netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Cong Wang <xiyou.wangcong@gmail.com>
To: Jakub Sitnicki <jakub@cloudflare.com>
Cc: Linux Kernel Network Developers <netdev@vger.kernel.org>,
	bpf <bpf@vger.kernel.org>,
	duanxiongchun@bytedance.com,
	Dongdong Wang <wangdongdong.6@bytedance.com>,
	Jiang Wang <jiang.wang@bytedance.com>,
	Cong Wang <cong.wang@bytedance.com>,
	John Fastabend <john.fastabend@gmail.com>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Lorenz Bauer <lmb@cloudflare.com>
Subject: Re: [Patch bpf-next v4 04/11] skmsg: avoid lock_sock() in sk_psock_backlog()
Date: Sat, 13 Mar 2021 09:32:17 -0800	[thread overview]
Message-ID: <CAM_iQpVmtHPqzGHEUPhtVroxCeWSBvahKMrbLrEq4gNNVGq2zg@mail.gmail.com> (raw)
In-Reply-To: <87y2es37i3.fsf@cloudflare.com>

On Fri, Mar 12, 2021 at 4:02 AM Jakub Sitnicki <jakub@cloudflare.com> wrote:
>
> On Wed, Mar 10, 2021 at 06:32 AM CET, Cong Wang wrote:
> > diff --git a/net/core/sock_map.c b/net/core/sock_map.c
> > index dd53a7771d7e..26ba47b099f1 100644
> > --- a/net/core/sock_map.c
> > +++ b/net/core/sock_map.c
> > @@ -1540,6 +1540,7 @@ void sock_map_close(struct sock *sk, long timeout)
> >       saved_close = psock->saved_close;
> >       sock_map_remove_links(sk, psock);
> >       rcu_read_unlock();
> > +     sk_psock_purge(psock);
> >       release_sock(sk);
> >       saved_close(sk, timeout);
> >  }
>
> Nothing stops sk_psock_backlog from running after sk_psock_purge:
>
>
> CPU 1                                                   CPU 2
>
> sk_psock_skb_redirect()
>   sk_psock(sk_other)
>   sock_flag(sk_other, SOCK_DEAD)
>   sk_psock_test_state(psock_other,
>                       SK_PSOCK_TX_ENABLED)
>                                                         sk_psock_purge()
>   skb_queue_tail(&psock_other->ingress_skb, skb)
>   schedule_work(&psock_other->work)
>
>
> And sock_orphan can run while we're in sendmsg/sendpage_unlocked:
>
>
> CPU 1                                                   CPU 2
>
> sk_psock_backlog
>   ...
>   sendmsg_unlocked
>     sock = sk->sk_socket
>                                                         tcp_close
>                                                           __tcp_close
>                                                             sock_orphan
>     kernel_sendmsg(sock, msg, vec, num, size)
>
>
> So, after this change, without lock_sock in sk_psock_backlog, we will
> not block tcp_close from running.
>
> This makes me think that the process socket can get released from under
> us, before kernel_sendmsg/sendpage runs.

I think you are right, I thought socket is orphaned in inet_release(), clearly
I was wrong. But, I'd argue in the above scenario, the packet should not
be even queued in the first place, as SK_PSOCK_TX_ENABLED is going
to be cleared, so I think the right fix is probably to make clearing psock
state and queuing the packet under a spinlock.

Thanks.

  reply	other threads:[~2021-03-13 17:33 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-03-10  5:32 [Patch bpf-next v4 00/11] sockmap: introduce BPF_SK_SKB_VERDICT and support UDP Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 01/11] skmsg: lock ingress_skb when purging Cong Wang
2021-03-11 10:52   ` Jakub Sitnicki
2021-03-10  5:32 ` [Patch bpf-next v4 02/11] skmsg: introduce a spinlock to protect ingress_msg Cong Wang
2021-03-11 11:28   ` Jakub Sitnicki
2021-03-12  0:45     ` Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 03/11] skmsg: introduce skb_send_sock() for sock_map Cong Wang
2021-03-11 11:42   ` Jakub Sitnicki
2021-03-12  0:47     ` Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 04/11] skmsg: avoid lock_sock() in sk_psock_backlog() Cong Wang
2021-03-12 12:02   ` Jakub Sitnicki
2021-03-13 17:32     ` Cong Wang [this message]
2021-03-15 20:55       ` Jakub Sitnicki
2021-03-10  5:32 ` [Patch bpf-next v4 05/11] sock_map: introduce BPF_SK_SKB_VERDICT Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 06/11] sock: introduce sk->sk_prot->psock_update_sk_prot() Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 07/11] udp: implement ->read_sock() for sockmap Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 08/11] skmsg: extract __tcp_bpf_recvmsg() and tcp_bpf_wait_data() Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 09/11] udp: implement udp_bpf_recvmsg() for sockmap Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 10/11] sock_map: update sock type checks for UDP Cong Wang
2021-03-10  5:32 ` [Patch bpf-next v4 11/11] selftests/bpf: add a test case for udp sockmap Cong Wang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAM_iQpVmtHPqzGHEUPhtVroxCeWSBvahKMrbLrEq4gNNVGq2zg@mail.gmail.com \
    --to=xiyou.wangcong@gmail.com \
    --cc=bpf@vger.kernel.org \
    --cc=cong.wang@bytedance.com \
    --cc=daniel@iogearbox.net \
    --cc=duanxiongchun@bytedance.com \
    --cc=jakub@cloudflare.com \
    --cc=jiang.wang@bytedance.com \
    --cc=john.fastabend@gmail.com \
    --cc=lmb@cloudflare.com \
    --cc=netdev@vger.kernel.org \
    --cc=wangdongdong.6@bytedance.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).