Re: Regression in linux-next

From: Thomas Gleixner
Date: Mon Oct 16 2017 - 07:05:56 EST


On Mon, 16 Oct 2017, Petri Latvala wrote:

Please CC LKML next time when you report a problem. I almost missed this
thread.

> [  174.561006] kernel BUG at arch/x86/kernel/apic/vector.c:154!
> [  174.716682]  assign_vector_locked+0x9c/0x150
> [  174.721229]  apic_set_affinity+0x47/0x70
> [  174.725440]  ioapic_set_affinity+0x1a/0x60
> [  174.729838]  irq_do_set_affinity+0x18/0x60
> [  174.734206]  irq_migrate_all_off_this_cpu+0x136/0x270
> [  174.739647]  fixup_irqs+0x2f/0x130
> [  174.743312]  cpu_disable_common+0x1c7/0x1e0
> [  174.747792]  native_cpu_disable+0x20/0x30
> [  174.752073]  take_cpu_down+0x3c/0xa0
> [  174.755869]  multi_cpu_stop+0x8e/0xb0
> [  174.759762]  ? cpu_stop_queue_work+0x90/0x90
> [  174.764360]  cpu_stopper_thread+0x8a/0x100
> [  174.768758]  smpboot_thread_fn+0x165/0x230
> [  174.773168]  kthread+0x10c/0x140

Can you please pick up linux-next of today and merge the x86/apic branch
from the tip-tree into it:

git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git x86/apic

There is a fix which might be related. Though I have no idea yet how the
above can happen.

Please enable tracing in the kernel configuration and enable the vector
tracepoints.

# for D in `ls -d /sys/kernel/debug/tracing/events/irq_vectors/vector_*`; do echo 1 > $D/enable; done

Also please do

# echo 1 > /proc/sys/kernel/ftrace_dump_on_oops

That should dump the trace buffer when the bug hits. I try to reproduce
myself.

Thanks,

tglx