Re: [PATCH] perf, x86: Optimize intel_pmu_pebs_fixup_ip()

From: Peter Zijlstra
Date: Wed Oct 16 2013 - 17:03:35 EST


On Wed, Oct 16, 2013 at 01:52:27PM -0700, Andi Kleen wrote:
> > So avoid having to call copy_from_user_nmi() for every instruction.
> > Since we already limit the max basic block size, we can easily
> > pre-allocate a piece of memory to copy the entire thing into in one
> > go.
>
> It would be better/more generic if you split copy_from_user_nmi() into
> init() copy() end()
>
> (and some state that checks if the underlying page changes)
>
> Then first you don't need the buffer and it could be also
> be applied to the other cases, like the stack unwinding,
> where copying everything is likely too slow.

You'd need to make an iterator interface because of the kmap_atomic crap
needed for 32bit.

But yes, something like that might work, it shouldn't be that hard to
cobble on top of that GUP patch I send out the other day.

The only real nasty part is where an instruction straddles a page
boundary, in that case the iterator stuff fails to be fully transparant
and you need a temp copy of sorts.

Anyway; if you want to have a go at this, feel free.
--
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/