From: David Teigland <teigland@redhat.com>
Date: Thu, 29 Oct 2009 10:37:47 -0500
Subject: [dlm] use GFP_NOFS on all lockspaces
Message-id: 20091029153747.GC13474@redhat.com
O-Subject: [RHEL5.5 PATCH] dlm: use GFP_NOFS on all lockspaces
Bugzilla: 530537
RH-Acked-by: Christine Caulfield <ccaulfie@redhat.com>
RH-Acked-by: Rik van Riel <riel@redhat.com>

bz 530537

not upstream; with more time I may come up with a different solution

dlm memory allocations for any locks from any lockspace can cause
kernel memory reclamation to call into gfs which can then call into
the dlm.  If this happens to be in the context of the dlm_recv
thread, that thread will deadlock, blocking all further dlm activity.
This means all dlm memory allocations must use GFP_NOFS, not just
allocations related to locks belonging to gfs.

Signed-off-by: David Teigland <teigland@redhat.com>

diff --git a/fs/dlm/lockspace.c b/fs/dlm/lockspace.c
index d2a1df1..27ce1d5 100644
--- a/fs/dlm/lockspace.c
+++ b/fs/dlm/lockspace.c
@@ -450,10 +450,7 @@ static int new_lockspace(char *name, int namelen, void **lockspace,
 	if (flags & DLM_LSFL_TIMEWARN)
 		set_bit(LSFL_TIMEWARN, &ls->ls_flags);
 
-	if (flags & DLM_LSFL_FS)
-		ls->ls_allocation = GFP_NOFS;
-	else
-		ls->ls_allocation = GFP_KERNEL;
+	ls->ls_allocation = GFP_NOFS;
 
 	/* ls_exflags are forced to match among nodes, and we don't
 	   need to require all nodes to have TIMEWARN or FS set */