From 2b190ce9bf923fb1fd8ccff18e9bab72da40da06 Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Tue, 14 Dec 2010 14:14:27 -0800
Subject: ocfs2/cluster: Pin the remote node item in configfs

o2net pins the node item of the remote node in configfs before initiating
the connection. It is unpinned on disconnect. This is to prevent the node
item from being unlinked while it is still in use.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/tcp.c | 9 +++++++++
 1 file changed, 9 insertions(+)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 9aa426e42123..92de96cd247d 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -355,6 +355,7 @@ static void sc_kref_release(struct kref *kref)
 		sc->sc_sock = NULL;
 	}
 
+	o2nm_undepend_item(&sc->sc_node->nd_item);
 	o2nm_node_put(sc->sc_node);
 	sc->sc_node = NULL;
 
@@ -376,6 +377,7 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node)
 {
 	struct o2net_sock_container *sc, *ret = NULL;
 	struct page *page = NULL;
+	int status = 0;
 
 	page = alloc_page(GFP_NOFS);
 	sc = kzalloc(sizeof(*sc), GFP_NOFS);
@@ -386,6 +388,13 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node)
 	o2nm_node_get(node);
 	sc->sc_node = node;
 
+	/* pin the node item of the remote node */
+	status = o2nm_depend_item(&node->nd_item);
+	if (status) {
+		mlog_errno(status);
+		o2nm_node_put(node);
+		goto out;
+	}
 	INIT_WORK(&sc->sc_connect_work, o2net_sc_connect_completed);
 	INIT_WORK(&sc->sc_rx_work, o2net_rx_until_empty);
 	INIT_WORK(&sc->sc_shutdown_work, o2net_shutdown_sc);
-- 
cgit v1.2.3


From ffee223a9af4c5124beb56fa5c84132949923d23 Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Tue, 14 Dec 2010 14:14:28 -0800
Subject: ocfs2/cluster: Remove dropped region from o2hb quorum region bitmap

Patch removes a dropped region from the quorum region bitmap maintained by o2hb.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/heartbeat.c | 1 +
 1 file changed, 1 insertion(+)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index 9f26ac9be2a4..9f21dd785364 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -2009,6 +2009,7 @@ static void o2hb_heartbeat_group_drop_item(struct config_group *group,
 	if (o2hb_global_heartbeat_active()) {
 		clear_bit(reg->hr_region_num, o2hb_region_bitmap);
 		clear_bit(reg->hr_region_num, o2hb_live_region_bitmap);
+		clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
 	}
 	hb_task = reg->hr_task;
 	reg->hr_task = NULL;
-- 
cgit v1.2.3


From 58a3158a5d17ddf4894db9e8ccaf92093ff8e42e Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Tue, 14 Dec 2010 14:14:29 -0800
Subject: ocfs2/cluster: Pin/unpin o2hb regions

This patch adds support for pinning o2hb regions in configfs. Pinning disallows
a region to be cleanly stopped as long as it has an active dependent user
(read o2dlm).

In local heartbeat mode, the region uuid matching the domain name is pinned as
long as the o2dlm domain is active.

In global heartbeat mode, all regions are pinned as long as there is atleast
one dependent user and the region count is 3 or less. All regions are unpinned
if the number of dependent users is zero or region count is greater than 3.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/heartbeat.c | 216 +++++++++++++++++++++++++++++++++++--------
 1 file changed, 178 insertions(+), 38 deletions(-)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index 9f21dd785364..ad2e41d6879d 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -132,6 +132,33 @@ char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = {
 unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
 unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL;
 
+/*
+ * o2hb_dependent_users tracks the number of registered callbacks that depend
+ * on heartbeat. o2net and o2dlm are two entities that register this callback.
+ * However only o2dlm depends on the heartbeat. It does not want the heartbeat
+ * to stop while a dlm domain is still active.
+ */
+unsigned int o2hb_dependent_users;
+
+/*
+ * In global heartbeat mode, all regions are pinned if there are one or more
+ * dependent users and the quorum region count is <= O2HB_PIN_CUT_OFF. All
+ * regions are unpinned if the region count exceeds the cut off or the number
+ * of dependent users falls to zero.
+ */
+#define O2HB_PIN_CUT_OFF		3
+
+/*
+ * In local heartbeat mode, we assume the dlm domain name to be the same as
+ * region uuid. This is true for domains created for the file system but not
+ * necessarily true for userdlm domains. This is a known limitation.
+ *
+ * In global heartbeat mode, we pin/unpin all o2hb regions. This solution
+ * works for both file system and userdlm domains.
+ */
+static int o2hb_region_pin(const char *region_uuid);
+static void o2hb_region_unpin(const char *region_uuid);
+
 /* Only sets a new threshold if there are no active regions.
  *
  * No locking or otherwise interesting code is required for reading
@@ -186,7 +213,9 @@ struct o2hb_region {
 	struct config_item	hr_item;
 
 	struct list_head	hr_all_item;
-	unsigned		hr_unclean_stop:1;
+	unsigned		hr_unclean_stop:1,
+				hr_item_pinned:1,
+				hr_item_dropped:1;
 
 	/* protected by the hr_callback_sem */
 	struct task_struct 	*hr_task;
@@ -702,6 +731,14 @@ static void o2hb_set_quorum_device(struct o2hb_region *reg,
 	       config_item_name(&reg->hr_item));
 
 	set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
+
+	/*
+	 * If global heartbeat active, unpin all regions if the
+	 * region count > CUT_OFF
+	 */
+	if (o2hb_pop_count(&o2hb_quorum_region_bitmap,
+			   O2NM_MAX_REGIONS) > O2HB_PIN_CUT_OFF)
+		o2hb_region_unpin(NULL);
 }
 
 static int o2hb_check_slot(struct o2hb_region *reg,
@@ -1316,6 +1353,8 @@ int o2hb_init(void)
 	memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap));
 	memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap));
 
+	o2hb_dependent_users = 0;
+
 	return o2hb_debug_init();
 }
 
@@ -2003,16 +2042,20 @@ static void o2hb_heartbeat_group_drop_item(struct config_group *group,
 {
 	struct task_struct *hb_task;
 	struct o2hb_region *reg = to_o2hb_region(item);
+	int quorum_region = 0;
 
 	/* stop the thread when the user removes the region dir */
 	spin_lock(&o2hb_live_lock);
 	if (o2hb_global_heartbeat_active()) {
 		clear_bit(reg->hr_region_num, o2hb_region_bitmap);
 		clear_bit(reg->hr_region_num, o2hb_live_region_bitmap);
+		if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
+			quorum_region = 1;
 		clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
 	}
 	hb_task = reg->hr_task;
 	reg->hr_task = NULL;
+	reg->hr_item_dropped = 1;
 	spin_unlock(&o2hb_live_lock);
 
 	if (hb_task)
@@ -2030,7 +2073,27 @@ static void o2hb_heartbeat_group_drop_item(struct config_group *group,
 	if (o2hb_global_heartbeat_active())
 		printk(KERN_NOTICE "o2hb: Heartbeat stopped on region %s\n",
 		       config_item_name(&reg->hr_item));
+
 	config_item_put(item);
+
+	if (!o2hb_global_heartbeat_active() || !quorum_region)
+		return;
+
+	/*
+	 * If global heartbeat active and there are dependent users,
+	 * pin all regions if quorum region count <= CUT_OFF
+	 */
+	spin_lock(&o2hb_live_lock);
+
+	if (!o2hb_dependent_users)
+		goto unlock;
+
+	if (o2hb_pop_count(&o2hb_quorum_region_bitmap,
+			   O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
+		o2hb_region_pin(NULL);
+
+unlock:
+	spin_unlock(&o2hb_live_lock);
 }
 
 struct o2hb_heartbeat_group_attribute {
@@ -2216,63 +2279,138 @@ void o2hb_setup_callback(struct o2hb_callback_func *hc,
 }
 EXPORT_SYMBOL_GPL(o2hb_setup_callback);
 
-static struct o2hb_region *o2hb_find_region(const char *region_uuid)
+/*
+ * In local heartbeat mode, region_uuid passed matches the dlm domain name.
+ * In global heartbeat mode, region_uuid passed is NULL.
+ *
+ * In local, we only pin the matching region. In global we pin all the active
+ * regions.
+ */
+static int o2hb_region_pin(const char *region_uuid)
 {
-	struct o2hb_region *p, *reg = NULL;
+	int ret = 0, found = 0;
+	struct o2hb_region *reg;
+	char *uuid;
 
 	assert_spin_locked(&o2hb_live_lock);
 
-	list_for_each_entry(p, &o2hb_all_regions, hr_all_item) {
-		if (!strcmp(region_uuid, config_item_name(&p->hr_item))) {
-			reg = p;
-			break;
+	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
+		uuid = config_item_name(&reg->hr_item);
+
+		/* local heartbeat */
+		if (region_uuid) {
+			if (strcmp(region_uuid, uuid))
+				continue;
+			found = 1;
 		}
+
+		if (reg->hr_item_pinned || reg->hr_item_dropped)
+			goto skip_pin;
+
+		/* Ignore ENOENT only for local hb (userdlm domain) */
+		ret = o2nm_depend_item(&reg->hr_item);
+		if (!ret) {
+			mlog(ML_CLUSTER, "Pin region %s\n", uuid);
+			reg->hr_item_pinned = 1;
+		} else {
+			if (ret == -ENOENT && found)
+				ret = 0;
+			else {
+				mlog(ML_ERROR, "Pin region %s fails with %d\n",
+				     uuid, ret);
+				break;
+			}
+		}
+skip_pin:
+		if (found)
+			break;
 	}
 
-	return reg;
+	return ret;
 }
 
-static int o2hb_region_get(const char *region_uuid)
+/*
+ * In local heartbeat mode, region_uuid passed matches the dlm domain name.
+ * In global heartbeat mode, region_uuid passed is NULL.
+ *
+ * In local, we only unpin the matching region. In global we unpin all the
+ * active regions.
+ */
+static void o2hb_region_unpin(const char *region_uuid)
 {
-	int ret = 0;
 	struct o2hb_region *reg;
+	char *uuid;
+	int found = 0;
 
-	spin_lock(&o2hb_live_lock);
+	assert_spin_locked(&o2hb_live_lock);
 
-	reg = o2hb_find_region(region_uuid);
-	if (!reg)
-		ret = -ENOENT;
-	spin_unlock(&o2hb_live_lock);
+	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
+		uuid = config_item_name(&reg->hr_item);
+		if (region_uuid) {
+			if (strcmp(region_uuid, uuid))
+				continue;
+			found = 1;
+		}
 
-	if (ret)
-		goto out;
+		if (reg->hr_item_pinned) {
+			mlog(ML_CLUSTER, "Unpin region %s\n", uuid);
+			o2nm_undepend_item(&reg->hr_item);
+			reg->hr_item_pinned = 0;
+		}
+		if (found)
+			break;
+	}
+}
 
-	ret = o2nm_depend_this_node();
-	if (ret)
-		goto out;
+static int o2hb_region_inc_user(const char *region_uuid)
+{
+	int ret = 0;
 
-	ret = o2nm_depend_item(&reg->hr_item);
-	if (ret)
-		o2nm_undepend_this_node();
+	spin_lock(&o2hb_live_lock);
 
-out:
+	/* local heartbeat */
+	if (!o2hb_global_heartbeat_active()) {
+	    ret = o2hb_region_pin(region_uuid);
+	    goto unlock;
+	}
+
+	/*
+	 * if global heartbeat active and this is the first dependent user,
+	 * pin all regions if quorum region count <= CUT_OFF
+	 */
+	o2hb_dependent_users++;
+	if (o2hb_dependent_users > 1)
+		goto unlock;
+
+	if (o2hb_pop_count(&o2hb_quorum_region_bitmap,
+			   O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
+		ret = o2hb_region_pin(NULL);
+
+unlock:
+	spin_unlock(&o2hb_live_lock);
 	return ret;
 }
 
-static void o2hb_region_put(const char *region_uuid)
+void o2hb_region_dec_user(const char *region_uuid)
 {
-	struct o2hb_region *reg;
-
 	spin_lock(&o2hb_live_lock);
 
-	reg = o2hb_find_region(region_uuid);
+	/* local heartbeat */
+	if (!o2hb_global_heartbeat_active()) {
+	    o2hb_region_unpin(region_uuid);
+	    goto unlock;
+	}
 
-	spin_unlock(&o2hb_live_lock);
+	/*
+	 * if global heartbeat active and there are no dependent users,
+	 * unpin all quorum regions
+	 */
+	o2hb_dependent_users--;
+	if (!o2hb_dependent_users)
+		o2hb_region_unpin(NULL);
 
-	if (reg) {
-		o2nm_undepend_item(&reg->hr_item);
-		o2nm_undepend_this_node();
-	}
+unlock:
+	spin_unlock(&o2hb_live_lock);
 }
 
 int o2hb_register_callback(const char *region_uuid,
@@ -2293,9 +2431,11 @@ int o2hb_register_callback(const char *region_uuid,
 	}
 
 	if (region_uuid) {
-		ret = o2hb_region_get(region_uuid);
-		if (ret)
+		ret = o2hb_region_inc_user(region_uuid);
+		if (ret) {
+			mlog_errno(ret);
 			goto out;
+		}
 	}
 
 	down_write(&o2hb_callback_sem);
@@ -2313,7 +2453,7 @@ int o2hb_register_callback(const char *region_uuid,
 	up_write(&o2hb_callback_sem);
 	ret = 0;
 out:
-	mlog(ML_HEARTBEAT, "returning %d on behalf of %p for funcs %p\n",
+	mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n",
 	     ret, __builtin_return_address(0), hc);
 	return ret;
 }
@@ -2324,7 +2464,7 @@ void o2hb_unregister_callback(const char *region_uuid,
 {
 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
 
-	mlog(ML_HEARTBEAT, "on behalf of %p for funcs %p\n",
+	mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n",
 	     __builtin_return_address(0), hc);
 
 	/* XXX Can this happen _with_ a region reference? */
@@ -2332,7 +2472,7 @@ void o2hb_unregister_callback(const char *region_uuid,
 		return;
 
 	if (region_uuid)
-		o2hb_region_put(region_uuid);
+		o2hb_region_dec_user(region_uuid);
 
 	down_write(&o2hb_callback_sem);
 
-- 
cgit v1.2.3


From cb0586bd4c77c531fe0be4ae860ec642450eeda5 Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Tue, 14 Dec 2010 14:14:30 -0800
Subject: ocfs2/cluster: Show pin state for each o2hb region

This patch adds a per o2hb region debugfs file that shows whether that region
is pinned or not.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/heartbeat.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index ad2e41d6879d..3722e3850a1f 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -82,6 +82,7 @@ static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
 #define O2HB_DB_TYPE_REGION_LIVENODES	4
 #define O2HB_DB_TYPE_REGION_NUMBER	5
 #define O2HB_DB_TYPE_REGION_ELAPSED_TIME	6
+#define O2HB_DB_TYPE_REGION_PINNED	7
 struct o2hb_debug_buf {
 	int db_type;
 	int db_size;
@@ -101,6 +102,7 @@ static struct o2hb_debug_buf *o2hb_db_failedregions;
 #define O2HB_DEBUG_FAILEDREGIONS	"failed_regions"
 #define O2HB_DEBUG_REGION_NUMBER	"num"
 #define O2HB_DEBUG_REGION_ELAPSED_TIME	"elapsed_time_in_ms"
+#define O2HB_DEBUG_REGION_PINNED	"pinned"
 
 static struct dentry *o2hb_debug_dir;
 static struct dentry *o2hb_debug_livenodes;
@@ -241,9 +243,11 @@ struct o2hb_region {
 	struct dentry		*hr_debug_livenodes;
 	struct dentry		*hr_debug_regnum;
 	struct dentry		*hr_debug_elapsed_time;
+	struct dentry		*hr_debug_pinned;
 	struct o2hb_debug_buf	*hr_db_livenodes;
 	struct o2hb_debug_buf	*hr_db_regnum;
 	struct o2hb_debug_buf	*hr_db_elapsed_time;
+	struct o2hb_debug_buf	*hr_db_pinned;
 
 	/* let the person setting up hb wait for it to return until it
 	 * has reached a 'steady' state.  This will be fixed when we have
@@ -1180,6 +1184,12 @@ static int o2hb_debug_open(struct inode *inode, struct file *file)
 						 reg->hr_last_timeout_start));
 		goto done;
 
+	case O2HB_DB_TYPE_REGION_PINNED:
+		reg = (struct o2hb_region *)db->db_data;
+		out += snprintf(buf + out, PAGE_SIZE - out, "%u\n",
+				!!reg->hr_item_pinned);
+		goto done;
+
 	default:
 		goto done;
 	}
@@ -1424,6 +1434,7 @@ static void o2hb_region_release(struct config_item *item)
 	debugfs_remove(reg->hr_debug_livenodes);
 	debugfs_remove(reg->hr_debug_regnum);
 	debugfs_remove(reg->hr_debug_elapsed_time);
+	debugfs_remove(reg->hr_debug_pinned);
 	debugfs_remove(reg->hr_debug_dir);
 
 	spin_lock(&o2hb_live_lock);
@@ -1988,6 +1999,18 @@ static int o2hb_debug_region_init(struct o2hb_region *reg, struct dentry *dir)
 		goto bail;
 	}
 
+	reg->hr_debug_pinned =
+			o2hb_debug_create(O2HB_DEBUG_REGION_PINNED,
+					  reg->hr_debug_dir,
+					  &(reg->hr_db_pinned),
+					  sizeof(*(reg->hr_db_pinned)),
+					  O2HB_DB_TYPE_REGION_PINNED,
+					  0, 0, reg);
+	if (!reg->hr_debug_pinned) {
+		mlog_errno(ret);
+		goto bail;
+	}
+
 	ret = 0;
 bail:
 	return ret;
-- 
cgit v1.2.3


From cfc069d3fa24d7c6357e3b731e97f94db495bf0c Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Tue, 14 Dec 2010 14:14:31 -0800
Subject: ocfs2/cluster: Pin the local node when o2hb thread starts

The patch pins the node item of the local node when the o2hb thread
starts and unpins on stop.

An earlier patch pinned the node item of the remote node on o2net
connect and unpinned on disconnect.

Signed-off-by Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/heartbeat.c | 6 ++++++
 1 file changed, 6 insertions(+)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index 3722e3850a1f..8b50c1ad7a69 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -1083,6 +1083,9 @@ static int o2hb_thread(void *data)
 
 	set_user_nice(current, -20);
 
+	/* Pin node */
+	o2nm_depend_this_node();
+
 	while (!kthread_should_stop() && !reg->hr_unclean_stop) {
 		/* We track the time spent inside
 		 * o2hb_do_disk_heartbeat so that we avoid more than
@@ -1132,6 +1135,9 @@ static int o2hb_thread(void *data)
 		mlog_errno(ret);
 	}
 
+	/* Unpin node */
+	o2nm_undepend_this_node();
+
 	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread exiting\n");
 
 	return 0;
-- 
cgit v1.2.3


From 37096a7927decb0b1d3c2514b8adb4583a834112 Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Mon, 20 Dec 2010 16:35:00 -0800
Subject: ocfs2/dlm: Minor cleanup

Patch makes use of task_pid_nr(). Also removes the null check before calling
debugfs_remove().

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/netdebug.c | 20 +++++++-------------
 1 file changed, 7 insertions(+), 13 deletions(-)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c
index a3f150e52b02..0edf836b42d5 100644
--- a/fs/ocfs2/cluster/netdebug.c
+++ b/fs/ocfs2/cluster/netdebug.c
@@ -141,7 +141,7 @@ static int nst_seq_show(struct seq_file *seq, void *v)
 			   "  sock acquiry: %lu.%ld\n"
 			   "  send start:   %lu.%ld\n"
 			   "  wait start:   %lu.%ld\n",
-			   nst, (unsigned long)nst->st_task->pid,
+			   nst, (unsigned long)task_pid_nr(nst->st_task),
 			   (unsigned long)nst->st_task->tgid,
 			   nst->st_task->comm, nst->st_node,
 			   nst->st_sc, nst->st_id, nst->st_msg_type,
@@ -421,23 +421,17 @@ int o2net_debugfs_init(void)
 
 	return 0;
 bail:
-	if (sc_dentry)
-		debugfs_remove(sc_dentry);
-	if (nst_dentry)
-		debugfs_remove(nst_dentry);
-	if (o2net_dentry)
-		debugfs_remove(o2net_dentry);
+	debugfs_remove(sc_dentry);
+	debugfs_remove(nst_dentry);
+	debugfs_remove(o2net_dentry);
 	return -ENOMEM;
 }
 
 void o2net_debugfs_exit(void)
 {
-	if (sc_dentry)
-		debugfs_remove(sc_dentry);
-	if (nst_dentry)
-		debugfs_remove(nst_dentry);
-	if (o2net_dentry)
-		debugfs_remove(o2net_dentry);
+	debugfs_remove(sc_dentry);
+	debugfs_remove(nst_dentry);
+	debugfs_remove(o2net_dentry);
 }
 
 #endif	/* CONFIG_DEBUG_FS */
-- 
cgit v1.2.3


From 3f9c14fab0a2e90af9995f261a123f59e0b41141 Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Wed, 22 Dec 2010 12:39:38 -0800
Subject: ocfs2/cluster: Replace timeval with ktime in struct
 o2net_send_tracking

Replace time trackers in struct o2net_send_tracking from struct timeval to
union ktime.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/netdebug.c     | 22 +++++++++++++---------
 fs/ocfs2/cluster/tcp.c          |  6 +++---
 fs/ocfs2/cluster/tcp_internal.h |  6 +++---
 3 files changed, 19 insertions(+), 15 deletions(-)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c
index 0edf836b42d5..2b986aa82299 100644
--- a/fs/ocfs2/cluster/netdebug.c
+++ b/fs/ocfs2/cluster/netdebug.c
@@ -123,10 +123,17 @@ static void *nst_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 static int nst_seq_show(struct seq_file *seq, void *v)
 {
 	struct o2net_send_tracking *nst, *dummy_nst = seq->private;
+	ktime_t now;
+	s64 sock, send, status;
 
 	spin_lock(&o2net_debug_lock);
 	nst = next_nst(dummy_nst);
 
+	now = ktime_get();
+	sock = ktime_to_us(ktime_sub(now, nst->st_sock_time));
+	send = ktime_to_us(ktime_sub(now, nst->st_send_time));
+	status = ktime_to_us(ktime_sub(now, nst->st_status_time));
+
 	if (nst != NULL) {
 		/* get_task_comm isn't exported.  oh well. */
 		seq_printf(seq, "%p:\n"
@@ -138,20 +145,17 @@ static int nst_seq_show(struct seq_file *seq, void *v)
 			   "  message id:   %d\n"
 			   "  message type: %u\n"
 			   "  message key:  0x%08x\n"
-			   "  sock acquiry: %lu.%ld\n"
-			   "  send start:   %lu.%ld\n"
-			   "  wait start:   %lu.%ld\n",
+			   "  sock acquiry: %lld usecs ago\n"
+			   "  send start:   %lld usecs ago\n"
+			   "  wait start:   %lld usecs ago\n",
 			   nst, (unsigned long)task_pid_nr(nst->st_task),
 			   (unsigned long)nst->st_task->tgid,
 			   nst->st_task->comm, nst->st_node,
 			   nst->st_sc, nst->st_id, nst->st_msg_type,
 			   nst->st_msg_key,
-			   nst->st_sock_time.tv_sec,
-			   (long)nst->st_sock_time.tv_usec,
-			   nst->st_send_time.tv_sec,
-			   (long)nst->st_send_time.tv_usec,
-			   nst->st_status_time.tv_sec,
-			   (long)nst->st_status_time.tv_usec);
+			   (long long)sock,
+			   (long long)send,
+			   (long long)status);
 	}
 
 	spin_unlock(&o2net_debug_lock);
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 92de96cd247d..49c1a95e352e 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -155,17 +155,17 @@ static void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype,
 
 static void o2net_set_nst_sock_time(struct o2net_send_tracking *nst)
 {
-	do_gettimeofday(&nst->st_sock_time);
+	nst->st_sock_time = ktime_get();
 }
 
 static void o2net_set_nst_send_time(struct o2net_send_tracking *nst)
 {
-	do_gettimeofday(&nst->st_send_time);
+	nst->st_send_time = ktime_get();
 }
 
 static void o2net_set_nst_status_time(struct o2net_send_tracking *nst)
 {
-	do_gettimeofday(&nst->st_status_time);
+	nst->st_status_time = ktime_get();
 }
 
 static void o2net_set_nst_sock_container(struct o2net_send_tracking *nst,
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
index 15fdbdf9eb4b..b613aaaf3e5c 100644
--- a/fs/ocfs2/cluster/tcp_internal.h
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -220,9 +220,9 @@ struct o2net_send_tracking {
 	u32				st_msg_type;
 	u32				st_msg_key;
 	u8				st_node;
-	struct timeval			st_sock_time;
-	struct timeval			st_send_time;
-	struct timeval			st_status_time;
+	ktime_t				st_sock_time;
+	ktime_t				st_send_time;
+	ktime_t				st_status_time;
 };
 #else
 struct o2net_send_tracking {
-- 
cgit v1.2.3


From ff1becbf85bf4d4d4652915b7ab27db949585f6b Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Wed, 22 Dec 2010 12:39:39 -0800
Subject: ocfs2/cluster: Use ktime instead of timeval in struct
 o2net_sock_container

Replace time trackers in struct o2net_sock_container from struct timeval to
union ktime.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/netdebug.c     | 28 ++++++-------
 fs/ocfs2/cluster/tcp.c          | 91 ++++++++++++++++++++++++-----------------
 fs/ocfs2/cluster/tcp_internal.h | 19 +++++----
 3 files changed, 76 insertions(+), 62 deletions(-)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c
index 2b986aa82299..536a93d13a06 100644
--- a/fs/ocfs2/cluster/netdebug.c
+++ b/fs/ocfs2/cluster/netdebug.c
@@ -280,8 +280,6 @@ static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 	return sc; /* unused, just needs to be null when done */
 }
 
-#define TV_SEC_USEC(TV) TV.tv_sec, (long)TV.tv_usec
-
 static int sc_seq_show(struct seq_file *seq, void *v)
 {
 	struct o2net_sock_container *sc, *dummy_sc = seq->private;
@@ -313,13 +311,13 @@ static int sc_seq_show(struct seq_file *seq, void *v)
 			   "  remote node:     %s\n"
 			   "  page off:        %zu\n"
 			   "  handshake ok:    %u\n"
-			   "  timer:           %lu.%ld\n"
-			   "  data ready:      %lu.%ld\n"
-			   "  advance start:   %lu.%ld\n"
-			   "  advance stop:    %lu.%ld\n"
-			   "  func start:      %lu.%ld\n"
-			   "  func stop:       %lu.%ld\n"
-			   "  func key:        %u\n"
+			   "  timer:           %lld usecs\n"
+			   "  data ready:      %lld usecs\n"
+			   "  advance start:   %lld usecs\n"
+			   "  advance stop:    %lld usecs\n"
+			   "  func start:      %lld usecs\n"
+			   "  func stop:       %lld usecs\n"
+			   "  func key:        0x%08x\n"
 			   "  func type:       %u\n",
 			   sc,
 			   atomic_read(&sc->sc_kref.refcount),
@@ -328,12 +326,12 @@ static int sc_seq_show(struct seq_file *seq, void *v)
 			   sc->sc_node->nd_name,
 			   sc->sc_page_off,
 			   sc->sc_handshake_ok,
-			   TV_SEC_USEC(sc->sc_tv_timer),
-			   TV_SEC_USEC(sc->sc_tv_data_ready),
-			   TV_SEC_USEC(sc->sc_tv_advance_start),
-			   TV_SEC_USEC(sc->sc_tv_advance_stop),
-			   TV_SEC_USEC(sc->sc_tv_func_start),
-			   TV_SEC_USEC(sc->sc_tv_func_stop),
+			   (long long)ktime_to_us(sc->sc_tv_timer),
+			   (long long)ktime_to_us(sc->sc_tv_data_ready),
+			   (long long)ktime_to_us(sc->sc_tv_advance_start),
+			   (long long)ktime_to_us(sc->sc_tv_advance_stop),
+			   (long long)ktime_to_us(sc->sc_tv_func_start),
+			   (long long)ktime_to_us(sc->sc_tv_func_stop),
 			   sc->sc_msg_key,
 			   sc->sc_msg_type);
 	}
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 49c1a95e352e..efd848334052 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -153,61 +153,75 @@ static void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype,
 	nst->st_node = node;
 }
 
-static void o2net_set_nst_sock_time(struct o2net_send_tracking *nst)
+static inline void o2net_set_nst_sock_time(struct o2net_send_tracking *nst)
 {
 	nst->st_sock_time = ktime_get();
 }
 
-static void o2net_set_nst_send_time(struct o2net_send_tracking *nst)
+static inline void o2net_set_nst_send_time(struct o2net_send_tracking *nst)
 {
 	nst->st_send_time = ktime_get();
 }
 
-static void o2net_set_nst_status_time(struct o2net_send_tracking *nst)
+static inline void o2net_set_nst_status_time(struct o2net_send_tracking *nst)
 {
 	nst->st_status_time = ktime_get();
 }
 
-static void o2net_set_nst_sock_container(struct o2net_send_tracking *nst,
-					 struct o2net_sock_container *sc)
+static inline void o2net_set_nst_sock_container(struct o2net_send_tracking *nst,
+						struct o2net_sock_container *sc)
 {
 	nst->st_sc = sc;
 }
 
-static void o2net_set_nst_msg_id(struct o2net_send_tracking *nst, u32 msg_id)
+static inline void o2net_set_nst_msg_id(struct o2net_send_tracking *nst,
+					u32 msg_id)
 {
 	nst->st_id = msg_id;
 }
 
-#else  /* CONFIG_DEBUG_FS */
-
-static inline void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype,
-				  u32 msgkey, struct task_struct *task, u8 node)
+static inline void o2net_set_sock_timer(struct o2net_sock_container *sc)
 {
+	sc->sc_tv_timer = ktime_get();
 }
 
-static inline void o2net_set_nst_sock_time(struct o2net_send_tracking *nst)
+static inline void o2net_set_data_ready_time(struct o2net_sock_container *sc)
 {
+	sc->sc_tv_data_ready = ktime_get();
 }
 
-static inline void o2net_set_nst_send_time(struct o2net_send_tracking *nst)
+static inline void o2net_set_advance_start_time(struct o2net_sock_container *sc)
 {
+	sc->sc_tv_advance_start = ktime_get();
 }
 
-static inline void o2net_set_nst_status_time(struct o2net_send_tracking *nst)
+static inline void o2net_set_advance_stop_time(struct o2net_sock_container *sc)
 {
+	sc->sc_tv_advance_stop = ktime_get();
 }
 
-static inline void o2net_set_nst_sock_container(struct o2net_send_tracking *nst,
-						struct o2net_sock_container *sc)
+static inline void o2net_set_func_start_time(struct o2net_sock_container *sc)
 {
+	sc->sc_tv_func_start = ktime_get();
 }
 
-static inline void o2net_set_nst_msg_id(struct o2net_send_tracking *nst,
-					u32 msg_id)
+static inline void o2net_set_func_stop_time(struct o2net_sock_container *sc)
 {
+	sc->sc_tv_func_stop = ktime_get();
 }
-
+#else  /* CONFIG_DEBUG_FS */
+# define o2net_init_nst(a, b, c, d, e)
+# define o2net_set_nst_sock_time(a)
+# define o2net_set_nst_send_time(a)
+# define o2net_set_nst_status_time(a)
+# define o2net_set_nst_sock_container(a, b)
+# define o2net_set_nst_msg_id(a, b)
+# define o2net_set_sock_timer(a)
+# define o2net_set_data_ready_time(a)
+# define o2net_set_advance_start_time(a)
+# define o2net_set_advance_stop_time(a)
+# define o2net_set_func_start_time(a)
+# define o2net_set_func_stop_time(a)
 #endif /* CONFIG_DEBUG_FS */
 
 static inline int o2net_reconnect_delay(void)
@@ -555,7 +569,7 @@ static void o2net_data_ready(struct sock *sk, int bytes)
 	if (sk->sk_user_data) {
 		struct o2net_sock_container *sc = sk->sk_user_data;
 		sclog(sc, "data_ready hit\n");
-		do_gettimeofday(&sc->sc_tv_data_ready);
+		o2net_set_data_ready_time(sc);
 		o2net_sc_queue_work(sc, &sc->sc_rx_work);
 		ready = sc->sc_data_ready;
 	} else {
@@ -1192,13 +1206,13 @@ static int o2net_process_message(struct o2net_sock_container *sc,
 	if (syserr != O2NET_ERR_NONE)
 		goto out_respond;
 
-	do_gettimeofday(&sc->sc_tv_func_start);
+	o2net_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 o2net_msg) +
 					     be16_to_cpu(hdr->data_len),
 					nmh->nh_func_data, &ret_data);
-	do_gettimeofday(&sc->sc_tv_func_stop);
+	o2net_set_func_stop_time(sc);
 
 out_respond:
 	/* this destroys the hdr, so don't use it after this */
@@ -1309,7 +1323,7 @@ static int o2net_advance_rx(struct o2net_sock_container *sc)
 	size_t datalen;
 
 	sclog(sc, "receiving\n");
-	do_gettimeofday(&sc->sc_tv_advance_start);
+	o2net_set_advance_start_time(sc);
 
 	if (unlikely(sc->sc_handshake_ok == 0)) {
 		if(sc->sc_page_off < sizeof(struct o2net_handshake)) {
@@ -1384,7 +1398,7 @@ static int o2net_advance_rx(struct o2net_sock_container *sc)
 
 out:
 	sclog(sc, "ret = %d\n", ret);
-	do_gettimeofday(&sc->sc_tv_advance_stop);
+	o2net_set_advance_stop_time(sc);
 	return ret;
 }
 
@@ -1484,27 +1498,28 @@ static void o2net_idle_timer(unsigned long data)
 {
 	struct o2net_sock_container *sc = (struct o2net_sock_container *)data;
 	struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
-	struct timeval now;
 
-	do_gettimeofday(&now);
+#ifdef CONFIG_DEBUG_FS
+	ktime_t now = ktime_get();
+#endif
 
 	printk(KERN_NOTICE "o2net: connection to " SC_NODEF_FMT " has been idle for %u.%u "
 	     "seconds, shutting it down.\n", SC_NODEF_ARGS(sc),
 		     o2net_idle_timeout() / 1000,
 		     o2net_idle_timeout() % 1000);
-	mlog(ML_NOTICE, "here are some times that might help debug the "
-	     "situation: (tmr %ld.%ld now %ld.%ld dr %ld.%ld adv "
-	     "%ld.%ld:%ld.%ld func (%08x:%u) %ld.%ld:%ld.%ld)\n",
-	     sc->sc_tv_timer.tv_sec, (long) sc->sc_tv_timer.tv_usec,
-	     now.tv_sec, (long) now.tv_usec,
-	     sc->sc_tv_data_ready.tv_sec, (long) sc->sc_tv_data_ready.tv_usec,
-	     sc->sc_tv_advance_start.tv_sec,
-	     (long) sc->sc_tv_advance_start.tv_usec,
-	     sc->sc_tv_advance_stop.tv_sec,
-	     (long) sc->sc_tv_advance_stop.tv_usec,
+
+#ifdef CONFIG_DEBUG_FS
+	mlog(ML_NOTICE, "Here are some times that might help debug the "
+	     "situation: (Timer: %lld, Now %lld, DataReady %lld, Advance %lld-%lld, "
+	     "Key 0x%08x, Func %u, FuncTime %lld-%lld)\n",
+	     (long long)ktime_to_us(sc->sc_tv_timer), (long long)ktime_to_us(now),
+	     (long long)ktime_to_us(sc->sc_tv_data_ready),
+	     (long long)ktime_to_us(sc->sc_tv_advance_start),
+	     (long long)ktime_to_us(sc->sc_tv_advance_stop),
 	     sc->sc_msg_key, sc->sc_msg_type,
-	     sc->sc_tv_func_start.tv_sec, (long) sc->sc_tv_func_start.tv_usec,
-	     sc->sc_tv_func_stop.tv_sec, (long) sc->sc_tv_func_stop.tv_usec);
+	     (long long)ktime_to_us(sc->sc_tv_func_start),
+	     (long long)ktime_to_us(sc->sc_tv_func_stop));
+#endif
 
 	/*
 	 * Initialize the nn_timeout so that the next connection attempt
@@ -1520,7 +1535,7 @@ static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc)
 	o2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work);
 	o2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work,
 		      msecs_to_jiffies(o2net_keepalive_delay()));
-	do_gettimeofday(&sc->sc_tv_timer);
+	o2net_set_sock_timer(sc);
 	mod_timer(&sc->sc_idle_timeout,
 	       jiffies + msecs_to_jiffies(o2net_idle_timeout()));
 }
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
index b613aaaf3e5c..f81576333911 100644
--- a/fs/ocfs2/cluster/tcp_internal.h
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -166,18 +166,19 @@ struct o2net_sock_container {
 	/* original handlers for the sockets */
 	void			(*sc_state_change)(struct sock *sk);
 	void			(*sc_data_ready)(struct sock *sk, int bytes);
-#ifdef CONFIG_DEBUG_FS
-	struct list_head        sc_net_debug_item;
-#endif
-	struct timeval 		sc_tv_timer;
-	struct timeval 		sc_tv_data_ready;
-	struct timeval 		sc_tv_advance_start;
-	struct timeval 		sc_tv_advance_stop;
-	struct timeval 		sc_tv_func_start;
-	struct timeval 		sc_tv_func_stop;
+
 	u32			sc_msg_key;
 	u16			sc_msg_type;
 
+#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
 	struct mutex		sc_send_lock;
 };
 
-- 
cgit v1.2.3


From 3c193b3807e933cf2a16d55a38debbe549195847 Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Wed, 22 Dec 2010 12:39:40 -0800
Subject: ocfs2/cluster: Track send message timing stats for each socket

Tracks total send and status times for all messages sent on a socket.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/tcp.c          | 24 ++++++++++++++++++++++++
 fs/ocfs2/cluster/tcp_internal.h |  6 ++++++
 2 files changed, 30 insertions(+)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index efd848334052..4d61e19d6146 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -224,6 +224,28 @@ static inline void o2net_set_func_stop_time(struct o2net_sock_container *sc)
 # define o2net_set_func_stop_time(a)
 #endif /* CONFIG_DEBUG_FS */
 
+#ifdef CONFIG_OCFS2_FS_STATS
+static void o2net_update_send_stats(struct o2net_send_tracking *nst,
+				    struct o2net_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++;
+}
+
+#else
+
+# define o2net_update_send_stats(a, b)
+
+#endif /* CONFIG_OCFS2_FS_STATS */
+
 static inline int o2net_reconnect_delay(void)
 {
 	return o2nm_single_cluster->cl_reconnect_delay_ms;
@@ -1093,6 +1115,8 @@ int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec,
 	o2net_set_nst_status_time(&nst);
 	wait_event(nsw.ns_wq, o2net_nsw_completed(nn, &nsw));
 
+	o2net_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. */
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
index f81576333911..640c6fcef720 100644
--- a/fs/ocfs2/cluster/tcp_internal.h
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -178,6 +178,12 @@ struct o2net_sock_container {
 	ktime_t			sc_tv_advance_stop;
 	ktime_t			sc_tv_func_start;
 	ktime_t			sc_tv_func_stop;
+#endif
+#ifdef CONFIG_OCFS2_FS_STATS
+	ktime_t			sc_tv_acquiry_total;
+	ktime_t			sc_tv_send_total;
+	ktime_t			sc_tv_status_total;
+	u32			sc_send_count;
 #endif
 	struct mutex		sc_send_lock;
 };
-- 
cgit v1.2.3


From e453039f8bf44abf82f3ecfb34177e0cb04bce12 Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Wed, 22 Dec 2010 12:39:41 -0800
Subject: ocfs2/cluster: Track process message timing stats for each socket

Tracks total time taken to process messages received on a socket.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/tcp.c          | 15 +++++++++++++++
 fs/ocfs2/cluster/tcp_internal.h |  2 ++
 2 files changed, 17 insertions(+)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 4d61e19d6146..bc2309554d0b 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -209,6 +209,11 @@ static inline void o2net_set_func_stop_time(struct o2net_sock_container *sc)
 {
 	sc->sc_tv_func_stop = ktime_get();
 }
+
+static ktime_t o2net_get_func_run_time(struct o2net_sock_container *sc)
+{
+	return ktime_sub(sc->sc_tv_func_stop, sc->sc_tv_func_start);
+}
 #else  /* CONFIG_DEBUG_FS */
 # define o2net_init_nst(a, b, c, d, e)
 # define o2net_set_nst_sock_time(a)
@@ -222,6 +227,7 @@ static inline void o2net_set_func_stop_time(struct o2net_sock_container *sc)
 # define o2net_set_advance_stop_time(a)
 # define o2net_set_func_start_time(a)
 # define o2net_set_func_stop_time(a)
+# define o2net_get_func_run_time(a)		(ktime_t)0
 #endif /* CONFIG_DEBUG_FS */
 
 #ifdef CONFIG_OCFS2_FS_STATS
@@ -240,6 +246,13 @@ static void o2net_update_send_stats(struct o2net_send_tracking *nst,
 	sc->sc_send_count++;
 }
 
+static void o2net_update_recv_stats(struct o2net_sock_container *sc)
+{
+	sc->sc_tv_process_total = ktime_add(sc->sc_tv_process_total,
+					    o2net_get_func_run_time(sc));
+	sc->sc_recv_count++;
+}
+
 #else
 
 # define o2net_update_send_stats(a, b)
@@ -1238,6 +1251,8 @@ static int o2net_process_message(struct o2net_sock_container *sc,
 					nmh->nh_func_data, &ret_data);
 	o2net_set_func_stop_time(sc);
 
+	o2net_update_recv_stats(sc);
+
 out_respond:
 	/* this destroys the hdr, so don't use it after this */
 	mutex_lock(&sc->sc_send_lock);
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
index 640c6fcef720..4cbcb65784a3 100644
--- a/fs/ocfs2/cluster/tcp_internal.h
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -184,6 +184,8 @@ struct o2net_sock_container {
 	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;
 };
-- 
cgit v1.2.3


From db02754c8a1205b24beac70562c45ca5d671151f Mon Sep 17 00:00:00 2001
From: Sunil Mushran <sunil.mushran@oracle.com>
Date: Wed, 22 Dec 2010 12:39:42 -0800
Subject: ocfs2/cluster: Show o2net timing statistics

Adds debugfs dentry o2net/stats to show the o2net timing statistics.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/netdebug.c | 211 ++++++++++++++++++++++++++++++++------------
 1 file changed, 157 insertions(+), 54 deletions(-)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c
index 536a93d13a06..61df89cedded 100644
--- a/fs/ocfs2/cluster/netdebug.c
+++ b/fs/ocfs2/cluster/netdebug.c
@@ -46,10 +46,15 @@
 #define O2NET_DEBUG_DIR		"o2net"
 #define SC_DEBUG_NAME		"sock_containers"
 #define NST_DEBUG_NAME		"send_tracking"
+#define STATS_DEBUG_NAME	"stats"
+
+#define SHOW_SOCK_CONTAINERS	0
+#define SHOW_SOCK_STATS		1
 
 static struct dentry *o2net_dentry;
 static struct dentry *sc_dentry;
 static struct dentry *nst_dentry;
+static struct dentry *stats_dentry;
 
 static DEFINE_SPINLOCK(o2net_debug_lock);
 
@@ -232,6 +237,11 @@ void o2net_debug_del_sc(struct o2net_sock_container *sc)
 	spin_unlock(&o2net_debug_lock);
 }
 
+struct o2net_sock_debug {
+	int dbg_ctxt;
+	struct o2net_sock_container *dbg_sock;
+};
+
 static struct o2net_sock_container
 			*next_sc(struct o2net_sock_container *sc_start)
 {
@@ -257,7 +267,8 @@ static struct o2net_sock_container
 
 static void *sc_seq_start(struct seq_file *seq, loff_t *pos)
 {
-	struct o2net_sock_container *sc, *dummy_sc = seq->private;
+	struct o2net_sock_debug *sd = seq->private;
+	struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock;
 
 	spin_lock(&o2net_debug_lock);
 	sc = next_sc(dummy_sc);
@@ -268,7 +279,8 @@ static void *sc_seq_start(struct seq_file *seq, loff_t *pos)
 
 static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 {
-	struct o2net_sock_container *sc, *dummy_sc = seq->private;
+	struct o2net_sock_debug *sd = seq->private;
+	struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock;
 
 	spin_lock(&o2net_debug_lock);
 	sc = next_sc(dummy_sc);
@@ -280,63 +292,107 @@ static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 	return sc; /* unused, just needs to be null when done */
 }
 
-static int sc_seq_show(struct seq_file *seq, void *v)
+#ifdef CONFIG_OCFS2_FS_STATS
+# define sc_send_count(_s)		((_s)->sc_send_count)
+# define sc_recv_count(_s)		((_s)->sc_recv_count)
+# define sc_tv_acquiry_total_ns(_s)	(ktime_to_ns((_s)->sc_tv_acquiry_total))
+# define sc_tv_send_total_ns(_s)	(ktime_to_ns((_s)->sc_tv_send_total))
+# define sc_tv_status_total_ns(_s)	(ktime_to_ns((_s)->sc_tv_status_total))
+# define sc_tv_process_total_ns(_s)	(ktime_to_ns((_s)->sc_tv_process_total))
+#else
+# define sc_send_count(_s)		(0U)
+# define sc_recv_count(_s)		(0U)
+# define sc_tv_acquiry_total_ns(_s)	(0LL)
+# define sc_tv_send_total_ns(_s)	(0LL)
+# define sc_tv_status_total_ns(_s)	(0LL)
+# define sc_tv_process_total_ns(_s)	(0LL)
+#endif
+
+/* So that debugfs.ocfs2 can determine which format is being used */
+#define O2NET_STATS_STR_VERSION		1
+static void sc_show_sock_stats(struct seq_file *seq,
+			       struct o2net_sock_container *sc)
 {
-	struct o2net_sock_container *sc, *dummy_sc = seq->private;
+	if (!sc)
+		return;
+
+	seq_printf(seq, "%d,%u,%lu,%lld,%lld,%lld,%lu,%lld\n", O2NET_STATS_STR_VERSION,
+		   sc->sc_node->nd_num, (unsigned long)sc_send_count(sc),
+		   (long long)sc_tv_acquiry_total_ns(sc),
+		   (long long)sc_tv_send_total_ns(sc),
+		   (long long)sc_tv_status_total_ns(sc),
+		   (unsigned long)sc_recv_count(sc),
+		   (long long)sc_tv_process_total_ns(sc));
+}
 
-	spin_lock(&o2net_debug_lock);
-	sc = next_sc(dummy_sc);
+static void sc_show_sock_container(struct seq_file *seq,
+				   struct o2net_sock_container *sc)
+{
+	struct inet_sock *inet = NULL;
+	__be32 saddr = 0, daddr = 0;
+	__be16 sport = 0, dport = 0;
+
+	if (!sc)
+		return;
+
+	if (sc->sc_sock) {
+		inet = inet_sk(sc->sc_sock->sk);
+		/* the stack's structs aren't sparse endian clean */
+		saddr = (__force __be32)inet->inet_saddr;
+		daddr = (__force __be32)inet->inet_daddr;
+		sport = (__force __be16)inet->inet_sport;
+		dport = (__force __be16)inet->inet_dport;
+	}
 
-	if (sc != NULL) {
-		struct inet_sock *inet = NULL;
+	/* XXX sigh, inet-> doesn't have sparse annotation so any
+	 * use of it here generates a warning with -Wbitwise */
+	seq_printf(seq, "%p:\n"
+		   "  krefs:           %d\n"
+		   "  sock:            %pI4:%u -> "
+				      "%pI4:%u\n"
+		   "  remote node:     %s\n"
+		   "  page off:        %zu\n"
+		   "  handshake ok:    %u\n"
+		   "  timer:           %lld usecs\n"
+		   "  data ready:      %lld usecs\n"
+		   "  advance start:   %lld usecs\n"
+		   "  advance stop:    %lld usecs\n"
+		   "  func start:      %lld usecs\n"
+		   "  func stop:       %lld usecs\n"
+		   "  func key:        0x%08x\n"
+		   "  func type:       %u\n",
+		   sc,
+		   atomic_read(&sc->sc_kref.refcount),
+		   &saddr, inet ? ntohs(sport) : 0,
+		   &daddr, inet ? ntohs(dport) : 0,
+		   sc->sc_node->nd_name,
+		   sc->sc_page_off,
+		   sc->sc_handshake_ok,
+		   (long long)ktime_to_us(sc->sc_tv_timer),
+		   (long long)ktime_to_us(sc->sc_tv_data_ready),
+		   (long long)ktime_to_us(sc->sc_tv_advance_start),
+		   (long long)ktime_to_us(sc->sc_tv_advance_stop),
+		   (long long)ktime_to_us(sc->sc_tv_func_start),
+		   (long long)ktime_to_us(sc->sc_tv_func_stop),
+		   sc->sc_msg_key,
+		   sc->sc_msg_type);
+}
 
-		__be32 saddr = 0, daddr = 0;
-		__be16 sport = 0, dport = 0;
+static int sc_seq_show(struct seq_file *seq, void *v)
+{
+	struct o2net_sock_debug *sd = seq->private;
+	struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock;
 
-		if (sc->sc_sock) {
-			inet = inet_sk(sc->sc_sock->sk);
-			/* the stack's structs aren't sparse endian clean */
-			saddr = (__force __be32)inet->inet_saddr;
-			daddr = (__force __be32)inet->inet_daddr;
-			sport = (__force __be16)inet->inet_sport;
-			dport = (__force __be16)inet->inet_dport;
-		}
+	spin_lock(&o2net_debug_lock);
+	sc = next_sc(dummy_sc);
 
-		/* XXX sigh, inet-> doesn't have sparse annotation so any
-		 * use of it here generates a warning with -Wbitwise */
-		seq_printf(seq, "%p:\n"
-			   "  krefs:           %d\n"
-			   "  sock:            %pI4:%u -> "
-					      "%pI4:%u\n"
-			   "  remote node:     %s\n"
-			   "  page off:        %zu\n"
-			   "  handshake ok:    %u\n"
-			   "  timer:           %lld usecs\n"
-			   "  data ready:      %lld usecs\n"
-			   "  advance start:   %lld usecs\n"
-			   "  advance stop:    %lld usecs\n"
-			   "  func start:      %lld usecs\n"
-			   "  func stop:       %lld usecs\n"
-			   "  func key:        0x%08x\n"
-			   "  func type:       %u\n",
-			   sc,
-			   atomic_read(&sc->sc_kref.refcount),
-			   &saddr, inet ? ntohs(sport) : 0,
-			   &daddr, inet ? ntohs(dport) : 0,
-			   sc->sc_node->nd_name,
-			   sc->sc_page_off,
-			   sc->sc_handshake_ok,
-			   (long long)ktime_to_us(sc->sc_tv_timer),
-			   (long long)ktime_to_us(sc->sc_tv_data_ready),
-			   (long long)ktime_to_us(sc->sc_tv_advance_start),
-			   (long long)ktime_to_us(sc->sc_tv_advance_stop),
-			   (long long)ktime_to_us(sc->sc_tv_func_start),
-			   (long long)ktime_to_us(sc->sc_tv_func_stop),
-			   sc->sc_msg_key,
-			   sc->sc_msg_type);
+	if (sc) {
+		if (sd->dbg_ctxt == SHOW_SOCK_CONTAINERS)
+			sc_show_sock_container(seq, sc);
+		else
+			sc_show_sock_stats(seq, sc);
 	}
 
-
 	spin_unlock(&o2net_debug_lock);
 
 	return 0;
@@ -353,7 +409,7 @@ static const struct seq_operations sc_seq_ops = {
 	.show = sc_seq_show,
 };
 
-static int sc_fop_open(struct inode *inode, struct file *file)
+static int sc_common_open(struct file *file, struct o2net_sock_debug *sd)
 {
 	struct o2net_sock_container *dummy_sc;
 	struct seq_file *seq;
@@ -371,7 +427,8 @@ static int sc_fop_open(struct inode *inode, struct file *file)
 		goto out;
 
 	seq = file->private_data;
-	seq->private = dummy_sc;
+	seq->private = sd;
+	sd->dbg_sock = dummy_sc;
 	o2net_debug_add_sc(dummy_sc);
 
 	dummy_sc = NULL;
@@ -384,12 +441,48 @@ out:
 static int sc_fop_release(struct inode *inode, struct file *file)
 {
 	struct seq_file *seq = file->private_data;
-	struct o2net_sock_container *dummy_sc = seq->private;
+	struct o2net_sock_debug *sd = seq->private;
+	struct o2net_sock_container *dummy_sc = sd->dbg_sock;
 
 	o2net_debug_del_sc(dummy_sc);
 	return seq_release_private(inode, file);
 }
 
+static int stats_fop_open(struct inode *inode, struct file *file)
+{
+	struct o2net_sock_debug *sd;
+
+	sd = kmalloc(sizeof(struct o2net_sock_debug), GFP_KERNEL);
+	if (sd == NULL)
+		return -ENOMEM;
+
+	sd->dbg_ctxt = SHOW_SOCK_STATS;
+	sd->dbg_sock = NULL;
+
+	return sc_common_open(file, sd);
+}
+
+static const struct file_operations stats_seq_fops = {
+	.open = stats_fop_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = sc_fop_release,
+};
+
+static int sc_fop_open(struct inode *inode, struct file *file)
+{
+	struct o2net_sock_debug *sd;
+
+	sd = kmalloc(sizeof(struct o2net_sock_debug), GFP_KERNEL);
+	if (sd == NULL)
+		return -ENOMEM;
+
+	sd->dbg_ctxt = SHOW_SOCK_CONTAINERS;
+	sd->dbg_sock = NULL;
+
+	return sc_common_open(file, sd);
+}
+
 static const struct file_operations sc_seq_fops = {
 	.open = sc_fop_open,
 	.read = seq_read,
@@ -421,8 +514,17 @@ int o2net_debugfs_init(void)
 		goto bail;
 	}
 
+	stats_dentry = debugfs_create_file(STATS_DEBUG_NAME, S_IFREG|S_IRUSR,
+					   o2net_dentry, NULL,
+					   &stats_seq_fops);
+	if (!stats_dentry) {
+		mlog_errno(-ENOMEM);
+		goto bail;
+	}
+
 	return 0;
 bail:
+	debugfs_remove(stats_dentry);
 	debugfs_remove(sc_dentry);
 	debugfs_remove(nst_dentry);
 	debugfs_remove(o2net_dentry);
@@ -431,6 +533,7 @@ bail:
 
 void o2net_debugfs_exit(void)
 {
+	debugfs_remove(stats_dentry);
 	debugfs_remove(sc_dentry);
 	debugfs_remove(nst_dentry);
 	debugfs_remove(o2net_dentry);
-- 
cgit v1.2.3


From e70d84501b630e390b4242fb4ac629b694c11484 Mon Sep 17 00:00:00 2001
From: Randy Dunlap <randy.dunlap@oracle.com>
Date: Mon, 27 Dec 2010 11:55:08 -0800
Subject: ocfs2: fix build for OCFS2_FS_STATS not enabled

When CONFIG_OCFS2_FS_STATS is not enabled:

fs/ocfs2/cluster/tcp.c:1254: error: implicit declaration of function 'o2net_update_recv_stats'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc:	Mark Fasheh <mfasheh@suse.com>
Cc:	Joel Becker <joel.becker@oracle.com>
Cc:	ocfs2-devel@oss.oracle.com
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/tcp.c | 2 ++
 1 file changed, 2 insertions(+)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index bc2309554d0b..3b11cb1e38fc 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -257,6 +257,8 @@ static void o2net_update_recv_stats(struct o2net_sock_container *sc)
 
 # define o2net_update_send_stats(a, b)
 
+# define o2net_update_recv_stats(sc)
+
 #endif /* CONFIG_OCFS2_FS_STATS */
 
 static inline int o2net_reconnect_delay(void)
-- 
cgit v1.2.3


From cc548166b2aadba7a566fb0a42884a9d2ff684b0 Mon Sep 17 00:00:00 2001
From: Dan Carpenter <error27@gmail.com>
Date: Mon, 3 Jan 2011 09:00:20 +0300
Subject: ocfs2/cluster: dereferencing before checking in nst_seq_show()

In the original code, we dereferenced "nst" before checking that it was
non-NULL.  I moved the check forward and pulled the code in an indent
level.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
 fs/ocfs2/cluster/netdebug.c | 47 +++++++++++++++++++++++----------------------
 1 file changed, 24 insertions(+), 23 deletions(-)

(limited to 'fs/ocfs2/cluster')

diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c
index 61df89cedded..3a5835904b3d 100644
--- a/fs/ocfs2/cluster/netdebug.c
+++ b/fs/ocfs2/cluster/netdebug.c
@@ -133,36 +133,37 @@ static int nst_seq_show(struct seq_file *seq, void *v)
 
 	spin_lock(&o2net_debug_lock);
 	nst = next_nst(dummy_nst);
+	if (!nst)
+		goto out;
 
 	now = ktime_get();
 	sock = ktime_to_us(ktime_sub(now, nst->st_sock_time));
 	send = ktime_to_us(ktime_sub(now, nst->st_send_time));
 	status = ktime_to_us(ktime_sub(now, nst->st_status_time));
 
-	if (nst != NULL) {
-		/* get_task_comm isn't exported.  oh well. */
-		seq_printf(seq, "%p:\n"
-			   "  pid:          %lu\n"
-			   "  tgid:         %lu\n"
-			   "  process name: %s\n"
-			   "  node:         %u\n"
-			   "  sc:           %p\n"
-			   "  message id:   %d\n"
-			   "  message type: %u\n"
-			   "  message key:  0x%08x\n"
-			   "  sock acquiry: %lld usecs ago\n"
-			   "  send start:   %lld usecs ago\n"
-			   "  wait start:   %lld usecs ago\n",
-			   nst, (unsigned long)task_pid_nr(nst->st_task),
-			   (unsigned long)nst->st_task->tgid,
-			   nst->st_task->comm, nst->st_node,
-			   nst->st_sc, nst->st_id, nst->st_msg_type,
-			   nst->st_msg_key,
-			   (long long)sock,
-			   (long long)send,
-			   (long long)status);
-	}
+	/* get_task_comm isn't exported.  oh well. */
+	seq_printf(seq, "%p:\n"
+		   "  pid:          %lu\n"
+		   "  tgid:         %lu\n"
+		   "  process name: %s\n"
+		   "  node:         %u\n"
+		   "  sc:           %p\n"
+		   "  message id:   %d\n"
+		   "  message type: %u\n"
+		   "  message key:  0x%08x\n"
+		   "  sock acquiry: %lld usecs ago\n"
+		   "  send start:   %lld usecs ago\n"
+		   "  wait start:   %lld usecs ago\n",
+		   nst, (unsigned long)task_pid_nr(nst->st_task),
+		   (unsigned long)nst->st_task->tgid,
+		   nst->st_task->comm, nst->st_node,
+		   nst->st_sc, nst->st_id, nst->st_msg_type,
+		   nst->st_msg_key,
+		   (long long)sock,
+		   (long long)send,
+		   (long long)status);
 
+out:
 	spin_unlock(&o2net_debug_lock);
 
 	return 0;
-- 
cgit v1.2.3