[PATCH resend] vsock/virtio: avoid potential deadlock when vsock device remove

From: Longpeng(Mike)
Date: Thu Aug 12 2021 - 01:31:17 EST


There's a potential deadlock case when remove the vsock device or
process the RESET event:

vsock_for_each_connected_socket:
spin_lock_bh(&vsock_table_lock) ----------- (1)
...
virtio_vsock_reset_sock:
lock_sock(sk) --------------------- (2)
...
spin_unlock_bh(&vsock_table_lock)

lock_sock() may do initiative schedule when the 'sk' is owned by
other thread at the same time, we would receivce a warning message
that "scheduling while atomic".

Even worse, if the next task (selected by the scheduler) try to
release a 'sk', it need to request vsock_table_lock and the deadlock
occur, cause the system into softlockup state.
Call trace:
queued_spin_lock_slowpath
vsock_remove_bound
vsock_remove_sock
virtio_transport_release
__vsock_release
vsock_release
__sock_release
sock_close
__fput
____fput

So we should not require sk_lock in this case, just like the behavior
in vhost_vsock or vmci.

Cc: Stefan Hajnoczi <stefanha@xxxxxxxxxx>
Cc: Stefano Garzarella <sgarzare@xxxxxxxxxx>
Cc: "David S. Miller" <davem@xxxxxxxxxxxxx>
Cc: Jakub Kicinski <kuba@xxxxxxxxxx>
Signed-off-by: Longpeng(Mike) <longpeng2@xxxxxxxxxx>
---
net/vmw_vsock/virtio_transport.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index e0c2c99..4f7c99d 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c
@@ -357,11 +357,14 @@ static void virtio_vsock_event_fill(struct virtio_vsock *vsock)

static void virtio_vsock_reset_sock(struct sock *sk)
{
- lock_sock(sk);
+ /* vmci_transport.c doesn't take sk_lock here either. At least we're
+ * under vsock_table_lock so the sock cannot disappear while we're
+ * executing.
+ */
+
sk->sk_state = TCP_CLOSE;
sk->sk_err = ECONNRESET;
sk_error_report(sk);
- release_sock(sk);
}

static void virtio_vsock_update_guest_cid(struct virtio_vsock *vsock)
--
1.8.3.1