On Mon, 21 Apr 2025 13:50:52 +0200 Holger Hoffstätte holger@applied-asynchrony.com wrote:
On 2025-04-21 11:40, Alan J. Wylie wrote:
#regzbot introduced: 6.14.2..6.14.3
Since 6.14.3 I have been seeing random panics, all in htb_dequeue. 6.14.2 was fine.
6.14.3 contains: "codel: remove sch->q.qlen check before qdisc_tree_reduce_backlog()" aka https://web.git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/...
Is your HTB backed by fq_codel by any chance?
Yes
# grep fq 91-tc.sh quantum=300 # fq_codel quantum 300 gives a boost to interactive flows modprobe sch_fq_codel tc qdisc add dev "$ext_ingress" parent 1:11 fq_codel quantum $quantum ecn tc qdisc add dev "$ext" parent 1:11 fq_codel quantum $quantum noecn
If so, try either reverting the above or adding: "sch_htb: make htb_qlen_notify() idempotent" aka https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commi...
which was successfully not added to 6.14.3, along with the rest of the series: https://lore.kernel.org/all/20250403211033.166059-2-xiyou.wangcong@gmail.com...
"successfully not added"?
$ git cherry-pick 5ba8b837b522d7051ef81bacf3d95383ff8edce5 [linux-6.14.y 2285c724bf7d] sch_htb: make htb_qlen_notify() idempotent Author: Cong Wang xiyou.wangcong@gmail.com Date: Thu Apr 3 14:10:23 2025 -0700 1 file changed, 2 insertions(+)
It will take a while (perhaps days?) before I can confirm success.
Thanks Alan