Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Thu, 7 Jan 2021 13:20:00 -0500 (EST)
From:   Bob Peterson <rpeterso@redhat.com>
To:     linux-fsdevel@vger.kernel.org, Jens Axboe <axboe@kernel.dk>
Cc:     Satya Tangirala <satyat@google.com>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        linux-kernel@vger.kernel.org, Christoph Hellwig <hch@lst.de>
Message-ID: <1630504905.43082455.1610043600869.JavaMail.zimbra@redhat.com>
In-Reply-To: <1137375419.42956970.1610036857271.JavaMail.zimbra@redhat.com>
References: <20201224044954.1349459-1-satyat@google.com> <20210107162000.GA2693@lst.de> <1137375419.42956970.1610036857271.JavaMail.zimbra@redhat.com>
Subject: [fs PATCH] fs: fix freeze count problem in freeze_bdev
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Thread-Topic: Fix freeze_bdev()/thaw_bdev() accounting of bd_fsfreeze_sb
Thread-Index: QLSEzKoRlbffxOodTzmv1w/O4NuJkjkVwvLu
Precedence: bulk

Hi,

I wrote this patch to fix the freeze/thaw device problem before I saw
the patch "fs: Fix freeze_bdev()/thaw_bdev() accounting of bd_fsfreeze_sb"
from Satya Tangirala. That one, however, does not fix the bd_freeze_count
problem and this patch does. Jens, Christoph, what do you think?
This is a very recreatable problem via repeated runs of generic/085,
at least on gfs2.

Description:

Before this patch, if you tried to freeze a device (function freeze_bdev)
while it was being unmounted, it would get NULL back from get_active_super
and correctly bypass the freeze calls. Unfortunately, it forgot to decrement
its bd_fsfreeze_count. Subsequent calls to device thaw (thaw_bdev) would
see the non-zero bd_fsfreeze_count and assume the bd_fsfreeze_sb value was
still valid. That's not a safe assumption and resulted in use-after-free,
which often caused fatal kernel errors like: "unable to handle page fault
for address."

This patch adds the necessary decrement of bd_fsfreeze_count for that
error path. It also adds code to set the bd_fsfreeze_sb to NULL when the
last reference is reached in thaw_bdev.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/block_dev.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/block_dev.c b/fs/block_dev.c
index 9e56ee1f2652..c6daf7d12546 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -555,8 +555,10 @@ int freeze_bdev(struct block_device *bdev)
 		goto done;
 
 	sb = get_active_super(bdev);
-	if (!sb)
+	if (!sb) {
+		bdev->bd_fsfreeze_count--;
 		goto sync;
+	}
 	if (sb->s_op->freeze_super)
 		error = sb->s_op->freeze_super(sb);
 	else
@@ -600,6 +602,7 @@ int thaw_bdev(struct block_device *bdev)
 	if (!sb)
 		goto out;
 
+	bdev->bd_fsfreeze_sb = NULL;
 	if (sb->s_op->thaw_super)
 		error = sb->s_op->thaw_super(sb);
 	else