Commit graph

30689 commits

Author SHA1 Message Date
Kurt Hackel
0032abd674 ocfs2: remove whitespace in dlmunlock.c
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:19 -07:00
Kurt Hackel
3156d26701 ocfs2: move dlm work to a private work queue
The work that is done can block for long periods of time and so is not
appropriate for keventd.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:18 -07:00
Kurt Hackel
495ac96e63 ocfs2: fix incorrect error returns
Use DLM_REJECTED instead of DLM_RECOVERING.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:17 -07:00
Kurt Hackel
3b3b84a892 ocfs2: tune down some noisy messages during dlm recovery
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:16 -07:00
Kurt Hackel
56a7c104bc ocfs2: display message before waiting for recovery to complete
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:15 -07:00
Kurt Hackel
44a7f1d063 ocfs2: mlog in dlm_convert_lock_handler() should be ML_ERROR
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:15 -07:00
Kurt Hackel
b220532a71 ocfs2: retry operations when a lock is marked in recovery
Before checking for a nonexistent lock, make sure the lockres is not marked
RECOVERING. The caller will just retry and the state should be fixed up when
recovery completes.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:14 -07:00
Kurt Hackel
f85cd47a58 ocfs2: use cond_resched() in dlm_thread()
yield() does not yield.  cond_resched() does.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:13 -07:00
Kurt Hackel
ad8100e0d2 ocfs2: use GFP_NOFS in some dlm operations
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:12 -07:00
Kurt Hackel
b7084ab538 ocfs2: wait for recovery when starting lock mastery
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:12 -07:00
Kurt Hackel
c27069e6cf ocfs2: continue recovery when a dead node is encountered
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:11 -07:00
Kurt Hackel
67a187412b ocfs2: remove unneccesary spin_unlock() in dlm_remaster_locks()
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:10 -07:00
Kurt Hackel
6a41321121 ocfs2: dlm_remaster_locks() should never exit without completing
We cannot restart recovery. Once we begin to recover a node, keep the state
of the recovery intact and follow through, regardless of any other node
deaths that may occur.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:09 -07:00
Kurt Hackel
c8df412e1c ocfs2: special case recovery lock in dlmlock_remote()
If the previous master of the recovery lock dies, let calc_usage take it
down completely and let the caller completely redo the dlmlock() call.
Otherwise, there will never be an opportunity to re-master the lockres and
recovery wont be able to progress.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:08 -07:00
Kurt Hackel
36407488b1 ocfs2: pending mastery asserts and migrations should block each other
Use the existing structure for blocking migrations when ASTs are pending to
achieve the same result. If we can catch the assert before it goes on the
wire, just cancel it and let the migration continue.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:08 -07:00
Kurt Hackel
c87a9ae705 ocfs2: temporarily disable automatic lock migration
Now we never change the owner of a lock resource until unmount or node
death. This will be re-enabled once some issues in the algorithm used have
been resolved.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:07 -07:00
Kurt Hackel
2abaf97e62 ocfs2: do not unconditionally purge the lockres in dlmlock_remote()
In dlmlock_remote(), do not call purge_lockres until the lock resource
actually changes. otherwise, the mastery info on the lockres will go away
underneath the caller.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:06 -07:00
Kurt Hackel
aa087b8497 ocfs2: increase backoff before waiting for recovery
When mastering non-recovery lock resources, additional time was frequently
needed to allow the disk heartbeat to catch up with the network timeout. the
recovery lock resource is time critical and avoids this path.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:05 -07:00
Kurt Hackel
f42a100b22 ocfs2: have dlm_pre_master_reco_lockres() ignore dead nodes
Recovery will spin in dlm_pre_master_reco_lockres if we do not ignore
timed-out network responses from dead nodes.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:05 -07:00
Kurt Hackel
6ff06a9391 ocfs2: give the dlm dirty list a reference on the lockres
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:04 -07:00
Kurt Hackel
e7e69eb389 ocfs2: teach dlm_restart_lock_mastery() to wait on recovery
Change behavior of dlm_restart_lock_mastery() when a node goes down.  Dump
all responses that have been collected and start over.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:03 -07:00
Kurt Hackel
e4eb03681a ocfs2: gracefully handle stale create_lock messages.
This is an error on the sending side, so gracefully error out on the
receiving end.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:02 -07:00
Kurt Hackel
ccd8b1f916 ocfs2: update lvb immediately during recovery
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:02 -07:00
Kurt Hackel
588e00902b ocfs2: do not send master requests to localhost
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:01 -07:00
Kurt Hackel
8b2198097a ocfs2: purge lockres' sooner
Immediately purge a lockress that the local node is not the master of.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:43:00 -07:00
Kurt Hackel
343e26a400 ocfs2: dump mismatching migrated lvbs before BUG()
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:59 -07:00
Kurt Hackel
466d1a4591 ocfs2: make dlm recovery finalization 2 stage
Makes it easier for the recovery process to deal with node death.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:58 -07:00
Kurt Hackel
69d72b066c ocfs2: dlm recovery / lockres reference count fix
Take a reference on lockres structures while they are on the recovery list.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:58 -07:00
Kurt Hackel
a9ee4c8a67 ocfs2: better error handling during assert master message
handle errors during lock assert master by either killing self or other node

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:57 -07:00
Kurt Hackel
a7f90d83ea ocfs2: dump lockres info before we BUG() on a bad reference
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:56 -07:00
Mark Fasheh
c0a8520c73 ocfs2: do LVB puts in place
Don't wait until the AST will be fired to do the LVB copy into the lock
resource.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:55 -07:00
Kurt Hackel
aa85235427 ocfs2: mle ref count debugging
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:55 -07:00
Kurt Hackel
dc2ed195dd ocfs2: allow for an assert message during lock mastery
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:54 -07:00
Kurt Hackel
2d1a868c56 ocfs2: take mle reference during migration
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:53 -07:00
Kurt Hackel
41b8c8a101 ocfs2: properly initialize the mle structure
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:52 -07:00
Kurt Hackel
da01ad0552 ocfs2: detach mle from heartbeat events
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:52 -07:00
Kurt Hackel
a2bf04774b ocfs2: mle ref counting fixes
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:51 -07:00
Kurt Hackel
958837197e ocfs2: better mle debugging
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:50 -07:00
Kurt Hackel
d6dea6e973 ocfs2: clean up recovery related messages
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:49 -07:00
Kurt Hackel
29c0fa0f56 ocfs2: handle network errors during recovery
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:49 -07:00
Kurt Hackel
c3187ce5e3 ocfs2: only recover one dead node at a time
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:48 -07:00
Kurt Hackel
ab27eb6f47 ocfs2: Better tracking for recovery state changes
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:47 -07:00
Kurt Hackel
8bc674cb48 ocfs2: Fix empty lvb check
The check for an empty lvb should check the entire buffer not just the first
byte.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:46 -07:00
Kurt Hackel
aba9aac788 ocfs2: fix inverted logic in dlm_is_node_dead
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:45 -07:00
Kurt Hackel
2580a580e0 ocfs2: recheck lockres master before sending an unlock request.
Recovery may have happened and it may now be mastered locally.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:45 -07:00
Kurt Hackel
8d79d088e8 ocfs2: add a small delay after a failed migration
Otherwise we risk starving other threads.

Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:44 -07:00
Mark Fasheh
685f1adb38 ocfs2: silence a compile warning in dlm_alloc_pagevec()
Reported by Andrew Morton.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:43 -07:00
Joel Becker
c8f33b6e86 [PATCH] ocfs2: Alloc at least a page for the DLM hash
The OCFS2 DLM allocates a number of pages for a hash to lookup locks.
There was a bug where a PAGE_SIZE bigger than the hash size (eg, 64K
pages) would result in zero pages allocated.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:42 -07:00
Daniel Phillips
03d864c02c ocfs2: allocate lockres hash pages in an array
This allows us to have a hash table greater than a single page which greatly
improves dlm performance on some tests.

Signed-off-by: Daniel Phillips <phillips@google.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:42 -07:00
Mark Fasheh
95c4f581d6 ocfs2: inline dlm_lockres_get()
It's called on every lookup so this might help performance a bit.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
2006-06-26 14:42:41 -07:00