Re: [PATCH] net: fix setsockopt() locking errors
From: Jarek Poplawski <hidden>
Date: 2009-01-26 11:50:23
On 24-01-2009 23:49, Vegard Nossum wrote:
Hi, This survives basic testing here, but I don't know what that counts for when I couldn't reproduce the lockdep report in the first place. Please review. Vegard From cc8bcd1c4fd219a31d6d191aefa4b4b57dadb9b0 Mon Sep 17 00:00:00 2001 From: Vegard Nossum <redacted> Date: Sat, 24 Jan 2009 22:44:16 +0100 Subject: [PATCH] net: fix setsockopt() locking errors MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit Martin MOKREJ. [off-list ref] reported:quoted
======================================================= [ INFO: possible circular locking dependency detected ] 2.6.29-rc2-git1 #1 ------------------------------------------------------- tcpdump/3734 is trying to acquire lock: (&mm->mmap_sem){----}, at: [<c1053294>] might_fault+0x30/0x6b but task is already holding lock: (sk_lock-AF_PACKET){--..}, at: [<c12798c8>] sock_setsockopt+0x12b/0x4a4 which lock already depends on the new lock.It turns out that sock_setsockopt() is calling copy_from_user() while holding the lock on the socket.
I guess it has been like this for some time, so it would be nice to mention what scenario happens here, or IOW what exactly needs to get these locks in reverse order.
quoted hunk ↗ jump to hunk
We fix it by splitting the ioctl code so that one switch handles the ioctls that have their own code for reading from userspace, and one switch handles the cases that require no additional reading. Reported-by: Martin MOKREJ. <redacted> Signed-off-by: Vegard Nossum <redacted> --- net/core/sock.c | 134 +++++++++++++++++++++++++++++++++++------------------- 1 files changed, 87 insertions(+), 47 deletions(-)diff --git a/net/core/sock.c b/net/core/sock.c index f3a0d08..6bd618d 100644 --- a/net/core/sock.c +++ b/net/core/sock.c@@ -424,6 +424,80 @@ out: return ret; } +static int sock_linger(struct sock *sk, char __user *optval, int optlen)
...
+static int sock_set_rcvtimeo(struct sock *sk, char __user *optval, int optlen)
+{
+ int ret;
+ long rcvtimeo;
+
+ ret = sock_set_timeout(&rcvtimeo, optval, optlen);A check for error is needed here and below.
+
+ lock_sock(sk);
+ sk->sk_rcvtimeo = rcvtimeo;
+ release_sock(sk);
+
+ return ret;
+}
+
+static int sock_set_sndtimeo(struct sock *sk, char __user *optval, int optlen)
+{
+ int ret;
+ long sndtimeo;
+
+ ret = sock_set_timeout(&sndtimeo, optval, optlen);
+
+ lock_sock(sk);
+ sk->sk_sndtimeo = sndtimeo;
+ release_sock(sk);
+
+ return ret;
+}... Jarek P.