Re: [PATCH] drivers: ipa: Add missing IRQF_ONESHOT

[Alex Elder]

On 2/23/21 3:09 AM, Yang Li wrote:
> fixed the following coccicheck:
> ./drivers/net/ipa/ipa_smp2p.c:186:7-27: ERROR: Threaded IRQ with no
> primary handler requested without IRQF_ONESHOT
>
> Make sure threaded IRQs without a primary handler are always request
> with IRQF_ONESHOT

SMP2P interrupts are handled as nested interrupts.  The hard
handler for a registered SMP2P interrupt is never called, and
is in fact ignored (it should really be NULL when registering).

The "main" SMP2P interrupt handler is a ONESHOT threaded
interrupt handler, and all registered SMP2P interrupts are
handled within (called by) that main handler thread function.

It is not *necessary* to provide the IRQF_ONESHOT flag when
registering an SMP2P interrupt handler.  I don't think it
does real harm to add it, but unless I'm mistaken, the
interrupt handling is actually simpler if ONESHOT is *not*
set in this case.

I could be convinced otherwise, but until I would rather
not accept this patch.

					-Alex

> Reported-by: Abaci Robot <abaci@xxxxxxxxxxxxxxxxx>
> Signed-off-by: Yang Li <yang.lee@xxxxxxxxxxxxxxxxx>
> ---
>   drivers/net/ipa/ipa_smp2p.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/ipa/ipa_smp2p.c b/drivers/net/ipa/ipa_smp2p.c
> index a5f7a79..1149ed8 100644
> --- a/drivers/net/ipa/ipa_smp2p.c
> +++ b/drivers/net/ipa/ipa_smp2p.c
> @@ -183,7 +183,7 @@ static int ipa_smp2p_irq_init(struct ipa_smp2p *smp2p, const char *name,
>   	}
>   	irq = ret;
>   
> -	ret = request_threaded_irq(irq, NULL, handler, 0, name, smp2p);
> +	ret = request_threaded_irq(irq, NULL, handler, IRQF_ONESHOT, name, smp2p);
>   	if (ret) {
>   		dev_err(dev, "error %d requesting \"%s\" IRQ\n", ret, name);
>   		return ret;