[PATCH 1/2] perf, x86: Streamline LBR MSR handling in PMI

[Andi Kleen <andi@firstfloor.org>]

From: Andi Kleen <ak@linux.intel.com>

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>
---
 arch/x86/kernel/cpu/perf_event.h           |  2 +-
 arch/x86/kernel/cpu/perf_event_intel.c     | 23 ++++++++++++++++++-----
 arch/x86/kernel/cpu/perf_event_intel_lbr.c | 12 ++++++++----
 3 files changed, 27 insertions(+), 10 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index aa316c0..2d65de4 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -739,7 +739,7 @@ void intel_pmu_lbr_enable(struct perf_event *event);
 
 void intel_pmu_lbr_disable(struct perf_event *event);
 
-void intel_pmu_lbr_enable_all(void);
+void intel_pmu_lbr_enable_all(bool pmi);
 
 void intel_pmu_lbr_disable_all(void);
 
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 9f1dd18..ea81a21 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -1029,7 +1029,10 @@ static __initconst const u64 slm_hw_cache_event_ids
  },
 };
 
-static void intel_pmu_disable_all(void)
+/*
+ * Use from PMIs where the LBRs are already disabled.
+ */
+static void __intel_pmu_disable_all(void)
 {
 	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
 
@@ -1039,15 +1042,20 @@ static void intel_pmu_disable_all(void)
 		intel_pmu_disable_bts();
 
 	intel_pmu_pebs_disable_all();
+}
+
+static void intel_pmu_disable_all(void)
+{
+	__intel_pmu_disable_all();
 	intel_pmu_lbr_disable_all();
 }
 
-static void intel_pmu_enable_all(int added)
+static void __intel_pmu_enable_all(int added, bool pmi)
 {
 	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
 
 	intel_pmu_pebs_enable_all();
-	intel_pmu_lbr_enable_all();
+	intel_pmu_lbr_enable_all(pmi);
 	wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL,
 			x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_guest_mask);
 
@@ -1062,6 +1070,11 @@ static void intel_pmu_enable_all(int added)
 	}
 }
 
+static void intel_pmu_enable_all(int added)
+{
+	__intel_pmu_enable_all(added, false);
+}
+
 /*
  * Workaround for:
  *   Intel Errata AAK100 (model 26)
@@ -1343,7 +1356,7 @@ static int intel_pmu_handle_irq(struct pt_regs *regs)
 	 */
 	if (!x86_pmu.late_ack)
 		apic_write(APIC_LVTPC, APIC_DM_NMI);
-	intel_pmu_disable_all();
+	__intel_pmu_disable_all();
 	handled = intel_pmu_drain_bts_buffer();
 	status = intel_pmu_get_status();
 	if (!status)
@@ -1419,7 +1432,7 @@ again:
 		goto again;
 
 done:
-	intel_pmu_enable_all(0);
+	__intel_pmu_enable_all(0, true);
 	/*
 	 * Only unmask the NMI after the overflow counters
 	 * have been reset. This avoids spurious NMIs on
diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index 0473874..3d53725 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -132,12 +132,16 @@ static void intel_pmu_lbr_filter(struct cpu_hw_events *cpuc);
  * otherwise it becomes near impossible to get a reliable stack.
  */
 
-static void __intel_pmu_lbr_enable(void)
+static void __intel_pmu_lbr_enable(bool pmi)
 {
 	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
 	u64 debugctl, lbr_select = 0;
 
-	if (cpuc->lbr_sel) {
+	/*
+	 * No need to reprogram LBR_SELECT in a PMI, as it
+	 * did not change.
+	 */
+	if (cpuc->lbr_sel && !pmi) {
 		lbr_select = cpuc->lbr_sel->config;
 		wrmsrl(MSR_LBR_SELECT, lbr_select);
 	}
@@ -351,12 +355,12 @@ void intel_pmu_lbr_disable(struct perf_event *event)
 	}
 }
 
-void intel_pmu_lbr_enable_all(void)
+void intel_pmu_lbr_enable_all(bool pmi)
 {
 	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
 
 	if (cpuc->lbr_users)
-		__intel_pmu_lbr_enable();
+		__intel_pmu_lbr_enable(pmi);
 }
 
 void intel_pmu_lbr_disable_all(void)
-- 
1.9.3