Thread (28 messages) 28 messages, 4 authors, 2017-12-21

Re: RCU callback crashes

flat view

From: Jakub Kicinski <hidden>
Date: 2017-12-21 00:37:17

On Wed, 20 Dec 2017 16:03:49 -0800, Cong Wang wrote:
quoted hunk ↗ jump to hunk
On Wed, Dec 20, 2017 at 10:31 AM, Cong Wang [off-list ref] wrote:
quoted
On Wed, Dec 20, 2017 at 10:17 AM, Cong Wang [off-list ref] wrote:  
quoted
I guess it is q->miniqp which is freed in qdisc_graft() without properly
waiting for rcu readers?  
It is probably so, the call_rcu_bh(&miniq_old->rcu, mini_qdisc_rcu_func)
in the end of mini_qdisc_pair_swap() is invoked on miniq_old->rcu,
but miniq is being freed, no rcu barrier waits for it...

You can try to add a rcu_barrier_bh() at the end to see if this crash
is gone, but I don't think people like adding yet another rcu barrier...  
Hi, Jakub

Can you test the following fix? I am not a fan of rcu barrier but we
already have one so...
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index 876fab2604b8..1b68fedea124 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -1240,6 +1240,8 @@ void mini_qdisc_pair_swap(struct mini_Qdisc_pair *miniqp,

        if (!tp_head) {
                RCU_INIT_POINTER(*miniqp->p_miniq, NULL);
+               /* Wait for existing flying RCU callback before being freed. */
+               rcu_barrier_bh();
                return;
        }
Looks good after 30 minutes, feel free to add if you post officially:

Tested-by: Jakub Kicinski <redacted>
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help