On Fri, Jun 25, 2021 at 02:23:16PM +0530, Bharata B Rao wrote:

On Fri, Jun 25, 2021 at 09:28:09AM +0200, Peter Zijlstra wrote:

quoted

On Fri, Jun 25, 2021 at 11:16:08AM +0530, Srikar Dronamraju wrote:

quoted

* Bharata B Rao [off-list ref] [2021-06-24 21:25:09]:

quoted

A PowerPC KVM guest gets the following BUG message when booting
linux-next-20210623:

smp: Bringing up secondary CPUs ...
BUG: scheduling while atomic: swapper/1/0/0x00000000

'funny', your preempt_count is actually too low. The check here is for
preempt_count() == DISABLE_OFFSET (aka. 1 when PREEMPT=y), but you have
0.

quoted

quoted

no locks held by swapper/1/0.
Modules linked in:
CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.13.0-rc7-next-20210623
Call Trace:
[c00000000ae5bc20] [c000000000badc64] dump_stack_lvl+0x98/0xe0 (unreliable)
[c00000000ae5bc60] [c000000000210200] __schedule_bug+0xb0/0xe0
[c00000000ae5bcd0] [c000000001609e28] __schedule+0x1788/0x1c70
[c00000000ae5be20] [c00000000160a8cc] schedule_idle+0x3c/0x70
[c00000000ae5be50] [c00000000022984c] do_idle+0x2bc/0x420
[c00000000ae5bf00] [c000000000229d88] cpu_startup_entry+0x38/0x40
[c00000000ae5bf30] [c0000000000666c0] start_secondary+0x290/0x2a0
[c00000000ae5bf90] [c00000000000be54] start_secondary_prolog+0x10/0x14

<The above repeats for all the secondary CPUs>

smp: Brought up 2 nodes, 16 CPUs
numa: Node 0 CPUs: 0-7
numa: Node 1 CPUs: 8-15

This seems to have started from next-20210521 and isn't seen on
next-20210511.

Bharata,

I think the regression is due to Commit f1a0a376ca0c ("sched/core:
Initialize the idle task with preemption disabled")

So that extra preempt_disable() that got removed would've incremented it
to 1 and then things would've been fine.

Except.. Valentin changed things such that preempt_count() should've
been inittialized to 1, instead of 0, but for some raisin that didn't
stick.. what gives.

So we have init_idle(p) -> init_idle_preempt_count(p) ->
task_thread_info(p)->preempt_count = PREEMPT_DISABLED;

But somehow, by the time you're running start_secondary(), that's gotten
to be 0 again. Does DEBUG_PREEMPT give more clues?

PREEMPTION is off here.

You mean: CONFIG_PREEMPTION=n, what about CONFIG_PREEMPT_COUNT?

Because if both are =n, then I don't see how that warning could trigger.
in_atomic_preempt_off() would then result in prempt_count() == 0, and
per the print above, it *is* 0.