[PATCH] xfs: recalculate leaf entry pointer after compacting a dir2 block

[Mark Tinguely]

On 01/10/13 16:47, Dave Chinner wrote:
> On Thu, Jan 10, 2013 at 10:41:48AM -0600, Eric Sandeen wrote:
>> Dave Jones hit this assert when doing a compile on recent git, with
>> CONFIG_XFS_DEBUG enabled:
>>
>> XFS: Assertion failed: (char *)dup - (char *)hdr == be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)), file: fs/xfs/xfs_dir2_data.c, line: 828
>>
>> Upon further digging, the tag found by xfs_dir2_data_unused_tag_p(dup)
>> contained "2" and not the proper offset, and I found that this value was
>> changed after the memmoves under "Use a stale leaf for our new entry."
>> in xfs_dir2_block_addname(), i.e.
>>
>>                          memmove(&blp[mid + 1],&blp[mid],
>>                                  (highstale - mid) * sizeof(*blp));
>>
>> overwrote it.
>>
>> What has happened is that the previous call to xfs_dir2_block_compact()
>> has rearranged things; it changes btp->count as well as the
>> blp array.  So after we make that call, we must recalculate the
>> proper pointer to the leaf entries by making another call to
>> xfs_dir2_block_leaf_p().
>
> Perhaps a bit of ascii art to explain it for other reviewers to save
> them having to spend a long time trying to understand the dir2 code?
>
> Pre-compact:
>
>   dup	+   unused data pointer - size 72 bytes at offset 3560
> 	|
> 	|
>   blp	+   leaf entry [0] - good - at offset 3632
> 	+   leaf entry [1] - good
> 	+   leaf entry [2] - stale
> 	+   leaf entry [3] - good
> 	+   leaf entry [4] - good
> 	+   leaf entry [5] - stale
> 	+   leaf entry [6] - good
> 	.....
> 	+   leaf entry [n] - stale
> 	+   leaf entry [n+1] - good
> 	+   block tail - points to offset 3632
>
> Post compact (removes all but the highest stale leaf entry, adds the
> free space to the last unused data region):
>
>   dup	+   unused data pointer - size 112 bytes at offset 3560
> 	|
> 	|
>   blp	|   points into unused region, not to start of leaf array
> 	|
> 	|
> 	+   leaf entry [0] - good - at offset 3672
> 	+   leaf entry [1] - good
> 	+   leaf entry [2] - good
> 	+   leaf entry [3] - good
> 	.....
> 	+   leaf entry [n] - stale
> 	+   leaf entry [n+1] - good
> 	+   block tail - points to offset 3672
>
> So as you can see, blp needs to be updated to point at offset 3672...

Thanks for the study aid - it did help explain the 
xfs_dir2_block_compact() code.

Reviewed-by: Mark Tinguely <tinguely at sgi.com>