b=2313

r=shaver
This bug happens when a file is opened twice for write, then both close it
at the same time.  If they both drop the writecount, then race to
compare it against 0, one will free the fsdata and the other will assert.

This looks like a big patch, but it's mostly plumbing.  I had to do some
different argument passing, in order to keep everything protected under the
same lock.

I removed the writecount spinlock, and use the epoch semaphore for all three
things: management of the epoch, protection of the writecount, and atomicity of
writecount modifications which result in allocation or freeing of the fsdata.

diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index 95d6dd9..1264aa7 100644 (file)
--- a/lustre/ChangeLog
+++ b/lustre/ChangeLog
@@ -17,6 +17,8 @@ tbd         Cluster File Systems, Inc. <info@clusterfs.com>
        - fix gateway LMC keyword conflict (2318)
        - fix MDS lock inversions in getattr/reint paths (1844)
        - fix a rare lock re-ordering bug, which caused deadlock (2322)
+       - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
+       - protect MDS inode fsdata with stronger locking (2313)
        * miscellania
        - allow configurable automake binary, for testing new versions
        - small update to the lfs documentation