From: Bryn M. Reeves <bmr@redhat.com>
Date: Thu, 24 Jul 2008 17:47:12 +0100
Subject: [fs] fix softlockups when repeatedly dropping caches
Message-id: 4888B210.8080300@redhat.com
O-Subject: [RHEL 5.3 PATCH] Fix softlockups when repeatedly dropping caches (bz 444961)
Bugzilla: 444961
RH-Acked-by: Josef Bacik <jbacik@redhat.com>
RH-Acked-by: Pete Zaitcev <zaitcev@redhat.com>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

LLNL noticed that while repeatedly dropping page caches with file system
activity going on the system will eventually start producing softlockup
warnings and everything starts to contend for the inode_lock.

This was reported on lkml (while running XFS stress tests):

http://lkml.org/lkml/2008/3/18/150

This can be easily reproduced, e.g. running a kernel build over NFS and
dropping the page cache in a shell loop:

make -j8 a kernel in a directory that is backed by NFS
# while true;do echo 1 > /proc/sys/vm/drop_caches;sleep 2;done

The problem is a lock inversion caused by holding inode_lock across
calls to __invalidate_mapping_pages().

Jan Kara proposed a fix (drop the lock across the call but keep a
reference on the inode until the inode list scanning has resumed) that
was merged into Linus' tree in commit the following commit:

commit eccb95cee4f0d56faa46ef22fb94dd4a3578d3eb
Author: Jan Kara <jack@suse.cz>
Date:   Tue Apr 29 00:59:37 2008 -0700

    vfs: fix lock inversion in drop_pagecache_sb()

    Fix longstanding lock inversion in drop_pagecache_sb by dropping
    inode_lock before calling __invalidate_mapping_pages().  We just
    have to make sure inode won't go away from under us by keeping
    reference to it and putting the reference only after we have safely
    resumed the scan of the inode list.  A bit tricky but not too bad...

    Signed-off-by: Jan Kara <jack@suse.cz>
    Cc: Fengguang Wu <wfg@mail.ustc.edu.cn>
    Cc: David Chinner <dgc@sgi.com>
    Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This patch is a straightforward re-diff of the upstream patch. Tested
here and at LLNL applied to 2.6.18-89. I'm waiting for a build against
2.6.18-99 to complete now.

Regards,
Bryn.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org

iD8DBQFIiLIQ6YSQoMYUY94RAqW8AKDAX8cSXkDOmkSpWZosX7bqNJ6ymwCg38Z1
WWXDDCtDyxWgiR0ok9v7bZM=
=yHbT
-----END PGP SIGNATURE-----

diff --git a/fs/drop_caches.c b/fs/drop_caches.c
index 59375ef..f5aae26 100644
--- a/fs/drop_caches.c
+++ b/fs/drop_caches.c
@@ -14,15 +14,21 @@ int sysctl_drop_caches;
 
 static void drop_pagecache_sb(struct super_block *sb)
 {
-	struct inode *inode;
+	struct inode *inode, *toput_inode = NULL;
 
 	spin_lock(&inode_lock);
 	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
 		if (inode->i_state & (I_FREEING|I_WILL_FREE))
 			continue;
+		__iget(inode);
+		spin_unlock(&inode_lock);
 		__invalidate_mapping_pages(inode->i_mapping, 0, -1, true);
+		iput(toput_inode);
+		toput_inode = inode;
+		spin_lock(&inode_lock);
 	}
 	spin_unlock(&inode_lock);
+	iput(toput_inode);
 }
 
 void drop_pagecache(void)