On Sun, 20 Jul 2008, David Miller wrote:

Alexey, please try this patch:

atl1: Do not wake queue before queue has been started.

David, can we please make this all a bit less fragile?

There are _millions_ of network drivers, and these changes seem to have 
broken not just common drivers, but also drivers that "work" seem to now 
have broken suspend/resume.

There's at least one report of suspend apparently oopsing now, and I 
assume it's basically the same thing - it was bisected down to that same 
37437bb2e1ae8af470dfcd5b4ff454110894ccaf commit ("pkt_sched: Schedule 
qdiscs instead of netdev_queue.")

Why is it so unnecessarily fragile to begin with? Especially for stuff 
that happens at bootup or suspend, doing a BUG_ON() is _particularly_ 
painful, because a dead machine means that you cannot get any logs or 
anything else.

So wouldn't it be *much* better to do something like the appended, and at 
least try to limp on, and maybe have a system that people can get logs 
out of?

			Linus

---
 net/core/dev.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/net/core/dev.c b/net/core/dev.c
index 2eed17b..43ab4f5 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c

@@ -1325,7 +1325,8 @@ static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
 
 void __netif_schedule(struct Qdisc *q)
 {
-	BUG_ON(q == &noop_qdisc);
+	if (WARN_ON_ONCE(q == &noop_qdisc))
+		return;
 
 	if (!test_and_set_bit(__QDISC_STATE_SCHED, &q->state)) {
 		struct softnet_data *sd;

---