glusterfs.git, branch v3.3.0qa16

mount/fuse: export PATH for which in mount script

2011-12-13T17:43:36+00:00

exporting PATH environment variable for mount.glusterfs.in
to correct the "which: no getfattr in ((null))" error
during fuse mount.

Change-Id: Id7d024c0d1cf3d265489557897e9e1f8e7ce4ce4
BUG: 765561
Signed-off-by: Rajesh Amaravathi 
Reviewed-on: http://review.gluster.com/782
Tested-by: Gluster Build System 
Reviewed-by: Jeff Darcy 
Reviewed-by: Amar Tumballi 
Reviewed-by: Vijay Bellur

cluster/afr: Double the call count if transaction is for rename

2011-12-13T11:40:09+00:00

In rename the changelog modification needs to happen both on
old parent-dir and new parent-dir, so 2 stack winds are
done per brick.

Change-Id: I43f34661e397c4288162213944529e18b7724b1d
BUG: 766603
Signed-off-by: Pranith Kumar K 
Reviewed-on: http://review.gluster.com/783
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

fuse: bring in the reverse invalidation

2011-12-12T10:15:12+00:00

Thanks to Csaba Henk  for the patch

Currently one can invalidate the inodes using 'setxattr()' with
key 'inode-invalidate' (and any value). This can be further
extended to do a purge of inode table itself.

Change-Id: I165d5d585ed808b9e463ac0aad859ec64568c7a2
BUG: 762277
Signed-off-by: Amar Tumballi 
Reviewed-on: http://review.gluster.com/324
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

iobuf: fix a crash in iobuf when statedump is taken

2011-12-08T06:27:10+00:00

With the previous patch this change was missed.

Change-Id: If536cef3fa423415eaa4104e6c3e5e72c5d0a22d
BUG: 3854
Signed-off-by: Raghavendra Bhat 
Reviewed-on: http://review.gluster.com/775
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

libglusterfs/iobuf: have fixed number of arenas

2011-12-07T13:01:52+00:00

* so overall memory usage will be in limit.
* the array is hard-coded, need to improve upon this.
* need more benchmarking to tune the proper values to the array
* fixed the issue of pruning of arenas.

Change-Id: I38a8ffab37378c25d78f77a2d412b1b8935c67d3
BUG: 3474
Signed-off-by: Amar Tumballi 
Reviewed-on: http://review.gluster.com/543
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

cluster/distribute: Assert checks at known locations

2011-12-06T10:02:02+00:00

and new function dict_get_ptr_and_len().

Change-Id: I653a1cc8123baa36d750250d02721aa98b196f38
BUG: 3158
Signed-off-by: Harshavardhana 
Reviewed-on: http://review.gluster.com/744
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

CONTRIBUTING: add a DCO

2011-12-05T11:25:29+00:00

The new CONTRIBUTING file has a Developer Certificate of Origin with
instructions for contributors to ensure 'Signed-off-by:' line in the
commit log (git commit -s). Corresponding changes also done in Gerrit
to reject patches which do not have Signed-off-by: line.

Change-Id: Ia3e1e8d3cfd4b32e4cfd4d2df91c6dbd57e2f60e
BUG: 3234
Signed-off-by: Anand Avati 
Reviewed-on: http://review.gluster.com/767
Tested-by: Gluster Build System

Changes all parent values for quota_check_limit

2011-12-05T10:47:32+00:00

and quota_update_size from 0x0 to NULL to make
sure uuid_copy happens between pointers

Change-Id: I73ef5f9054f972fce00b8f42125dc49c9d86c3f2
BUG: 3830
Reviewed-on: http://review.gluster.com/756
Tested-by: Gluster Build System 
Reviewed-by: Mohammed Junaid 
Reviewed-by: Amar Tumballi

bz 3826, fix for parallel make in fedora build system

2011-12-02T07:19:32+00:00

builds of glusterfs in the fedora build system often fail due to a
race condition between running yacc and starting to compile the y.tab.c
produced by yacc

We found that the previous fix would still occasionally trip the race
condition. This revised patch changes the automake Makefile.am to generate
the parser files without incurring the race condition. An extra dimension
of the problem is that the tarball from
  http://download.gluster.com/pub/gluster/glusterfs/3.2/3.2.5/glusterfs-3.2.5.tar.gz
contains files that you don't get when you clone from the git repo (e.g.
libgluster/src/{graph.lex.c,y.tab.c}, and all the Makefile.in files) so
build issues on fedora build systems do not manifest themselves on jenkins
and vice versa.

This works on jenkins, the fedora build system, and my f16 vm/guest
machines. (Finding the right combination that works on all three was an
exercise to say the least. I'm open to other suggestions for avoiding
the race condition.)

Run autogen.sh to (re)generate the Makefile.ins. Then run configure to
produce all Makefiles, followed by `make -j X` where X>1

see also https://bugzilla.redhat.com/show_bug.cgi?id=756510

BUG: 3826
Change-Id: Iaeecb59c61a77bf3927da18253c83cf5ffed4254
Reviewed-on: http://review.gluster.com/765
Tested-by: Gluster Build System 
Reviewed-by: Jeff Darcy 
Reviewed-by: Anand Avati

mount/fuse: Inherit direct-io-mode values from fds already

2011-12-02T06:02:39+00:00

opened, for a new fd being opened.

When an fd is being opened, it inherits direct-io-mode characterstics
(either enabled or disabled) from the fds already opened on inode. If none
are opened and user has not specified the mode through cmdline options,
default mode is used.

Change-Id: I0c9e959100e9130e46bbd16d63eca278260635b4
BUG: 801
Reviewed-on: http://review.gluster.com/55
Tested-by: Gluster Build System 
Reviewed-by: Csaba Henk 
Reviewed-by: Anand Avati