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;
Content-Type: text/plain; charset="UTF-8"
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
MIME-Version: 1.0
From:   Ben Hutchings <ben@decadent.org.uk>
To:     linux-kernel@vger.kernel.org, stable@vger.kernel.org
CC:     akpm@linux-foundation.org, Denis Kirjanov <kda@linux-powerpc.org>,
        "Cengiz Can" <cengiz@kernel.wtf>, "Bob Liu" <bob.liu@oracle.com>,
        "Steven Rostedt (VMware)" <rostedt@goodmis.org>,
        "Ming Lei" <ming.lei@redhat.com>, "Jens Axboe" <axboe@kernel.dk>
Date:   Fri, 24 Apr 2020 00:07:20 +0100
Message-ID: <lsq.1587683028.504333663@decadent.org.uk>
Subject: [PATCH 3.16 213/245] blktrace: fix dereference after null check
In-Reply-To: <lsq.1587683027.831233700@decadent.org.uk>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

3.16.83-rc1 review patch.  If anyone has any objections, please let me know.

------------------

From: Cengiz Can <cengiz@kernel.wtf>

commit 153031a301bb07194e9c37466cfce8eacb977621 upstream.

There was a recent change in blktrace.c that added a RCU protection to
`q->blk_trace` in order to fix a use-after-free issue during access.

However the change missed an edge case that can lead to dereferencing of
`bt` pointer even when it's NULL:

Coverity static analyzer marked this as a FORWARD_NULL issue with CID
1460458.

```
/kernel/trace/blktrace.c: 1904 in sysfs_blk_trace_attr_store()
1898            ret = 0;
1899            if (bt == NULL)
1900                    ret = blk_trace_setup_queue(q, bdev);
1901
1902            if (ret == 0) {
1903                    if (attr == &dev_attr_act_mask)
>>>     CID 1460458:  Null pointer dereferences  (FORWARD_NULL)
>>>     Dereferencing null pointer "bt".
1904                            bt->act_mask = value;
1905                    else if (attr == &dev_attr_pid)
1906                            bt->pid = value;
1907                    else if (attr == &dev_attr_start_lba)
1908                            bt->start_lba = value;
1909                    else if (attr == &dev_attr_end_lba)
```

Added a reassignment with RCU annotation to fix the issue.

Fixes: c780e86dd48 ("blktrace: Protect q->blk_trace with RCU")
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Bob Liu <bob.liu@oracle.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Cengiz Can <cengiz@kernel.wtf>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
---
 kernel/trace/blktrace.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

--- a/kernel/trace/blktrace.c
+++ b/kernel/trace/blktrace.c
@@ -1834,8 +1834,11 @@ static ssize_t sysfs_blk_trace_attr_stor
 	}
 
 	ret = 0;
-	if (bt == NULL)
+	if (bt == NULL) {
 		ret = blk_trace_setup_queue(q, bdev);
+		bt = rcu_dereference_protected(q->blk_trace,
+				lockdep_is_held(&q->blk_trace_mutex));
+	}
 
 	if (ret == 0) {
 		if (attr == &dev_attr_act_mask)