Re: [patch 13/26] Xen-paravirt_ops: Consistently wrap paravirt opscallsites to make them patchable

From: Zachary Amsden
Date: Tue Mar 20 2007 - 18:09:51 EST


Jeremy Fitzhardinge wrote:

Yeah, disable interrupts, and set a flag that the fake "sti" can test, and just return without doing anything.

(You may or may not also need to do extra work to Ack the hardware interrupt etc, which may be irq-controller specific. Once the CPU has accepted the interrupt, you may not be able to just leave it dangling)

So it would be something like:

pda.intr_mask = 1; /* disable interrupts */
...
pda.intr_mask = 0; /* enable interrupts */
if (xchg(&pda.intr_pending, 0)) /* check pending */

Well, can't do xchg, since it implies #LOCK, and you'll lose more than you gain on the processors where it matters. Cmpxchg is fine, but processor dependent.

Or, just make the interrupt handlers use software resend for IRQs when pda.intr_mask is set to zero. Now, local_irq_save / restore are very pretty:

int local_irq_save(void)
{
int tmp = pda.intr_mask;
pda.intr_mask = 0;
/*
* note there is a window here where local IRQs notice intr_mask == 0
* in that case, they will attempt to resend the IRQ via a tasklet,
* and will succeed, albeit through a slightly longer path
*/
local_bh_disable();
return tmp;
}

void local_irq_restore(int enabled)
{
pda.intr_mask = enabled;
/*
* note there is a window here where softirqs are not processed by
* the interrupt handler, but that is not a problem, since it will
* get done here in the outer enable of any nested pair.
*/
if (enabled)
local_bh_enable();
}

I think Ingo's suggestion of using the hardirq tracing is another way that could work, but it seems to be too heavyweight and tied too much to the lockdep verification code - plus it inserts additional raw_irq_disable in places that seem counter to the goal of getting rid of them in the first place. Perhaps I'm misunderstanding the code, though.

Zach
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/