[PATCH 0/5] xfs: various 2.6.38 candidate bug fixes

[Dave Chinner <david@xxxxxxxxxxxxx>]

There's the fix for the specualtive allocation regression that arekm
and hch reported different symptoms of. Some extsize tweaks to
prevent assert failures and btree corruptions. An allocation size
fix to prevent wildly inefficient handling of nullfb allocation
group scanning (demonstrated by low space scanning or single large
preallocation that spans hundreds/thousands of AGs).  Finally,
another a fix for another delayed logging commit path error handling
problem that I noticed by inspection.