net_sched: prio: fix a race in prio_tune()
authorEric Dumazet <edumazet@google.com>
Wed, 11 Jun 2025 11:15:11 +0000 (11:15 +0000)
committerJakub Kicinski <kuba@kernel.org>
Thu, 12 Jun 2025 15:05:49 +0000 (08:05 -0700)
commitd35acc1be3480505b5931f17e4ea9b7617fea4d3
treed44cad1a7efeca952eafc00c513ad15c50758e29
parent7ca52541c05c832d32b112274f81a985101f9ba8
net_sched: prio: fix a race in prio_tune()

Gerrard Tai reported a race condition in PRIO, whenever SFQ perturb timer
fires at the wrong time.

The race is as follows:

CPU 0                                 CPU 1
[1]: lock root
[2]: qdisc_tree_flush_backlog()
[3]: unlock root
 |
 |                                    [5]: lock root
 |                                    [6]: rehash
 |                                    [7]: qdisc_tree_reduce_backlog()
 |
[4]: qdisc_put()

This can be abused to underflow a parent's qlen.

Calling qdisc_purge_queue() instead of qdisc_tree_flush_backlog()
should fix the race, because all packets will be purged from the qdisc
before releasing the lock.

Fixes: 7b8e0b6e6599 ("net: sched: prio: delay destroying child qdiscs on change")
Reported-by: Gerrard Tai <gerrard.tai@starlabs.sg>
Suggested-by: Gerrard Tai <gerrard.tai@starlabs.sg>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250611111515.1983366-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/sched/sch_prio.c