glusterfs.git/xlators/cluster, branch v3.0.2rc1

dht: initialize layout type

2010-01-23T19:07:05+00:00

Instead of relying on CALLOC's zero initialization which happens
to be the value of DHT_HASH_TYPE_DM, explicitly set the value

Signed-off-by: Anand V. Avati 
Signed-off-by: Anand V. Avati 

BUG: 523 (Hash type is not initialized in dht_layout_new)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=523

cluster/afr: Send the struct flock returned by the server to the user.

2010-01-21T14:40:59+00:00

Signed-off-by: Vikas Gorur 
Signed-off-by: Anand V. Avati 

BUG: 521 (SPECFS validation fails over distribute + replicate)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=521

stripe entry self heal

2010-01-19T13:26:02+00:00

Create the entry (either regular file or directory), when its
missing in some of the subvolumes and available on some.

Signed-off-by: Amar Tumballi 
Signed-off-by: Anand V. Avati 

BUG: 535 (stripe entry self-heal..)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=535

cluster/afr: Pick a source for metadata self-heal even if all nodes are innocent.

2010-01-15T07:23:49+00:00

If metadata changelog has been disabled, all subvolumes
will be innocent. In that case, simply pick the subvolume
on which the file has the lowest uid as the source and
sync other subvolumes to it.

Signed-off-by: Vikas Gorur 
Signed-off-by: Anand V. Avati 

BUG: 451 (metadata self-heal does not a pick a source if mode/times have been changed at the backend)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=451

cluster/unify: don't call xl_init_tree on namespace node.

2010-01-15T07:23:42+00:00

- just before xl_init_tree call on namespace, unify node is set as parent of
    namespace node and since xl->ready of unify node is set only after return
    from init call, calling xl_init_tree on namespace node will result in
    execution of init of unify node for the second time. This results in
    following bugs:

    * since during second execution, parent of namespace node is unify, init of
      unify xlator fails.
    * even if init did not fail, re-execution of init in itself is bad, since it
      results in memory leaks etc. Worse, it can result in infinite recursive
      calls into init of unify node.

Signed-off-by: Raghavendra G 
Signed-off-by: Anand V. Avati 

BUG: 518 (unify does not start)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=518

cluster/afr: Use dict_ref instead of dict_copy_with_ref.

2010-01-08T11:51:18+00:00

Signed-off-by: Vikas Gorur 
Signed-off-by: Anand V. Avati 

BUG: 509 (Crash in afr_local_cleanup ())
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=509

cluster/afr: Allocate diff algorithm loop_state structures only once.

2009-12-28T12:39:24+00:00

Instead of CALLOC'ing a loop_state structure at the beginning
of every loop, keep a table of allocated structures and reuse
them.

Signed-off-by: Vikas Gorur 
Signed-off-by: Anand V. Avati 

BUG: 320 (Improve self-heal performance)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=320

cluster/afr: Report number of blocks healed during diff self-heal.

2009-12-28T12:39:20+00:00

Signed-off-by: Vikas Gorur 
Signed-off-by: Anand V. Avati 

BUG: 320 (Improve self-heal performance)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=320

properly ref/unref fd in stripe_opendir_cbk

2009-12-22T10:42:33+00:00

fixes the crash due to dereferencing NULL fd pointer

Signed-off-by: Amar Tumballi 
Signed-off-by: Vijay Bellur 

BUG: 497 (Client crashes in fuse_fd_cbk while running ltp tests)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=497

distribute: Dont dereference sbuf on failed readlink_cbk

2009-12-18T16:41:33+00:00

This fixes a crash that was caused due to a NULL pointer deref
because a NULL sbuf is returned if the readlink fop fails.

Signed-off-by: Shehjar Tikoo 
Signed-off-by: Vijay Bellur 

BUG: 492 (distribute crashes in readlink)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=492