perf/x86/intel: Streamline LBR MSR handling in PMI
authorAndi Kleen <ak@linux.intel.com>
Fri, 20 Mar 2015 17:11:23 +0000 (10:11 -0700)
committerIngo Molnar <mingo@kernel.org>
Thu, 2 Apr 2015 15:33:19 +0000 (17:33 +0200)
commit1a78d93750bb5f61abdc59a91fc3bd06a214542a
treee8307794f69c68367d42c7a0cef8c9a1d5b89461
parent15fde1101a1aed11958e0d86bc360f01866a74b1
perf/x86/intel: Streamline LBR MSR handling in PMI

The perf PMI currently does unnecessary MSR accesses when
LBRs are enabled. We use LBR freezing, or when in callstack
mode force the LBRs to only filter on ring 3.

So there is no need to disable the LBRs explicitely in the
PMI handler.

Also we always unnecessarily rewrite LBR_SELECT in the LBR
handler, even though it can never change.

 5)               |  /* write_msr: MSR_LBR_SELECT(1c8), value 0 */
 5)               |  /* read_msr: MSR_IA32_DEBUGCTLMSR(1d9), value 1801 */
 5)               |  /* write_msr: MSR_IA32_DEBUGCTLMSR(1d9), value 1801 */
 5)               |  /* write_msr: MSR_CORE_PERF_GLOBAL_CTRL(38f), value 70000000f */
 5)               |  /* write_msr: MSR_CORE_PERF_GLOBAL_CTRL(38f), value 0 */
 5)               |  /* write_msr: MSR_LBR_SELECT(1c8), value 0 */
 5)               |  /* read_msr: MSR_IA32_DEBUGCTLMSR(1d9), value 1801 */
 5)               |  /* write_msr: MSR_IA32_DEBUGCTLMSR(1d9), value 1801 */

This patch:

  - Avoids disabling already frozen LBRs unnecessarily in the PMI
  - Avoids changing LBR_SELECT in the PMI

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/1426871484-21285-1-git-send-email-andi@firstfloor.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/kernel/cpu/perf_event.h
arch/x86/kernel/cpu/perf_event_intel.c
arch/x86/kernel/cpu/perf_event_intel_lbr.c