xfstest generic/089 failed on NFSv2

xfstest generic/089 failed on NFSv2

[zhangxiaoxu (A)]

Hi All,
I run the xfstest on nfsv2, the case generic/089 will failed with timeout sometimes.

The top of the stack always on 'nlmclnt_block' function.
And the 'nlmclnt_lock' function will retry every 30s.
So, the test case is killed by the timer and failed.

We open the nlm_debug, found that:
The client send unlock message to server, and the server handle it correct.
But no 'NLM_LCK_GRANTED' message send to the client.

The configuration about the xfstest and nfs server:

The local.config of the xfstest is:
export FSTYP=nfs
export TEST_DEV=192.168.240.61:/export/ext4/test
export TEST_DIR=/nfs/test
export SCRATCH_DEV=192.168.240.61:/export/ext4/scratch
export SCRATCH_MNT=/nfs/scratch
export NFS_MOUNT_OPTIONS="-o acl,rw,relatime,vers=2,lock"

The nfs server just export the share directory.

I don't know why this case is skip from https://wiki.linux-nfs.org/wiki/index.php/Xfstests
If any problem about the fcntl(F_SETLKW) wait so long time?

Re: xfstest generic/089 failed on NFSv2

[J. Bruce Fields]

On Mon, Dec 10, 2018 at 01:21:43AM +0800, zhangxiaoxu (A) wrote:
> I run the xfstest on nfsv2, the case generic/089 will failed with timeout sometimes.
> 
> The top of the stack always on 'nlmclnt_block' function.
> And the 'nlmclnt_lock' function will retry every 30s.
> So, the test case is killed by the timer and failed.
> 
> We open the nlm_debug, found that:
> The client send unlock message to server, and the server handle it correct.
> But no 'NLM_LCK_GRANTED' message send to the client.
> 
> The configuration about the xfstest and nfs server:
> 
> The local.config of the xfstest is:
> export FSTYP=nfs
> export TEST_DEV=192.168.240.61:/export/ext4/test
> export TEST_DIR=/nfs/test
> export SCRATCH_DEV=192.168.240.61:/export/ext4/scratch
> export SCRATCH_MNT=/nfs/scratch
> export NFS_MOUNT_OPTIONS="-o acl,rw,relatime,vers=2,lock"
> 
> The nfs server just export the share directory.
> 
> I don't know why this case is skip from https://wiki.linux-nfs.org/wiki/index.php/Xfstests
> If any problem about the fcntl(F_SETLKW) wait so long time?

I've seen intermittent generic/089 failures as well and haven't had the
time to track them down.  From your description it does sound like a
potentialy server bug.

--b.