Merge tag 'xfs-for-linus-v3.14-rc1' of git://oss.sgi.com/xfs/xfs

Pull xfs update from Ben Myers:
 "This is primarily bug fixes, many of which you already have.  New
  stuff includes a series to decouple the in-memory and on-disk log
  format, helpers in the area of inode clusters, and i_version handling.

  We decided to try to use more topic branches this release, so there
  are some merge commits in there on account of that.  I'm afraid I
  didn't do a good job of putting meaningful comments in the first
  couple of merges.  Sorry about that.  I think I have the hang of it
  now.

  For 3.14-rc1 there are fixes in the areas of remote attributes,
  discard, growfs, memory leaks in recovery, directory v2, quotas, the
  MAINTAINERS file, allocation alignment, extent list locking, and in
  xfs_bmapi_allocate.  There are cleanups in xfs_setsize_buftarg,
  removing unused macros, quotas, setattr, and freeing of inode
  clusters.  The in-memory and on-disk log format have been decoupled, a
  common helper to calculate the number of blocks in an inode cluster
  has been added, and handling of i_version has been pulled into the
  filesystems that use it.

   - cleanup in xfs_setsize_buftarg
   - removal of remaining unused flags for vop toss/flush/flushinval
   - fix for memory corruption in xfs_attrlist_by_handle
   - fix for out-of-date comment in xfs_trans_dqlockedjoin
   - fix for discard if range length is less than one block
   - fix for overrun of agfl buffer using growfs on v4 superblock
     filesystems
   - pull i_version handling out into the filesystems that use it
   - don't leak recovery items on error
   - fix for memory leak in xfs_dir2_node_removename
   - several cleanups for quotas
   - fix bad assertion in xfs_qm_vop_create_dqattach
   - cleanup for xfs_setattr_mode, and add xfs_setattr_time
   - fix quota assert in xfs_setattr_nonsize
   - fix an infinite loop when turning off group/project quota before
     user quota
   - fix for temporary buffer allocation failure in xfs_dir2_block_to_sf
     with large directory block sizes
   - fix Dave's email address in MAINTAINERS
   - cleanup calculation of freed inode cluster blocks
   - fix alignment of initial file allocations to match filesystem
     geometry
   - decouple in-memory and on-disk log format
   - introduce a common helper to calculate the number of filesystem
     blocks in an inode cluster
   - fixes for extent list locking
   - fix for off-by-one in xfs_attr3_rmt_verify
   - fix for missing destroy_work_on_stack in xfs_bmapi_allocate"

* tag 'xfs-for-linus-v3.14-rc1' of git://oss.sgi.com/xfs/xfs: (51 commits)
  xfs: Calling destroy_work_on_stack() to pair with INIT_WORK_ONSTACK()
  xfs: fix off-by-one error in xfs_attr3_rmt_verify
  xfs: assert that we hold the ilock for extent map access
  xfs: use xfs_ilock_attr_map_shared in xfs_attr_list_int
  xfs: use xfs_ilock_attr_map_shared in xfs_attr_get
  xfs: use xfs_ilock_data_map_shared in xfs_qm_dqiterate
  xfs: use xfs_ilock_data_map_shared in xfs_qm_dqtobp
  xfs: take the ilock around xfs_bmapi_read in xfs_zero_remaining_bytes
  xfs: reinstate the ilock in xfs_readdir
  xfs: add xfs_ilock_attr_map_shared
  xfs: rename xfs_ilock_map_shared
  xfs: remove xfs_iunlock_map_shared
  xfs: no need to lock the inode in xfs_find_handle
  xfs: use xfs_icluster_size_fsb in xfs_imap
  xfs: use xfs_icluster_size_fsb in xfs_ifree_cluster
  xfs: use xfs_icluster_size_fsb in xfs_ialloc_inode_init
  xfs: use xfs_icluster_size_fsb in xfs_bulkstat
  xfs: introduce a common helper xfs_icluster_size_fsb
  xfs: get rid of XFS_IALLOC_BLOCKS macros
  xfs: get rid of XFS_INODE_CLUSTER_SIZE macros
  ...

1 files changed, 23 insertions, 12 deletions

diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c
index 82e0dab46ee5..f264616080ca 100644
--- a/fs/xfs/xfs_bmap_util.c
+++ b/fs/xfs/xfs_bmap_util.c
@@ -618,22 +618,27 @@ xfs_getbmap(
                 return XFS_ERROR(ENOMEM);
 
         xfs_ilock(ip, XFS_IOLOCK_SHARED);
-        if (whichfork == XFS_DATA_FORK && !(iflags & BMV_IF_DELALLOC)) {
-                if (ip->i_delayed_blks || XFS_ISIZE(ip) > ip->i_d.di_size) {
+        if (whichfork == XFS_DATA_FORK) {
+                if (!(iflags & BMV_IF_DELALLOC) &&
+                    (ip->i_delayed_blks || XFS_ISIZE(ip) > ip->i_d.di_size)) {
                         error = -filemap_write_and_wait(VFS_I(ip)->i_mapping);
                         if (error)
                                 goto out_unlock_iolock;
+
+                        /*
+                         * Even after flushing the inode, there can still be
+                         * delalloc blocks on the inode beyond EOF due to
+                         * speculative preallocation.  These are not removed
+                         * until the release function is called or the inode
+                         * is inactivated.  Hence we cannot assert here that
+                         * ip->i_delayed_blks == 0.
+                         */
                 }
-                /*
-                 * even after flushing the inode, there can still be delalloc
-                 * blocks on the inode beyond EOF due to speculative
-                 * preallocation. These are not removed until the release
-                 * function is called or the inode is inactivated. Hence we
-                 * cannot assert here that ip->i_delayed_blks == 0.
-                 */
-        }
 
-        lock = xfs_ilock_map_shared(ip);
+                lock = xfs_ilock_data_map_shared(ip);
+        } else {
+                lock = xfs_ilock_attr_map_shared(ip);
+        }
 
         /*
          * Don't let nex be bigger than the number of extents
@@ -738,7 +743,7 @@ xfs_getbmap(
  out_free_map:
         kmem_free(map);
  out_unlock_ilock:
-        xfs_iunlock_map_shared(ip, lock);
+        xfs_iunlock(ip, lock);
  out_unlock_iolock:
         xfs_iunlock(ip, XFS_IOLOCK_SHARED);
 
@@ -1169,9 +1174,15 @@ xfs_zero_remaining_bytes(
         xfs_buf_unlock(bp);
 
         for (offset = startoff; offset <= endoff; offset = lastoffset + 1) {
+                uint lock_mode;
+
                 offset_fsb = XFS_B_TO_FSBT(mp, offset);
                 nimap = 1;
+
+                lock_mode = xfs_ilock_data_map_shared(ip);
                 error = xfs_bmapi_read(ip, offset_fsb, 1, &imap, &nimap, 0);
+                xfs_iunlock(ip, lock_mode);
+
                 if (error || nimap < 1)
                         break;
                 ASSERT(imap.br_blockcount >= 1);