On 12/5/18 10:48 AM, Darrick J. Wong wrote:
> On Tue, Dec 04, 2018 at 10:11:50PM -0600, Eric Sandeen wrote:
>>
>>
>> On 12/4/18 9:48 PM, Eric Sandeen wrote:
>>> On 11/22/18 12:13 PM, Darrick J. Wong wrote:
>>>> From: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
>>>>
>>>> In process_node_dir2, we need to distinguish between a directory with a
>>>> single leafn block (yes, they exist) having no interior da nodes, and a
>>>> directory with a da tree that incorrectly points to dablk 0. If we
>>>> happened to fill out any part of the da cursor then we have a da btree
>>>> with garbage in it; otherwise, we have a single leafn block.
>>>>
>>>> This was found by repair repeatedly rebuilding a directory containing a
>>>> single leafn block (xfs/495).
>>>>
>>>> Fixes: 67a79e2cc932 ("xfs_repair: treat zero da btree pointers as corruption")
>>>> Reported-by: Dave Chinner <david@xxxxxxxxxxxxx>
>>>> Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
>>>
>>> I'm finding the commit log hard to parse/understand.
>>>
>>> Let's reference
>>>
>>> 71a6af8 Revert "xfs_repair: treat zero da btree pointers as corruption"
>>>
>>> for starters, but can we do something like this....?
>>>
>>> ===
>>>
>>> As explained in
>>>
>>> 71a6af8 Revert "xfs_repair: treat zero da btree pointers as corruption"
>>>
>>> a single root LEAFN block can exist in a directory until it grows further.
>>>
>>> This is why, normally, we skip directories with a root marked
>>> XFS_DIR2_LEAFN_MAGIC, as detected by the left-most leaf block being
>>> found at file block 0.
>>>
>>> However, if we traversed any level of a btree to get here (as
>>> indicated by da_cursor.active > 0), then a leaf block claiming block
>>> 0 indicates corruption, and we should handle it as such, and rebuild
>>> the directory.
>>>
>>> This was found by repair repeatedly rebuilding a directory containing a
>>> single leafn block (xfs/495).
>>>
>>> Fixes: 67a79e2cc932 ("xfs_repair: treat zero da btree pointers as corruption")
>>> Reported-by: Dave Chinner <david@xxxxxxxxxxxxx>
>>> Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
>>>
>>> ===
>>>
>>> I'm not hung up on rewriting the commit log, but if you like it give me
>>> thumbs up. I needed to think it through that way to grok the change.
>>> Anyway, for the change now that I do grok it,
>>>
>>> Reviewed-by: Eric Sandeen <sandeen@xxxxxxxxxx>
>>
>> Uh, the summary is wrong too, o? We /already/ "don't error out"
>> on dirs with a single leafn block. Maybe:
>>
>> xfs_repair: rebuild directory when non-root leafn blocks claim block 0
>>
>> ?
>
> Works for me. Want a resend or are you fixing it up on the way in?
I'll do it.
-Eric
