RE: [2.6.25-rc5-mm1][regression] ia64: hackbench doesn't finish>12hour

[Peter Zijlstra]

On Thu, 2008-03-20 at 09:04 -0700, Luck, Tony wrote:
> > Which makes me wonder, why did you ever use a semaphore here? Looking at
> > the code its a straight forward mutex. And when you would have used a
> > mutex lockdep would have warned about this.
> 
> The functionality that we are trying to add is to allow up to N
> simultaneous processors to execute the critical region.  On current
> processors/platforms N=1 so a spinlock or mutex would be fine, but
> there will be platforms for which N is a small integer greater than
> one.  Semaphore initialized to N looked to be the ideal primitive
> for this (until Motohiro-san ran the test case that showed the path
> where we call this code with a spinlock held).

Right, no alternative there.

> Next question is whether it is reasonable to get to this code
> while holding a spinlock.  Isn't this a problem for architectures
> that need to use cross-processor interrupts to do a global TLB
> shootdown?

Yeah, semaphores can't be used from hardirq contexts for much the same
reasons. But its all ia64 code, right? So I'm not directly seeing how
other archs are affected here.

--
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/