[PATCH 2/2] drm/i915/pmu: Handle CPU hotplug more efficiently

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin at intel.com>

Dynamic CPU hotplug notification slots are a scarce resource and given
how we already register the multi instance type state, we can and should
add multiple instance of the i915 PMU to this same state and not allocate
a new one every time.

At the same time this also fixes the event migration when multiple GPUs
are present by making sure events are migrated for not just the first
callback against i915 cpu hotplug state.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
Suggested-by: Daniel Vetter <daniel.vetter at intel.com>
---
 drivers/gpu/drm/i915/i915_pci.c |  5 ++++
 drivers/gpu/drm/i915/i915_pmu.c | 43 +++++++++++++++++++--------------
 drivers/gpu/drm/i915/i915_pmu.h |  9 ++++++-
 3 files changed, 38 insertions(+), 19 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_pci.c b/drivers/gpu/drm/i915/i915_pci.c
index 366ddfc8df6b..58271ec7143b 100644
--- a/drivers/gpu/drm/i915/i915_pci.c
+++ b/drivers/gpu/drm/i915/i915_pci.c
@@ -1129,6 +1129,10 @@ static int __init i915_init(void)
 		return 0;
 	}
 
+	err = i915_pmu_init();
+	if (err)
+		return err;
+
 	err = pci_register_driver(&i915_pci_driver);
 	if (err)
 		return err;
@@ -1145,6 +1149,7 @@ static void __exit i915_exit(void)
 	i915_perf_sysctl_unregister();
 	pci_unregister_driver(&i915_pci_driver);
 	i915_globals_exit();
+	i915_pmu_exit();
 }
 
 module_init(i915_init);
diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c
index 5d1b38e66afb..5b1469154428 100644
--- a/drivers/gpu/drm/i915/i915_pmu.c
+++ b/drivers/gpu/drm/i915/i915_pmu.c
@@ -30,6 +30,7 @@
 #define ENGINE_SAMPLE_BITS (1 << I915_PMU_SAMPLE_BITS)
 
 static cpumask_t i915_pmu_cpumask;
+static unsigned int i915_pmu_target_cpu;
 
 static u8 engine_config_sample(u64 config)
 {
@@ -1050,25 +1051,32 @@ static int i915_pmu_cpu_online(unsigned int cpu, struct hlist_node *node)
 static int i915_pmu_cpu_offline(unsigned int cpu, struct hlist_node *node)
 {
 	struct i915_pmu *pmu = hlist_entry_safe(node, typeof(*pmu), cpuhp.node);
-	unsigned int target;
+	unsigned int target = -1;
 
 	GEM_BUG_ON(!pmu->base.event_init);
 
 	if (cpumask_test_and_clear_cpu(cpu, &i915_pmu_cpumask)) {
 		target = cpumask_any_but(topology_sibling_cpumask(cpu), cpu);
+
 		/* Migrate events if there is a valid target */
 		if (target < nr_cpu_ids) {
 			cpumask_set_cpu(target, &i915_pmu_cpumask);
-			perf_pmu_migrate_context(&pmu->base, cpu, target);
+			i915_pmu_target_cpu = target;
 		}
 	}
 
+	if (target < nr_cpu_ids && target != pmu->cpuhp.cpu) {
+		perf_pmu_migrate_context(&pmu->base, cpu, target);
+		pmu->cpuhp.cpu = target;
+	}
+
 	return 0;
 }
 
-static int i915_pmu_register_cpuhp_state(struct i915_pmu *pmu)
+static enum cpuhp_state cpuhp_slot = CPUHP_INVALID;
+
+int i915_pmu_init(void)
 {
-	enum cpuhp_state slot;
 	int ret;
 
 	ret = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN,
@@ -1078,25 +1086,24 @@ static int i915_pmu_register_cpuhp_state(struct i915_pmu *pmu)
 	if (ret < 0)
 		return ret;
 
-	slot = ret;
-	ret = cpuhp_state_add_instance(slot, &pmu->cpuhp.node);
-	if (ret) {
-		cpuhp_remove_multi_state(slot);
-		return ret;
-	}
+	cpuhp_slot = ret;
 
-	pmu->cpuhp.slot = slot;
 	return 0;
 }
 
-static void i915_pmu_unregister_cpuhp_state(struct i915_pmu *pmu)
+void i915_pmu_exit(void)
 {
-	struct drm_i915_private *i915 = container_of(pmu, typeof(*i915), pmu);
+	cpuhp_remove_multi_state(cpuhp_slot);
+}
+
+static int i915_pmu_register_cpuhp_state(struct i915_pmu *pmu)
+{
+	return cpuhp_state_add_instance(cpuhp_slot, &pmu->cpuhp.node);
+}
 
-	drm_WARN_ON(&i915->drm, pmu->cpuhp.slot == CPUHP_INVALID);
-	drm_WARN_ON(&i915->drm, cpuhp_state_remove_instance(pmu->cpuhp.slot, &pmu->cpuhp.node));
-	cpuhp_remove_multi_state(pmu->cpuhp.slot);
-	pmu->cpuhp.slot = CPUHP_INVALID;
+static void i915_pmu_unregister_cpuhp_state(struct i915_pmu *pmu)
+{
+	cpuhp_state_remove_instance(cpuhp_slot, &pmu->cpuhp.node);
 }
 
 static bool is_igp(struct drm_i915_private *i915)
@@ -1130,7 +1137,7 @@ void i915_pmu_register(struct drm_i915_private *i915)
 	spin_lock_init(&pmu->lock);
 	hrtimer_init(&pmu->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
 	pmu->timer.function = i915_sample;
-	pmu->cpuhp.slot = CPUHP_INVALID;
+	pmu->cpuhp.cpu = -1;
 
 	if (!is_igp(i915)) {
 		pmu->name = kasprintf(GFP_KERNEL,
diff --git a/drivers/gpu/drm/i915/i915_pmu.h b/drivers/gpu/drm/i915/i915_pmu.h
index 59a0d19afb67..a45039b09f0d 100644
--- a/drivers/gpu/drm/i915/i915_pmu.h
+++ b/drivers/gpu/drm/i915/i915_pmu.h
@@ -43,8 +43,9 @@ struct i915_pmu {
 	 */
 	struct {
 		struct hlist_node node;
-		enum cpuhp_state slot;
+		unsigned int cpu;
 	} cpuhp;
+
 	/**
 	 * @base: PMU base.
 	 */
@@ -126,11 +127,17 @@ struct i915_pmu {
 };
 
 #ifdef CONFIG_PERF_EVENTS
+int i915_pmu_init(void);
+void i915_pmu_exit(void);
+
 void i915_pmu_register(struct drm_i915_private *i915);
 void i915_pmu_unregister(struct drm_i915_private *i915);
 void i915_pmu_gt_parked(struct drm_i915_private *i915);
 void i915_pmu_gt_unparked(struct drm_i915_private *i915);
 #else
+static inline int i915_pmu_init(void) { return 0; }
+static inline void i915_pmu_exit(void) {}
+
 static inline void i915_pmu_register(struct drm_i915_private *i915) {}
 static inline void i915_pmu_unregister(struct drm_i915_private *i915) {}
 static inline void i915_pmu_gt_parked(struct drm_i915_private *i915) {}
-- 
2.25.1