[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Z1n9v7Z6iNJ-wKmq@slm.duckdns.org>
Date: Wed, 11 Dec 2024 11:01:51 -1000
From: Tejun Heo <tj@...nel.org>
To: Ihor Solodrai <ihor.solodrai@...me>
Cc: David Vernet <void@...ifault.com>, sched-ext@...a.com,
kernel-team@...a.com, linux-kernel@...r.kernel.org,
bpf <bpf@...r.kernel.org>
Subject: [PATCH sched_ext/for-6.13-fixes] sched_ext: Fix invalid irq restore
in scx_ops_bypass()
While adding outer irqsave/restore locking, 0e7ffff1b811 ("scx: Fix raciness
in scx_ops_bypass()") forgot to convert an inner rq_unlock_irqrestore() to
rq_unlock() which could re-enable IRQ prematurely leading to the following
warning:
raw_local_irq_restore() called with IRQs enabled
WARNING: CPU: 1 PID: 96 at kernel/locking/irqflag-debug.c:10 warn_bogus_irq_restore+0x30/0x40
...
Sched_ext: create_dsq (enabling)
pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : warn_bogus_irq_restore+0x30/0x40
lr : warn_bogus_irq_restore+0x30/0x40
...
Call trace:
warn_bogus_irq_restore+0x30/0x40 (P)
warn_bogus_irq_restore+0x30/0x40 (L)
scx_ops_bypass+0x224/0x3b8
scx_ops_enable.isra.0+0x2c8/0xaa8
bpf_scx_reg+0x18/0x30
...
irq event stamp: 33739
hardirqs last enabled at (33739): [<ffff8000800b699c>] scx_ops_bypass+0x174/0x3b8
hardirqs last disabled at (33738): [<ffff800080d48ad4>] _raw_spin_lock_irqsave+0xb4/0xd8
Drop the stray _irqrestore().
Signed-off-by: Tejun Heo <tj@...nel.org>
Reported-by: Ihor Solodrai <ihor.solodrai@...me>
Link: http://lkml.kernel.org/r/qC39k3UsonrBYD_SmuxHnZIQLsuuccoCrkiqb_BT7DvH945A1_LZwE4g-5Pu9FcCtqZt4lY1HhIPi0homRuNWxkgo1rgP3bkxa0donw8kV4=@pm.me
Fixes: 0e7ffff1b811 ("scx: Fix raciness in scx_ops_bypass()")
Cc: stable@...r.kernel.org # v6.12
---
kernel/sched/ext.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 7fff1d045477..98519e6d0dcd 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -4763,7 +4763,7 @@ static void scx_ops_bypass(bool bypass)
* sees scx_rq_bypassing() before moving tasks to SCX.
*/
if (!scx_enabled()) {
- rq_unlock_irqrestore(rq, &rf);
+ rq_unlock(rq, &rf);
continue;
}
Powered by blists - more mailing lists