summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--MAINTAINERS5
-rw-r--r--fs/ocfs2/Kconfig2
-rw-r--r--fs/ocfs2/alloc.c77
-rw-r--r--fs/ocfs2/alloc.h4
-rw-r--r--fs/ocfs2/aops.c59
-rw-r--r--fs/ocfs2/cluster/heartbeat.c246
-rw-r--r--fs/ocfs2/cluster/netdebug.c286
-rw-r--r--fs/ocfs2/cluster/tcp.c145
-rw-r--r--fs/ocfs2/cluster/tcp_internal.h33
-rw-r--r--fs/ocfs2/dlm/dlmast.c76
-rw-r--r--fs/ocfs2/dlm/dlmcommon.h86
-rw-r--r--fs/ocfs2/dlm/dlmdebug.c200
-rw-r--r--fs/ocfs2/dlm/dlmdebug.h5
-rw-r--r--fs/ocfs2/dlm/dlmdomain.c10
-rw-r--r--fs/ocfs2/dlm/dlmlock.c3
-rw-r--r--fs/ocfs2/dlm/dlmthread.c132
-rw-r--r--fs/ocfs2/namei.c5
-rw-r--r--fs/ocfs2/ocfs2.h5
18 files changed, 845 insertions, 534 deletions
diff --git a/MAINTAINERS b/MAINTAINERS
index bb6c1ac85138..42f991e5a85d 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -1785,7 +1785,8 @@ S: Maintained
F: drivers/usb/atm/cxacru.c
CONFIGFS
-M: Joel Becker <joel.becker@oracle.com>
+M: Joel Becker <jlbec@evilplan.org>
+T: git git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/configfs.git
S: Supported
F: fs/configfs/
F: include/linux/configfs.h
@@ -4549,7 +4550,7 @@ F: include/linux/oprofile.h
ORACLE CLUSTER FILESYSTEM 2 (OCFS2)
M: Mark Fasheh <mfasheh@suse.com>
-M: Joel Becker <joel.becker@oracle.com>
+M: Joel Becker <jlbec@evilplan.org>
L: ocfs2-devel@oss.oracle.com (moderated for non-subscribers)
W: http://oss.oracle.com/projects/ocfs2/
T: git git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2.git
diff --git a/fs/ocfs2/Kconfig b/fs/ocfs2/Kconfig
index 0d840669698e..ab152c00cd3a 100644
--- a/fs/ocfs2/Kconfig
+++ b/fs/ocfs2/Kconfig
@@ -51,7 +51,7 @@ config OCFS2_FS_USERSPACE_CLUSTER
config OCFS2_FS_STATS
bool "OCFS2 statistics"
- depends on OCFS2_FS
+ depends on OCFS2_FS && DEBUG_FS
default y
help
This option allows some fs statistics to be captured. Enabling
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index 592fae5007d1..e4984e259cb6 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -565,7 +565,6 @@ static inline int ocfs2_et_sanity_check(struct ocfs2_extent_tree *et)
return ret;
}
-static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc);
static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
struct ocfs2_extent_block *eb);
static void ocfs2_adjust_rightmost_records(handle_t *handle,
@@ -5858,6 +5857,7 @@ int ocfs2_truncate_log_append(struct ocfs2_super *osb,
ocfs2_journal_dirty(handle, tl_bh);
+ osb->truncated_clusters += num_clusters;
bail:
mlog_exit(status);
return status;
@@ -5929,6 +5929,8 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
i--;
}
+ osb->truncated_clusters = 0;
+
bail:
mlog_exit(status);
return status;
@@ -7139,64 +7141,6 @@ bail:
}
/*
- * Expects the inode to already be locked.
- */
-int ocfs2_prepare_truncate(struct ocfs2_super *osb,
- struct inode *inode,
- struct buffer_head *fe_bh,
- struct ocfs2_truncate_context **tc)
-{
- int status;
- unsigned int new_i_clusters;
- struct ocfs2_dinode *fe;
- struct ocfs2_extent_block *eb;
- struct buffer_head *last_eb_bh = NULL;
-
- mlog_entry_void();
-
- *tc = NULL;
-
- new_i_clusters = ocfs2_clusters_for_bytes(osb->sb,
- i_size_read(inode));
- fe = (struct ocfs2_dinode *) fe_bh->b_data;
-
- mlog(0, "fe->i_clusters = %u, new_i_clusters = %u, fe->i_size ="
- "%llu\n", le32_to_cpu(fe->i_clusters), new_i_clusters,
- (unsigned long long)le64_to_cpu(fe->i_size));
-
- *tc = kzalloc(sizeof(struct ocfs2_truncate_context), GFP_KERNEL);
- if (!(*tc)) {
- status = -ENOMEM;
- mlog_errno(status);
- goto bail;
- }
- ocfs2_init_dealloc_ctxt(&(*tc)->tc_dealloc);
-
- if (fe->id2.i_list.l_tree_depth) {
- status = ocfs2_read_extent_block(INODE_CACHE(inode),
- le64_to_cpu(fe->i_last_eb_blk),
- &last_eb_bh);
- if (status < 0) {
- mlog_errno(status);
- goto bail;
- }
- eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
- }
-
- (*tc)->tc_last_eb_bh = last_eb_bh;
-
- status = 0;
-bail:
- if (status < 0) {
- if (*tc)
- ocfs2_free_truncate_context(*tc);
- *tc = NULL;
- }
- mlog_exit_void();
- return status;
-}
-
-/*
* 'start' is inclusive, 'end' is not.
*/
int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
@@ -7270,18 +7214,3 @@ out_commit:
out:
return ret;
}
-
-static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc)
-{
- /*
- * The caller is responsible for completing deallocation
- * before freeing the context.
- */
- if (tc->tc_dealloc.c_first_suballocator != NULL)
- mlog(ML_NOTICE,
- "Truncate completion has non-empty dealloc context\n");
-
- brelse(tc->tc_last_eb_bh);
-
- kfree(tc);
-}
diff --git a/fs/ocfs2/alloc.h b/fs/ocfs2/alloc.h
index 55762b554b99..3bd08a03251c 100644
--- a/fs/ocfs2/alloc.h
+++ b/fs/ocfs2/alloc.h
@@ -228,10 +228,6 @@ struct ocfs2_truncate_context {
int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
u64 range_start, u64 range_end);
-int ocfs2_prepare_truncate(struct ocfs2_super *osb,
- struct inode *inode,
- struct buffer_head *fe_bh,
- struct ocfs2_truncate_context **tc);
int ocfs2_commit_truncate(struct ocfs2_super *osb,
struct inode *inode,
struct buffer_head *di_bh);
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c
index 0d7c5540ad66..1fbb0e20131b 100644
--- a/fs/ocfs2/aops.c
+++ b/fs/ocfs2/aops.c
@@ -1630,6 +1630,43 @@ static int ocfs2_zero_tail(struct inode *inode, struct buffer_head *di_bh,
return ret;
}
+/*
+ * Try to flush truncate logs if we can free enough clusters from it.
+ * As for return value, "< 0" means error, "0" no space and "1" means
+ * we have freed enough spaces and let the caller try to allocate again.
+ */
+static int ocfs2_try_to_free_truncate_log(struct ocfs2_super *osb,
+ unsigned int needed)
+{
+ tid_t target;
+ int ret = 0;
+ unsigned int truncated_clusters;
+
+ mutex_lock(&osb->osb_tl_inode->i_mutex);
+ truncated_clusters = osb->truncated_clusters;
+ mutex_unlock(&osb->osb_tl_inode->i_mutex);
+
+ /*
+ * Check whether we can succeed in allocating if we free
+ * the truncate log.
+ */
+ if (truncated_clusters < needed)
+ goto out;
+
+ ret = ocfs2_flush_truncate_log(osb);
+ if (ret) {
+ mlog_errno(ret);
+ goto out;
+ }
+
+ if (jbd2_journal_start_commit(osb->journal->j_journal, &target)) {
+ jbd2_log_wait_commit(osb->journal->j_journal, target);
+ ret = 1;
+ }
+out:
+ return ret;
+}
+
int ocfs2_write_begin_nolock(struct file *filp,
struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
@@ -1637,7 +1674,7 @@ int ocfs2_write_begin_nolock(struct file *filp,
struct buffer_head *di_bh, struct page *mmap_page)
{
int ret, cluster_of_pages, credits = OCFS2_INODE_UPDATE_CREDITS;
- unsigned int clusters_to_alloc, extents_to_split;
+ unsigned int clusters_to_alloc, extents_to_split, clusters_need = 0;
struct ocfs2_write_ctxt *wc;
struct inode *inode = mapping->host;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
@@ -1646,7 +1683,9 @@ int ocfs2_write_begin_nolock(struct file *filp,
struct ocfs2_alloc_context *meta_ac = NULL;
handle_t *handle;
struct ocfs2_extent_tree et;
+ int try_free = 1, ret1;
+try_again:
ret = ocfs2_alloc_write_ctxt(&wc, osb, pos, len, di_bh);
if (ret) {
mlog_errno(ret);
@@ -1681,6 +1720,7 @@ int ocfs2_write_begin_nolock(struct file *filp,
mlog_errno(ret);
goto out;
} else if (ret == 1) {
+ clusters_need = wc->w_clen;
ret = ocfs2_refcount_cow(inode, filp, di_bh,
wc->w_cpos, wc->w_clen, UINT_MAX);
if (ret) {
@@ -1695,6 +1735,7 @@ int ocfs2_write_begin_nolock(struct file *filp,
mlog_errno(ret);
goto out;
}
+ clusters_need += clusters_to_alloc;
di = (struct ocfs2_dinode *)wc->w_di_bh->b_data;
@@ -1817,6 +1858,22 @@ out:
ocfs2_free_alloc_context(data_ac);
if (meta_ac)
ocfs2_free_alloc_context(meta_ac);
+
+ if (ret == -ENOSPC && try_free) {
+ /*
+ * Try to free some truncate log so that we can have enough
+ * clusters to allocate.
+ */
+ try_free = 0;
+
+ ret1 = ocfs2_try_to_free_truncate_log(osb, clusters_need);
+ if (ret1 == 1)
+ goto try_again;
+
+ if (ret1 < 0)
+ mlog_errno(ret1);
+ }
+
return ret;
}
diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index 9e3d45bcb5fd..a6cc05302e9f 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;
@@ -132,6 +134,33 @@ 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 +215,9 @@ 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;
@@ -212,9 +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
@@ -701,6 +734,14 @@ 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,
@@ -1041,6 +1082,9 @@ 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
@@ -1090,6 +1134,9 @@ 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;
@@ -1142,6 +1189,12 @@ 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;
}
@@ -1315,6 +1368,8 @@ 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();
}
@@ -1384,6 +1439,7 @@ 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);
@@ -1948,6 +2004,18 @@ 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;
@@ -2002,15 +2070,20 @@ 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)
@@ -2028,7 +2101,27 @@ 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 {
@@ -2214,63 +2307,138 @@ 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,
@@ -2291,9 +2459,11 @@ 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);
@@ -2311,7 +2481,7 @@ 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;
}
@@ -2322,7 +2492,7 @@ 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? */
@@ -2330,7 +2500,7 @@ 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);
diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c
index a3f150e52b02..3a5835904b3d 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);
@@ -123,37 +128,42 @@ 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);
+ if (!nst)
+ goto out;
- 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: %lu.%ld\n"
- " send start: %lu.%ld\n"
- " wait start: %lu.%ld\n",
- nst, (unsigned long)nst->st_task->pid,
- (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);
- }
+ 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));
+
+ /* 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;
@@ -228,6 +238,11 @@ 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)
{
@@ -253,7 +268,8 @@ 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);
@@ -264,7 +280,8 @@ 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);
@@ -276,65 +293,107 @@ static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
return sc; /* unused, just needs to be null when done */
}
-#define TV_SEC_USEC(TV) TV.tv_sec, (long)TV.tv_usec
+#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)
+{
+ 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));
+}
+
+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;
+ }
+
+ /* 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);
+}
static int sc_seq_show(struct seq_file *seq, void *v)
{
- 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);
- if (sc != NULL) {
- struct inet_sock *inet = NULL;
-
- __be32 saddr = 0, daddr = 0;
- __be16 sport = 0, dport = 0;
-
- 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;
- }
-
- /* 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: %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"
- " 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,
- 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),
- 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;
@@ -351,7 +410,7 @@ 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;
@@ -369,7 +428,8 @@ 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;
@@ -382,12 +442,48 @@ 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);