Hello Boqun,
On Tue, Jun 24, 2025 at 08:11:01PM -0700, Boqun Feng wrote:
> Erik Lundgren and Breno Leitao reported [1] a case where
> lockdep_unregister_key() can be called from time critical code pathes
> where rntl_lock() may be held. And the synchronize_rcu() in it can slow
> down operations such as using tc to replace a qdisc in a network device.
>
> In fact the synchronize_rcu() in lockdep_unregister_key() is to wait for
> all is_dynamic_key() callers to finish so that removing a key from the
> key hashlist, and we can use shazptr to protect the hashlist as well.
>
> Compared to the proposed solution which replaces synchronize_rcu() with
> synchronize_rcu_expedited(), using shazptr here can achieve the
> same/better synchronization time without the need to send IPI. Hence use
> shazptr here.
>
> Reported-by: Erik Lundgren <elundg...@meta.com>
> Reported-by: Breno Leitao <lei...@debian.org>
> Link:
> https://lore.kernel.org/all/20250321-lockdep-v1-1-78b732d19...@debian.org/ [1]
> Signed-off-by: Boqun Feng <boqun.f...@gmail.com>
First of all, thanks for working to fix the origianl issue. I've been
able to test this in my host, and the gain is impressive.
Before:
# time /usr/sbin/tc qdisc replace dev eth0 root handle 0x1234: mq
real 0m13.195s
user 0m0.001s
sys 0m2.746s
With your patch:
# time /usr/sbin/tc qdisc replace dev eth0 root handle 0x1234: mq
real 0m0.135s
user 0m0.002s
sys 0m0.116s
# time /usr/sbin/tc qdisc replace dev eth0 root handle 0x1: mq
real 0m0.127s
user 0m0.001s
sys 0m0.112s
Please add the following to the series:
Tested-by: Breno Leitao <lei...@debian.org>
