Re: [PATCH bpf v1] unix: fix an issue in unix_shutdown causing the other end read/write failures

From: Song Liu <songliubraving@fb.com>
To: Casey Schaufler <casey@schaufler-ca.com>
Cc: Jiang Wang <jiang.wang@bytedance.com>, bpf <bpf@vger.kernel.org>,
	"Cong Wang" <cong.wang@bytedance.com>,
	"David S. Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>,
	Alexei Starovoitov <ast@kernel.org>,
	"Daniel Borkmann" <daniel@iogearbox.net>,
	Andrii Nakryiko <andrii@kernel.org>, "Martin Lau" <kafai@fb.com>,
	Yonghong Song <yhs@fb.com>,
	John Fastabend <john.fastabend@gmail.com>,
	KP Singh <kpsingh@kernel.org>, Al Viro <viro@zeniv.linux.org.uk>,
	Eric Dumazet <edumazet@google.com>,
	"Christian Brauner" <christian.brauner@ubuntu.com>,
	Rao Shoaib <Rao.Shoaib@oracle.com>,
	Kuniyuki Iwashima <kuniyu@amazon.co.jp>,
	Jakub Sitnicki <jakub@cloudflare.com>,
	"netdev@vger.kernel.org" <netdev@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH bpf v1] unix: fix an issue in unix_shutdown causing the other end read/write failures
Date: Tue, 5 Oct 2021 22:17:37 +0000	[thread overview]
Message-ID: <774AADB0-5AF7-413E-8046-B863826565CB@fb.com> (raw)
In-Reply-To: <f7338e3e-8798-478f-bac9-a86e247e3a13@schaufler-ca.com>


> On Oct 4, 2021, at 5:04 PM, Casey Schaufler <casey@schaufler-ca.com> wrote:
> 
> On 10/4/2021 4:25 PM, Jiang Wang wrote:
>> Commit 94531cfcbe79 ("af_unix: Add unix_stream_proto for sockmap")
>> sets unix domain socket peer state to TCP_CLOSE
>> in unix_shutdown. This could happen when the local end is shutdown
>> but the other end is not. Then the other end will get read or write
>> failures which is not expected.
>> 
>> Fix the issue by setting the local state to shutdown.
>> 
>> Fixes: 94531cfcbe79 (af_unix: Add unix_stream_proto for sockmap)
>> Suggested-by: Cong Wang <cong.wang@bytedance.com>
>> Reported-by: Casey Schaufler <casey@schaufler-ca.com>
>> Signed-off-by: Jiang Wang <jiang.wang@bytedance.com>
> 
> This patch looks like it has fixed the problem. My test cases
> are now getting expected results consistently. Please add any
> or all of:
> 
> Tested-by: Casey Schaufler <casey@schaufler-ca.com>
> Reviewed-by: Casey Schaufler <casey@schaufler-ca.com>

Acked-by: Song Liu <songliubraving@fb.com>

> 
>> ---
>> net/unix/af_unix.c | 9 +++++----
>> 1 file changed, 5 insertions(+), 4 deletions(-)
>> 
>> diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
>> index efac5989edb5..0878ab86597b 100644
>> --- a/net/unix/af_unix.c
>> +++ b/net/unix/af_unix.c
>> @@ -2882,6 +2882,9 @@ static int unix_shutdown(struct socket *sock, int mode)
>> 
>> 	unix_state_lock(sk);
>> 	sk->sk_shutdown |= mode;
>> +	if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) &&
>> +	    mode == SHUTDOWN_MASK)
>> +		sk->sk_state = TCP_CLOSE;
>> 	other = unix_peer(sk);
>> 	if (other)
>> 		sock_hold(other);
>> @@ -2904,12 +2907,10 @@ static int unix_shutdown(struct socket *sock, int mode)
>> 		other->sk_shutdown |= peer_mode;
>> 		unix_state_unlock(other);
>> 		other->sk_state_change(other);
>> -		if (peer_mode == SHUTDOWN_MASK) {
>> +		if (peer_mode == SHUTDOWN_MASK)
>> 			sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP);
>> -			other->sk_state = TCP_CLOSE;
>> -		} else if (peer_mode & RCV_SHUTDOWN) {
>> +		else if (peer_mode & RCV_SHUTDOWN)
>> 			sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN);
>> -		}
>> 	}
>> 	if (other)
>> 		sock_put(other);