bpf.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: John Fastabend <john.fastabend@gmail.com>
To: Liu Jian <liujian56@huawei.com>,
	john.fastabend@gmail.com, daniel@iogearbox.net,
	jakub@cloudflare.com, lmb@cloudflare.com, edumazet@google.com,
	davem@davemloft.net, yoshfuji@linux-ipv6.org, dsahern@kernel.org,
	kuba@kernel.org, ast@kernel.org, andrii@kernel.org, kafai@fb.com,
	songliubraving@fb.com, yhs@fb.com, kpsingh@kernel.org,
	netdev@vger.kernel.org, bpf@vger.kernel.org
Cc: liujian56@huawei.com
Subject: RE: [PATCH] tcp_bpf: Fix one concurrency problem in the tcp_bpf_send_verdict function
Date: Thu, 30 Sep 2021 15:25:23 -0700	[thread overview]
Message-ID: <61563953e731f_6c4e420814@john-XPS-13-9370.notmuch> (raw)
In-Reply-To: <20210929084529.96583-1-liujian56@huawei.com>

Liu Jian wrote:
> In the following cases:
> We need to redirect the first msg to sock1 and the second msg to sock2.
> The sock lock needs to be released at __SK_REDIRECT and to get another
> sock lock, this will cause the probability that psock->eval is not set to
> __SK_NONE when the second msg comes.
> 
> If psock does not set apple bytes, fix this by do the cleanup before
> releasing the sock lock. And keep the original logic in other cases.

It took me sometime to figure out the above description. Please include
a bit more details here this needs to be backported so we want to be
very clear what the error  is and how to trigger it.

In this case we should list the flow to show how the interleaving of
msgs breaks.

"
With two Msgs, msgA and msgB and a user doing nonblocking sendmsg calls
(or multiple cores) on a single socket 'sk' we could get the following
flow.

 msgA, sk                               msgB, sk
 -----------                            ---------------
 tcp_bpf_sendmsg()
 lock(sk)
 psock = sk->psock
                                        tcp_bpf_sendmsg()
                                        lock(sk) ... blocking
 tcp_bpf_send_verdict
 if (psock->eval == NONE)
   psock->eval = sk_psock_msg_verdict
 ..
 < handle SK_REDIRECT case >
   release_sock(sk)                     < lock dropped so grab here >
   ret = tcp_bpf_sendmsg_redir
                                        psock = sk->psock
                                        tcp_bpf_send_verdict
 lock_sock(sk) ... blocking on B
                                        if (psock->eval == NONE) <- boom.
                                         psock->eval will have msgA state

The problem here is we dropped the lock on msgA and grabbed it with msgB.
Now we have old state in psock and importantly psock->eval has not
been cleared. So msgB will run whatever action was done on A and the
verdict program may never see it.
"

Showing the flow makes it painfully obvious why dropping that lock
with old state is broken.


> 
> Signed-off-by: Liu Jian <liujian56@huawei.com>

We need a fixes tag as well so we can backport this.

> ---
>  net/ipv4/tcp_bpf.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 
> diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c
> index d3e9386b493e..02442e43ac4d 100644
> --- a/net/ipv4/tcp_bpf.c
> +++ b/net/ipv4/tcp_bpf.c
> @@ -232,6 +232,7 @@ static int tcp_bpf_send_verdict(struct sock *sk, struct sk_psock *psock,
>  	bool cork = false, enospc = sk_msg_full(msg);
>  	struct sock *sk_redir;
>  	u32 tosend, delta = 0;
> +	u32 eval = __SK_NONE;
>  	int ret;
>  
>  more_data:
> @@ -274,6 +275,12 @@ static int tcp_bpf_send_verdict(struct sock *sk, struct sk_psock *psock,
>  		break;
>  	case __SK_REDIRECT:
>  		sk_redir = psock->sk_redir;
> +		if (!psock->apply_bytes) {
> +			/* Clean up before releasing the sock lock. */
> +			eval = psock->eval;
> +			psock->eval = __SK_NONE;
> +			psock->sk_redir = NULL;
> +		}

We need to move above chunk below sk_msg_apply_bytes() so we account for
the bytes and if we zero apply bytes with this send we clear the psock
state. Otherwise we could have the same issue with stale state on the
boundary where apply bytes is met.

>  		sk_msg_apply_bytes(psock, tosend);

<-- put above chunk here.

>  		if (psock->cork) {
>  			cork = true;

Interestingly, I caught the race with cork state, but missed it with
the eval case. Likely because our program redirected to a single sk.

> @@ -281,7 +288,12 @@ static int tcp_bpf_send_verdict(struct sock *sk, struct sk_psock *psock,
>  		}
>  		sk_msg_return(sk, msg, tosend);
>  		release_sock(sk);
> +
>  		ret = tcp_bpf_sendmsg_redir(sk_redir, msg, tosend, flags);
> +
> +		if (eval == __SK_REDIRECT)

Is the 'if' needed? we are in this case because eval is SK_REDIRECT.

> +			sock_put(sk_redir);
> +
>  		lock_sock(sk);
>  		if (unlikely(ret < 0)) {
>  			int free = sk_msg_free_nocharge(sk, msg);
> -- 
> 2.17.1
> 

  reply	other threads:[~2021-09-30 22:25 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-29  8:45 Liu Jian
2021-09-30 22:25 ` John Fastabend [this message]
2021-10-04  4:21   ` liujian (CE)

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=61563953e731f_6c4e420814@john-XPS-13-9370.notmuch \
    --to=john.fastabend@gmail.com \
    --cc=andrii@kernel.org \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=davem@davemloft.net \
    --cc=dsahern@kernel.org \
    --cc=edumazet@google.com \
    --cc=jakub@cloudflare.com \
    --cc=kafai@fb.com \
    --cc=kpsingh@kernel.org \
    --cc=kuba@kernel.org \
    --cc=liujian56@huawei.com \
    --cc=lmb@cloudflare.com \
    --cc=netdev@vger.kernel.org \
    --cc=songliubraving@fb.com \
    --cc=yhs@fb.com \
    --cc=yoshfuji@linux-ipv6.org \
    --subject='RE: [PATCH] tcp_bpf: Fix one concurrency problem in the tcp_bpf_send_verdict function' \
    /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

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).