xfs: switch to NOFS allocation under i_lock in xfs_da_buf_make
authorChristoph Hellwig <hch@infradead.org>
Sat, 18 Jul 2009 22:14:56 +0000 (18:14 -0400)
committerFelix Blyakher <felixb@sgi.com>
Tue, 11 Aug 2009 15:46:24 +0000 (10:46 -0500)
i_lock is taken in the reclaim context so all allocations under it
must avoid recursions into the filesystem.

Reported by the new reclaim context tracing in lockdep.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Felix Blyakher <felixb@sgi.com>
Signed-off-by: Felix Blyakher <felixb@sgi.com>
fs/xfs/xfs_da_btree.c

index bd0bb6dfcdfadc97c653baa17f206872b5734741..2847bbc1c534db4b9e2b3bd49645773678d15e07 100644 (file)
@@ -2261,9 +2261,9 @@ xfs_da_buf_make(int nbuf, xfs_buf_t **bps, inst_t *ra)
        int             off;
 
        if (nbuf == 1)
-               dabuf = kmem_zone_alloc(xfs_dabuf_zone, KM_SLEEP);
+               dabuf = kmem_zone_alloc(xfs_dabuf_zone, KM_NOFS);
        else
-               dabuf = kmem_alloc(XFS_DA_BUF_SIZE(nbuf), KM_SLEEP);
+               dabuf = kmem_alloc(XFS_DA_BUF_SIZE(nbuf), KM_NOFS);
        dabuf->dirty = 0;
 #ifdef XFS_DABUF_DEBUG
        dabuf->ra = ra;