Merge tag 'perf-core-2023-04-27' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull perf updates from Ingo Molnar:

 - Add Intel Granite Rapids support

 - Add uncore events for Intel SPR IMC PMU

 - Fix perf IRQ throttling bug

* tag 'perf-core-2023-04-27' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/intel/uncore: Add events for Intel SPR IMC PMU
  perf/core: Fix hardlockup failure caused by perf throttle
  perf/x86/cstate: Add Granite Rapids support
  perf/x86/msr: Add Granite Rapids
  perf/x86/intel: Add Granite Rapids

Author: torvalds

arch/x86/events/intel/core.c

@@ -5469,6 +5469,15 @@ pebs_is_visible(struct kobject *kobj, struct attribute *attr, int i)
 	return x86_pmu.pebs ? attr->mode : 0;
 }
 
+static umode_t
+mem_is_visible(struct kobject *kobj, struct attribute *attr, int i)
+{
+	if (attr == &event_attr_mem_ld_aux.attr.attr)
+		return x86_pmu.flags & PMU_FL_MEM_LOADS_AUX ? attr->mode : 0;
+
+	return pebs_is_visible(kobj, attr, i);
+}
+
 static umode_t
 lbr_is_visible(struct kobject *kobj, struct attribute *attr, int i)
 {
@@ -5496,7 +5505,7 @@ static struct attribute_group group_events_td  = {
 
 static struct attribute_group group_events_mem = {
 	.name       = "events",
-	.is_visible = pebs_is_visible,
+	.is_visible = mem_is_visible,
 };
 
 static struct attribute_group group_events_tsx = {
@@ -6486,6 +6495,10 @@ __init int intel_pmu_init(void)
 
 	case INTEL_FAM6_SAPPHIRERAPIDS_X:
 	case INTEL_FAM6_EMERALDRAPIDS_X:
+		x86_pmu.flags |= PMU_FL_MEM_LOADS_AUX;
+		fallthrough;
+	case INTEL_FAM6_GRANITERAPIDS_X:
+	case INTEL_FAM6_GRANITERAPIDS_D:
 		pmem = true;
 		x86_pmu.late_ack = true;
 		memcpy(hw_cache_event_ids, spr_hw_cache_event_ids, sizeof(hw_cache_event_ids));
@@ -6502,7 +6515,6 @@ __init int intel_pmu_init(void)
 		x86_pmu.flags |= PMU_FL_HAS_RSP_1;
 		x86_pmu.flags |= PMU_FL_NO_HT_SHARING;
 		x86_pmu.flags |= PMU_FL_INSTR_LATENCY;
-		x86_pmu.flags |= PMU_FL_MEM_LOADS_AUX;
 
 		x86_pmu.hw_config = hsw_hw_config;
 		x86_pmu.get_event_constraints = spr_get_event_constraints;

arch/x86/events/intel/cstate.c

@@ -678,6 +678,8 @@ static const struct x86_cpu_id intel_cstates_match[] __initconst = {
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_D,		&icx_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(SAPPHIRERAPIDS_X,	&icx_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(EMERALDRAPIDS_X,	&icx_cstates),
+	X86_MATCH_INTEL_FAM6_MODEL(GRANITERAPIDS_X,	&icx_cstates),
+	X86_MATCH_INTEL_FAM6_MODEL(GRANITERAPIDS_D,	&icx_cstates),
 
 	X86_MATCH_INTEL_FAM6_MODEL(TIGERLAKE_L,		&icl_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(TIGERLAKE,		&icl_cstates),

arch/x86/events/intel/uncore_snbep.c

@@ -6068,13 +6068,25 @@ static struct intel_uncore_ops spr_uncore_mmio_ops = {
 	.read_counter		= uncore_mmio_read_counter,
 };
 
+static struct uncore_event_desc spr_uncore_imc_events[] = {
+	INTEL_UNCORE_EVENT_DESC(clockticks,      "event=0x01,umask=0x00"),
+	INTEL_UNCORE_EVENT_DESC(cas_count_read,  "event=0x05,umask=0xcf"),
+	INTEL_UNCORE_EVENT_DESC(cas_count_read.scale, "6.103515625e-5"),
+	INTEL_UNCORE_EVENT_DESC(cas_count_read.unit, "MiB"),
+	INTEL_UNCORE_EVENT_DESC(cas_count_write, "event=0x05,umask=0xf0"),
+	INTEL_UNCORE_EVENT_DESC(cas_count_write.scale, "6.103515625e-5"),
+	INTEL_UNCORE_EVENT_DESC(cas_count_write.unit, "MiB"),
+	{ /* end: all zeroes */ },
+};
+
 static struct intel_uncore_type spr_uncore_imc = {
 	SPR_UNCORE_COMMON_FORMAT(),
 	.name			= "imc",
 	.fixed_ctr_bits		= 48,
 	.fixed_ctr		= SNR_IMC_MMIO_PMON_FIXED_CTR,
 	.fixed_ctl		= SNR_IMC_MMIO_PMON_FIXED_CTL,
 	.ops			= &spr_uncore_mmio_ops,
+	.event_descs		= spr_uncore_imc_events,
 };
 
 static void spr_uncore_pci_enable_event(struct intel_uncore_box *box,

arch/x86/events/msr.c

@@ -70,6 +70,8 @@ static bool test_intel(int idx, void *data)
 	case INTEL_FAM6_BROADWELL_X:
 	case INTEL_FAM6_SAPPHIRERAPIDS_X:
 	case INTEL_FAM6_EMERALDRAPIDS_X:
+	case INTEL_FAM6_GRANITERAPIDS_X:
+	case INTEL_FAM6_GRANITERAPIDS_D:
 
 	case INTEL_FAM6_ATOM_SILVERMONT:
 	case INTEL_FAM6_ATOM_SILVERMONT_D:

kernel/events/core.c

@@ -9433,8 +9433,8 @@ __perf_event_account_interrupt(struct perf_event *event, int throttle)
 		hwc->interrupts = 1;
 	} else {
 		hwc->interrupts++;
-		if (unlikely(throttle
-			     && hwc->interrupts >= max_samples_per_tick)) {
+		if (unlikely(throttle &&
+			     hwc->interrupts > max_samples_per_tick)) {
 			__this_cpu_inc(perf_throttled_count);
 			tick_dep_set_cpu(smp_processor_id(), TICK_DEP_BIT_PERF_EVENTS);
 			hwc->interrupts = MAX_INTERRUPTS;