Re: [PATCH net] ppp: don't override sk->sk_state in pppoe_flush_dev()

[Guillaume Nault <g.nault@alphalink.fr>]

On Tue, Oct 13, 2015 at 05:13:54AM +0300, Denys Fedoryshchenko wrote:
> On 2015-10-07 15:12, Guillaume Nault wrote:
> >On Mon, Oct 05, 2015 at 02:08:44PM +0200, Guillaume Nault wrote:
> >> 	if (po) {
> >> 		struct sock *sk = sk_pppox(po);
> >>
> >>-		bh_lock_sock(sk);
> >>-
> >>-		/* If the user has locked the socket, just ignore
> >>-		 * the packet.  With the way two rcv protocols hook into
> >>-		 * one socket family type, we cannot (easily) distinguish
> >>-		 * what kind of SKB it is during backlog rcv.
> >>-		 */
> >>-		if (sock_owned_by_user(sk) == 0) {
> >>-			/* We're no longer connect at the PPPOE layer,
> >>-			 * and must wait for ppp channel to disconnect us.
> >>-			 */
> >>-			sk->sk_state = PPPOX_ZOMBIE;
> >>-		}
> >>-
> >>-		bh_unlock_sock(sk);
> >> 		if (!schedule_work(&po->proto.pppoe.padt_work))
> >> 			sock_put(sk);
> >> 	}
> >>
> >Finally, I think I'll keep this approach for net-next, to completely
> >remove PPPOX_ZOMBIE.
> >For now, let's just avoid any assumption about the relationship between
> >the PPPOX_ZOMBIE state and the value of po->pppoe_dev, as suggested by
> >Matt.
> >
> >Denys, can you let me know if your issue goes away with the following
> >patch?
> >---
> >diff --git a/drivers/net/ppp/pppoe.c b/drivers/net/ppp/pppoe.c
> >index 2ed7506..5e0b432 100644
> >--- a/drivers/net/ppp/pppoe.c
> >+++ b/drivers/net/ppp/pppoe.c
> >@@ -589,7 +589,7 @@ static int pppoe_release(struct socket *sock)
> >
> > 	po = pppox_sk(sk);
> >
> >-	if (sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND | PPPOX_ZOMBIE)) {
> >+	if (po->pppoe_dev) {
> > 		dev_put(po->pppoe_dev);
> > 		po->pppoe_dev = NULL;
> > 	}
> I just got OK to upgrade server yesterday, for now around 12 hours working
> fine. I need 1-2 more days, and maybe will upgrade few more servers to say
> for sure, if it is ok or not.
> Sorry for delay, just it is production servers and at current situation they
> cannot tolerate significant downtime.
> 
That's ok. I'll send an official patch when you consider the issue to
be definitely fixed.