Re: Syncing up libxfs to kernel source

[Dave Chinner <david@fromorbit.com>]

On Wed, Jul 23, 2008 at 05:52:24PM +1000, Barry Naujok wrote:
> This one is mainly for Dave...
>
> In xfs_alloc_read_agf() in xfs_alloc.c, the current user-space version
> checks:
>
> agf_ok = ...
> 	be32_to_cpu(agf->agf_btreeblks) <= be32_to_cpu(agf->agf_length) &&
> 	...
>
> but the kernel version doesn't.
>
> Is this an oversight with the lazy-counter code in the kernel, remove from
> user-space or leave them different?

Oversight in the kernel code. Patch below.

Cheers,

Dave.

-- 
Dave Chinner
david@fromorbit.com


Check that the number of AGF btree blocks is within a sane bound
when reading the AGF. Noticed by Barry Naujok.

Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/xfs_alloc.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index 1956f83..a85bf20 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -2185,6 +2185,7 @@ xfs_alloc_read_agf(
 		be32_to_cpu(agf->agf_magicnum) == XFS_AGF_MAGIC &&
 		XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)) &&
 		be32_to_cpu(agf->agf_freeblks) <= be32_to_cpu(agf->agf_length) &&
+		be32_to_cpu(agf->agf_btreeblks) <= be32_to_cpu(agf->agf_length) &&
 		be32_to_cpu(agf->agf_flfirst) < XFS_AGFL_SIZE(mp) &&
 		be32_to_cpu(agf->agf_fllast) < XFS_AGFL_SIZE(mp) &&
 		be32_to_cpu(agf->agf_flcount) <= XFS_AGFL_SIZE(mp);