Hi folks, These are the fixes for the attr fork regressions in the current for-next tree. The got through testing because none of my test regression VMs had selinux enabled on them, while my perf test VMs did. Hence it was never exercised by the fstests runs I did. I've re-enabled selinux on some of my test VMs, and run it through fstests ofor v4 and v5 filesystems with selinux enabled. The first 3 patches are the fixes that address the regressions. The last patch is an optimisation I noticed that avoids recalculating a static, fixed value on every call to xfs_default_attroffset(). It is not needed for the bugs to be fixed, but is definitely a for-next candidate while I'm touching that code again. Cheers, Dave.
