linux-kernel - RE: [PATCH V5 4/8] perf/x86/intel/uncore: add new data structures for free running counters

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]

Date:   Tue, 23 Jan 2018 22:00:58 +0000
From:   "Liang, Kan" <kan.liang@...el.com>
To:     Peter Zijlstra <peterz@...radead.org>,
        Andi Kleen <ak@...ux.intel.com>
CC:     Stephane Eranian <eranian@...gle.com>,
        "tglx@...utronix.de" <tglx@...utronix.de>,
        "mingo@...hat.com" <mingo@...hat.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "acme@...nel.org" <acme@...nel.org>
Subject: RE: [PATCH V5 4/8] perf/x86/intel/uncore: add new data structures
 for free running counters

> On Fri, Jan 19, 2018 at 12:24:17PM -0800, Andi Kleen wrote:
> > > Oh, think a bit more.
> > > I think we cannot do the same thing as we did for CPU PMU's fixed
> counters.
> > >
> > > The counters here are free running counters. They cannot be start/stop.
> >
> > Yes free running counter have completely different semantics. They
> > need a separate event code.
> 
> The only thing that matters is if they count the same thing or not.
>

Hi Peter,

There is NO event available on the GPs, that is exactly the same as
the free-running counters.

For example, the BW free-running counters count the requests associated
with writes and completions.
The most similar events on the GPs are DATA_REQ_{OF,BY}_CPU.* events.
Except that some of their sub-events count requests which not completions.
There are also other minor differences.
So we don't have alternative events for the free-running counters.
I think we have to use 0xff.

For details, please refer to the uncore PMU guide. 
https://www.intel.com/content/www/us/en/processors/xeon/scalable/xeon-scalable-uncore-performance-monitoring-manual.html

Thanks,
Kan

> The not start/stop thing is not important. See arch/x86/events/msr.c on how
> to deal with that. The short story is that you simply ignore stop and update
> the prev_count on start. Then any next update will increment with the
> correct delta.
> 
> (if the counter is short you also need to run a timer to deal with wraps).