search for: dlm_lock_resourc

backgroud: ocfs2 dlm uses a hash table to store dlm_lock_resource objects. the often used lookup is performed on the hash table. problem: for usages that there are huge number of inodes(thus huge number of dlm_lock_resource objects) in a ocfs2 volume, the lookup performance becomes a problem. the lookup holds spin_lock which could put all others cpus into the...

...uct dlm_lock, list); + list_for_each_entry(lock, queue, list) { ++count; if (lock->ml.node == dlm->node_num) { mlog(0, "found a lock owned by this node still " @@ -2923,18 +2912,16 @@ again: static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { - struct list_head *iter, *iter2; struct list_head *queue = &res->granted; int i, bit; - struct dlm_lock *lock; + struct dlm_lock *lock, *next; assert_spin_locked(&res->spinlock); BUG_ON(res->owner == dlm->node_num); for (i=0; i<3; i++) { - list_for...

Wether the dlm_lock needs to access lvb or not depends on dlm_lock_resource it belongs to. So a new parameter "struct dlm_lock_resource *res" is added to dlm_new_lock() so that we can know if we need to allocate lvb for the dlm_lock. And we have to make the lockres availale for calling dlm_new_lock(). Signed-off-by: Wengang Wang <wen.gang.wang at oracle.com&...

The changes from the last drop are: 1. Patch 11 removes struct dlm_lock_name. 2. Patch 12 is an unrelated bugfix. Actually is related to a bugfix that we are retracting in mainline currently. The patch may need more testing. While I did hit the condition in my testing, Marcos hasn't. I am sending it because it can be queued for 2.6.30. Give us more time to test. 3. Patch 13 will be useful

These patches convert the mle list to a hash. The same patches apply on ocfs2 1.4 too. Currently, we use the same number of hash pages for mles and lockres'. This will be addressed in a future patch that will make both of them configurable. Sunil

...tx q = &res->lockname; q->hash = full_name_hash(q->name, q->len); - bucket = &(dlm->lockres_hash[q->hash % DLM_HASH_BUCKETS]); + bucket = dlm_lockres_hash(dlm, q->hash); /* get a reference for our hashtable */ dlm_lockres_get(res); @@ -151,7 +179,7 @@ struct dlm_lock_resource * __dlm_lookup_ hash = full_name_hash(name, len); - bucket = &(dlm->lockres_hash[hash % DLM_HASH_BUCKETS]); + bucket = dlm_lockres_hash(dlm, hash); /* check for pre-existing lock */ hlist_for_each(iter, bucket) { @@ -394,7 +422,7 @@ static int dlm_migrate_all_locks(struct for...

Please review the list of patches being applied to the ocfs2 1.4 tree. All patches list the mainline commit hash. Thanks Sunil

...8 @@ #define MLOG_MASK_PREFIX ML_DLM #include "cluster/masklog.h" -int stringify_lockname(const char *lockname, int locklen, char *buf, int len); +static int stringify_lockname(const char *lockname, int locklen, char *buf, + int len); void dlm_print_one_lock_resource(struct dlm_lock_resource *res) { @@ -251,7 +252,8 @@ EXPORT_SYMBOL_GPL(dlm_errname); * * For more on lockname formats, please refer to dlmglue.c and ocfs2_lockid.h. */ -int stringify_lockname(const char *lockname, int locklen, char *buf, int len) +static int stringify_lockname(const char *lockname, int locklen, cha...

Found 15 patches (out of 162) that appeared relevant to ocfs2 1.4. Please review. Sunil

...nds, resend msg works well. We are forward to appreciate another better way to avoid it. Thanks. --- ocfs2/dlm/dlmthread.c 2014-06-07 10:40:09.000000000 +0800 +++ ocfs2/dlm/dlmthread.c 2014-09-26 16:42:36.000000000 +0800 @@ -517,6 +517,9 @@ static void dlm_flush_asts(struct dlm_ct struct dlm_lock_resource *res; u8 hi; + /* resend the msg again */ + int send_times = 0; + spin_lock(&dlm->ast_lock); while (!list_empty(&dlm->pending_asts)) { lock = list_entry(dlm->pending_asts.next, @@ -539,9 +542,16 @@ static void dlm_flush_asts(struct dl...

...LOCK_RES_DROPPING_REF; spin_unlock(&res->spinlock); wake_up(&res->wq); - } + } else + spin_unlock(&res->spinlock); return 0; } @@ -251,17 +246,7 @@ static void dlm_run_purge_list(struct dlm_ctxt *dlm, lockres = list_entry(dlm->purge_list.next, struct dlm_lock_resource, purge); - /* Status of the lockres *might* change so double - * check. If the lockres is unused, holding the dlm - * spinlock will prevent people from getting and more - * refs on it -- there's no need to keep the lockres - * spinlock. */ spin_lock(&lockres->spinlock); -...

...tx q = &res->lockname; q->hash = full_name_hash(q->name, q->len); - bucket = &(dlm->lockres_hash[q->hash % DLM_HASH_BUCKETS]); + bucket = dlm_lockres_hash(dlm, q->hash); /* get a reference for our hashtable */ dlm_lockres_get(res); @@ -151,7 +178,7 @@ struct dlm_lock_resource * __dlm_lookup_ hash = full_name_hash(name, len); - bucket = &(dlm->lockres_hash[hash % DLM_HASH_BUCKETS]); + bucket = dlm_lockres_hash(dlm, hash); /* check for pre-existing lock */ hlist_for_each(iter, bucket) { @@ -262,7 +289,7 @@ static void dlm_free_ctxt_mem(struct dlm dlm...

Hi, The following patches comprise the bulk of Ocfs2 updates for the 2.6.30 merge window. Aside from larger, more involved fixes, we're adding the following features, which I will describe in the order their patches are mailed. Sunil's exported some more state to our debugfs files, and consolidated some other aspects of our debugfs infrastructure. This will further aid us in debugging

The following series of patches comprises the bulk of our outstanding changes for Ocfs2. Aside from the usual set of cleanups and fixes that were inappropriate for 2.6.25, there are a few highlights: The '/sys/o2cb' directory has been moved to '/sys/fs/o2cb'. The new location meshes better with modern sysfs layout. A symbolic link has been placed in the old location so as to