On Wed, Mar 23, 2011 at 4:57 PM, Eric Sandeen <sandeen@xxxxxxxxxx> wrote:
> On 3/23/11 9:36 AM, Amir Goldstein wrote:
>> On Wed, Mar 23, 2011 at 4:03 PM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
>>> I've been running xfstests on ext4 and I always get annoying failures to
>>> 213 8s ... 31s
>>> umount: /mnt/test/ext4: device is busy.
>>> (In some cases useful info about processes that use
>>> the device is found by lsof(8) or fuser(1))
>>> post 198 mostly fails to umount as well.
>>> and post 124 always fails to umount.
>>> My server is Ubuntu 10.10, running kernel 2.6.38.
>>> Is there any known issue about this? because I could find it on google
>>> or on the XFS list.
>> Hi All,
>> Has anyone seen these failures when running xfstests on ext4?
>> My test and scratch partitions are dedicated for xfstests.
> I've not seen it, can you investigate with lsof etc?
OK. I have not given complete information.
It does not fail indefinably. I can umount the fs immediately after the failure.
Only the test fails because fsck sees a mounted fs.
There must be a bit of a delay in the umount that causes the problem.
> They both work for me on a quick test, anyway, on a .38
Problem existed for me in .37 as well. I don't think it is kernel (or
I think it is a system problem, but I fall short of ideas how to fix it.
> What are you using for your partitions under test?
Currently, I have /dev/sda5,6,7,8.
I used to work with LVM, but lstat64 didn't like the /dev/mapper links,
so I switched to raw partitions.
Although I was puzzled that xfstests could not work over LVM partitions
It could when I changed lstat64 to stat64, but then I changed it back
thinking it was causing my problems.