[PATCH v4 5/6] xen/PMU: Intercept PMU-related MSR and APIC accesses

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

Provide interfaces for recognizing accesses to PMU-related MSRs and LVTPC APIC
and process these accesses in Xen PMU code.

(The interrupt handler performs XENPMU_flush right away in the beginning since
no PMU emulation is available. It will be added with a later patch).

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: David Vrabel <david.vrabel@citrix.com>
---
 arch/x86/xen/enlighten.c       | 26 ++++++++----
 arch/x86/xen/pmu.c             | 95 +++++++++++++++++++++++++++++++++++++++++-
 arch/x86/xen/pmu.h             |  4 ++
 include/xen/interface/xenpmu.h |  2 +
 4 files changed, 118 insertions(+), 9 deletions(-)

diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 57764ce..9ab5f76 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -82,6 +82,7 @@
 #include "mmu.h"
 #include "smp.h"
 #include "multicalls.h"
+#include "pmu.h"
 
 EXPORT_SYMBOL_GPL(hypercall_page);
 
@@ -962,15 +963,12 @@ static u32 xen_apic_read(u32 reg)
 	return op.u.pcpu_info.apic_id << 24;
 }
 
-unsigned long long xen_read_pmc(int counter)
-{
-	return 0;
-}
-
 static void xen_apic_write(u32 reg, u32 val)
 {
-	if (reg == APIC_LVTPC)
+	if (reg == APIC_LVTPC) {
+		(void)pmu_apic_update(reg);
 		return;
+	}
 
 	/* Warn to see if there's any stray references */
 	WARN_ON(1);
@@ -1076,6 +1074,17 @@ static inline void xen_write_cr8(unsigned long val)
 	BUG_ON(val);
 }
 #endif
+
+static u64 xen_read_msr_safe(unsigned int msr, int *err)
+{
+	u64 val;
+
+	if (pmu_msr_read(msr, &val, err))
+		return val;
+
+	return native_read_msr_safe(msr, err);
+}
+
 static int xen_write_msr_safe(unsigned int msr, unsigned low, unsigned high)
 {
 	int ret;
@@ -1116,7 +1125,8 @@ static int xen_write_msr_safe(unsigned int msr, unsigned low, unsigned high)
 		break;
 
 	default:
-		ret = native_write_msr_safe(msr, low, high);
+		if (!pmu_msr_write(msr, low, high, &ret))
+			ret = native_write_msr_safe(msr, low, high);
 	}
 
 	return ret;
@@ -1252,7 +1262,7 @@ static const struct pv_cpu_ops xen_cpu_ops __initconst = {
 
 	.wbinvd = native_wbinvd,
 
-	.read_msr = native_read_msr_safe,
+	.read_msr = xen_read_msr_safe,
 	.write_msr = xen_write_msr_safe,
 
 	.read_tsc = native_read_tsc,
diff --git a/arch/x86/xen/pmu.c b/arch/x86/xen/pmu.c
index f893631..959172e 100644
--- a/arch/x86/xen/pmu.c
+++ b/arch/x86/xen/pmu.c
@@ -51,6 +51,8 @@ static __read_mostly int amd_num_counters;
 /* Alias registers (0x4c1) for full-width writes to PMCs */
 #define MSR_PMC_ALIAS_MASK          (~(MSR_IA32_PERFCTR0 ^ MSR_IA32_PMC0))
 
+#define INTEL_PMC_TYPE_SHIFT        30
+
 static __read_mostly int intel_num_arch_counters, intel_num_fixed_counters;
 
 
@@ -167,6 +169,91 @@ static int is_intel_pmu_msr(u32 msr_index, int *type, int *index)
 	}
 }
 
+bool pmu_msr_read(unsigned int msr, uint64_t *val, int *err)
+{
+
+	if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
+		if (is_amd_pmu_msr(msr)) {
+			*val = native_read_msr_safe(msr, err);
+			return true;
+		}
+	} else {
+		int type, index;
+
+		if (is_intel_pmu_msr(msr, &type, &index)) {
+			*val = native_read_msr_safe(msr, err);
+			return true;
+		}
+	}
+
+	return false;
+}
+
+bool pmu_msr_write(unsigned int msr, uint32_t low, uint32_t high, int *err)
+{
+	if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
+		if (is_amd_pmu_msr(msr)) {
+			*err = native_write_msr_safe(msr, low, high);
+			return true;
+		}
+	} else {
+		int type, index;
+
+		if (is_intel_pmu_msr(msr, &type, &index)) {
+			*err = native_write_msr_safe(msr, low, high);
+			return true;
+		}
+	}
+
+	return false;
+}
+
+unsigned long long xen_amd_read_pmc(int counter)
+{
+	uint32_t msr;
+	int err;
+
+	msr = amd_counters_base + (counter * amd_msr_step);
+	return native_read_msr_safe(msr, &err);
+}
+
+unsigned long long xen_intel_read_pmc(int counter)
+{
+	int err;
+	uint32_t msr;
+
+	if (counter & (1<<INTEL_PMC_TYPE_SHIFT))
+		msr = MSR_CORE_PERF_FIXED_CTR0 + (counter & 0xffff);
+	else
+		msr = MSR_IA32_PERFCTR0 + counter;
+
+	return native_read_msr_safe(msr, &err);
+}
+
+unsigned long long xen_read_pmc(int counter)
+{
+	if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
+		return xen_amd_read_pmc(counter);
+	else
+		return xen_intel_read_pmc(counter);
+}
+
+int pmu_apic_update(uint32_t val)
+{
+	int ret;
+	struct xen_pmu_data *xenpmu_data = get_xenpmu_data();
+
+	if (!xenpmu_data) {
+		WARN_ONCE(1, "%s: pmudata not initialized\n", __func__);
+		return -EINVAL;
+	}
+
+	xenpmu_data->pmu.lapic_lvtpc = val;
+	ret = HYPERVISOR_xenpmu_op(XENPMU_lvtpc_set, NULL);
+
+	return ret;
+}
+
 /* perf callbacks*/
 int xen_is_in_guest(void)
 {
@@ -223,7 +310,7 @@ static void xen_convert_regs(const struct cpu_user_regs *xen_regs,
 
 irqreturn_t xen_pmu_irq_handler(int irq, void *dev_id)
 {
-	int ret = IRQ_NONE;
+	int err, ret = IRQ_NONE;
 	struct pt_regs regs;
 	const struct xen_pmu_data *xenpmu_data = get_xenpmu_data();
 
@@ -232,6 +319,12 @@ irqreturn_t xen_pmu_irq_handler(int irq, void *dev_id)
 		return ret;
 	}
 
+	err = HYPERVISOR_xenpmu_op(XENPMU_flush, NULL);
+	if (err) {
+		WARN_ONCE(1, "%s failed hypercall, err: %d\n", __func__, err);
+		return ret;
+	}
+
 	xen_convert_regs(&xenpmu_data->pmu.regs, &regs);
 	if (x86_pmu.handle_irq(&regs))
 		ret = IRQ_HANDLED;
diff --git a/arch/x86/xen/pmu.h b/arch/x86/xen/pmu.h
index d52e8db..30bfbcf 100644
--- a/arch/x86/xen/pmu.h
+++ b/arch/x86/xen/pmu.h
@@ -7,5 +7,9 @@ irqreturn_t xen_pmu_irq_handler(int irq, void *dev_id);
 int xen_pmu_init(int cpu);
 void xen_pmu_finish(int cpu);
 bool is_xen_pmu(int cpu);
+bool pmu_msr_read(unsigned int msr, uint64_t *val, int *err);
+bool pmu_msr_write(unsigned int msr, uint32_t low, uint32_t high, int *err);
+int pmu_apic_update(uint32_t reg);
+unsigned long long xen_read_pmc(int counter);
 
 #endif /* __XEN_PMU_H */
diff --git a/include/xen/interface/xenpmu.h b/include/xen/interface/xenpmu.h
index 78073dc..59aaedd 100644
--- a/include/xen/interface/xenpmu.h
+++ b/include/xen/interface/xenpmu.h
@@ -13,6 +13,8 @@
 #define XENPMU_feature_set     3
 #define XENPMU_init            4
 #define XENPMU_finish          5
+#define XENPMU_lvtpc_set       6
+#define XENPMU_flush           7
 
 /* Parameter structure for HYPERVISOR_xenpmu_op call */
 struct xen_pmu_params {
-- 
1.8.1.4