Re: [PATCH v2 04/11] sched: Simplify wake_up_*idle*()

[Marek Szyprowski <m.szyprowski@samsung.com>]

Hi

On 29.09.2021 17:17, Peter Zijlstra wrote:
> Simplify and make wake_up_if_idle() more robust, also don't iterate
> the whole machine with preempt_disable() in it's caller:
> wake_up_all_idle_cpus().
>
> This prepares for another wake_up_if_idle() user that needs a full
> do_idle() cycle.
>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>

This patch landed recently in linux-next as commit 8850cb663b5c ("sched: 
Simplify wake_up_*idle*()"). It causes the following warning on the 
arm64 virt machine under qemu during the system suspend/resume cycle:

--->8---

  printk: Suspending console(s) (use no_console_suspend to debug)

  ============================================
  WARNING: possible recursive locking detected
  5.15.0-rc6-next-20211022 #10905 Not tainted
  --------------------------------------------
  rtcwake/1326 is trying to acquire lock:
  ffffd4e9192e8130 (cpu_hotplug_lock){++++}-{0:0}, at: 
wake_up_all_idle_cpus+0x24/0x98

  but task is already holding lock:
  ffffd4e9192e8130 (cpu_hotplug_lock){++++}-{0:0}, at: 
suspend_devices_and_enter+0x740/0x9f0

  other info that might help us debug this:
   Possible unsafe locking scenario:

         CPU0
         ----
    lock(cpu_hotplug_lock);
    lock(cpu_hotplug_lock);

   *** DEADLOCK ***

   May be due to missing lock nesting notation

  5 locks held by rtcwake/1326:
   #0: ffff54ad86a78438 (sb_writers#7){.+.+}-{0:0}, at: ksys_write+0x64/0xf0
   #1: ffff54ad84094a88 (&of->mutex){+.+.}-{3:3}, at: 
kernfs_fop_write_iter+0xf4/0x1a8
   #2: ffff54ad83b17a88 (kn->active#43){.+.+}-{0:0}, at: 
kernfs_fop_write_iter+0xfc/0x1a8
   #3: ffffd4e9192efab0 (system_transition_mutex){+.+.}-{3:3}, at: 
pm_suspend+0x214/0x3d0
   #4: ffffd4e9192e8130 (cpu_hotplug_lock){++++}-{0:0}, at: 
suspend_devices_and_enter+0x740/0x9f0

  stack backtrace:
  CPU: 0 PID: 1326 Comm: rtcwake Not tainted 5.15.0-rc6-next-20211022 #10905
  Hardware name: linux,dummy-virt (DT)
  Call trace:
   dump_backtrace+0x0/0x1d0
   show_stack+0x14/0x20
   dump_stack_lvl+0x88/0xb0
   dump_stack+0x14/0x2c
   __lock_acquire+0x171c/0x17b8
   lock_acquire+0x234/0x378
   cpus_read_lock+0x5c/0x150
   wake_up_all_idle_cpus+0x24/0x98
   suspend_devices_and_enter+0x748/0x9f0
   pm_suspend+0x2b0/0x3d0
   state_store+0x84/0x108
   kobj_attr_store+0x14/0x28
   sysfs_kf_write+0x60/0x70
   kernfs_fop_write_iter+0x124/0x1a8
   new_sync_write+0xe8/0x1b0
   vfs_write+0x1d0/0x408
   ksys_write+0x64/0xf0
   __arm64_sys_write+0x14/0x20
   invoke_syscall+0x40/0xf8
   el0_svc_common.constprop.3+0x8c/0x120
   do_el0_svc_compat+0x18/0x48
   el0_svc_compat+0x48/0x100
   el0t_32_sync_handler+0xec/0x140
   el0t_32_sync+0x170/0x174
  OOM killer enabled.
  Restarting tasks ... done.
  PM: suspend exit

--->8---

Let me know if there is anything I can help to debug and fix this issue.

> ---
>   kernel/sched/core.c |   14 +++++---------
>   kernel/smp.c        |    6 +++---
>   2 files changed, 8 insertions(+), 12 deletions(-)
>
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -3691,15 +3691,11 @@ void wake_up_if_idle(int cpu)
>   	if (!is_idle_task(rcu_dereference(rq->curr)))
>   		goto out;
>   
> -	if (set_nr_if_polling(rq->idle)) {
> -		trace_sched_wake_idle_without_ipi(cpu);
> -	} else {
> -		rq_lock_irqsave(rq, &rf);
> -		if (is_idle_task(rq->curr))
> -			smp_send_reschedule(cpu);
> -		/* Else CPU is not idle, do nothing here: */
> -		rq_unlock_irqrestore(rq, &rf);
> -	}
> +	rq_lock_irqsave(rq, &rf);
> +	if (is_idle_task(rq->curr))
> +		resched_curr(rq);
> +	/* Else CPU is not idle, do nothing here: */
> +	rq_unlock_irqrestore(rq, &rf);
>   
>   out:
>   	rcu_read_unlock();
> --- a/kernel/smp.c
> +++ b/kernel/smp.c
> @@ -1170,14 +1170,14 @@ void wake_up_all_idle_cpus(void)
>   {
>   	int cpu;
>   
> -	preempt_disable();
> +	cpus_read_lock();
>   	for_each_online_cpu(cpu) {
> -		if (cpu == smp_processor_id())
> +		if (cpu == raw_smp_processor_id())
>   			continue;
>   
>   		wake_up_if_idle(cpu);
>   	}
> -	preempt_enable();
> +	cpus_read_unlock();
>   }
>   EXPORT_SYMBOL_GPL(wake_up_all_idle_cpus);
>   
>
>
>
Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland