Re: [PATCH] softirq: Remove raise_softirq from tasklet_action_common()

[Frederic Weisbecker]

On Sat, Feb 05, 2022 at 06:43:25PM +0530, Mukesh Ojha wrote:
> Think about a scenario when all other cores are in suspend
> and one core is only running ksoftirqd and it is because
> some client has invoked tasklet_hi_schedule() only once
> during that phase.
> 
> tasklet_action_common() handles that softirq and marks the
> same softirq as pending again. And due to that core keeps
> running the softirq handler [1] forever and it is not able to
> go to suspend.
> 
> We can get rid of raising softirq from tasklet handler.
> 
> [1]
> <idle>-0    [003]   13058.769081:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13058.769085: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13058.769087:  softirq_exit   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13058.769091:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13058.769094: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13058.769097:  softirq_exit   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13058.769100:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13058.769103: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13058.769106:  softirq_exit   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13058.769109:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13059.058923:  softirq_entry   vec=0  action=HI_SOFTIRQ
> ...
> ..
> ..
> ..
> 
> <idle>-0    [003]   13059.058951:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13059.058954: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13059.058957:  softirq_exit   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13059.058960:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13059.058963: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13059.058966:  softirq_exit   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13059.058969:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13059.058972: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13059.058975:  softirq_exit   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13059.058978:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13059.058981: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13059.058984:  softirq_exit   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13059.058987:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13059.058990: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13059.058993:  softirq_exit   vec=0  action=HI_SOFTIRQ
> <idle>-0    [003]   13059.058996:  softirq_entry   vec=0  action=HI_SOFTIRQ
> <idle>-0     [003]  13059.059000: softirq_raise:        vec=0 [action=HI_SOFTIRQ]
> <idle>-0    [003]   13059.059002:  softirq_exit   vec=0  action=HI_SOFTIRQ
> 
> Signed-off-by: Mukesh Ojha <quic_mojha@xxxxxxxxxxx>
> ---
>  kernel/softirq.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/kernel/softirq.c b/kernel/softirq.c
> index 41f4709..d3e6fb9 100644
> --- a/kernel/softirq.c
> +++ b/kernel/softirq.c
> @@ -795,7 +795,6 @@ static void tasklet_action_common(struct softirq_action *a,
>  		t->next = NULL;
>  		*tl_head->tail = t;
>  		tl_head->tail = &t->next;
> -		__raise_softirq_irqoff(softirq_nr);
>  		local_irq_enable();

That requeue happens when the tasklet is already executing on some other CPU
or when it has been disabled through tasklet_disable().

So you can't just remove that line or you'll break everything.

It would be nice to identify which tasklet keeps being requeued. Is it because
something called tasklet_disable() to it and never called back tasklet_enable() ?

Thanks.