glusterfs.git/xlators/storage, branch v3.0.3rc2

posix: gen number fixes on non-Linux systems

2010-03-03T14:40:05+00:00

Most non-Linux platforms do not support extended attributes on special
files and symlinks. Use the mtime as the generation number for these
files on such systems

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

BUG: 621 (3.0.2 GlusterFS fails on Solaris 10)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=621

storage/posix: Don't use FTW_CONTINUE

2010-02-24T05:37:16+00:00

FTW_CONTINUE is not defined on Mac, and POSIX only specifies
that the walker function must return 0. So just return 0.

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

BUG: 625 (Gluster Fails to build on OS X 10.6.2)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=625

storage/posix: Check for path before logging it in case of an error in xattrop FOP.

2010-02-22T11:40:08+00:00

Signed-off-by: Pavan Vilas Sondur 
Signed-off-by: Anand V. Avati 

BUG: 614 (GlusterFS crashes while printing error if fsetxattr call fails in posix)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=614

storage/posix: Fix device number handling.

2010-02-08T07:27:52+00:00

There are two fixes in this patch:

1) If the device number has changed, do a fresh stat on the export
   directory and if it matches the device number of the file,
   assume a remount happened and remember the new device number
   as official (this helps automounted export volumes).

2) Don't log the "device number changed" message if it is due
   to a stat on ".." (in posix_readdirp), as it could be because
   ".." leads us out of the exported volume.

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

BUG: 576 (crossing device (2056) + fuse LOOKUP error)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=576

Revert "Server backend storage hang should not cause the mount point to hang."

2010-01-26T15:23:19+00:00

This reverts commit a0b148ea4e2a0163548eeb89b7580be4adbb8070.

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

BUG: 272 (Server backend storage hang should not cause the mount point to hang)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=272

storage/posix: Fix an access-after-free bug in release() and releasedir().

2010-01-25T12:39:09+00:00

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

BUG: 533 (close() should not block main loop)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=533

Server backend storage hang should not cause the mount point to hang.

2010-01-23T10:58:34+00:00

Submitted-by: Krishna Srinivas 
NOTE: fixed compilation issues in posix.c introduced while merging

storage/posix polls for FS/kernel being functional by issuing
statvfs() call. In case statvfs expires the timer, storage/posix will
send CHILD_DOWN to upper translator. Ultimately this will cause
protocol/server to disconnect all clients connected and also cleans up
the data structures. Hence if soft lockup or other kernel bug causes
backend FS to hang, the clients will not be hung.

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

BUG: 272 (Server backend storage hang should not cause the mount point to hang)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=272

storage/posix: Make the janitor sleep duration configurable.

2010-01-21T12:25:45+00:00

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

BUG: 533 (close() should not block main loop)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=533

storage/posix: Hand off calling close() and closedir() to the janitor thread.

2010-01-19T13:56:35+00:00

Let the janitor thread handle calling close() and closedir() on fd's and DIR *
when posix_release is called. This avoids blocking the main server loop
due to close().

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

BUG: 533 (close() should not block main loop)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=533

storage/posix: Zero out the checksum array in rchecksum.

2009-12-28T12:39:27+00:00

Set the checksum bytes to zero to ensure random junk
is not returned as the checksum.

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