On Thu, Aug 04, 2022 at 05:31:01AM +0000, xuyang2018.jy@xxxxxxxxxxx wrote:
>
>
> on 2022/08/04 13:15, Darrick J. Wong wrote:
> > On Wed, Aug 03, 2022 at 10:06:16PM -0700, Darrick J. Wong wrote:
> >> On Thu, Aug 04, 2022 at 01:53:31AM +0000, xuyang2018.jy@xxxxxxxxxxx wrote:
> >>> on 2022/08/03 12:21, Darrick J. Wong wrote:
> >>>> From: Darrick J. Wong <djwong@xxxxxxxxxx>
> >>>>
> >>>> Not sure what's up with this new test, but the golden output isn't right
> >>>> for upstream xfsprogs for-next. Change it to pass there...
> >>>
> >>> It failed becuase libxfs code validates v5 feature fields.
> >>>
> >>> b12d5ae5d ("xfs: validate v5 feature fields")
> >>>>
> >>>> Signed-off-by: Darrick J. Wong <djwong@xxxxxxxxxx>
> >>>> ---
> >>>> tests/xfs/533.out | 2 +-
> >>>> 1 file changed, 1 insertion(+), 1 deletion(-)
> >>>>
> >>>>
> >>>> diff --git a/tests/xfs/533.out b/tests/xfs/533.out
> >>>> index 7deb78a3..439fb16e 100644
> >>>> --- a/tests/xfs/533.out
> >>>> +++ b/tests/xfs/533.out
> >>>> @@ -1,5 +1,5 @@
> >>>> QA output created by 533
> >>>> Allowing write of corrupted data with good CRC
> >>>> magicnum = 0
> >>>> -bad magic number
> >>
> >> Ohhh, so this is a V4 output.
> >>
> >>>> +Superblock has bad magic number 0x0. Not an XFS filesystem?
> >>>
> >>> Since this case is designed to detect xfs_db bug, should we filter the
> >>> output?
> >>
> >> Yep. I'll rework this patch to handle V4 and V5. Well, thanks for
> >> keeping me on my toes! ;)
> >
> > Hmm, V4 produces this:
> >
> > --- /tmp/fstests/tests/xfs/533.out 2022-08-02 19:02:12.876335795 -0700
> > +++ /var/tmp/fstests/xfs/533.out.bad 2022-08-03 22:12:43.596000000 -0700
> > @@ -1,5 +1,3 @@
> > QA output created by 533
> > -Allowing write of corrupted data with good CRC
> > magicnum = 0
> > -Superblock has bad magic number 0x0. Not an XFS filesystem?
> > 0
> >
> > So I guess this isn't a V4 output. Which version of xfsprogs and what
> > MKFS_OPTIONS did you use to make this to produce 'bad magic number'?
>
> I remember I used xfsprogs master branch and I don't add any special
> MKFS_OPTIONS.
>
> meta-data=/dev/sda7 isize=512 agcount=4, agsize=3276800 blks
> = sectsz=4096 attr=2, projid32bit=1
> = crc=1 finobt=1, sparse=1, rmapbt=0
> = reflink=1 bigtime=1 inobtcount=1
> data = bsize=4096 blocks=13107200, imaxpct=25
> = sunit=0 swidth=0 blks
> naming =version 2 bsize=4096 ascii-ci=0, ftype=1
> log =internal log bsize=4096 blocks=16384, version=2
> = sectsz=4096 sunit=1 blks, lazy-count=1
> realtime =none extsz=4096 blocks=0, rtextents=0
>
> local.config
> MODULAR=0
> export TEST_DIR=/mnt/xfstests/test
> export TEST_DEV=/dev/sda6
> export SCRATCH_MNT=/mnt/xfstests/scratch
> export SCRATCH_DEV=/dev/sda7
>
> I have mentioned that xfsprogs commit b12d5ae5d ("xfs: validate v5
> feature fields") will change output to "-Superblock has bad magic
> number 0x0. Not an XFS filesystem". And this commit is belong to
> for-next branch that is why I write this case doesn't find this because
> I use master branch that time.
Ah, sorry, I missed your mention of this commit. Ok, so the test was
based on master (aka xfsprogs 5.18) and my corrections are based on
for-next (future xfsprogs 5.19).
Yes, I think the correct fix here is to filter the new output to match
the old. I might throw in a _require_scratch_xfs_crc so this test
doesn't fail on (deprecated) V4 setups.
--D
> Best Regards
> Yang Xu
> >
> > --D
> >
> >> --D
> >>
> >>> Best Regards
> >>> Yang Xu
> >>>> 0
> >>>>
