[PATCH V7 2/5] perf/x86/kvm: Avoid unnecessary work in guest filtering

[kan . liang]

From: Andi Kleen <ak@xxxxxxxxxxxxxxx>

KVM added a workaround for PEBS events leaking into guests with
commit 26a4f3c08de4 ("perf/x86: disable PEBS on a guest entry.")
This uses the VT entry/exit list to add an extra disable of the
PEBS_ENABLE MSR.

Intel also added a fix for this issue to microcode updates on
Haswell/Broadwell/Skylake.

It turns out using the MSR entry/exit list makes VM exits
significantly slower. The list is only needed for disabling
PEBS, because the GLOBAL_CTRL change gets optimized by
KVM into changing the VMCS.

Check for the microcode updates that have the microcode
fix for leaking PEBS, and disable the extra entry/exit list
entry for PEBS_ENABLE. In addition we always clear the
GLOBAL_CTRL for the PEBS counter while running in the guest,
which is enough to make them never fire at the wrong
side of the host/guest transition.

The overhead for VM exits with the filtering active with the patch is
reduced from 8% to 4%.

The microcode patch has already been merged into future platforms.
This patch is one-off thing. The quirks is used here.

For other old platforms which doesn't have microcode patch and quirks,
extra disable of the PEBS_ENABLE MSR is still required.

Signed-off-by: Andi Kleen <ak@xxxxxxxxxxxxxxx>
Signed-off-by: Kan Liang <kan.liang@xxxxxxxxxxxxxxx>
---

Changes since V6:
- Remove check for CANNONLAKE
- Move pebs_no_isolation to intel_ds_init() to reduce the noisy
- Rename intel_check_isolation() to intel_check_pebs_isolation()
  and intel_isolation_quirk() to intel_pebs_isolation_quirk()

Keep the KABYLAKE check as is. Because it is one off.

 arch/x86/events/intel/core.c | 74 ++++++++++++++++++++++++++++++++++++++------
 arch/x86/events/intel/ds.c   |  2 ++
 arch/x86/events/perf_event.h | 15 ++++-----
 3 files changed, 75 insertions(+), 16 deletions(-)

diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c
index 40e12cf..8d94f40 100644
--- a/arch/x86/events/intel/core.c
+++ b/arch/x86/events/intel/core.c
@@ -18,6 +18,7 @@
 #include <asm/hardirq.h>
 #include <asm/intel-family.h>
 #include <asm/apic.h>
+#include <asm/cpu_device_id.h>
 
 #include "../perf_event.h"
 
@@ -3206,16 +3207,27 @@ static struct perf_guest_switch_msr *intel_guest_get_msrs(int *nr)
 	arr[0].msr = MSR_CORE_PERF_GLOBAL_CTRL;
 	arr[0].host = x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_guest_mask;
 	arr[0].guest = x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_host_mask;
-	/*
-	 * If PMU counter has PEBS enabled it is not enough to disable counter
-	 * on a guest entry since PEBS memory write can overshoot guest entry
-	 * and corrupt guest memory. Disabling PEBS solves the problem.
-	 */
-	arr[1].msr = MSR_IA32_PEBS_ENABLE;
-	arr[1].host = cpuc->pebs_enabled;
-	arr[1].guest = 0;
+	if (x86_pmu.flags & PMU_FL_PEBS_ALL)
+		arr[0].guest &= ~cpuc->pebs_enabled;
+	else
+		arr[0].guest &= ~(cpuc->pebs_enabled & PEBS_COUNTER_MASK);
+	*nr = 1;
+
+	if (x86_pmu.pebs && x86_pmu.pebs_no_isolation) {
+		/*
+		 * If PMU counter has PEBS enabled it is not enough to
+		 * disable counter on a guest entry since PEBS memory
+		 * write can overshoot guest entry and corrupt guest
+		 * memory. Disabling PEBS solves the problem.
+		 *
+		 * Don't do this if the CPU already enforces it.
+		 */
+		arr[1].msr = MSR_IA32_PEBS_ENABLE;
+		arr[1].host = cpuc->pebs_enabled;
+		arr[1].guest = 0;
+		*nr = 2;
+	}
 
-	*nr = 2;
 	return arr;
 }
 
@@ -3733,6 +3745,47 @@ static __init void intel_clovertown_quirk(void)
 	x86_pmu.pebs_constraints = NULL;
 }
 
+static const struct x86_cpu_desc isolation_ucodes[] = {
+	INTEL_CPU_DESC(INTEL_FAM6_HASWELL_CORE,		 3, 0x0000001f),
+	INTEL_CPU_DESC(INTEL_FAM6_HASWELL_ULT,		 1, 0x0000001e),
+	INTEL_CPU_DESC(INTEL_FAM6_HASWELL_GT3E,		 1, 0x00000015),
+	INTEL_CPU_DESC(INTEL_FAM6_HASWELL_X,		 2, 0x00000037),
+	INTEL_CPU_DESC(INTEL_FAM6_HASWELL_X,		 4, 0x0000000a),
+	INTEL_CPU_DESC(INTEL_FAM6_BROADWELL_CORE,	 4, 0x00000023),
+	INTEL_CPU_DESC(INTEL_FAM6_BROADWELL_GT3E,	 1, 0x00000014),
+	INTEL_CPU_DESC(INTEL_FAM6_BROADWELL_XEON_D,	 2, 0x00000010),
+	INTEL_CPU_DESC(INTEL_FAM6_BROADWELL_XEON_D,	 3, 0x07000009),
+	INTEL_CPU_DESC(INTEL_FAM6_BROADWELL_XEON_D,	 4, 0x0f000009),
+	INTEL_CPU_DESC(INTEL_FAM6_BROADWELL_XEON_D,	 5, 0x0e000002),
+	INTEL_CPU_DESC(INTEL_FAM6_BROADWELL_X,		 2, 0x0b000014),
+	INTEL_CPU_DESC(INTEL_FAM6_SKYLAKE_X,		 3, 0x00000021),
+	INTEL_CPU_DESC(INTEL_FAM6_SKYLAKE_X,		 4, 0x00000000),
+	INTEL_CPU_DESC(INTEL_FAM6_SKYLAKE_MOBILE,	 3, 0x0000007c),
+	INTEL_CPU_DESC(INTEL_FAM6_SKYLAKE_DESKTOP,	 3, 0x0000007c),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_DESKTOP,	 9, 0x0000004e),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_MOBILE,	 9, 0x0000004e),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_MOBILE,	10, 0x0000004e),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_MOBILE,	11, 0x0000004e),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_MOBILE,	12, 0x0000004e),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_DESKTOP,	10, 0x0000004e),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_DESKTOP,	11, 0x0000004e),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_DESKTOP,	12, 0x0000004e),
+	INTEL_CPU_DESC(INTEL_FAM6_KABYLAKE_DESKTOP,	13, 0x0000004e),
+	{}
+};
+
+static void intel_check_pebs_isolation(void)
+{
+	x86_pmu.pebs_no_isolation = !x86_cpu_has_min_microcode_rev(isolation_ucodes);
+}
+
+static __init void intel_pebs_isolation_quirk(void)
+{
+	WARN_ON_ONCE(x86_pmu.check_microcode);
+	x86_pmu.check_microcode = intel_check_pebs_isolation;
+	intel_check_pebs_isolation();
+}
+
 static int intel_snb_pebs_broken(int cpu)
 {
 	u32 rev = UINT_MAX; /* default to broken for unknown models */
@@ -4425,6 +4478,7 @@ __init int intel_pmu_init(void)
 	case INTEL_FAM6_HASWELL_ULT:
 	case INTEL_FAM6_HASWELL_GT3E:
 		x86_add_quirk(intel_ht_bug);
+		x86_add_quirk(intel_pebs_isolation_quirk);
 		x86_pmu.late_ack = true;
 		memcpy(hw_cache_event_ids, hsw_hw_cache_event_ids, sizeof(hw_cache_event_ids));
 		memcpy(hw_cache_extra_regs, hsw_hw_cache_extra_regs, sizeof(hw_cache_extra_regs));
@@ -4456,6 +4510,7 @@ __init int intel_pmu_init(void)
 	case INTEL_FAM6_BROADWELL_XEON_D:
 	case INTEL_FAM6_BROADWELL_GT3E:
 	case INTEL_FAM6_BROADWELL_X:
+		x86_add_quirk(intel_pebs_isolation_quirk);
 		x86_pmu.late_ack = true;
 		memcpy(hw_cache_event_ids, hsw_hw_cache_event_ids, sizeof(hw_cache_event_ids));
 		memcpy(hw_cache_extra_regs, hsw_hw_cache_extra_regs, sizeof(hw_cache_extra_regs));
@@ -4518,6 +4573,7 @@ __init int intel_pmu_init(void)
 	case INTEL_FAM6_SKYLAKE_X:
 	case INTEL_FAM6_KABYLAKE_MOBILE:
 	case INTEL_FAM6_KABYLAKE_DESKTOP:
+		x86_add_quirk(intel_pebs_isolation_quirk);
 		x86_pmu.late_ack = true;
 		memcpy(hw_cache_event_ids, skl_hw_cache_event_ids, sizeof(hw_cache_event_ids));
 		memcpy(hw_cache_extra_regs, skl_hw_cache_extra_regs, sizeof(hw_cache_extra_regs));
diff --git a/arch/x86/events/intel/ds.c b/arch/x86/events/intel/ds.c
index e9acf1d..450ea6b 100644
--- a/arch/x86/events/intel/ds.c
+++ b/arch/x86/events/intel/ds.c
@@ -1628,6 +1628,8 @@ void __init intel_ds_init(void)
 	x86_pmu.bts  = boot_cpu_has(X86_FEATURE_BTS);
 	x86_pmu.pebs = boot_cpu_has(X86_FEATURE_PEBS);
 	x86_pmu.pebs_buffer_size = PEBS_BUFFER_SIZE;
+	if (x86_pmu.version < 5)
+		x86_pmu.pebs_no_isolation = 1;
 	if (x86_pmu.pebs) {
 		char pebs_type = x86_pmu.intel_cap.pebs_trap ?  '+' : '-';
 		int format = x86_pmu.intel_cap.pebs_format;
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index 78d7b70..dea716e 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -601,13 +601,14 @@ struct x86_pmu {
 	/*
 	 * Intel DebugStore bits
 	 */
-	unsigned int	bts		:1,
-			bts_active	:1,
-			pebs		:1,
-			pebs_active	:1,
-			pebs_broken	:1,
-			pebs_prec_dist	:1,
-			pebs_no_tlb	:1;
+	unsigned int	bts			:1,
+			bts_active		:1,
+			pebs			:1,
+			pebs_active		:1,
+			pebs_broken		:1,
+			pebs_prec_dist		:1,
+			pebs_no_tlb		:1,
+			pebs_no_isolation	:1;
 	int		pebs_record_size;
 	int		pebs_buffer_size;
 	void		(*drain_pebs)(struct pt_regs *regs);
-- 
2.7.4