Re: [PATCH net] tcp: Remove sock_owned_by_user() test in tcp_child_process().

From:   Eric Dumazet <redacted>
Date:   Wed, 8 Dec 2021 10:30:49 -0800

On Tue, Dec 7, 2021 at 9:16 PM Kuniyuki Iwashima [off-list ref] wrote:

quoted

While creating a child socket, before v2.3.41, we used to call
bh_lock_sock() later than now; it was called just before
tcp_rcv_state_process().  The full socket was put into an accept queue
and exposed to other CPUs before bh_lock_sock() so that process context
might have acquired the lock by then.  Thus, we had to check if any
process context was accessing the socket before tcp_rcv_state_process().

We can see this code in tcp_v4_do_rcv() of v2.3.14. [0]

        if (sk->state == TCP_LISTEN) {
                struct sock *nsk;

                nsk = tcp_v4_hnd_req(sk, skb);
                ...
                if (nsk != sk) {
                        bh_lock_sock(nsk);
                        if (nsk->lock.users != 0) {
                                ...
                                sk_add_backlog(nsk, skb);
                                bh_unlock_sock(nsk);
                                return 0;
                        }
                        ...
                }
        }

        if (tcp_rcv_state_process(sk, skb, skb->h.th, skb->len))
                goto reset;

However, in 2.3.15, this lock.users test was replaced with BUG_TRAP() by
mistake. [1]

                if (nsk != sk) {
                        ...
                        BUG_TRAP(nsk->lock.users == 0);
                        ...
                        ret = tcp_rcv_state_process(nsk, skb, skb->h.th, skb->len);
                        ...
                        bh_unlock_sock(nsk);
                        ...
                        return 0;
                }

Fortunately, the test was back in 2.3.41. [2]  Then, related code was
packed into tcp_child_process() with comments, which remains until now.

What is interesting in v2.3.41 is that the bh_lock_sock() was moved to
tcp_create_openreq_child() and placed just after sock_lock_init().
Thus, the lock is never acquired until tcp_rcv_state_process() by process
contexts.  The bh_lock_sock() is now in sk_clone_lock() and the rule does
not change.

As of now, alas, it is not possible to reach the commented path by the
change.  Let's remove the remnant of the old days.

[0]: https://cdn.kernel.org/pub/linux/kernel/v2.3/linux-2.3.14.tar.gz
[1]: https://cdn.kernel.org/pub/linux/kernel/v2.3/patch-2.3.15.gz
[2]: https://cdn.kernel.org/pub/linux/kernel/v2.3/patch-2.3.41.gz

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")

I do not think this patch qualifies as a stable candidate.

At best this is a cleanup.

At worst this could add a bug.

I would advise adding a WARN_ON_ONCE() there for at least one release
so that syzbot can validate for you if this is dead code or not.

Thanks for review.
I will add a WARN_ON_ONCE() and respin for net-next.

TCP_SYN_RECV is not TCP_NEW_SYN_RECV

Right, TCP_SYN_RECV is not the case.
"While creating a child socket," was a bit misleading.
I will clarify that is for TCP_NEW_SYN_RECV case and SYN cookie case.

Thanks.

quoted

Signed-off-by: Kuniyuki Iwashima <redacted>
---
 net/ipv4/tcp_minisocks.c | 18 ++++++------------
 1 file changed, 6 insertions(+), 12 deletions(-)

diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index 7c2d3ac2363a..b4a1f8728093 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c

@@ -833,18 +833,12 @@ int tcp_child_process(struct sock *parent, struct sock *child,
        sk_mark_napi_id_set(child, skb);

        tcp_segs_in(tcp_sk(child), skb);
-       if (!sock_owned_by_user(child)) {
-               ret = tcp_rcv_state_process(child, skb);
-               /* Wakeup parent, send SIGIO */
-               if (state == TCP_SYN_RECV && child->sk_state != state)
-                       parent->sk_data_ready(parent);
-       } else {
-               /* Alas, it is possible again, because we do lookup
-                * in main socket hash table and lock on listening
-                * socket does not protect us more.
-                */
-               __sk_add_backlog(child, skb);
-       }
+
+       ret = tcp_rcv_state_process(child, skb);
+
+       /* Wakeup parent, send SIGIO */
+       if (state == TCP_SYN_RECV && child->sk_state != state)
+               parent->sk_data_ready(parent);

        bh_unlock_sock(child);
        sock_put(child);
--

2.30.2