[Gluster-devel] tests/basic/ec/ec-3-1.t generated core

[Xavier Hernandez]

Is it possible that volume mount returns before fuse_init() is executed 
? if that's true, then the core is generated because just after mounting 
the volume, statedumps are requested to determine when all ec childs are 
up. The code in fuse's dump assumes that fuse_init() has already been 
called when a statedump is generated.

#0  0x00007f75dae83137 in fuse_itable_dump (this=0x2079be0) at 
/home/jenkins/root/workspace/rackspace-regression-2GB-triggered/xlators/mount/fuse/src/fuse-bridge.c:4988
4988            inode_table_dump(priv->active_subvol->itable,
(gdb) print priv->init_recvd
$14 = 0 '\000'

Xavi

On 14/01/16 08:33, Xavier Hernandez wrote:
> The failure happens when a statedump is generated. For some reason
> priv->active_subvol is NULL, causing a segmentation fault:
>
> (gdb) t 1
> [Switching to thread 1 (LWP 4179)]
> #0  0x00007f75dae83137 in fuse_itable_dump (this=0x2079be0) at
> /home/jenkins/root/workspace/rackspace-regression-2GB-triggered/xlators/mount/fuse/src/fuse-bridge.c:4988
>
> 4988            inode_table_dump(priv->active_subvol->itable,
> (gdb) bt
> #0  0x00007f75dae83137 in fuse_itable_dump (this=0x2079be0) at
> /home/jenkins/root/workspace/rackspace-regression-2GB-triggered/xlators/mount/fuse/src/fuse-bridge.c:4988
>
> #1  0x00007f75e30f8a11 in gf_proc_dump_xlator_info (top=0x2079be0) at
> /home/jenkins/root/workspace/rackspace-regression-2GB-triggered/libglusterfs/src/statedump.c:506
>
> #2  0x00007f75e30f96e9 in gf_proc_dump_info (signum=10, ctx=0x2055010)
> at
> /home/jenkins/root/workspace/rackspace-regression-2GB-triggered/libglusterfs/src/statedump.c:832
>
> #3  0x0000000000409894 in glusterfs_sigwaiter (arg=0x7ffceb7dba50) at
> /home/jenkins/root/workspace/rackspace-regression-2GB-triggered/glusterfsd/src/glusterfsd.c:2015
>
> #4  0x00007f75e23a6a51 in start_thread () from ./lib64/libpthread.so.0
> #5  0x00007f75e1d1093d in clone () from ./lib64/libc.so.6
> (gdb) list
> 4983                     return -1;
> 4984
> 4985            priv = this->private;
> 4986
> 4987            gf_proc_dump_add_section("xlator.mount.fuse.itable");
> 4988            inode_table_dump(priv->active_subvol->itable,
> 4989                             "xlator.mount.fuse.itable");
> 4990
> 4991            return 0;
> 4992    }
> (gdb) print priv->active_subvol
> $5 = (xlator_t *) 0x0
>
> Does this sound familiar to anyone ?
>
> Xavi
>
> On 14/01/16 08:08, Xavier Hernandez wrote:
>> I'm looking it.
>>
>> On 14/01/16 08:03, Atin Mukherjee wrote:
>>> [1] has caused a regression failure with a core from the mentioned test.
>>> Mind having a look?
>>>
>>> [1]
>>> https://build.gluster.org/job/rackspace-regression-2GB-triggered/17579/consoleFull
>>>
>>>
>>>
>>> Thanks,
>>> Atin
>>>
>> _______________________________________________
>> Gluster-devel mailing list
>> Gluster-devel at gluster.org
>> http://www.gluster.org/mailman/listinfo/gluster-devel
> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-devel