How to run xfstests against multipath disks on SuSE?

[Rich Johnston]

On 11/28/2012 08:10 AM, Ryan Wang wrote:
> 2012/11/28 Rich Johnston <rjohnston at sgi.com>:
>> On 11/27/2012 02:37 PM, Eric Sandeen wrote:
>>>
>>> On 11/27/12 2:35 PM, Dave Chinner wrote:
>>>>
>>>> On Tue, Nov 27, 2012 at 11:08:30AM -0600, Eric Sandeen wrote:
>>>>>
>>>>> On 11/26/12 9:14 PM, Dave Chinner wrote:
>>>>>>
>>>>>> On Tue, Nov 27, 2012 at 09:20:47AM +0800, Ryan Wang wrote:
>>>>>>>
>>>>>>> 2012/11/26 Theodore Ts'o <tytso at mit.edu>:
>>>>>>>>
>>>>>>>> On Mon, Nov 26, 2012 at 02:43:40PM +0100, Lukáš Czerner wrote:
>>>>>>>>>>
>>>>>>>>>> The I checked the script common.rc, and found that it uses
>>>>>>>>>> df to determine the filesystem type and it uses the realpath
>>>>>>>>>> "/dev/dm-0" instead of "/dev/mapper/mpatha", while df always
>>>>>>>>>> shows "devtmpfs" instead of "ext4" for /dev/dm-0.
>>>>>>>>>>
>>>>>>>>>> I run "mount" and found that devtmpfs is mount on /dev/ on opensuse
>>>>>>>>>> ***by default***. Even though  I run "mount -t btrfs /dev/dm-0
>>>>>>>>>> /somedir",
>>>>>>>>>> df still showed devtmpfs.
>>>>>>>>>>
>>>>>>>>>> Any ideas to kick off the xfstests?
>>>>>>>>
>>>>>>>>
>>>>>>>> At least with the df which ships with Debian Testing, df will display
>>>>>>>> the file system in use if it matches the device name which is
>>>>>>>> /etc/mtab.  It looks the issue is that mount is canonicalizing the
>>>>>>>> device name to be the human friendly version:
>>>>>>>>
>>>>>>>> # ls -l /dev/mapper/closure-bigscratch
>>>>>>>> 0 lrwxrwxrwx 1 root root 7 Nov 16 15:15
>>>>>>>> /dev/mapper/closure-bigscratch -> ../dm-7
>>>>>>>> <tytso.root at closure> {/home/tytso/bin}, level 2
>>>>>>>> # mount /dev/dm-7 /u2
>>>>>>>> # df /u2
>>>>>>>> Filesystem                     1K-blocks   Used Available Use%
>>>>>>>> Mounted on
>>>>>>>> /dev/mapper/closure-bigscratch  52403200 224772  52178428   1% /u2
>>>>>>>> # grep /u2 /etc/mtab
>>>>>>>> /dev/mapper/closure-bigscratch /u2 xfs
>>>>>>>> rw,relatime,attr2,inode64,noquota 0 0
>>>>>>>>
>>>>>>>> So the simplest solution is to just use /dev/mapper/mpatha instead of
>>>>>>>> insisting on using /dev/dm-0.  That's what I do....
>>>>>>>
>>>>>>>
>>>>>>> In xfstests/common.rc, it will use realpath to obtain /dev/dm-0, even
>>>>>>> if you
>>>>>>> specified /dev/mapper/mpatha.
>>>>>>> My problem is in SuSE, the devtmpfs is mounted on /dev by default,
>>>>>>> seems
>>>>>>> used by udev. And df reports devtmpfs on /dev/dm-0, instead of ext4.
>>>>>>
>>>>>>
>>>>>> What version of xfstests are you using? Do you have a
>>>>>> "README.device-mapper" file in the base directory? If not, upgrade
>>>>>> to the latest version and retry.
>>>>>>
>>>>>> $ gl -n 1 -p 0a81886
>>>>>> commit 0a818862bb74932e15971ec8bedb55765ec3a3f6
>>>>>> Author: Christoph Hellwig <hch at lst.de>
>>>>>> Date:   Wed May 4 16:28:32 2011 +0200
>>>>>>
>>>>>>       xfstests: support post-udev device mapper nodes
>>>>>>
>>>>>>       Because of udevs complaining device mapper now creates /dev/dm-N
>>>>>> as the real
>>>>>>       device nodes, and just symlinks the /dev/mapper/ names to it.
>>>>>> This would be
>>>>>>       easy if everything used the /dev/mapper clear names, but most
>>>>>> system utilities
>>>>>>       translate them back to the /dev/mapper/ names and thus confuse
>>>>>> various test
>>>>>>       cases.  Add support to _is_block_dev to read symlinks, and add
>>>>>> documentation
>>>>>>       on how to run xfstests on device mapper volumes.
>>>>>>
>>>>>>       Signed-off-by: Christoph Hellwig <hch at lst.de>
>>>>>>       Reviewed-by: Alex Elder <aelder at sgi.com>
>>>>>>
>>>>>> Cheers,
>>>>>>
>>>>>> Dave.
>>>>>>
>>>>>
>>>>> Recent "realpath" changes also went in:
>>>>>
>>>>> b03c662399956609934ff063e9064e15dc7459ac xfstests: fix to build
>>>>> src/realpath and the correct the existence of target devices
>>>>> d5ea873fcbf00b841f0511a26fb341364a5c6425 xfstests: resolve symlinked
>>>>> devices to real paths
>>>>>
>>>>> so either of those could have broken and/or fixed it for you :)
>>>>
>>>>
>>>> Yup, those are almost certainly the patches that broke device mapper
>>>> support. There' aren't changes to common.rc, so I didn't notice them.
>>>> They basically override the fix that Christoph made to support
>>>> /dev/mapper links properly.
>>>>
>>>> I'd suggest a revert is in order, the two patches above were part
>>>> of Rich's hovering of old, unreviewed patches. This problem was
>>>> already fixed in a different way by Christoph, so adding these
>>>> patches was wrong.
>>>
>>>
>>> Fine with me ...
>>>
>>> Ryan, can you test w/o those, and if it works -
>>>
>>> Rich, can you revert those please?
>>
>>
>> Sure, I will wait to hear from Ryan first.
>>
>
> Sorry for my late response.
>
> I reverted the 2 patches and rebuilt xfstests on my box.
> It can work now on the multipath disks.
>
> Thanks for your support!
>
> Ryan

Patches c7459ac and 5c6425 have been reverted on
git://oss.sgi.com/xfs/cmds/xfstests master branch, commits 86cf77 and 
82e889.

--Rich

>
>>>
>>> -Eric
>>>
>>>> Cheers,
>>>>
>>>> Dave.
>>>>
>>>
>>> _______________________________________________
>>> xfs mailing list
>>> xfs at oss.sgi.com
>>> http://oss.sgi.com/mailman/listinfo/xfs
>>>
>>
>>