Re: [PATCH 2/4, v2] x86: enlightenment for ticket spin locks - Xen implementation
From: Jan Beulich
Date: Wed Jun 30 2010 - 05:03:57 EST
>>> On 30.06.10 at 10:56, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> On Wed, 2010-06-30 at 09:52 +0100, Jan Beulich wrote:
>> > why is that a rwlock?, those things are useless.
>> Because potentially each CPU's lock gets acquired for reading during
>> unlock, while only the locking CPU's one needs to be acquired for
>> writing during lock.
> Can you say: scalability nightmare? but then its Xen code so who cares..
Yes, this is a scalability problem. And yes, I'm open to suggestions.
But no, I didn't have any better idea myself. And for the time being
I don't foresee this to be a problem, as VMs tend to have much fewer
CPUs than physical machines. And I think a performance win of 25%
justifies a sub-optimal implementation (with the perspective of people
smarter than me replacing it with a better one over time).
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/