Le Mon, Feb 10, 2025 at 12:59:42PM +0100, Paul Menzel a écrit :
> Dear Michał,
>
>
> Thank you for your reply.
>
> Am 10.02.25 um 12:45 schrieb Michał Pecio:
>
> > > > > > > > > On Dell XPS 13 9360/0596KF, BIOS 2.21.0 06/02/2022, with Linux
> > > > > > > > > 6.9-rc2+
> >
> > > Just for the record, I am still seeing this with 6.14.0-rc1
> >
> > Is this a regression? If so, which versions were not affected?
>
> Unfortunately, I do not know. Right now, my logs go back until September
> 2024.
>
> Sep 22 13:08:04 abreu kernel: Linux version 6.11.0-07273-g1e7530883cd2
> (build@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx) (gcc (Debian 14.2.0-5) 14.2.0, GNU ld
> (GNU Binutils for Debian) 2.43.1) #12 SMP PREEMPT_DYNAMIC Sun Sep 22
> 09:57:36 CEST 2024
>
> > How hard to reproduce? Wasn't it during resume from hibernation?
>
> It’s not easy to reproduce, and I believe it’s not related with resuming
> from hibernation (which I do not use) or ACPI S3 suspend. I think, I can
> force it more, when having the USB-C adapter with only the network cable
> plugged into it, and then running `sudo powertop --auto-tune`. But sometimes
> it seems unrelated.
>
> > IRQ isuses may be a red herring, this code here is a busy wait under
> > spinlock. There are a few of those, they cause various problems.
> >
> > if (xhci_handshake(&xhci->op_regs->status,
> > STS_RESTORE, 0, 100 * 1000)) {
> > xhci_warn(xhci, "WARN: xHC restore state timeout\n");
> > spin_unlock_irq(&xhci->lock);
> > return -ETIMEDOUT;
> > }
> >
> > This thing timing out may be close to the root cause of everything.
>
> Interesting. Hopefully the USB folks have an idea.
Handler #08 is NET_RX. So something raised the NET_RX on some non-appropriate
place, perhaps...
Can I ask you one more trace dump?
I need:
echo 1 > /sys/kernel/tracing/events/irq/softirq_raise/enable
echo 1 > /sys/kernel/tracing/options/stacktrace
Unfortunately this will also involve a small patch:
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index fa058510af9c..accd2eb8c927 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -1159,6 +1159,9 @@ static bool report_idle_softirq(void)
if (local_bh_blocked())
return false;
+ trace_printk("STOP\n");
+ trace_dump_stack(0);
+ tracing_off();
pr_warn("NOHZ tick-stop error: local softirq work is pending, handler #%02x!!!\n",
pending);
ratelimit++;
Thanks.
