On Mon, Jan 04, 2016 at 02:12:18PM -0800, Darrick J. Wong wrote:
> On Sun, Jan 03, 2016 at 04:15:25AM -0800, Christoph Hellwig wrote:
> > On Sat, Dec 19, 2015 at 12:56:55AM -0800, Darrick J. Wong wrote:
> > > Because struct xfs_agfl is 36 bytes long and has a 64-bit integer
> > > inside it, gcc will quietly round the structure size up to the nearest
> > > 64 bits -- in this case, 40 bytes. This results in the XFS_AGFL_SIZE
> > > macro returning incorrect results for v5 filesystems on 64-bit
> > > machines (118 items instead of 119). As a result, a 32-bit xfs_repair
> > > will see garbage in AGFL item 119 and complain.
> > >
> > > Therefore, tell gcc not to pad the structure so that the AGFL size
> > > calculation is correct.
> >
> > Do you have a testcase for this?
>
> Not much aside from:
>
> 0. Build kernel/xfsprogs with RFCv4 patches on a 64bit machine.
> 1. Build kernel/xfsprogs with RFCv4 patches on a 32bit machine.
> 2. Format a XFS with reflink and rmap on a 64-bit machine, so that the AGFL
> size is maximized.
> 3. Mount FS and create a reflinked file.
> 4. Unmount and xfs_repair with the 32-bit build.
>
> I guess we could create a program that compares all the known sizeof(struct
> xfs_disk_object) values against known good values and stuff that into the
> xfsprogs build process.
There's an xfstest for that: xfs/122. It notruns on my systems since
the big xfsprogs build/header rework, and I haven't found the time
to work out what it needs to run again. Also, I don't think it
covers the AGFL structure, because that it relatively new and the
test doesn't check any of the v5 specific structures....
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
|