On 29/08/17 21:09, Eric Dumazet wrote:
> On Tue, 2017-08-29 at 20:58 +0300, Nikolay Aleksandrov wrote:
>> The commit below added a call to the ->destroy() callback for all qdiscs
>> which failed in their ->init(), but some were not prepared for such
>> change and can't handle partially initialized qdisc. HTB is one of them
>> and if any error occurs before the qdisc watchdog timer and qdisc work are
>> initialized then we can hit either a null ptr deref (timer->base) when
>> canceling in ->destroy or lockdep error info about trying to register
>> a non-static key and a stack dump. So to fix these two move the watchdog
>> timer and workqueue init before anything that can err out.
>> To reproduce userspace needs to send broken htb qdisc create request,
>> tested with a modified tc (q_htb.c).
>
>> Note that probably this bug goes further back because the default qdisc
>> handling always calls ->destroy on init failure too.
>>
>> Fixes: 87b60cfacf9f ("net_sched: fix error recovery at qdisc creation")
>> Signed-off-by: Nikolay Aleksandrov <niko...@cumulusnetworks.com>
>> ---
>> Always calling qdisc destroy on init failure in the default qdisc handling
>> was added in commit 0fbbeb1ba43b. I'm not sure if I should include that
>> one as fixes tag.
>
> Well, we probably need to audit init/destroy not only in net/sched, but
> other parts of networking stack.
>
I'm not sure I follow, I hit this while working on a net/sched/ patch and had
to error
out in the init() function.
> What about the qdisc_skb_head_init(&q->direct_queue) call ?
>
> I am surprised you do not crash in __skb_queue_purge(&q->direct_queue);
>
Hm, do you mean in __qdisc_reset_queue() ?
I have only tried/seen the crash happen on qdisc add.
A much simpler and easier bug is sch_multiq (ethX non-multiq):
$ tc qdisc add dev ethX root multiq
(error EOPNOTSUPP + double free of ->queues due to free in init())
and e.g. $ ip l add dumdum type dummy
to see a crash due to the corrupted memory.
