Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Peilin Ye <yepeilin.cs@gmail.com>
To:     Stefano Garzarella <sgarzare@redhat.com>,
        "David S. Miller" <davem@davemloft.net>,
        Eric Dumazet <edumazet@google.com>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>
Cc:     Peilin Ye <peilin.ye@bytedance.com>,
        virtualization@lists.linux-foundation.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org, Peilin Ye <yepeilin.cs@gmail.com>
Subject: [PATCH RFC net-next] vsock: Reschedule connect_work for O_NONBLOCK connect() requests
Date:   Wed,  3 Aug 2022 19:09:25 -0700
Message-Id: <20220804020925.32167-1-yepeilin.cs@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

From: Peilin Ye <peilin.ye@bytedance.com>

An O_NONBLOCK vsock_connect() request may try to reschedule
@connect_work.  Consider the following vsock_connect() requests:

  1. The 1st, non-blocking request schedules @connect_work, which will
     expire after, say, 200 jiffies.  Socket state is now SS_CONNECTING;

  2. Later, the 2nd, blocking request gets interrupted by a signal after
     5 jiffies while waiting for the connection to be established.
     Socket state is back to SS_UNCONNECTED, and @connect_work will
     expire after 100 jiffies;

  3. Now, the 3rd, non-blocking request tries to schedule @connect_work
     again, but @connect_work has already been scheduled, and will
     expire in, say, 50 jiffies.

In this scenario, currently this 3rd request simply decreases the sock
reference count and returns.  Instead, let it reschedules @connect_work
and resets the timeout back to @connect_timeout.

Signed-off-by: Peilin Ye <peilin.ye@bytedance.com>
---
Hi all,

This patch is RFC because it bases on Stefano's WIP fix [1] for a bug [2]
reported by syzbot, and it won't apply on current net-next.  I think it
solves a separate issue.

Please advise, thanks!
Peilin Ye

[1] https://gitlab.com/sgarzarella/linux/-/commit/2d0f0b9cbbb30d58fdcbca7c1a857fd8f3110d61
[2] https://syzkaller.appspot.com/bug?id=cd9103dc63346d26acbbdbf5c6ba9bd74e48c860

 net/vmw_vsock/af_vsock.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
index 194d22291d8b..417e4ad17c03 100644
--- a/net/vmw_vsock/af_vsock.c
+++ b/net/vmw_vsock/af_vsock.c
@@ -1395,7 +1395,7 @@ static int vsock_connect(struct socket *sock, struct sockaddr *addr,
 			/* If the timeout function is already scheduled, ungrab
 			 * the socket refcount to not leave it unbalanced.
 			 */
-			if (!schedule_delayed_work(&vsk->connect_work, timeout))
+			if (mod_delayed_work(system_wq, &vsk->connect_work, timeout))
 				sock_put(sk);
 
 			/* Skip ahead to preserve error code set above. */
-- 
2.20.1