linux-kernel - Re: [PATCH] POWER: perf_event: Skip updating kernel counters if register value shrinks

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] [day] [month] [year] [list]

Message-ID: <20110407161611.GC3119@mgebm.net>
Date:	Thu, 7 Apr 2011 12:16:11 -0400
From:	Eric B Munson <emunson@...bm.net>
To:	Benjamin Herrenschmidt <benh@...nel.crashing.org>
Cc:	a.p.zijlstra@...llo.nl, paulus@...ba.org, mingo@...e.hu,
	acme@...stprotocols.net, linuxppc-dev@...ts.ozlabs.org,
	linux-kernel@...r.kernel.org, anton@...ba.org
Subject: Re: [PATCH] POWER: perf_event: Skip updating kernel counters if
 register value shrinks

On Thu, 07 Apr 2011, Benjamin Herrenschmidt wrote:

> 
> > > Doesn't that mean that power_pmu_read() can only ever increase the value of
> > > the perf_event and so will essentially -stop- once the counter rolls over ?
> > > 
> > > Similar comments every where you do this type of comparison.
> > > 
> > > Cheers,
> > > Ben.
> > 
> > Sorry for the nag, but am I missing something about the way the register and
> > the previous values are reset in the overflow interrupt handler?
> 
> Well, not all counters get interrupts right ? Some counters are just
> free running... I'm not sure when that power_pmu_read() function is
> actually used by the core, I'm not that familiar with perf, but I'd say
> better safe than sorry. When comparing counter values, doing in a way
> that is generally safe vs. wraparounds. Eventually do a helper for that.
> 
> Cheers,
> Ben.

I am honestly not sure, I was under the assumption that all counters would
generate an interrupt if they overflowed.  I do not have the hardware docs to
prove this, so I will have a V3 that (I think/hope) addresses your concerns out
momentarily.

Eric

Download attachment "signature.asc" of type "application/pgp-signature" (491 bytes)