linux-kernel - [RFC 0/6] perf, tool: Allow to use hw events in PMU syntax

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-next>] [day] [month] [year] [list]

Message-Id: <1339706321-8802-1-git-send-email-jolsa@redhat.com>
Date:	Thu, 14 Jun 2012 22:38:35 +0200
From:	Jiri Olsa <jolsa@...hat.com>
To:	acme@...hat.com, a.p.zijlstra@...llo.nl, mingo@...e.hu,
	paulus@...ba.org, cjashfor@...ux.vnet.ibm.com, fweisbec@...il.com,
	eranian@...gle.com
Cc:	linux-kernel@...r.kernel.org
Subject: [RFC 0/6] perf, tool: Allow to use hw events in PMU syntax

hi,
here's the change to make following syntax available:
  perf stat -e cpu/event=instructions/u ls

this is identical to:
  perf stat -e instructions:u ls

The patch 1 makes PERF_TYPE_HARDWARE events translations available
by the sysfs 'events' group attribute under the PMU record.

Once we hit the PERF_TYPE_HARDWARE event mnemonic within the
PMU event term parsing like:
   'event=cycles'
we translate the 'cycles' string into specific PERF_TYPE_HARDWARE
translation and assign this value to the 'event' term value.

So given that:
  $ cat /sys/devices/cpu/events/instructions
  0xc0

following examples are identical:
  perf stat -e cpu/event=0xc0/u ls
  perf stat -e cpu/event=instructions/u ls

It feels like a hack since it relies on:
  - kernel does no more than translating the 'cycles' value and putting it
    to the 'attr.config' - I'm not sure this is allways true.
  - user uses this translation with terms that fill the attr.config same way
    as kernel does, like for 'event':
      $ cat /sys/devices/cpu/format/event
      config:0-7

    This could be ensured by perf tool.. but it smells like CPU specific again.

note, it's not tested much.. ;)

attached patches:
  1/6 perf, x86: Making hardware events tranlations sysfs available
  2/6 perf tools: Fix generation of pmu list
  3/6 perf, tool: Properly free format data
  4/6 perf, tool: Add events support for pmu
  5/6 perf, tool: event parsing - split PE_VALUE_SYM to SW and HW tokens
  6/6 perf, tool: Support translate terms for hw events

thoughts? ;)
jirka
---
 arch/x86/kernel/cpu/perf_event.c |   44 +++++++++
 tools/perf/util/parse-events.c   |    8 ++-
 tools/perf/util/parse-events.h   |    3 +
 tools/perf/util/parse-events.l   |    2 +-
 tools/perf/util/parse-events.y   |   26 +++++-
 tools/perf/util/pmu.c            |  179 ++++++++++++++++++++++++++++++++-----
 tools/perf/util/pmu.h            |    2 +
 7 files changed, 233 insertions(+), 31 deletions(-)
--
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/