_TIMEOUT_MS) {
+		printk(KERN_NOTICE "ramster: " SC_NODEF_FMT " uses a heartbeat "
+		       "timeout of %u ms, but we use %u ms locally. "
+		       "Disconnecting.\n", sc->sc_node->nd_name,
+			sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
+			ntohs(sc->sc_node->nd_ipv4_port),
+		       be32_to_cpu(hand->r2hb_heartbeat_timeout_ms),
+		       R2HB_MAX_WRITE_TIMEOUT_MS);
+		r2net_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) {
+		r2net_sc_reset_idle_timer(sc);
+		atomic_set(&nn->nn_timeout, 0);
+		r2net_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 r2net_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 r2net_advance_rx(struct r2net_sock_container *sc)
+{
+	struct r2net_msg *hdr;
+	int ret = 0;
+	void *data;
+	size_t datalen;
+
+	sclog(sc, "receiving\n");
+	r2net_set_advance_start_time(sc);
+
+	if (unlikely(sc->sc_handshake_ok == 0)) {
+		if (sc->sc_page_off < sizeof(struct r2net_handshake)) {
+			data = page_address(sc->sc_page) + sc->sc_page_off;
+			datalen = sizeof(struct r2net_handshake) -
+							sc->sc_page_off;
+			ret = r2net_recv_tcp_msg(sc->sc_sock, data, datalen);
+			if (ret > 0)
+				sc->sc_page_off += ret;
+		}
+
+		if (sc->sc_page_off == sizeof(struct r2net_handshake)) {
+			r2net_check_handshake(sc);
+			if (unlikely(sc->sc_handshake_ok == 0))
+				ret = -EPROTO;
+		}
+		goto out;
+	}
+
+	/* do we need more header? */
+	if (sc->sc_page_off < sizeof(struct r2net_msg)) {
+		data = page_address(sc->sc_page) + sc->sc_page_off;
+		datalen = sizeof(struct r2net_msg) - sc->sc_page_off;
+		ret = r2net_recv_tcp_msg(sc->sc_sock, data, datalen);
+		if (ret > 0) {
+			sc->sc_page_off += ret;
+			/* 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 r2net_msg)) {
+				hdr = page_address(sc->sc_page);
+				if (be16_to_cpu(hdr->data_len) >
+				    R2NET_MAX_PAYLOAD_BYTES)
+					ret = -EOVERFLOW;
+			}
+		}
+		if (ret <= 0)
+			goto out;
+	}
+
+	if (sc->sc_page_off < sizeof(struct r2net_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 r2net_msg) <
+					be16_to_cpu(hdr->data_len)) {
+		/* need more payload */
+		data = page_address(sc->sc_page) + sc->sc_page_off;
+		datalen = (sizeof(struct r2net_msg) +
+				be16_to_cpu(hdr->data_len)) -
+				sc->sc_page_off;
+		ret = r2net_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 r2net_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 = r2net_process_message(sc, hdr);
+		if (ret == 0)
+			ret = 1;
+		sc->sc_page_off = 0;
+	}
+
+out:
+	sclog(sc, "ret = %d\n", ret);
+	r2net_set_advance_stop_time(sc);
+	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 r2net_rx_until_empty(struct work_struct *work)
+{
+	struct r2net_sock_container *sc =
+		container_of(work, struct r2net_sock_container, sc_rx_work);
+	int ret;
+
+	do {
+		ret = r2net_advance_rx(sc);
+	} while (ret > 0);
+
+	if (ret <= 0 && ret != -EAGAIN) {
+		struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
+		sclog(sc, "saw error %d, closing\n", ret);
+		/* not permanent so read failed handshake can retry */
+		r2net_ensure_shutdown(nn, sc, 0);
+	}
+
+	sc_put(sc);
+}
+
+static int r2net_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;
+}
+
+static void r2net_initialize_handshake(void)
+{
+	r2net_hand->r2hb_heartbeat_timeout_ms = cpu_to_be32(
+		R2HB_MAX_WRITE_TIMEOUT_MS);
+	r2net_hand->r2net_idle_timeout_ms = cpu_to_be32(r2net_idle_timeout());
+	r2net_hand->r2net_keepalive_delay_ms = cpu_to_be32(
+		r2net_keepalive_delay());
+	r2net_hand->r2net_reconnect_delay_ms = cpu_to_be32(
+		r2net_reconnect_delay());
+}
+
+/* ------------------------------------------------------------ */
+
+/* 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 r2net_sc_connect_completed(struct work_struct *work)
+{
+	struct r2net_sock_container *sc =
+			container_of(work, struct r2net_sock_container,
+			     sc_connect_work);
+
+	mlog(ML_MSG, "sc sending handshake with ver %llu id %llx\n",
+		(unsigned long long)R2NET_PROTOCOL_VERSION,
+		(unsigned long long)be64_to_cpu(r2net_hand->connector_id));
+
+	r2net_initialize_handshake();
+	r2net_sendpage(sc, r2net_hand, sizeof(*r2net_hand));
+	sc_put(sc);
+}
+
+/* this is called as a work_struct func. */
+static void r2net_sc_send_keep_req(struct work_struct *work)
+{
+	struct r2net_sock_container *sc =
+		container_of(work, struct r2net_sock_container,
+			     sc_keepalive_work.work);
+
+	r2net_sendpage(sc, r2net_keep_req, sizeof(*r2net_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 r2net_idle_timer(unsigned long data)
+{
+	struct r2net_sock_container *sc = (struct r2net_sock_container *)data;
+#ifdef CONFIG_DEBUG_FS
+	unsigned long msecs = ktime_to_ms(ktime_get()) -
+		ktime_to_ms(sc->sc_tv_timer);
+#else
+	unsigned long msecs = r2net_idle_timeout();
+#endif
+
+	printk(KERN_NOTICE "ramster: Connection to " SC_NODEF_FMT " has been "
+	       "idle for %lu.%lu secs, shutting it down.\n",
+		sc->sc_node->nd_name, sc->sc_node->nd_num,
+		&sc->sc_node->nd_ipv4_address, ntohs(sc->sc_node->nd_ipv4_port),
+	       msecs / 1000, msecs % 1000);
+
+	/*
+	 * Initialize the nn_timeout so that the next connection attempt
+	 * will continue in r2net_start_connect.
+	 */
+	/* Avoid spurious shutdowns... not sure if this is still necessary */
+	pr_err("ramster_idle_timer, skipping shutdown work\n");
+#if 0
+	/* old code used to do these two lines */
+	atomic_set(&nn->nn_timeout, 1);
+	r2net_sc_queue_work(sc, &sc->sc_shutdown_work);
+#endif
+}
+
+static void r2net_sc_reset_idle_timer(struct r2net_sock_container *sc)
+{
+	r2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work);
+	r2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work,
+		      msecs_to_jiffies(r2net_keepalive_delay()));
+	r2net_set_sock_timer(sc);
+	mod_timer(&sc->sc_idle_timeout,
+	       jiffies + msecs_to_jiffies(r2net_idle_timeout()));
+}
+
+static void r2net_sc_postpone_idle(struct r2net_sock_container *sc)
+{
+	/* Only push out an existing timer */
+	if (timer_pending(&sc->sc_idle_timeout))
+		r2net_sc_reset_idle_timer(sc);
+}
+
+/* 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 r2net_start_connect(struct work_struct *work)
+{
+	struct r2net_node *nn =
+		container_of(work, struct r2net_node, nn_connect_work.work);
+	struct r2net_sock_container *sc = NULL;
+	struct r2nm_node *node = NULL, *mynode = NULL;
+	struct socket *sock = NULL;
+	struct sockaddr_in myaddr = {0, }, remoteaddr = {0, };
+	int ret = 0, stop;
+	unsigned int timeout;
+
+	/* if we're greater we initiate tx, otherwise we accept */
+	if (r2nm_this_node() <= r2net_num_from_nn(nn))
+		goto out;
+
+	/* watch for racing with tearing a node down */
+	node = r2nm_get_node_by_num(r2net_num_from_nn(nn));
+	if (node == NULL) {
+		ret = 0;
+		goto out;
+	}
+
+	mynode = r2nm_get_node_by_num(r2nm_this_node());
+	if (mynode == NULL) {
+		ret = 0;
+		goto out;
+	}
+
+	spin_lock(&nn->nn_lock);
+	/*
+	 * see if we already have one pending or have given up.
+	 * For nn_timeout, it is set when we close the connection
+	 * because of the idle time out. So it means that we have
+	 * at least connected to that node successfully once,
+	 * now try to connect to it again.
+	 */
+	timeout = atomic_read(&nn->nn_timeout);
+	stop = (nn->nn_sc ||
+		(nn->nn_persistent_error &&
+		(nn->nn_persistent_error != -ENOTCONN || timeout == 0)));
+	spin_unlock(&nn->nn_lock);
+	if (stop)
+		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_addr.s_addr = mynode->nd_ipv4_address;
+	myaddr.sin_port = htons(0); /* any port */
+
+	ret = sock->ops->bind(sock, (struct sockaddr *)&myaddr,
+			      sizeof(myaddr));
+	if (ret) {
+		mlog(ML_ERROR, "bind failed with %d at address %pI4\n",
+		     ret, &mynode->nd_ipv4_address);
+		goto out;
+	}
+
+	ret = r2net_set_nodelay(sc->sc_sock);
+	if (ret) {
+		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
+		goto out;
+	}
+
+	r2net_register_callbacks(sc->sc_sock->sk, sc);
+
+	spin_lock(&nn->nn_lock);
+	/* handshake completion will set nn->nn_sc_valid */
+	r2net_set_nn_state(nn, sc, 0, 0);
+	spin_unlock(&nn->nn_lock);
+
+	remoteaddr.sin_family = AF_INET;
+	remoteaddr.sin_addr.s_addr = node->nd_ipv4_address;
+	remoteaddr.sin_port = 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) {
+		printk(KERN_NOTICE "ramster: Connect attempt to " SC_NODEF_FMT
+		       " failed with errno %d\n", sc->sc_node->nd_name,
+			sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
+			ntohs(sc->sc_node->nd_ipv4_port), ret);
+		/* 0 err so that another will be queued and attempted
+		 * from set_nn_state */
+		if (sc)
+			r2net_ensure_shutdown(nn, sc, 0);
+	}
+	if (sc)
+		sc_put(sc);
+	if (node)
+		r2nm_node_put(node);
+	if (mynode)
+		r2nm_node_put(mynode);
+
+	return;
+}
+
+static void r2net_connect_expired(struct work_struct *work)
+{
+	struct r2net_node *nn =
+		container_of(work, struct r2net_node, nn_connect_expired.work);
+
+	spin_lock(&nn->nn_lock);
+	if (!nn->nn_sc_valid) {
+		printk(KERN_NOTICE "ramster: No connection established with "
+		       "node %u after %u.%u seconds, giving up.\n",
+		     r2net_num_from_nn(nn),
+		     r2net_idle_timeout() / 1000,
+		     r2net_idle_timeout() % 1000);
+
+		r2net_set_nn_state(nn, NULL, 0, -ENOTCONN);
+	}
+	spin_unlock(&nn->nn_lock);
+}
+
+static void r2net_still_up(struct work_struct *work)
+{
+}
+
+/* ------------------------------------------------------------ */
+
+void r2net_disconnect_node(struct r2nm_node *node)
+{
+	struct r2net_node *nn = r2net_nn_from_num(node->nd_num);
+
+	/* don't reconnect until it's heartbeating again */
+	spin_lock(&nn->nn_lock);
+	atomic_set(&nn->nn_timeout, 0);
+	r2net_set_nn_state(nn, NULL, 0, -ENOTCONN);
+	spin_unlock(&nn->nn_lock);
+
+	if (r2net_wq) {
+		cancel_delayed_work(&nn->nn_connect_expired);
+		cancel_delayed_work(&nn->nn_connect_work);
+		cancel_delayed_work(&nn->nn_still_up);
+		flush_workqueue(r2net_wq);
+	}
+}
+
+static void r2net_hb_node_down_cb(struct r2nm_node *node, int node_num,
+				  void *data)
+{
+	if (!node)
+		return;
+
+	if (node_num != r2nm_this_node())
+		r2net_disconnect_node(node);
+
+	BUG_ON(atomic_read(&r2net_connected_peers) < 0);
+}
+
+static void r2net_hb_node_up_cb(struct r2nm_node *node, int node_num,
+				void *data)
+{
+	struct r2net_node *nn = r2net_nn_from_num(node_num);
+
+	BUG_ON(!node);
+
+	/* ensure an immediate connect attempt */
+	nn->nn_last_connect_attempt = jiffies -
+		(msecs_to_jiffies(r2net_reconnect_delay()) + 1);
+
+	if (node_num != r2nm_this_node()) {
+		/* 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);
+		atomic_set(&nn->nn_timeout, 0);
+		if (nn->nn_persistent_error)
+			r2net_set_nn_state(nn, NULL, 0, 0);
+		spin_unlock(&nn->nn_lock);
+	}
+}
+
+void r2net_unregister_hb_callbacks(void)
+{
+	r2hb_unregister_callback(NULL, &r2net_hb_up);
+	r2hb_unregister_callback(NULL, &r2net_hb_down);
+}
+
+int r2net_register_hb_callbacks(void)
+{
+	int ret;
+
+	r2hb_setup_callback(&r2net_hb_down, R2HB_NODE_DOWN_CB,
+			    r2net_hb_node_down_cb, NULL, R2NET_HB_PRI);
+	r2hb_setup_callback(&r2net_hb_up, R2HB_NODE_UP_CB,
+			    r2net_hb_node_up_cb, NULL, R2NET_HB_PRI);
+
+	ret = r2hb_register_callback(NULL, &r2net_hb_up);
+	if (ret == 0)
+		ret = r2hb_register_callback(NULL, &r2net_hb_down);
+
+	if (ret)
+		r2net_unregister_hb_callbacks();
+
+	return ret;
+}
+
+/* ------------------------------------------------------------ */
+
+static int r2net_accept_one(struct socket *sock)
+{
+	int ret, slen;
+	struct sockaddr_in sin;
+	struct socket *new_sock = NULL;
+	struct r2nm_node *node = NULL;
+	struct r2nm_node *local_node = NULL;
+	struct r2net_sock_container *sc = NULL;
+	struct r2net_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 = r2net_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 = r2nm_get_node_by_ip(sin.sin_addr.s_addr);
+	if (node == NULL) {
+		printk(KERN_NOTICE "ramster: Attempt to connect from unknown "
+		       "node at %pI4:%d\n", &sin.sin_addr.s_addr,
+		       ntohs(sin.sin_port));
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (r2nm_this_node() >= node->nd_num) {
+		local_node = r2nm_get_node_by_num(r2nm_this_node());
+		printk(KERN_NOTICE "ramster: Unexpected connect attempt seen "
+		       "at node '%s' (%u, %pI4:%d) from node '%s' (%u, "
+		       "%pI4:%d)\n", local_node->nd_name, local_node->nd_num,
+		       &(local_node->nd_ipv4_address),
+		       ntohs(local_node->nd_ipv4_port), node->nd_name,
+		       node->nd_num, &sin.sin_addr.s_addr, ntohs(sin.sin_port));
+		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 (!r2hb_check_node_heartbeating_from_callback(node->nd_num)) {
+		mlog(ML_CONN, "attempt to connect from node '%s' at "
+		     "%pI4:%d but it isn't heartbeating\n",
+		     node->nd_name, &sin.sin_addr.s_addr,
+		     ntohs(sin.sin_port));
+		ret = -EINVAL;
+		goto out;
+	}
+
+	nn = r2net_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) {
+		printk(KERN_NOTICE "ramster: Attempt to connect from node '%s' "
+		       "at %pI4:%d but it already has an open connection\n",
+		       node->nd_name, &sin.sin_addr.s_addr,
+		       ntohs(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);
+	atomic_set(&nn->nn_timeout, 0);
+	r2net_set_nn_state(nn, sc, 0, 0);
+	spin_unlock(&nn->nn_lock);
+
+	r2net_register_callbacks(sc->sc_sock->sk, sc);
+	r2net_sc_queue_work(sc, &sc->sc_rx_work);
+
+	r2net_initialize_handshake();
+	r2net_sendpage(sc, r2net_hand, sizeof(*r2net_hand));
+
+out:
+	if (new_sock)
+		sock_release(new_sock);
+	if (node)
+		r2nm_node_put(node);
+	if (local_node)
+		r2nm_node_put(local_node);
+	if (sc)
+		sc_put(sc);
+	return ret;
+}
+
+static void r2net_accept_many(struct work_struct *work)
+{
+	struct socket *sock = r2net_listen_sock;
+	while (r2net_accept_one(sock) == 0)
+		cond_resched();
+}
+
+static void r2net_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(r2net_wq, &r2net_listen_work);
+	}
+
+out:
+	read_unlock(&sk->sk_callback_lock);
+	ready(sk, bytes);
+}
+
+static int r2net_open_listening_sock(__be32 addr, __be16 port)
+{
+	struct socket *sock = NULL;
+	int ret;
+	struct sockaddr_in sin = {
+		.sin_family = PF_INET,
+		.sin_addr = { .s_addr = addr },
+		.sin_port = port,
+	};
+
+	ret = sock_create(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
+	if (ret < 0) {
+		printk(KERN_ERR "ramster: Error %d while creating socket\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 = r2net_listen_data_ready;
+	write_unlock_bh(&sock->sk->sk_callback_lock);
+
+	r2net_listen_sock = sock;
+	INIT_WORK(&r2net_listen_work, r2net_accept_many);
+
+	sock->sk->sk_reuse = 1;
+	ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
+	if (ret < 0) {
+		printk(KERN_ERR "ramster: Error %d while binding socket at "
+			"%pI4:%u\n", ret, &addr, ntohs(port));
+		goto out;
+	}
+
+	ret = sock->ops->listen(sock, 64);
+	if (ret < 0)
+		printk(KERN_ERR "ramster: Error %d while listening on %pI4:%u\n",
+		       ret, &addr, ntohs(port));
+
+out:
+	if (ret) {
+		r2net_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 r2net_stop_listening().  note,
+ * r2nm_this_node() doesn't work yet as we're being called while it
+ * is being set up.
+ */
+int r2net_start_listening(struct r2nm_node *node)
+{
+	int ret = 0;
+
+	BUG_ON(r2net_wq != NULL);
+	BUG_ON(r2net_listen_sock != NULL);
+
+	mlog(ML_KTHREAD, "starting r2net thread...\n");
+	r2net_wq = create_singlethread_workqueue("r2net");
+	if (r2net_wq == NULL) {
+		mlog(ML_ERROR, "unable to launch r2net thread\n");
+		return -ENOMEM; /* ? */
+	}
+
+	ret = r2net_open_listening_sock(node->nd_ipv4_address,
+					node->nd_ipv4_port);
+	if (ret) {
+		destroy_workqueue(r2net_wq);
+		r2net_wq = NULL;
+	}
+
+	return ret;
+}
+
+/* again, r2nm_this_node() doesn't work here as we're involved in
+ * tearing it down */
+void r2net_stop_listening(struct r2nm_node *node)
+{
+	struct socket *sock = r2net_listen_sock;
+	size_t i;
+
+	BUG_ON(r2net_wq == NULL);
+	BUG_ON(r2net_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(r2net_nodes); i++) {
+		struct r2nm_node *node = r2nm_get_node_by_num(i);
+		if (node) {
+			r2net_disconnect_node(node);
+			r2nm_node_put(node);
+		}
+	}
+
+	/* finish all work and tear down the work queue */
+	mlog(ML_KTHREAD, "waiting for r2net thread to exit....\n");
+	destroy_workqueue(r2net_wq);
+	r2net_wq = NULL;
+
+	sock_release(r2net_listen_sock);
+	r2net_listen_sock = NULL;
+}
+
+void r2net_hb_node_up_manual(int node_num)
+{
+	struct r2nm_node dummy;
+	if (r2nm_single_cluster == NULL)
+		pr_err("ramster: cluster not alive, node_up_manual ignored\n");
+	else {
+		r2hb_manual_set_node_heartbeating(node_num);
+		r2net_hb_node_up_cb(&dummy, node_num, NULL);
+	}
+}
+
+/* ------------------------------------------------------------ */
+
+int r2net_init(void)
+{
+	unsigned long i;
+
+	if (r2net_debugfs_init())
+		return -ENOMEM;
+
+	r2net_hand = kzalloc(sizeof(struct r2net_handshake), GFP_KERNEL);
+	r2net_keep_req = kzalloc(sizeof(struct r2net_msg), GFP_KERNEL);
+	r2net_keep_resp = kzalloc(sizeof(struct r2net_msg), GFP_KERNEL);
+	if (!r2net_hand || !r2net_keep_req || !r2net_keep_resp) {
+		kfree(r2net_hand);
+		kfree(r2net_keep_req);
+		kfree(r2net_keep_resp);
+		return -ENOMEM;
+	}
+
+	r2net_hand->protocol_version = cpu_to_be64(R2NET_PROTOCOL_VERSION);
+	r2net_hand->connector_id = cpu_to_be64(1);
+
+	r2net_keep_req->magic = cpu_to_be16(R2NET_MSG_KEEP_REQ_MAGIC);
+	r2net_keep_resp->magic = cpu_to_be16(R2NET_MSG_KEEP_RESP_MAGIC);
+
+	for (i = 0; i < ARRAY_SIZE(r2net_nodes); i++) {
+		struct r2net_node *nn = r2net_nn_from_num(i);
+
+		atomic_set(&nn->nn_timeout, 0);
+		spin_lock_init(&nn->nn_lock);
+		INIT_DELAYED_WORK(&nn->nn_connect_work, r2net_start_connect);
+		INIT_DELAYED_WORK(&nn->nn_connect_expired,
+				  r2net_connect_expired);
+		INIT_DELAYED_WORK(&nn->nn_still_up, r2net_still_up);
+		/* 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 r2net_exit(void)
+{
+	kfree(r2net_hand);
+	kfree(r2net_keep_req);
+	kfree(r2net_keep_resp);
+	r2net_debugfs_exit();
+}
diff --git a/drivers/staging/ramster/cluster/tcp.h b/drivers/staging/ramster/cluster/tcp.h
new file mode 100644
index 0000000..9d05833
--- /dev/null
+++ b/drivers/staging/ramster/cluster/tcp.h
@@ -0,0 +1,159 @@
+/* -*- 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 R2CLUSTER_TCP_H
+#define R2CLUSTER_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 r2net_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 (r2net_msg_handler_func)(struct r2net_msg *msg, u32 len, void *data,
+				     void **ret_data);
+typedef void (r2net_post_msg_handler_func)(int status, void *data,
+					   void *ret_data);
+
+#define R2NET_MAX_PAYLOAD_BYTES  (4096 - sizeof(struct r2net_msg))
+
+/* same as hb delay, we're waiting for another node to recognize our hb */
+#define R2NET_RECONNECT_DELAY_MS_DEFAULT	2000
+
+#define R2NET_KEEPALIVE_DELAY_MS_DEFAULT	2000
+#define R2NET_IDLE_TIMEOUT_MS_DEFAULT		30000
+
+
+/* TODO: figure this out.... */
+static inline int r2net_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 {
+	R2NET_DRIVER_UNINITED,
+	R2NET_DRIVER_READY,
+};
+
+int r2net_send_message(u32 msg_type, u32 key, void *data, u32 len,
+		       u8 target_node, int *status);
+int r2net_send_message_vec(u32 msg_type, u32 key, struct kvec *vec,
+			   size_t veclen, u8 target_node, int *status);
+
+int r2net_register_handler(u32 msg_type, u32 key, u32 max_len,
+			   r2net_msg_handler_func *func, void *data,
+			   r2net_post_msg_handler_func *post_func,
+			   struct list_head *unreg_list);
+void r2net_unregister_handler_list(struct list_head *list);
+
+void r2net_fill_node_map(unsigned long *map, unsigned bytes);
+
+void r2net_force_data_magic(struct r2net_msg *, u16, u32);
+void r2net_hb_node_up_manual(int);
+struct r2net_node *r2net_nn_from_num(u8);
+
+struct r2nm_node;
+int r2net_register_hb_callbacks(void);
+void r2net_unregister_hb_callbacks(void);
+int r2net_start_listening(struct r2nm_node *node);
+void r2net_stop_listening(struct r2nm_node *node);
+void r2net_disconnect_node(struct r2nm_node *node);
+int r2net_num_connected_peers(void);
+
+int r2net_init(void);
+void r2net_exit(void);
+
+struct r2net_send_tracking;
+struct r2net_sock_container;
+
+#if 0
+int r2net_debugfs_init(void);
+void r2net_debugfs_exit(void);
+void r2net_debug_add_nst(struct r2net_send_tracking *nst);
+void r2net_debug_del_nst(struct r2net_send_tracking *nst);
+void r2net_debug_add_sc(struct r2net_sock_container *sc);
+void r2net_debug_del_sc(struct r2net_sock_container *sc);
+#else
+static inline int r2net_debugfs_init(void)
+{
+	return 0;
+}
+static inline void r2net_debugfs_exit(void)
+{
+}
+static inline void r2net_debug_add_nst(struct r2net_send_tracking *nst)
+{
+}
+static inline void r2net_debug_del_nst(struct r2net_send_tracking *nst)
+{
+}
+static inline void r2net_debug_add_sc(struct r2net_sock_container *sc)
+{
+}
+static inline void r2net_debug_del_sc(struct r2net_sock_container *sc)
+{
+}
+#endif	/* CONFIG_DEBUG_FS */
+
+#endif /* R2CLUSTER_TCP_H */
diff --git a/drivers/staging/ramster/cluster/tcp_internal.h b/drivers/staging/ramster/cluster/tcp_internal.h
new file mode 100644
index 0000000..4d8cc9f
--- /dev/null
+++ b/drivers/staging/ramster/cluster/tcp_internal.h
@@ -0,0 +1,248 @@
+/* -*- 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 R2CLUSTER_TCP_INTERNAL_H
+#define R2CLUSTER_TCP_INTERNAL_H
+
+#define R2NET_MSG_MAGIC           ((u16)0xfa55)
+#define R2NET_MSG_STATUS_MAGIC    ((u16)0xfa56)
+#define R2NET_MSG_KEEP_REQ_MAGIC  ((u16)0xfa57)
+#define R2NET_MSG_KEEP_RESP_MAGIC ((u16)0xfa58)
+/*
+ * "data magic" is a long version of "status magic" where the message
+ * payload actually contains data to be passed in reply to certain messages
+ */
+#define R2NET_MSG_DATA_MAGIC      ((u16)0xfa59)
+
+/* 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 R2NET_QUORUM_DELAY_MS	\
+		((r2hb_dead_threshold + 2) * R2HB_REGION_TIMEOUT_MS)
+
+/*
+ * 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.
+ *
+ * With version 11, we separate out the filesystem locking portion.  The
+ * filesystem now has a major.minor version it negotiates.  Version 11
+ * introduces this negotiation to the r2dlm protocol, and as such the
+ * version here in tcp_internal.h should not need to be bumped for
+ * filesystem locking changes.
+ *
+ * New in version 11
+ *	- Negotiation of filesystem locking in the dlm join.
+ *
+ * New in version 10:
+ *	- Meta/data locks combined
+ *
+ * New in version 9:
+ *	- All votes removed
+ *
+ * New in version 8:
+ *	- Replace delete inode votes with a cluster lock
+ *
+ * New in version 7:
+ *	- DLM join domain includes the live nodemap
+ *
+ * New in version 6:
+ *	- DLM lockres remote refcount fixes.
+ *
+ * New in version 5:
+ *	- Network timeout checking protocol
+ *
+ * New in version 4:
+ *	- Remove i_generation from lock names for better stat performance.
+ *
+ * New in version 3:
+ *	- Replace dentry votes with a cluster lock
+ *
+ * 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 R2NET_PROTOCOL_VERSION 11ULL
+struct r2net_handshake {
+	__be64	protocol_version;
+	__be64	connector_id;
+	__be32  r2hb_heartbeat_timeout_ms;
+	__be32  r2net_idle_timeout_ms;
+	__be32  r2net_keepalive_delay_ms;
+	__be32  r2net_reconnect_delay_ms;
+};
+
+struct r2net_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 r2net_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;
+	/* It is only set to 1 after the idle time out. */
+	atomic_t			nn_timeout;
+
+	/* 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 R2NET_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 delayed_work		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 delayed_work		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 delayed_work		nn_still_up;
+};
+
+struct r2net_sock_container {
+	struct kref		sc_kref;
+	/* the next two are valid for the life time of the sc */
+	struct socket		*sc_sock;
+	struct r2nm_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 r2net_wq destruction in r2net_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 delayed_work	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);
+
+	u32			sc_msg_key;
+	u16			sc_msg_type;
+
+#ifdef CONFIG_DEBUG_FS
+	struct list_head        sc_net_debug_item;
+	ktime_t			sc_tv_timer;
+	ktime_t			sc_tv_data_ready;
+	ktime_t			sc_tv_advance_start;
+	ktime_t			sc_tv_advance_stop;
+	ktime_t			sc_tv_func_start;
+	ktime_t			sc_tv_func_stop;
+#endif
+#ifdef CONFIG_RAMSTER_FS_STATS
+	ktime_t			sc_tv_acquiry_total;
+	ktime_t			sc_tv_send_total;
+	ktime_t			sc_tv_status_total;
+	u32			sc_send_count;
+	u32			sc_recv_count;
+	ktime_t			sc_tv_process_total;
+#endif
+	struct mutex		sc_send_lock;
+};
+
+struct r2net_msg_handler {
+	struct rb_node		nh_node;
+	u32			nh_max_len;
+	u32			nh_msg_type;
+	u32			nh_key;
+	r2net_msg_handler_func	*nh_func;
+	r2net_msg_handler_func	*nh_func_data;
+	r2net_post_msg_handler_func
+				*nh_post_func;
+	struct kref		nh_kref;
+	struct list_head	nh_unregister_item;
+};
+
+enum r2net_system_error {
+	R2NET_ERR_NONE = 0,
+	R2NET_ERR_NO_HNDLR,
+	R2NET_ERR_OVERFLOW,
+	R2NET_ERR_DIED,
+	R2NET_ERR_MAX
+};
+
+struct r2net_status_wait {
+	enum r2net_system_error	ns_sys_status;
+	s32			ns_status;
+	int			ns_id;
+	wait_queue_head_t	ns_wq;
+	struct list_head	ns_node_item;
+};
+
+#ifdef CONFIG_DEBUG_FS
+/* just for state dumps */
+struct r2net_send_tracking {
+	struct list_head		st_net_debug_item;
+	struct task_struct		*st_task;
+	struct r2net_sock_container	*st_sc;
+	u32				st_id;
+	u32				st_msg_type;
+	u32				st_msg_key;
+	u8				st_node;
+	ktime_t				st_sock_time;
+	ktime_t				st_send_time;
+	ktime_t				st_status_time;
+};
+#else
+struct r2net_send_tracking {
+	u32	dummy;
+};
+#endif	/* CONFIG_DEBUG_FS */
+
+#endif /* R2CLUSTER_TCP_INTERNAL_H */
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH 2/6] staging: ramster: local compression + tmem
  2012-02-15 15:54 [PATCH V5r1 0/6] staging: ramster: multi-machine memory capacity management Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 1/6] staging: ramster: cluster/messaging foundation Dan Magenheimer
@ 2012-02-15 15:54 ` Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 3/6] staging: ramster: xvmalloc allocation files Dan Magenheimer
                   ` (3 subsequent siblings)
  5 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-02-15 15:54 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, dan.magenheimer

RAMster implements peer-to-peer transcendent memory, allowing a "cluster"
of kernels to dynamically pool their RAM.

This patch copies files from drivers/staging/zcache.  RAMster compresses
pages locally before transmitting them to another node, so we can
leverage the zcache and tmem code directly.  Note: there are
no ramster-specific changes yet to these files.

(Why copy?  The ramster tmem.c/tmem.h changes are definitely shareable
between zcache and ramster; the eventual destination for tmem.c
is the linux lib directory.  Ramster changes to zcache are more substantial
and zcache is currently undergoing some significant unrelated changes
(including a new allocator and breaking zcache-main.c into smaller files),
so it seemed best to branch temporarily and merge later.)

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/ramster/Kconfig       |   13 +
 drivers/staging/ramster/Makefile      |    3 +
 drivers/staging/ramster/tmem.c        |  770 +++++++++++++
 drivers/staging/ramster/tmem.h        |  206 ++++
 drivers/staging/ramster/zcache-main.c | 1988 +++++++++++++++++++++++++++++++++
 5 files changed, 2980 insertions(+), 0 deletions(-)
 create mode 100644 drivers/staging/ramster/Kconfig
 create mode 100644 drivers/staging/ramster/Makefile
 create mode 100644 drivers/staging/ramster/tmem.c
 create mode 100644 drivers/staging/ramster/tmem.h
 create mode 100644 drivers/staging/ramster/zcache-main.c

diff --git a/drivers/staging/ramster/Kconfig b/drivers/staging/ramster/Kconfig
new file mode 100644
index 0000000..7fabcb2
--- /dev/null
+++ b/drivers/staging/ramster/Kconfig
@@ -0,0 +1,13 @@
+config ZCACHE
+	tristate "Dynamic compression of swap pages and clean pagecache pages"
+	depends on CLEANCACHE || FRONTSWAP
+	select XVMALLOC
+	select LZO_COMPRESS
+	select LZO_DECOMPRESS
+	default n
+	help
+	  Zcache doubles RAM efficiency while providing a significant
+	  performance boosts on many workloads.  Zcache uses lzo1x
+	  compression and an in-kernel implementation of transcendent
+	  memory to store clean page cache pages and swap in RAM,
+	  providing a noticeable reduction in disk I/O.
diff --git a/drivers/staging/ramster/Makefile b/drivers/staging/ramster/Makefile
new file mode 100644
index 0000000..60daa27
--- /dev/null
+++ b/drivers/staging/ramster/Makefile
@@ -0,0 +1,3 @@
+zcache-y	:=	zcache-main.o tmem.o
+
+obj-$(CONFIG_ZCACHE)	+=	zcache.o
diff --git a/drivers/staging/ramster/tmem.c b/drivers/staging/ramster/tmem.c
new file mode 100644
index 0000000..1ca66ea
--- /dev/null
+++ b/drivers/staging/ramster/tmem.c
@@ -0,0 +1,770 @@
+/*
+ * In-kernel transcendent memory (generic implementation)
+ *
+ * Copyright (c) 2009-2011, Dan Magenheimer, Oracle Corp.
+ *
+ * The primary purpose of Transcedent Memory ("tmem") is to map object-oriented
+ * "handles" (triples containing a pool id, and object id, and an index), to
+ * pages in a page-accessible memory (PAM).  Tmem references the PAM pages via
+ * an abstract "pampd" (PAM page-descriptor), which can be operated on by a
+ * set of functions (pamops).  Each pampd contains some representation of
+ * PAGE_SIZE bytes worth of data. Tmem must support potentially millions of
+ * pages and must be able to insert, find, and delete these pages at a
+ * potential frequency of thousands per second concurrently across many CPUs,
+ * (and, if used with KVM, across many vcpus across many guests).
+ * Tmem is tracked with a hierarchy of data structures, organized by
+ * the elements in a handle-tuple: pool_id, object_id, and page index.
+ * One or more "clients" (e.g. guests) each provide one or more tmem_pools.
+ * Each pool, contains a hash table of rb_trees of tmem_objs.  Each
+ * tmem_obj contains a radix-tree-like tree of pointers, with intermediate
+ * nodes called tmem_objnodes.  Each leaf pointer in this tree points to
+ * a pampd, which is accessible only through a small set of callbacks
+ * registered by the PAM implementation (see tmem_register_pamops). Tmem
+ * does all memory allocation via a set of callbacks registered by the tmem
+ * host implementation (e.g. see tmem_register_hostops).
+ */
+
+#include <linux/list.h>
+#include <linux/spinlock.h>
+#include <linux/atomic.h>
+
+#include "tmem.h"
+
+/* data structure sentinels used for debugging... see tmem.h */
+#define POOL_SENTINEL 0x87658765
+#define OBJ_SENTINEL 0x12345678
+#define OBJNODE_SENTINEL 0xfedcba09
+
+/*
+ * A tmem host implementation must use this function to register callbacks
+ * for memory allocation.
+ */
+static struct tmem_hostops tmem_hostops;
+
+static void tmem_objnode_tree_init(void);
+
+void tmem_register_hostops(struct tmem_hostops *m)
+{
+	tmem_objnode_tree_init();
+	tmem_hostops = *m;
+}
+
+/*
+ * A tmem host implementation must use this function to register
+ * callbacks for a page-accessible memory (PAM) implementation
+ */
+static struct tmem_pamops tmem_pamops;
+
+void tmem_register_pamops(struct tmem_pamops *m)
+{
+	tmem_pamops = *m;
+}
+
+/*
+ * Oid's are potentially very sparse and tmem_objs may have an indeterminately
+ * short life, being added and deleted at a relatively high frequency.
+ * So an rb_tree is an ideal data structure to manage tmem_objs.  But because
+ * of the potentially huge number of tmem_objs, each pool manages a hashtable
+ * of rb_trees to reduce search, insert, delete, and rebalancing time.
+ * Each hashbucket also has a lock to manage concurrent access.
+ *
+ * The following routines manage tmem_objs.  When any tmem_obj is accessed,
+ * the hashbucket lock must be held.
+ */
+
+/* searches for object==oid in pool, returns locked object if found */
+static struct tmem_obj *tmem_obj_find(struct tmem_hashbucket *hb,
+					struct tmem_oid *oidp)
+{
+	struct rb_node *rbnode;
+	struct tmem_obj *obj;
+
+	rbnode = hb->obj_rb_root.rb_node;
+	while (rbnode) {
+		BUG_ON(RB_EMPTY_NODE(rbnode));
+		obj = rb_entry(rbnode, struct tmem_obj, rb_tree_node);
+		switch (tmem_oid_compare(oidp, &obj->oid)) {
+		case 0: /* equal */
+			goto out;
+		case -1:
+			rbnode = rbnode->rb_left;
+			break;
+		case 1:
+			rbnode = rbnode->rb_right;
+			break;
+		}
+	}
+	obj = NULL;
+out:
+	return obj;
+}
+
+static void tmem_pampd_destroy_all_in_obj(struct tmem_obj *);
+
+/* free an object that has no more pampds in it */
+static void tmem_obj_free(struct tmem_obj *obj, struct tmem_hashbucket *hb)
+{
+	struct tmem_pool *pool;
+
+	BUG_ON(obj == NULL);
+	ASSERT_SENTINEL(obj, OBJ);
+	BUG_ON(obj->pampd_count > 0);
+	pool = obj->pool;
+	BUG_ON(pool == NULL);
+	if (obj->objnode_tree_root != NULL) /* may be "stump" with no leaves */
+		tmem_pampd_destroy_all_in_obj(obj);
+	BUG_ON(obj->objnode_tree_root != NULL);
+	BUG_ON((long)obj->objnode_count != 0);
+	atomic_dec(&pool->obj_count);
+	BUG_ON(atomic_read(&pool->obj_count) < 0);
+	INVERT_SENTINEL(obj, OBJ);
+	obj->pool = NULL;
+	tmem_oid_set_invalid(&obj->oid);
+	rb_erase(&obj->rb_tree_node, &hb->obj_rb_root);
+}
+
+/*
+ * initialize, and insert an tmem_object_root (called only if find failed)
+ */
+static void tmem_obj_init(struct tmem_obj *obj, struct tmem_hashbucket *hb,
+					struct tmem_pool *pool,
+					struct tmem_oid *oidp)
+{
+	struct rb_root *root = &hb->obj_rb_root;
+	struct rb_node **new = &(root->rb_node), *parent = NULL;
+	struct tmem_obj *this;
+
+	BUG_ON(pool == NULL);
+	atomic_inc(&pool->obj_count);
+	obj->objnode_tree_height = 0;
+	obj->objnode_tree_root = NULL;
+	obj->pool = pool;
+	obj->oid = *oidp;
+	obj->objnode_count = 0;
+	obj->pampd_count = 0;
+	(*tmem_pamops.new_obj)(obj);
+	SET_SENTINEL(obj, OBJ);
+	while (*new) {
+		BUG_ON(RB_EMPTY_NODE(*new));
+		this = rb_entry(*new, struct tmem_obj, rb_tree_node);
+		parent = *new;
+		switch (tmem_oid_compare(oidp, &this->oid)) {
+		case 0:
+			BUG(); /* already present; should never happen! */
+			break;
+		case -1:
+			new = &(*new)->rb_left;
+			break;
+		case 1:
+			new = &(*new)->rb_right;
+			break;
+		}
+	}
+	rb_link_node(&obj->rb_tree_node, parent, new);
+	rb_insert_color(&obj->rb_tree_node, root);
+}
+
+/*
+ * Tmem is managed as a set of tmem_pools with certain attributes, such as
+ * "ephemeral" vs "persistent".  These attributes apply to all tmem_objs
+ * and all pampds that belong to a tmem_pool.  A tmem_pool is created
+ * or deleted relatively rarely (for example, when a filesystem is
+ * mounted or unmounted.
+ */
+
+/* flush all data from a pool and, optionally, free it */
+static void tmem_pool_flush(struct tmem_pool *pool, bool destroy)
+{
+	struct rb_node *rbnode;
+	struct tmem_obj *obj;
+	struct tmem_hashbucket *hb = &pool->hashbucket[0];
+	int i;
+
+	BUG_ON(pool == NULL);
+	for (i = 0; i < TMEM_HASH_BUCKETS; i++, hb++) {
+		spin_lock(&hb->lock);
+		rbnode = rb_first(&hb->obj_rb_root);
+		while (rbnode != NULL) {
+			obj = rb_entry(rbnode, struct tmem_obj, rb_tree_node);
+			rbnode = rb_next(rbnode);
+			tmem_pampd_destroy_all_in_obj(obj);
+			tmem_obj_free(obj, hb);
+			(*tmem_hostops.obj_free)(obj, pool);
+		}
+		spin_unlock(&hb->lock);
+	}
+	if (destroy)
+		list_del(&pool->pool_list);
+}
+
+/*
+ * A tmem_obj contains a radix-tree-like tree in which the intermediate
+ * nodes are called tmem_objnodes.  (The kernel lib/radix-tree.c implementation
+ * is very specialized and tuned for specific uses and is not particularly
+ * suited for use from this code, though some code from the core algorithms has
+ * been reused, thus the copyright notices below).  Each tmem_objnode contains
+ * a set of pointers which point to either a set of intermediate tmem_objnodes
+ * or a set of of pampds.
+ *
+ * Portions Copyright (C) 2001 Momchil Velikov
+ * Portions Copyright (C) 2001 Christoph Hellwig
+ * Portions Copyright (C) 2005 SGI, Christoph Lameter <clameter@sgi.com>
+ */
+
+struct tmem_objnode_tree_path {
+	struct tmem_objnode *objnode;
+	int offset;
+};
+
+/* objnode height_to_maxindex translation */
+static unsigned long tmem_objnode_tree_h2max[OBJNODE_TREE_MAX_PATH + 1];
+
+static void tmem_objnode_tree_init(void)
+{
+	unsigned int ht, tmp;
+
+	for (ht = 0; ht < ARRAY_SIZE(tmem_objnode_tree_h2max); ht++) {
+		tmp = ht * OBJNODE_TREE_MAP_SHIFT;
+		if (tmp >= OBJNODE_TREE_INDEX_BITS)
+			tmem_objnode_tree_h2max[ht] = ~0UL;
+		else
+			tmem_objnode_tree_h2max[ht] =
+			    (~0UL >> (OBJNODE_TREE_INDEX_BITS - tmp - 1)) >> 1;
+	}
+}
+
+static struct tmem_objnode *tmem_objnode_alloc(struct tmem_obj *obj)
+{
+	struct tmem_objnode *objnode;
+
+	ASSERT_SENTINEL(obj, OBJ);
+	BUG_ON(obj->pool == NULL);
+	ASSERT_SENTINEL(obj->pool, POOL);
+	objnode = (*tmem_hostops.objnode_alloc)(obj->pool);
+	if (unlikely(objnode == NULL))
+		goto out;
+	objnode->obj = obj;
+	SET_SENTINEL(objnode, OBJNODE);
+	memset(&objnode->slots, 0, sizeof(objnode->slots));
+	objnode->slots_in_use = 0;
+	obj->objnode_count++;
+out:
+	return objnode;
+}
+
+static void tmem_objnode_free(struct tmem_objnode *objnode)
+{
+	struct tmem_pool *pool;
+	int i;
+
+	BUG_ON(objnode == NULL);
+	for (i = 0; i < OBJNODE_TREE_MAP_SIZE; i++)
+		BUG_ON(objnode->slots[i] != NULL);
+	ASSERT_SENTINEL(objnode, OBJNODE);
+	INVERT_SENTINEL(objnode, OBJNODE);
+	BUG_ON(objnode->obj == NULL);
+	ASSERT_SENTINEL(objnode->obj, OBJ);
+	pool = objnode->obj->pool;
+	BUG_ON(pool == NULL);
+	ASSERT_SENTINEL(pool, POOL);
+	objnode->obj->objnode_count--;
+	objnode->obj = NULL;
+	(*tmem_hostops.objnode_free)(objnode, pool);
+}
+
+/*
+ * lookup index in object and return associated pampd (or NULL if not found)
+ */
+static void **__tmem_pampd_lookup_in_obj(struct tmem_obj *obj, uint32_t index)
+{
+	unsigned int height, shift;
+	struct tmem_objnode **slot = NULL;
+
+	BUG_ON(obj == NULL);
+	ASSERT_SENTINEL(obj, OBJ);
+	BUG_ON(obj->pool == NULL);
+	ASSERT_SENTINEL(obj->pool, POOL);
+
+	height = obj->objnode_tree_height;
+	if (index > tmem_objnode_tree_h2max[obj->objnode_tree_height])
+		goto out;
+	if (height == 0 && obj->objnode_tree_root) {
+		slot = &obj->objnode_tree_root;
+		goto out;
+	}
+	shift = (height-1) * OBJNODE_TREE_MAP_SHIFT;
+	slot = &obj->objnode_tree_root;
+	while (height > 0) {
+		if (*slot == NULL)
+			goto out;
+		slot = (struct tmem_objnode **)
+			((*slot)->slots +
+			 ((index >> shift) & OBJNODE_TREE_MAP_MASK));
+		shift -= OBJNODE_TREE_MAP_SHIFT;
+		height--;
+	}
+out:
+	return slot != NULL ? (void **)slot : NULL;
+}
+
+static void *tmem_pampd_lookup_in_obj(struct tmem_obj *obj, uint32_t index)
+{
+	struct tmem_objnode **slot;
+
+	slot = (struct tmem_objnode **)__tmem_pampd_lookup_in_obj(obj, index);
+	return slot != NULL ? *slot : NULL;
+}
+
+static void *tmem_pampd_replace_in_obj(struct tmem_obj *obj, uint32_t index,
+					void *new_pampd)
+{
+	struct tmem_objnode **slot;
+	void *ret = NULL;
+
+	slot = (struct tmem_objnode **)__tmem_pampd_lookup_in_obj(obj, index);
+	if ((slot != NULL) && (*slot != NULL)) {
+		void *old_pampd = *(void **)slot;
+		*(void **)slot = new_pampd;
+		(*tmem_pamops.free)(old_pampd, obj->pool, NULL, 0);
+		ret = new_pampd;
+	}
+	return ret;
+}
+
+static int tmem_pampd_add_to_obj(struct tmem_obj *obj, uint32_t index,
+					void *pampd)
+{
+	int ret = 0;
+	struct tmem_objnode *objnode = NULL, *newnode, *slot;
+	unsigned int height, shift;
+	int offset = 0;
+
+	/* if necessary, extend the tree to be higher  */
+	if (index > tmem_objnode_tree_h2max[obj->objnode_tree_height]) {
+		height = obj->objnode_tree_height + 1;
+		if (index > tmem_objnode_tree_h2max[height])
+			while (index > tmem_objnode_tree_h2max[height])
+				height++;
+		if (obj->objnode_tree_root == NULL) {
+			obj->objnode_tree_height = height;
+			goto insert;
+		}
+		do {
+			newnode = tmem_objnode_alloc(obj);
+			if (!newnode) {
+				ret = -ENOMEM;
+				goto out;
+			}
+			newnode->slots[0] = obj->objnode_tree_root;
+			newnode->slots_in_use = 1;
+			obj->objnode_tree_root = newnode;
+			obj->objnode_tree_height++;
+		} while (height > obj->objnode_tree_height);
+	}
+insert:
+	slot = obj->objnode_tree_root;
+	height = obj->objnode_tree_height;
+	shift = (height-1) * OBJNODE_TREE_MAP_SHIFT;
+	while (height > 0) {
+		if (slot == NULL) {
+			/* add a child objnode.  */
+			slot = tmem_objnode_alloc(obj);
+			if (!slot) {
+				ret = -ENOMEM;
+				goto out;
+			}
+			if (objnode) {
+
+				objnode->slots[offset] = slot;
+				objnode->slots_in_use++;
+			} else
+				obj->objnode_tree_root = slot;
+		}
+		/* go down a level */
+		offset = (index >> shift) & OBJNODE_TREE_MAP_MASK;
+		objnode = slot;
+		slot = objnode->slots[offset];
+		shift -= OBJNODE_TREE_MAP_SHIFT;
+		height--;
+	}
+	BUG_ON(slot != NULL);
+	if (objnode) {
+		objnode->slots_in_use++;
+		objnode->slots[offset] = pampd;
+	} else
+		obj->objnode_tree_root = pampd;
+	obj->pampd_count++;
+out:
+	return ret;
+}
+
+static void *tmem_pampd_delete_from_obj(struct tmem_obj *obj, uint32_t index)
+{
+	struct tmem_objnode_tree_path path[OBJNODE_TREE_MAX_PATH + 1];
+	struct tmem_objnode_tree_path *pathp = path;
+	struct tmem_objnode *slot = NULL;
+	unsigned int height, shift;
+	int offset;
+
+	BUG_ON(obj == NULL);
+	ASSERT_SENTINEL(obj, OBJ);
+	BUG_ON(obj->pool == NULL);
+	ASSERT_SENTINEL(obj->pool, POOL);
+	height = obj->objnode_tree_height;
+	if (index > tmem_objnode_tree_h2max[height])
+		goto out;
+	slot = obj->objnode_tree_root;
+	if (height == 0 && obj->objnode_tree_root) {
+		obj->objnode_tree_root = NULL;
+		goto out;
+	}
+	shift = (height - 1) * OBJNODE_TREE_MAP_SHIFT;
+	pathp->objnode = NULL;
+	do {
+		if (slot == NULL)
+			goto out;
+		pathp++;
+		offset = (index >> shift) & OBJNODE_TREE_MAP_MASK;
+		pathp->offset = offset;
+		pathp->objnode = slot;
+		slot = slot->slots[offset];
+		shift -= OBJNODE_TREE_MAP_SHIFT;
+		height--;
+	} while (height > 0);
+	if (slot == NULL)
+		goto out;
+	while (pathp->objnode) {
+		pathp->objnode->slots[pathp->offset] = NULL;
+		pathp->objnode->slots_in_use--;
+		if (pathp->objnode->slots_in_use) {
+			if (pathp->objnode == obj->objnode_tree_root) {
+				while (obj->objnode_tree_height > 0 &&
+				  obj->objnode_tree_root->slots_in_use == 1 &&
+				  obj->objnode_tree_root->slots[0]) {
+					struct tmem_objnode *to_free =
+						obj->objnode_tree_root;
+
+					obj->objnode_tree_root =
+							to_free->slots[0];
+					obj->objnode_tree_height--;
+					to_free->slots[0] = NULL;
+					to_free->slots_in_use = 0;
+					tmem_objnode_free(to_free);
+				}
+			}
+			goto out;
+		}
+		tmem_objnode_free(pathp->objnode); /* 0 slots used, free it */
+		pathp--;
+	}
+	obj->objnode_tree_height = 0;
+	obj->objnode_tree_root = NULL;
+
+out:
+	if (slot != NULL)
+		obj->pampd_count--;
+	BUG_ON(obj->pampd_count < 0);
+	return slot;
+}
+
+/* recursively walk the objnode_tree destroying pampds and objnodes */
+static void tmem_objnode_node_destroy(struct tmem_obj *obj,
+					struct tmem_objnode *objnode,
+					unsigned int ht)
+{
+	int i;
+
+	if (ht == 0)
+		return;
+	for (i = 0; i < OBJNODE_TREE_MAP_SIZE; i++) {
+		if (objnode->slots[i]) {
+			if (ht == 1) {
+				obj->pampd_count--;
+				(*tmem_pamops.free)(objnode->slots[i],
+						obj->pool, NULL, 0);
+				objnode->slots[i] = NULL;
+				continue;
+			}
+			tmem_objnode_node_destroy(obj, objnode->slots[i], ht-1);
+			tmem_objnode_free(objnode->slots[i]);
+			objnode->slots[i] = NULL;
+		}
+	}
+}
+
+static void tmem_pampd_destroy_all_in_obj(struct tmem_obj *obj)
+{
+	if (obj->objnode_tree_root == NULL)
+		return;
+	if (obj->objnode_tree_height == 0) {
+		obj->pampd_count--;
+		(*tmem_pamops.free)(obj->objnode_tree_root, obj->pool, NULL, 0);
+	} else {
+		tmem_objnode_node_destroy(obj, obj->objnode_tree_root,
+					obj->objnode_tree_height);
+		tmem_objnode_free(obj->objnode_tree_root);
+		obj->objnode_tree_height = 0;
+	}
+	obj->objnode_tree_root = NULL;
+	(*tmem_pamops.free_obj)(obj->pool, obj);
+}
+
+/*
+ * Tmem is operated on by a set of well-defined actions:
+ * "put", "get", "flush", "flush_object", "new pool" and "destroy pool".
+ * (The tmem ABI allows for subpages and exchanges but these operations
+ * are not included in this implementation.)
+ *
+ * These "tmem core" operations are implemented in the following functions.
+ */
+
+/*
+ * "Put" a page, e.g. copy a page from the kernel into newly allocated
+ * PAM space (if such space is available).  Tmem_put is complicated by
+ * a corner case: What if a page with matching handle already exists in
+ * tmem?  To guarantee coherency, one of two actions is necessary: Either
+ * the data for the page must be overwritten, or the page must be
+ * "flushed" so that the data is not accessible to a subsequent "get".
+ * Since these "duplicate puts" are relatively rare, this implementation
+ * always flushes for simplicity.
+ */
+int tmem_put(struct tmem_pool *pool, struct tmem_oid *oidp, uint32_t index,
+		char *data, size_t size, bool raw, bool ephemeral)
+{
+	struct tmem_obj *obj = NULL, *objfound = NULL, *objnew = NULL;
+	void *pampd = NULL, *pampd_del = NULL;
+	int ret = -ENOMEM;
+	struct tmem_hashbucket *hb;
+
+	hb = &pool->hashbucket[tmem_oid_hash(oidp)];
+	spin_lock(&hb->lock);
+	obj = objfound = tmem_obj_find(hb, oidp);
+	if (obj != NULL) {
+		pampd = tmem_pampd_lookup_in_obj(objfound, index);
+		if (pampd != NULL) {
+			/* if found, is a dup put, flush the old one */
+			pampd_del = tmem_pampd_delete_from_obj(obj, index);
+			BUG_ON(pampd_del != pampd);
+			(*tmem_pamops.free)(pampd, pool, oidp, index);
+			if (obj->pampd_count == 0) {
+				objnew = obj;
+				objfound = NULL;
+			}
+			pampd = NULL;
+		}
+	} else {
+		obj = objnew = (*tmem_hostops.obj_alloc)(pool);
+		if (unlikely(obj == NULL)) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		tmem_obj_init(obj, hb, pool, oidp);
+	}
+	BUG_ON(obj == NULL);
+	BUG_ON(((objnew != obj) && (objfound != obj)) || (objnew == objfound));
+	pampd = (*tmem_pamops.create)(data, size, raw, ephemeral,
+					obj->pool, &obj->oid, index);
+	if (unlikely(pampd == NULL))
+		goto free;
+	ret = tmem_pampd_add_to_obj(obj, index, pampd);
+	if (unlikely(ret == -ENOMEM))
+		/* may have partially built objnode tree ("stump") */
+		goto delete_and_free;
+	goto out;
+
+delete_and_free:
+	(void)tmem_pampd_delete_from_obj(obj, index);
+free:
+	if (pampd)
+		(*tmem_pamops.free)(pampd, pool, NULL, 0);
+	if (objnew) {
+		tmem_obj_free(objnew, hb);
+		(*tmem_hostops.obj_free)(objnew, pool);
+	}
+out:
+	spin_unlock(&hb->lock);
+	return ret;
+}
+
+/*
+ * "Get" a page, e.g. if one can be found, copy the tmem page with the
+ * matching handle from PAM space to the kernel.  By tmem definition,
+ * when a "get" is successful on an ephemeral page, the page is "flushed",
+ * and when a "get" is successful on a persistent page, the page is retained
+ * in tmem.  Note that to preserve
+ * coherency, "get" can never be skipped if tmem contains the data.
+ * That is, if a get is done with a certain handle and fails, any
+ * subsequent "get" must also fail (unless of course there is a
+ * "put" done with the same handle).
+
+ */
+int tmem_get(struct tmem_pool *pool, struct tmem_oid *oidp, uint32_t index,
+		char *data, size_t *size, bool raw, int get_and_free)
+{
+	struct tmem_obj *obj;
+	void *pampd;
+	bool ephemeral = is_ephemeral(pool);
+	int ret = -1;
+	struct tmem_hashbucket *hb;
+	bool free = (get_and_free == 1) || ((get_and_free == 0) && ephemeral);
+	bool lock_held = false;
+
+	hb = &pool->hashbucket[tmem_oid_hash(oidp)];
+	spin_lock(&hb->lock);
+	lock_held = true;
+	obj = tmem_obj_find(hb, oidp);
+	if (obj == NULL)
+		goto out;
+	if (free)
+		pampd = tmem_pampd_delete_from_obj(obj, index);
+	else
+		pampd = tmem_pampd_lookup_in_obj(obj, index);
+	if (pampd == NULL)
+		goto out;
+	if (free) {
+		if (obj->pampd_count == 0) {
+			tmem_obj_free(obj, hb);
+			(*tmem_hostops.obj_free)(obj, pool);
+			obj = NULL;
+		}
+	}
+	if (tmem_pamops.is_remote(pampd)) {
+		lock_held = false;
+		spin_unlock(&hb->lock);
+	}
+	if (free)
+		ret = (*tmem_pamops.get_data_and_free)(
+				data, size, raw, pampd, pool, oidp, index);
+	else
+		ret = (*tmem_pamops.get_data)(
+				data, size, raw, pampd, pool, oidp, index);
+	if (ret < 0)
+		goto out;
+	ret = 0;
+out:
+	if (lock_held)
+		spin_unlock(&hb->lock);
+	return ret;
+}
+
+/*
+ * If a page in tmem matches the handle, "flush" this page from tmem such
+ * that any subsequent "get" does not succeed (unless, of course, there
+ * was another "put" with the same handle).
+ */
+int tmem_flush_page(struct tmem_pool *pool,
+				struct tmem_oid *oidp, uint32_t index)
+{
+	struct tmem_obj *obj;
+	void *pampd;
+	int ret = -1;
+	struct tmem_hashbucket *hb;
+
+	hb = &pool->hashbucket[tmem_oid_hash(oidp)];
+	spin_lock(&hb->lock);
+	obj = tmem_obj_find(hb, oidp);
+	if (obj == NULL)
+		goto out;
+	pampd = tmem_pampd_delete_from_obj(obj, index);
+	if (pampd == NULL)
+		goto out;
+	(*tmem_pamops.free)(pampd, pool, oidp, index);
+	if (obj->pampd_count == 0) {
+		tmem_obj_free(obj, hb);
+		(*tmem_hostops.obj_free)(obj, pool);
+	}
+	ret = 0;
+
+out:
+	spin_unlock(&hb->lock);
+	return ret;
+}
+
+/*
+ * If a page in tmem matches the handle, replace the page so that any
+ * subsequent "get" gets the new page.  Returns 0 if
+ * there was a page to replace, else returns -1.
+ */
+int tmem_replace(struct tmem_pool *pool, struct tmem_oid *oidp,
+			uint32_t index, void *new_pampd)
+{
+	struct tmem_obj *obj;
+	int ret = -1;
+	struct tmem_hashbucket *hb;
+
+	hb = &pool->hashbucket[tmem_oid_hash(oidp)];
+	spin_lock(&hb->lock);
+	obj = tmem_obj_find(hb, oidp);
+	if (obj == NULL)
+		goto out;
+	new_pampd = tmem_pampd_replace_in_obj(obj, index, new_pampd);
+	ret = (*tmem_pamops.replace_in_obj)(new_pampd, obj);
+out:
+	spin_unlock(&hb->lock);
+	return ret;
+}
+
+/*
+ * "Flush" all pages in tmem matching this oid.
+ */
+int tmem_flush_object(struct tmem_pool *pool, struct tmem_oid *oidp)
+{
+	struct tmem_obj *obj;
+	struct tmem_hashbucket *hb;
+	int ret = -1;
+
+	hb = &pool->hashbucket[tmem_oid_hash(oidp)];
+	spin_lock(&hb->lock);
+	obj = tmem_obj_find(hb, oidp);
+	if (obj == NULL)
+		goto out;
+	tmem_pampd_destroy_all_in_obj(obj);
+	tmem_obj_free(obj, hb);
+	(*tmem_hostops.obj_free)(obj, pool);
+	ret = 0;
+
+out:
+	spin_unlock(&hb->lock);
+	return ret;
+}
+
+/*
+ * "Flush" all pages (and tmem_objs) from this tmem_pool and disable
+ * all subsequent access to this tmem_pool.
+ */
+int tmem_destroy_pool(struct tmem_pool *pool)
+{
+	int ret = -1;
+
+	if (pool == NULL)
+		goto out;
+	tmem_pool_flush(pool, 1);
+	ret = 0;
+out:
+	return ret;
+}
+
+static LIST_HEAD(tmem_global_pool_list);
+
+/*
+ * Create a new tmem_pool with the provided flag and return
+ * a pool id provided by the tmem host implementation.
+ */
+void tmem_new_pool(struct tmem_pool *pool, uint32_t flags)
+{
+	int persistent = flags & TMEM_POOL_PERSIST;
+	int shared = flags & TMEM_POOL_SHARED;
+	struct tmem_hashbucket *hb = &pool->hashbucket[0];
+	int i;
+
+	for (i = 0; i < TMEM_HASH_BUCKETS; i++, hb++) {
+		hb->obj_rb_root = RB_ROOT;
+		spin_lock_init(&hb->lock);
+	}
+	INIT_LIST_HEAD(&pool->pool_list);
+	atomic_set(&pool->obj_count, 0);
+	SET_SENTINEL(pool, POOL);
+	list_add_tail(&pool->pool_list, &tmem_global_pool_list);
+	pool->persistent = persistent;
+	pool->shared = shared;
+}
diff --git a/drivers/staging/ramster/tmem.h b/drivers/staging/ramster/tmem.h
new file mode 100644
index 0000000..ed147c4
--- /dev/null
+++ b/drivers/staging/ramster/tmem.h
@@ -0,0 +1,206 @@
+/*
+ * tmem.h
+ *
+ * Transcendent memory
+ *
+ * Copyright (c) 2009-2011, Dan Magenheimer, Oracle Corp.
+ */
+
+#ifndef _TMEM_H_
+#define _TMEM_H_
+
+#include <linux/types.h>
+#include <linux/highmem.h>
+#include <linux/hash.h>
+#include <linux/atomic.h>
+
+/*
+ * These are pre-defined by the Xen<->Linux ABI
+ */
+#define TMEM_PUT_PAGE			4
+#define TMEM_GET_PAGE			5
+#define TMEM_FLUSH_PAGE			6
+#define TMEM_FLUSH_OBJECT		7
+#define TMEM_POOL_PERSIST		1
+#define TMEM_POOL_SHARED		2
+#define TMEM_POOL_PRECOMPRESSED		4
+#define TMEM_POOL_PAGESIZE_SHIFT	4
+#define TMEM_POOL_PAGESIZE_MASK		0xf
+#define TMEM_POOL_RESERVED_BITS		0x00ffff00
+
+/*
+ * sentinels have proven very useful for debugging but can be removed
+ * or disabled before final merge.
+ */
+#define SENTINELS
+#ifdef SENTINELS
+#define DECL_SENTINEL uint32_t sentinel;
+#define SET_SENTINEL(_x, _y) (_x->sentinel = _y##_SENTINEL)
+#define INVERT_SENTINEL(_x, _y) (_x->sentinel = ~_y##_SENTINEL)
+#define ASSERT_SENTINEL(_x, _y) WARN_ON(_x->sentinel != _y##_SENTINEL)
+#define ASSERT_INVERTED_SENTINEL(_x, _y) WARN_ON(_x->sentinel != ~_y##_SENTINEL)
+#else
+#define DECL_SENTINEL
+#define SET_SENTINEL(_x, _y) do { } while (0)
+#define INVERT_SENTINEL(_x, _y) do { } while (0)
+#define ASSERT_SENTINEL(_x, _y) do { } while (0)
+#define ASSERT_INVERTED_SENTINEL(_x, _y) do { } while (0)
+#endif
+
+#define ASSERT_SPINLOCK(_l)	WARN_ON(!spin_is_locked(_l))
+
+/*
+ * A pool is the highest-level data structure managed by tmem and
+ * usually corresponds to a large independent set of pages such as
+ * a filesystem.  Each pool has an id, and certain attributes and counters.
+ * It also contains a set of hash buckets, each of which contains an rbtree
+ * of objects and a lock to manage concurrency within the pool.
+ */
+
+#define TMEM_HASH_BUCKET_BITS	8
+#define TMEM_HASH_BUCKETS	(1<<TMEM_HASH_BUCKET_BITS)
+
+struct tmem_hashbucket {
+	struct rb_root obj_rb_root;
+	spinlock_t lock;
+};
+
+struct tmem_pool {
+	void *client; /* "up" for some clients, avoids table lookup */
+	struct list_head pool_list;
+	uint32_t pool_id;
+	bool persistent;
+	bool shared;
+	atomic_t obj_count;
+	atomic_t refcount;
+	struct tmem_hashbucket hashbucket[TMEM_HASH_BUCKETS];
+	DECL_SENTINEL
+};
+
+#define is_persistent(_p)  (_p->persistent)
+#define is_ephemeral(_p)   (!(_p->persistent))
+
+/*
+ * An object id ("oid") is large: 192-bits (to ensure, for example, files
+ * in a modern filesystem can be uniquely identified).
+ */
+
+struct tmem_oid {
+	uint64_t oid[3];
+};
+
+static inline void tmem_oid_set_invalid(struct tmem_oid *oidp)
+{
+	oidp->oid[0] = oidp->oid[1] = oidp->oid[2] = -1UL;
+}
+
+static inline bool tmem_oid_valid(struct tmem_oid *oidp)
+{
+	return oidp->oid[0] != -1UL || oidp->oid[1] != -1UL ||
+		oidp->oid[2] != -1UL;
+}
+
+static inline int tmem_oid_compare(struct tmem_oid *left,
+					struct tmem_oid *right)
+{
+	int ret;
+
+	if (left->oid[2] == right->oid[2]) {
+		if (left->oid[1] == right->oid[1]) {
+			if (left->oid[0] == right->oid[0])
+				ret = 0;
+			else if (left->oid[0] < right->oid[0])
+				ret = -1;
+			else
+				return 1;
+		} else if (left->oid[1] < right->oid[1])
+			ret = -1;
+		else
+			ret = 1;
+	} else if (left->oid[2] < right->oid[2])
+		ret = -1;
+	else
+		ret = 1;
+	return ret;
+}
+
+static inline unsigned tmem_oid_hash(struct tmem_oid *oidp)
+{
+	return hash_long(oidp->oid[0] ^ oidp->oid[1] ^ oidp->oid[2],
+				TMEM_HASH_BUCKET_BITS);
+}
+
+/*
+ * A tmem_obj contains an identifier (oid), pointers to the parent
+ * pool and the rb_tree to which it belongs, counters, and an ordered
+ * set of pampds, structured in a radix-tree-like tree.  The intermediate
+ * nodes of the tree are called tmem_objnodes.
+ */
+
+struct tmem_objnode;
+
+struct tmem_obj {
+	struct tmem_oid oid;
+	struct tmem_pool *pool;
+	struct rb_node rb_tree_node;
+	struct tmem_objnode *objnode_tree_root;
+	unsigned int objnode_tree_height;
+	unsigned long objnode_count;
+	long pampd_count;
+	void *extra; /* for private use by pampd implementation */
+	DECL_SENTINEL
+};
+
+#define OBJNODE_TREE_MAP_SHIFT 6
+#define OBJNODE_TREE_MAP_SIZE (1UL << OBJNODE_TREE_MAP_SHIFT)
+#define OBJNODE_TREE_MAP_MASK (OBJNODE_TREE_MAP_SIZE-1)
+#define OBJNODE_TREE_INDEX_BITS (8 /* CHAR_BIT */ * sizeof(unsigned long))
+#define OBJNODE_TREE_MAX_PATH \
+		(OBJNODE_TREE_INDEX_BITS/OBJNODE_TREE_MAP_SHIFT + 2)
+
+struct tmem_objnode {
+	struct tmem_obj *obj;
+	DECL_SENTINEL
+	void *slots[OBJNODE_TREE_MAP_SIZE];
+	unsigned int slots_in_use;
+};
+
+/* pampd abstract datatype methods provided by the PAM implementation */
+struct tmem_pamops {
+	void *(*create)(char *, size_t, bool, int,
+			struct tmem_pool *, struct tmem_oid *, uint32_t);
+	int (*get_data)(char *, size_t *, bool, void *, struct tmem_pool *,
+				struct tmem_oid *, uint32_t);
+	int (*get_data_and_free)(char *, size_t *, bool, void *,
+				struct tmem_pool *, struct tmem_oid *,
+				uint32_t);
+	void (*free)(void *, struct tmem_pool *, struct tmem_oid *, uint32_t);
+	void (*free_obj)(struct tmem_pool *, struct tmem_obj *);
+	bool (*is_remote)(void *);
+	void (*new_obj)(struct tmem_obj *);
+	int (*replace_in_obj)(void *, struct tmem_obj *);
+};
+extern void tmem_register_pamops(struct tmem_pamops *m);
+
+/* memory allocation methods provided by the host implementation */
+struct tmem_hostops {
+	struct tmem_obj *(*obj_alloc)(struct tmem_pool *);
+	void (*obj_free)(struct tmem_obj *, struct tmem_pool *);
+	struct tmem_objnode *(*objnode_alloc)(struct tmem_pool *);
+	void (*objnode_free)(struct tmem_objnode *, struct tmem_pool *);
+};
+extern void tmem_register_hostops(struct tmem_hostops *m);
+
+/* core tmem accessor functions */
+extern int tmem_put(struct tmem_pool *, struct tmem_oid *, uint32_t index,
+			char *, size_t, bool, bool);
+extern int tmem_get(struct tmem_pool *, struct tmem_oid *, uint32_t index,
+			char *, size_t *, bool, int);
+extern int tmem_replace(struct tmem_pool *, struct tmem_oid *, uint32_t index,
+			void *);
+extern int tmem_flush_page(struct tmem_pool *, struct tmem_oid *,
+			uint32_t index);
+extern int tmem_flush_object(struct tmem_pool *, struct tmem_oid *);
+extern int tmem_destroy_pool(struct tmem_pool *);
+extern void tmem_new_pool(struct tmem_pool *, uint32_t);
+#endif /* _TMEM_H */
diff --git a/drivers/staging/ramster/zcache-main.c b/drivers/staging/ramster/zcache-main.c
new file mode 100644
index 0000000..49c8791
--- /dev/null
+++ b/drivers/staging/ramster/zcache-main.c
@@ -0,0 +1,1988 @@
+/*
+ * zcache.c
+ *
+ * Copyright (c) 2010,2011, Dan Magenheimer, Oracle Corp.
+ * Copyright (c) 2010,2011, Nitin Gupta
+ *
+ * Zcache provides an in-kernel "host implementation" for transcendent memory
+ * and, thus indirectly, for cleancache and frontswap.  Zcache includes two
+ * page-accessible memory [1] interfaces, both utilizing lzo1x compression:
+ * 1) "compression buddies" ("zbud") is used for ephemeral pages
+ * 2) xvmalloc is used for persistent pages.
+ * Xvmalloc (based on the TLSF allocator) has very low fragmentation
+ * so maximizes space efficiency, while zbud allows pairs (and potentially,
+ * in the future, more than a pair of) compressed pages to be closely linked
+ * so that reclaiming can be done via the kernel's physical-page-oriented
+ * "shrinker" interface.
+ *
+ * [1] For a definition of page-accessible memory (aka PAM), see:
+ *   http://marc.info/?l=linux-mm&m=127811271605009
+ */
+
+#include <linux/module.h>
+#include <linux/cpu.h>
+#include <linux/highmem.h>
+#include <linux/list.h>
+#include <linux/lzo.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+#include <linux/atomic.h>
+#include <linux/math64.h>
+#include "tmem.h"
+
+#include "../zram/xvmalloc.h" /* if built in drivers/staging */
+
+#if (!defined(CONFIG_CLEANCACHE) && !defined(CONFIG_FRONTSWAP))
+#error "zcache is useless without CONFIG_CLEANCACHE or CONFIG_FRONTSWAP"
+#endif
+#ifdef CONFIG_CLEANCACHE
+#include <linux/cleancache.h>
+#endif
+#ifdef CONFIG_FRONTSWAP
+#include <linux/frontswap.h>
+#endif
+
+#if 0
+/* this is more aggressive but may cause other problems? */
+#define ZCACHE_GFP_MASK	(GFP_ATOMIC | __GFP_NORETRY | __GFP_NOWARN)
+#else
+#define ZCACHE_GFP_MASK \
+	(__GFP_FS | __GFP_NORETRY | __GFP_NOWARN | __GFP_NOMEMALLOC)
+#endif
+
+#define MAX_POOLS_PER_CLIENT 16
+
+#define MAX_CLIENTS 16
+#define LOCAL_CLIENT ((uint16_t)-1)
+
+MODULE_LICENSE("GPL");
+
+struct zcache_client {
+	struct tmem_pool *tmem_pools[MAX_POOLS_PER_CLIENT];
+	struct xv_pool *xvpool;
+	bool allocated;
+	atomic_t refcount;
+};
+
+static struct zcache_client zcache_host;
+static struct zcache_client zcache_clients[MAX_CLIENTS];
+
+static inline uint16_t get_client_id_from_client(struct zcache_client *cli)
+{
+	BUG_ON(cli == NULL);
+	if (cli == &zcache_host)
+		return LOCAL_CLIENT;
+	return cli - &zcache_clients[0];
+}
+
+static inline bool is_local_client(struct zcache_client *cli)
+{
+	return cli == &zcache_host;
+}
+
+/**********
+ * Compression buddies ("zbud") provides for packing two (or, possibly
+ * in the future, more) compressed ephemeral pages into a single "raw"
+ * (physical) page and tracking them with data structures so that
+ * the raw pages can be easily reclaimed.
+ *
+ * A zbud page ("zbpg") is an aligned page containing a list_head,
+ * a lock, and two "zbud headers".  The remainder of the physical
+ * page is divided up into aligned 64-byte "chunks" which contain
+ * the compressed data for zero, one, or two zbuds.  Each zbpg
+ * resides on: (1) an "unused list" if it has no zbuds; (2) a
+ * "buddied" list if it is fully populated  with two zbuds; or
+ * (3) one of PAGE_SIZE/64 "unbuddied" lists indexed by how many chunks
+ * the one unbuddied zbud uses.  The data inside a zbpg cannot be
+ * read or written unless the zbpg's lock is held.
+ */
+
+#define ZBH_SENTINEL  0x43214321
+#define ZBPG_SENTINEL  0xdeadbeef
+
+#define ZBUD_MAX_BUDS 2
+
+struct zbud_hdr {
+	uint16_t client_id;
+	uint16_t pool_id;
+	struct tmem_oid oid;
+	uint32_t index;
+	uint16_t size; /* compressed size in bytes, zero means unused */
+	DECL_SENTINEL
+};
+
+struct zbud_page {
+	struct list_head bud_list;
+	spinlock_t lock;
+	struct zbud_hdr buddy[ZBUD_MAX_BUDS];
+	DECL_SENTINEL
+	/* followed by NUM_CHUNK aligned CHUNK_SIZE-byte chunks */
+};
+
+#define CHUNK_SHIFT	6
+#define CHUNK_SIZE	(1 << CHUNK_SHIFT)
+#define CHUNK_MASK	(~(CHUNK_SIZE-1))
+#define NCHUNKS		(((PAGE_SIZE - sizeof(struct zbud_page)) & \
+				CHUNK_MASK) >> CHUNK_SHIFT)
+#define MAX_CHUNK	(NCHUNKS-1)
+
+static struct {
+	struct list_head list;
+	unsigned count;
+} zbud_unbuddied[NCHUNKS];
+/* list N contains pages with N chunks USED and NCHUNKS-N unused */
+/* element 0 is never used but optimizing that isn't worth it */
+static unsigned long zbud_cumul_chunk_counts[NCHUNKS];
+
+struct list_head zbud_buddied_list;
+static unsigned long zcache_zbud_buddied_count;
+
+/* protects the buddied list and all unbuddied lists */
+static DEFINE_SPINLOCK(zbud_budlists_spinlock);
+
+static LIST_HEAD(zbpg_unused_list);
+static unsigned long zcache_zbpg_unused_list_count;
+
+/* protects the unused page list */
+static DEFINE_SPINLOCK(zbpg_unused_list_spinlock);
+
+static atomic_t zcache_zbud_curr_raw_pages;
+static atomic_t zcache_zbud_curr_zpages;
+static unsigned long zcache_zbud_curr_zbytes;
+static unsigned long zcache_zbud_cumul_zpages;
+static unsigned long zcache_zbud_cumul_zbytes;
+static unsigned long zcache_compress_poor;
+static unsigned long zcache_mean_compress_poor;
+
+/* forward references */
+static void *zcache_get_free_page(void);
+static void zcache_free_page(void *p);
+
+/*
+ * zbud helper functions
+ */
+
+static inline unsigned zbud_max_buddy_size(void)
+{
+	return MAX_CHUNK << CHUNK_SHIFT;
+}
+
+static inline unsigned zbud_size_to_chunks(unsigned size)
+{
+	BUG_ON(size == 0 || size > zbud_max_buddy_size());
+	return (size + CHUNK_SIZE - 1) >> CHUNK_SHIFT;
+}
+
+static inline int zbud_budnum(struct zbud_hdr *zh)
+{
+	unsigned offset = (unsigned long)zh & (PAGE_SIZE - 1);
+	struct zbud_page *zbpg = NULL;
+	unsigned budnum = -1U;
+	int i;
+
+	for (i = 0; i < ZBUD_MAX_BUDS; i++)
+		if (offset == offsetof(typeof(*zbpg), buddy[i])) {
+			budnum = i;
+			break;
+		}
+	BUG_ON(budnum == -1U);
+	return budnum;
+}
+
+static char *zbud_data(struct zbud_hdr *zh, unsigned size)
+{
+	struct zbud_page *zbpg;
+	char *p;
+	unsigned budnum;
+
+	ASSERT_SENTINEL(zh, ZBH);
+	budnum = zbud_budnum(zh);
+	BUG_ON(size == 0 || size > zbud_max_buddy_size());
+	zbpg = container_of(zh, struct zbud_page, buddy[budnum]);
+	ASSERT_SPINLOCK(&zbpg->lock);
+	p = (char *)zbpg;
+	if (budnum == 0)
+		p += ((sizeof(struct zbud_page) + CHUNK_SIZE - 1) &
+							CHUNK_MASK);
+	else if (budnum == 1)
+		p += PAGE_SIZE - ((size + CHUNK_SIZE - 1) & CHUNK_MASK);
+	return p;
+}
+
+/*
+ * zbud raw page management
+ */
+
+static struct zbud_page *zbud_alloc_raw_page(void)
+{
+	struct zbud_page *zbpg = NULL;
+	struct zbud_hdr *zh0, *zh1;
+	bool recycled = 0;
+
+	/* if any pages on the zbpg list, use one */
+	spin_lock(&zbpg_unused_list_spinlock);
+	if (!list_empty(&zbpg_unused_list)) {
+		zbpg = list_first_entry(&zbpg_unused_list,
+				struct zbud_page, bud_list);
+		list_del_init(&zbpg->bud_list);
+		zcache_zbpg_unused_list_count--;
+		recycled = 1;
+	}
+	spin_unlock(&zbpg_unused_list_spinlock);
+	if (zbpg == NULL)
+		/* none on zbpg list, try to get a kernel page */
+		zbpg = zcache_get_free_page();
+	if (likely(zbpg != NULL)) {
+		INIT_LIST_HEAD(&zbpg->bud_list);
+		zh0 = &zbpg->buddy[0]; zh1 = &zbpg->buddy[1];
+		spin_lock_init(&zbpg->lock);
+		if (recycled) {
+			ASSERT_INVERTED_SENTINEL(zbpg, ZBPG);
+			SET_SENTINEL(zbpg, ZBPG);
+			BUG_ON(zh0->size != 0 || tmem_oid_valid(&zh0->oid));
+			BUG_ON(zh1->size != 0 || tmem_oid_valid(&zh1->oid));
+		} else {
+			atomic_inc(&zcache_zbud_curr_raw_pages);
+			INIT_LIST_HEAD(&zbpg->bud_list);
+			SET_SENTINEL(zbpg, ZBPG);
+			zh0->size = 0; zh1->size = 0;
+			tmem_oid_set_invalid(&zh0->oid);
+			tmem_oid_set_invalid(&zh1->oid);
+		}
+	}
+	return zbpg;
+}
+
+static void zbud_free_raw_page(struct zbud_page *zbpg)
+{
+	struct zbud_hdr *zh0 = &zbpg->buddy[0], *zh1 = &zbpg->buddy[1];
+
+	ASSERT_SENTINEL(zbpg, ZBPG);
+	BUG_ON(!list_empty(&zbpg->bud_list));
+	ASSERT_SPINLOCK(&zbpg->lock);
+	BUG_ON(zh0->size != 0 || tmem_oid_valid(&zh0->oid));
+	BUG_ON(zh1->size != 0 || tmem_oid_valid(&zh1->oid));
+	INVERT_SENTINEL(zbpg, ZBPG);
+	spin_unlock(&zbpg->lock);
+	spin_lock(&zbpg_unused_list_spinlock);
+	list_add(&zbpg->bud_list, &zbpg_unused_list);
+	zcache_zbpg_unused_list_count++;
+	spin_unlock(&zbpg_unused_list_spinlock);
+}
+
+/*
+ * core zbud handling routines
+ */
+
+static unsigned zbud_free(struct zbud_hdr *zh)
+{
+	unsigned size;
+
+	ASSERT_SENTINEL(zh, ZBH);
+	BUG_ON(!tmem_oid_valid(&zh->oid));
+	size = zh->size;
+	BUG_ON(zh->size == 0 || zh->size > zbud_max_buddy_size());
+	zh->size = 0;
+	tmem_oid_set_invalid(&zh->oid);
+	INVERT_SENTINEL(zh, ZBH);
+	zcache_zbud_curr_zbytes -= size;
+	atomic_dec(&zcache_zbud_curr_zpages);
+	return size;
+}
+
+static void zbud_free_and_delist(struct zbud_hdr *zh)
+{
+	unsigned chunks;
+	struct zbud_hdr *zh_other;
+	unsigned budnum = zbud_budnum(zh), size;
+	struct zbud_page *zbpg =
+		container_of(zh, struct zbud_page, buddy[budnum]);
+
+	spin_lock(&zbpg->lock);
+	if (list_empty(&zbpg->bud_list)) {
+		/* ignore zombie page... see zbud_evict_pages() */
+		spin_unlock(&zbpg->lock);
+		return;
+	}
+	size = zbud_free(zh);
+	ASSERT_SPINLOCK(&zbpg->lock);
+	zh_other = &zbpg->buddy[(budnum == 0) ? 1 : 0];
+	if (zh_other->size == 0) { /* was unbuddied: unlist and free */
+		chunks = zbud_size_to_chunks(size) ;
+		spin_lock(&zbud_budlists_spinlock);
+		BUG_ON(list_empty(&zbud_unbuddied[chunks].list));
+		list_del_init(&zbpg->bud_list);
+		zbud_unbuddied[chunks].count--;
+		spin_unlock(&zbud_budlists_spinlock);
+		zbud_free_raw_page(zbpg);
+	} else { /* was buddied: move remaining buddy to unbuddied list */
+		chunks = zbud_size_to_chunks(zh_other->size) ;
+		spin_lock(&zbud_budlists_spinlock);
+		list_del_init(&zbpg->bud_list);
+		zcache_zbud_buddied_count--;
+		list_add_tail(&zbpg->bud_list, &zbud_unbuddied[chunks].list);
+		zbud_unbuddied[chunks].count++;
+		spin_unlock(&zbud_budlists_spinlock);
+		spin_unlock(&zbpg->lock);
+	}
+}
+
+static struct zbud_hdr *zbud_create(uint16_t client_id, uint16_t pool_id,
+					struct tmem_oid *oid,
+					uint32_t index, struct page *page,
+					void *cdata, unsigned size)
+{
+	struct zbud_hdr *zh0, *zh1, *zh = NULL;
+	struct zbud_page *zbpg = NULL, *ztmp;
+	unsigned nchunks;
+	char *to;
+	int i, found_good_buddy = 0;
+
+	nchunks = zbud_size_to_chunks(size) ;
+	for (i = MAX_CHUNK - nchunks + 1; i > 0; i--) {
+		spin_lock(&zbud_budlists_spinlock);
+		if (!list_empty(&zbud_unbuddied[i].list)) {
+			list_for_each_entry_safe(zbpg, ztmp,
+				    &zbud_unbuddied[i].list, bud_list) {
+				if (spin_trylock(&zbpg->lock)) {
+					found_good_buddy = i;
+					goto found_unbuddied;
+				}
+			}
+		}
+		spin_unlock(&zbud_budlists_spinlock);
+	}
+	/* didn't find a good buddy, try allocating a new page */
+	zbpg = zbud_alloc_raw_page();
+	if (unlikely(zbpg == NULL))
+		goto out;
+	/* ok, have a page, now compress the data before taking locks */
+	spin_lock(&zbpg->lock);
+	spin_lock(&zbud_budlists_spinlock);
+	list_add_tail(&zbpg->bud_list, &zbud_unbuddied[nchunks].list);
+	zbud_unbuddied[nchunks].count++;
+	zh = &zbpg->buddy[0];
+	goto init_zh;
+
+found_unbuddied:
+	ASSERT_SPINLOCK(&zbpg->lock);
+	zh0 = &zbpg->buddy[0]; zh1 = &zbpg->buddy[1];
+	BUG_ON(!((zh0->size == 0) ^ (zh1->size == 0)));
+	if (zh0->size != 0) { /* buddy0 in use, buddy1 is vacant */
+		ASSERT_SENTINEL(zh0, ZBH);
+		zh = zh1;
+	} else if (zh1->size != 0) { /* buddy1 in use, buddy0 is vacant */
+		ASSERT_SENTINEL(zh1, ZBH);
+		zh = zh0;
+	} else
+		BUG();
+	list_del_init(&zbpg->bud_list);
+	zbud_unbuddied[found_good_buddy].count--;
+	list_add_tail(&zbpg->bud_list, &zbud_buddied_list);
+	zcache_zbud_buddied_count++;
+
+init_zh:
+	SET_SENTINEL(zh, ZBH);
+	zh->size = size;
+	zh->index = index;
+	zh->oid = *oid;
+	zh->pool_id = pool_id;
+	zh->client_id = client_id;
+	/* can wait to copy the data until the list locks are dropped */
+	spin_unlock(&zbud_budlists_spinlock);
+
+	to = zbud_data(zh, size);
+	memcpy(to, cdata, size);
+	spin_unlock(&zbpg->lock);
+	zbud_cumul_chunk_counts[nchunks]++;
+	atomic_inc(&zcache_zbud_curr_zpages);
+	zcache_zbud_cumul_zpages++;
+	zcache_zbud_curr_zbytes += size;
+	zcache_zbud_cumul_zbytes += size;
+out:
+	return zh;
+}
+
+static int zbud_decompress(struct page *page, struct zbud_hdr *zh)
+{
+	struct zbud_page *zbpg;
+	unsigned budnum = zbud_budnum(zh);
+	size_t out_len = PAGE_SIZE;
+	char *to_va, *from_va;
+	unsigned size;
+	int ret = 0;
+
+	zbpg = container_of(zh, struct zbud_page, buddy[budnum]);
+	spin_lock(&zbpg->lock);
+	if (list_empty(&zbpg->bud_list)) {
+		/* ignore zombie page... see zbud_evict_pages() */
+		ret = -EINVAL;
+		goto out;
+	}
+	ASSERT_SENTINEL(zh, ZBH);
+	BUG_ON(zh->size == 0 || zh->size > zbud_max_buddy_size());
+	to_va = kmap_atomic(page, KM_USER0);
+	size = zh->size;
+	from_va = zbud_data(zh, size);
+	ret = lzo1x_decompress_safe(from_va, size, to_va, &out_len);
+	BUG_ON(ret != LZO_E_OK);
+	BUG_ON(out_len != PAGE_SIZE);
+	kunmap_atomic(to_va, KM_USER0);
+out:
+	spin_unlock(&zbpg->lock);
+	return ret;
+}
+
+/*
+ * The following routines handle shrinking of ephemeral pages by evicting
+ * pages "least valuable" first.
+ */
+
+static unsigned long zcache_evicted_raw_pages;
+static unsigned long zcache_evicted_buddied_pages;
+static unsigned long zcache_evicted_unbuddied_pages;
+
+static struct tmem_pool *zcache_get_pool_by_id(uint16_t cli_id,
+						uint16_t poolid);
+static void zcache_put_pool(struct tmem_pool *pool);
+
+/*
+ * Flush and free all zbuds in a zbpg, then free the pageframe
+ */
+static void zbud_evict_zbpg(struct zbud_page *zbpg)
+{
+	struct zbud_hdr *zh;
+	int i, j;
+	uint32_t pool_id[ZBUD_MAX_BUDS], client_id[ZBUD_MAX_BUDS];
+	uint32_t index[ZBUD_MAX_BUDS];
+	struct tmem_oid oid[ZBUD_MAX_BUDS];
+	struct tmem_pool *pool;
+
+	ASSERT_SPINLOCK(&zbpg->lock);
+	BUG_ON(!list_empty(&zbpg->bud_list));
+	for (i = 0, j = 0; i < ZBUD_MAX_BUDS; i++) {
+		zh = &zbpg->buddy[i];
+		if (zh->size) {
+			client_id[j] = zh->client_id;
+			pool_id[j] = zh->pool_id;
+			oid[j] = zh->oid;
+			index[j] = zh->index;
+			j++;
+			zbud_free(zh);
+		}
+	}
+	spin_unlock(&zbpg->lock);
+	for (i = 0; i < j; i++) {
+		pool = zcache_get_pool_by_id(client_id[i], pool_id[i]);
+		if (pool != NULL) {
+			tmem_flush_page(pool, &oid[i], index[i]);
+			zcache_put_pool(pool);
+		}
+	}
+	ASSERT_SENTINEL(zbpg, ZBPG);
+	spin_lock(&zbpg->lock);
+	zbud_free_raw_page(zbpg);
+}
+
+/*
+ * Free nr pages.  This code is funky because we want to hold the locks
+ * protecting various lists for as short a time as possible, and in some
+ * circumstances the list may change asynchronously when the list lock is
+ * not held.  In some cases we also trylock not only to avoid waiting on a
+ * page in use by another cpu, but also to avoid potential deadlock due to
+ * lock inversion.
+ */
+static void zbud_evict_pages(int nr)
+{
+	struct zbud_page *zbpg;
+	int i;
+
+	/* first try freeing any pages on unused list */
+retry_unused_list:
+	spin_lock_bh(&zbpg_unused_list_spinlock);
+	if (!list_empty(&zbpg_unused_list)) {
+		/* can't walk list here, since it may change when unlocked */
+		zbpg = list_first_entry(&zbpg_unused_list,
+				struct zbud_page, bud_list);
+		list_del_init(&zbpg->bud_list);
+		zcache_zbpg_unused_list_count--;
+		atomic_dec(&zcache_zbud_curr_raw_pages);
+		spin_unlock_bh(&zbpg_unused_list_spinlock);
+		zcache_free_page(zbpg);
+		zcache_evicted_raw_pages++;
+		if (--nr <= 0)
+			goto out;
+		goto retry_unused_list;
+	}
+	spin_unlock_bh(&zbpg_unused_list_spinlock);
+
+	/* now try freeing unbuddied pages, starting with least space avail */
+	for (i = 0; i < MAX_CHUNK; i++) {
+retry_unbud_list_i:
+		spin_lock_bh(&zbud_budlists_spinlock);
+		if (list_empty(&zbud_unbuddied[i].list)) {
+			spin_unlock_bh(&zbud_budlists_spinlock);
+			continue;
+		}
+		list_for_each_entry(zbpg, &zbud_unbuddied[i].list, bud_list) {
+			if (unlikely(!spin_trylock(&zbpg->lock)))
+				continue;
+			list_del_init(&zbpg->bud_list);
+			zbud_unbuddied[i].count--;
+			spin_unlock(&zbud_budlists_spinlock);
+			zcache_evicted_unbuddied_pages++;
+			/* want budlists unlocked when doing zbpg eviction */
+			zbud_evict_zbpg(zbpg);
+			local_bh_enable();
+			if (--nr <= 0)
+				goto out;
+			goto retry_unbud_list_i;
+		}
+		spin_unlock_bh(&zbud_budlists_spinlock);
+	}
+
+	/* as a last resort, free buddied pages */
+retry_bud_list:
+	spin_lock_bh(&zbud_budlists_spinlock);
+	if (list_empty(&zbud_buddied_list)) {
+		spin_unlock_bh(&zbud_budlists_spinlock);
+		goto out;
+	}
+	list_for_each_entry(zbpg, &zbud_buddied_list, bud_list) {
+		if (unlikely(!spin_trylock(&zbpg->lock)))
+			continue;
+		list_del_init(&zbpg->bud_list);
+		zcache_zbud_buddied_count--;
+		spin_unlock(&zbud_budlists_spinlock);
+		zcache_evicted_buddied_pages++;
+		/* want budlists unlocked when doing zbpg eviction */
+		zbud_evict_zbpg(zbpg);
+		local_bh_enable();
+		if (--nr <= 0)
+			goto out;
+		goto retry_bud_list;
+	}
+	spin_unlock_bh(&zbud_budlists_spinlock);
+out:
+	return;
+}
+
+static void zbud_init(void)
+{
+	int i;
+
+	INIT_LIST_HEAD(&zbud_buddied_list);
+	zcache_zbud_buddied_count = 0;
+	for (i = 0; i < NCHUNKS; i++) {
+		INIT_LIST_HEAD(&zbud_unbuddied[i].list);
+		zbud_unbuddied[i].count = 0;
+	}
+}
+
+#ifdef CONFIG_SYSFS
+/*
+ * These sysfs routines show a nice distribution of how many zbpg's are
+ * currently (and have ever been placed) in each unbuddied list.  It's fun
+ * to watch but can probably go away before final merge.
+ */
+static int zbud_show_unbuddied_list_counts(char *buf)
+{
+	int i;
+	char *p = buf;
+
+	for (i = 0; i < NCHUNKS; i++)
+		p += sprintf(p, "%u ", zbud_unbuddied[i].count);
+	return p - buf;
+}
+
+static int zbud_show_cumul_chunk_counts(char *buf)
+{
+	unsigned long i, chunks = 0, total_chunks = 0, sum_total_chunks = 0;
+	unsigned long total_chunks_lte_21 = 0, total_chunks_lte_32 = 0;
+	unsigned long total_chunks_lte_42 = 0;
+	char *p = buf;
+
+	for (i = 0; i < NCHUNKS; i++) {
+		p += sprintf(p, "%lu ", zbud_cumul_chunk_counts[i]);
+		chunks += zbud_cumul_chunk_counts[i];
+		total_chunks += zbud_cumul_chunk_counts[i];
+		sum_total_chunks += i * zbud_cumul_chunk_counts[i];
+		if (i == 21)
+			total_chunks_lte_21 = total_chunks;
+		if (i == 32)
+			total_chunks_lte_32 = total_chunks;
+		if (i == 42)
+			total_chunks_lte_42 = total_chunks;
+	}
+	p += sprintf(p, "<=21:%lu <=32:%lu <=42:%lu, mean:%lu\n",
+		total_chunks_lte_21, total_chunks_lte_32, total_chunks_lte_42,
+		chunks == 0 ? 0 : sum_total_chunks / chunks);
+	return p - buf;
+}
+#endif
+
+/**********
+ * This "zv" PAM implementation combines the TLSF-based xvMalloc
+ * with lzo1x compression to maximize the amount of data that can
+ * be packed into a physical page.
+ *
+ * Zv represents a PAM page with the index and object (plus a "size" value
+ * necessary for decompression) immediately preceding the compressed data.
+ */
+
+#define ZVH_SENTINEL  0x43214321
+
+struct zv_hdr {
+	uint32_t pool_id;
+	struct tmem_oid oid;
+	uint32_t index;
+	DECL_SENTINEL
+};
+
+/* rudimentary policy limits */
+/* total number of persistent pages may not exceed this percentage */
+static unsigned int zv_page_count_policy_percent = 75;
+/*
+ * byte count defining poor compression; pages with greater zsize will be
+ * rejected
+ */
+static unsigned int zv_max_zsize = (PAGE_SIZE / 8) * 7;
+/*
+ * byte count defining poor *mean* compression; pages with greater zsize
+ * will be rejected until sufficient better-compressed pages are accepted
+ * driving the mean below this threshold
+ */
+static unsigned int zv_max_mean_zsize = (PAGE_SIZE / 8) * 5;
+
+static unsigned long zv_curr_dist_counts[NCHUNKS];
+static unsigned long zv_cumul_dist_counts[NCHUNKS];
+
+static struct zv_hdr *zv_create(struct xv_pool *xvpool, uint32_t pool_id,
+				struct tmem_oid *oid, uint32_t index,
+				void *cdata, unsigned clen)
+{
+	struct page *page;
+	struct zv_hdr *zv = NULL;
+	uint32_t offset;
+	int alloc_size = clen + sizeof(struct zv_hdr);
+	int chunks = (alloc_size + (CHUNK_SIZE - 1)) >> CHUNK_SHIFT;
+	int ret;
+
+	BUG_ON(!irqs_disabled());
+	BUG_ON(chunks >= NCHUNKS);
+	ret = xv_malloc(xvpool, alloc_size,
+			&page, &offset, ZCACHE_GFP_MASK);
+	if (unlikely(ret))
+		goto out;
+	zv_curr_dist_counts[chunks]++;
+	zv_cumul_dist_counts[chunks]++;
+	zv = kmap_atomic(page, KM_USER0) + offset;
+	zv->index = index;
+	zv->oid = *oid;
+	zv->pool_id = pool_id;
+	SET_SENTINEL(zv, ZVH);
+	memcpy((char *)zv + sizeof(struct zv_hdr), cdata, clen);
+	kunmap_atomic(zv, KM_USER0);
+out:
+	return zv;
+}
+
+static void zv_free(struct xv_pool *xvpool, struct zv_hdr *zv)
+{
+	unsigned long flags;
+	struct page *page;
+	uint32_t offset;
+	uint16_t size = xv_get_object_size(zv);
+	int chunks = (size + (CHUNK_SIZE - 1)) >> CHUNK_SHIFT;
+
+	ASSERT_SENTINEL(zv, ZVH);
+	BUG_ON(chunks >= NCHUNKS);
+	zv_curr_dist_counts[chunks]--;
+	size -= sizeof(*zv);
+	BUG_ON(size == 0);
+	INVERT_SENTINEL(zv, ZVH);
+	page = virt_to_page(zv);
+	offset = (unsigned long)zv & ~PAGE_MASK;
+	local_irq_save(flags);
+	xv_free(xvpool, page, offset);
+	local_irq_restore(flags);
+}
+
+static void zv_decompress(struct page *page, struct zv_hdr *zv)
+{
+	size_t clen = PAGE_SIZE;
+	char *to_va;
+	unsigned size;
+	int ret;
+
+	ASSERT_SENTINEL(zv, ZVH);
+	size = xv_get_object_size(zv) - sizeof(*zv);
+	BUG_ON(size == 0);
+	to_va = kmap_atomic(page, KM_USER0);
+	ret = lzo1x_decompress_safe((char *)zv + sizeof(*zv),
+					size, to_va, &clen);
+	kunmap_atomic(to_va, KM_USER0);
+	BUG_ON(ret != LZO_E_OK);
+	BUG_ON(clen != PAGE_SIZE);
+}
+
+#ifdef CONFIG_SYSFS
+/*
+ * show a distribution of compression stats for zv pages.
+ */
+
+static int zv_curr_dist_counts_show(char *buf)
+{
+	unsigned long i, n, chunks = 0, sum_total_chunks = 0;
+	char *p = buf;
+
+	for (i = 0; i < NCHUNKS; i++) {
+		n = zv_curr_dist_counts[i];
+		p += sprintf(p, "%lu ", n);
+		chunks += n;
+		sum_total_chunks += i * n;
+	}
+	p += sprintf(p, "mean:%lu\n",
+		chunks == 0 ? 0 : sum_total_chunks / chunks);
+	return p - buf;
+}
+
+static int zv_cumul_dist_counts_show(char *buf)
+{
+	unsigned long i, n, chunks = 0, sum_total_chunks = 0;
+	char *p = buf;
+
+	for (i = 0; i < NCHUNKS; i++) {
+		n = zv_cumul_dist_counts[i];
+		p += sprintf(p, "%lu ", n);
+		chunks += n;
+		sum_total_chunks += i * n;
+	}
+	p += sprintf(p, "mean:%lu\n",
+		chunks == 0 ? 0 : sum_total_chunks / chunks);
+	return p - buf;
+}
+
+/*
+ * setting zv_max_zsize via sysfs causes all persistent (e.g. swap)
+ * pages that don't compress to less than this value (including metadata
+ * overhead) to be rejected.  We don't allow the value to get too close
+ * to PAGE_SIZE.
+ */
+static ssize_t zv_max_zsize_show(struct kobject *kobj,
+				    struct kobj_attribute *attr,
+				    char *buf)
+{
+	return sprintf(buf, "%u\n", zv_max_zsize);
+}
+
+static ssize_t zv_max_zsize_store(struct kobject *kobj,
+				    struct kobj_attribute *attr,
+				    const char *buf, size_t count)
+{
+	unsigned long val;
+	int err;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	err = strict_strtoul(buf, 10, &val);
+	if (err || (val == 0) || (val > (PAGE_SIZE / 8) * 7))
+		return -EINVAL;
+	zv_max_zsize = val;
+	return count;
+}
+
+/*
+ * setting zv_max_mean_zsize via sysfs causes all persistent (e.g. swap)
+ * pages that don't compress to less than this value (including metadata
+ * overhead) to be rejected UNLESS the mean compression is also smaller
+ * than this value.  In other words, we are load-balancing-by-zsize the
+ * accepted pages.  Again, we don't allow the value to get too close
+ * to PAGE_SIZE.
+ */
+static ssize_t zv_max_mean_zsize_show(struct kobject *kobj,
+				    struct kobj_attribute *attr,
+				    char *buf)
+{
+	return sprintf(buf, "%u\n", zv_max_mean_zsize);
+}
+
+static ssize_t zv_max_mean_zsize_store(struct kobject *kobj,
+				    struct kobj_attribute *attr,
+				    const char *buf, size_t count)
+{
+	unsigned long val;
+	int err;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	err = strict_strtoul(buf, 10, &val);
+	if (err || (val == 0) || (val > (PAGE_SIZE / 8) * 7))
+		return -EINVAL;
+	zv_max_mean_zsize = val;
+	return count;
+}
+
+/*
+ * setting zv_page_count_policy_percent via sysfs sets an upper bound of
+ * persistent (e.g. swap) pages that will be retained according to:
+ *     (zv_page_count_policy_percent * totalram_pages) / 100)
+ * when that limit is reached, further puts will be rejected (until
+ * some pages have been flushed).  Note that, due to compression,
+ * this number may exceed 100; it defaults to 75 and we set an
+ * arbitary limit of 150.  A poor choice will almost certainly result
+ * in OOM's, so this value should only be changed prudently.
+ */
+static ssize_t zv_page_count_policy_percent_show(struct kobject *kobj,
+						 struct kobj_attribute *attr,
+						 char *buf)
+{
+	return sprintf(buf, "%u\n", zv_page_count_policy_percent);
+}
+
+static ssize_t zv_page_count_policy_percent_store(struct kobject *kobj,
+						  struct kobj_attribute *attr,
+						  const char *buf, size_t count)
+{
+	unsigned long val;
+	int err;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	err = strict_strtoul(buf, 10, &val);
+	if (err || (val == 0) || (val > 150))
+		return -EINVAL;
+	zv_page_count_policy_percent = val;
+	return count;
+}
+
+static struct kobj_attribute zcache_zv_max_zsize_attr = {
+		.attr = { .name = "zv_max_zsize", .mode = 0644 },
+		.show = zv_max_zsize_show,
+		.store = zv_max_zsize_store,
+};
+
+static struct kobj_attribute zcache_zv_max_mean_zsize_attr = {
+		.attr = { .name = "zv_max_mean_zsize", .mode = 0644 },
+		.show = zv_max_mean_zsize_show,
+		.store = zv_max_mean_zsize_store,
+};
+
+static struct kobj_attribute zcache_zv_page_count_policy_percent_attr = {
+		.attr = { .name = "zv_page_count_policy_percent",
+			  .mode = 0644 },
+		.show = zv_page_count_policy_percent_show,
+		.store = zv_page_count_policy_percent_store,
+};
+#endif
+
+/*
+ * zcache core code starts here
+ */
+
+/* useful stats not collected by cleancache or frontswap */
+static unsigned long zcache_flush_total;
+static unsigned long zcache_flush_found;
+static unsigned long zcache_flobj_total;
+static unsigned long zcache_flobj_found;
+static unsigned long zcache_failed_eph_puts;
+static unsigned long zcache_failed_pers_puts;
+
+/*
+ * Tmem operations assume the poolid implies the invoking client.
+ * Zcache only has one client (the kernel itself): LOCAL_CLIENT.
+ * RAMster has each client numbered by cluster node, and a KVM version
+ * of zcache would have one client per guest and each client might
+ * have a poolid==N.
+ */
+static struct tmem_pool *zcache_get_pool_by_id(uint16_t cli_id, uint16_t poolid)
+{
+	struct tmem_pool *pool = NULL;
+	struct zcache_client *cli = NULL;
+
+	if (cli_id == LOCAL_CLIENT)
+		cli = &zcache_host;
+	else {
+		if (cli_id >= MAX_CLIENTS)
+			goto out;
+		cli = &zcache_clients[cli_id];
+		if (cli == NULL)
+			goto out;
+		atomic_inc(&cli->refcount);
+	}
+	if (poolid < MAX_POOLS_PER_CLIENT) {
+		pool = cli->tmem_pools[poolid];
+		if (pool != NULL)
+			atomic_inc(&pool->refcount);
+	}
+out:
+	return pool;
+}
+
+static void zcache_put_pool(struct tmem_pool *pool)
+{
+	struct zcache_client *cli = NULL;
+
+	if (pool == NULL)
+		BUG();
+	cli = pool->client;
+	atomic_dec(&pool->refcount);
+	atomic_dec(&cli->refcount);
+}
+
+int zcache_new_client(uint16_t cli_id)
+{
+	struct zcache_client *cli = NULL;
+	int ret = -1;
+
+	if (cli_id == LOCAL_CLIENT)
+		cli = &zcache_host;
+	else if ((unsigned int)cli_id < MAX_CLIENTS)
+		cli = &zcache_clients[cli_id];
+	if (cli == NULL)
+		goto out;
+	if (cli->allocated)
+		goto out;
+	cli->allocated = 1;
+#ifdef CONFIG_FRONTSWAP
+	cli->xvpool = xv_create_pool();
+	if (cli->xvpool == NULL)
+		goto out;
+#endif
+	ret = 0;
+out:
+	return ret;
+}
+
+/* counters for debugging */
+static unsigned long zcache_failed_get_free_pages;
+static unsigned long zcache_failed_alloc;
+static unsigned long zcache_put_to_flush;
+
+/*
+ * for now, used named slabs so can easily track usage; later can
+ * either just use kmalloc, or perhaps add a slab-like allocator
+ * to more carefully manage total memory utilization
+ */
+static struct kmem_cache *zcache_objnode_cache;
+static struct kmem_cache *zcache_obj_cache;
+static atomic_t zcache_curr_obj_count = ATOMIC_INIT(0);
+static unsigned long zcache_curr_obj_count_max;
+static atomic_t zcache_curr_objnode_count = ATOMIC_INIT(0);
+static unsigned long zcache_curr_objnode_count_max;
+
+/*
+ * to avoid memory allocation recursion (e.g. due to direct reclaim), we
+ * preload all necessary data structures so the hostops callbacks never
+ * actually do a malloc
+ */
+struct zcache_preload {
+	void *page;
+	struct tmem_obj *obj;
+	int nr;
+	struct tmem_objnode *objnodes[OBJNODE_TREE_MAX_PATH];
+};
+static DEFINE_PER_CPU(struct zcache_preload, zcache_preloads) = { 0, };
+
+static int zcache_do_preload(struct tmem_pool *pool)
+{
+	struct zcache_preload *kp;
+	struct tmem_objnode *objnode;
+	struct tmem_obj *obj;
+	void *page;
+	int ret = -ENOMEM;
+
+	if (unlikely(zcache_objnode_cache == NULL))
+		goto out;
+	if (unlikely(zcache_obj_cache == NULL))
+		goto out;
+	preempt_disable();
+	kp = &__get_cpu_var(zcache_preloads);
+	while (kp->nr < ARRAY_SIZE(kp->objnodes)) {
+		preempt_enable_no_resched();
+		objnode = kmem_cache_alloc(zcache_objnode_cache,
+				ZCACHE_GFP_MASK);
+		if (unlikely(objnode == NULL)) {
+			zcache_failed_alloc++;
+			goto out;
+		}
+		preempt_disable();
+		kp = &__get_cpu_var(zcache_preloads);
+		if (kp->nr < ARRAY_SIZE(kp->objnodes))
+			kp->objnodes[kp->nr++] = objnode;
+		else
+			kmem_cache_free(zcache_objnode_cache, objnode);
+	}
+	preempt_enable_no_resched();
+	obj = kmem_cache_alloc(zcache_obj_cache, ZCACHE_GFP_MASK);
+	if (unlikely(obj == NULL)) {
+		zcache_failed_alloc++;
+		goto out;
+	}
+	page = (void *)__get_free_page(ZCACHE_GFP_MASK);
+	if (unlikely(page == NULL)) {
+		zcache_failed_get_free_pages++;
+		kmem_cache_free(zcache_obj_cache, obj);
+		goto out;
+	}
+	preempt_disable();
+	kp = &__get_cpu_var(zcache_preloads);
+	if (kp->obj == NULL)
+		kp->obj = obj;
+	else
+		kmem_cache_free(zcache_obj_cache, obj);
+	if (kp->page == NULL)
+		kp->page = page;
+	else
+		free_page((unsigned long)page);
+	ret = 0;
+out:
+	return ret;
+}
+
+static void *zcache_get_free_page(void)
+{
+	struct zcache_preload *kp;
+	void *page;
+
+	kp = &__get_cpu_var(zcache_preloads);
+	page = kp->page;
+	BUG_ON(page == NULL);
+	kp->page = NULL;
+	return page;
+}
+
+static void zcache_free_page(void *p)
+{
+	free_page((unsigned long)p);
+}
+
+/*
+ * zcache implementation for tmem host ops
+ */
+
+static struct tmem_objnode *zcache_objnode_alloc(struct tmem_pool *pool)
+{
+	struct tmem_objnode *objnode = NULL;
+	unsigned long count;
+	struct zcache_preload *kp;
+
+	kp = &__get_cpu_var(zcache_preloads);
+	if (kp->nr <= 0)
+		goto out;
+	objnode = kp->objnodes[kp->nr - 1];
+	BUG_ON(objnode == NULL);
+	kp->objnodes[kp->nr - 1] = NULL;
+	kp->nr--;
+	count = atomic_inc_return(&zcache_curr_objnode_count);
+	if (count > zcache_curr_objnode_count_max)
+		zcache_curr_objnode_count_max = count;
+out:
+	return objnode;
+}
+
+static void zcache_objnode_free(struct tmem_objnode *objnode,
+					struct tmem_pool *pool)
+{
+	atomic_dec(&zcache_curr_objnode_count);
+	BUG_ON(atomic_read(&zcache_curr_objnode_count) < 0);
+	kmem_cache_free(zcache_objnode_cache, objnode);
+}
+
+static struct tmem_obj *zcache_obj_alloc(struct tmem_pool *pool)
+{
+	struct tmem_obj *obj = NULL;
+	unsigned long count;
+	struct zcache_preload *kp;
+
+	kp = &__get_cpu_var(zcache_preloads);
+	obj = kp->obj;
+	BUG_ON(obj == NULL);
+	kp->obj = NULL;
+	count = atomic_inc_return(&zcache_curr_obj_count);
+	if (count > zcache_curr_obj_count_max)
+		zcache_curr_obj_count_max = count;
+	return obj;
+}
+
+static void zcache_obj_free(struct tmem_obj *obj, struct tmem_pool *pool)
+{
+	atomic_dec(&zcache_curr_obj_count);
+	BUG_ON(atomic_read(&zcache_curr_obj_count) < 0);
+	kmem_cache_free(zcache_obj_cache, obj);
+}
+
+static struct tmem_hostops zcache_hostops = {
+	.obj_alloc = zcache_obj_alloc,
+	.obj_free = zcache_obj_free,
+	.objnode_alloc = zcache_objnode_alloc,
+	.objnode_free = zcache_objnode_free,
+};
+
+/*
+ * zcache implementations for PAM page descriptor ops
+ */
+
+static atomic_t zcache_curr_eph_pampd_count = ATOMIC_INIT(0);
+static unsigned long zcache_curr_eph_pampd_count_max;
+static atomic_t zcache_curr_pers_pampd_count = ATOMIC_INIT(0);
+static unsigned long zcache_curr_pers_pampd_count_max;
+
+/* forward reference */
+static int zcache_compress(struct page *from, void **out_va, size_t *out_len);
+
+static void *zcache_pampd_create(char *data, size_t size, bool raw, int eph,
+				struct tmem_pool *pool, struct tmem_oid *oid,
+				 uint32_t index)
+{
+	void *pampd = NULL, *cdata;
+	size_t clen;
+	int ret;
+	unsigned long count;
+	struct page *page = (struct page *)(data);
+	struct zcache_client *cli = pool->client;
+	uint16_t client_id = get_client_id_from_client(cli);
+	unsigned long zv_mean_zsize;
+	unsigned long curr_pers_pampd_count;
+	u64 total_zsize;
+
+	if (eph) {
+		ret = zcache_compress(page, &cdata, &clen);
+		if (ret == 0)
+			goto out;
+		if (clen == 0 || clen > zbud_max_buddy_size()) {
+			zcache_compress_poor++;
+			goto out;
+		}
+		pampd = (void *)zbud_create(client_id, pool->pool_id, oid,
+						index, page, cdata, clen);
+		if (pampd != NULL) {
+			count = atomic_inc_return(&zcache_curr_eph_pampd_count);
+			if (count > zcache_curr_eph_pampd_count_max)
+				zcache_curr_eph_pampd_count_max = count;
+		}
+	} else {
+		curr_pers_pampd_count =
+			atomic_read(&zcache_curr_pers_pampd_count);
+		if (curr_pers_pampd_count >
+		    (zv_page_count_policy_percent * totalram_pages) / 100)
+			goto out;
+		ret = zcache_compress(page, &cdata, &clen);
+		if (ret == 0)
+			goto out;
+		/* reject if compression is too poor */
+		if (clen > zv_max_zsize) {
+			zcache_compress_poor++;
+			goto out;
+		}
+		/* reject if mean compression is too poor */
+		if ((clen > zv_max_mean_zsize) && (curr_pers_pampd_count > 0)) {
+			total_zsize = xv_get_total_size_bytes(cli->xvpool);
+			zv_mean_zsize = div_u64(total_zsize,
+						curr_pers_pampd_count);
+			if (zv_mean_zsize > zv_max_mean_zsize) {
+				zcache_mean_compress_poor++;
+				goto out;
+			}
+		}
+		pampd = (void *)zv_create(cli->xvpool, pool->pool_id,
+						oid, index, cdata, clen);
+		if (pampd == NULL)
+			goto out;
+		count = atomic_inc_return(&zcache_curr_pers_pampd_count);
+		if (count > zcache_curr_pers_pampd_count_max)
+			zcache_curr_pers_pampd_count_max = count;
+	}
+out:
+	return pampd;
+}
+
+/*
+ * fill the pageframe corresponding to the struct page with the data
+ * from the passed pampd
+ */
+static int zcache_pampd_get_data(char *data, size_t *bufsize, bool raw,
+					void *pampd, struct tmem_pool *pool,
+					struct tmem_oid *oid, uint32_t index)
+{
+	int ret = 0;
+
+	BUG_ON(is_ephemeral(pool));
+	zv_decompress((struct page *)(data), pampd);
+	return ret;
+}
+
+/*
+ * fill the pageframe corresponding to the struct page with the data
+ * from the passed pampd
+ */
+static int zcache_pampd_get_data_and_free(char *data, size_t *bufsize, bool raw,
+					void *pampd, struct tmem_pool *pool,
+					struct tmem_oid *oid, uint32_t index)
+{
+	int ret = 0;
+
+	BUG_ON(!is_ephemeral(pool));
+	zbud_decompress((struct page *)(data), pampd);
+	zbud_free_and_delist((struct zbud_hdr *)pampd);
+	atomic_dec(&zcache_curr_eph_pampd_count);
+	return ret;
+}
+
+/*
+ * free the pampd and remove it from any zcache lists
+ * pampd must no longer be pointed to from any tmem data structures!
+ */
+static void zcache_pampd_free(void *pampd, struct tmem_pool *pool,
+				struct tmem_oid *oid, uint32_t index)
+{
+	struct zcache_client *cli = pool->client;
+
+	if (is_ephemeral(pool)) {
+		zbud_free_and_delist((struct zbud_hdr *)pampd);
+		atomic_dec(&zcache_curr_eph_pampd_count);
+		BUG_ON(atomic_read(&zcache_curr_eph_pampd_count) < 0);
+	} else {
+		zv_free(cli->xvpool, (struct zv_hdr *)pampd);
+		atomic_dec(&zcache_curr_pers_pampd_count);
+		BUG_ON(atomic_read(&zcache_curr_pers_pampd_count) < 0);
+	}
+}
+
+static void zcache_pampd_free_obj(struct tmem_pool *pool, struct tmem_obj *obj)
+{
+}
+
+static void zcache_pampd_new_obj(struct tmem_obj *obj)
+{
+}
+
+static int zcache_pampd_replace_in_obj(void *pampd, struct tmem_obj *obj)
+{
+	return -1;
+}
+
+static bool zcache_pampd_is_remote(void *pampd)
+{
+	return 0;
+}
+
+static struct tmem_pamops zcache_pamops = {
+	.create = zcache_pampd_create,
+	.get_data = zcache_pampd_get_data,
+	.get_data_and_free = zcache_pampd_get_data_and_free,
+	.free = zcache_pampd_free,
+	.free_obj = zcache_pampd_free_obj,
+	.new_obj = zcache_pampd_new_obj,
+	.replace_in_obj = zcache_pampd_replace_in_obj,
+	.is_remote = zcache_pampd_is_remote,
+};
+
+/*
+ * zcache compression/decompression and related per-cpu stuff
+ */
+
+#define LZO_WORKMEM_BYTES LZO1X_1_MEM_COMPRESS
+#define LZO_DSTMEM_PAGE_ORDER 1
+static DEFINE_PER_CPU(unsigned char *, zcache_workmem);
+static DEFINE_PER_CPU(unsigned char *, zcache_dstmem);
+
+static int zcache_compress(struct page *from, void **out_va, size_t *out_len)
+{
+	int ret = 0;
+	unsigned char *dmem = __get_cpu_var(zcache_dstmem);
+	unsigned char *wmem = __get_cpu_var(zcache_workmem);
+	char *from_va;
+
+	BUG_ON(!irqs_disabled());
+	if (unlikely(dmem == NULL || wmem == NULL))
+		goto out;  /* no buffer, so can't compress */
+	from_va = kmap_atomic(from, KM_USER0);
+	mb();
+	ret = lzo1x_1_compress(from_va, PAGE_SIZE, dmem, out_len, wmem);
+	BUG_ON(ret != LZO_E_OK);
+	*out_va = dmem;
+	kunmap_atomic(from_va, KM_USER0);
+	ret = 1;
+out:
+	return ret;
+}
+
+
+static int zcache_cpu_notifier(struct notifier_block *nb,
+				unsigned long action, void *pcpu)
+{
+	int cpu = (long)pcpu;
+	struct zcache_preload *kp;
+
+	switch (action) {
+	case CPU_UP_PREPARE:
+		per_cpu(zcache_dstmem, cpu) = (void *)__get_free_pages(
+			GFP_KERNEL | __GFP_REPEAT,
+			LZO_DSTMEM_PAGE_ORDER),
+		per_cpu(zcache_workmem, cpu) =
+			kzalloc(LZO1X_MEM_COMPRESS,
+				GFP_KERNEL | __GFP_REPEAT);
+		break;
+	case CPU_DEAD:
+	case CPU_UP_CANCELED:
+		free_pages((unsigned long)per_cpu(zcache_dstmem, cpu),
+				LZO_DSTMEM_PAGE_ORDER);
+		per_cpu(zcache_dstmem, cpu) = NULL;
+		kfree(per_cpu(zcache_workmem, cpu));
+		per_cpu(zcache_workmem, cpu) = NULL;
+		kp = &per_cpu(zcache_preloads, cpu);
+		while (kp->nr) {
+			kmem_cache_free(zcache_objnode_cache,
+					kp->objnodes[kp->nr - 1]);
+			kp->objnodes[kp->nr - 1] = NULL;
+			kp->nr--;
+		}
+		if (kp->obj) {
+			kmem_cache_free(zcache_obj_cache, kp->obj);
+			kp->obj = NULL;
+		}
+		if (kp->page) {
+			free_page((unsigned long)kp->page);
+			kp->page = NULL;
+		}
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_OK;
+}
+
+static struct notifier_block zcache_cpu_notifier_block = {
+	.notifier_call = zcache_cpu_notifier
+};
+
+#ifdef CONFIG_SYSFS
+#define ZCACHE_SYSFS_RO(_name) \
+	static ssize_t zcache_##_name##_show(struct kobject *kobj, \
+				struct kobj_attribute *attr, char *buf) \
+	{ \
+		return sprintf(buf, "%lu\n", zcache_##_name); \
+	} \
+	static struct kobj_attribute zcache_##_name##_attr = { \
+		.attr = { .name = __stringify(_name), .mode = 0444 }, \
+		.show = zcache_##_name##_show, \
+	}
+
+#define ZCACHE_SYSFS_RO_ATOMIC(_name) \
+	static ssize_t zcache_##_name##_show(struct kobject *kobj, \
+				struct kobj_attribute *attr, char *buf) \
+	{ \
+	    return sprintf(buf, "%d\n", atomic_read(&zcache_##_name)); \
+	} \
+	static struct kobj_attribute zcache_##_name##_attr = { \
+		.attr = { .name = __stringify(_name), .mode = 0444 }, \
+		.show = zcache_##_name##_show, \
+	}
+
+#define ZCACHE_SYSFS_RO_CUSTOM(_name, _func) \
+	static ssize_t zcache_##_name##_show(struct kobject *kobj, \
+				struct kobj_attribute *attr, char *buf) \
+	{ \
+	    return _func(buf); \
+	} \
+	static struct kobj_attribute zcache_##_name##_attr = { \
+		.attr = { .name = __stringify(_name), .mode = 0444 }, \
+		.show = zcache_##_name##_show, \
+	}
+
+ZCACHE_SYSFS_RO(curr_obj_count_max);
+ZCACHE_SYSFS_RO(curr_objnode_count_max);
+ZCACHE_SYSFS_RO(flush_total);
+ZCACHE_SYSFS_RO(flush_found);
+ZCACHE_SYSFS_RO(flobj_total);
+ZCACHE_SYSFS_RO(flobj_found);
+ZCACHE_SYSFS_RO(failed_eph_puts);
+ZCACHE_SYSFS_RO(failed_pers_puts);
+ZCACHE_SYSFS_RO(zbud_curr_zbytes);
+ZCACHE_SYSFS_RO(zbud_cumul_zpages);
+ZCACHE_SYSFS_RO(zbud_cumul_zbytes);
+ZCACHE_SYSFS_RO(zbud_buddied_count);
+ZCACHE_SYSFS_RO(zbpg_unused_list_count);
+ZCACHE_SYSFS_RO(evicted_raw_pages);
+ZCACHE_SYSFS_RO(evicted_unbuddied_pages);
+ZCACHE_SYSFS_RO(evicted_buddied_pages);
+ZCACHE_SYSFS_RO(failed_get_free_pages);
+ZCACHE_SYSFS_RO(failed_alloc);
+ZCACHE_SYSFS_RO(put_to_flush);
+ZCACHE_SYSFS_RO(compress_poor);
+ZCACHE_SYSFS_RO(mean_compress_poor);
+ZCACHE_SYSFS_RO_ATOMIC(zbud_curr_raw_pages);
+ZCACHE_SYSFS_RO_ATOMIC(zbud_curr_zpages);
+ZCACHE_SYSFS_RO_ATOMIC(curr_obj_count);
+ZCACHE_SYSFS_RO_ATOMIC(curr_objnode_count);
+ZCACHE_SYSFS_RO_CUSTOM(zbud_unbuddied_list_counts,
+			zbud_show_unbuddied_list_counts);
+ZCACHE_SYSFS_RO_CUSTOM(zbud_cumul_chunk_counts,
+			zbud_show_cumul_chunk_counts);
+ZCACHE_SYSFS_RO_CUSTOM(zv_curr_dist_counts,
+			zv_curr_dist_counts_show);
+ZCACHE_SYSFS_RO_CUSTOM(zv_cumul_dist_counts,
+			zv_cumul_dist_counts_show);
+
+static struct attribute *zcache_attrs[] = {
+	&zcache_curr_obj_count_attr.attr,
+	&zcache_curr_obj_count_max_attr.attr,
+	&zcache_curr_objnode_count_attr.attr,
+	&zcache_curr_objnode_count_max_attr.attr,
+	&zcache_flush_total_attr.attr,
+	&zcache_flobj_total_attr.attr,
+	&zcache_flush_found_attr.attr,
+	&zcache_flobj_found_attr.attr,
+	&zcache_failed_eph_puts_attr.attr,
+	&zcache_failed_pers_puts_attr.attr,
+	&zcache_compress_poor_attr.attr,
+	&zcache_mean_compress_poor_attr.attr,
+	&zcache_zbud_curr_raw_pages_attr.attr,
+	&zcache_zbud_curr_zpages_attr.attr,
+	&zcache_zbud_curr_zbytes_attr.attr,
+	&zcache_zbud_cumul_zpages_attr.attr,
+	&zcache_zbud_cumul_zbytes_attr.attr,
+	&zcache_zbud_buddied_count_attr.attr,
+	&zcache_zbpg_unused_list_count_attr.attr,
+	&zcache_evicted_raw_pages_attr.attr,
+	&zcache_evicted_unbuddied_pages_attr.attr,
+	&zcache_evicted_buddied_pages_attr.attr,
+	&zcache_failed_get_free_pages_attr.attr,
+	&zcache_failed_alloc_attr.attr,
+	&zcache_put_to_flush_attr.attr,
+	&zcache_zbud_unbuddied_list_counts_attr.attr,
+	&zcache_zbud_cumul_chunk_counts_attr.attr,
+	&zcache_zv_curr_dist_counts_attr.attr,
+	&zcache_zv_cumul_dist_counts_attr.attr,
+	&zcache_zv_max_zsize_attr.attr,
+	&zcache_zv_max_mean_zsize_attr.attr,
+	&zcache_zv_page_count_policy_percent_attr.attr,
+	NULL,
+};
+
+static struct attribute_group zcache_attr_group = {
+	.attrs = zcache_attrs,
+	.name = "zcache",
+};
+
+#endif /* CONFIG_SYSFS */
+/*
+ * When zcache is disabled ("frozen"), pools can be created and destroyed,
+ * but all puts (and thus all other operations that require memory allocation)
+ * must fail.  If zcache is unfrozen, accepts puts, then frozen again,
+ * data consistency requires all puts while frozen to be converted into
+ * flushes.
+ */
+static bool zcache_freeze;
+
+/*
+ * zcache shrinker interface (only useful for ephemeral pages, so zbud only)
+ */
+static int shrink_zcache_memory(struct shrinker *shrink,
+				struct shrink_control *sc)
+{
+	int ret = -1;
+	int nr = sc->nr_to_scan;
+	gfp_t gfp_mask = sc->gfp_mask;
+
+	if (nr >= 0) {
+		if (!(gfp_mask & __GFP_FS))
+			/* does this case really need to be skipped? */
+			goto out;
+		zbud_evict_pages(nr);
+	}
+	ret = (int)atomic_read(&zcache_zbud_curr_raw_pages);
+out:
+	return ret;
+}
+
+static struct shrinker zcache_shrinker = {
+	.shrink = shrink_zcache_memory,
+	.seeks = DEFAULT_SEEKS,
+};
+
+/*
+ * zcache shims between cleancache/frontswap ops and tmem
+ */
+
+static int zcache_put_page(int cli_id, int pool_id, struct tmem_oid *oidp,
+				uint32_t index, struct page *page)
+{
+	struct tmem_pool *pool;
+	int ret = -1;
+
+	BUG_ON(!irqs_disabled());
+	pool = zcache_get_pool_by_id(cli_id, pool_id);
+	if (unlikely(pool == NULL))
+		goto out;
+	if (!zcache_freeze && zcache_do_preload(pool) == 0) {
+		/* preload does preempt_disable on success */
+		ret = tmem_put(pool, oidp, index, (char *)(page),
+				PAGE_SIZE, 0, is_ephemeral(pool));
+		if (ret < 0) {
+			if (is_ephemeral(pool))
+				zcache_failed_eph_puts++;
+			else
+				zcache_failed_pers_puts++;
+		}
+		zcache_put_pool(pool);
+		preempt_enable_no_resched();
+	} else {
+		zcache_put_to_flush++;
+		if (atomic_read(&pool->obj_count) > 0)
+			/* the put fails whether the flush succeeds or not */
+			(void)tmem_flush_page(pool, oidp, index);
+		zcache_put_pool(pool);
+	}
+out:
+	return ret;
+}
+
+static int zcache_get_page(int cli_id, int pool_id, struct tmem_oid *oidp,
+				uint32_t index, struct page *page)
+{
+	struct tmem_pool *pool;
+	int ret = -1;
+	unsigned long flags;
+	size_t size = PAGE_SIZE;
+
+	local_irq_save(flags);
+	pool = zcache_get_pool_by_id(cli_id, pool_id);
+	if (likely(pool != NULL)) {
+		if (atomic_read(&pool->obj_count) > 0)
+			ret = tmem_get(pool, oidp, index, (char *)(page),
+					&size, 0, is_ephemeral(pool));
+		zcache_put_pool(pool);
+	}
+	local_irq_restore(flags);
+	return ret;
+}
+
+static int zcache_flush_page(int cli_id, int pool_id,
+				struct tmem_oid *oidp, uint32_t index)
+{
+	struct tmem_pool *pool;
+	int ret = -1;
+	unsigned long flags;
+
+	local_irq_save(flags);
+	zcache_flush_total++;
+	pool = zcache_get_pool_by_id(cli_id, pool_id);
+	if (likely(pool != NULL)) {
+		if (atomic_read(&pool->obj_count) > 0)
+			ret = tmem_flush_page(pool, oidp, index);
+		zcache_put_pool(pool);
+	}
+	if (ret >= 0)
+		zcache_flush_found++;
+	local_irq_restore(flags);
+	return ret;
+}
+
+static int zcache_flush_object(int cli_id, int pool_id,
+				struct tmem_oid *oidp)
+{
+	struct tmem_pool *pool;
+	int ret = -1;
+	unsigned long flags;
+
+	local_irq_save(flags);
+	zcache_flobj_total++;
+	pool = zcache_get_pool_by_id(cli_id, pool_id);
+	if (likely(pool != NULL)) {
+		if (atomic_read(&pool->obj_count) > 0)
+			ret = tmem_flush_object(pool, oidp);
+		zcache_put_pool(pool);
+	}
+	if (ret >= 0)
+		zcache_flobj_found++;
+	local_irq_restore(flags);
+	return ret;
+}
+
+static int zcache_destroy_pool(int cli_id, int pool_id)
+{
+	struct tmem_pool *pool = NULL;
+	struct zcache_client *cli = NULL;
+	int ret = -1;
+
+	if (pool_id < 0)
+		goto out;
+	if (cli_id == LOCAL_CLIENT)
+		cli = &zcache_host;
+	else if ((unsigned int)cli_id < MAX_CLIENTS)
+		cli = &zcache_clients[cli_id];
+	if (cli == NULL)
+		goto out;
+	atomic_inc(&cli->refcount);
+	pool = cli->tmem_pools[pool_id];
+	if (pool == NULL)
+		goto out;
+	cli->tmem_pools[pool_id] = NULL;
+	/* wait for pool activity on other cpus to quiesce */
+	while (atomic_read(&pool->refcount) != 0)
+		;
+	atomic_dec(&cli->refcount);
+	local_bh_disable();
+	ret = tmem_destroy_pool(pool);
+	local_bh_enable();
+	kfree(pool);
+	pr_info("zcache: destroyed pool id=%d, cli_id=%d\n",
+			pool_id, cli_id);
+out:
+	return ret;
+}
+
+static int zcache_new_pool(uint16_t cli_id, uint32_t flags)
+{
+	int poolid = -1;
+	struct tmem_pool *pool;
+	struct zcache_client *cli = NULL;
+
+	if (cli_id == LOCAL_CLIENT)
+		cli = &zcache_host;
+	else if ((unsigned int)cli_id < MAX_CLIENTS)
+		cli = &zcache_clients[cli_id];
+	if (cli == NULL)
+		goto out;
+	atomic_inc(&cli->refcount);
+	pool = kmalloc(sizeof(struct tmem_pool), GFP_ATOMIC);
+	if (pool == NULL) {
+		pr_info("zcache: pool creation failed: out of memory\n");
+		goto out;
+	}
+
+	for (poolid = 0; poolid < MAX_POOLS_PER_CLIENT; poolid++)
+		if (cli->tmem_pools[poolid] == NULL)
+			break;
+	if (poolid >= MAX_POOLS_PER_CLIENT) {
+		pr_info("zcache: pool creation failed: max exceeded\n");
+		kfree(pool);
+		poolid = -1;
+		goto out;
+	}
+	atomic_set(&pool->refcount, 0);
+	pool->client = cli;
+	pool->pool_id = poolid;
+	tmem_new_pool(pool, flags);
+	cli->tmem_pools[poolid] = pool;
+	pr_info("zcache: created %s tmem pool, id=%d, client=%d\n",
+		flags & TMEM_POOL_PERSIST ? "persistent" : "ephemeral",
+		poolid, cli_id);
+out:
+	if (cli != NULL)
+		atomic_dec(&cli->refcount);
+	return poolid;
+}
+
+/**********
+ * Two kernel functionalities currently can be layered on top of tmem.
+ * These are "cleancache" which is used as a second-chance cache for clean
+ * page cache pages; and "frontswap" which is used for swap pages
+ * to avoid writes to disk.  A generic "shim" is provided here for each
+ * to translate in-kernel semantics to zcache semantics.
+ */
+
+#ifdef CONFIG_CLEANCACHE
+static void zcache_cleancache_put_page(int pool_id,
+					struct cleancache_filekey key,
+					pgoff_t index, struct page *page)
+{
+	u32 ind = (u32) index;
+	struct tmem_oid oid = *(struct tmem_oid *)&key;
+
+	if (likely(ind == index))
+		(void)zcache_put_page(LOCAL_CLIENT, pool_id, &oid, index, page);
+}
+
+static int zcache_cleancache_get_page(int pool_id,
+					struct cleancache_filekey key,
+					pgoff_t index, struct page *page)
+{
+	u32 ind = (u32) index;
+	struct tmem_oid oid = *(struct tmem_oid *)&key;
+	int ret = -1;
+
+	if (likely(ind == index))
+		ret = zcache_get_page(LOCAL_CLIENT, pool_id, &oid, index, page);
+	return ret;
+}
+
+static void zcache_cleancache_flush_page(int pool_id,
+					struct cleancache_filekey key,
+					pgoff_t index)
+{
+	u32 ind = (u32) index;
+	struct tmem_oid oid = *(struct tmem_oid *)&key;
+
+	if (likely(ind == index))
+		(void)zcache_flush_page(LOCAL_CLIENT, pool_id, &oid, ind);
+}
+
+static void zcache_cleancache_flush_inode(int pool_id,
+					struct cleancache_filekey key)
+{
+	struct tmem_oid oid = *(struct tmem_oid *)&key;
+
+	(void)zcache_flush_object(LOCAL_CLIENT, pool_id, &oid);
+}
+
+static void zcache_cleancache_flush_fs(int pool_id)
+{
+	if (pool_id >= 0)
+		(void)zcache_destroy_pool(LOCAL_CLIENT, pool_id);
+}
+
+static int zcache_cleancache_init_fs(size_t pagesize)
+{
+	BUG_ON(sizeof(struct cleancache_filekey) !=
+				sizeof(struct tmem_oid));
+	BUG_ON(pagesize != PAGE_SIZE);
+	return zcache_new_pool(LOCAL_CLIENT, 0);
+}
+
+static int zcache_cleancache_init_shared_fs(char *uuid, size_t pagesize)
+{
+	/* shared pools are unsupported and map to private */
+	BUG_ON(sizeof(struct cleancache_filekey) !=
+				sizeof(struct tmem_oid));
+	BUG_ON(pagesize != PAGE_SIZE);
+	return zcache_new_pool(LOCAL_CLIENT, 0);
+}
+
+static struct cleancache_ops zcache_cleancache_ops = {
+	.put_page = zcache_cleancache_put_page,
+	.get_page = zcache_cleancache_get_page,
+	.invalidate_page = zcache_cleancache_flush_page,
+	.invalidate_inode = zcache_cleancache_flush_inode,
+	.invalidate_fs = zcache_cleancache_flush_fs,
+	.init_shared_fs = zcache_cleancache_init_shared_fs,
+	.init_fs = zcache_cleancache_init_fs
+};
+
+struct cleancache_ops zcache_cleancache_register_ops(void)
+{
+	struct cleancache_ops old_ops =
+		cleancache_register_ops(&zcache_cleancache_ops);
+
+	return old_ops;
+}
+#endif
+
+#ifdef CONFIG_FRONTSWAP
+/* a single tmem poolid is used for all frontswap "types" (swapfiles) */
+static int zcache_frontswap_poolid = -1;
+
+/*
+ * Swizzling increases objects per swaptype, increasing tmem concurrency
+ * for heavy swaploads.  Later, larger nr_cpus -> larger SWIZ_BITS
+ * Setting SWIZ_BITS to 27 basically reconstructs the swap entry from
+ * frontswap_get_page()
+ */
+#define SWIZ_BITS		27
+#define SWIZ_MASK		((1 << SWIZ_BITS) - 1)
+#define _oswiz(_type, _ind)	((_type << SWIZ_BITS) | (_ind & SWIZ_MASK))
+#define iswiz(_ind)		(_ind >> SWIZ_BITS)
+
+static inline struct tmem_oid oswiz(unsigned type, u32 ind)
+{
+	struct tmem_oid oid = { .oid = { 0 } };
+	oid.oid[0] = _oswiz(type, ind);
+	return oid;
+}
+
+static int zcache_frontswap_put_page(unsigned type, pgoff_t offset,
+				   struct page *page)
+{
+	u64 ind64 = (u64)offset;
+	u32 ind = (u32)offset;
+	struct tmem_oid oid = oswiz(type, ind);
+	int ret = -1;
+	unsigned long flags;
+
+	BUG_ON(!PageLocked(page));
+	if (likely(ind64 == ind)) {
+		local_irq_save(flags);
+		ret = zcache_put_page(LOCAL_CLIENT, zcache_frontswap_poolid,
+					&oid, iswiz(ind), page);
+		local_irq_restore(flags);
+	}
+	return ret;
+}
+
+/* returns 0 if the page was successfully gotten from frontswap, -1 if
+ * was not present (should never happen!) */
+static int zcache_frontswap_get_page(unsigned type, pgoff_t offset,
+				   struct page *page)
+{
+	u64 ind64 = (u64)offset;
+	u32 ind = (u32)offset;
+	struct tmem_oid oid = oswiz(type, ind);
+	int ret = -1;
+
+	BUG_ON(!PageLocked(page));
+	if (likely(ind64 == ind))
+		ret = zcache_get_page(LOCAL_CLIENT, zcache_frontswap_poolid,
+					&oid, iswiz(ind), page);
+	return ret;
+}
+
+/* flush a single page from frontswap */
+static void zcache_frontswap_flush_page(unsigned type, pgoff_t offset)
+{
+	u64 ind64 = (u64)offset;
+	u32 ind = (u32)offset;
+	struct tmem_oid oid = oswiz(type, ind);
+
+	if (likely(ind64 == ind))
+		(void)zcache_flush_page(LOCAL_CLIENT, zcache_frontswap_poolid,
+					&oid, iswiz(ind));
+}
+
+/* flush all pages from the passed swaptype */
+static void zcache_frontswap_flush_area(unsigned type)
+{
+	struct tmem_oid oid;
+	int ind;
+
+	for (ind = SWIZ_MASK; ind >= 0; ind--) {
+		oid = oswiz(type, ind);
+		(void)zcache_flush_object(LOCAL_CLIENT,
+						zcache_frontswap_poolid, &oid);
+	}
+}
+
+static void zcache_frontswap_init(unsigned ignored)
+{
+	/* a single tmem poolid is used for all frontswap "types" (swapfiles) */
+	if (zcache_frontswap_poolid < 0)
+		zcache_frontswap_poolid =
+			zcache_new_pool(LOCAL_CLIENT, TMEM_POOL_PERSIST);
+}
+
+static struct frontswap_ops zcache_frontswap_ops = {
+	.put_page = zcache_frontswap_put_page,
+	.get_page = zcache_frontswap_get_page,
+	.invalidate_page = zcache_frontswap_flush_page,
+	.invalidate_area = zcache_frontswap_flush_area,
+	.init = zcache_frontswap_init
+};
+
+struct frontswap_ops zcache_frontswap_register_ops(void)
+{
+	struct frontswap_ops old_ops =
+		frontswap_register_ops(&zcache_frontswap_ops);
+
+	return old_ops;
+}
+#endif
+
+/*
+ * zcache initialization
+ * NOTE FOR NOW zcache MUST BE PROVIDED AS A KERNEL BOOT PARAMETER OR
+ * NOTHING HAPPENS!
+ */
+
+static int zcache_enabled;
+
+static int __init enable_zcache(char *s)
+{
+	zcache_enabled = 1;
+	return 1;
+}
+__setup("zcache", enable_zcache);
+
+/* allow independent dynamic disabling of cleancache and frontswap */
+
+static int use_cleancache = 1;
+
+static int __init no_cleancache(char *s)
+{
+	use_cleancache = 0;
+	return 1;
+}
+
+__setup("nocleancache", no_cleancache);
+
+static int use_frontswap = 1;
+
+static int __init no_frontswap(char *s)
+{
+	use_frontswap = 0;
+	return 1;
+}
+
+__setup("nofrontswap", no_frontswap);
+
+static int __init zcache_init(void)
+{
+	int ret = 0;
+
+#ifdef CONFIG_SYSFS
+	ret = sysfs_create_group(mm_kobj, &zcache_attr_group);
+	if (ret) {
+		pr_err("zcache: can't create sysfs\n");
+		goto out;
+	}
+#endif /* CONFIG_SYSFS */
+#if defined(CONFIG_CLEANCACHE) || defined(CONFIG_FRONTSWAP)
+	if (zcache_enabled) {
+		unsigned int cpu;
+
+		tmem_register_hostops(&zcache_hostops);
+		tmem_register_pamops(&zcache_pamops);
+		ret = register_cpu_notifier(&zcache_cpu_notifier_block);
+		if (ret) {
+			pr_err("zcache: can't register cpu notifier\n");
+			goto out;
+		}
+		for_each_online_cpu(cpu) {
+			void *pcpu = (void *)(long)cpu;
+			zcache_cpu_notifier(&zcache_cpu_notifier_block,
+				CPU_UP_PREPARE, pcpu);
+		}
+	}
+	zcache_objnode_cache = kmem_cache_create("zcache_objnode",
+				sizeof(struct tmem_objnode), 0, 0, NULL);
+	zcache_obj_cache = kmem_cache_create("zcache_obj",
+				sizeof(struct tmem_obj), 0, 0, NULL);
+	ret = zcache_new_client(LOCAL_CLIENT);
+	if (ret) {
+		pr_err("zcache: can't create client\n");
+		goto out;
+	}
+#endif
+#ifdef CONFIG_CLEANCACHE
+	if (zcache_enabled && use_cleancache) {
+		struct cleancache_ops old_ops;
+
+		zbud_init();
+		register_shrinker(&zcache_shrinker);
+		old_ops = zcache_cleancache_register_ops();
+		pr_info("zcache: cleancache enabled using kernel "
+			"transcendent memory and compression buddies\n");
+		if (old_ops.init_fs != NULL)
+			pr_warning("zcache: cleancache_ops overridden");
+	}
+#endif
+#ifdef CONFIG_FRONTSWAP
+	if (zcache_enabled && use_frontswap) {
+		struct frontswap_ops old_ops;
+
+		old_ops = zcache_frontswap_register_ops();
+		pr_info("zcache: frontswap enabled using kernel "
+			"transcendent memory and xvmalloc\n");
+		if (old_ops.init != NULL)
+			pr_warning("zcache: frontswap_ops overridden");
+	}
+#endif
+out:
+	return ret;
+}
+
+module_init(zcache_init)
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH 3/6] staging: ramster: xvmalloc allocation files
  2012-02-15 15:54 [PATCH V5r1 0/6] staging: ramster: multi-machine memory capacity management Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 1/6] staging: ramster: cluster/messaging foundation Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 2/6] staging: ramster: local compression + tmem Dan Magenheimer
@ 2012-02-15 15:54 ` Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 4/6] staging: ramster: ramster-specific changes to zcache/tmem Dan Magenheimer
                   ` (2 subsequent siblings)
  5 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-02-15 15:54 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, dan.magenheimer

RAMster implements peer-to-peer transcendent memory, allowing a "cluster"
of kernels to dynamically pool their RAM.

Zcache is in the process of converting allocators, from xvmalloc to zsmalloc.
Further, RAMster V5 testing to date has been done only with xvmalloc.
To avoid merging problems, a linux-3.2 copy of xvmalloc is incorporated by
this patch.  Later patches will be able to eliminate xvmalloc and use zsmalloc.

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/ramster/xvmalloc.c     |  510 ++++++++++++++++++++++++++++++++
 drivers/staging/ramster/xvmalloc.h     |   30 ++
 drivers/staging/ramster/xvmalloc_int.h |   95 ++++++
 3 files changed, 635 insertions(+), 0 deletions(-)
 create mode 100644 drivers/staging/ramster/xvmalloc.c
 create mode 100644 drivers/staging/ramster/xvmalloc.h
 create mode 100644 drivers/staging/ramster/xvmalloc_int.h

diff --git a/drivers/staging/ramster/xvmalloc.c b/drivers/staging/ramster/xvmalloc.c
new file mode 100644
index 0000000..1f9c508
--- /dev/null
+++ b/drivers/staging/ramster/xvmalloc.c
@@ -0,0 +1,510 @@
+/*
+ * xvmalloc memory allocator
+ *
+ * Copyright (C) 2008, 2009, 2010  Nitin Gupta
+ *
+ * This code is released using a dual license strategy: BSD/GPL
+ * You can choose the licence that better fits your requirements.
+ *
+ * Released under the terms of 3-clause BSD License
+ * Released under the terms of GNU General Public License Version 2.0
+ */
+
+#ifdef CONFIG_ZRAM_DEBUG
+#define DEBUG
+#endif
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/bitops.h>
+#include <linux/errno.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/string.h>
+#include <linux/slab.h>
+
+#include "xvmalloc.h"
+#include "xvmalloc_int.h"
+
+static void stat_inc(u64 *value)
+{
+	*value = *value + 1;
+}
+
+static void stat_dec(u64 *value)
+{
+	*value = *value - 1;
+}
+
+static int test_flag(struct block_header *block, enum blockflags flag)
+{
+	return block->prev & BIT(flag);
+}
+
+static void set_flag(struct block_header *block, enum blockflags flag)
+{
+	block->prev |= BIT(flag);
+}
+
+static void clear_flag(struct block_header *block, enum blockflags flag)
+{
+	block->prev &= ~BIT(flag);
+}
+
+/*
+ * Given <page, offset> pair, provide a dereferencable pointer.
+ * This is called from xv_malloc/xv_free path, so it
+ * needs to be fast.
+ */
+static void *get_ptr_atomic(struct page *page, u16 offset, enum km_type type)
+{
+	unsigned char *base;
+
+	base = kmap_atomic(page, type);
+	return base + offset;
+}
+
+static void put_ptr_atomic(void *ptr, enum km_type type)
+{
+	kunmap_atomic(ptr, type);
+}
+
+static u32 get_blockprev(struct block_header *block)
+{
+	return block->prev & PREV_MASK;
+}
+
+static void set_blockprev(struct block_header *block, u16 new_offset)
+{
+	block->prev = new_offset | (block->prev & FLAGS_MASK);
+}
+
+static struct block_header *BLOCK_NEXT(struct block_header *block)
+{
+	return (struct block_header *)
+		((char *)block + block->size + XV_ALIGN);
+}
+
+/*
+ * Get index of free list containing blocks of maximum size
+ * which is less than or equal to given size.
+ */
+static u32 get_index_for_insert(u32 size)
+{
+	if (unlikely(size > XV_MAX_ALLOC_SIZE))
+		size = XV_MAX_ALLOC_SIZE;
+	size &= ~FL_DELTA_MASK;
+	return (size - XV_MIN_ALLOC_SIZE) >> FL_DELTA_SHIFT;
+}
+
+/*
+ * Get index of free list having blocks of size greater than
+ * or equal to requested size.
+ */
+static u32 get_index(u32 size)
+{
+	if (unlikely(size < XV_MIN_ALLOC_SIZE))
+		size = XV_MIN_ALLOC_SIZE;
+	size = ALIGN(size, FL_DELTA);
+	return (size - XV_MIN_ALLOC_SIZE) >> FL_DELTA_SHIFT;
+}
+
+/**
+ * find_block - find block of at least given size
+ * @pool: memory pool to search from
+ * @size: size of block required
+ * @page: page containing required block
+ * @offset: offset within the page where block is located.
+ *
+ * Searches two level bitmap to locate block of at least
+ * the given size. If such a block is found, it provides
+ * <page, offset> to identify this block and returns index
+ * in freelist where we found this block.
+ * Otherwise, returns 0 and <page, offset> params are not touched.
+ */
+static u32 find_block(struct xv_pool *pool, u32 size,
+			struct page **page, u32 *offset)
+{
+	ulong flbitmap, slbitmap;
+	u32 flindex, slindex, slbitstart;
+
+	/* There are no free blocks in this pool */
+	if (!pool->flbitmap)
+		return 0;
+
+	/* Get freelist index correspoding to this size */
+	slindex = get_index(size);
+	slbitmap = pool->slbitmap[slindex / BITS_PER_LONG];
+	slbitstart = slindex % BITS_PER_LONG;
+
+	/*
+	 * If freelist is not empty at this index, we found the
+	 * block - head of this list. This is approximate best-fit match.
+	 */
+	if (test_bit(slbitstart, &slbitmap)) {
+		*page = pool->freelist[slindex].page;
+		*offset = pool->freelist[slindex].offset;
+		return slindex;
+	}
+
+	/*
+	 * No best-fit found. Search a bit further in bitmap for a free block.
+	 * Second level bitmap consists of series of 32-bit chunks. Search
+	 * further in the chunk where we expected a best-fit, starting from
+	 * index location found above.
+	 */
+	slbitstart++;
+	slbitmap >>= slbitstart;
+
+	/* Skip this search if we were already at end of this bitmap chunk */
+	if ((slbitstart != BITS_PER_LONG) && slbitmap) {
+		slindex += __ffs(slbitmap) + 1;
+		*page = pool->freelist[slindex].page;
+		*offset = pool->freelist[slindex].offset;
+		return slindex;
+	}
+
+	/* Now do a full two-level bitmap search to find next nearest fit */
+	flindex = slindex / BITS_PER_LONG;
+
+	flbitmap = (pool->flbitmap) >> (flindex + 1);
+	if (!flbitmap)
+		return 0;
+
+	flindex += __ffs(flbitmap) + 1;
+	slbitmap = pool->slbitmap[flindex];
+	slindex = (flindex * BITS_PER_LONG) + __ffs(slbitmap);
+	*page = pool->freelist[slindex].page;
+	*offset = pool->freelist[slindex].offset;
+
+	return slindex;
+}
+
+/*
+ * Insert block at <page, offset> in freelist of given pool.
+ * freelist used depends on block size.
+ */
+static void insert_block(struct xv_pool *pool, struct page *page, u32 offset,
+			struct block_header *block)
+{
+	u32 flindex, slindex;
+	struct block_header *nextblock;
+
+	slindex = get_index_for_insert(block->size);
+	flindex = slindex / BITS_PER_LONG;
+
+	block->link.prev_page = NULL;
+	block->link.prev_offset = 0;
+	block->link.next_page = pool->freelist[slindex].page;
+	block->link.next_offset = pool->freelist[slindex].offset;
+	pool->freelist[slindex].page = page;
+	pool->freelist[slindex].offset = offset;
+
+	if (block->link.next_page) {
+		nextblock = get_ptr_atomic(block->link.next_page,
+					block->link.next_offset, KM_USER1);
+		nextblock->link.prev_page = page;
+		nextblock->link.prev_offset = offset;
+		put_ptr_atomic(nextblock, KM_USER1);
+		/* If there was a next page then the free bits are set. */
+		return;
+	}
+
+	__set_bit(slindex % BITS_PER_LONG, &pool->slbitmap[flindex]);
+	__set_bit(flindex, &pool->flbitmap);
+}
+
+/*
+ * Remove block from freelist. Index 'slindex' identifies the freelist.
+ */
+static void remove_block(struct xv_pool *pool, struct page *page, u32 offset,
+			struct block_header *block, u32 slindex)
+{
+	u32 flindex = slindex / BITS_PER_LONG;
+	struct block_header *tmpblock;
+
+	if (block->link.prev_page) {
+		tmpblock = get_ptr_atomic(block->link.prev_page,
+				block->link.prev_offset, KM_USER1);
+		tmpblock->link.next_page = block->link.next_page;
+		tmpblock->link.next_offset = block->link.next_offset;
+		put_ptr_atomic(tmpblock, KM_USER1);
+	}
+
+	if (block->link.next_page) {
+		tmpblock = get_ptr_atomic(block->link.next_page,
+				block->link.next_offset, KM_USER1);
+		tmpblock->link.prev_page = block->link.prev_page;
+		tmpblock->link.prev_offset = block->link.prev_offset;
+		put_ptr_atomic(tmpblock, KM_USER1);
+	}
+
+	/* Is this block is at the head of the freelist? */
+	if (pool->freelist[slindex].page == page
+	   && pool->freelist[slindex].offset == offset) {
+
+		pool->freelist[slindex].page = block->link.next_page;
+		pool->freelist[slindex].offset = block->link.next_offset;
+
+		if (pool->freelist[slindex].page) {
+			struct block_header *tmpblock;
+			tmpblock = get_ptr_atomic(pool->freelist[slindex].page,
+					pool->freelist[slindex].offset,
+					KM_USER1);
+			tmpblock->link.prev_page = NULL;
+			tmpblock->link.prev_offset = 0;
+			put_ptr_atomic(tmpblock, KM_USER1);
+		} else {
+			/* This freelist bucket is empty */
+			__clear_bit(slindex % BITS_PER_LONG,
+				    &pool->slbitmap[flindex]);
+			if (!pool->slbitmap[flindex])
+				__clear_bit(flindex, &pool->flbitmap);
+		}
+	}
+
+	block->link.prev_page = NULL;
+	block->link.prev_offset = 0;
+	block->link.next_page = NULL;
+	block->link.next_offset = 0;
+}
+
+/*
+ * Allocate a page and add it to freelist of given pool.
+ */
+static int grow_pool(struct xv_pool *pool, gfp_t flags)
+{
+	struct page *page;
+	struct block_header *block;
+
+	page = alloc_page(flags);
+	if (unlikely(!page))
+		return -ENOMEM;
+
+	stat_inc(&pool->total_pages);
+
+	spin_lock(&pool->lock);
+	block = get_ptr_atomic(page, 0, KM_USER0);
+
+	block->size = PAGE_SIZE - XV_ALIGN;
+	set_flag(block, BLOCK_FREE);
+	clear_flag(block, PREV_FREE);
+	set_blockprev(block, 0);
+
+	insert_block(pool, page, 0, block);
+
+	put_ptr_atomic(block, KM_USER0);
+	spin_unlock(&pool->lock);
+
+	return 0;
+}
+
+/*
+ * Create a memory pool. Allocates freelist, bitmaps and other
+ * per-pool metadata.
+ */
+struct xv_pool *xv_create_pool(void)
+{
+	u32 ovhd_size;
+	struct xv_pool *pool;
+
+	ovhd_size = roundup(sizeof(*pool), PAGE_SIZE);
+	pool = kzalloc(ovhd_size, GFP_KERNEL);
+	if (!pool)
+		return NULL;
+
+	spin_lock_init(&pool->lock);
+
+	return pool;
+}
+EXPORT_SYMBOL_GPL(xv_create_pool);
+
+void xv_destroy_pool(struct xv_pool *pool)
+{
+	kfree(pool);
+}
+EXPORT_SYMBOL_GPL(xv_destroy_pool);
+
+/**
+ * xv_malloc - Allocate block of given size from pool.
+ * @pool: pool to allocate from
+ * @size: size of block to allocate
+ * @page: page no. that holds the object
+ * @offset: location of object within page
+ *
+ * On success, <page, offset> identifies block allocated
+ * and 0 is returned. On failure, <page, offset> is set to
+ * 0 and -ENOMEM is returned.
+ *
+ * Allocation requests with size > XV_MAX_ALLOC_SIZE will fail.
+ */
+int xv_malloc(struct xv_pool *pool, u32 size, struct page **page,
+		u32 *offset, gfp_t flags)
+{
+	int error;
+	u32 index, tmpsize, origsize, tmpoffset;
+	struct block_header *block, *tmpblock;
+
+	*page = NULL;
+	*offset = 0;
+	origsize = size;
+
+	if (unlikely(!size || size > XV_MAX_ALLOC_SIZE))
+		return -ENOMEM;
+
+	size = ALIGN(size, XV_ALIGN);
+
+	spin_lock(&pool->lock);
+
+	index = find_block(pool, size, page, offset);
+
+	if (!*page) {
+		spin_unlock(&pool->lock);
+		if (flags & GFP_NOWAIT)
+			return -ENOMEM;
+		error = grow_pool(pool, flags);
+		if (unlikely(error))
+			return error;
+
+		spin_lock(&pool->lock);
+		index = find_block(pool, size, page, offset);
+	}
+
+	if (!*page) {
+		spin_unlock(&pool->lock);
+		return -ENOMEM;
+	}
+
+	block = get_ptr_atomic(*page, *offset, KM_USER0);
+
+	remove_block(pool, *page, *offset, block, index);
+
+	/* Split the block if required */
+	tmpoffset = *offset + size + XV_ALIGN;
+	tmpsize = block->size - size;
+	tmpblock = (struct block_header *)((char *)block + size + XV_ALIGN);
+	if (tmpsize) {
+		tmpblock->size = tmpsize - XV_ALIGN;
+		set_flag(tmpblock, BLOCK_FREE);
+		clear_flag(tmpblock, PREV_FREE);
+
+		set_blockprev(tmpblock, *offset);
+		if (tmpblock->size >= XV_MIN_ALLOC_SIZE)
+			insert_block(pool, *page, tmpoffset, tmpblock);
+
+		if (tmpoffset + XV_ALIGN + tmpblock->size != PAGE_SIZE) {
+			tmpblock = BLOCK_NEXT(tmpblock);
+			set_blockprev(tmpblock, tmpoffset);
+		}
+	} else {
+		/* This block is exact fit */
+		if (tmpoffset != PAGE_SIZE)
+			clear_flag(tmpblock, PREV_FREE);
+	}
+
+	block->size = origsize;
+	clear_flag(block, BLOCK_FREE);
+
+	put_ptr_atomic(block, KM_USER0);
+	spin_unlock(&pool->lock);
+
+	*offset += XV_ALIGN;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xv_malloc);
+
+/*
+ * Free block identified with <page, offset>
+ */
+void xv_free(struct xv_pool *pool, struct page *page, u32 offset)
+{
+	void *page_start;
+	struct block_header *block, *tmpblock;
+
+	offset -= XV_ALIGN;
+
+	spin_lock(&pool->lock);
+
+	page_start = get_ptr_atomic(page, 0, KM_USER0);
+	block = (struct block_header *)((char *)page_start + offset);
+
+	/* Catch double free bugs */
+	BUG_ON(test_flag(block, BLOCK_FREE));
+
+	block->size = ALIGN(block->size, XV_ALIGN);
+
+	tmpblock = BLOCK_NEXT(block);
+	if (offset + block->size + XV_ALIGN == PAGE_SIZE)
+		tmpblock = NULL;
+
+	/* Merge next block if its free */
+	if (tmpblock && test_flag(tmpblock, BLOCK_FREE)) {
+		/*
+		 * Blocks smaller than XV_MIN_ALLOC_SIZE
+		 * are not inserted in any free list.
+		 */
+		if (tmpblock->size >= XV_MIN_ALLOC_SIZE) {
+			remove_block(pool, page,
+				    offset + block->size + XV_ALIGN, tmpblock,
+				    get_index_for_insert(tmpblock->size));
+		}
+		block->size += tmpblock->size + XV_ALIGN;
+	}
+
+	/* Merge previous block if its free */
+	if (test_flag(block, PREV_FREE)) {
+		tmpblock = (struct block_header *)((char *)(page_start) +
+						get_blockprev(block));
+		offset = offset - tmpblock->size - XV_ALIGN;
+
+		if (tmpblock->size >= XV_MIN_ALLOC_SIZE)
+			remove_block(pool, page, offset, tmpblock,
+				    get_index_for_insert(tmpblock->size));
+
+		tmpblock->size += block->size + XV_ALIGN;
+		block = tmpblock;
+	}
+
+	/* No used objects in this page. Free it. */
+	if (block->size == PAGE_SIZE - XV_ALIGN) {
+		put_ptr_atomic(page_start, KM_USER0);
+		spin_unlock(&pool->lock);
+
+		__free_page(page);
+		stat_dec(&pool->total_pages);
+		return;
+	}
+
+	set_flag(block, BLOCK_FREE);
+	if (block->size >= XV_MIN_ALLOC_SIZE)
+		insert_block(pool, page, offset, block);
+
+	if (offset + block->size + XV_ALIGN != PAGE_SIZE) {
+		tmpblock = BLOCK_NEXT(block);
+		set_flag(tmpblock, PREV_FREE);
+		set_blockprev(tmpblock, offset);
+	}
+
+	put_ptr_atomic(page_start, KM_USER0);
+	spin_unlock(&pool->lock);
+}
+EXPORT_SYMBOL_GPL(xv_free);
+
+u32 xv_get_object_size(void *obj)
+{
+	struct block_header *blk;
+
+	blk = (struct block_header *)((char *)(obj) - XV_ALIGN);
+	return blk->size;
+}
+EXPORT_SYMBOL_GPL(xv_get_object_size);
+
+/*
+ * Returns total memory used by allocator (userdata + metadata)
+ */
+u64 xv_get_total_size_bytes(struct xv_pool *pool)
+{
+	return pool->total_pages << PAGE_SHIFT;
+}
+EXPORT_SYMBOL_GPL(xv_get_total_size_bytes);
diff --git a/drivers/staging/ramster/xvmalloc.h b/drivers/staging/ramster/xvmalloc.h
new file mode 100644
index 0000000..5b1a81a
--- /dev/null
+++ b/drivers/staging/ramster/xvmalloc.h
@@ -0,0 +1,30 @@
+/*
+ * xvmalloc memory allocator
+ *
+ * Copyright (C) 2008, 2009, 2010  Nitin Gupta
+ *
+ * This code is released using a dual license strategy: BSD/GPL
+ * You can choose the licence that better fits your requirements.
+ *
+ * Released under the terms of 3-clause BSD License
+ * Released under the terms of GNU General Public License Version 2.0
+ */
+
+#ifndef _XV_MALLOC_H_
+#define _XV_MALLOC_H_
+
+#include <linux/types.h>
+
+struct xv_pool;
+
+struct xv_pool *xv_create_pool(void);
+void xv_destroy_pool(struct xv_pool *pool);
+
+int xv_malloc(struct xv_pool *pool, u32 size, struct page **page,
+			u32 *offset, gfp_t flags);
+void xv_free(struct xv_pool *pool, struct page *page, u32 offset);
+
+u32 xv_get_object_size(void *obj);
+u64 xv_get_total_size_bytes(struct xv_pool *pool);
+
+#endif
diff --git a/drivers/staging/ramster/xvmalloc_int.h b/drivers/staging/ramster/xvmalloc_int.h
new file mode 100644
index 0000000..b5f1f7f
--- /dev/null
+++ b/drivers/staging/ramster/xvmalloc_int.h
@@ -0,0 +1,95 @@
+/*
+ * xvmalloc memory allocator
+ *
+ * Copyright (C) 2008, 2009, 2010  Nitin Gupta
+ *
+ * This code is released using a dual license strategy: BSD/GPL
+ * You can choose the licence that better fits your requirements.
+ *
+ * Released under the terms of 3-clause BSD License
+ * Released under the terms of GNU General Public License Version 2.0
+ */
+
+#ifndef _XV_MALLOC_INT_H_
+#define _XV_MALLOC_INT_H_
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+
+/* User configurable params */
+
+/* Must be power of two */
+#ifdef CONFIG_64BIT
+#define XV_ALIGN_SHIFT 3
+#else
+#define XV_ALIGN_SHIFT	2
+#endif
+#define XV_ALIGN	(1 << XV_ALIGN_SHIFT)
+#define XV_ALIGN_MASK	(XV_ALIGN - 1)
+
+/* This must be greater than sizeof(link_free) */
+#define XV_MIN_ALLOC_SIZE	32
+#define XV_MAX_ALLOC_SIZE	(PAGE_SIZE - XV_ALIGN)
+
+/*
+ * Free lists are separated by FL_DELTA bytes
+ * This value is 3 for 4k pages and 4 for 64k pages, for any
+ * other page size, a conservative (PAGE_SHIFT - 9) is used.
+ */
+#if PAGE_SHIFT == 16
+#define FL_DELTA_SHIFT 4
+#else
+#define FL_DELTA_SHIFT (PAGE_SHIFT - 9)
+#endif
+#define FL_DELTA	(1 << FL_DELTA_SHIFT)
+#define FL_DELTA_MASK	(FL_DELTA - 1)
+#define NUM_FREE_LISTS	((XV_MAX_ALLOC_SIZE - XV_MIN_ALLOC_SIZE) \
+				/ FL_DELTA + 1)
+
+#define MAX_FLI		DIV_ROUND_UP(NUM_FREE_LISTS, BITS_PER_LONG)
+
+/* End of user params */
+
+enum blockflags {
+	BLOCK_FREE,
+	PREV_FREE,
+	__NR_BLOCKFLAGS,
+};
+
+#define FLAGS_MASK	XV_ALIGN_MASK
+#define PREV_MASK	(~FLAGS_MASK)
+
+struct freelist_entry {
+	struct page *page;
+	u16 offset;
+	u16 pad;
+};
+
+struct link_free {
+	struct page *prev_page;
+	struct page *next_page;
+	u16 prev_offset;
+	u16 next_offset;
+};
+
+struct block_header {
+	union {
+		/* This common header must be XV_ALIGN bytes */
+		u8 common[XV_ALIGN];
+		struct {
+			u16 size;
+			u16 prev;
+		};
+	};
+	struct link_free link;
+};
+
+struct xv_pool {
+	ulong flbitmap;
+	ulong slbitmap[MAX_FLI];
+	u64 total_pages;	/* stats */
+	struct freelist_entry freelist[NUM_FREE_LISTS];
+	spinlock_t lock;
+};
+
+#endif
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH 4/6] staging: ramster: ramster-specific changes to zcache/tmem
  2012-02-15 15:54 [PATCH V5r1 0/6] staging: ramster: multi-machine memory capacity management Dan Magenheimer
                   ` (2 preceding siblings ...)
  2012-02-15 15:54 ` [PATCH 3/6] staging: ramster: xvmalloc allocation files Dan Magenheimer
@ 2012-02-15 15:54 ` Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 5/6] staging: ramster: ramster-specific new files Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 6/6] staging: ramster: enable as staging driver Dan Magenheimer
  5 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-02-15 15:54 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, dan.magenheimer

RAMster implements peer-to-peer transcendent memory, allowing a "cluster"
of kernels to dynamically pool their RAM.

This patch incorporates changes transforming zcache to work with
a remote store.

In tmem.[ch], new "repatriate" (provoke async get) and "localify" (handle
incoming data resulting from an async get) routines combine with a handful
of changes to existing pamops interfaces allow the generic tmem code
to support asynchronous operations.  Also, a new tmem_xhandle struct
groups together key information that must be passed to remote tmem stores.

Zcache-main.c is augmented with a large amount of ramster-specific code
to handle remote operations and "foreign" pages on both ends of the
"remotify" protocol.  New "foreign" pools are auto-created on demand.
A "selfshrinker" thread periodically repatriates remote persistent pages
when local memory conditions allow.  For certain operations, a queue is
necessary to guarantee strict ordering as out-of-order puts/flushes can
cause strange race conditions.  Pampd pointers now either point to local
memory OR describe a remote page; to allow the same 64-bits to describe
either, the LSB is used to differentiate.  Some acrobatics must be performed
to ensure local memory is available to handle a remote persistent get,
or deal with the data directly anyway if the malloc failed.  Lots
of ramster-specific statistics are available via sysfs.

Note: Some debug ifdefs left in for now.
Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/ramster/Kconfig       |   18 +-
 drivers/staging/ramster/Makefile      |    4 +-
 drivers/staging/ramster/tmem.c        |  115 ++-
 drivers/staging/ramster/tmem.h        |   46 +-
 drivers/staging/ramster/zcache-main.c | 1844 ++++++++++++++++++++++++++++-----
 5 files changed, 1738 insertions(+), 289 deletions(-)

diff --git a/drivers/staging/ramster/Kconfig b/drivers/staging/ramster/Kconfig
index 7fabcb2..acfb974 100644
--- a/drivers/staging/ramster/Kconfig
+++ b/drivers/staging/ramster/Kconfig
@@ -1,13 +1,13 @@
-config ZCACHE
-	tristate "Dynamic compression of swap pages and clean pagecache pages"
-	depends on CLEANCACHE || FRONTSWAP
-	select XVMALLOC
+config RAMSTER
+	bool "Cross-machine RAM capacity sharing, aka peer-to-peer tmem"
+	depends on (CLEANCACHE || FRONTSWAP) && CONFIGFS_FS && !ZCACHE && !XVMALLOC && !HIGHMEM
 	select LZO_COMPRESS
 	select LZO_DECOMPRESS
 	default n
 	help
-	  Zcache doubles RAM efficiency while providing a significant
-	  performance boosts on many workloads.  Zcache uses lzo1x
-	  compression and an in-kernel implementation of transcendent
-	  memory to store clean page cache pages and swap in RAM,
-	  providing a noticeable reduction in disk I/O.
+	  RAMster allows RAM on other machines in a cluster to be utilized
+	  dynamically and symmetrically instead of swapping to a local swap
+	  disk, thus improving performance on memory-constrained workloads
+	  while minimizing total RAM across the cluster.  RAMster, like
+	  zcache, compresses swap pages into local RAM, but then remotifies
+	  the compressed pages to another node in the RAMster cluster.
diff --git a/drivers/staging/ramster/Makefile b/drivers/staging/ramster/Makefile
index 60daa27..bcc13c8 100644
--- a/drivers/staging/ramster/Makefile
+++ b/drivers/staging/ramster/Makefile
@@ -1,3 +1 @@
-zcache-y	:=	zcache-main.o tmem.o
-
-obj-$(CONFIG_ZCACHE)	+=	zcache.o
+obj-$(CONFIG_RAMSTER)	+=	zcache-main.o tmem.o r2net.o xvmalloc.o cluster/
diff --git a/drivers/staging/ramster/tmem.c b/drivers/staging/ramster/tmem.c
index 1ca66ea..8f2f689 100644
--- a/drivers/staging/ramster/tmem.c
+++ b/drivers/staging/ramster/tmem.c
@@ -27,6 +27,7 @@
 #include <linux/list.h>
 #include <linux/spinlock.h>
 #include <linux/atomic.h>
+#include <linux/delay.h>
 
 #include "tmem.h"
 
@@ -316,7 +317,7 @@ static void *tmem_pampd_lookup_in_obj(struct tmem_obj *obj, uint32_t index)
 }
 
 static void *tmem_pampd_replace_in_obj(struct tmem_obj *obj, uint32_t index,
-					void *new_pampd)
+					void *new_pampd, bool no_free)
 {
 	struct tmem_objnode **slot;
 	void *ret = NULL;
@@ -325,7 +326,9 @@ static void *tmem_pampd_replace_in_obj(struct tmem_obj *obj, uint32_t index,
 	if ((slot != NULL) && (*slot != NULL)) {
 		void *old_pampd = *(void **)slot;
 		*(void **)slot = new_pampd;
-		(*tmem_pamops.free)(old_pampd, obj->pool, NULL, 0);
+		if (!no_free)
+			(*tmem_pamops.free)(old_pampd, obj->pool,
+						NULL, 0, false);
 		ret = new_pampd;
 	}
 	return ret;
@@ -481,7 +484,7 @@ static void tmem_objnode_node_destroy(struct tmem_obj *obj,
 			if (ht == 1) {
 				obj->pampd_count--;
 				(*tmem_pamops.free)(objnode->slots[i],
-						obj->pool, NULL, 0);
+						obj->pool, NULL, 0, true);
 				objnode->slots[i] = NULL;
 				continue;
 			}
@@ -498,7 +501,8 @@ static void tmem_pampd_destroy_all_in_obj(struct tmem_obj *obj)
 		return;
 	if (obj->objnode_tree_height == 0) {
 		obj->pampd_count--;
-		(*tmem_pamops.free)(obj->objnode_tree_root, obj->pool, NULL, 0);
+		(*tmem_pamops.free)(obj->objnode_tree_root,
+					obj->pool, NULL, 0, true);
 	} else {
 		tmem_objnode_node_destroy(obj, obj->objnode_tree_root,
 					obj->objnode_tree_height);
@@ -529,7 +533,7 @@ static void tmem_pampd_destroy_all_in_obj(struct tmem_obj *obj)
  * always flushes for simplicity.
  */
 int tmem_put(struct tmem_pool *pool, struct tmem_oid *oidp, uint32_t index,
-		char *data, size_t size, bool raw, bool ephemeral)
+		char *data, size_t size, bool raw, int ephemeral)
 {
 	struct tmem_obj *obj = NULL, *objfound = NULL, *objnew = NULL;
 	void *pampd = NULL, *pampd_del = NULL;
@@ -545,7 +549,7 @@ int tmem_put(struct tmem_pool *pool, struct tmem_oid *oidp, uint32_t index,
 			/* if found, is a dup put, flush the old one */
 			pampd_del = tmem_pampd_delete_from_obj(obj, index);
 			BUG_ON(pampd_del != pampd);
-			(*tmem_pamops.free)(pampd, pool, oidp, index);
+			(*tmem_pamops.free)(pampd, pool, oidp, index, true);
 			if (obj->pampd_count == 0) {
 				objnew = obj;
 				objfound = NULL;
@@ -576,7 +580,7 @@ delete_and_free:
 	(void)tmem_pampd_delete_from_obj(obj, index);
 free:
 	if (pampd)
-		(*tmem_pamops.free)(pampd, pool, NULL, 0);
+		(*tmem_pamops.free)(pampd, pool, NULL, 0, true);
 	if (objnew) {
 		tmem_obj_free(objnew, hb);
 		(*tmem_hostops.obj_free)(objnew, pool);
@@ -586,6 +590,65 @@ out:
 	return ret;
 }
 
+void *tmem_localify_get_pampd(struct tmem_pool *pool, struct tmem_oid *oidp,
+				uint32_t index, struct tmem_obj **ret_obj,
+				void **saved_hb)
+{
+	struct tmem_hashbucket *hb;
+	struct tmem_obj *obj = NULL;
+	void *pampd = NULL;
+
+	hb = &pool->hashbucket[tmem_oid_hash(oidp)];
+	spin_lock(&hb->lock);
+	obj = tmem_obj_find(hb, oidp);
+	if (likely(obj != NULL))
+		pampd = tmem_pampd_lookup_in_obj(obj, index);
+	*ret_obj = obj;
+	*saved_hb = (void *)hb;
+	/* note, hashbucket remains locked */
+	return pampd;
+}
+
+void tmem_localify_finish(struct tmem_obj *obj, uint32_t index,
+			  void *pampd, void *saved_hb, bool delete)
+{
+	struct tmem_hashbucket *hb = (struct tmem_hashbucket *)saved_hb;
+
+	BUG_ON(!spin_is_locked(&hb->lock));
+	if (pampd != NULL) {
+		BUG_ON(obj == NULL);
+		(void)tmem_pampd_replace_in_obj(obj, index, pampd, 1);
+	} else if (delete) {
+		BUG_ON(obj == NULL);
+		(void)tmem_pampd_delete_from_obj(obj, index);
+	}
+	spin_unlock(&hb->lock);
+}
+
+static int tmem_repatriate(void **ppampd, struct tmem_hashbucket *hb,
+				struct tmem_pool *pool, struct tmem_oid *oidp,
+				uint32_t index, bool free, char *data)
+{
+	void *old_pampd = *ppampd, *new_pampd = NULL;
+	bool intransit = false;
+	int ret = 0;
+
+
+	if (!is_ephemeral(pool))
+		new_pampd = (*tmem_pamops.repatriate_preload)(
+				old_pampd, pool, oidp, index, &intransit);
+	if (intransit)
+		ret = -EAGAIN;
+	else if (new_pampd != NULL)
+		*ppampd = new_pampd;
+	/* must release the hb->lock else repatriate can't sleep */
+	spin_unlock(&hb->lock);
+	if (!intransit)
+		ret = (*tmem_pamops.repatriate)(old_pampd, new_pampd, pool,
+						oidp, index, free, data);
+	return ret;
+}
+
 /*
  * "Get" a page, e.g. if one can be found, copy the tmem page with the
  * matching handle from PAM space to the kernel.  By tmem definition,
@@ -607,14 +670,36 @@ int tmem_get(struct tmem_pool *pool, struct tmem_oid *oidp, uint32_t index,
 	int ret = -1;
 	struct tmem_hashbucket *hb;
 	bool free = (get_and_free == 1) || ((get_and_free == 0) && ephemeral);
-	bool lock_held = false;
+	bool lock_held = 0;
+	void **ppampd;
 
+again:
 	hb = &pool->hashbucket[tmem_oid_hash(oidp)];
 	spin_lock(&hb->lock);
-	lock_held = true;
+	lock_held = 1;
 	obj = tmem_obj_find(hb, oidp);
 	if (obj == NULL)
 		goto out;
+	ppampd = __tmem_pampd_lookup_in_obj(obj, index);
+	if (ppampd == NULL)
+		goto out;
+	if (tmem_pamops.is_remote(*ppampd)) {
+		ret = tmem_repatriate(ppampd, hb, pool, oidp,
+					index, free, data);
+		lock_held = 0; /* note hb->lock has been unlocked */
+		if (ret == -EAGAIN) {
+			/* rare I think, but should cond_resched()??? */
+			usleep_range(10, 1000);
+			goto again;
+		} else if (ret != 0) {
+			if (ret != -ENOENT)
+				pr_err("UNTESTED case in tmem_get, ret=%d\n",
+						ret);
+			ret = -1;
+			goto out;
+		}
+		goto out;
+	}
 	if (free)
 		pampd = tmem_pampd_delete_from_obj(obj, index);
 	else
@@ -628,10 +713,6 @@ int tmem_get(struct tmem_pool *pool, struct tmem_oid *oidp, uint32_t index,
 			obj = NULL;
 		}
 	}
-	if (tmem_pamops.is_remote(pampd)) {
-		lock_held = false;
-		spin_unlock(&hb->lock);
-	}
 	if (free)
 		ret = (*tmem_pamops.get_data_and_free)(
 				data, size, raw, pampd, pool, oidp, index);
@@ -668,7 +749,7 @@ int tmem_flush_page(struct tmem_pool *pool,
 	pampd = tmem_pampd_delete_from_obj(obj, index);
 	if (pampd == NULL)
 		goto out;
-	(*tmem_pamops.free)(pampd, pool, oidp, index);
+	(*tmem_pamops.free)(pampd, pool, oidp, index, true);
 	if (obj->pampd_count == 0) {
 		tmem_obj_free(obj, hb);
 		(*tmem_hostops.obj_free)(obj, pool);
@@ -682,8 +763,8 @@ out:
 
 /*
  * If a page in tmem matches the handle, replace the page so that any
- * subsequent "get" gets the new page.  Returns 0 if
- * there was a page to replace, else returns -1.
+ * subsequent "get" gets the new page.  Returns the new page if
+ * there was a page to replace, else returns NULL.
  */
 int tmem_replace(struct tmem_pool *pool, struct tmem_oid *oidp,
 			uint32_t index, void *new_pampd)
@@ -697,7 +778,7 @@ int tmem_replace(struct tmem_pool *pool, struct tmem_oid *oidp,
 	obj = tmem_obj_find(hb, oidp);
 	if (obj == NULL)
 		goto out;
-	new_pampd = tmem_pampd_replace_in_obj(obj, index, new_pampd);
+	new_pampd = tmem_pampd_replace_in_obj(obj, index, new_pampd, 0);
 	ret = (*tmem_pamops.replace_in_obj)(new_pampd, obj);
 out:
 	spin_unlock(&hb->lock);
diff --git a/drivers/staging/ramster/tmem.h b/drivers/staging/ramster/tmem.h
index ed147c4..47f1918 100644
--- a/drivers/staging/ramster/tmem.h
+++ b/drivers/staging/ramster/tmem.h
@@ -9,7 +9,6 @@
 #ifndef _TMEM_H_
 #define _TMEM_H_
 
-#include <linux/types.h>
 #include <linux/highmem.h>
 #include <linux/hash.h>
 #include <linux/atomic.h>
@@ -89,6 +88,31 @@ struct tmem_oid {
 	uint64_t oid[3];
 };
 
+struct tmem_xhandle {
+	uint8_t client_id;
+	uint8_t xh_data_cksum;
+	uint16_t xh_data_size;
+	uint16_t pool_id;
+	struct tmem_oid oid;
+	uint32_t index;
+	void *extra;
+};
+
+static inline struct tmem_xhandle tmem_xhandle_fill(uint16_t client_id,
+					struct tmem_pool *pool,
+					struct tmem_oid *oidp,
+					uint32_t index)
+{
+	struct tmem_xhandle xh;
+	xh.client_id = client_id;
+	xh.xh_data_cksum = (uint8_t)-1;
+	xh.xh_data_size = (uint16_t)-1;
+	xh.pool_id = pool->pool_id;
+	xh.oid = *oidp;
+	xh.index = index;
+	return xh;
+}
+
 static inline void tmem_oid_set_invalid(struct tmem_oid *oidp)
 {
 	oidp->oid[0] = oidp->oid[1] = oidp->oid[2] = -1UL;
@@ -147,7 +171,11 @@ struct tmem_obj {
 	unsigned int objnode_tree_height;
 	unsigned long objnode_count;
 	long pampd_count;
-	void *extra; /* for private use by pampd implementation */
+	/* for current design of ramster, all pages belonging to
+	 * an object reside on the same remotenode and extra is
+	 * used to record the number of the remotenode so a
+	 * flush-object operation can specify it */
+	void *extra; /* for use by pampd implementation */
 	DECL_SENTINEL
 };
 
@@ -174,9 +202,14 @@ struct tmem_pamops {
 	int (*get_data_and_free)(char *, size_t *, bool, void *,
 				struct tmem_pool *, struct tmem_oid *,
 				uint32_t);
-	void (*free)(void *, struct tmem_pool *, struct tmem_oid *, uint32_t);
+	void (*free)(void *, struct tmem_pool *,
+				struct tmem_oid *, uint32_t, bool);
 	void (*free_obj)(struct tmem_pool *, struct tmem_obj *);
 	bool (*is_remote)(void *);
+	void *(*repatriate_preload)(void *, struct tmem_pool *,
+					struct tmem_oid *, uint32_t, bool *);
+	int (*repatriate)(void *, void *, struct tmem_pool *,
+				struct tmem_oid *, uint32_t, bool, void *);
 	void (*new_obj)(struct tmem_obj *);
 	int (*replace_in_obj)(void *, struct tmem_obj *);
 };
@@ -193,11 +226,16 @@ extern void tmem_register_hostops(struct tmem_hostops *m);
 
 /* core tmem accessor functions */
 extern int tmem_put(struct tmem_pool *, struct tmem_oid *, uint32_t index,
-			char *, size_t, bool, bool);
+			char *, size_t, bool, int);
 extern int tmem_get(struct tmem_pool *, struct tmem_oid *, uint32_t index,
 			char *, size_t *, bool, int);
 extern int tmem_replace(struct tmem_pool *, struct tmem_oid *, uint32_t index,
 			void *);
+extern void *tmem_localify_get_pampd(struct tmem_pool *, struct tmem_oid *,
+				   uint32_t index, struct tmem_obj **,
+				   void **);
+extern void tmem_localify_finish(struct tmem_obj *, uint32_t index,
+				 void *, void *, bool);
 extern int tmem_flush_page(struct tmem_pool *, struct tmem_oid *,
 			uint32_t index);
 extern int tmem_flush_object(struct tmem_pool *, struct tmem_oid *);
diff --git a/drivers/staging/ramster/zcache-main.c b/drivers/staging/ramster/zcache-main.c
index 49c8791..36d53ed 100644
--- a/drivers/staging/ramster/zcache-main.c
+++ b/drivers/staging/ramster/zcache-main.c
@@ -1,7 +1,7 @@
 /*
  * zcache.c
  *
- * Copyright (c) 2010,2011, Dan Magenheimer, Oracle Corp.
+ * Copyright (c) 2010-2012, Dan Magenheimer, Oracle Corp.
  * Copyright (c) 2010,2011, Nitin Gupta
  *
  * Zcache provides an in-kernel "host implementation" for transcendent memory
@@ -17,6 +17,9 @@
  *
  * [1] For a definition of page-accessible memory (aka PAM), see:
  *   http://marc.info/?l=linux-mm&m=127811271605009
+ *  RAMSTER TODO:
+ *   - handle remotifying of buddied pages (see zbud_remotify_zbpg)
+ *   - kernel boot params: nocleancache/nofrontswap don't always work?!?
  */
 
 #include <linux/module.h>
@@ -30,11 +33,16 @@
 #include <linux/atomic.h>
 #include <linux/math64.h>
 #include "tmem.h"
+#include "zcache.h"
+#include "ramster.h"
+#include "cluster/tcp.h"
 
-#include "../zram/xvmalloc.h" /* if built in drivers/staging */
+#include "xvmalloc.h"	/* temporary until change to zsmalloc */
+
+#define	RAMSTER_TESTING
 
 #if (!defined(CONFIG_CLEANCACHE) && !defined(CONFIG_FRONTSWAP))
-#error "zcache is useless without CONFIG_CLEANCACHE or CONFIG_FRONTSWAP"
+#error "ramster is useless without CONFIG_CLEANCACHE or CONFIG_FRONTSWAP"
 #endif
 #ifdef CONFIG_CLEANCACHE
 #include <linux/cleancache.h>
@@ -43,6 +51,61 @@
 #include <linux/frontswap.h>
 #endif
 
+enum ramster_remotify_op {
+	RAMSTER_REMOTIFY_EPH_PUT,
+	RAMSTER_REMOTIFY_PERS_PUT,
+	RAMSTER_REMOTIFY_FLUSH_PAGE,
+	RAMSTER_REMOTIFY_FLUSH_OBJ,
+	RAMSTER_INTRANSIT_PERS
+};
+
+struct ramster_remotify_hdr {
+	enum ramster_remotify_op op;
+	struct list_head list;
+};
+
+#define ZBH_SENTINEL  0x43214321
+#define ZBPG_SENTINEL  0xdeadbeef
+
+#define ZBUD_MAX_BUDS 2
+
+struct zbud_hdr {
+	struct ramster_remotify_hdr rem_op;
+	uint16_t client_id;
+	uint16_t pool_id;
+	struct tmem_oid oid;
+	uint32_t index;
+	uint16_t size; /* compressed size in bytes, zero means unused */
+	DECL_SENTINEL
+};
+
+#define ZVH_SENTINEL  0x43214321
+static const int zv_max_page_size = (PAGE_SIZE / 8) * 7;
+
+struct zv_hdr {
+	struct ramster_remotify_hdr rem_op;
+	uint16_t client_id;
+	uint16_t pool_id;
+	struct tmem_oid oid;
+	uint32_t index;
+	DECL_SENTINEL
+};
+
+struct flushlist_node {
+	struct ramster_remotify_hdr rem_op;
+	struct tmem_xhandle xh;
+};
+
+union {
+	struct ramster_remotify_hdr rem_op;
+	struct zv_hdr zv;
+	struct zbud_hdr zbud;
+	struct flushlist_node flist;
+} remotify_list_node;
+
+static LIST_HEAD(zcache_rem_op_list);
+static DEFINE_SPINLOCK(zcache_rem_op_list_lock);
+
 #if 0
 /* this is more aggressive but may cause other problems? */
 #define ZCACHE_GFP_MASK	(GFP_ATOMIC | __GFP_NORETRY | __GFP_NOWARN)
@@ -98,20 +161,6 @@ static inline bool is_local_client(struct zcache_client *cli)
  * read or written unless the zbpg's lock is held.
  */
 
-#define ZBH_SENTINEL  0x43214321
-#define ZBPG_SENTINEL  0xdeadbeef
-
-#define ZBUD_MAX_BUDS 2
-
-struct zbud_hdr {
-	uint16_t client_id;
-	uint16_t pool_id;
-	struct tmem_oid oid;
-	uint32_t index;
-	uint16_t size; /* compressed size in bytes, zero means unused */
-	DECL_SENTINEL
-};
-
 struct zbud_page {
 	struct list_head bud_list;
 	spinlock_t lock;
@@ -141,20 +190,43 @@ static unsigned long zcache_zbud_buddied_count;
 /* protects the buddied list and all unbuddied lists */
 static DEFINE_SPINLOCK(zbud_budlists_spinlock);
 
-static LIST_HEAD(zbpg_unused_list);
-static unsigned long zcache_zbpg_unused_list_count;
-
-/* protects the unused page list */
-static DEFINE_SPINLOCK(zbpg_unused_list_spinlock);
-
 static atomic_t zcache_zbud_curr_raw_pages;
 static atomic_t zcache_zbud_curr_zpages;
 static unsigned long zcache_zbud_curr_zbytes;
 static unsigned long zcache_zbud_cumul_zpages;
 static unsigned long zcache_zbud_cumul_zbytes;
 static unsigned long zcache_compress_poor;
+static unsigned long zcache_policy_percent_exceeded;
 static unsigned long zcache_mean_compress_poor;
 
+/*
+ * RAMster counters
+ * - Remote pages are pages with a local pampd but the data is remote
+ * - Foreign pages are pages stored locally but belonging to another node
+ */
+static atomic_t ramster_remote_pers_pages = ATOMIC_INIT(0);
+static unsigned long ramster_pers_remotify_enable;
+static unsigned long ramster_eph_remotify_enable;
+static unsigned long ramster_eph_pages_remoted;
+static unsigned long ramster_eph_pages_remote_failed;
+static unsigned long ramster_pers_pages_remoted;
+static unsigned long ramster_pers_pages_remote_failed;
+static unsigned long ramster_pers_pages_remote_nomem;
+static unsigned long ramster_remote_objects_flushed;
+static unsigned long ramster_remote_object_flushes_failed;
+static unsigned long ramster_remote_pages_flushed;
+static unsigned long ramster_remote_page_flushes_failed;
+static unsigned long ramster_remote_eph_pages_succ_get;
+static unsigned long ramster_remote_pers_pages_succ_get;
+static unsigned long ramster_remote_eph_pages_unsucc_get;
+static unsigned long ramster_remote_pers_pages_unsucc_get;
+static atomic_t ramster_curr_flnode_count = ATOMIC_INIT(0);
+static unsigned long ramster_curr_flnode_count_max;
+static atomic_t ramster_foreign_eph_pampd_count = ATOMIC_INIT(0);
+static unsigned long ramster_foreign_eph_pampd_count_max;
+static atomic_t ramster_foreign_pers_pampd_count = ATOMIC_INIT(0);
+static unsigned long ramster_foreign_pers_pampd_count_max;
+
 /* forward references */
 static void *zcache_get_free_page(void);
 static void zcache_free_page(void *p);
@@ -210,6 +282,29 @@ static char *zbud_data(struct zbud_hdr *zh, unsigned size)
 	return p;
 }
 
+static void zbud_copy_from_pampd(char *data, size_t *size, struct zbud_hdr *zh)
+{
+	struct zbud_page *zbpg;
+	char *p;
+	unsigned budnum;
+
+	ASSERT_SENTINEL(zh, ZBH);
+	budnum = zbud_budnum(zh);
+	zbpg = container_of(zh, struct zbud_page, buddy[budnum]);
+	spin_lock(&zbpg->lock);
+	BUG_ON(zh->size > *size);
+	p = (char *)zbpg;
+	if (budnum == 0)
+		p += ((sizeof(struct zbud_page) + CHUNK_SIZE - 1) &
+							CHUNK_MASK);
+	else if (budnum == 1)
+		p += PAGE_SIZE - ((zh->size + CHUNK_SIZE - 1) & CHUNK_MASK);
+	/* client should be filled in by caller */
+	memcpy(data, p, zh->size);
+	*size = zh->size;
+	spin_unlock(&zbpg->lock);
+}
+
 /*
  * zbud raw page management
  */
@@ -218,38 +313,17 @@ static struct zbud_page *zbud_alloc_raw_page(void)
 {
 	struct zbud_page *zbpg = NULL;
 	struct zbud_hdr *zh0, *zh1;
-	bool recycled = 0;
-
-	/* if any pages on the zbpg list, use one */
-	spin_lock(&zbpg_unused_list_spinlock);
-	if (!list_empty(&zbpg_unused_list)) {
-		zbpg = list_first_entry(&zbpg_unused_list,
-				struct zbud_page, bud_list);
-		list_del_init(&zbpg->bud_list);
-		zcache_zbpg_unused_list_count--;
-		recycled = 1;
-	}
-	spin_unlock(&zbpg_unused_list_spinlock);
-	if (zbpg == NULL)
-		/* none on zbpg list, try to get a kernel page */
 		zbpg = zcache_get_free_page();
 	if (likely(zbpg != NULL)) {
 		INIT_LIST_HEAD(&zbpg->bud_list);
 		zh0 = &zbpg->buddy[0]; zh1 = &zbpg->buddy[1];
 		spin_lock_init(&zbpg->lock);
-		if (recycled) {
-			ASSERT_INVERTED_SENTINEL(zbpg, ZBPG);
-			SET_SENTINEL(zbpg, ZBPG);
-			BUG_ON(zh0->size != 0 || tmem_oid_valid(&zh0->oid));
-			BUG_ON(zh1->size != 0 || tmem_oid_valid(&zh1->oid));
-		} else {
-			atomic_inc(&zcache_zbud_curr_raw_pages);
-			INIT_LIST_HEAD(&zbpg->bud_list);
-			SET_SENTINEL(zbpg, ZBPG);
-			zh0->size = 0; zh1->size = 0;
-			tmem_oid_set_invalid(&zh0->oid);
-			tmem_oid_set_invalid(&zh1->oid);
-		}
+		atomic_inc(&zcache_zbud_curr_raw_pages);
+		INIT_LIST_HEAD(&zbpg->bud_list);
+		SET_SENTINEL(zbpg, ZBPG);
+		zh0->size = 0; zh1->size = 0;
+		tmem_oid_set_invalid(&zh0->oid);
+		tmem_oid_set_invalid(&zh1->oid);
 	}
 	return zbpg;
 }
@@ -265,10 +339,8 @@ static void zbud_free_raw_page(struct zbud_page *zbpg)
 	BUG_ON(zh1->size != 0 || tmem_oid_valid(&zh1->oid));
 	INVERT_SENTINEL(zbpg, ZBPG);
 	spin_unlock(&zbpg->lock);
-	spin_lock(&zbpg_unused_list_spinlock);
-	list_add(&zbpg->bud_list, &zbpg_unused_list);
-	zcache_zbpg_unused_list_count++;
-	spin_unlock(&zbpg_unused_list_spinlock);
+	atomic_dec(&zcache_zbud_curr_raw_pages);
+	zcache_free_page(zbpg);
 }
 
 /*
@@ -299,6 +371,10 @@ static void zbud_free_and_delist(struct zbud_hdr *zh)
 	struct zbud_page *zbpg =
 		container_of(zh, struct zbud_page, buddy[budnum]);
 
+	/* FIXME, should be BUG_ON, pool destruction path doesn't disable
+	 * interrupts tmem_destroy_pool()->tmem_pampd_destroy_all_in_obj()->
+	 * tmem_objnode_node_destroy()-> zcache_pampd_free() */
+	WARN_ON(!irqs_disabled());
 	spin_lock(&zbpg->lock);
 	if (list_empty(&zbpg->bud_list)) {
 		/* ignore zombie page... see zbud_evict_pages() */
@@ -358,8 +434,8 @@ static struct zbud_hdr *zbud_create(uint16_t client_id, uint16_t pool_id,
 	if (unlikely(zbpg == NULL))
 		goto out;
 	/* ok, have a page, now compress the data before taking locks */
-	spin_lock(&zbpg->lock);
 	spin_lock(&zbud_budlists_spinlock);
+	spin_lock(&zbpg->lock);
 	list_add_tail(&zbpg->bud_list, &zbud_unbuddied[nchunks].list);
 	zbud_unbuddied[nchunks].count++;
 	zh = &zbpg->buddy[0];
@@ -389,12 +465,10 @@ init_zh:
 	zh->oid = *oid;
 	zh->pool_id = pool_id;
 	zh->client_id = client_id;
-	/* can wait to copy the data until the list locks are dropped */
-	spin_unlock(&zbud_budlists_spinlock);
-
 	to = zbud_data(zh, size);
 	memcpy(to, cdata, size);
 	spin_unlock(&zbpg->lock);
+	spin_unlock(&zbud_budlists_spinlock);
 	zbud_cumul_chunk_counts[nchunks]++;
 	atomic_inc(&zcache_zbud_curr_zpages);
 	zcache_zbud_cumul_zpages++;
@@ -458,9 +532,9 @@ static void zbud_evict_zbpg(struct zbud_page *zbpg)
 	uint32_t index[ZBUD_MAX_BUDS];
 	struct tmem_oid oid[ZBUD_MAX_BUDS];
 	struct tmem_pool *pool;
+	unsigned long flags;
 
 	ASSERT_SPINLOCK(&zbpg->lock);
-	BUG_ON(!list_empty(&zbpg->bud_list));
 	for (i = 0, j = 0; i < ZBUD_MAX_BUDS; i++) {
 		zh = &zbpg->buddy[i];
 		if (zh->size) {
@@ -469,20 +543,18 @@ static void zbud_evict_zbpg(struct zbud_page *zbpg)
 			oid[j] = zh->oid;
 			index[j] = zh->index;
 			j++;
-			zbud_free(zh);
 		}
 	}
 	spin_unlock(&zbpg->lock);
 	for (i = 0; i < j; i++) {
 		pool = zcache_get_pool_by_id(client_id[i], pool_id[i]);
-		if (pool != NULL) {
-			tmem_flush_page(pool, &oid[i], index[i]);
-			zcache_put_pool(pool);
-		}
+		BUG_ON(pool == NULL);
+		local_irq_save(flags);
+		/* these flushes should dispose of any local storage */
+		tmem_flush_page(pool, &oid[i], index[i]);
+		local_irq_restore(flags);
+		zcache_put_pool(pool);
 	}
-	ASSERT_SENTINEL(zbpg, ZBPG);
-	spin_lock(&zbpg->lock);
-	zbud_free_raw_page(zbpg);
 }
 
 /*
@@ -496,26 +568,8 @@ static void zbud_evict_zbpg(struct zbud_page *zbpg)
 static void zbud_evict_pages(int nr)
 {
 	struct zbud_page *zbpg;
-	int i;
+	int i, newly_unused_pages = 0;
 
-	/* first try freeing any pages on unused list */
-retry_unused_list:
-	spin_lock_bh(&zbpg_unused_list_spinlock);
-	if (!list_empty(&zbpg_unused_list)) {
-		/* can't walk list here, since it may change when unlocked */
-		zbpg = list_first_entry(&zbpg_unused_list,
-				struct zbud_page, bud_list);
-		list_del_init(&zbpg->bud_list);
-		zcache_zbpg_unused_list_count--;
-		atomic_dec(&zcache_zbud_curr_raw_pages);
-		spin_unlock_bh(&zbpg_unused_list_spinlock);
-		zcache_free_page(zbpg);
-		zcache_evicted_raw_pages++;
-		if (--nr <= 0)
-			goto out;
-		goto retry_unused_list;
-	}
-	spin_unlock_bh(&zbpg_unused_list_spinlock);
 
 	/* now try freeing unbuddied pages, starting with least space avail */
 	for (i = 0; i < MAX_CHUNK; i++) {
@@ -528,15 +582,15 @@ retry_unbud_list_i:
 		list_for_each_entry(zbpg, &zbud_unbuddied[i].list, bud_list) {
 			if (unlikely(!spin_trylock(&zbpg->lock)))
 				continue;
-			list_del_init(&zbpg->bud_list);
 			zbud_unbuddied[i].count--;
 			spin_unlock(&zbud_budlists_spinlock);
 			zcache_evicted_unbuddied_pages++;
 			/* want budlists unlocked when doing zbpg eviction */
 			zbud_evict_zbpg(zbpg);
+			newly_unused_pages++;
 			local_bh_enable();
 			if (--nr <= 0)
-				goto out;
+				goto evict_unused;
 			goto retry_unbud_list_i;
 		}
 		spin_unlock_bh(&zbud_budlists_spinlock);
@@ -547,27 +601,413 @@ retry_bud_list:
 	spin_lock_bh(&zbud_budlists_spinlock);
 	if (list_empty(&zbud_buddied_list)) {
 		spin_unlock_bh(&zbud_budlists_spinlock);
-		goto out;
+		goto evict_unused;
 	}
 	list_for_each_entry(zbpg, &zbud_buddied_list, bud_list) {
 		if (unlikely(!spin_trylock(&zbpg->lock)))
 			continue;
-		list_del_init(&zbpg->bud_list);
 		zcache_zbud_buddied_count--;
 		spin_unlock(&zbud_budlists_spinlock);
 		zcache_evicted_buddied_pages++;
 		/* want budlists unlocked when doing zbpg eviction */
 		zbud_evict_zbpg(zbpg);
+		newly_unused_pages++;
 		local_bh_enable();
 		if (--nr <= 0)
-			goto out;
+			goto evict_unused;
 		goto retry_bud_list;
 	}
 	spin_unlock_bh(&zbud_budlists_spinlock);
+
+evict_unused:
+	return;
+}
+
+static DEFINE_PER_CPU(unsigned char *, zcache_remoteputmem);
+
+static int zbud_remotify_zbud(struct tmem_xhandle *xh, char *data,
+				size_t size)
+{
+	struct tmem_pool *pool;
+	int i, remotenode, ret = -1;
+	unsigned char cksum, *p;
+	unsigned long flags;
+
+	for (p = data, cksum = 0, i = 0; i < size; i++)
+		cksum += *p;
+	ret = ramster_remote_put(xh, data, size, true, &remotenode);
+	if (ret == 0) {
+		/* data was successfully remoted so change the local version
+		 * to point to the remote node where it landed */
+		pool = zcache_get_pool_by_id(LOCAL_CLIENT, xh->pool_id);
+		BUG_ON(pool == NULL);
+		local_irq_save(flags);
+		/* tmem_replace will also free up any local space */
+		(void)tmem_replace(pool, &xh->oid, xh->index,
+			pampd_make_remote(remotenode, size, cksum));
+		local_irq_restore(flags);
+		zcache_put_pool(pool);
+		ramster_eph_pages_remoted++;
+		ret = 0;
+	} else
+		ramster_eph_pages_remote_failed++;
+	return ret;
+}
+
+static int zbud_remotify_zbpg(struct zbud_page *zbpg)
+{
+	struct zbud_hdr *zh1, *zh2 = NULL;
+	struct tmem_xhandle xh1, xh2 = { 0 };
+	char *data1 = NULL, *data2 = NULL;
+	size_t size1 = 0, size2 = 0;
+	int ret = 0;
+	unsigned char *tmpmem = __get_cpu_var(zcache_remoteputmem);
+
+	ASSERT_SPINLOCK(&zbpg->lock);
+	if (zbpg->buddy[0].size == 0)
+		zh1 = &zbpg->buddy[1];
+	else if (zbpg->buddy[1].size == 0)
+		zh1 = &zbpg->buddy[0];
+	else {
+		zh1 = &zbpg->buddy[0];
+		zh2 = &zbpg->buddy[1];
+	}
+	/* don't remotify pages that are already remotified */
+	if (zh1->client_id != LOCAL_CLIENT)
+		zh1 = NULL;
+	if ((zh2 != NULL) && (zh2->client_id != LOCAL_CLIENT))
+		zh2 = NULL;
+
+	/* copy the data and metadata so can release lock */
+	if (zh1 != NULL) {
+		xh1.client_id = zh1->client_id;
+		xh1.pool_id = zh1->pool_id;
+		xh1.oid = zh1->oid;
+		xh1.index = zh1->index;
+		size1 = zh1->size;
+		data1 = zbud_data(zh1, size1);
+		memcpy(tmpmem, zbud_data(zh1, size1), size1);
+		data1 = tmpmem;
+		tmpmem += size1;
+	}
+	if (zh2 != NULL) {
+		xh2.client_id = zh2->client_id;
+		xh2.pool_id = zh2->pool_id;
+		xh2.oid = zh2->oid;
+		xh2.index = zh2->index;
+		size2 = zh2->size;
+		memcpy(tmpmem, zbud_data(zh2, size2), size2);
+		data2 = tmpmem;
+	}
+	spin_unlock(&zbpg->lock);
+	preempt_enable();
+
+	/* OK, no locks held anymore, remotify one or both zbuds */
+	if (zh1 != NULL)
+		ret = zbud_remotify_zbud(&xh1, data1, size1);
+	if (zh2 != NULL)
+		ret |= zbud_remotify_zbud(&xh2, data2, size2);
+	return ret;
+}
+
+void zbud_remotify_pages(int nr)
+{
+	struct zbud_page *zbpg;
+	int i, ret;
+
+	/*
+	 * for now just try remotifying unbuddied pages, starting with
+	 * least space avail
+	 */
+	for (i = 0; i < MAX_CHUNK; i++) {
+retry_unbud_list_i:
+		preempt_disable();  /* enable in zbud_remotify_zbpg */
+		spin_lock_bh(&zbud_budlists_spinlock);
+		if (list_empty(&zbud_unbuddied[i].list)) {
+			spin_unlock_bh(&zbud_budlists_spinlock);
+			preempt_enable();
+			continue; /* next i in for loop */
+		}
+		list_for_each_entry(zbpg, &zbud_unbuddied[i].list, bud_list) {
+			if (unlikely(!spin_trylock(&zbpg->lock)))
+				continue; /* next list_for_each_entry */
+			zbud_unbuddied[i].count--;
+			/* want budlists unlocked when doing zbpg remotify */
+			spin_unlock_bh(&zbud_budlists_spinlock);
+			ret = zbud_remotify_zbpg(zbpg);
+			/* preemption is re-enabled in zbud_remotify_zbpg */
+			if (ret == 0) {
+				if (--nr <= 0)
+					goto out;
+				goto retry_unbud_list_i;
+			}
+			/* if fail to remotify any page, quit */
+			pr_err("TESTING zbud_remotify_pages failed on page,"
+				" trying to re-add\n");
+			spin_lock_bh(&zbud_budlists_spinlock);
+			spin_lock(&zbpg->lock);
+			list_add_tail(&zbpg->bud_list, &zbud_unbuddied[i].list);
+			zbud_unbuddied[i].count++;
+			spin_unlock(&zbpg->lock);
+			spin_unlock_bh(&zbud_budlists_spinlock);
+			pr_err("TESTING zbud_remotify_pages failed on page,"
+				" finished re-add\n");
+			goto out;
+		}
+		spin_unlock_bh(&zbud_budlists_spinlock);
+		preempt_enable();
+	}
+
+next_buddied_zbpg:
+	preempt_disable();  /* enable in zbud_remotify_zbpg */
+	spin_lock_bh(&zbud_budlists_spinlock);
+	if (list_empty(&zbud_buddied_list))
+		goto unlock_out;
+	list_for_each_entry(zbpg, &zbud_buddied_list, bud_list) {
+		if (unlikely(!spin_trylock(&zbpg->lock)))
+			continue; /* next list_for_each_entry */
+		zcache_zbud_buddied_count--;
+		/* want budlists unlocked when doing zbpg remotify */
+		spin_unlock_bh(&zbud_budlists_spinlock);
+		ret = zbud_remotify_zbpg(zbpg);
+		/* preemption is re-enabled in zbud_remotify_zbpg */
+		if (ret == 0) {
+			if (--nr <= 0)
+				goto out;
+			goto next_buddied_zbpg;
+		}
+		/* if fail to remotify any page, quit */
+		pr_err("TESTING zbud_remotify_pages failed on BUDDIED page,"
+			" trying to re-add\n");
+		spin_lock_bh(&zbud_budlists_spinlock);
+		spin_lock(&zbpg->lock);
+		list_add_tail(&zbpg->bud_list, &zbud_buddied_list);
+		zcache_zbud_buddied_count++;
+		spin_unlock(&zbpg->lock);
+		spin_unlock_bh(&zbud_budlists_spinlock);
+		pr_err("TESTING zbud_remotify_pages failed on BUDDIED page,"
+			" finished re-add\n");
+		goto out;
+	}
+unlock_out:
+	spin_unlock_bh(&zbud_budlists_spinlock);
+	preempt_enable();
 out:
 	return;
 }
 
+/* the "flush list" asynchronously collects pages to remotely flush */
+#define FLUSH_ENTIRE_OBJECT ((uint32_t)-1)
+static void ramster_flnode_free(struct flushlist_node *,
+				struct tmem_pool *);
+
+static void zcache_remote_flush_page(struct flushlist_node *flnode)
+{
+	struct tmem_xhandle *xh;
+	int remotenode, ret;
+
+	preempt_disable();
+	xh = &flnode->xh;
+	remotenode = flnode->xh.client_id;
+	ret = ramster_remote_flush(xh, remotenode);
+	if (ret >= 0)
+		ramster_remote_pages_flushed++;
+	else
+		ramster_remote_page_flushes_failed++;
+	preempt_enable_no_resched();
+	ramster_flnode_free(flnode, NULL);
+}
+
+static void zcache_remote_flush_object(struct flushlist_node *flnode)
+{
+	struct tmem_xhandle *xh;
+	int remotenode, ret;
+
+	preempt_disable();
+	xh = &flnode->xh;
+	remotenode = flnode->xh.client_id;
+	ret = ramster_remote_flush_object(xh, remotenode);
+	if (ret >= 0)
+		ramster_remote_objects_flushed++;
+	else
+		ramster_remote_object_flushes_failed++;
+	preempt_enable_no_resched();
+	ramster_flnode_free(flnode, NULL);
+}
+
+static void zcache_remote_eph_put(struct zbud_hdr *zbud)
+{
+	/* FIXME */
+}
+
+static void zcache_remote_pers_put(struct zv_hdr *zv)
+{
+	struct tmem_xhandle xh;
+	uint16_t size;
+	bool ephemeral;
+	int remotenode, ret = -1;
+	char *data;
+	struct tmem_pool *pool;
+	unsigned long flags;
+	unsigned char cksum;
+	char *p;
+	int i;
+	unsigned char *tmpmem = __get_cpu_var(zcache_remoteputmem);
+
+	ASSERT_SENTINEL(zv, ZVH);
+	BUG_ON(zv->client_id != LOCAL_CLIENT);
+	local_bh_disable();
+	xh.client_id = zv->client_id;
+	xh.pool_id = zv->pool_id;
+	xh.oid = zv->oid;
+	xh.index = zv->index;
+	size = xv_get_object_size(zv) - sizeof(*zv);
+	BUG_ON(size == 0 || size > zv_max_page_size);
+	data = (char *)zv + sizeof(*zv);
+	for (p = data, cksum = 0, i = 0; i < size; i++)
+		cksum += *p;
+	memcpy(tmpmem, data, size);
+	data = tmpmem;
+	pool = zcache_get_pool_by_id(zv->client_id, zv->pool_id);
+	ephemeral = is_ephemeral(pool);
+	zcache_put_pool(pool);
+	/* now OK to release lock set in caller */
+	spin_unlock(&zcache_rem_op_list_lock);
+	local_bh_enable();
+	preempt_disable();
+	ret = ramster_remote_put(&xh, data, size, ephemeral, &remotenode);
+	preempt_enable_no_resched();
+	if (ret != 0) {
+		/*
+		 * This is some form of a memory leak... if the remote put
+		 * fails, there will never be another attempt to remotify
+		 * this page.  But since we've dropped the zv pointer,
+		 * the page may have been freed or the data replaced
+		 * so we can't just "put it back" in the remote op list.
+		 * Even if we could, not sure where to put it in the list
+		 * because there may be flushes that must be strictly
+		 * ordered vs the put.  So leave this as a FIXME for now.
+		 * But count them so we know if it becomes a problem.
+		 */
+		ramster_pers_pages_remote_failed++;
+		goto out;
+	} else
+		atomic_inc(&ramster_remote_pers_pages);
+	ramster_pers_pages_remoted++;
+	/*
+	 * data was successfully remoted so change the local version to
+	 * point to the remote node where it landed
+	 */
+	local_bh_disable();
+	pool = zcache_get_pool_by_id(LOCAL_CLIENT, xh.pool_id);
+	local_irq_save(flags);
+	(void)tmem_replace(pool, &xh.oid, xh.index,
+			pampd_make_remote(remotenode, size, cksum));
+	local_irq_restore(flags);
+	zcache_put_pool(pool);
+	local_bh_enable();
+out:
+	return;
+}
+
+static void zcache_do_remotify_ops(int nr)
+{
+	struct ramster_remotify_hdr *rem_op;
+	union remotify_list_node *u;
+
+	while (1) {
+		if (!nr)
+			goto out;
+		spin_lock(&zcache_rem_op_list_lock);
+		if (list_empty(&zcache_rem_op_list)) {
+			spin_unlock(&zcache_rem_op_list_lock);
+			goto out;
+		}
+		rem_op = list_first_entry(&zcache_rem_op_list,
+				struct ramster_remotify_hdr, list);
+		list_del_init(&rem_op->list);
+		if (rem_op->op != RAMSTER_REMOTIFY_PERS_PUT)
+			spin_unlock(&zcache_rem_op_list_lock);
+		u = (union remotify_list_node *)rem_op;
+		switch (rem_op->op) {
+		case RAMSTER_REMOTIFY_EPH_PUT:
+BUG();
+			zcache_remote_eph_put((struct zbud_hdr *)rem_op);
+			break;
+		case RAMSTER_REMOTIFY_PERS_PUT:
+			zcache_remote_pers_put((struct zv_hdr *)rem_op);
+			break;
+		case RAMSTER_REMOTIFY_FLUSH_PAGE:
+			zcache_remote_flush_page((struct flushlist_node *)u);
+			break;
+		case RAMSTER_REMOTIFY_FLUSH_OBJ:
+			zcache_remote_flush_object((struct flushlist_node *)u);
+			break;
+		default:
+			BUG();
+		}
+	}
+out:
+	return;
+}
+
+/*
+ * Communicate interface revision with userspace
+ */
+#include "cluster/ramster_nodemanager.h"
+static unsigned long ramster_interface_revision  = R2NM_API_VERSION;
+
+/*
+ * For now, just push over a few pages every few seconds to
+ * ensure that it basically works
+ */
+static struct workqueue_struct *ramster_remotify_workqueue;
+static void ramster_remotify_process(struct work_struct *work);
+static DECLARE_DELAYED_WORK(ramster_remotify_worker,
+		ramster_remotify_process);
+
+static void ramster_remotify_queue_delayed_work(unsigned long delay)
+{
+	if (!queue_delayed_work(ramster_remotify_workqueue,
+				&ramster_remotify_worker, delay))
+		pr_err("ramster_remotify: bad workqueue\n");
+}
+
+
+static int use_frontswap;
+static int use_cleancache;
+static int ramster_remote_target_nodenum = -1;
+static void ramster_remotify_process(struct work_struct *work)
+{
+	static bool remotify_in_progress;
+
+	BUG_ON(irqs_disabled());
+	if (remotify_in_progress)
+		ramster_remotify_queue_delayed_work(HZ);
+	else if (ramster_remote_target_nodenum != -1) {
+		remotify_in_progress = true;
+#ifdef CONFIG_CLEANCACHE
+	if (use_cleancache && ramster_eph_remotify_enable)
+		zbud_remotify_pages(5000); /* FIXME is this a good number? */
+#endif
+#ifdef CONFIG_FRONTSWAP
+	if (use_frontswap && ramster_pers_remotify_enable)
+		zcache_do_remotify_ops(500); /* FIXME is this a good number? */
+#endif
+		remotify_in_progress = false;
+		ramster_remotify_queue_delayed_work(HZ);
+	}
+}
+
+static void ramster_remotify_init(void)
+{
+	unsigned long n = 60UL;
+	ramster_remotify_workqueue =
+		create_singlethread_workqueue("ramster_remotify");
+	ramster_remotify_queue_delayed_work(n * HZ);
+}
+
+
 static void zbud_init(void)
 {
 	int i;
@@ -631,15 +1071,6 @@ static int zbud_show_cumul_chunk_counts(char *buf)
  * necessary for decompression) immediately preceding the compressed data.
  */
 
-#define ZVH_SENTINEL  0x43214321
-
-struct zv_hdr {
-	uint32_t pool_id;
-	struct tmem_oid oid;
-	uint32_t index;
-	DECL_SENTINEL
-};
-
 /* rudimentary policy limits */
 /* total number of persistent pages may not exceed this percentage */
 static unsigned int zv_page_count_policy_percent = 75;
@@ -655,10 +1086,11 @@ static unsigned int zv_max_zsize = (PAGE_SIZE / 8) * 7;
  */
 static unsigned int zv_max_mean_zsize = (PAGE_SIZE / 8) * 5;
 
-static unsigned long zv_curr_dist_counts[NCHUNKS];
-static unsigned long zv_cumul_dist_counts[NCHUNKS];
+static atomic_t zv_curr_dist_counts[NCHUNKS];
+static atomic_t zv_cumul_dist_counts[NCHUNKS];
+
 
-static struct zv_hdr *zv_create(struct xv_pool *xvpool, uint32_t pool_id,
+static struct zv_hdr *zv_create(struct zcache_client *cli, uint32_t pool_id,
 				struct tmem_oid *oid, uint32_t index,
 				void *cdata, unsigned clen)
 {
@@ -671,23 +1103,61 @@ static struct zv_hdr *zv_create(struct xv_pool *xvpool, uint32_t pool_id,
 
 	BUG_ON(!irqs_disabled());
 	BUG_ON(chunks >= NCHUNKS);
-	ret = xv_malloc(xvpool, alloc_size,
+	ret = xv_malloc(cli->xvpool, clen + sizeof(struct zv_hdr),
 			&page, &offset, ZCACHE_GFP_MASK);
 	if (unlikely(ret))
 		goto out;
-	zv_curr_dist_counts[chunks]++;
-	zv_cumul_dist_counts[chunks]++;
+	atomic_inc(&zv_curr_dist_counts[chunks]);
+	atomic_inc(&zv_cumul_dist_counts[chunks]);
 	zv = kmap_atomic(page, KM_USER0) + offset;
 	zv->index = index;
 	zv->oid = *oid;
 	zv->pool_id = pool_id;
 	SET_SENTINEL(zv, ZVH);
+	INIT_LIST_HEAD(&zv->rem_op.list);
+	zv->client_id = get_client_id_from_client(cli);
+	zv->rem_op.op = RAMSTER_REMOTIFY_PERS_PUT;
+	if (zv->client_id == LOCAL_CLIENT) {
+		spin_lock(&zcache_rem_op_list_lock);
+		list_add_tail(&zv->rem_op.list, &zcache_rem_op_list);
+		spin_unlock(&zcache_rem_op_list_lock);
+	}
 	memcpy((char *)zv + sizeof(struct zv_hdr), cdata, clen);
 	kunmap_atomic(zv, KM_USER0);
 out:
 	return zv;
 }
 
+/* similar to zv_create, but just reserve space, no data yet */
+static struct zv_hdr *zv_alloc(struct tmem_pool *pool,
+				struct tmem_oid *oid, uint32_t index,
+				unsigned clen)
+{
+	struct zcache_client *cli = pool->client;
+	struct page *page;
+	struct zv_hdr *zv = NULL;
+	uint32_t offset;
+	int ret;
+
+	BUG_ON(!irqs_disabled());
+	BUG_ON(!is_local_client(pool->client));
+	ret = xv_malloc(cli->xvpool, clen + sizeof(struct zv_hdr),
+			&page, &offset, ZCACHE_GFP_MASK);
+	if (unlikely(ret))
+		goto out;
+	zv = kmap_atomic(page, KM_USER0) + offset;
+	SET_SENTINEL(zv, ZVH);
+	INIT_LIST_HEAD(&zv->rem_op.list);
+	zv->client_id = LOCAL_CLIENT;
+	zv->rem_op.op = RAMSTER_INTRANSIT_PERS;
+	zv->index = index;
+	zv->oid = *oid;
+	zv->pool_id = pool->pool_id;
+	kunmap_atomic(zv, KM_USER0);
+out:
+	return zv;
+}
+
 static void zv_free(struct xv_pool *xvpool, struct zv_hdr *zv)
 {
 	unsigned long flags;
@@ -698,10 +1168,15 @@ static void zv_free(struct xv_pool *xvpool, struct zv_hdr *zv)
 
 	ASSERT_SENTINEL(zv, ZVH);
 	BUG_ON(chunks >= NCHUNKS);
-	zv_curr_dist_counts[chunks]--;
+	atomic_dec(&zv_curr_dist_counts[chunks]);
 	size -= sizeof(*zv);
+	spin_lock(&zcache_rem_op_list_lock);
+	size = xv_get_object_size(zv) - sizeof(*zv);
 	BUG_ON(size == 0);
 	INVERT_SENTINEL(zv, ZVH);
+	if (!list_empty(&zv->rem_op.list))
+		list_del_init(&zv->rem_op.list);
+	spin_unlock(&zcache_rem_op_list_lock);
 	page = virt_to_page(zv);
 	offset = (unsigned long)zv & ~PAGE_MASK;
 	local_irq_save(flags);
@@ -727,6 +1202,29 @@ static void zv_decompress(struct page *page, struct zv_hdr *zv)
 	BUG_ON(clen != PAGE_SIZE);
 }
 
+static void zv_copy_from_pampd(char *data, size_t *bufsize, struct zv_hdr *zv)
+{
+	unsigned size;
+
+	ASSERT_SENTINEL(zv, ZVH);
+	size = xv_get_object_size(zv) - sizeof(*zv);
+	BUG_ON(size == 0 || size > zv_max_page_size);
+	BUG_ON(size > *bufsize);
+	memcpy(data, (char *)zv + sizeof(*zv), size);
+	*bufsize = size;
+}
+
+static void zv_copy_to_pampd(struct zv_hdr *zv, char *data, size_t size)
+{
+	unsigned zv_size;
+
+	ASSERT_SENTINEL(zv, ZVH);
+	zv_size = xv_get_object_size(zv) - sizeof(*zv);
+	BUG_ON(zv_size != size);
+	BUG_ON(zv_size == 0 || zv_size > zv_max_page_size);
+	memcpy((char *)zv + sizeof(*zv), data, size);
+}
+
 #ifdef CONFIG_SYSFS
 /*
  * show a distribution of compression stats for zv pages.
@@ -738,7 +1236,7 @@ static int zv_curr_dist_counts_show(char *buf)
 	char *p = buf;
 
 	for (i = 0; i < NCHUNKS; i++) {
-		n = zv_curr_dist_counts[i];
+		n = atomic_read(&zv_curr_dist_counts[i]);
 		p += sprintf(p, "%lu ", n);
 		chunks += n;
 		sum_total_chunks += i * n;
@@ -754,7 +1252,7 @@ static int zv_cumul_dist_counts_show(char *buf)
 	char *p = buf;
 
 	for (i = 0; i < NCHUNKS; i++) {
-		n = zv_cumul_dist_counts[i];
+		n = atomic_read(&zv_cumul_dist_counts[i]);
 		p += sprintf(p, "%lu ", n);
 		chunks += n;
 		sum_total_chunks += i * n;
@@ -787,7 +1285,7 @@ static ssize_t zv_max_zsize_store(struct kobject *kobj,
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	err = strict_strtoul(buf, 10, &val);
+	err = kstrtoul(buf, 10, &val);
 	if (err || (val == 0) || (val > (PAGE_SIZE / 8) * 7))
 		return -EINVAL;
 	zv_max_zsize = val;
@@ -819,7 +1317,7 @@ static ssize_t zv_max_mean_zsize_store(struct kobject *kobj,
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	err = strict_strtoul(buf, 10, &val);
+	err = kstrtoul(buf, 10, &val);
 	if (err || (val == 0) || (val > (PAGE_SIZE / 8) * 7))
 		return -EINVAL;
 	zv_max_mean_zsize = val;
@@ -853,7 +1351,7 @@ static ssize_t zv_page_count_policy_percent_store(struct kobject *kobj,
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	err = strict_strtoul(buf, 10, &val);
+	err = kstrtoul(buf, 10, &val);
 	if (err || (val == 0) || (val > 150))
 		return -EINVAL;
 	zv_page_count_policy_percent = val;
@@ -890,6 +1388,7 @@ static unsigned long zcache_flush_found;
 static unsigned long zcache_flobj_total;
 static unsigned long zcache_flobj_found;
 static unsigned long zcache_failed_eph_puts;
+static unsigned long zcache_nonactive_puts;
 static unsigned long zcache_failed_pers_puts;
 
 /*
@@ -970,6 +1469,7 @@ static unsigned long zcache_put_to_flush;
  */
 static struct kmem_cache *zcache_objnode_cache;
 static struct kmem_cache *zcache_obj_cache;
+static struct kmem_cache *ramster_flnode_cache;
 static atomic_t zcache_curr_obj_count = ATOMIC_INIT(0);
 static unsigned long zcache_curr_obj_count_max;
 static atomic_t zcache_curr_objnode_count = ATOMIC_INIT(0);
@@ -985,6 +1485,7 @@ struct zcache_preload {
 	struct tmem_obj *obj;
 	int nr;
 	struct tmem_objnode *objnodes[OBJNODE_TREE_MAX_PATH];
+	struct flushlist_node *flnode;
 };
 static DEFINE_PER_CPU(struct zcache_preload, zcache_preloads) = { 0, };
 
@@ -993,6 +1494,7 @@ static int zcache_do_preload(struct tmem_pool *pool)
 	struct zcache_preload *kp;
 	struct tmem_objnode *objnode;
 	struct tmem_obj *obj;
+	struct flushlist_node *flnode;
 	void *page;
 	int ret = -ENOMEM;
 
@@ -1023,27 +1525,61 @@ static int zcache_do_preload(struct tmem_pool *pool)
 		zcache_failed_alloc++;
 		goto out;
 	}
-	page = (void *)__get_free_page(ZCACHE_GFP_MASK);
-	if (unlikely(page == NULL)) {
-		zcache_failed_get_free_pages++;
-		kmem_cache_free(zcache_obj_cache, obj);
+	flnode = kmem_cache_alloc(ramster_flnode_cache, ZCACHE_GFP_MASK);
+	if (unlikely(flnode == NULL)) {
+		zcache_failed_alloc++;
 		goto out;
 	}
+	if (is_ephemeral(pool)) {
+		page = (void *)__get_free_page(ZCACHE_GFP_MASK);
+		if (unlikely(page == NULL)) {
+			zcache_failed_get_free_pages++;
+			kmem_cache_free(zcache_obj_cache, obj);
+			kmem_cache_free(ramster_flnode_cache, flnode);
+			goto out;
+		}
+	}
 	preempt_disable();
 	kp = &__get_cpu_var(zcache_preloads);
 	if (kp->obj == NULL)
 		kp->obj = obj;
 	else
 		kmem_cache_free(zcache_obj_cache, obj);
-	if (kp->page == NULL)
-		kp->page = page;
+	if (kp->flnode == NULL)
+		kp->flnode = flnode;
 	else
-		free_page((unsigned long)page);
+		kmem_cache_free(ramster_flnode_cache, flnode);
+	if (is_ephemeral(pool)) {
+		if (kp->page == NULL)
+			kp->page = page;
+		else
+			free_page((unsigned long)page);
+	}
 	ret = 0;
 out:
 	return ret;
 }
 
+static int ramster_do_preload_flnode_only(struct tmem_pool *pool)
+{
+	struct zcache_preload *kp;
+	struct flushlist_node *flnode;
+	int ret = -ENOMEM;
+
+	BUG_ON(!irqs_disabled());
+	if (unlikely(ramster_flnode_cache == NULL))
+		BUG();
+	kp = &__get_cpu_var(zcache_preloads);
+	flnode = kmem_cache_alloc(ramster_flnode_cache, GFP_ATOMIC);
+	if (unlikely(flnode == NULL) && kp->flnode == NULL)
+		BUG();  /* FIXME handle more gracefully, but how??? */
+	else if (kp->flnode == NULL)
+		kp->flnode = flnode;
+	else
+		kmem_cache_free(ramster_flnode_cache, flnode);
+	return ret;
+}
+
 static void *zcache_get_free_page(void)
 {
 	struct zcache_preload *kp;
@@ -1116,6 +1652,30 @@ static void zcache_obj_free(struct tmem_obj *obj, struct tmem_pool *pool)
 	kmem_cache_free(zcache_obj_cache, obj);
 }
 
+static struct flushlist_node *ramster_flnode_alloc(struct tmem_pool *pool)
+{
+	struct flushlist_node *flnode = NULL;
+	struct zcache_preload *kp;
+	int count;
+
+	kp = &__get_cpu_var(zcache_preloads);
+	flnode = kp->flnode;
+	BUG_ON(flnode == NULL);
+	kp->flnode = NULL;
+	count = atomic_inc_return(&ramster_curr_flnode_count);
+	if (count > ramster_curr_flnode_count_max)
+		ramster_curr_flnode_count_max = count;
+	return flnode;
+}
+
+static void ramster_flnode_free(struct flushlist_node *flnode,
+				struct tmem_pool *pool)
+{
+	atomic_dec(&ramster_curr_flnode_count);
+	BUG_ON(atomic_read(&ramster_curr_flnode_count) < 0);
+	kmem_cache_free(ramster_flnode_cache, flnode);
+}
+
 static struct tmem_hostops zcache_hostops = {
 	.obj_alloc = zcache_obj_alloc,
 	.obj_free = zcache_obj_free,
@@ -1127,6 +1687,14 @@ static struct tmem_hostops zcache_hostops = {
  * zcache implementations for PAM page descriptor ops
  */
 
+
+static inline void dec_and_check(atomic_t *pvar)
+{
+	atomic_dec(pvar);
+	/* later when all accounting is fixed, make this a BUG */
+	WARN_ON_ONCE(atomic_read(pvar) < 0);
+}
+
 static atomic_t zcache_curr_eph_pampd_count = ATOMIC_INIT(0);
 static unsigned long zcache_curr_eph_pampd_count_max;
 static atomic_t zcache_curr_pers_pampd_count = ATOMIC_INIT(0);
@@ -1135,22 +1703,20 @@ static unsigned long zcache_curr_pers_pampd_count_max;
 /* forward reference */
 static int zcache_compress(struct page *from, void **out_va, size_t *out_len);
 
-static void *zcache_pampd_create(char *data, size_t size, bool raw, int eph,
+static int zcache_pampd_eph_create(char *data, size_t size, bool raw,
 				struct tmem_pool *pool, struct tmem_oid *oid,
-				 uint32_t index)
+				uint32_t index, void **pampd)
 {
-	void *pampd = NULL, *cdata;
-	size_t clen;
-	int ret;
-	unsigned long count;
-	struct page *page = (struct page *)(data);
+	int ret = -1;
+	void *cdata = data;
+	size_t clen = size;
 	struct zcache_client *cli = pool->client;
 	uint16_t client_id = get_client_id_from_client(cli);
-	unsigned long zv_mean_zsize;
-	unsigned long curr_pers_pampd_count;
-	u64 total_zsize;
+	struct page *page = NULL;
+	unsigned long count;
 
-	if (eph) {
+	if (!raw) {
+		page = virt_to_page(data);
 		ret = zcache_compress(page, &cdata, &clen);
 		if (ret == 0)
 			goto out;
@@ -1158,46 +1724,114 @@ static void *zcache_pampd_create(char *data, size_t size, bool raw, int eph,
 			zcache_compress_poor++;
 			goto out;
 		}
-		pampd = (void *)zbud_create(client_id, pool->pool_id, oid,
-						index, page, cdata, clen);
-		if (pampd != NULL) {
-			count = atomic_inc_return(&zcache_curr_eph_pampd_count);
-			if (count > zcache_curr_eph_pampd_count_max)
-				zcache_curr_eph_pampd_count_max = count;
-		}
-	} else {
-		curr_pers_pampd_count =
-			atomic_read(&zcache_curr_pers_pampd_count);
-		if (curr_pers_pampd_count >
-		    (zv_page_count_policy_percent * totalram_pages) / 100)
-			goto out;
-		ret = zcache_compress(page, &cdata, &clen);
-		if (ret == 0)
-			goto out;
-		/* reject if compression is too poor */
-		if (clen > zv_max_zsize) {
-			zcache_compress_poor++;
+	}
+	*pampd = (void *)zbud_create(client_id, pool->pool_id, oid,
+					index, page, cdata, clen);
+	if (*pampd == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	ret = 0;
+	count = atomic_inc_return(&zcache_curr_eph_pampd_count);
+	if (count > zcache_curr_eph_pampd_count_max)
+		zcache_curr_eph_pampd_count_max = count;
+	if (client_id != LOCAL_CLIENT) {
+		count = atomic_inc_return(&ramster_foreign_eph_pampd_count);
+		if (count > ramster_foreign_eph_pampd_count_max)
+			ramster_foreign_eph_pampd_count_max = count;
+	}
+out:
+	return ret;
+}
+
+static int zcache_pampd_pers_create(char *data, size_t size, bool raw,
+				struct tmem_pool *pool, struct tmem_oid *oid,
+				uint32_t index, void **pampd)
+{
+	int ret = -1;
+	void *cdata = data;
+	size_t clen = size;
+	struct zcache_client *cli = pool->client;
+	struct page *page;
+	unsigned long count;
+	unsigned long zv_mean_zsize;
+	struct zv_hdr *zv;
+	long curr_pers_pampd_count;
+	u64 total_zsize;
+#ifdef RAMSTER_TESTING
+	static bool pampd_neg_warned;
+#endif
+
+	curr_pers_pampd_count = atomic_read(&zcache_curr_pers_pampd_count) -
+			atomic_read(&ramster_remote_pers_pages);
+#ifdef RAMSTER_TESTING
+	/* should always be positive, but warn if accounting is off */
+	if (!pampd_neg_warned) {
+		pr_warn("ramster: bad accounting for curr_pers_pampd_count\n");
+		pampd_neg_warned = true;
+	}
+#endif
+	if (curr_pers_pampd_count >
+		    (zv_page_count_policy_percent * totalram_pages) / 100) {
+		zcache_policy_percent_exceeded++;
+		goto out;
+	}
+	if (raw)
+		goto ok_to_create;
+	page = virt_to_page(data);
+	if (zcache_compress(page, &cdata, &clen) == 0)
+		goto out;
+	/* reject if compression is too poor */
+	if (clen > zv_max_zsize) {
+		zcache_compress_poor++;
+		goto out;
+	}
+	/* reject if mean compression is too poor */
+	if ((clen > zv_max_mean_zsize) && (curr_pers_pampd_count > 0)) {
+		total_zsize = xv_get_total_size_bytes(cli->xvpool);
+		zv_mean_zsize = div_u64(total_zsize, curr_pers_pampd_count);
+		if (zv_mean_zsize > zv_max_mean_zsize) {
+			zcache_mean_compress_poor++;
 			goto out;
 		}
-		/* reject if mean compression is too poor */
-		if ((clen > zv_max_mean_zsize) && (curr_pers_pampd_count > 0)) {
-			total_zsize = xv_get_total_size_bytes(cli->xvpool);
-			zv_mean_zsize = div_u64(total_zsize,
-						curr_pers_pampd_count);
-			if (zv_mean_zsize > zv_max_mean_zsize) {
-				zcache_mean_compress_poor++;
-				goto out;
-			}
-		}
-		pampd = (void *)zv_create(cli->xvpool, pool->pool_id,
-						oid, index, cdata, clen);
-		if (pampd == NULL)
-			goto out;
-		count = atomic_inc_return(&zcache_curr_pers_pampd_count);
-		if (count > zcache_curr_pers_pampd_count_max)
-			zcache_curr_pers_pampd_count_max = count;
 	}
+ok_to_create:
+	*pampd = (void *)zv_create(cli, pool->pool_id, oid, index, cdata, clen);
+	if (*pampd == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	ret = 0;
+	count = atomic_inc_return(&zcache_curr_pers_pampd_count);
+	if (count > zcache_curr_pers_pampd_count_max)
+		zcache_curr_pers_pampd_count_max = count;
+	if (is_local_client(cli))
+		goto out;
+	zv = *(struct zv_hdr **)pampd;
+	count = atomic_inc_return(&ramster_foreign_pers_pampd_count);
+	if (count > ramster_foreign_pers_pampd_count_max)
+		ramster_foreign_pers_pampd_count_max = count;
 out:
+	return ret;
+}
+
+static void *zcache_pampd_create(char *data, size_t size, bool raw, int eph,
+				struct tmem_pool *pool, struct tmem_oid *oid,
+				uint32_t index)
+{
+	void *pampd = NULL;
+	int ret;
+	bool ephemeral;
+
+	BUG_ON(preemptible());
+	ephemeral = (eph == 1) || ((eph == 0) && is_ephemeral(pool));
+	if (ephemeral)
+		ret = zcache_pampd_eph_create(data, size, raw, pool,
+						oid, index, &pampd);
+	else
+		ret = zcache_pampd_pers_create(data, size, raw, pool,
+						oid, index, &pampd);
+	/* FIXME add some counters here for failed creates? */
 	return pampd;
 }
 
@@ -1211,75 +1845,343 @@ static int zcache_pampd_get_data(char *data, size_t *bufsize, bool raw,
 {
 	int ret = 0;
 
-	BUG_ON(is_ephemeral(pool));
-	zv_decompress((struct page *)(data), pampd);
+	BUG_ON(preemptible());
+	BUG_ON(is_ephemeral(pool)); /* Fix later for shared pools? */
+	BUG_ON(pampd_is_remote(pampd));
+	if (raw)
+		zv_copy_from_pampd(data, bufsize, pampd);
+	else
+		zv_decompress(virt_to_page(data), pampd);
 	return ret;
 }
 
-/*
- * fill the pageframe corresponding to the struct page with the data
- * from the passed pampd
- */
 static int zcache_pampd_get_data_and_free(char *data, size_t *bufsize, bool raw,
 					void *pampd, struct tmem_pool *pool,
 					struct tmem_oid *oid, uint32_t index)
 {
 	int ret = 0;
+	unsigned long flags;
+	struct zcache_client *cli = pool->client;
 
-	BUG_ON(!is_ephemeral(pool));
-	zbud_decompress((struct page *)(data), pampd);
-	zbud_free_and_delist((struct zbud_hdr *)pampd);
-	atomic_dec(&zcache_curr_eph_pampd_count);
+	BUG_ON(preemptible());
+	BUG_ON(pampd_is_remote(pampd));
+	if (is_ephemeral(pool)) {
+		local_irq_save(flags);
+		if (raw)
+			zbud_copy_from_pampd(data, bufsize, pampd);
+		else
+			ret = zbud_decompress(virt_to_page(data), pampd);
+		zbud_free_and_delist((struct zbud_hdr *)pampd);
+		local_irq_restore(flags);
+		if (!is_local_client(cli))
+			dec_and_check(&ramster_foreign_eph_pampd_count);
+		dec_and_check(&zcache_curr_eph_pampd_count);
+	} else {
+		if (is_local_client(cli))
+			BUG();
+		if (raw)
+			zv_copy_from_pampd(data, bufsize, pampd);
+		else
+			zv_decompress(virt_to_page(data), pampd);
+		zv_free(cli->xvpool, pampd);
+		if (!is_local_client(cli))
+			dec_and_check(&ramster_foreign_pers_pampd_count);
+		dec_and_check(&zcache_curr_pers_pampd_count);
+		ret = 0;
+	}
 	return ret;
 }
 
+static bool zcache_pampd_is_remote(void *pampd)
+{
+	return pampd_is_remote(pampd);
+}
+
 /*
  * free the pampd and remove it from any zcache lists
  * pampd must no longer be pointed to from any tmem data structures!
  */
 static void zcache_pampd_free(void *pampd, struct tmem_pool *pool,
-				struct tmem_oid *oid, uint32_t index)
+			      struct tmem_oid *oid, uint32_t index, bool acct)
 {
 	struct zcache_client *cli = pool->client;
-
-	if (is_ephemeral(pool)) {
+	bool eph = is_ephemeral(pool);
+	struct zv_hdr *zv;
+
+	BUG_ON(preemptible());
+	if (pampd_is_remote(pampd)) {
+		WARN_ON(acct == false);
+		if (oid == NULL) {
+			/*
+			 * a NULL oid means to ignore this pampd free
+			 * as the remote freeing will be handled elsewhere
+			 */
+		} else if (eph) {
+			/* FIXME remote flush optional but probably good idea */
+			/* FIXME get these working properly again */
+			dec_and_check(&zcache_curr_eph_pampd_count);
+		} else if (pampd_is_intransit(pampd)) {
+			/* did a pers remote get_and_free, so just free local */
+			pampd = pampd_mask_intransit_and_remote(pampd);
+			goto local_pers;
+		} else {
+			struct flushlist_node *flnode =
+				ramster_flnode_alloc(pool);
+
+			flnode->xh.client_id = pampd_remote_node(pampd);
+			flnode->xh.pool_id = pool->pool_id;
+			flnode->xh.oid = *oid;
+			flnode->xh.index = index;
+			flnode->rem_op.op = RAMSTER_REMOTIFY_FLUSH_PAGE;
+			spin_lock(&zcache_rem_op_list_lock);
+			list_add(&flnode->rem_op.list, &zcache_rem_op_list);
+			spin_unlock(&zcache_rem_op_list_lock);
+			dec_and_check(&zcache_curr_pers_pampd_count);
+			dec_and_check(&ramster_remote_pers_pages);
+		}
+	} else if (eph) {
 		zbud_free_and_delist((struct zbud_hdr *)pampd);
-		atomic_dec(&zcache_curr_eph_pampd_count);
-		BUG_ON(atomic_read(&zcache_curr_eph_pampd_count) < 0);
+		if (!is_local_client(pool->client))
+			dec_and_check(&ramster_foreign_eph_pampd_count);
+		if (acct)
+			/* FIXME get these working properly again */
+			dec_and_check(&zcache_curr_eph_pampd_count);
 	} else {
-		zv_free(cli->xvpool, (struct zv_hdr *)pampd);
-		atomic_dec(&zcache_curr_pers_pampd_count);
-		BUG_ON(atomic_read(&zcache_curr_pers_pampd_count) < 0);
+local_pers:
+		zv = (struct zv_hdr *)pampd;
+		if (!is_local_client(pool->client))
+			dec_and_check(&ramster_foreign_pers_pampd_count);
+		zv_free(cli->xvpool, zv);
+		if (acct)
+			/* FIXME get these working properly again */
+			dec_and_check(&zcache_curr_pers_pampd_count);
 	}
 }
 
-static void zcache_pampd_free_obj(struct tmem_pool *pool, struct tmem_obj *obj)
+static void zcache_pampd_free_obj(struct tmem_pool *pool,
+					struct tmem_obj *obj)
 {
+	struct flushlist_node *flnode;
+
+	BUG_ON(preemptible());
+	if (obj->extra == NULL)
+		return;
+	BUG_ON(!pampd_is_remote(obj->extra));
+	flnode = ramster_flnode_alloc(pool);
+	flnode->xh.client_id = pampd_remote_node(obj->extra);
+	flnode->xh.pool_id = pool->pool_id;
+	flnode->xh.oid = obj->oid;
+	flnode->xh.index = FLUSH_ENTIRE_OBJECT;
+	flnode->rem_op.op = RAMSTER_REMOTIFY_FLUSH_OBJ;
+	spin_lock(&zcache_rem_op_list_lock);
+	list_add(&flnode->rem_op.list, &zcache_rem_op_list);
+	spin_unlock(&zcache_rem_op_list_lock);
 }
 
-static void zcache_pampd_new_obj(struct tmem_obj *obj)
+void zcache_pampd_new_obj(struct tmem_obj *obj)
 {
+	obj->extra = NULL;
 }
 
-static int zcache_pampd_replace_in_obj(void *pampd, struct tmem_obj *obj)
+int zcache_pampd_replace_in_obj(void *new_pampd, struct tmem_obj *obj)
 {
-	return -1;
+	int ret = -1;
+
+	if (new_pampd != NULL) {
+		if (obj->extra == NULL)
+			obj->extra = new_pampd;
+		/* enforce that all remote pages in an object reside
+		 * in the same node! */
+		else if (pampd_remote_node(new_pampd) !=
+				pampd_remote_node((void *)(obj->extra)))
+			BUG();
+		ret = 0;
+	}
+	return ret;
 }
 
-static bool zcache_pampd_is_remote(void *pampd)
+/*
+ * Called by the message handler after a (still compressed) page has been
+ * fetched from the remote machine in response to an "is_remote" tmem_get
+ * or persistent tmem_localify.  For a tmem_get, "extra" is the address of
+ * the page that is to be filled to succesfully resolve the tmem_get; for
+ * a (persistent) tmem_localify, "extra" is NULL (as the data is placed only
+ * in the local zcache).  "data" points to "size" bytes of (compressed) data
+ * passed in the message.  In the case of a persistent remote get, if
+ * pre-allocation was successful (see zcache_repatriate_preload), the page
+ * is placed into both local zcache and at "extra".
+ */
+int zcache_localify(int pool_id, struct tmem_oid *oidp,
+			uint32_t index, char *data, size_t size,
+			void *extra)
 {
-	return 0;
+	int ret = -ENOENT;
+	unsigned long flags;
+	struct tmem_pool *pool;
+	bool ephemeral, delete = false;
+	size_t clen = PAGE_SIZE;
+	void *pampd, *saved_hb;
+	struct tmem_obj *obj;
+
+	pool = zcache_get_pool_by_id(LOCAL_CLIENT, pool_id);
+	if (unlikely(pool == NULL))
+		/* pool doesn't exist anymore */
+		goto out;
+	ephemeral = is_ephemeral(pool);
+	local_irq_save(flags);  /* FIXME: maybe only disable softirqs? */
+	pampd = tmem_localify_get_pampd(pool, oidp, index, &obj, &saved_hb);
+	if (pampd == NULL) {
+		/* hmmm... must have been a flush while waiting */
+#ifdef RAMSTER_TESTING
+		pr_err("UNTESTED pampd==NULL in zcache_localify\n");
+#endif
+		if (ephemeral)
+			ramster_remote_eph_pages_unsucc_get++;
+		else
+			ramster_remote_pers_pages_unsucc_get++;
+		obj = NULL;
+		goto finish;
+	} else if (unlikely(!pampd_is_remote(pampd))) {
+		/* hmmm... must have been a dup put while waiting */
+#ifdef RAMSTER_TESTING
+		pr_err("UNTESTED dup while waiting in zcache_localify\n");
+#endif
+		if (ephemeral)
+			ramster_remote_eph_pages_unsucc_get++;
+		else
+			ramster_remote_pers_pages_unsucc_get++;
+		obj = NULL;
+		pampd = NULL;
+		ret = -EEXIST;
+		goto finish;
+	} else if (size == 0) {
+		/* no remote data, delete the local is_remote pampd */
+		pampd = NULL;
+		if (ephemeral)
+			ramster_remote_eph_pages_unsucc_get++;
+		else
+			BUG();
+		delete = true;
+		goto finish;
+	}
+	if (!ephemeral && pampd_is_intransit(pampd)) {
+		/* localify to zcache */
+		pampd = pampd_mask_intransit_and_remote(pampd);
+		zv_copy_to_pampd(pampd, data, size);
+	} else {
+		pampd = NULL;
+		obj = NULL;
+	}
+	if (extra != NULL) {
+		/* decompress direct-to-memory to complete remotify */
+		ret = lzo1x_decompress_safe((char *)data, size,
+						(char *)extra, &clen);
+		BUG_ON(ret != LZO_E_OK);
+		BUG_ON(clen != PAGE_SIZE);
+	}
+	if (ephemeral)
+		ramster_remote_eph_pages_succ_get++;
+	else
+		ramster_remote_pers_pages_succ_get++;
+	ret = 0;
+finish:
+	tmem_localify_finish(obj, index, pampd, saved_hb, delete);
+	zcache_put_pool(pool);
+	local_irq_restore(flags);
+out:
+	return ret;
+}
+
+/*
+ * Called on a remote persistent tmem_get to attempt to preallocate
+ * local storage for the data contained in the remote persistent page.
+ * If succesfully preallocated, returns the pampd, marked as remote and
+ * in_transit.  Else returns NULL.  Note that the appropriate tmem data
+ * structure must be locked.
+ */
+static void *zcache_pampd_repatriate_preload(void *pampd,
+						struct tmem_pool *pool,
+						struct tmem_oid *oid,
+						uint32_t index,
+						bool *intransit)
+{
+	int clen = pampd_remote_size(pampd);
+	void *ret_pampd = NULL;
+	unsigned long flags;
+
+	if (!pampd_is_remote(pampd))
+		BUG();
+	if (is_ephemeral(pool))
+		BUG();
+	if (pampd_is_intransit(pampd)) {
+		/*
+		 * to avoid multiple allocations (and maybe a memory leak)
+		 * don't preallocate if already in the process of being
+		 * repatriated
+		 */
+		*intransit = true;
+		goto out;
+	}
+	*intransit = false;
+	local_irq_save(flags);
+	ret_pampd = (void *)zv_alloc(pool, oid, index, clen);
+	if (ret_pampd != NULL) {
+		/*
+		 *  a pampd is marked intransit if it is remote and space has
+		 *  been allocated for it locally (note, only happens for
+		 *  persistent pages, in which case the remote copy is freed)
+		 */
+		ret_pampd = pampd_mark_intransit(ret_pampd);
+		dec_and_check(&ramster_remote_pers_pages);
+	} else
+		ramster_pers_pages_remote_nomem++;
+	local_irq_restore(flags);
+out:
+	return ret_pampd;
+}
+
+/*
+ * Called on a remote tmem_get to invoke a message to fetch the page.
+ * Might sleep so no tmem locks can be held.  "extra" is passed
+ * all the way through the round-trip messaging to zcache_localify.
+ */
+static int zcache_pampd_repatriate(void *fake_pampd, void *real_pampd,
+				   struct tmem_pool *pool,
+				   struct tmem_oid *oid, uint32_t index,
+				   bool free, void *extra)
+{
+	struct tmem_xhandle xh;
+	int ret;
+
+	if (pampd_is_intransit(real_pampd))
+		/* have local space pre-reserved, so free remote copy */
+		free = true;
+	xh = tmem_xhandle_fill(LOCAL_CLIENT, pool, oid, index);
+	/* unreliable request/response for now */
+	ret = ramster_remote_async_get(&xh, free,
+					pampd_remote_node(fake_pampd),
+					pampd_remote_size(fake_pampd),
+					pampd_remote_cksum(fake_pampd),
+					extra);
+#ifdef RAMSTER_TESTING
+	if (ret != 0 && ret != -ENOENT)
+		pr_err("TESTING zcache_pampd_repatriate returns, ret=%d\n",
+			ret);
+#endif
+	return ret;
 }
 
 static struct tmem_pamops zcache_pamops = {
 	.create = zcache_pampd_create,
 	.get_data = zcache_pampd_get_data,
-	.get_data_and_free = zcache_pampd_get_data_and_free,
 	.free = zcache_pampd_free,
+	.get_data_and_free = zcache_pampd_get_data_and_free,
 	.free_obj = zcache_pampd_free_obj,
+	.is_remote = zcache_pampd_is_remote,
+	.repatriate_preload = zcache_pampd_repatriate_preload,
+	.repatriate = zcache_pampd_repatriate,
 	.new_obj = zcache_pampd_new_obj,
 	.replace_in_obj = zcache_pampd_replace_in_obj,
-	.is_remote = zcache_pampd_is_remote,
 };
 
 /*
@@ -1327,9 +2229,13 @@ static int zcache_cpu_notifier(struct notifier_block *nb,
 		per_cpu(zcache_workmem, cpu) =
 			kzalloc(LZO1X_MEM_COMPRESS,
 				GFP_KERNEL | __GFP_REPEAT);
+		per_cpu(zcache_remoteputmem, cpu) =
+			kzalloc(PAGE_SIZE, GFP_KERNEL | __GFP_REPEAT);
 		break;
 	case CPU_DEAD:
 	case CPU_UP_CANCELED:
+		kfree(per_cpu(zcache_remoteputmem, cpu));
+		per_cpu(zcache_remoteputmem, cpu) = NULL;
 		free_pages((unsigned long)per_cpu(zcache_dstmem, cpu),
 				LZO_DSTMEM_PAGE_ORDER);
 		per_cpu(zcache_dstmem, cpu) = NULL;
@@ -1346,6 +2252,10 @@ static int zcache_cpu_notifier(struct notifier_block *nb,
 			kmem_cache_free(zcache_obj_cache, kp->obj);
 			kp->obj = NULL;
 		}
+		if (kp->flnode) {
+			kmem_cache_free(ramster_flnode_cache, kp->flnode);
+			kp->flnode = NULL;
+		}
 		if (kp->page) {
 			free_page((unsigned long)kp->page);
 			kp->page = NULL;
@@ -1402,12 +2312,12 @@ ZCACHE_SYSFS_RO(flush_found);
 ZCACHE_SYSFS_RO(flobj_total);
 ZCACHE_SYSFS_RO(flobj_found);
 ZCACHE_SYSFS_RO(failed_eph_puts);
+ZCACHE_SYSFS_RO(nonactive_puts);
 ZCACHE_SYSFS_RO(failed_pers_puts);
 ZCACHE_SYSFS_RO(zbud_curr_zbytes);
 ZCACHE_SYSFS_RO(zbud_cumul_zpages);
 ZCACHE_SYSFS_RO(zbud_cumul_zbytes);
 ZCACHE_SYSFS_RO(zbud_buddied_count);
-ZCACHE_SYSFS_RO(zbpg_unused_list_count);
 ZCACHE_SYSFS_RO(evicted_raw_pages);
 ZCACHE_SYSFS_RO(evicted_unbuddied_pages);
 ZCACHE_SYSFS_RO(evicted_buddied_pages);
@@ -1416,6 +2326,7 @@ ZCACHE_SYSFS_RO(failed_alloc);
 ZCACHE_SYSFS_RO(put_to_flush);
 ZCACHE_SYSFS_RO(compress_poor);
 ZCACHE_SYSFS_RO(mean_compress_poor);
+ZCACHE_SYSFS_RO(policy_percent_exceeded);
 ZCACHE_SYSFS_RO_ATOMIC(zbud_curr_raw_pages);
 ZCACHE_SYSFS_RO_ATOMIC(zbud_curr_zpages);
 ZCACHE_SYSFS_RO_ATOMIC(curr_obj_count);
@@ -1439,7 +2350,9 @@ static struct attribute *zcache_attrs[] = {
 	&zcache_flush_found_attr.attr,
 	&zcache_flobj_found_attr.attr,
 	&zcache_failed_eph_puts_attr.attr,
+	&zcache_nonactive_puts_attr.attr,
 	&zcache_failed_pers_puts_attr.attr,
+	&zcache_policy_percent_exceeded_attr.attr,
 	&zcache_compress_poor_attr.attr,
 	&zcache_mean_compress_poor_attr.attr,
 	&zcache_zbud_curr_raw_pages_attr.attr,
@@ -1448,7 +2361,6 @@ static struct attribute *zcache_attrs[] = {
 	&zcache_zbud_cumul_zpages_attr.attr,
 	&zcache_zbud_cumul_zbytes_attr.attr,
 	&zcache_zbud_buddied_count_attr.attr,
-	&zcache_zbpg_unused_list_count_attr.attr,
 	&zcache_evicted_raw_pages_attr.attr,
 	&zcache_evicted_unbuddied_pages_attr.attr,
 	&zcache_evicted_buddied_pages_attr.attr,
@@ -1470,6 +2382,202 @@ static struct attribute_group zcache_attr_group = {
 	.name = "zcache",
 };
 
+#define RAMSTER_SYSFS_RO(_name) \
+	static ssize_t ramster_##_name##_show(struct kobject *kobj, \
+				struct kobj_attribute *attr, char *buf) \
+	{ \
+		return sprintf(buf, "%lu\n", ramster_##_name); \
+	} \
+	static struct kobj_attribute ramster_##_name##_attr = { \
+		.attr = { .name = __stringify(_name), .mode = 0444 }, \
+		.show = ramster_##_name##_show, \
+	}
+
+#define RAMSTER_SYSFS_RW(_name) \
+	static ssize_t ramster_##_name##_show(struct kobject *kobj, \
+				struct kobj_attribute *attr, char *buf) \
+	{ \
+		return sprintf(buf, "%lu\n", ramster_##_name); \
+	} \
+	static ssize_t ramster_##_name##_store(struct kobject *kobj, \
+		struct kobj_attribute *attr, const char *buf, size_t count) \
+	{ \
+		int err; \
+		unsigned long enable; \
+		err = kstrtoul(buf, 10, &enable); \
+		if (err) \
+			return -EINVAL; \
+		ramster_##_name = enable; \
+		return count; \
+	} \
+	static struct kobj_attribute ramster_##_name##_attr = { \
+		.attr = { .name = __stringify(_name), .mode = 0644 }, \
+		.show = ramster_##_name##_show, \
+		.store = ramster_##_name##_store, \
+	}
+
+#define RAMSTER_SYSFS_RO_ATOMIC(_name) \
+	static ssize_t ramster_##_name##_show(struct kobject *kobj, \
+				struct kobj_attribute *attr, char *buf) \
+	{ \
+	    return sprintf(buf, "%d\n", atomic_read(&ramster_##_name)); \
+	} \
+	static struct kobj_attribute ramster_##_name##_attr = { \
+		.attr = { .name = __stringify(_name), .mode = 0444 }, \
+		.show = ramster_##_name##_show, \
+	}
+
+RAMSTER_SYSFS_RO(interface_revision);
+RAMSTER_SYSFS_RO_ATOMIC(remote_pers_pages);
+RAMSTER_SYSFS_RW(pers_remotify_enable);
+RAMSTER_SYSFS_RW(eph_remotify_enable);
+RAMSTER_SYSFS_RO(eph_pages_remoted);
+RAMSTER_SYSFS_RO(eph_pages_remote_failed);
+RAMSTER_SYSFS_RO(pers_pages_remoted);
+RAMSTER_SYSFS_RO(pers_pages_remote_failed);
+RAMSTER_SYSFS_RO(pers_pages_remote_nomem);
+RAMSTER_SYSFS_RO(remote_pages_flushed);
+RAMSTER_SYSFS_RO(remote_page_flushes_failed);
+RAMSTER_SYSFS_RO(remote_objects_flushed);
+RAMSTER_SYSFS_RO(remote_object_flushes_failed);
+RAMSTER_SYSFS_RO(remote_eph_pages_succ_get);
+RAMSTER_SYSFS_RO(remote_eph_pages_unsucc_get);
+RAMSTER_SYSFS_RO(remote_pers_pages_succ_get);
+RAMSTER_SYSFS_RO(remote_pers_pages_unsucc_get);
+RAMSTER_SYSFS_RO_ATOMIC(foreign_eph_pampd_count);
+RAMSTER_SYSFS_RO(foreign_eph_pampd_count_max);
+RAMSTER_SYSFS_RO_ATOMIC(foreign_pers_pampd_count);
+RAMSTER_SYSFS_RO(foreign_pers_pampd_count_max);
+RAMSTER_SYSFS_RO_ATOMIC(curr_flnode_count);
+RAMSTER_SYSFS_RO(curr_flnode_count_max);
+
+#define MANUAL_NODES 8
+static bool ramster_nodes_manual_up[MANUAL_NODES];
+static ssize_t ramster_manual_node_up_show(struct kobject *kobj,
+				struct kobj_attribute *attr, char *buf)
+{
+	int i;
+	char *p = buf;
+	for (i = 0; i < MANUAL_NODES; i++)
+		if (ramster_nodes_manual_up[i])
+			p += sprintf(p, "%d ", i);
+	p += sprintf(p, "\n");
+	return p - buf;
+}
+
+static ssize_t ramster_manual_node_up_store(struct kobject *kobj,
+		struct kobj_attribute *attr, const char *buf, size_t count)
+{
+	int err;
+	unsigned long node_num;
+
+	err = kstrtoul(buf, 10, &node_num);
+	if (err) {
+		pr_err("ramster: bad strtoul?\n");
+		return -EINVAL;
+	}
+	if (node_num >= MANUAL_NODES) {
+		pr_err("ramster: bad node_num=%lu?\n", node_num);
+		return -EINVAL;
+	}
+	if (ramster_nodes_manual_up[node_num]) {
+		pr_err("ramster: node %d already up, ignoring\n",
+							(int)node_num);
+	} else {
+		ramster_nodes_manual_up[node_num] = true;
+		r2net_hb_node_up_manual((int)node_num);
+	}
+	return count;
+}
+
+static struct kobj_attribute ramster_manual_node_up_attr = {
+	.attr = { .name = "manual_node_up", .mode = 0644 },
+	.show = ramster_manual_node_up_show,
+	.store = ramster_manual_node_up_store,
+};
+
+static ssize_t ramster_remote_target_nodenum_show(struct kobject *kobj,
+				struct kobj_attribute *attr, char *buf)
+{
+	if (ramster_remote_target_nodenum == -1UL)
+		return sprintf(buf, "unset\n");
+	else
+		return sprintf(buf, "%d\n", ramster_remote_target_nodenum);
+}
+
+static ssize_t ramster_remote_target_nodenum_store(struct kobject *kobj,
+		struct kobj_attribute *attr, const char *buf, size_t count)
+{
+	int err;
+	unsigned long node_num;
+
+	err = kstrtoul(buf, 10, &node_num);
+	if (err) {
+		pr_err("ramster: bad strtoul?\n");
+		return -EINVAL;
+	} else if (node_num == -1UL) {
+		pr_err("ramster: disabling all remotification, "
+			"data may still reside on remote nodes however\n");
+		return -EINVAL;
+	} else if (node_num >= MANUAL_NODES) {
+		pr_err("ramster: bad node_num=%lu?\n", node_num);
+		return -EINVAL;
+	} else if (!ramster_nodes_manual_up[node_num]) {
+		pr_err("ramster: node %d not up, ignoring setting "
+			"of remotification target\n", (int)node_num);
+	} else if (r2net_remote_target_node_set((int)node_num) >= 0) {
+		pr_info("ramster: node %d set as remotification target\n",
+				(int)node_num);
+		ramster_remote_target_nodenum = (int)node_num;
+	} else {
+		pr_err("ramster: bad num to node node_num=%d?\n",
+				(int)node_num);
+		return -EINVAL;
+	}
+	return count;
+}
+
+static struct kobj_attribute ramster_remote_target_nodenum_attr = {
+	.attr = { .name = "remote_target_nodenum", .mode = 0644 },
+	.show = ramster_remote_target_nodenum_show,
+	.store = ramster_remote_target_nodenum_store,
+};
+
+
+static struct attribute *ramster_attrs[] = {
+	&ramster_interface_revision_attr.attr,
+	&ramster_pers_remotify_enable_attr.attr,
+	&ramster_eph_remotify_enable_attr.attr,
+	&ramster_remote_pers_pages_attr.attr,
+	&ramster_eph_pages_remoted_attr.attr,
+	&ramster_eph_pages_remote_failed_attr.attr,
+	&ramster_pers_pages_remoted_attr.attr,
+	&ramster_pers_pages_remote_failed_attr.attr,
+	&ramster_pers_pages_remote_nomem_attr.attr,
+	&ramster_remote_pages_flushed_attr.attr,
+	&ramster_remote_page_flushes_failed_attr.attr,
+	&ramster_remote_objects_flushed_attr.attr,
+	&ramster_remote_object_flushes_failed_attr.attr,
+	&ramster_remote_eph_pages_succ_get_attr.attr,
+	&ramster_remote_eph_pages_unsucc_get_attr.attr,
+	&ramster_remote_pers_pages_succ_get_attr.attr,
+	&ramster_remote_pers_pages_unsucc_get_attr.attr,
+	&ramster_foreign_eph_pampd_count_attr.attr,
+	&ramster_foreign_eph_pampd_count_max_attr.attr,
+	&ramster_foreign_pers_pampd_count_attr.attr,
+	&ramster_foreign_pers_pampd_count_max_attr.attr,
+	&ramster_curr_flnode_count_attr.attr,
+	&ramster_curr_flnode_count_max_attr.attr,
+	&ramster_manual_node_up_attr.attr,
+	&ramster_remote_target_nodenum_attr.attr,
+	NULL,
+};
+
+static struct attribute_group ramster_attr_group = {
+	.attrs = ramster_attrs,
+	.name = "ramster",
+};
+
 #endif /* CONFIG_SYSFS */
 /*
  * When zcache is disabled ("frozen"), pools can be created and destroyed,
@@ -1510,8 +2618,9 @@ static struct shrinker zcache_shrinker = {
  * zcache shims between cleancache/frontswap ops and tmem
  */
 
-static int zcache_put_page(int cli_id, int pool_id, struct tmem_oid *oidp,
-				uint32_t index, struct page *page)
+int zcache_put(int cli_id, int pool_id, struct tmem_oid *oidp,
+			uint32_t index, char *data, size_t size,
+			bool raw, int ephemeral)
 {
 	struct tmem_pool *pool;
 	int ret = -1;
@@ -1522,8 +2631,7 @@ static int zcache_put_page(int cli_id, int pool_id, struct tmem_oid *oidp,
 		goto out;
 	if (!zcache_freeze && zcache_do_preload(pool) == 0) {
 		/* preload does preempt_disable on success */
-		ret = tmem_put(pool, oidp, index, (char *)(page),
-				PAGE_SIZE, 0, is_ephemeral(pool));
+		ret = tmem_put(pool, oidp, index, data, size, raw, ephemeral);
 		if (ret < 0) {
 			if (is_ephemeral(pool))
 				zcache_failed_eph_puts++;
@@ -1543,27 +2651,38 @@ out:
 	return ret;
 }
 
-static int zcache_get_page(int cli_id, int pool_id, struct tmem_oid *oidp,
-				uint32_t index, struct page *page)
+int zcache_get(int cli_id, int pool_id, struct tmem_oid *oidp,
+			uint32_t index, char *data, size_t *sizep,
+			bool raw, int get_and_free)
 {
 	struct tmem_pool *pool;
 	int ret = -1;
-	unsigned long flags;
-	size_t size = PAGE_SIZE;
+	bool eph;
 
-	local_irq_save(flags);
+	if (!raw) {
+		BUG_ON(irqs_disabled());
+		BUG_ON(in_softirq());
+	}
 	pool = zcache_get_pool_by_id(cli_id, pool_id);
+	eph = is_ephemeral(pool);
 	if (likely(pool != NULL)) {
 		if (atomic_read(&pool->obj_count) > 0)
-			ret = tmem_get(pool, oidp, index, (char *)(page),
-					&size, 0, is_ephemeral(pool));
+			ret = tmem_get(pool, oidp, index, data, sizep,
+					raw, get_and_free);
 		zcache_put_pool(pool);
 	}
-	local_irq_restore(flags);
+	WARN_ONCE((!eph && (ret != 0)), "zcache_get fails on persistent pool, "
+			  "bad things are very likely to happen soon\n");
+#ifdef RAMSTER_TESTING
+	if (ret != 0 && ret != -1 && !(ret == -EINVAL && is_ephemeral(pool)))
+		pr_err("TESTING zcache_get tmem_get returns ret=%d\n", ret);
+#endif
+	if (ret == -EAGAIN)
+		BUG(); /* FIXME... don't need this anymore??? let's ensure */
 	return ret;
 }
 
-static int zcache_flush_page(int cli_id, int pool_id,
+int zcache_flush(int cli_id, int pool_id,
 				struct tmem_oid *oidp, uint32_t index)
 {
 	struct tmem_pool *pool;
@@ -1573,6 +2692,7 @@ static int zcache_flush_page(int cli_id, int pool_id,
 	local_irq_save(flags);
 	zcache_flush_total++;
 	pool = zcache_get_pool_by_id(cli_id, pool_id);
+	ramster_do_preload_flnode_only(pool);
 	if (likely(pool != NULL)) {
 		if (atomic_read(&pool->obj_count) > 0)
 			ret = tmem_flush_page(pool, oidp, index);
@@ -1584,8 +2704,7 @@ static int zcache_flush_page(int cli_id, int pool_id,
 	return ret;
 }
 
-static int zcache_flush_object(int cli_id, int pool_id,
-				struct tmem_oid *oidp)
+int zcache_flush_object(int cli_id, int pool_id, struct tmem_oid *oidp)
 {
 	struct tmem_pool *pool;
 	int ret = -1;
@@ -1594,6 +2713,7 @@ static int zcache_flush_object(int cli_id, int pool_id,
 	local_irq_save(flags);
 	zcache_flobj_total++;
 	pool = zcache_get_pool_by_id(cli_id, pool_id);
+	ramster_do_preload_flnode_only(pool);
 	if (likely(pool != NULL)) {
 		if (atomic_read(&pool->obj_count) > 0)
 			ret = tmem_flush_object(pool, oidp);
@@ -1605,7 +2725,7 @@ static int zcache_flush_object(int cli_id, int pool_id,
 	return ret;
 }
 
-static int zcache_destroy_pool(int cli_id, int pool_id)
+int zcache_client_destroy_pool(int cli_id, int pool_id)
 {
 	struct tmem_pool *pool = NULL;
 	struct zcache_client *cli = NULL;
@@ -1632,13 +2752,17 @@ static int zcache_destroy_pool(int cli_id, int pool_id)
 	ret = tmem_destroy_pool(pool);
 	local_bh_enable();
 	kfree(pool);
-	pr_info("zcache: destroyed pool id=%d, cli_id=%d\n",
-			pool_id, cli_id);
+	pr_info("ramster: destroyed pool id=%d cli_id=%d\n", pool_id, cli_id);
 out:
 	return ret;
 }
 
-static int zcache_new_pool(uint16_t cli_id, uint32_t flags)
+static int zcache_destroy_pool(int pool_id)
+{
+	return zcache_client_destroy_pool(LOCAL_CLIENT, pool_id);
+}
+
+int zcache_new_pool(uint16_t cli_id, uint32_t flags)
 {
 	int poolid = -1;
 	struct tmem_pool *pool;
@@ -1653,7 +2777,7 @@ static int zcache_new_pool(uint16_t cli_id, uint32_t flags)
 	atomic_inc(&cli->refcount);
 	pool = kmalloc(sizeof(struct tmem_pool), GFP_ATOMIC);
 	if (pool == NULL) {
-		pr_info("zcache: pool creation failed: out of memory\n");
+		pr_info("ramster: pool creation failed: out of memory\n");
 		goto out;
 	}
 
@@ -1661,7 +2785,7 @@ static int zcache_new_pool(uint16_t cli_id, uint32_t flags)
 		if (cli->tmem_pools[poolid] == NULL)
 			break;
 	if (poolid >= MAX_POOLS_PER_CLIENT) {
-		pr_info("zcache: pool creation failed: max exceeded\n");
+		pr_info("ramster: pool creation failed: max exceeded\n");
 		kfree(pool);
 		poolid = -1;
 		goto out;
@@ -1671,15 +2795,78 @@ static int zcache_new_pool(uint16_t cli_id, uint32_t flags)
 	pool->pool_id = poolid;
 	tmem_new_pool(pool, flags);
 	cli->tmem_pools[poolid] = pool;
-	pr_info("zcache: created %s tmem pool, id=%d, client=%d\n",
-		flags & TMEM_POOL_PERSIST ? "persistent" : "ephemeral",
-		poolid, cli_id);
+	if (cli_id == LOCAL_CLIENT)
+		pr_info("ramster: created %s tmem pool, id=%d, local client\n",
+			flags & TMEM_POOL_PERSIST ? "persistent" : "ephemeral",
+			poolid);
+	else
+		pr_info("ramster: created %s tmem pool, id=%d, client=%d\n",
+			flags & TMEM_POOL_PERSIST ? "persistent" : "ephemeral",
+			poolid, cli_id);
 out:
 	if (cli != NULL)
 		atomic_dec(&cli->refcount);
 	return poolid;
 }
 
+static int zcache_local_new_pool(uint32_t flags)
+{
+	return zcache_new_pool(LOCAL_CLIENT, flags);
+}
+
+int zcache_autocreate_pool(int cli_id, int pool_id, bool ephemeral)
+{
+	struct tmem_pool *pool;
+	struct zcache_client *cli = NULL;
+	uint32_t flags = ephemeral ? 0 : TMEM_POOL_PERSIST;
+	int ret = -1;
+
+	if (cli_id == LOCAL_CLIENT)
+		goto out;
+	if (pool_id >= MAX_POOLS_PER_CLIENT)
+		goto out;
+	else if ((unsigned int)cli_id < MAX_CLIENTS)
+		cli = &zcache_clients[cli_id];
+	if ((ephemeral && !use_cleancache) || (!ephemeral && !use_frontswap))
+		BUG(); /* FIXME, handle more gracefully later */
+	if (!cli->allocated) {
+		if (zcache_new_client(cli_id))
+			BUG(); /* FIXME, handle more gracefully later */
+		cli = &zcache_clients[cli_id];
+	}
+	atomic_inc(&cli->refcount);
+	pool = cli->tmem_pools[pool_id];
+	if (pool != NULL) {
+		if (pool->persistent && ephemeral) {
+			pr_err("zcache_autocreate_pool: type mismatch\n");
+			goto out;
+		}
+		ret = 0;
+		goto out;
+	}
+	pool = kmalloc(sizeof(struct tmem_pool), GFP_KERNEL);
+	if (pool == NULL) {
+		pr_info("ramster: pool creation failed: out of memory\n");
+		goto out;
+	}
+	atomic_set(&pool->refcount, 0);
+	pool->client = cli;
+	pool->pool_id = pool_id;
+	tmem_new_pool(pool, flags);
+	cli->tmem_pools[pool_id] = pool;
+	pr_info("ramster: AUTOcreated %s tmem poolid=%d, for remote client=%d\n",
+		flags & TMEM_POOL_PERSIST ? "persistent" : "ephemeral",
+		pool_id, cli_id);
+	ret = 0;
+out:
+	if (cli == NULL)
+		BUG(); /* FIXME, handle more gracefully later */
+		/* pr_err("zcache_autocreate_pool: failed\n"); */
+	if (cli != NULL)
+		atomic_dec(&cli->refcount);
+	return ret;
+}
+
 /**********
  * Two kernel functionalities currently can be layered on top of tmem.
  * These are "cleancache" which is used as a second-chance cache for clean
@@ -1696,8 +2883,18 @@ static void zcache_cleancache_put_page(int pool_id,
 	u32 ind = (u32) index;
 	struct tmem_oid oid = *(struct tmem_oid *)&key;
 
-	if (likely(ind == index))
-		(void)zcache_put_page(LOCAL_CLIENT, pool_id, &oid, index, page);
+#ifdef __PG_WAS_ACTIVE
+	if (!PageWasActive(page)) {
+		zcache_nonactive_puts++;
+		return;
+	}
+#endif
+	if (likely(ind == index)) {
+		char *kva = page_address(page);
+
+		(void)zcache_put(LOCAL_CLIENT, pool_id, &oid, index,
+			kva, PAGE_SIZE, 0, 1);
+	}
 }
 
 static int zcache_cleancache_get_page(int pool_id,
@@ -1708,8 +2905,19 @@ static int zcache_cleancache_get_page(int pool_id,
 	struct tmem_oid oid = *(struct tmem_oid *)&key;
 	int ret = -1;
 
-	if (likely(ind == index))
-		ret = zcache_get_page(LOCAL_CLIENT, pool_id, &oid, index, page);
+	preempt_disable();
+	if (likely(ind == index)) {
+		char *kva = page_address(page);
+		size_t size = PAGE_SIZE;
+
+		ret = zcache_get(LOCAL_CLIENT, pool_id, &oid, index,
+			kva, &size, 0, 0);
+#ifdef __PG_WAS_ACTIVE
+		if (ret == 0)
+			SetPageWasActive(page);
+#endif
+	}
+	preempt_enable();
 	return ret;
 }
 
@@ -1721,7 +2929,7 @@ static void zcache_cleancache_flush_page(int pool_id,
 	struct tmem_oid oid = *(struct tmem_oid *)&key;
 
 	if (likely(ind == index))
-		(void)zcache_flush_page(LOCAL_CLIENT, pool_id, &oid, ind);
+		(void)zcache_flush(LOCAL_CLIENT, pool_id, &oid, ind);
 }
 
 static void zcache_cleancache_flush_inode(int pool_id,
@@ -1735,7 +2943,7 @@ static void zcache_cleancache_flush_inode(int pool_id,
 static void zcache_cleancache_flush_fs(int pool_id)
 {
 	if (pool_id >= 0)
-		(void)zcache_destroy_pool(LOCAL_CLIENT, pool_id);
+		(void)zcache_destroy_pool(pool_id);
 }
 
 static int zcache_cleancache_init_fs(size_t pagesize)
@@ -1743,7 +2951,7 @@ static int zcache_cleancache_init_fs(size_t pagesize)
 	BUG_ON(sizeof(struct cleancache_filekey) !=
 				sizeof(struct tmem_oid));
 	BUG_ON(pagesize != PAGE_SIZE);
-	return zcache_new_pool(LOCAL_CLIENT, 0);
+	return zcache_local_new_pool(0);
 }
 
 static int zcache_cleancache_init_shared_fs(char *uuid, size_t pagesize)
@@ -1752,7 +2960,7 @@ static int zcache_cleancache_init_shared_fs(char *uuid, size_t pagesize)
 	BUG_ON(sizeof(struct cleancache_filekey) !=
 				sizeof(struct tmem_oid));
 	BUG_ON(pagesize != PAGE_SIZE);
-	return zcache_new_pool(LOCAL_CLIENT, 0);
+	return zcache_local_new_pool(0);
 }
 
 static struct cleancache_ops zcache_cleancache_ops = {
@@ -1781,10 +2989,8 @@ static int zcache_frontswap_poolid = -1;
 /*
  * Swizzling increases objects per swaptype, increasing tmem concurrency
  * for heavy swaploads.  Later, larger nr_cpus -> larger SWIZ_BITS
- * Setting SWIZ_BITS to 27 basically reconstructs the swap entry from
- * frontswap_get_page()
  */
-#define SWIZ_BITS		27
+#define SWIZ_BITS		8
 #define SWIZ_MASK		((1 << SWIZ_BITS) - 1)
 #define _oswiz(_type, _ind)	((_type << SWIZ_BITS) | (_ind & SWIZ_MASK))
 #define iswiz(_ind)		(_ind >> SWIZ_BITS)
@@ -1804,12 +3010,14 @@ static int zcache_frontswap_put_page(unsigned type, pgoff_t offset,
 	struct tmem_oid oid = oswiz(type, ind);
 	int ret = -1;
 	unsigned long flags;
+	char *kva;
 
 	BUG_ON(!PageLocked(page));
 	if (likely(ind64 == ind)) {
 		local_irq_save(flags);
-		ret = zcache_put_page(LOCAL_CLIENT, zcache_frontswap_poolid,
-					&oid, iswiz(ind), page);
+		kva = page_address(page);
+		ret = zcache_put(LOCAL_CLIENT, zcache_frontswap_poolid,
+				&oid, iswiz(ind), kva, PAGE_SIZE, 0, 0);
 		local_irq_restore(flags);
 	}
 	return ret;
@@ -1825,10 +3033,16 @@ static int zcache_frontswap_get_page(unsigned type, pgoff_t offset,
 	struct tmem_oid oid = oswiz(type, ind);
 	int ret = -1;
 
+	preempt_disable(); /* FIXME, remove this? */
 	BUG_ON(!PageLocked(page));
-	if (likely(ind64 == ind))
-		ret = zcache_get_page(LOCAL_CLIENT, zcache_frontswap_poolid,
-					&oid, iswiz(ind), page);
+	if (likely(ind64 == ind)) {
+		char *kva = page_address(page);
+		size_t size = PAGE_SIZE;
+
+		ret = zcache_get(LOCAL_CLIENT, zcache_frontswap_poolid,
+					&oid, iswiz(ind), kva, &size, 0, -1);
+	}
+	preempt_enable(); /* FIXME, remove this? */
 	return ret;
 }
 
@@ -1840,7 +3054,7 @@ static void zcache_frontswap_flush_page(unsigned type, pgoff_t offset)
 	struct tmem_oid oid = oswiz(type, ind);
 
 	if (likely(ind64 == ind))
-		(void)zcache_flush_page(LOCAL_CLIENT, zcache_frontswap_poolid,
+		(void)zcache_flush(LOCAL_CLIENT, zcache_frontswap_poolid,
 					&oid, iswiz(ind));
 }
 
@@ -1862,7 +3076,7 @@ static void zcache_frontswap_init(unsigned ignored)
 	/* a single tmem poolid is used for all frontswap "types" (swapfiles) */
 	if (zcache_frontswap_poolid < 0)
 		zcache_frontswap_poolid =
-			zcache_new_pool(LOCAL_CLIENT, TMEM_POOL_PERSIST);
+				zcache_local_new_pool(TMEM_POOL_PERSIST);
 }
 
 static struct frontswap_ops zcache_frontswap_ops = {
@@ -1883,19 +3097,125 @@ struct frontswap_ops zcache_frontswap_register_ops(void)
 #endif
 
 /*
+ * frontswap selfshrinking
+ */
+
+#ifdef CONFIG_FRONTSWAP
+/* In HZ, controls frequency of worker invocation. */
+static unsigned int selfshrink_interval __read_mostly = 5;
+
+static void selfshrink_process(struct work_struct *work);
+static DECLARE_DELAYED_WORK(selfshrink_worker, selfshrink_process);
+
+/* Enable/disable with sysfs. */
+static bool frontswap_selfshrinking __read_mostly;
+
+/* Enable/disable with kernel boot option. */
+static bool use_frontswap_selfshrink __initdata = true;
+
+/*
+ * The default values for the following parameters were deemed reasonable
+ * by experimentation, may be workload-dependent, and can all be
+ * adjusted via sysfs.
+ */
+
+/* Control rate for frontswap shrinking. Higher hysteresis is slower. */
+static unsigned int frontswap_hysteresis __read_mostly = 20;
+
+/*
+ * Number of selfshrink worker invocations to wait before observing that
+ * frontswap selfshrinking should commence. Note that selfshrinking does
+ * not use a separate worker thread.
+ */
+static unsigned int frontswap_inertia __read_mostly = 3;
+
+/* Countdown to next invocation of frontswap_shrink() */
+static unsigned long frontswap_inertia_counter;
+
+/*
+ * Invoked by the selfshrink worker thread, uses current number of pages
+ * in frontswap (frontswap_curr_pages()), previous status, and control
+ * values (hysteresis and inertia) to determine if frontswap should be
+ * shrunk and what the new frontswap size should be.  Note that
+ * frontswap_shrink is essentially a partial swapoff that immediately
+ * transfers pages from the "swap device" (frontswap) back into kernel
+ * RAM; despite the name, frontswap "shrinking" is very different from
+ * the "shrinker" interface used by the kernel MM subsystem to reclaim
+ * memory.
+ */
+static void frontswap_selfshrink(void)
+{
+	static unsigned long cur_frontswap_pages;
+	static unsigned long last_frontswap_pages;
+	static unsigned long tgt_frontswap_pages;
+
+	last_frontswap_pages = cur_frontswap_pages;
+	cur_frontswap_pages = frontswap_curr_pages();
+	if (!cur_frontswap_pages ||
+			(cur_frontswap_pages > last_frontswap_pages)) {
+		frontswap_inertia_counter = frontswap_inertia;
+		return;
+	}
+	if (frontswap_inertia_counter && --frontswap_inertia_counter)
+		return;
+	if (cur_frontswap_pages <= frontswap_hysteresis)
+		tgt_frontswap_pages = 0;
+	else
+		tgt_frontswap_pages = cur_frontswap_pages -
+			(cur_frontswap_pages / frontswap_hysteresis);
+	frontswap_shrink(tgt_frontswap_pages);
+}
+
+static int __init ramster_nofrontswap_selfshrink_setup(char *s)
+{
+	use_frontswap_selfshrink = false;
+	return 1;
+}
+
+__setup("noselfshrink", ramster_nofrontswap_selfshrink_setup);
+
+static void selfshrink_process(struct work_struct *work)
+{
+	if (frontswap_selfshrinking && frontswap_enabled) {
+		frontswap_selfshrink();
+		schedule_delayed_work(&selfshrink_worker,
+			selfshrink_interval * HZ);
+	}
+}
+
+static int ramster_enabled;
+
+static int __init ramster_selfshrink_init(void)
+{
+	frontswap_selfshrinking = ramster_enabled && use_frontswap_selfshrink;
+	if (frontswap_selfshrinking)
+		pr_info("ramster: Initializing frontswap "
+					"selfshrinking driver.\n");
+	else
+		return -ENODEV;
+
+	schedule_delayed_work(&selfshrink_worker, selfshrink_interval * HZ);
+
+	return 0;
+}
+
+subsys_initcall(ramster_selfshrink_init);
+#endif
+
+/*
  * zcache initialization
- * NOTE FOR NOW zcache MUST BE PROVIDED AS A KERNEL BOOT PARAMETER OR
+ * NOTE FOR NOW ramster MUST BE PROVIDED AS A KERNEL BOOT PARAMETER OR
  * NOTHING HAPPENS!
  */
 
-static int zcache_enabled;
+static int ramster_enabled;
 
-static int __init enable_zcache(char *s)
+static int __init enable_ramster(char *s)
 {
-	zcache_enabled = 1;
+	ramster_enabled = 1;
 	return 1;
 }
-__setup("zcache", enable_zcache);
+__setup("ramster", enable_ramster);
 
 /* allow independent dynamic disabling of cleancache and frontswap */
 
@@ -1903,16 +3223,22 @@ static int use_cleancache = 1;
 
 static int __init no_cleancache(char *s)
 {
+	pr_info("INIT no_cleancache called\n");
 	use_cleancache = 0;
 	return 1;
 }
 
-__setup("nocleancache", no_cleancache);
+/*
+ * FIXME: need to guarantee this gets checked before zcache_init is called
+ * What is the correct way to achieve this?
+ */
+early_param("nocleancache", no_cleancache);
 
 static int use_frontswap = 1;
 
 static int __init no_frontswap(char *s)
 {
+	pr_info("INIT no_frontswap called\n");
 	use_frontswap = 0;
 	return 1;
 }
@@ -1925,20 +3251,22 @@ static int __init zcache_init(void)
 
 #ifdef CONFIG_SYSFS
 	ret = sysfs_create_group(mm_kobj, &zcache_attr_group);
+	ret = sysfs_create_group(mm_kobj, &ramster_attr_group);
 	if (ret) {
-		pr_err("zcache: can't create sysfs\n");
+		pr_err("ramster: can't create sysfs\n");
 		goto out;
 	}
 #endif /* CONFIG_SYSFS */
 #if defined(CONFIG_CLEANCACHE) || defined(CONFIG_FRONTSWAP)
-	if (zcache_enabled) {
+	if (ramster_enabled) {
 		unsigned int cpu;
 
+		(void)r2net_register_handlers();
 		tmem_register_hostops(&zcache_hostops);
 		tmem_register_pamops(&zcache_pamops);
 		ret = register_cpu_notifier(&zcache_cpu_notifier_block);
 		if (ret) {
-			pr_err("zcache: can't register cpu notifier\n");
+			pr_err("ramster: can't register cpu notifier\n");
 			goto out;
 		}
 		for_each_online_cpu(cpu) {
@@ -1951,35 +3279,39 @@ static int __init zcache_init(void)
 				sizeof(struct tmem_objnode), 0, 0, NULL);
 	zcache_obj_cache = kmem_cache_create("zcache_obj",
 				sizeof(struct tmem_obj), 0, 0, NULL);
-	ret = zcache_new_client(LOCAL_CLIENT);
-	if (ret) {
-		pr_err("zcache: can't create client\n");
-		goto out;
-	}
+	ramster_flnode_cache = kmem_cache_create("ramster_flnode",
+				sizeof(struct flushlist_node), 0, 0, NULL);
 #endif
 #ifdef CONFIG_CLEANCACHE
-	if (zcache_enabled && use_cleancache) {
+	pr_info("INIT ramster_enabled=%d use_cleancache=%d\n",
+					ramster_enabled, use_cleancache);
+	if (ramster_enabled && use_cleancache) {
 		struct cleancache_ops old_ops;
 
 		zbud_init();
 		register_shrinker(&zcache_shrinker);
 		old_ops = zcache_cleancache_register_ops();
-		pr_info("zcache: cleancache enabled using kernel "
+		pr_info("ramster: cleancache enabled using kernel "
 			"transcendent memory and compression buddies\n");
 		if (old_ops.init_fs != NULL)
-			pr_warning("zcache: cleancache_ops overridden");
+			pr_warning("ramster: cleancache_ops overridden");
 	}
 #endif
 #ifdef CONFIG_FRONTSWAP
-	if (zcache_enabled && use_frontswap) {
+	pr_info("INIT ramster_enabled=%d use_frontswap=%d\n",
+					ramster_enabled, use_frontswap);
+	if (ramster_enabled && use_frontswap) {
 		struct frontswap_ops old_ops;
 
+		zcache_new_client(LOCAL_CLIENT);
 		old_ops = zcache_frontswap_register_ops();
-		pr_info("zcache: frontswap enabled using kernel "
+		pr_info("ramster: frontswap enabled using kernel "
 			"transcendent memory and xvmalloc\n");
 		if (old_ops.init != NULL)
-			pr_warning("zcache: frontswap_ops overridden");
+			pr_warning("ramster: frontswap_ops overridden");
 	}
+	if (ramster_enabled && (use_frontswap || use_cleancache))
+		ramster_remotify_init();
 #endif
 out:
 	return ret;
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH 5/6] staging: ramster: ramster-specific new files
  2012-02-15 15:54 [PATCH V5r1 0/6] staging: ramster: multi-machine memory capacity management Dan Magenheimer
                   ` (3 preceding siblings ...)
  2012-02-15 15:54 ` [PATCH 4/6] staging: ramster: ramster-specific changes to zcache/tmem Dan Magenheimer
@ 2012-02-15 15:54 ` Dan Magenheimer
  2012-02-15 15:54 ` [PATCH 6/6] staging: ramster: enable as staging driver Dan Magenheimer
  5 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-02-15 15:54 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, dan.magenheimer

RAMster implements peer-to-peer transcendent memory, allowing a "cluster"
of kernels to dynamically pool their RAM.

This patch adds new files necessary for ramster support:  The file
ramster.h declares externs and some pampd bitfield manipulation.  The
file zcache.h declares some zcache functions that now must be accessed
from the ramster glue code.  The file r2net.c is the glue between zcache
and the messaging layer, providing routines called from zcache that
initiate messages, and routines that handle messages by calling zcache.
TODO explains future plans for merging.

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/ramster/TODO      |   13 ++
 drivers/staging/ramster/r2net.c   |  401 +++++++++++++++++++++++++++++++++++++
 drivers/staging/ramster/ramster.h |  118 +++++++++++
 drivers/staging/ramster/zcache.h  |   22 ++
 4 files changed, 554 insertions(+), 0 deletions(-)
 create mode 100644 drivers/staging/ramster/TODO
 create mode 100644 drivers/staging/ramster/r2net.c
 create mode 100644 drivers/staging/ramster/ramster.h
 create mode 100644 drivers/staging/ramster/zcache.h

diff --git a/drivers/staging/ramster/TODO b/drivers/staging/ramster/TODO
new file mode 100644
index 0000000..46fcf0c
--- /dev/null
+++ b/drivers/staging/ramster/TODO
@@ -0,0 +1,13 @@
+For this staging driver, RAMster duplicates code from drivers/staging/zcache
+then incorporates changes to the local copy of the code.  For V5, it also
+directly incorporates the soon-to-be-removed drivers/staging/zram/xvmalloc.[ch]
+as all testing has been done with xvmalloc rather than the new zsmalloc.
+Before RAMster can be promoted from staging, the zcache and RAMster drivers
+should be either merged or reorganized to separate out common code.
+
+Until V4, RAMster duplicated code from fs/ocfs2/cluster, but this made
+RAMster incompatible with ocfs2 running in the same kernel and included
+lots of code that could be removed.  As of V5, the ocfs2 code has been
+mined and made RAMster-specific, made to communicate with a userland
+ramster-tools package rather than ocfs2-tools, and can co-exist with ocfs2
+both in the same kernel and in userland on the same machine.
diff --git a/drivers/staging/ramster/r2net.c b/drivers/staging/ramster/r2net.c
new file mode 100644
index 0000000..2ee0220
--- /dev/null
+++ b/drivers/staging/ramster/r2net.c
@@ -0,0 +1,401 @@
+/*
+ * r2net.c
+ *
+ * Copyright (c) 2011, Dan Magenheimer, Oracle Corp.
+ *
+ * Ramster_r2net provides an interface between zcache and r2net.
+ *
+ * FIXME: support more than two nodes
+ */
+
+#include <linux/list.h>
+#include "cluster/tcp.h"
+#include "cluster/nodemanager.h"
+#include "tmem.h"
+#include "zcache.h"
+#include "ramster.h"
+
+#define RAMSTER_TESTING
+
+#define RMSTR_KEY	0x77347734
+
+enum {
+	RMSTR_TMEM_PUT_EPH = 100,
+	RMSTR_TMEM_PUT_PERS,
+	RMSTR_TMEM_ASYNC_GET_REQUEST,
+	RMSTR_TMEM_ASYNC_GET_AND_FREE_REQUEST,
+	RMSTR_TMEM_ASYNC_GET_REPLY,
+	RMSTR_TMEM_FLUSH,
+	RMSTR_TMEM_FLOBJ,
+	RMSTR_TMEM_DESTROY_POOL,
+};
+
+#define RMSTR_R2NET_MAX_LEN \
+		(R2NET_MAX_PAYLOAD_BYTES - sizeof(struct tmem_xhandle))
+
+#include "cluster/tcp_internal.h"
+
+static struct r2nm_node *r2net_target_node;
+static int r2net_target_nodenum;
+
+int r2net_remote_target_node_set(int node_num)
+{
+	int ret = -1;
+
+	r2net_target_node = r2nm_get_node_by_num(node_num);
+	if (r2net_target_node != NULL) {
+		r2net_target_nodenum = node_num;
+		r2nm_node_put(r2net_target_node);
+		ret = 0;
+	}
+	return ret;
+}
+
+/* FIXME following buffer should be per-cpu, protected by preempt_disable */
+static char ramster_async_get_buf[R2NET_MAX_PAYLOAD_BYTES];
+
+static int ramster_remote_async_get_request_handler(struct r2net_msg *msg,
+				u32 len, void *data, void **ret_data)
+{
+	char *pdata;
+	struct tmem_xhandle xh;
+	int found;
+	size_t size = RMSTR_R2NET_MAX_LEN;
+	u16 msgtype = be16_to_cpu(msg->msg_type);
+	bool get_and_free = (msgtype == RMSTR_TMEM_ASYNC_GET_AND_FREE_REQUEST);
+	unsigned long flags;
+
+	xh = *(struct tmem_xhandle *)msg->buf;
+	if (xh.xh_data_size > RMSTR_R2NET_MAX_LEN)
+		BUG();
+	pdata = ramster_async_get_buf;
+	*(struct tmem_xhandle *)pdata = xh;
+	pdata += sizeof(struct tmem_xhandle);
+	local_irq_save(flags);
+	found = zcache_get(xh.client_id, xh.pool_id, &xh.oid, xh.index,
+				pdata, &size, 1, get_and_free ? 1 : -1);
+	local_irq_restore(flags);
+	if (found < 0) {
+		/* a zero size indicates the get failed */
+		size = 0;
+	}
+	if (size > RMSTR_R2NET_MAX_LEN)
+		BUG();
+	*ret_data = pdata - sizeof(struct tmem_xhandle);
+	/* now make caller (r2net_process_message) handle specially */
+	r2net_force_data_magic(msg, RMSTR_TMEM_ASYNC_GET_REPLY, RMSTR_KEY);
+	return size + sizeof(struct tmem_xhandle);
+}
+
+static int ramster_remote_async_get_reply_handler(struct r2net_msg *msg,
+				u32 len, void *data, void **ret_data)
+{
+	char *in = (char *)msg->buf;
+	int datalen = len - sizeof(struct r2net_msg);
+	int ret = -1;
+	struct tmem_xhandle *xh = (struct tmem_xhandle *)in;
+
+	in += sizeof(struct tmem_xhandle);
+	datalen -= sizeof(struct tmem_xhandle);
+	BUG_ON(datalen < 0 || datalen > PAGE_SIZE);
+	ret = zcache_localify(xh->pool_id, &xh->oid, xh->index,
+				in, datalen, xh->extra);
+#ifdef RAMSTER_TESTING
+	if (ret == -EEXIST)
+		pr_err("TESTING ArrgREP, aborted overwrite on racy put\n");
+#endif
+	return ret;
+}
+
+int ramster_remote_put_handler(struct r2net_msg *msg,
+				u32 len, void *data, void **ret_data)
+{
+	struct tmem_xhandle *xh;
+	char *p = (char *)msg->buf;
+	int datalen = len - sizeof(struct r2net_msg) -
+				sizeof(struct tmem_xhandle);
+	u16 msgtype = be16_to_cpu(msg->msg_type);
+	bool ephemeral = (msgtype == RMSTR_TMEM_PUT_EPH);
+	unsigned long flags;
+	int ret;
+
+	xh = (struct tmem_xhandle *)p;
+	p += sizeof(struct tmem_xhandle);
+	zcache_autocreate_pool(xh->client_id, xh->pool_id, ephemeral);
+	local_irq_save(flags);
+	ret = zcache_put(xh->client_id, xh->pool_id, &xh->oid, xh->index,
+				p, datalen, 1, ephemeral ? 1 : -1);
+	local_irq_restore(flags);
+	return ret;
+}
+
+int ramster_remote_flush_handler(struct r2net_msg *msg,
+				u32 len, void *data, void **ret_data)
+{
+	struct tmem_xhandle *xh;
+	char *p = (char *)msg->buf;
+
+	xh = (struct tmem_xhandle *)p;
+	p += sizeof(struct tmem_xhandle);
+	(void)zcache_flush(xh->client_id, xh->pool_id, &xh->oid, xh->index);
+	return 0;
+}
+
+int ramster_remote_flobj_handler(struct r2net_msg *msg,
+				u32 len, void *data, void **ret_data)
+{
+	struct tmem_xhandle *xh;
+	char *p = (char *)msg->buf;
+
+	xh = (struct tmem_xhandle *)p;
+	p += sizeof(struct tmem_xhandle);
+	(void)zcache_flush_object(xh->client_id, xh->pool_id, &xh->oid);
+	return 0;
+}
+
+int ramster_remote_async_get(struct tmem_xhandle *xh, bool free, int remotenode,
+				size_t expect_size, uint8_t expect_cksum,
+				void *extra)
+{
+	int ret = -1, status;
+	struct r2nm_node *node = NULL;
+	struct kvec vec[1];
+	size_t veclen = 1;
+	u32 msg_type;
+
+	node = r2nm_get_node_by_num(remotenode);
+	if (node == NULL)
+		goto out;
+	xh->client_id = r2nm_this_node(); /* which node is getting */
+	xh->xh_data_cksum = expect_cksum;
+	xh->xh_data_size = expect_size;
+	xh->extra = extra;
+	vec[0].iov_len = sizeof(*xh);
+	vec[0].iov_base = xh;
+	if (free)
+		msg_type = RMSTR_TMEM_ASYNC_GET_AND_FREE_REQUEST;
+	else
+		msg_type = RMSTR_TMEM_ASYNC_GET_REQUEST;
+	ret = r2net_send_message_vec(msg_type, RMSTR_KEY,
+					vec, veclen, remotenode, &status);
+	r2nm_node_put(node);
+	if (ret < 0) {
+		/* FIXME handle bad message possibilities here? */
+		pr_err("UNTESTED ret<0 in ramster_remote_async_get\n");
+	}
+	ret = status;
+out:
+	return ret;
+}
+
+#ifdef RAMSTER_TESTING
+/* leave me here to see if it catches a weird crash */
+static void ramster_check_irq_counts(void)
+{
+	static int last_hardirq_cnt, last_softirq_cnt, last_preempt_cnt;
+	int cur_hardirq_cnt, cur_softirq_cnt, cur_preempt_cnt;
+
+	cur_hardirq_cnt = hardirq_count() >> HARDIRQ_SHIFT;
+	if (cur_hardirq_cnt > last_hardirq_cnt) {
+		last_hardirq_cnt = cur_hardirq_cnt;
+		if (!(last_hardirq_cnt&(last_hardirq_cnt-1)))
+			pr_err("RAMSTER TESTING RRP hardirq_count=%d\n",
+				last_hardirq_cnt);
+	}
+	cur_softirq_cnt = softirq_count() >> SOFTIRQ_SHIFT;
+	if (cur_softirq_cnt > last_softirq_cnt) {
+		last_softirq_cnt = cur_softirq_cnt;
+		if (!(last_softirq_cnt&(last_softirq_cnt-1)))
+			pr_err("RAMSTER TESTING RRP softirq_count=%d\n",
+				last_softirq_cnt);
+	}
+	cur_preempt_cnt = preempt_count() & PREEMPT_MASK;
+	if (cur_preempt_cnt > last_preempt_cnt) {
+		last_preempt_cnt = cur_preempt_cnt;
+		if (!(last_preempt_cnt&(last_preempt_cnt-1)))
+			pr_err("RAMSTER TESTING RRP preempt_count=%d\n",
+				last_preempt_cnt);
+	}
+}
+#endif
+
+int ramster_remote_put(struct tmem_xhandle *xh, char *data, size_t size,
+				bool ephemeral, int *remotenode)
+{
+	int nodenum, ret = -1, status;
+	struct r2nm_node *node = NULL;
+	struct kvec vec[2];
+	size_t veclen = 2;
+	u32 msg_type;
+#ifdef RAMSTER_TESTING
+	struct r2net_node *nn;
+#endif
+
+	BUG_ON(size > RMSTR_R2NET_MAX_LEN);
+	xh->client_id = r2nm_this_node(); /* which node is putting */
+	vec[0].iov_len = sizeof(*xh);
+	vec[0].iov_base = xh;
+	vec[1].iov_len = size;
+	vec[1].iov_base = data;
+	node = r2net_target_node;
+	if (!node)
+		goto out;
+
+	nodenum = r2net_target_nodenum;
+
+	r2nm_node_get(node);
+
+#ifdef RAMSTER_TESTING
+	nn = r2net_nn_from_num(nodenum);
+	WARN_ON_ONCE(nn->nn_persistent_error || !nn->nn_sc_valid);
+#endif
+
+	if (ephemeral)
+		msg_type = RMSTR_TMEM_PUT_EPH;
+	else
+		msg_type = RMSTR_TMEM_PUT_PERS;
+#ifdef RAMSTER_TESTING
+	/* leave me here to see if it catches a weird crash */
+	ramster_check_irq_counts();
+#endif
+
+	ret = r2net_send_message_vec(msg_type, RMSTR_KEY, vec, veclen,
+						nodenum, &status);
+#ifdef RAMSTER_TESTING
+	if (ret != 0) {
+		static unsigned long cnt;
+		cnt++;
+		if (!(cnt&(cnt-1)))
+			pr_err("ramster_remote_put: message failed, "
+				"ret=%d, cnt=%lu\n", ret, cnt);
+		ret = -1;
+	}
+#endif
+	if (ret < 0)
+		ret = -1;
+	else {
+		ret = status;
+		*remotenode = nodenum;
+	}
+
+	r2nm_node_put(node);
+out:
+	return ret;
+}
+
+int ramster_remote_flush(struct tmem_xhandle *xh, int remotenode)
+{
+	int ret = -1, status;
+	struct r2nm_node *node = NULL;
+	struct kvec vec[1];
+	size_t veclen = 1;
+
+	node = r2nm_get_node_by_num(remotenode);
+	BUG_ON(node == NULL);
+	xh->client_id = r2nm_this_node(); /* which node is flushing */
+	vec[0].iov_len = sizeof(*xh);
+	vec[0].iov_base = xh;
+	BUG_ON(irqs_disabled());
+	BUG_ON(in_softirq());
+	ret = r2net_send_message_vec(RMSTR_TMEM_FLUSH, RMSTR_KEY,
+					vec, veclen, remotenode, &status);
+	r2nm_node_put(node);
+	return ret;
+}
+
+int ramster_remote_flush_object(struct tmem_xhandle *xh, int remotenode)
+{
+	int ret = -1, status;
+	struct r2nm_node *node = NULL;
+	struct kvec vec[1];
+	size_t veclen = 1;
+
+	node = r2nm_get_node_by_num(remotenode);
+	BUG_ON(node == NULL);
+	xh->client_id = r2nm_this_node(); /* which node is flobjing */
+	vec[0].iov_len = sizeof(*xh);
+	vec[0].iov_base = xh;
+	ret = r2net_send_message_vec(RMSTR_TMEM_FLOBJ, RMSTR_KEY,
+					vec, veclen, remotenode, &status);
+	r2nm_node_put(node);
+	return ret;
+}
+
+/*
+ * Handler registration
+ */
+
+static LIST_HEAD(r2net_unreg_list);
+
+static void r2net_unregister_handlers(void)
+{
+	r2net_unregister_handler_list(&r2net_unreg_list);
+}
+
+int r2net_register_handlers(void)
+{
+	int status;
+
+	status = r2net_register_handler(RMSTR_TMEM_PUT_EPH, RMSTR_KEY,
+				RMSTR_R2NET_MAX_LEN,
+				ramster_remote_put_handler,
+				NULL, NULL, &r2net_unreg_list);
+	if (status)
+		goto bail;
+
+	status = r2net_register_handler(RMSTR_TMEM_PUT_PERS, RMSTR_KEY,
+				RMSTR_R2NET_MAX_LEN,
+				ramster_remote_put_handler,
+				NULL, NULL, &r2net_unreg_list);
+	if (status)
+		goto bail;
+
+	status = r2net_register_handler(RMSTR_TMEM_ASYNC_GET_REQUEST, RMSTR_KEY,
+				RMSTR_R2NET_MAX_LEN,
+				ramster_remote_async_get_request_handler,
+				NULL, NULL,
+				&r2net_unreg_list);
+	if (status)
+		goto bail;
+
+	status = r2net_register_handler(RMSTR_TMEM_ASYNC_GET_AND_FREE_REQUEST,
+				RMSTR_KEY, RMSTR_R2NET_MAX_LEN,
+				ramster_remote_async_get_request_handler,
+				NULL, NULL,
+				&r2net_unreg_list);
+	if (status)
+		goto bail;
+
+	status = r2net_register_handler(RMSTR_TMEM_ASYNC_GET_REPLY, RMSTR_KEY,
+				RMSTR_R2NET_MAX_LEN,
+				ramster_remote_async_get_reply_handler,
+				NULL, NULL,
+				&r2net_unreg_list);
+	if (status)
+		goto bail;
+
+	status = r2net_register_handler(RMSTR_TMEM_FLUSH, RMSTR_KEY,
+				RMSTR_R2NET_MAX_LEN,
+				ramster_remote_flush_handler,
+				NULL, NULL,
+				&r2net_unreg_list);
+	if (status)
+		goto bail;
+
+	status = r2net_register_handler(RMSTR_TMEM_FLOBJ, RMSTR_KEY,
+				RMSTR_R2NET_MAX_LEN,
+				ramster_remote_flobj_handler,
+				NULL, NULL,
+				&r2net_unreg_list);
+	if (status)
+		goto bail;
+
+	pr_info("ramster: r2net handlers registered\n");
+
+bail:
+	if (status) {
+		r2net_unregister_handlers();
+		pr_err("ramster: couldn't register r2net handlers\n");
+	}
+	return status;
+}
diff --git a/drivers/staging/ramster/ramster.h b/drivers/staging/ramster/ramster.h
new file mode 100644
index 0000000..0c9455e
--- /dev/null
+++ b/drivers/staging/ramster/ramster.h
@@ -0,0 +1,118 @@
+/*
+ * ramster.h
+ *
+ * Peer-to-peer transcendent memory
+ *
+ * Copyright (c) 2009-2012, Dan Magenheimer, Oracle Corp.
+ */
+
+#ifndef _RAMSTER_H_
+#define _RAMSTER_H_
+
+/*
+ * format of remote pampd:
+ *   bit 0 == intransit
+ *   bit 1 == is_remote... if this bit is set, then
+ *   bit 2-9 == remotenode
+ *   bit 10-22 == size
+ *   bit 23-30 == cksum
+ */
+#define FAKE_PAMPD_INTRANSIT_BITS	1
+#define FAKE_PAMPD_ISREMOTE_BITS	1
+#define FAKE_PAMPD_REMOTENODE_BITS	8
+#define FAKE_PAMPD_REMOTESIZE_BITS	13
+#define FAKE_PAMPD_CHECKSUM_BITS	8
+
+#define FAKE_PAMPD_INTRANSIT_SHIFT	0
+#define FAKE_PAMPD_ISREMOTE_SHIFT	(FAKE_PAMPD_INTRANSIT_SHIFT + \
+					 FAKE_PAMPD_INTRANSIT_BITS)
+#define FAKE_PAMPD_REMOTENODE_SHIFT	(FAKE_PAMPD_ISREMOTE_SHIFT + \
+					 FAKE_PAMPD_ISREMOTE_BITS)
+#define FAKE_PAMPD_REMOTESIZE_SHIFT	(FAKE_PAMPD_REMOTENODE_SHIFT + \
+					 FAKE_PAMPD_REMOTENODE_BITS)
+#define FAKE_PAMPD_CHECKSUM_SHIFT	(FAKE_PAMPD_REMOTESIZE_SHIFT + \
+					 FAKE_PAMPD_REMOTESIZE_BITS)
+
+#define FAKE_PAMPD_MASK(x)		((1UL << (x)) - 1)
+
+static inline void *pampd_make_remote(int remotenode, size_t size,
+					unsigned char cksum)
+{
+	unsigned long fake_pampd = 0;
+	fake_pampd |= 1UL << FAKE_PAMPD_ISREMOTE_SHIFT;
+	fake_pampd |= ((unsigned long)remotenode &
+			FAKE_PAMPD_MASK(FAKE_PAMPD_REMOTENODE_BITS)) <<
+				FAKE_PAMPD_REMOTENODE_SHIFT;
+	fake_pampd |= ((unsigned long)size &
+			FAKE_PAMPD_MASK(FAKE_PAMPD_REMOTESIZE_BITS)) <<
+				FAKE_PAMPD_REMOTESIZE_SHIFT;
+	fake_pampd |= ((unsigned long)cksum &
+			FAKE_PAMPD_MASK(FAKE_PAMPD_CHECKSUM_BITS)) <<
+				FAKE_PAMPD_CHECKSUM_SHIFT;
+	return (void *)fake_pampd;
+}
+
+static inline unsigned int pampd_remote_node(void *pampd)
+{
+	unsigned long fake_pampd = (unsigned long)pampd;
+	return (fake_pampd >> FAKE_PAMPD_REMOTENODE_SHIFT) &
+		FAKE_PAMPD_MASK(FAKE_PAMPD_REMOTENODE_BITS);
+}
+
+static inline unsigned int pampd_remote_size(void *pampd)
+{
+	unsigned long fake_pampd = (unsigned long)pampd;
+	return (fake_pampd >> FAKE_PAMPD_REMOTESIZE_SHIFT) &
+		FAKE_PAMPD_MASK(FAKE_PAMPD_REMOTESIZE_BITS);
+}
+
+static inline unsigned char pampd_remote_cksum(void *pampd)
+{
+	unsigned long fake_pampd = (unsigned long)pampd;
+	return (fake_pampd >> FAKE_PAMPD_CHECKSUM_SHIFT) &
+		FAKE_PAMPD_MASK(FAKE_PAMPD_CHECKSUM_BITS);
+}
+
+static inline bool pampd_is_remote(void *pampd)
+{
+	unsigned long fake_pampd = (unsigned long)pampd;
+	return (fake_pampd >> FAKE_PAMPD_ISREMOTE_SHIFT) &
+		FAKE_PAMPD_MASK(FAKE_PAMPD_ISREMOTE_BITS);
+}
+
+static inline bool pampd_is_intransit(void *pampd)
+{
+	unsigned long fake_pampd = (unsigned long)pampd;
+	return (fake_pampd >> FAKE_PAMPD_INTRANSIT_SHIFT) &
+		FAKE_PAMPD_MASK(FAKE_PAMPD_INTRANSIT_BITS);
+}
+
+/* note that it is a BUG for intransit to be set without isremote also set */
+static inline void *pampd_mark_intransit(void *pampd)
+{
+	unsigned long fake_pampd = (unsigned long)pampd;
+
+	fake_pampd |= 1UL << FAKE_PAMPD_ISREMOTE_SHIFT;
+	fake_pampd |= 1UL << FAKE_PAMPD_INTRANSIT_SHIFT;
+	return (void *)fake_pampd;
+}
+
+static inline void *pampd_mask_intransit_and_remote(void *marked_pampd)
+{
+	unsigned long pampd = (unsigned long)marked_pampd;
+
+	pampd &= ~(1UL << FAKE_PAMPD_INTRANSIT_SHIFT);
+	pampd &= ~(1UL << FAKE_PAMPD_ISREMOTE_SHIFT);
+	return (void *)pampd;
+}
+
+extern int ramster_remote_async_get(struct tmem_xhandle *,
+				bool, int, size_t, uint8_t, void *extra);
+extern int ramster_remote_put(struct tmem_xhandle *, char *, size_t,
+				bool, int *);
+extern int ramster_remote_flush(struct tmem_xhandle *, int);
+extern int ramster_remote_flush_object(struct tmem_xhandle *, int);
+extern int r2net_register_handlers(void);
+extern int r2net_remote_target_node_set(int);
+
+#endif /* _TMEM_H */
diff --git a/drivers/staging/ramster/zcache.h b/drivers/staging/ramster/zcache.h
new file mode 100644
index 0000000..250b121
--- /dev/null
+++ b/drivers/staging/ramster/zcache.h
@@ -0,0 +1,22 @@
+/*
+ * zcache.h
+ *
+ * External zcache functions
+ *
+ * Copyright (c) 2009-2012, Dan Magenheimer, Oracle Corp.
+ */
+
+#ifndef _ZCACHE_H_
+#define _ZCACHE_H_
+
+extern int zcache_put(int, int, struct tmem_oid *, uint32_t,
+			char *, size_t, bool, int);
+extern int zcache_autocreate_pool(int, int, bool);
+extern int zcache_get(int, int, struct tmem_oid *, uint32_t,
+			char *, size_t *, bool, int);
+extern int zcache_flush(int, int, struct tmem_oid *, uint32_t);
+extern int zcache_flush_object(int, int, struct tmem_oid *);
+extern int zcache_localify(int, struct tmem_oid *, uint32_t,
+			char *, size_t, void *);
+
+#endif /* _ZCACHE_H */
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH 6/6] staging: ramster: enable as staging driver
  2012-02-15 15:54 [PATCH V5r1 0/6] staging: ramster: multi-machine memory capacity management Dan Magenheimer
                   ` (4 preceding siblings ...)
  2012-02-15 15:54 ` [PATCH 5/6] staging: ramster: ramster-specific new files Dan Magenheimer
@ 2012-02-15 15:54 ` Dan Magenheimer
  5 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-02-15 15:54 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, dan.magenheimer

RAMster implements peer-to-peer transcendent memory, allowing a "cluster"
of kernels to dynamically pool their RAM.

Enable build of ramster as a staging driver

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/Kconfig  |    2 ++
 drivers/staging/Makefile |    1 +
 2 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/drivers/staging/Kconfig b/drivers/staging/Kconfig
index b5a311b..c37dbc5 100644
--- a/drivers/staging/Kconfig
+++ b/drivers/staging/Kconfig
@@ -130,4 +130,6 @@ source "drivers/staging/android/Kconfig"
 
 source "drivers/staging/telephony/Kconfig"
 
+source "drivers/staging/ramster/Kconfig"
+
 endif # STAGING
diff --git a/drivers/staging/Makefile b/drivers/staging/Makefile
index 4908e46..035a035 100644
--- a/drivers/staging/Makefile
+++ b/drivers/staging/Makefile
@@ -55,3 +55,4 @@ obj-$(CONFIG_MFD_NVEC)		+= nvec/
 obj-$(CONFIG_DRM_OMAP)		+= omapdrm/
 obj-$(CONFIG_ANDROID)		+= android/
 obj-$(CONFIG_PHONE)		+= telephony/
+obj-$(CONFIG_RAMSTER)		+= ramster/
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH 6/6] staging: ramster: enable as staging driver
@ 2012-02-14 23:36 Dan Magenheimer
  0 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-02-14 23:36 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, ngupta, konrad.wilk,
	kurt.hackel, sjenning, chris.mason, dan.magenheimer

RAMster implements peer-to-peer transcendent memory, allowing a "cluster"
of kernels to dynamically pool their RAM.

Enable build of ramster as a staging driver

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/Kconfig  |    2 ++
 drivers/staging/Makefile |    1 +
 2 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/drivers/staging/Kconfig b/drivers/staging/Kconfig
index 25cdff3..5f7b5eb 100644
--- a/drivers/staging/Kconfig
+++ b/drivers/staging/Kconfig
@@ -132,4 +132,6 @@ source "drivers/staging/nvec/Kconfig"
 
 source "drivers/staging/media/Kconfig"
 
+source "drivers/staging/ramster/Kconfig"
+
 endif # STAGING
diff --git a/drivers/staging/Makefile b/drivers/staging/Makefile
index a25f3f2..3a7b2f6 100644
--- a/drivers/staging/Makefile
+++ b/drivers/staging/Makefile
@@ -57,3 +57,4 @@ obj-$(CONFIG_TOUCHSCREEN_SYNAPTICS_I2C_RMI4)	+= ste_rmi4/
 obj-$(CONFIG_DRM_PSB)		+= gma500/
 obj-$(CONFIG_INTEL_MEI)		+= mei/
 obj-$(CONFIG_MFD_NVEC)		+= nvec/
+obj-$(CONFIG_RAMSTER)		+= ramster/
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH 6/6] staging: ramster: enable as staging driver
@ 2012-01-30 22:39 Dan Magenheimer
  0 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2012-01-30 22:39 UTC (permalink / raw)
  To: devel, linux-kernel, gregkh, linux-mm, ngupta, konrad.wilk,
	kurt.hackel, sjenning, chris.mason, dan.magenheimer

Enable build of ramster as a staging driver

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/Kconfig  |    2 ++
 drivers/staging/Makefile |    1 +
 2 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/drivers/staging/Kconfig b/drivers/staging/Kconfig
index 25cdff3..5f7b5eb 100644
--- a/drivers/staging/Kconfig
+++ b/drivers/staging/Kconfig
@@ -132,4 +132,6 @@ source "drivers/staging/nvec/Kconfig"
 
 source "drivers/staging/media/Kconfig"
 
+source "drivers/staging/ramster/Kconfig"
+
 endif # STAGING
diff --git a/drivers/staging/Makefile b/drivers/staging/Makefile
index a25f3f2..3a7b2f6 100644
--- a/drivers/staging/Makefile
+++ b/drivers/staging/Makefile
@@ -57,3 +57,4 @@ obj-$(CONFIG_TOUCHSCREEN_SYNAPTICS_I2C_RMI4)	+= ste_rmi4/
 obj-$(CONFIG_DRM_PSB)		+= gma500/
 obj-$(CONFIG_INTEL_MEI)		+= mei/
 obj-$(CONFIG_MFD_NVEC)		+= nvec/
+obj-$(CONFIG_RAMSTER)		+= ramster/
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH 6/6] staging: ramster: enable as staging driver
@ 2011-12-27 18:50 Dan Magenheimer
  0 siblings, 0 replies; 10+ messages in thread
From: Dan Magenheimer @ 2011-12-27 18:50 UTC (permalink / raw)
  To: devel, linux-kernel, linux-mm, ngupta, konrad.wilk, kurt.hackel,
	sjenning, chris.mason, dan.magenheimer

Enable build of ramster as a staging driver

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
---
 drivers/staging/Kconfig  |    2 ++
 drivers/staging/Makefile |    1 +
 2 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/drivers/staging/Kconfig b/drivers/staging/Kconfig
index 06c9081..28819c3 100644
--- a/drivers/staging/Kconfig
+++ b/drivers/staging/Kconfig
@@ -150,4 +150,6 @@ source "drivers/staging/mei/Kconfig"
 
 source "drivers/staging/nvec/Kconfig"
 
+source "drivers/staging/ramster/Kconfig"
+
 endif # STAGING
diff --git a/drivers/staging/Makefile b/drivers/staging/Makefile
index f3c5e33..f1e6ec0 100644
--- a/drivers/staging/Makefile
+++ b/drivers/staging/Makefile
@@ -66,3 +66,4 @@ obj-$(CONFIG_TOUCHSCREEN_SYNAPTICS_I2C_RMI4)	+= ste_rmi4/
 obj-$(CONFIG_DRM_PSB)		+= gma500/
 obj-$(CONFIG_INTEL_MEI)		+= mei/
 obj-$(CONFIG_MFD_NVEC)		+= nvec/
+obj-$(CONFIG_RAMSTER)		+= ramster/
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2012-02-15 15:55 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-02-15 15:54 [PATCH V5r1 0/6] staging: ramster: multi-machine memory capacity management Dan Magenheimer
2012-02-15 15:54 ` [PATCH 1/6] staging: ramster: cluster/messaging foundation Dan Magenheimer
2012-02-15 15:54 ` [PATCH 2/6] staging: ramster: local compression + tmem Dan Magenheimer
2012-02-15 15:54 ` [PATCH 3/6] staging: ramster: xvmalloc allocation files Dan Magenheimer
2012-02-15 15:54 ` [PATCH 4/6] staging: ramster: ramster-specific changes to zcache/tmem Dan Magenheimer
2012-02-15 15:54 ` [PATCH 5/6] staging: ramster: ramster-specific new files Dan Magenheimer
2012-02-15 15:54 ` [PATCH 6/6] staging: ramster: enable as staging driver Dan Magenheimer
  -- strict thread matches above, loose matches on Subject: below --
2012-02-14 23:36 Dan Magenheimer
2012-01-30 22:39 Dan Magenheimer
2011-12-27 18:50 Dan Magenheimer

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).