* [PATCH 1/3] perf/x86/amd/uncore: Add common PMU helper functions
2026-07-01 3:44 [PATCH 0/3] perf/x86/amd/uncore: Add common helpers and Hygon support Qi Liu
@ 2026-07-01 3:44 ` Qi Liu
2026-07-01 4:04 ` sashiko-bot
2026-07-01 7:00 ` Peter Zijlstra
2026-07-01 3:44 ` [PATCH 2/3] perf/x86/amd/uncore: Convert AMD driver to common PMU helpers Qi Liu
2026-07-01 3:44 ` [PATCH 3/3] perf/x86/amd/uncore: Add Hygon uncore PMU support Qi Liu
2 siblings, 2 replies; 11+ messages in thread
From: Qi Liu @ 2026-07-01 3:44 UTC (permalink / raw)
To: peterz, mingo, namhyung, alexander.shishkin, jolsa, x86
Cc: linux-perf-users, linux-kernel, lijing, liuqi, Zhenglang Hu
Add common helper functions for AMD-family uncore PMU handling.
The helpers cover event initialization, counter allocation, counter
read/update, event start/stop and per-CPU context management. These
paths are not tied to a specific uncore unit and can be reused by
drivers with a similar uncore PMU programming model.
Signed-off-by: Qi Liu <liuqi@hygon.cn>
Tested-by: Zhenglang Hu <huzhenglang@hygon.cn>
---
arch/x86/events/Kconfig | 4 +
arch/x86/events/amd/Makefile | 2 +
arch/x86/events/amd/uncore_common.c | 403 ++++++++++++++++++++++++++++
arch/x86/events/amd/uncore_common.h | 114 ++++++++
4 files changed, 523 insertions(+)
create mode 100644 arch/x86/events/amd/uncore_common.c
create mode 100644 arch/x86/events/amd/uncore_common.h
diff --git a/arch/x86/events/Kconfig b/arch/x86/events/Kconfig
index dabdf3d7bf84..dafe189d3633 100644
--- a/arch/x86/events/Kconfig
+++ b/arch/x86/events/Kconfig
@@ -34,6 +34,10 @@ config PERF_EVENTS_AMD_POWER
(CPUID Fn8000_0007_EDX[12]) interface to calculate the
average power consumption on Family 15h processors.
+config PERF_EVENTS_UNCORE_COMMON
+ tristate
+ depends on PERF_EVENTS
+
config PERF_EVENTS_AMD_UNCORE
tristate "AMD Uncore performance events"
depends on PERF_EVENTS && CPU_SUP_AMD
diff --git a/arch/x86/events/amd/Makefile b/arch/x86/events/amd/Makefile
index 527d947eb76b..0ddaa049b3b7 100644
--- a/arch/x86/events/amd/Makefile
+++ b/arch/x86/events/amd/Makefile
@@ -3,6 +3,8 @@ obj-$(CONFIG_CPU_SUP_AMD) += core.o lbr.o
obj-$(CONFIG_PERF_EVENTS_AMD_BRS) += brs.o
obj-$(CONFIG_PERF_EVENTS_AMD_POWER) += power.o
obj-$(CONFIG_X86_LOCAL_APIC) += ibs.o
+obj-$(CONFIG_PERF_EVENTS_UNCORE_COMMON) += uncore-common.o
+uncore-common-objs := uncore_common.o
obj-$(CONFIG_PERF_EVENTS_AMD_UNCORE) += amd-uncore.o
amd-uncore-objs := uncore.o
ifdef CONFIG_AMD_IOMMU
diff --git a/arch/x86/events/amd/uncore_common.c b/arch/x86/events/amd/uncore_common.c
new file mode 100644
index 000000000000..81affaab319a
--- /dev/null
+++ b/arch/x86/events/amd/uncore_common.c
@@ -0,0 +1,403 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Common uncore PMU helpers for AMD-family x86 processors.
+ */
+
+#include <linux/cpumask.h>
+#include <linux/hrtimer.h>
+#include <linux/percpu.h>
+#include <linux/perf_event.h>
+#include <linux/slab.h>
+#include <linux/smp.h>
+
+#include <asm/msr.h>
+#include <asm/perf_event.h>
+
+#include "uncore_common.h"
+
+/* Interval for hrtimer, defaults to 60000 milliseconds */
+static unsigned int uncore_update_interval = 60 * MSEC_PER_SEC;
+
+void uncore_common_set_update_interval(unsigned int interval)
+{
+ uncore_update_interval = interval;
+}
+EXPORT_SYMBOL_GPL(uncore_common_set_update_interval);
+
+struct uncore_common_pmu *event_to_uncore_common_pmu(struct perf_event *event)
+{
+ return container_of(event->pmu, struct uncore_common_pmu, pmu);
+}
+EXPORT_SYMBOL_GPL(event_to_uncore_common_pmu);
+
+static ssize_t cpumask_show(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ struct pmu *ptr = dev_get_drvdata(dev);
+ struct uncore_common_pmu *pmu;
+
+ pmu = container_of(ptr, struct uncore_common_pmu, pmu);
+
+ return cpumap_print_to_pagebuf(true, buf, &pmu->active_mask);
+}
+static DEVICE_ATTR_RO(cpumask);
+
+static struct attribute *uncore_common_attrs[] = {
+ &dev_attr_cpumask.attr,
+ NULL,
+};
+
+struct attribute_group uncore_common_attr_group = {
+ .attrs = uncore_common_attrs,
+};
+EXPORT_SYMBOL_GPL(uncore_common_attr_group);
+
+static enum hrtimer_restart uncore_common_hrtimer(struct hrtimer *hrtimer)
+{
+ struct uncore_common_ctx *ctx;
+ struct perf_event *event;
+ int bit;
+
+ ctx = container_of(hrtimer, struct uncore_common_ctx, hrtimer);
+
+ if (!ctx->nr_active || ctx->cpu != smp_processor_id())
+ return HRTIMER_NORESTART;
+
+ for_each_set_bit(bit, ctx->active_mask, NUM_COUNTERS_MAX) {
+ event = ctx->events[bit];
+ event->pmu->read(event);
+ }
+
+ hrtimer_forward_now(hrtimer, ns_to_ktime(ctx->hrtimer_duration));
+
+ return HRTIMER_RESTART;
+}
+
+void uncore_common_start_hrtimer(struct uncore_common_ctx *ctx)
+{
+ hrtimer_start(&ctx->hrtimer, ns_to_ktime(ctx->hrtimer_duration),
+ HRTIMER_MODE_REL_PINNED_HARD);
+}
+EXPORT_SYMBOL_GPL(uncore_common_start_hrtimer);
+
+static void uncore_common_cancel_hrtimer(struct uncore_common_ctx *ctx)
+{
+ hrtimer_cancel(&ctx->hrtimer);
+}
+
+static void uncore_common_init_hrtimer(struct uncore_common_ctx *ctx)
+{
+ hrtimer_setup(&ctx->hrtimer, uncore_common_hrtimer, CLOCK_MONOTONIC,
+ HRTIMER_MODE_REL_HARD);
+}
+
+void uncore_common_read(struct perf_event *event)
+{
+ struct hw_perf_event *hwc = &event->hw;
+ u64 prev, new;
+ s64 delta;
+
+ /*
+ * since we do not enable counter overflow interrupts,
+ * we do not have to worry about prev_count changing on us
+ */
+ prev = local64_read(&hwc->prev_count);
+
+ /*
+ * Some uncore PMUs do not have RDPMC assignments. In such cases,
+ * read counts directly from the corresponding PERF_CTR.
+ */
+ if (hwc->event_base_rdpmc < 0)
+ rdmsrq(hwc->event_base, new);
+ else
+ new = rdpmc(hwc->event_base_rdpmc);
+
+ local64_set(&hwc->prev_count, new);
+
+ delta = (new << COUNTER_SHIFT) - (prev << COUNTER_SHIFT);
+ delta >>= COUNTER_SHIFT;
+
+ local64_add(delta, &event->count);
+}
+EXPORT_SYMBOL_GPL(uncore_common_read);
+
+void uncore_common_start(struct perf_event *event, int flags)
+{
+ struct uncore_common_pmu *pmu = event_to_uncore_common_pmu(event);
+ struct uncore_common_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
+ struct hw_perf_event *hwc = &event->hw;
+
+ if (!ctx->nr_active++)
+ uncore_common_start_hrtimer(ctx);
+
+ if (flags & PERF_EF_RELOAD)
+ wrmsrq(hwc->event_base, (u64)local64_read(&hwc->prev_count));
+
+ hwc->state = 0;
+
+ __set_bit(hwc->idx, ctx->active_mask);
+ wrmsrq(hwc->config_base, (hwc->config | ARCH_PERFMON_EVENTSEL_ENABLE));
+
+ perf_event_update_userpage(event);
+}
+EXPORT_SYMBOL_GPL(uncore_common_start);
+
+void uncore_common_stop(struct perf_event *event, int flags)
+{
+ struct uncore_common_pmu *pmu = event_to_uncore_common_pmu(event);
+ struct uncore_common_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
+ struct hw_perf_event *hwc = &event->hw;
+
+ wrmsrq(hwc->config_base, hwc->config);
+ hwc->state |= PERF_HES_STOPPED;
+
+ if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) {
+ event->pmu->read(event);
+ hwc->state |= PERF_HES_UPTODATE;
+ }
+
+ if (!--ctx->nr_active)
+ uncore_common_cancel_hrtimer(ctx);
+
+ __clear_bit(hwc->idx, ctx->active_mask);
+}
+EXPORT_SYMBOL_GPL(uncore_common_stop);
+
+int uncore_common_event_init(struct perf_event *event)
+{
+ struct uncore_common_pmu *pmu;
+ struct uncore_common_ctx *ctx;
+ struct hw_perf_event *hwc = &event->hw;
+
+ if (event->attr.type != event->pmu->type)
+ return -ENOENT;
+
+ if (event->cpu < 0)
+ return -EINVAL;
+
+ pmu = event_to_uncore_common_pmu(event);
+ ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
+ if (!ctx)
+ return -ENODEV;
+
+ hwc->config = event->attr.config;
+ hwc->idx = -1;
+
+ event->cpu = ctx->cpu;
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(uncore_common_event_init);
+
+int uncore_common_add(struct perf_event *event, int flags)
+{
+ struct uncore_common_pmu *pmu = event_to_uncore_common_pmu(event);
+ struct uncore_common_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
+ struct hw_perf_event *hwc = &event->hw;
+ int i;
+
+ /* are we already assigned? */
+ if (hwc->idx != -1 && ctx->events[hwc->idx] == event)
+ goto out;
+
+ for (i = 0; i < pmu->num_counters; i++) {
+ if (ctx->events[i] == event) {
+ hwc->idx = i;
+ goto out;
+ }
+ }
+
+ /* if not, take the first available counter */
+ hwc->idx = -1;
+
+ for (i = 0; i < pmu->num_counters; i++) {
+ struct perf_event *tmp = NULL;
+
+ if (try_cmpxchg(&ctx->events[i], &tmp, event)) {
+ hwc->idx = i;
+ break;
+ }
+ }
+
+out:
+ if (hwc->idx == -1)
+ return -EBUSY;
+
+ hwc->config_base = pmu->msr_base + (2 * hwc->idx);
+ hwc->event_base = pmu->msr_base + 1 + (2 * hwc->idx);
+ hwc->event_base_rdpmc = pmu->rdpmc_base + hwc->idx;
+ hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED;
+
+ if (pmu->rdpmc_base < 0)
+ hwc->event_base_rdpmc = -1;
+
+ if (flags & PERF_EF_START)
+ event->pmu->start(event, PERF_EF_RELOAD);
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(uncore_common_add);
+
+void uncore_common_del(struct perf_event *event, int flags)
+{
+ struct uncore_common_pmu *pmu = event_to_uncore_common_pmu(event);
+ struct uncore_common_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
+ struct hw_perf_event *hwc = &event->hw;
+ int i;
+
+ event->pmu->stop(event, PERF_EF_UPDATE);
+
+ for (i = 0; i < pmu->num_counters; i++) {
+ struct perf_event *tmp = event;
+
+ if (try_cmpxchg(&ctx->events[i], &tmp, NULL))
+ break;
+ }
+
+ hwc->idx = -1;
+}
+EXPORT_SYMBOL_GPL(uncore_common_del);
+
+int uncore_common_ctx_init(struct uncore_common *uncore, unsigned int cpu)
+{
+ struct uncore_common_ctx *curr, *prev;
+ struct uncore_common_pmu *pmu;
+ int node, cid, gid;
+ int i, j;
+
+ if (!uncore->init_done || !uncore->num_pmus)
+ return 0;
+
+ cid = uncore_common_ctx_cid(uncore, cpu);
+ gid = uncore_common_ctx_gid(uncore, cpu);
+
+ for (i = 0; i < uncore->num_pmus; i++) {
+ pmu = &uncore->pmus[i];
+ *per_cpu_ptr(pmu->ctx, cpu) = NULL;
+ curr = NULL;
+
+ if (gid != pmu->group)
+ continue;
+
+ for_each_online_cpu(j) {
+ if (cpu == j)
+ continue;
+
+ prev = *per_cpu_ptr(pmu->ctx, j);
+ if (!prev)
+ continue;
+
+ if (cid == uncore_common_ctx_cid(uncore, j)) {
+ curr = prev;
+ break;
+ }
+ }
+
+ if (!curr) {
+ node = cpu_to_node(cpu);
+
+ curr = kzalloc_node(sizeof(*curr), GFP_KERNEL, node);
+ if (!curr)
+ goto fail;
+
+ curr->cpu = cpu;
+ curr->events = kzalloc_node(sizeof(*curr->events) *
+ pmu->num_counters,
+ GFP_KERNEL, node);
+ if (!curr->events) {
+ kfree(curr);
+ goto fail;
+ }
+
+ uncore_common_init_hrtimer(curr);
+ curr->hrtimer_duration = (u64)uncore_update_interval * NSEC_PER_MSEC;
+
+ cpumask_set_cpu(cpu, &pmu->active_mask);
+ }
+
+ curr->refcnt++;
+ *per_cpu_ptr(pmu->ctx, cpu) = curr;
+ }
+
+ return 0;
+
+fail:
+ uncore_common_ctx_free(uncore, cpu);
+
+ return -ENOMEM;
+}
+EXPORT_SYMBOL_GPL(uncore_common_ctx_init);
+
+void uncore_common_ctx_free(struct uncore_common *uncore, unsigned int cpu)
+{
+ struct uncore_common_pmu *pmu;
+ struct uncore_common_ctx *ctx;
+ int i;
+
+ if (!uncore->init_done)
+ return;
+
+ for (i = 0; i < uncore->num_pmus; i++) {
+ pmu = &uncore->pmus[i];
+
+ if (!pmu->ctx)
+ continue;
+
+ ctx = *per_cpu_ptr(pmu->ctx, cpu);
+ if (!ctx)
+ continue;
+
+ if (cpu == ctx->cpu)
+ cpumask_clear_cpu(cpu, &pmu->active_mask);
+
+ if (!--ctx->refcnt) {
+ kfree(ctx->events);
+ kfree(ctx);
+ }
+
+ *per_cpu_ptr(pmu->ctx, cpu) = NULL;
+ }
+}
+EXPORT_SYMBOL_GPL(uncore_common_ctx_free);
+
+void uncore_common_ctx_move(struct uncore_common *uncore, unsigned int cpu)
+{
+ struct uncore_common_ctx *curr, *next;
+ struct uncore_common_pmu *pmu;
+ int i, j;
+
+ if (!uncore->init_done)
+ return;
+
+ for (i = 0; i < uncore->num_pmus; i++) {
+ pmu = &uncore->pmus[i];
+ if (!pmu->ctx)
+ continue;
+
+ curr = *per_cpu_ptr(pmu->ctx, cpu);
+ if (!curr)
+ continue;
+
+ for_each_online_cpu(j) {
+ if (cpu == j)
+ continue;
+
+ next = *per_cpu_ptr(pmu->ctx, j);
+ if (!next)
+ continue;
+
+ if (curr == next) {
+ perf_pmu_migrate_context(&pmu->pmu, cpu, j);
+ cpumask_clear_cpu(cpu, &pmu->active_mask);
+ cpumask_set_cpu(j, &pmu->active_mask);
+ next->cpu = j;
+ break;
+ }
+ }
+ }
+}
+EXPORT_SYMBOL_GPL(uncore_common_ctx_move);
+
+MODULE_DESCRIPTION("x86 Uncore common support");
+MODULE_LICENSE("GPL");
diff --git a/arch/x86/events/amd/uncore_common.h b/arch/x86/events/amd/uncore_common.h
new file mode 100644
index 000000000000..c4f0245069e6
--- /dev/null
+++ b/arch/x86/events/amd/uncore_common.h
@@ -0,0 +1,114 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef _X86_EVENTS_UNCORE_COMMON_H
+#define _X86_EVENTS_UNCORE_COMMON_H
+
+#include <linux/cpumask.h>
+#include <linux/device.h>
+#include <linux/hrtimer.h>
+#include <linux/perf_event.h>
+#include <linux/percpu.h>
+#include <linux/sysfs.h>
+#include <linux/types.h>
+
+#define UNCORE_NAME_LEN 16
+#define UNCORE_GROUP_MAX 256
+#define NUM_COUNTERS_MAX 64
+#define COUNTER_SHIFT 16
+
+#define DEFINE_UNCORE_FORMAT_ATTR(_var, _name, _format) \
+static ssize_t __uncore_##_var##_show(struct device *dev, \
+ struct device_attribute *attr, \
+ char *page) \
+{ \
+ BUILD_BUG_ON(sizeof(_format) >= PAGE_SIZE); \
+ return sprintf(page, _format "\n"); \
+} \
+static struct device_attribute format_attr_##_var = \
+ __ATTR(_name, 0444, __uncore_##_var##_show, NULL)
+
+union uncore_common_info {
+ struct {
+ u64 aux_data:32;
+ u64 num_pmcs:8;
+ u64 gid:8;
+ u64 cid:8;
+ u64 private:8;
+ } split;
+ u64 full;
+};
+
+struct uncore_common_ctx {
+ int refcnt;
+ int cpu;
+ struct perf_event **events;
+ unsigned long active_mask[BITS_TO_LONGS(NUM_COUNTERS_MAX)];
+ int nr_active;
+ struct hrtimer hrtimer;
+ u64 hrtimer_duration;
+};
+
+struct uncore_common_pmu {
+ char name[UNCORE_NAME_LEN];
+ int num_counters;
+ int rdpmc_base;
+ u32 msr_base;
+ int group;
+ cpumask_t active_mask;
+ struct pmu pmu;
+ struct uncore_common_ctx * __percpu *ctx;
+ void *private;
+};
+
+struct uncore_common {
+ union uncore_common_info __percpu *info;
+ struct uncore_common_pmu *pmus;
+ unsigned int num_pmus;
+ bool init_done;
+ void (*scan)(struct uncore_common *uncore, unsigned int cpu);
+ int (*init)(struct uncore_common *uncore, unsigned int cpu);
+ void (*move)(struct uncore_common *uncore, unsigned int cpu);
+ void (*free)(struct uncore_common *uncore, unsigned int cpu);
+};
+
+extern struct attribute_group uncore_common_attr_group;
+
+static inline int uncore_common_ctx_cid(struct uncore_common *uncore,
+ unsigned int cpu)
+{
+ union uncore_common_info *info = per_cpu_ptr(uncore->info, cpu);
+
+ return info->split.cid;
+}
+
+static inline int uncore_common_ctx_gid(struct uncore_common *uncore,
+ unsigned int cpu)
+{
+ union uncore_common_info *info = per_cpu_ptr(uncore->info, cpu);
+
+ return info->split.gid;
+}
+
+static inline int uncore_common_ctx_num_pmcs(struct uncore_common *uncore,
+ unsigned int cpu)
+{
+ union uncore_common_info *info = per_cpu_ptr(uncore->info, cpu);
+
+ return info->split.num_pmcs;
+}
+
+struct uncore_common_pmu *event_to_uncore_common_pmu(struct perf_event *event);
+
+void uncore_common_set_update_interval(unsigned int interval);
+int uncore_common_event_init(struct perf_event *event);
+int uncore_common_add(struct perf_event *event, int flags);
+void uncore_common_del(struct perf_event *event, int flags);
+void uncore_common_start(struct perf_event *event, int flags);
+void uncore_common_stop(struct perf_event *event, int flags);
+void uncore_common_read(struct perf_event *event);
+
+int uncore_common_ctx_init(struct uncore_common *uncore, unsigned int cpu);
+void uncore_common_ctx_free(struct uncore_common *uncore, unsigned int cpu);
+void uncore_common_ctx_move(struct uncore_common *uncore, unsigned int cpu);
+void uncore_common_start_hrtimer(struct uncore_common_ctx *ctx);
+
+#endif /* _X86_EVENTS_UNCORE_COMMON_H */
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCH 2/3] perf/x86/amd/uncore: Convert AMD driver to common PMU helpers
2026-07-01 3:44 [PATCH 0/3] perf/x86/amd/uncore: Add common helpers and Hygon support Qi Liu
2026-07-01 3:44 ` [PATCH 1/3] perf/x86/amd/uncore: Add common PMU helper functions Qi Liu
@ 2026-07-01 3:44 ` Qi Liu
2026-07-01 4:00 ` sashiko-bot
2026-07-01 3:44 ` [PATCH 3/3] perf/x86/amd/uncore: Add Hygon uncore PMU support Qi Liu
2 siblings, 1 reply; 11+ messages in thread
From: Qi Liu @ 2026-07-01 3:44 UTC (permalink / raw)
To: peterz, mingo, namhyung, alexander.shishkin, jolsa, x86
Cc: linux-perf-users, linux-kernel, lijing, liuqi, Zhenglang Hu
Use the common uncore PMU helpers for AMD uncore event handling and
per-CPU context management.
The AMD-specific DF, L3 and UMC discovery, event encoding and PMU setup
logic remains in the AMD driver. This only replaces duplicated common
operations with calls to the helper functions.
Signed-off-by: Qi Liu <liuqi@hygon.cn>
Tested-by: Zhenglang Hu <huzhenglang@hygon.cn>
---
arch/x86/events/Kconfig | 1 +
arch/x86/events/amd/uncore.c | 578 +++++------------------------------
2 files changed, 74 insertions(+), 505 deletions(-)
diff --git a/arch/x86/events/Kconfig b/arch/x86/events/Kconfig
index dafe189d3633..466581535ae1 100644
--- a/arch/x86/events/Kconfig
+++ b/arch/x86/events/Kconfig
@@ -41,6 +41,7 @@ config PERF_EVENTS_UNCORE_COMMON
config PERF_EVENTS_AMD_UNCORE
tristate "AMD Uncore performance events"
depends on PERF_EVENTS && CPU_SUP_AMD
+ select PERF_EVENTS_UNCORE_COMMON
default y
help
Include support for AMD uncore performance events for use with
diff --git a/arch/x86/events/amd/uncore.c b/arch/x86/events/amd/uncore.c
index dbc00b6dd69e..9cb8b0d1afe7 100644
--- a/arch/x86/events/amd/uncore.c
+++ b/arch/x86/events/amd/uncore.c
@@ -19,6 +19,8 @@
#include <asm/cpuid/api.h>
#include <asm/msr.h>
+#include "uncore_common.h"
+
#define NUM_COUNTERS_NB 4
#define NUM_COUNTERS_L2 4
#define NUM_COUNTERS_L3 6
@@ -36,27 +38,6 @@
static int pmu_version;
-struct amd_uncore_ctx {
- int refcnt;
- int cpu;
- struct perf_event **events;
- unsigned long active_mask[BITS_TO_LONGS(NUM_COUNTERS_MAX)];
- int nr_active;
- struct hrtimer hrtimer;
- u64 hrtimer_duration;
-};
-
-struct amd_uncore_pmu {
- char name[UNCORE_NAME_LEN];
- int num_counters;
- int rdpmc_base;
- u32 msr_base;
- int group;
- cpumask_t active_mask;
- struct pmu pmu;
- struct amd_uncore_ctx * __percpu *ctx;
-};
-
enum {
UNCORE_TYPE_DF,
UNCORE_TYPE_L3,
@@ -65,243 +46,12 @@ enum {
UNCORE_TYPE_MAX
};
-union amd_uncore_info {
- struct {
- u64 aux_data:32; /* auxiliary data */
- u64 num_pmcs:8; /* number of counters */
- u64 gid:8; /* group id */
- u64 cid:8; /* context id */
- } split;
- u64 full;
-};
-
-struct amd_uncore {
- union amd_uncore_info __percpu *info;
- struct amd_uncore_pmu *pmus;
- unsigned int num_pmus;
- bool init_done;
- void (*scan)(struct amd_uncore *uncore, unsigned int cpu);
- int (*init)(struct amd_uncore *uncore, unsigned int cpu);
- void (*move)(struct amd_uncore *uncore, unsigned int cpu);
- void (*free)(struct amd_uncore *uncore, unsigned int cpu);
-};
-
-static struct amd_uncore uncores[UNCORE_TYPE_MAX];
+static struct uncore_common uncores[UNCORE_TYPE_MAX];
/* Interval for hrtimer, defaults to 60000 milliseconds */
static unsigned int update_interval = 60 * MSEC_PER_SEC;
module_param(update_interval, uint, 0444);
-static struct amd_uncore_pmu *event_to_amd_uncore_pmu(struct perf_event *event)
-{
- return container_of(event->pmu, struct amd_uncore_pmu, pmu);
-}
-
-static enum hrtimer_restart amd_uncore_hrtimer(struct hrtimer *hrtimer)
-{
- struct amd_uncore_ctx *ctx;
- struct perf_event *event;
- int bit;
-
- ctx = container_of(hrtimer, struct amd_uncore_ctx, hrtimer);
-
- if (!ctx->nr_active || ctx->cpu != smp_processor_id())
- return HRTIMER_NORESTART;
-
- for_each_set_bit(bit, ctx->active_mask, NUM_COUNTERS_MAX) {
- event = ctx->events[bit];
- event->pmu->read(event);
- }
-
- hrtimer_forward_now(hrtimer, ns_to_ktime(ctx->hrtimer_duration));
- return HRTIMER_RESTART;
-}
-
-static void amd_uncore_start_hrtimer(struct amd_uncore_ctx *ctx)
-{
- hrtimer_start(&ctx->hrtimer, ns_to_ktime(ctx->hrtimer_duration),
- HRTIMER_MODE_REL_PINNED_HARD);
-}
-
-static void amd_uncore_cancel_hrtimer(struct amd_uncore_ctx *ctx)
-{
- hrtimer_cancel(&ctx->hrtimer);
-}
-
-static void amd_uncore_init_hrtimer(struct amd_uncore_ctx *ctx)
-{
- hrtimer_setup(&ctx->hrtimer, amd_uncore_hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_HARD);
-}
-
-static void amd_uncore_read(struct perf_event *event)
-{
- struct hw_perf_event *hwc = &event->hw;
- u64 prev, new;
- s64 delta;
-
- /*
- * since we do not enable counter overflow interrupts,
- * we do not have to worry about prev_count changing on us
- */
-
- prev = local64_read(&hwc->prev_count);
-
- /*
- * Some uncore PMUs do not have RDPMC assignments. In such cases,
- * read counts directly from the corresponding PERF_CTR.
- */
- if (hwc->event_base_rdpmc < 0)
- rdmsrq(hwc->event_base, new);
- else
- new = rdpmc(hwc->event_base_rdpmc);
-
- local64_set(&hwc->prev_count, new);
- delta = (new << COUNTER_SHIFT) - (prev << COUNTER_SHIFT);
- delta >>= COUNTER_SHIFT;
- local64_add(delta, &event->count);
-}
-
-static void amd_uncore_start(struct perf_event *event, int flags)
-{
- struct amd_uncore_pmu *pmu = event_to_amd_uncore_pmu(event);
- struct amd_uncore_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
- struct hw_perf_event *hwc = &event->hw;
-
- if (!ctx->nr_active++)
- amd_uncore_start_hrtimer(ctx);
-
- if (flags & PERF_EF_RELOAD)
- wrmsrq(hwc->event_base, (u64)local64_read(&hwc->prev_count));
-
- hwc->state = 0;
- __set_bit(hwc->idx, ctx->active_mask);
- wrmsrq(hwc->config_base, (hwc->config | ARCH_PERFMON_EVENTSEL_ENABLE));
- perf_event_update_userpage(event);
-}
-
-static void amd_uncore_stop(struct perf_event *event, int flags)
-{
- struct amd_uncore_pmu *pmu = event_to_amd_uncore_pmu(event);
- struct amd_uncore_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
- struct hw_perf_event *hwc = &event->hw;
-
- wrmsrq(hwc->config_base, hwc->config);
- hwc->state |= PERF_HES_STOPPED;
-
- if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) {
- event->pmu->read(event);
- hwc->state |= PERF_HES_UPTODATE;
- }
-
- if (!--ctx->nr_active)
- amd_uncore_cancel_hrtimer(ctx);
-
- __clear_bit(hwc->idx, ctx->active_mask);
-}
-
-static int amd_uncore_add(struct perf_event *event, int flags)
-{
- int i;
- struct amd_uncore_pmu *pmu = event_to_amd_uncore_pmu(event);
- struct amd_uncore_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
- struct hw_perf_event *hwc = &event->hw;
-
- /* are we already assigned? */
- if (hwc->idx != -1 && ctx->events[hwc->idx] == event)
- goto out;
-
- for (i = 0; i < pmu->num_counters; i++) {
- if (ctx->events[i] == event) {
- hwc->idx = i;
- goto out;
- }
- }
-
- /* if not, take the first available counter */
- hwc->idx = -1;
- for (i = 0; i < pmu->num_counters; i++) {
- struct perf_event *tmp = NULL;
-
- if (try_cmpxchg(&ctx->events[i], &tmp, event)) {
- hwc->idx = i;
- break;
- }
- }
-
-out:
- if (hwc->idx == -1)
- return -EBUSY;
-
- hwc->config_base = pmu->msr_base + (2 * hwc->idx);
- hwc->event_base = pmu->msr_base + 1 + (2 * hwc->idx);
- hwc->event_base_rdpmc = pmu->rdpmc_base + hwc->idx;
- hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED;
-
- if (pmu->rdpmc_base < 0)
- hwc->event_base_rdpmc = -1;
-
- if (flags & PERF_EF_START)
- event->pmu->start(event, PERF_EF_RELOAD);
-
- return 0;
-}
-
-static void amd_uncore_del(struct perf_event *event, int flags)
-{
- int i;
- struct amd_uncore_pmu *pmu = event_to_amd_uncore_pmu(event);
- struct amd_uncore_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
- struct hw_perf_event *hwc = &event->hw;
-
- event->pmu->stop(event, PERF_EF_UPDATE);
-
- for (i = 0; i < pmu->num_counters; i++) {
- struct perf_event *tmp = event;
-
- if (try_cmpxchg(&ctx->events[i], &tmp, NULL))
- break;
- }
-
- hwc->idx = -1;
-}
-
-static int amd_uncore_event_init(struct perf_event *event)
-{
- struct amd_uncore_pmu *pmu;
- struct amd_uncore_ctx *ctx;
- struct hw_perf_event *hwc = &event->hw;
-
- if (event->attr.type != event->pmu->type)
- return -ENOENT;
-
- if (event->cpu < 0)
- return -EINVAL;
-
- pmu = event_to_amd_uncore_pmu(event);
- ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
- if (!ctx)
- return -ENODEV;
-
- /*
- * NB and Last level cache counters (MSRs) are shared across all cores
- * that share the same NB / Last level cache. On family 16h and below,
- * Interrupts can be directed to a single target core, however, event
- * counts generated by processes running on other cores cannot be masked
- * out. So we do not support sampling and per-thread events via
- * CAP_NO_INTERRUPT, and we do not enable counter overflow interrupts:
- */
- hwc->config = event->attr.config;
- hwc->idx = -1;
-
- /*
- * since request can come in to any of the shared cores, we will remap
- * to a single common cpu.
- */
- event->cpu = ctx->cpu;
-
- return 0;
-}
-
static umode_t
amd_f17h_uncore_is_visible(struct kobject *kobj, struct attribute *attr, int i)
{
@@ -315,37 +65,6 @@ amd_f19h_uncore_is_visible(struct kobject *kobj, struct attribute *attr, int i)
return boot_cpu_data.x86 >= 0x19 ? attr->mode : 0;
}
-static ssize_t amd_uncore_attr_show_cpumask(struct device *dev,
- struct device_attribute *attr,
- char *buf)
-{
- struct pmu *ptr = dev_get_drvdata(dev);
- struct amd_uncore_pmu *pmu = container_of(ptr, struct amd_uncore_pmu, pmu);
-
- return cpumap_print_to_pagebuf(true, buf, &pmu->active_mask);
-}
-static DEVICE_ATTR(cpumask, S_IRUGO, amd_uncore_attr_show_cpumask, NULL);
-
-static struct attribute *amd_uncore_attrs[] = {
- &dev_attr_cpumask.attr,
- NULL,
-};
-
-static struct attribute_group amd_uncore_attr_group = {
- .attrs = amd_uncore_attrs,
-};
-
-#define DEFINE_UNCORE_FORMAT_ATTR(_var, _name, _format) \
-static ssize_t __uncore_##_var##_show(struct device *dev, \
- struct device_attribute *attr, \
- char *page) \
-{ \
- BUILD_BUG_ON(sizeof(_format) >= PAGE_SIZE); \
- return sprintf(page, _format "\n"); \
-} \
-static struct device_attribute format_attr_##_var = \
- __ATTR(_name, 0444, __uncore_##_var##_show, NULL)
-
DEFINE_UNCORE_FORMAT_ATTR(event12, event, "config:0-7,32-35");
DEFINE_UNCORE_FORMAT_ATTR(event14, event, "config:0-7,32-35,59-60"); /* F17h+ DF */
DEFINE_UNCORE_FORMAT_ATTR(event14v2, event, "config:0-7,32-37"); /* PerfMonV2 DF */
@@ -426,13 +145,13 @@ static struct attribute_group amd_uncore_umc_format_group = {
};
static const struct attribute_group *amd_uncore_df_attr_groups[] = {
- &amd_uncore_attr_group,
+ &uncore_common_attr_group,
&amd_uncore_df_format_group,
NULL,
};
static const struct attribute_group *amd_uncore_l3_attr_groups[] = {
- &amd_uncore_attr_group,
+ &uncore_common_attr_group,
&amd_uncore_l3_format_group,
NULL,
};
@@ -444,164 +163,14 @@ static const struct attribute_group *amd_uncore_l3_attr_update[] = {
};
static const struct attribute_group *amd_uncore_umc_attr_groups[] = {
- &amd_uncore_attr_group,
+ &uncore_common_attr_group,
&amd_uncore_umc_format_group,
NULL,
};
-static __always_inline
-int amd_uncore_ctx_cid(struct amd_uncore *uncore, unsigned int cpu)
-{
- union amd_uncore_info *info = per_cpu_ptr(uncore->info, cpu);
- return info->split.cid;
-}
-
-static __always_inline
-int amd_uncore_ctx_gid(struct amd_uncore *uncore, unsigned int cpu)
-{
- union amd_uncore_info *info = per_cpu_ptr(uncore->info, cpu);
- return info->split.gid;
-}
-
-static __always_inline
-int amd_uncore_ctx_num_pmcs(struct amd_uncore *uncore, unsigned int cpu)
-{
- union amd_uncore_info *info = per_cpu_ptr(uncore->info, cpu);
- return info->split.num_pmcs;
-}
-
-static void amd_uncore_ctx_free(struct amd_uncore *uncore, unsigned int cpu)
-{
- struct amd_uncore_pmu *pmu;
- struct amd_uncore_ctx *ctx;
- int i;
-
- if (!uncore->init_done)
- return;
-
- for (i = 0; i < uncore->num_pmus; i++) {
- pmu = &uncore->pmus[i];
- ctx = *per_cpu_ptr(pmu->ctx, cpu);
- if (!ctx)
- continue;
-
- if (cpu == ctx->cpu)
- cpumask_clear_cpu(cpu, &pmu->active_mask);
-
- if (!--ctx->refcnt) {
- kfree(ctx->events);
- kfree(ctx);
- }
-
- *per_cpu_ptr(pmu->ctx, cpu) = NULL;
- }
-}
-
-static int amd_uncore_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
-{
- struct amd_uncore_ctx *curr, *prev;
- struct amd_uncore_pmu *pmu;
- int node, cid, gid, i, j;
-
- if (!uncore->init_done || !uncore->num_pmus)
- return 0;
-
- cid = amd_uncore_ctx_cid(uncore, cpu);
- gid = amd_uncore_ctx_gid(uncore, cpu);
-
- for (i = 0; i < uncore->num_pmus; i++) {
- pmu = &uncore->pmus[i];
- *per_cpu_ptr(pmu->ctx, cpu) = NULL;
- curr = NULL;
-
- /* Check for group exclusivity */
- if (gid != pmu->group)
- continue;
-
- /* Find a sibling context */
- for_each_online_cpu(j) {
- if (cpu == j)
- continue;
-
- prev = *per_cpu_ptr(pmu->ctx, j);
- if (!prev)
- continue;
-
- if (cid == amd_uncore_ctx_cid(uncore, j)) {
- curr = prev;
- break;
- }
- }
-
- /* Allocate context if sibling does not exist */
- if (!curr) {
- node = cpu_to_node(cpu);
- curr = kzalloc_node(sizeof(*curr), GFP_KERNEL, node);
- if (!curr)
- goto fail;
-
- curr->cpu = cpu;
- curr->events = kzalloc_node(sizeof(*curr->events) *
- pmu->num_counters,
- GFP_KERNEL, node);
- if (!curr->events) {
- kfree(curr);
- goto fail;
- }
-
- amd_uncore_init_hrtimer(curr);
- curr->hrtimer_duration = (u64)update_interval * NSEC_PER_MSEC;
-
- cpumask_set_cpu(cpu, &pmu->active_mask);
- }
-
- curr->refcnt++;
- *per_cpu_ptr(pmu->ctx, cpu) = curr;
- }
-
- return 0;
-
-fail:
- amd_uncore_ctx_free(uncore, cpu);
-
- return -ENOMEM;
-}
-
-static void amd_uncore_ctx_move(struct amd_uncore *uncore, unsigned int cpu)
-{
- struct amd_uncore_ctx *curr, *next;
- struct amd_uncore_pmu *pmu;
- int i, j;
-
- if (!uncore->init_done)
- return;
-
- for (i = 0; i < uncore->num_pmus; i++) {
- pmu = &uncore->pmus[i];
- curr = *per_cpu_ptr(pmu->ctx, cpu);
- if (!curr)
- continue;
-
- /* Migrate to a shared sibling if possible */
- for_each_online_cpu(j) {
- next = *per_cpu_ptr(pmu->ctx, j);
- if (!next || cpu == j)
- continue;
-
- if (curr == next) {
- perf_pmu_migrate_context(&pmu->pmu, cpu, j);
- cpumask_clear_cpu(cpu, &pmu->active_mask);
- cpumask_set_cpu(j, &pmu->active_mask);
- next->cpu = j;
- break;
- }
- }
- }
-}
-
static int amd_uncore_cpu_starting(unsigned int cpu)
{
- struct amd_uncore *uncore;
+ struct uncore_common *uncore;
int i;
for (i = 0; i < UNCORE_TYPE_MAX; i++) {
@@ -614,7 +183,7 @@ static int amd_uncore_cpu_starting(unsigned int cpu)
static int amd_uncore_cpu_online(unsigned int cpu)
{
- struct amd_uncore *uncore;
+ struct uncore_common *uncore;
int i;
for (i = 0; i < UNCORE_TYPE_MAX; i++) {
@@ -628,7 +197,7 @@ static int amd_uncore_cpu_online(unsigned int cpu)
static int amd_uncore_cpu_down_prepare(unsigned int cpu)
{
- struct amd_uncore *uncore;
+ struct uncore_common *uncore;
int i;
for (i = 0; i < UNCORE_TYPE_MAX; i++) {
@@ -641,7 +210,7 @@ static int amd_uncore_cpu_down_prepare(unsigned int cpu)
static int amd_uncore_cpu_dead(unsigned int cpu)
{
- struct amd_uncore *uncore;
+ struct uncore_common *uncore;
int i;
for (i = 0; i < UNCORE_TYPE_MAX; i++) {
@@ -655,7 +224,7 @@ static int amd_uncore_cpu_dead(unsigned int cpu)
static int amd_uncore_df_event_init(struct perf_event *event)
{
struct hw_perf_event *hwc = &event->hw;
- int ret = amd_uncore_event_init(event);
+ int ret = uncore_common_event_init(event);
hwc->config = event->attr.config &
(pmu_version >= 2 ? AMD64_PERFMON_V2_RAW_EVENT_MASK_NB :
@@ -666,7 +235,7 @@ static int amd_uncore_df_event_init(struct perf_event *event)
static int amd_uncore_df_add(struct perf_event *event, int flags)
{
- int ret = amd_uncore_add(event, flags & ~PERF_EF_START);
+ int ret = uncore_common_add(event, flags & ~PERF_EF_START);
struct hw_perf_event *hwc = &event->hw;
if (ret)
@@ -684,16 +253,16 @@ static int amd_uncore_df_add(struct perf_event *event, int flags)
/* Delayed start after rdpmc base update */
if (flags & PERF_EF_START)
- amd_uncore_start(event, PERF_EF_RELOAD);
+ uncore_common_start(event, PERF_EF_RELOAD);
return 0;
}
static
-void amd_uncore_df_ctx_scan(struct amd_uncore *uncore, unsigned int cpu)
+void amd_uncore_df_ctx_scan(struct uncore_common *uncore, unsigned int cpu)
{
union cpuid_0x80000022_ebx ebx;
- union amd_uncore_info info;
+ union uncore_common_info info;
if (!boot_cpu_has(X86_FEATURE_PERFCTR_NB))
return;
@@ -712,17 +281,17 @@ void amd_uncore_df_ctx_scan(struct amd_uncore *uncore, unsigned int cpu)
}
static
-int amd_uncore_df_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
+int amd_uncore_df_ctx_init(struct uncore_common *uncore, unsigned int cpu)
{
struct attribute **df_attr = amd_uncore_df_format_attr;
- struct amd_uncore_pmu *pmu;
+ struct uncore_common_pmu *pmu;
int num_counters;
/* Run just once */
if (uncore->init_done)
- return amd_uncore_ctx_init(uncore, cpu);
+ return uncore_common_ctx_init(uncore, cpu);
- num_counters = amd_uncore_ctx_num_pmcs(uncore, cpu);
+ num_counters = uncore_common_ctx_num_pmcs(uncore, cpu);
if (!num_counters)
goto done;
@@ -742,7 +311,7 @@ int amd_uncore_df_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
pmu->num_counters = num_counters;
pmu->msr_base = MSR_F15H_NB_PERF_CTL;
pmu->rdpmc_base = RDPMC_BASE_NB;
- pmu->group = amd_uncore_ctx_gid(uncore, cpu);
+ pmu->group = uncore_common_ctx_gid(uncore, cpu);
if (pmu_version >= 2) {
*df_attr++ = &format_attr_event14v2.attr;
@@ -751,7 +320,7 @@ int amd_uncore_df_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
*df_attr = &format_attr_event14.attr;
}
- pmu->ctx = alloc_percpu(struct amd_uncore_ctx *);
+ pmu->ctx = alloc_percpu(struct uncore_common_ctx *);
if (!pmu->ctx)
goto done;
@@ -761,10 +330,10 @@ int amd_uncore_df_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
.name = pmu->name,
.event_init = amd_uncore_df_event_init,
.add = amd_uncore_df_add,
- .del = amd_uncore_del,
- .start = amd_uncore_start,
- .stop = amd_uncore_stop,
- .read = amd_uncore_read,
+ .del = uncore_common_del,
+ .start = uncore_common_start,
+ .stop = uncore_common_stop,
+ .read = uncore_common_read,
.capabilities = PERF_PMU_CAP_NO_EXCLUDE | PERF_PMU_CAP_NO_INTERRUPT,
.module = THIS_MODULE,
};
@@ -775,8 +344,7 @@ int amd_uncore_df_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
goto done;
}
- pr_info("%d %s%s counters detected\n", pmu->num_counters,
- boot_cpu_data.x86_vendor == X86_VENDOR_HYGON ? "HYGON " : "",
+ pr_info("%d %s counters detected\n", pmu->num_counters,
pmu->pmu.name);
uncore->num_pmus = 1;
@@ -784,12 +352,12 @@ int amd_uncore_df_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
done:
uncore->init_done = true;
- return amd_uncore_ctx_init(uncore, cpu);
+ return uncore_common_ctx_init(uncore, cpu);
}
static int amd_uncore_l3_event_init(struct perf_event *event)
{
- int ret = amd_uncore_event_init(event);
+ int ret = uncore_common_event_init(event);
struct hw_perf_event *hwc = &event->hw;
u64 config = event->attr.config;
u64 mask;
@@ -827,9 +395,9 @@ static int amd_uncore_l3_event_init(struct perf_event *event)
}
static
-void amd_uncore_l3_ctx_scan(struct amd_uncore *uncore, unsigned int cpu)
+void amd_uncore_l3_ctx_scan(struct uncore_common *uncore, unsigned int cpu)
{
- union amd_uncore_info info;
+ union uncore_common_info info;
if (!boot_cpu_has(X86_FEATURE_PERFCTR_LLC))
return;
@@ -846,17 +414,17 @@ void amd_uncore_l3_ctx_scan(struct amd_uncore *uncore, unsigned int cpu)
}
static
-int amd_uncore_l3_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
+int amd_uncore_l3_ctx_init(struct uncore_common *uncore, unsigned int cpu)
{
struct attribute **l3_attr = amd_uncore_l3_format_attr;
- struct amd_uncore_pmu *pmu;
+ struct uncore_common_pmu *pmu;
int num_counters;
/* Run just once */
if (uncore->init_done)
- return amd_uncore_ctx_init(uncore, cpu);
+ return uncore_common_ctx_init(uncore, cpu);
- num_counters = amd_uncore_ctx_num_pmcs(uncore, cpu);
+ num_counters = uncore_common_ctx_num_pmcs(uncore, cpu);
if (!num_counters)
goto done;
@@ -876,7 +444,7 @@ int amd_uncore_l3_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
pmu->num_counters = num_counters;
pmu->msr_base = MSR_F16H_L2I_PERF_CTL;
pmu->rdpmc_base = RDPMC_BASE_LLC;
- pmu->group = amd_uncore_ctx_gid(uncore, cpu);
+ pmu->group = uncore_common_ctx_gid(uncore, cpu);
if (boot_cpu_data.x86 >= 0x17) {
*l3_attr++ = &format_attr_event8.attr;
@@ -886,7 +454,7 @@ int amd_uncore_l3_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
&format_attr_threadmask8.attr;
}
- pmu->ctx = alloc_percpu(struct amd_uncore_ctx *);
+ pmu->ctx = alloc_percpu(struct uncore_common_ctx *);
if (!pmu->ctx)
goto done;
@@ -896,11 +464,11 @@ int amd_uncore_l3_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
.attr_update = amd_uncore_l3_attr_update,
.name = pmu->name,
.event_init = amd_uncore_l3_event_init,
- .add = amd_uncore_add,
- .del = amd_uncore_del,
- .start = amd_uncore_start,
- .stop = amd_uncore_stop,
- .read = amd_uncore_read,
+ .add = uncore_common_add,
+ .del = uncore_common_del,
+ .start = uncore_common_start,
+ .stop = uncore_common_stop,
+ .read = uncore_common_read,
.capabilities = PERF_PMU_CAP_NO_EXCLUDE | PERF_PMU_CAP_NO_INTERRUPT,
.module = THIS_MODULE,
};
@@ -911,8 +479,7 @@ int amd_uncore_l3_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
goto done;
}
- pr_info("%d %s%s counters detected\n", pmu->num_counters,
- boot_cpu_data.x86_vendor == X86_VENDOR_HYGON ? "HYGON " : "",
+ pr_info("%d %s counters detected\n", pmu->num_counters,
pmu->pmu.name);
uncore->num_pmus = 1;
@@ -920,14 +487,14 @@ int amd_uncore_l3_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
done:
uncore->init_done = true;
- return amd_uncore_ctx_init(uncore, cpu);
+ return uncore_common_ctx_init(uncore, cpu);
}
static int amd_uncore_umc_event_init(struct perf_event *event)
{
struct hw_perf_event *hwc = &event->hw;
- int ret = amd_uncore_event_init(event);
+ int ret = uncore_common_event_init(event);
if (ret)
return ret;
@@ -938,12 +505,12 @@ static int amd_uncore_umc_event_init(struct perf_event *event)
static void amd_uncore_umc_start(struct perf_event *event, int flags)
{
- struct amd_uncore_pmu *pmu = event_to_amd_uncore_pmu(event);
- struct amd_uncore_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
+ struct uncore_common_pmu *pmu = event_to_uncore_common_pmu(event);
+ struct uncore_common_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
struct hw_perf_event *hwc = &event->hw;
if (!ctx->nr_active++)
- amd_uncore_start_hrtimer(ctx);
+ uncore_common_start_hrtimer(ctx);
if (flags & PERF_EF_RELOAD)
wrmsrq(hwc->event_base, (u64)local64_read(&hwc->prev_count));
@@ -988,10 +555,10 @@ static void amd_uncore_umc_read(struct perf_event *event)
}
static
-void amd_uncore_umc_ctx_scan(struct amd_uncore *uncore, unsigned int cpu)
+void amd_uncore_umc_ctx_scan(struct uncore_common *uncore, unsigned int cpu)
{
union cpuid_0x80000022_ebx ebx;
- union amd_uncore_info info;
+ union uncore_common_info info;
unsigned int eax, ecx, edx;
if (pmu_version < 2)
@@ -1006,13 +573,13 @@ void amd_uncore_umc_ctx_scan(struct amd_uncore *uncore, unsigned int cpu)
}
static
-int amd_uncore_umc_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
+int amd_uncore_umc_ctx_init(struct uncore_common *uncore, unsigned int cpu)
{
DECLARE_BITMAP(gmask, UNCORE_GROUP_MAX) = { 0 };
u8 group_num_pmus[UNCORE_GROUP_MAX] = { 0 };
u8 group_num_pmcs[UNCORE_GROUP_MAX] = { 0 };
- union amd_uncore_info info;
- struct amd_uncore_pmu *pmu;
+ union uncore_common_info info;
+ struct uncore_common_pmu *pmu;
int gid, i;
u16 index = 0;
@@ -1021,12 +588,13 @@ int amd_uncore_umc_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
/* Run just once */
if (uncore->init_done)
- return amd_uncore_ctx_init(uncore, cpu);
+ return uncore_common_ctx_init(uncore, cpu);
/* Find unique groups */
for_each_online_cpu(i) {
info = *per_cpu_ptr(uncore->info, i);
gid = info.split.gid;
+
if (test_bit(gid, gmask))
continue;
@@ -1051,8 +619,7 @@ int amd_uncore_umc_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
pmu->msr_base = MSR_F19H_UMC_PERF_CTL + i * pmu->num_counters * 2;
pmu->rdpmc_base = -1;
pmu->group = gid;
-
- pmu->ctx = alloc_percpu(struct amd_uncore_ctx *);
+ pmu->ctx = alloc_percpu(struct uncore_common_ctx *);
if (!pmu->ctx)
goto done;
@@ -1061,10 +628,10 @@ int amd_uncore_umc_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
.attr_groups = amd_uncore_umc_attr_groups,
.name = pmu->name,
.event_init = amd_uncore_umc_event_init,
- .add = amd_uncore_add,
- .del = amd_uncore_del,
+ .add = uncore_common_add,
+ .del = uncore_common_del,
.start = amd_uncore_umc_start,
- .stop = amd_uncore_stop,
+ .stop = uncore_common_stop,
.read = amd_uncore_umc_read,
.capabilities = PERF_PMU_CAP_NO_EXCLUDE | PERF_PMU_CAP_NO_INTERRUPT,
.module = THIS_MODULE,
@@ -1087,41 +654,40 @@ int amd_uncore_umc_ctx_init(struct amd_uncore *uncore, unsigned int cpu)
uncore->num_pmus = index;
uncore->init_done = true;
- return amd_uncore_ctx_init(uncore, cpu);
+ return uncore_common_ctx_init(uncore, cpu);
}
-static struct amd_uncore uncores[UNCORE_TYPE_MAX] = {
+static struct uncore_common uncores[UNCORE_TYPE_MAX] = {
/* UNCORE_TYPE_DF */
{
.scan = amd_uncore_df_ctx_scan,
.init = amd_uncore_df_ctx_init,
- .move = amd_uncore_ctx_move,
- .free = amd_uncore_ctx_free,
+ .move = uncore_common_ctx_move,
+ .free = uncore_common_ctx_free,
},
/* UNCORE_TYPE_L3 */
{
.scan = amd_uncore_l3_ctx_scan,
.init = amd_uncore_l3_ctx_init,
- .move = amd_uncore_ctx_move,
- .free = amd_uncore_ctx_free,
+ .move = uncore_common_ctx_move,
+ .free = uncore_common_ctx_free,
},
/* UNCORE_TYPE_UMC */
{
.scan = amd_uncore_umc_ctx_scan,
.init = amd_uncore_umc_ctx_init,
- .move = amd_uncore_ctx_move,
- .free = amd_uncore_ctx_free,
+ .move = uncore_common_ctx_move,
+ .free = uncore_common_ctx_free,
},
};
static int __init amd_uncore_init(void)
{
- struct amd_uncore *uncore;
+ struct uncore_common *uncore;
int ret = -ENODEV;
int i;
- if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD &&
- boot_cpu_data.x86_vendor != X86_VENDOR_HYGON)
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD)
return -ENODEV;
if (!boot_cpu_has(X86_FEATURE_TOPOEXT))
@@ -1130,6 +696,8 @@ static int __init amd_uncore_init(void)
if (boot_cpu_has(X86_FEATURE_PERFMON_V2))
pmu_version = 2;
+ uncore_common_set_update_interval(update_interval);
+
for (i = 0; i < UNCORE_TYPE_MAX; i++) {
uncore = &uncores[i];
@@ -1138,7 +706,7 @@ static int __init amd_uncore_init(void)
BUG_ON(!uncore->move);
BUG_ON(!uncore->free);
- uncore->info = alloc_percpu(union amd_uncore_info);
+ uncore->info = alloc_percpu(union uncore_common_info);
if (!uncore->info) {
ret = -ENOMEM;
goto fail;
@@ -1187,8 +755,8 @@ static int __init amd_uncore_init(void)
static void __exit amd_uncore_exit(void)
{
- struct amd_uncore *uncore;
- struct amd_uncore_pmu *pmu;
+ struct uncore_common *uncore;
+ struct uncore_common_pmu *pmu;
int i, j;
cpuhp_remove_state(CPUHP_AP_PERF_X86_AMD_UNCORE_ONLINE);
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCH 3/3] perf/x86/amd/uncore: Add Hygon uncore PMU support
2026-07-01 3:44 [PATCH 0/3] perf/x86/amd/uncore: Add common helpers and Hygon support Qi Liu
2026-07-01 3:44 ` [PATCH 1/3] perf/x86/amd/uncore: Add common PMU helper functions Qi Liu
2026-07-01 3:44 ` [PATCH 2/3] perf/x86/amd/uncore: Convert AMD driver to common PMU helpers Qi Liu
@ 2026-07-01 3:44 ` Qi Liu
2026-07-01 3:56 ` sashiko-bot
2 siblings, 1 reply; 11+ messages in thread
From: Qi Liu @ 2026-07-01 3:44 UTC (permalink / raw)
To: peterz, mingo, namhyung, alexander.shishkin, jolsa, x86
Cc: linux-perf-users, linux-kernel, lijing, liuqi, Zhenglang Hu
Add uncore PMU support for Hygon processors based on the
shared AMD-family uncore framework.
Hygon processors implement uncore PMUs for Data Fabric units
with a programming model similar to AMD, but with several
differences:
- Different MSR base addresses
- Vendor-specific event encoding and masks
- DF IOD counter allocation semantics
- Additional format attributes (e.g. iod, constid)
Reuse the common uncore infrastructure introduced in the
previous patch, this avoids duplicating the full uncore driver
while keeping vendor-specific logic isolated.
Signed-off-by: Qi Liu <liuqi@hygon.cn>
Tested-by: Zhenglang Hu <huzhenglang@hygon.cn>
---
arch/x86/events/Kconfig | 12 +
arch/x86/events/amd/Makefile | 2 +
arch/x86/events/amd/hygon_uncore.c | 551 +++++++++++++++++++++++++++++
arch/x86/include/asm/msr-index.h | 2 +
arch/x86/include/asm/perf_event.h | 20 ++
include/linux/cpuhotplug.h | 3 +
6 files changed, 590 insertions(+)
create mode 100644 arch/x86/events/amd/hygon_uncore.c
diff --git a/arch/x86/events/Kconfig b/arch/x86/events/Kconfig
index 466581535ae1..ebc1491be4fc 100644
--- a/arch/x86/events/Kconfig
+++ b/arch/x86/events/Kconfig
@@ -50,6 +50,18 @@ config PERF_EVENTS_AMD_UNCORE
To compile this driver as a module, choose M here: the
module will be called 'amd-uncore'.
+config PERF_EVENTS_HYGON_UNCORE
+ tristate "Hygon Uncore performance events"
+ depends on PERF_EVENTS && CPU_SUP_HYGON
+ select PERF_EVENTS_UNCORE_COMMON
+ default y
+ help
+ Include support for Hygon uncore performance events for use with
+ e.g., perf stat -e hygon_df/.../.
+
+ To compile this driver as a module, choose M here: the
+ module will be called 'hygon-uncore'.
+
config PERF_EVENTS_AMD_BRS
depends on PERF_EVENTS && CPU_SUP_AMD
bool "AMD Zen3 Branch Sampling support"
diff --git a/arch/x86/events/amd/Makefile b/arch/x86/events/amd/Makefile
index 0ddaa049b3b7..9d90b85e08d9 100644
--- a/arch/x86/events/amd/Makefile
+++ b/arch/x86/events/amd/Makefile
@@ -7,6 +7,8 @@ obj-$(CONFIG_PERF_EVENTS_UNCORE_COMMON) += uncore-common.o
uncore-common-objs := uncore_common.o
obj-$(CONFIG_PERF_EVENTS_AMD_UNCORE) += amd-uncore.o
amd-uncore-objs := uncore.o
+obj-$(CONFIG_PERF_EVENTS_HYGON_UNCORE) += hygon-uncore.o
+hygon-uncore-objs := hygon_uncore.o
ifdef CONFIG_AMD_IOMMU
obj-$(CONFIG_CPU_SUP_AMD) += iommu.o
endif
diff --git a/arch/x86/events/amd/hygon_uncore.c b/arch/x86/events/amd/hygon_uncore.c
new file mode 100644
index 000000000000..7758a1dda21a
--- /dev/null
+++ b/arch/x86/events/amd/hygon_uncore.c
@@ -0,0 +1,551 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2026 Chengdu Haiguang IC Design Co., Ltd.
+ */
+
+#include <linux/cpu.h>
+#include <linux/cpufeature.h>
+#include <linux/cpumask.h>
+#include <linux/init.h>
+#include <linux/percpu.h>
+#include <linux/perf_event.h>
+#include <linux/slab.h>
+#include <linux/smp.h>
+#include <linux/types.h>
+
+#include <asm/amd/nb.h>
+#include <asm/cpuid/api.h>
+#include <asm/msr.h>
+#include <asm/perf_event.h>
+
+#include "uncore_common.h"
+
+#define NUM_COUNTERS_DF 4
+
+#undef pr_fmt
+#define pr_fmt(fmt) "hygon_uncore: " fmt
+
+enum {
+ HYGON_UNCORE_TYPE_DF,
+ HYGON_UNCORE_TYPE_DF_IOD,
+ HYGON_UNCORE_TYPE_MAX,
+};
+
+/* Interval for hrtimer, defaults to 60000 milliseconds */
+static unsigned int update_interval = 60 * MSEC_PER_SEC;
+module_param(update_interval, uint, 0444);
+
+static struct uncore_common hygon_uncores[HYGON_UNCORE_TYPE_MAX];
+
+static __always_inline bool hygon_uncore_is_df_iod(struct uncore_common_pmu *pmu)
+{
+ struct uncore_common *uncore = pmu->private;
+
+ return uncore == &hygon_uncores[HYGON_UNCORE_TYPE_DF_IOD];
+}
+
+static __always_inline int hygon_uncore_num_iods(struct uncore_common *uncore, unsigned int cpu)
+{
+ union uncore_common_info *info = per_cpu_ptr(uncore->info, cpu);
+
+ return info->split.private;
+}
+
+static u64 hygon_uncore_df_event_mask(void)
+{
+ if (boot_cpu_data.x86_model == 0x4 ||
+ boot_cpu_data.x86_model == 0x5)
+ return HYGON_F18H_M4H_RAW_EVENT_MASK_DF;
+
+ if (boot_cpu_data.x86_model >= 0x6 &&
+ boot_cpu_data.x86_model <= 0x18)
+ return HYGON_F18H_M6H_RAW_EVENT_MASK_DF;
+
+ return HYGON_F18H_RAW_EVENT_MASK_DF;
+}
+
+DEFINE_UNCORE_FORMAT_ATTR(event, event, "config:0-5");
+DEFINE_UNCORE_FORMAT_ATTR(umask8, umask, "config:8-15");
+DEFINE_UNCORE_FORMAT_ATTR(umask10, umask, "config:8-17");
+DEFINE_UNCORE_FORMAT_ATTR(umask12, umask, "config:8-19");
+DEFINE_UNCORE_FORMAT_ATTR(constid, constid, "config:6-7,32-35,61-62");
+DEFINE_UNCORE_FORMAT_ATTR(iod, iod, "config1:0-1");
+
+static struct attribute *hygon_uncore_df_format_attr[] = {
+ &format_attr_event.attr,
+ &format_attr_umask8.attr,
+ &format_attr_constid.attr,
+ NULL,
+};
+
+static struct attribute *hygon_uncore_df_iod_format_attr[] = {
+ &format_attr_event.attr,
+ &format_attr_umask10.attr,
+ &format_attr_constid.attr,
+ &format_attr_iod.attr,
+ NULL,
+};
+
+static struct attribute_group hygon_uncore_df_format_group = {
+ .name = "format",
+ .attrs = hygon_uncore_df_format_attr,
+};
+
+static struct attribute_group hygon_uncore_df_iod_format_group = {
+ .name = "format",
+ .attrs = hygon_uncore_df_iod_format_attr,
+};
+
+static const struct attribute_group *hygon_uncore_df_attr_groups[] = {
+ &uncore_common_attr_group,
+ &hygon_uncore_df_format_group,
+ NULL,
+};
+
+static const struct attribute_group *hygon_uncore_df_iod_attr_groups[] = {
+ &uncore_common_attr_group,
+ &hygon_uncore_df_iod_format_group,
+ NULL,
+};
+
+static int hygon_uncore_df_event_init(struct perf_event *event)
+{
+ struct hw_perf_event *hwc = &event->hw;
+ struct uncore_common_pmu *pmu;
+ struct uncore_common *uncore;
+ u64 event_mask;
+ int ret;
+
+ ret = uncore_common_event_init(event);
+ if (ret)
+ return ret;
+
+ pmu = event_to_uncore_common_pmu(event);
+ uncore = pmu->private;
+
+ if (hygon_uncore_is_df_iod(pmu) &&
+ event->attr.config1 >= hygon_uncore_num_iods(uncore, event->cpu))
+ return -EINVAL;
+
+ event_mask = hygon_uncore_df_event_mask();
+ hwc->config = event->attr.config & event_mask;
+
+ return 0;
+}
+
+static int hygon_uncore_df_iod_add(struct perf_event *event, int flags)
+{
+ struct uncore_common_pmu *pmu = event_to_uncore_common_pmu(event);
+ struct uncore_common_ctx *ctx = *per_cpu_ptr(pmu->ctx, event->cpu);
+ struct hw_perf_event *hwc = &event->hw;
+ int iod_idx;
+ int i;
+
+ if (hwc->idx != -1 && ctx->events[hwc->idx] == event)
+ goto out;
+
+ for (i = 0; i < pmu->num_counters; i++) {
+ if (ctx->events[i] == event) {
+ hwc->idx = i;
+ goto out;
+ }
+ }
+
+ hwc->idx = -1;
+ iod_idx = event->attr.config1;
+
+ if (iod_idx >= pmu->num_counters / NUM_COUNTERS_DF)
+ return -EINVAL;
+
+ for (i = iod_idx * NUM_COUNTERS_DF; i < (iod_idx + 1) * NUM_COUNTERS_DF; i++) {
+ struct perf_event *tmp = NULL;
+
+ if (try_cmpxchg(&ctx->events[i], &tmp, event)) {
+ hwc->idx = i;
+ break;
+ }
+ }
+
+out:
+ if (hwc->idx == -1)
+ return -EBUSY;
+
+ hwc->config_base = pmu->msr_base + 2 * hwc->idx;
+ hwc->event_base = pmu->msr_base + 1 + 2 * hwc->idx;
+ hwc->event_base_rdpmc = -1;
+ hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED;
+
+ if (flags & PERF_EF_START)
+ event->pmu->start(event, PERF_EF_RELOAD);
+
+ return 0;
+}
+
+static int hygon_uncore_cpu_starting(unsigned int cpu)
+{
+ struct uncore_common *uncore;
+ int i;
+
+ for (i = 0; i < HYGON_UNCORE_TYPE_MAX; i++) {
+ uncore = &hygon_uncores[i];
+ uncore->scan(uncore, cpu);
+ }
+
+ return 0;
+}
+
+static int hygon_uncore_cpu_online(unsigned int cpu)
+{
+ struct uncore_common *uncore;
+ int ret;
+ int i;
+
+ for (i = 0; i < HYGON_UNCORE_TYPE_MAX; i++) {
+ uncore = &hygon_uncores[i];
+
+ ret = uncore->init(uncore, cpu);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+static int hygon_uncore_cpu_down_prepare(unsigned int cpu)
+{
+ struct uncore_common *uncore;
+ int i;
+
+ for (i = 0; i < HYGON_UNCORE_TYPE_MAX; i++) {
+ uncore = &hygon_uncores[i];
+ uncore->move(uncore, cpu);
+ }
+
+ return 0;
+}
+
+static int hygon_uncore_cpu_dead(unsigned int cpu)
+{
+ struct uncore_common *uncore;
+ int i;
+
+ for (i = 0; i < HYGON_UNCORE_TYPE_MAX; i++) {
+ uncore = &hygon_uncores[i];
+ uncore->free(uncore, cpu);
+ }
+
+ return 0;
+}
+
+static int hygon_uncore_df_ctx_init(struct uncore_common *uncore,
+ unsigned int cpu)
+{
+ struct attribute *df_attr;
+ struct uncore_common_pmu *pmu;
+ int num_counters;
+
+ if (uncore->init_done)
+ return uncore_common_ctx_init(uncore, cpu);
+
+ num_counters = uncore_common_ctx_num_pmcs(uncore, cpu);
+ if (!num_counters)
+ goto done;
+
+ uncore->pmus = kzalloc_obj(*uncore->pmus);
+ if (!uncore->pmus)
+ goto done;
+
+ pmu = &uncore->pmus[0];
+ strscpy(pmu->name, "hygon_df", sizeof(pmu->name));
+ pmu->num_counters = num_counters;
+ pmu->msr_base = MSR_HYGON_F18H_DF_CTL;
+ pmu->rdpmc_base = -1;
+ pmu->group = uncore_common_ctx_gid(uncore, cpu);
+ pmu->private = uncore;
+
+ df_attr = &format_attr_umask8.attr;
+ if (boot_cpu_data.x86_model == 0x4 ||
+ boot_cpu_data.x86_model == 0x5)
+ df_attr = &format_attr_umask10.attr;
+ else if (boot_cpu_data.x86_model >= 0x6 &&
+ boot_cpu_data.x86_model <= 0x18)
+ df_attr = &format_attr_umask12.attr;
+ hygon_uncore_df_format_attr[1] = df_attr;
+
+ pmu->ctx = alloc_percpu(struct uncore_common_ctx *);
+ if (!pmu->ctx)
+ goto done;
+
+ pmu->pmu = (struct pmu) {
+ .task_ctx_nr = perf_invalid_context,
+ .attr_groups = hygon_uncore_df_attr_groups,
+ .name = pmu->name,
+ .event_init = hygon_uncore_df_event_init,
+ .add = uncore_common_add,
+ .del = uncore_common_del,
+ .start = uncore_common_start,
+ .stop = uncore_common_stop,
+ .read = uncore_common_read,
+ .capabilities = PERF_PMU_CAP_NO_EXCLUDE | PERF_PMU_CAP_NO_INTERRUPT,
+ .module = THIS_MODULE,
+ };
+
+ if (perf_pmu_register(&pmu->pmu, pmu->pmu.name, -1)) {
+ free_percpu(pmu->ctx);
+ pmu->ctx = NULL;
+ goto done;
+ }
+
+ pr_info("%d %s counters detected\n", pmu->num_counters, pmu->pmu.name);
+ uncore->num_pmus = 1;
+
+done:
+ uncore->init_done = true;
+ return uncore_common_ctx_init(uncore, cpu);
+}
+
+static void hygon_uncore_df_ctx_scan(struct uncore_common *uncore,
+ unsigned int cpu)
+{
+ unsigned int eax, ebx, ecx, edx;
+ union uncore_common_info info = {};
+
+ if (!boot_cpu_has(X86_FEATURE_PERFCTR_NB))
+ return;
+
+ info.split.gid = 0;
+ info.split.aux_data = 0;
+ info.split.num_pmcs = NUM_COUNTERS_DF;
+
+ cpuid(0x8000001e, &eax, &ebx, &ecx, &edx);
+ info.split.cid = ecx & 0xff;
+
+ *per_cpu_ptr(uncore->info, cpu) = info;
+}
+
+static void hygon_uncore_df_iod_ctx_scan(struct uncore_common *uncore,
+ unsigned int cpu)
+{
+ int num_packages, iods_per_package;
+ union uncore_common_info info = {};
+
+ if (!boot_cpu_has(X86_FEATURE_PERFCTR_NB))
+ return;
+
+ if (boot_cpu_data.x86_model < 0x4 || boot_cpu_data.x86_model == 0x6)
+ return;
+
+ num_packages = topology_max_packages();
+ iods_per_package = amd_nb_num() / num_packages - topology_max_dies_per_package();
+ /* Hardware does not support more than 2 IODs per package. */
+ if (iods_per_package <= 0 || iods_per_package > 2)
+ return;
+
+ info.split.cid = topology_physical_package_id(cpu);
+ info.split.gid = 0;
+ info.split.private = iods_per_package;
+ info.split.num_pmcs = NUM_COUNTERS_DF * iods_per_package;
+
+ *per_cpu_ptr(uncore->info, cpu) = info;
+}
+
+static int hygon_uncore_df_iod_ctx_init(struct uncore_common *uncore,
+ unsigned int cpu)
+{
+ struct uncore_common_pmu *pmu;
+ int num_counters;
+
+ if (uncore->init_done)
+ return uncore_common_ctx_init(uncore, cpu);
+
+ num_counters = uncore_common_ctx_num_pmcs(uncore, cpu);
+ if (!num_counters)
+ goto done;
+
+ uncore->pmus = kzalloc_obj(*uncore->pmus);
+ if (!uncore->pmus)
+ goto done;
+
+ pmu = &uncore->pmus[0];
+ strscpy(pmu->name, "hygon_df_iod", sizeof(pmu->name));
+ pmu->num_counters = num_counters;
+ pmu->msr_base = MSR_HYGON_F18H_DF_IOD_CTL;
+ pmu->rdpmc_base = -1;
+ pmu->group = uncore_common_ctx_gid(uncore, cpu);
+ pmu->private = uncore;
+
+ if (boot_cpu_data.x86_model >= 0x6 &&
+ boot_cpu_data.x86_model <= 0x18)
+ hygon_uncore_df_iod_format_attr[1] = &format_attr_umask12.attr;
+
+ pmu->ctx = alloc_percpu(struct uncore_common_ctx *);
+ if (!pmu->ctx)
+ goto done;
+
+ pmu->pmu = (struct pmu) {
+ .task_ctx_nr = perf_invalid_context,
+ .attr_groups = hygon_uncore_df_iod_attr_groups,
+ .name = pmu->name,
+ .event_init = hygon_uncore_df_event_init,
+ .add = hygon_uncore_df_iod_add,
+ .del = uncore_common_del,
+ .start = uncore_common_start,
+ .stop = uncore_common_stop,
+ .read = uncore_common_read,
+ .capabilities = PERF_PMU_CAP_NO_EXCLUDE | PERF_PMU_CAP_NO_INTERRUPT,
+ .module = THIS_MODULE,
+ };
+
+ if (perf_pmu_register(&pmu->pmu, pmu->pmu.name, -1)) {
+ free_percpu(pmu->ctx);
+ pmu->ctx = NULL;
+ goto done;
+ }
+
+ pr_info("%d %s counters detected\n", pmu->num_counters, pmu->pmu.name);
+ uncore->num_pmus = 1;
+
+done:
+ uncore->init_done = true;
+ return uncore_common_ctx_init(uncore, cpu);
+}
+
+static struct uncore_common hygon_uncores[HYGON_UNCORE_TYPE_MAX] = {
+ /* HYGON_UNCORE_TYPE_DF */
+ {
+ .scan = hygon_uncore_df_ctx_scan,
+ .init = hygon_uncore_df_ctx_init,
+ .move = uncore_common_ctx_move,
+ .free = uncore_common_ctx_free,
+ },
+ /* HYGON_UNCORE_TYPE_DF IOD */
+ {
+ .scan = hygon_uncore_df_iod_ctx_scan,
+ .init = hygon_uncore_df_iod_ctx_init,
+ .move = uncore_common_ctx_move,
+ .free = uncore_common_ctx_free,
+ },
+};
+
+static void hygon_uncore_pmus_unregister(void)
+{
+ struct uncore_common *uncore;
+ struct uncore_common_pmu *pmu;
+ int i, j;
+
+ for (i = 0; i < HYGON_UNCORE_TYPE_MAX; i++) {
+ uncore = &hygon_uncores[i];
+
+ for (j = 0; j < uncore->num_pmus; j++) {
+ pmu = &uncore->pmus[j];
+ if (!pmu->ctx)
+ continue;
+
+ perf_pmu_unregister(&pmu->pmu);
+ free_percpu(pmu->ctx);
+ pmu->ctx = NULL;
+ }
+
+ kfree(uncore->pmus);
+ uncore->pmus = NULL;
+ uncore->num_pmus = 0;
+ uncore->init_done = false;
+ }
+}
+
+static void hygon_uncore_infos_free(void)
+{
+ struct uncore_common *uncore;
+ int i;
+
+ for (i = 0; i < HYGON_UNCORE_TYPE_MAX; i++) {
+ uncore = &hygon_uncores[i];
+ if (!uncore->info)
+ continue;
+
+ free_percpu(uncore->info);
+ uncore->info = NULL;
+ }
+}
+
+static int __init hygon_uncore_init(void)
+{
+ struct uncore_common *uncore;
+ int ret = -ENODEV;
+ int i;
+
+ if (boot_cpu_data.x86_vendor != X86_VENDOR_HYGON)
+ return -ENODEV;
+
+ if (!boot_cpu_has(X86_FEATURE_TOPOEXT))
+ return -ENODEV;
+
+ uncore_common_set_update_interval(update_interval);
+
+ for (i = 0; i < HYGON_UNCORE_TYPE_MAX; i++) {
+ uncore = &hygon_uncores[i];
+
+ if (WARN_ON_ONCE(!uncore->scan ||
+ !uncore->init ||
+ !uncore->move ||
+ !uncore->free)) {
+ ret = -EINVAL;
+ goto fail;
+ }
+
+ uncore->info = alloc_percpu(union uncore_common_info);
+ if (!uncore->info) {
+ ret = -ENOMEM;
+ goto fail;
+ }
+ }
+
+ ret = cpuhp_setup_state(CPUHP_PERF_X86_HYGON_UNCORE_PREP,
+ "perf/x86/hygon/uncore:prepare",
+ NULL, hygon_uncore_cpu_dead);
+ if (ret)
+ goto fail;
+
+ ret = cpuhp_setup_state(CPUHP_AP_PERF_X86_HYGON_UNCORE_STARTING,
+ "perf/x86/hygon/uncore:starting",
+ hygon_uncore_cpu_starting, NULL);
+ if (ret)
+ goto fail_prep;
+
+ ret = cpuhp_setup_state(CPUHP_AP_PERF_X86_HYGON_UNCORE_ONLINE,
+ "perf/x86/hygon/uncore:online",
+ hygon_uncore_cpu_online,
+ hygon_uncore_cpu_down_prepare);
+ if (ret)
+ goto fail_start;
+
+ return 0;
+
+fail_start:
+ cpuhp_remove_state(CPUHP_AP_PERF_X86_HYGON_UNCORE_STARTING);
+ cpuhp_remove_state(CPUHP_PERF_X86_HYGON_UNCORE_PREP);
+ hygon_uncore_pmus_unregister();
+ goto fail;
+fail_prep:
+ cpuhp_remove_state(CPUHP_PERF_X86_HYGON_UNCORE_PREP);
+fail:
+ hygon_uncore_infos_free();
+
+ return ret;
+}
+
+static void __exit hygon_uncore_exit(void)
+{
+ cpuhp_remove_state(CPUHP_AP_PERF_X86_HYGON_UNCORE_ONLINE);
+ cpuhp_remove_state(CPUHP_AP_PERF_X86_HYGON_UNCORE_STARTING);
+ cpuhp_remove_state(CPUHP_PERF_X86_HYGON_UNCORE_PREP);
+
+ hygon_uncore_pmus_unregister();
+ hygon_uncore_infos_free();
+}
+
+module_init(hygon_uncore_init);
+module_exit(hygon_uncore_exit);
+
+MODULE_DESCRIPTION("Hygon Uncore Driver");
+MODULE_LICENSE("GPL");
diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 18c4be75e927..d20338f0b2b2 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -851,6 +851,8 @@
#define MSR_F15H_PTSC 0xc0010280
#define MSR_F15H_IC_CFG 0xc0011021
#define MSR_F15H_EX_CFG 0xc001102c
+#define MSR_HYGON_F18H_DF_CTL 0xc0010240
+#define MSR_HYGON_F18H_DF_IOD_CTL 0xc0010250
/* Fam 10h MSRs */
#define MSR_FAM10H_MMIO_CONF_BASE 0xc0010058
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 1eb13673e889..7c8c00ad56e9 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -124,6 +124,26 @@
(AMD64_PERFMON_V2_EVENTSEL_EVENT_NB | \
AMD64_PERFMON_V2_EVENTSEL_UMASK_NB)
+#define HYGON_F18H_EVENTSEL_EVENT_DF 0x0000003FULL
+#define HYGON_F18H_EVENTSEL_UMASK_DF 0x0000FF00ULL
+#define HYGON_F18H_M4H_EVENTSEL_UMASK_DF 0x0003FF00ULL
+#define HYGON_F18H_M6H_EVENTSEL_UMASK_DF 0x000FFF00ULL
+#define HYGON_F18H_EVENTSEL_CONSTID_DF \
+ (0x000000C0ULL | GENMASK_ULL(35, 32) | GENMASK_ULL(62, 61))
+
+#define HYGON_F18H_RAW_EVENT_MASK_DF \
+ (HYGON_F18H_EVENTSEL_EVENT_DF | \
+ HYGON_F18H_EVENTSEL_UMASK_DF | \
+ HYGON_F18H_EVENTSEL_CONSTID_DF)
+#define HYGON_F18H_M4H_RAW_EVENT_MASK_DF \
+ (HYGON_F18H_EVENTSEL_EVENT_DF | \
+ HYGON_F18H_M4H_EVENTSEL_UMASK_DF | \
+ HYGON_F18H_EVENTSEL_CONSTID_DF)
+#define HYGON_F18H_M6H_RAW_EVENT_MASK_DF \
+ (HYGON_F18H_EVENTSEL_EVENT_DF | \
+ HYGON_F18H_M6H_EVENTSEL_UMASK_DF | \
+ HYGON_F18H_EVENTSEL_CONSTID_DF)
+
#define AMD64_PERFMON_V2_ENABLE_UMC BIT_ULL(31)
#define AMD64_PERFMON_V2_EVENTSEL_EVENT_UMC GENMASK_ULL(7, 0)
#define AMD64_PERFMON_V2_EVENTSEL_RDWRMASK_UMC GENMASK_ULL(9, 8)
diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index 0fb3a2a62eb0..bccdb87a0c05 100644
--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -62,6 +62,7 @@ enum cpuhp_state {
CPUHP_CREATE_THREADS,
CPUHP_PERF_X86_PREPARE,
CPUHP_PERF_X86_AMD_UNCORE_PREP,
+ CPUHP_PERF_X86_HYGON_UNCORE_PREP,
CPUHP_PERF_POWER,
CPUHP_PERF_SUPERH,
CPUHP_X86_HPET_DEAD,
@@ -148,6 +149,7 @@ enum cpuhp_state {
CPUHP_AP_IRQ_RISCV_SBI_IPI_STARTING,
CPUHP_AP_ARM_MVEBU_COHERENCY,
CPUHP_AP_PERF_X86_AMD_UNCORE_STARTING,
+ CPUHP_AP_PERF_X86_HYGON_UNCORE_STARTING,
CPUHP_AP_PERF_X86_STARTING,
CPUHP_AP_PERF_X86_AMD_IBS_STARTING,
CPUHP_AP_PERF_XTENSA_STARTING,
@@ -205,6 +207,7 @@ enum cpuhp_state {
CPUHP_AP_PERF_X86_ONLINE,
CPUHP_AP_PERF_X86_UNCORE_ONLINE,
CPUHP_AP_PERF_X86_AMD_UNCORE_ONLINE,
+ CPUHP_AP_PERF_X86_HYGON_UNCORE_ONLINE,
CPUHP_AP_PERF_X86_AMD_POWER_ONLINE,
CPUHP_AP_PERF_S390_CF_ONLINE,
CPUHP_AP_PERF_S390_SF_ONLINE,
--
2.34.1
^ permalink raw reply related [flat|nested] 11+ messages in thread