[PATCH V2] xfstests: test EROFS vs. EEXIST when creating on an RO filesystem
Eric Sandeen
sandeen at sandeen.net
Tue Jan 8 15:11:52 CST 2013
On 1/8/13 2:48 PM, Dave Chinner wrote:
> On Tue, Jan 08, 2013 at 01:42:46PM -0600, Eric Sandeen wrote:
>> On 1/8/13 1:15 PM, Ben Myers wrote:
>>> Hey Eric,
>>>
>>> On Tue, Nov 27, 2012 at 09:19:12PM -0600, Eric Sandeen wrote:
>>>> TBH, I don't know if this is posix-specified, but I found out the
>>>> hard way that when trying to re-create existing files on a readonly
>>>> filesystem, some apps expect/handle EEXIST, but fail on EROFS.
>>>>
>>>> This will test mkdir, mknod, and symlinks for that behavior.
>>>>
>>>> Signed-off-by: Eric Sandeen <sandeen at redhat.com>
>>>> ---
>>>>
>>>> V2: Change test dir var name, do scratch mkfs, drop _fail
>>>
>>> This looked fine to me except that...
>>>
>>>> diff --git a/292.out b/292.out
>>>> new file mode 100644
>>>> index 0000000..485c567
>>>> --- /dev/null
>>>> +++ b/292.out
>>>> @@ -0,0 +1,5 @@
>>>> +QA output created by 292
>>>> +mknod: `SCRATCH_MNT/292.test/testnode': File exists
>>>> +mkdir: cannot create directory `SCRATCH_MNT/292.test/testdir': File exists
>>>> +touch: cannot touch `SCRATCH_MNT/292.test/testtarget': Read-only file system
>>>> +ln: creating symbolic link `SCRATCH_MNT/292.test/testlink': File exists
>>>
>>> ln output is different on my distro:
>>>
>>> ~/xfstests # diff 293.out 293.out.bad
>>> 5c5
>>> < ln: creating symbolic link `SCRATCH_MNT/293.test/testlink': File exists
>>> ---
>>>> ln: failed to create symbolic link `SCRATCH_MNT/293.test/testlink': File exists
>>
>> clearly using the wrong distro ;)
>>
>> How about something like:
>>
>> +_create_files 2>&1 | sed -e "s/failed to create/creating/" | _filter_scratch
>> +_scratch_mount -o remount,ro
>> +_create_files 2>&1 | sed -e "s/failed to create/creating/" | _filter_scratch
>
> There should already be a ln filter that handles this.
>
> Check test 103 - filter_ln.
yeah, it has: -e "s,failed to create,creating,"
so I guess it could be moved to common ...
-Eric
> Cheers,
>
> Dave.
>
More information about the xfs
mailing list