[PATCH v3 0/8] change sb_writers to use percpu_rw_semaphore

[Oleg Nesterov <oleg@redhat.com>]

On 08/13, Jan Kara wrote:
>
> Regarding the routing, ideally Al Viro should take these as a VFS
> maintainer.

Al, could you take these patches?

Only cosmetic changes in V3 to address the comments from Jan, I
preserved his acks.

In case you missed all the spam I sent before, let me repeat that
the awful (and currently unneeded) 7/8 will be reverted later. We
need it to ensure that other percpu_rw_semaphore changes routed
via another tree won't break fs/super.c. After that we will add
rcu_sync_dtor(s_writers->rw_sem) into deactivate_locked_super()
and revert this horror.

3/8 documents the lockdep problems we currently have. This is fixed
by the patch below but it depends on xfs ILOCK fixes from Dave, so
I will send it later. Plus another patch which removes the "trylock"
hack in __sb_start_write().

Oleg.

 arch/Kconfig                  |    1 -
 fs/btrfs/transaction.c        |    8 +--
 fs/super.c                    |  184 +++++++++++++++++++---------------------
 fs/xfs/xfs_aops.c             |    6 +-
 include/linux/fs.h            |   23 +++---
 include/linux/percpu-rwsem.h  |   20 +++++
 init/Kconfig                  |    1 -
 kernel/locking/Makefile       |    3 +-
 kernel/locking/percpu-rwsem.c |   13 +++
 lib/Kconfig                   |    3 -
 10 files changed, 136 insertions(+), 126 deletions(-)

--------------------------------------------------------------------------
[PATCH v3 9/8] don't fool lockdep in freeze_super() and thaw_super() paths

sb_wait_write()->percpu_rwsem_release() fools lockdep to avoid the
false-positives. Now that xfs was fixed by Dave we can remove it and
change freeze_super() and thaw_super() to run with s_writers.rw_sem
locks held; we add two trivial helpers for that, sb_freeze_release()
and sb_freeze_acquire().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Jan Kara <jack@suse.com>
---
 fs/super.c |   37 +++++++++++++++++++++++++------------
 1 files changed, 25 insertions(+), 12 deletions(-)

diff --git a/fs/super.c b/fs/super.c
index 4350ff4..91c9756 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -1213,25 +1213,34 @@ EXPORT_SYMBOL(__sb_start_write);
 static void sb_wait_write(struct super_block *sb, int level)
 {
 	percpu_down_write(sb->s_writers.rw_sem + level-1);
-	/*
-	 * We are going to return to userspace and forget about this lock, the
-	 * ownership goes to the caller of thaw_super() which does unlock.
-	 *
-	 * FIXME: we should do this before return from freeze_super() after we
-	 * called sync_filesystem(sb) and s_op->freeze_fs(sb), and thaw_super()
-	 * should re-acquire these locks before s_op->unfreeze_fs(sb). However
-	 * this leads to lockdep false-positives, so currently we do the early
-	 * release right after acquire.
-	 */
-	percpu_rwsem_release(sb->s_writers.rw_sem + level-1, 0, _THIS_IP_);
 }
 
-static void sb_freeze_unlock(struct super_block *sb)
+/*
+ * We are going to return to userspace and forget about these locks, the
+ * ownership goes to the caller of thaw_super() which does unlock().
+ */
+static void sb_freeze_release(struct super_block *sb)
+{
+	int level;
+
+	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
+		percpu_rwsem_release(sb->s_writers.rw_sem + level, 0, _THIS_IP_);
+}
+
+/*
+ * Tell lockdep we are holding these locks before we call ->unfreeze_fs(sb).
+ */
+static void sb_freeze_acquire(struct super_block *sb)
 {
 	int level;
 
 	for (level = 0; level < SB_FREEZE_LEVELS; ++level)
 		percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0, _THIS_IP_);
+}
+
+static void sb_freeze_unlock(struct super_block *sb)
+{
+	int level;
 
 	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
 		percpu_up_write(sb->s_writers.rw_sem + level);
@@ -1327,6 +1336,7 @@ int freeze_super(struct super_block *sb)
 	 * sees write activity when frozen is set to SB_FREEZE_COMPLETE.
 	 */
 	sb->s_writers.frozen = SB_FREEZE_COMPLETE;
+	sb_freeze_release(sb);
 	up_write(&sb->s_umount);
 	return 0;
 }
@@ -1353,11 +1363,14 @@ int thaw_super(struct super_block *sb)
 		goto out;
 	}
 
+	sb_freeze_acquire(sb);
+
 	if (sb->s_op->unfreeze_fs) {
 		error = sb->s_op->unfreeze_fs(sb);
 		if (error) {
 			printk(KERN_ERR
 				"VFS:Filesystem thaw failed\n");
+			sb_freeze_release(sb);
 			up_write(&sb->s_umount);
 			return error;
 		}
-- 
1.5.5.1