xfs
[Top] [All Lists]

Re: [PATCH 4/7] xfs: lockdep annotations for xfs_dqlock2

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>, xfs@xxxxxxxxxxx
Subject: Re: [PATCH 4/7] xfs: lockdep annotations for xfs_dqlock2
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Mon, 12 Jan 2009 10:15:44 -0500
In-reply-to: <20090111230637.GE8071@disturbed>
References: <20090109221104.237540000@xxxxxxxxxxxxxxxxxxxxxx> <20090109221300.520949000@xxxxxxxxxxxxxxxxxxxxxx> <20090111230637.GE8071@disturbed>
User-agent: Mutt/1.5.18 (2008-05-17)
On Mon, Jan 12, 2009 at 10:06:37AM +1100, Dave Chinner wrote:
> This looks a bit wierd.
> 
> Yes, xfs_dqlock() is just a wrapper around mutex_lock, but we should
> be consistent here. Can you add a xfs_dqlock_nested() wrapper to do
> this?

I don't think we should add more of the silly wrappers.  What about
the version below that always uses plain mutex_lock* in xfs_dqlock2?

---

Subject: xfs: lockdep annotations for xfs_dqlock2
From: Christoph Hellwig <hch@xxxxxx>

xfs_dqlock2 locks two xfs_dquots, which is fine as it always locks the
dquot with the lower id first.  Use mutex_lock_nested to tell lockdep 
about this fact.  Also clean up xfs_dqlock2 a bit by rationalizing
the conditionals and always using the mutex_lock family of functions
directly.


Signed-off-by: Christoph Hellwig <hch@xxxxxx>

Index: xfs/fs/xfs/quota/xfs_dquot.c
===================================================================
--- xfs.orig/fs/xfs/quota/xfs_dquot.c   2009-01-12 14:37:49.445796033 +0100
+++ xfs/fs/xfs/quota/xfs_dquot.c        2009-01-12 14:46:59.913795998 +0100
@@ -1383,6 +1383,12 @@ xfs_dqunlock_nonotify(
        mutex_unlock(&(dqp->q_qlock));
 }
 
+/*
+ * Lock two xfs_dquot structures.
+ *
+ * To avoid deadlocks we always lock the quota structure with
+ * the lowerd id first.
+ */
 void
 xfs_dqlock2(
        xfs_dquot_t     *d1,
@@ -1392,18 +1398,16 @@ xfs_dqlock2(
                ASSERT(d1 != d2);
                if (be32_to_cpu(d1->q_core.d_id) >
                    be32_to_cpu(d2->q_core.d_id)) {
-                       xfs_dqlock(d2);
-                       xfs_dqlock(d1);
+                       mutex_lock(&d2->q_qlock);
+                       mutex_lock_nested(&d1->q_qlock, XFS_QLOCK_NESTED);
                } else {
-                       xfs_dqlock(d1);
-                       xfs_dqlock(d2);
-               }
-       } else {
-               if (d1) {
-                       xfs_dqlock(d1);
-               } else if (d2) {
-                       xfs_dqlock(d2);
+                       mutex_lock(&d1->q_qlock);
+                       mutex_lock_nested(&d2->q_qlock, XFS_QLOCK_NESTED);
                }
+       } else if (d1) {
+               mutex_lock(&d1->q_qlock);
+       } else if (d2) {
+               mutex_lock(&d2->q_qlock);
        }
 }
 
Index: xfs/fs/xfs/quota/xfs_dquot.h
===================================================================
--- xfs.orig/fs/xfs/quota/xfs_dquot.h   2009-01-12 14:37:49.459795861 +0100
+++ xfs/fs/xfs/quota/xfs_dquot.h        2009-01-12 14:45:30.902817471 +0100
@@ -97,6 +97,16 @@ typedef struct xfs_dquot {
 #define dq_hashlist    q_lists.dqm_hashlist
 #define dq_flags       q_lists.dqm_flags
 
+/*
+ * Lock hierachy for q_qlock:
+ *     XFS_QLOCK_NORMAL is the implicit default,
+ *     XFS_QLOCK_NESTED is the dquot with the higher id in xfs_dqlock2
+ */
+enum {
+       XFS_QLOCK_NORMAL = 0,
+       XFS_QLOCK_NESTED,
+};
+
 #define XFS_DQHOLD(dqp)                ((dqp)->q_nrefs++)
 
 #ifdef DEBUG

<Prev in Thread] Current Thread [Next in Thread>