GFS2: Use rht_for_each_entry_rcu in glock_hash_walk
authorAndrew Price <anprice@redhat.com>
Thu, 12 Nov 2015 20:07:52 +0000 (14:07 -0600)
committerBob Peterson <rpeterso@redhat.com>
Mon, 16 Nov 2015 17:57:59 +0000 (11:57 -0600)
commit3dd1dd8c696bdb7c8dcc9456cb23558ad1b336b8
tree3ec8d2b7d442737c15f6b83999713c1c38cbdfc7
parent6fde22426be6af261816db5941744b8d3c4c7f96
GFS2: Use rht_for_each_entry_rcu in glock_hash_walk

This lockdep splat was being triggered on umount:

[55715.973122] ===============================
[55715.980169] [ INFO: suspicious RCU usage. ]
[55715.981021] 4.3.0-11553-g8d3de01-dirty #15 Tainted: G        W
[55715.982353] -------------------------------
[55715.983301] fs/gfs2/glock.c:1427 suspicious rcu_dereference_protected() usage!

The code it refers to is the rht_for_each_entry_safe usage in
glock_hash_walk. The condition that triggers the warning is
lockdep_rht_bucket_is_held(tbl, hash) which is checked in the
__rcu_dereference_protected macro.

The rhashtable buckets are not changed in glock_hash_walk so it's safe
to rely on the rcu protection. Replace the rht_for_each_entry_safe()
usage with rht_for_each_entry_rcu(), which doesn't care whether the
bucket lock is held if the rcu read lock is held.

Signed-off-by: Andrew Price <anprice@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Acked-by: Steven Whitehouse <swhiteho@redhat.com>
fs/gfs2/glock.c