On 01/31/2015 05:55 PM, David Miller wrote:

From: Eric Dumazet <redacted>
Date: Thu, 29 Jan 2015 17:30:12 -0800

quoted

From: Eric Dumazet <edumazet@google.com>

Doing the following commands on a non idle network device
panics the box instantly, because cpu_bstats gets overwritten
by stats.

tc qdisc add dev eth0 root <your_favorite_qdisc>
... some traffic (one packet is enough) ...
tc qdisc replace dev eth0 root est 1sec 4sec <your_favorite_qdisc>

  ...

quoted

Lets play safe and not use an union : percpu 'pointers' are mostly read
anyway, and we have typically few qdiscs per host.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Fixes: 22e0f8b9322c ("net: sched: make bstats per cpu and estimator RCU safe")

Applied and queued up for -stable, thanks Eric.

Thanks Eric, looks like my "replace" test scripts never sent traffic
before the replace call.

Also I haven't seen any RCU splats in awhile so I'll send out a patch
to drop the qdisc_lock on ing_filter shortly.

-- 
John Fastabend         Intel Corporation