Commit 97236ad5 authored by Andreas Gruenbacher's avatar Andreas Gruenbacher
Browse files

gfs2: Avoid dequeuing GL_ASYNC glock holders twice 



When a locking request fails, the associated glock holder is
automatically dequeued from the list of active and waiting holders.  For
GL_ASYNC locking requests, this will obviously happen asynchronously
and it can race with attempts to cancel that locking request via
gfs2_glock_dq().  Therefore, don't forget to check if a locking request
has already been dequeued in gfs2_glock_dq().

Signed-off-by: default avatarAndreas Gruenbacher <agruenba@redhat.com>
parent 4ad02083

fs/gfs2/glock.c

+8 −0
Original line number Diff line number Diff line
@@ -1707,6 +1707,13 @@ void gfs2_glock_dq(struct gfs2_holder *gh) 
	struct gfs2_glock *gl = gh->gh_gl;



	spin_lock(&gl->gl_lockref.lock);

	if (!gfs2_holder_queued(gh)) {

		/*

		 * May have already been dequeued because the locking request

		 * was GL_ASYNC and it has failed in the meantime.

		 */

		goto out;

	}

	if (list_is_first(&gh->gh_list, &gl->gl_holders) &&

	    !test_bit(HIF_HOLDER, &gh->gh_iflags)) {

		spin_unlock(&gl->gl_lockref.lock);
@@ -1716,6 +1723,7 @@ void gfs2_glock_dq(struct gfs2_holder *gh) 
	}



	__gfs2_glock_dq(gh);

out:

	spin_unlock(&gl->gl_lockref.lock);

}