linux-kernel - Re: [PATCH] perf, x86: Add workaround for MEM_*

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20130501090718.GA28253@dyad.programming.kicks-ass.net>
Date:	Wed, 1 May 2013 11:07:19 +0200
From:	Peter Zijlstra <peterz@...radead.org>
To:	Andi Kleen <andi@...stfloor.org>
Cc:	mingo@...e.hu, eranian@...gle.com, linux-kernel@...r.kernel.org,
	Andi Kleen <ak@...ux.intel.com>
Subject: Re: [PATCH] perf, x86: Add workaround for MEM_*_RETIRED errata BV98

On Mon, Apr 29, 2013 at 11:45:08AM -0700, Andi Kleen wrote:
> From: Andi Kleen <ak@...ux.intel.com>
> 
> On IvyBridge MEM_*_RETIRED can leak through to the same counter
> on the other thread. Add a dummy extra_reg to handle this case.
> The extra reg is allocated and makes sure nothing different
> runs on the same counter in the other thread.
> 
> This is normally only visible when multiplexing.
> 
> This patch doesn't 100% plug the hole, as the event may still get lost
> when the other CPU thread does not have an enabled counter
> (e.g. with per thread counting). However it fixes it well enough
> for the common global profiling case (e.g. perf top / perf stat -a)
> In theory it would be possible to handle the other cases
> too, but it would need far more changes.

So you're saying that if two SMT siblings count the same MEM_*_RETIRED event
(on the same counter?) events can get accounted to the wrong sibling?

And when the other sibling doesn't have (the same counter?) enabled we can
loose events?

This begs the question what happens when the sibling does have the (same?)
counter enabled but counting an all together different event; do we then still
'loose' events from the one sibling and add then to the other counter?


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