Re: [PATCH net-next 1/2] net: add a napi variant for RT-well-behaved drivers

On Sat, 15 May 2021 11:49:02 +0200 Thomas Gleixner wrote:

On Fri, May 14 2021 at 15:24, Jakub Kicinski wrote:

quoted

 /**
- *	napi_schedule_irqoff - schedule NAPI poll
- *	@n: NAPI context
+ * napi_schedule_irq() - schedule NAPI poll from hardware IRQ
+ * @n: NAPI context
  *
  * Variant of napi_schedule(), assuming hard irqs are masked.
+ * Hardware interrupt handler must be marked with IRQF_NO_THREAD
+ * to safely invoke this function on CONFIG_RT=y kernels (unless
+ * it manually masks the interrupts already).
  */
-static inline void napi_schedule_irqoff(struct napi_struct *n)
+static inline void napi_schedule_irq(struct napi_struct *n)
 {
 	if (napi_schedule_prep(n))
-		__napi_schedule_irqoff(n);
+		__napi_schedule_irq(n);  

As this is intended for the trivial

   irqhandler()
        napi_schedule_irq(n);
        return IRQ_HANDLED;

case, wouldn't it make sense to bring napi_schedule_irq() out of line
and have the prep invocation right there?

void napi_schedule_irq(struct napi_struct *n)
{
 	if (napi_schedule_prep(n))
		____napi_schedule(this_cpu_ptr(&softnet_data), n);
}
EXPORT_SYMBOL(napi_schedule_irq);

As that spares a function call and lets the compiler be smarter about
it. I might be missing something though, but at least brain is more
awake now :)

I'm just copying the existing two handlers (reviewer's favorite
response, I know) :)

My guess is .._prep() used to be a static inline itself, I can look 
at modifying all helpers if the assembly really looks better, but based
on Sebastian's email I'm not sure we're gonna go ahead with the new 
helper at all?