diff options
Diffstat (limited to 'drivers/gpu/drm/i915/i915_pmu.c')
| -rw-r--r-- | drivers/gpu/drm/i915/i915_pmu.c | 106 |
1 files changed, 56 insertions, 50 deletions
diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index e991a707bdb7..056994224c6b 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -269,12 +269,48 @@ static bool exclusive_mmio_access(const struct drm_i915_private *i915) return IS_GEN(i915, 7); } +static void engine_sample(struct intel_engine_cs *engine, unsigned int period_ns) +{ + struct intel_engine_pmu *pmu = &engine->pmu; + bool busy; + u32 val; + + val = ENGINE_READ_FW(engine, RING_CTL); + if (val == 0) /* powerwell off => engine idle */ + return; + + if (val & RING_WAIT) + add_sample(&pmu->sample[I915_SAMPLE_WAIT], period_ns); + if (val & RING_WAIT_SEMAPHORE) + add_sample(&pmu->sample[I915_SAMPLE_SEMA], period_ns); + + /* No need to sample when busy stats are supported. */ + if (intel_engine_supports_stats(engine)) + return; + + /* + * While waiting on a semaphore or event, MI_MODE reports the + * ring as idle. However, previously using the seqno, and with + * execlists sampling, we account for the ring waiting as the + * engine being busy. Therefore, we record the sample as being + * busy if either waiting or !idle. + */ + busy = val & (RING_WAIT_SEMAPHORE | RING_WAIT); + if (!busy) { + val = ENGINE_READ_FW(engine, RING_MI_MODE); + busy = !(val & MODE_IDLE); + } + if (busy) + add_sample(&pmu->sample[I915_SAMPLE_BUSY], period_ns); +} + static void engines_sample(struct intel_gt *gt, unsigned int period_ns) { struct drm_i915_private *i915 = gt->i915; struct intel_engine_cs *engine; enum intel_engine_id id; + unsigned long flags; if ((i915->pmu.enable & ENGINE_SAMPLE_MASK) == 0) return; @@ -283,53 +319,17 @@ engines_sample(struct intel_gt *gt, unsigned int period_ns) return; for_each_engine(engine, gt, id) { - struct intel_engine_pmu *pmu = &engine->pmu; - spinlock_t *mmio_lock; - unsigned long flags; - bool busy; - u32 val; - if (!intel_engine_pm_get_if_awake(engine)) continue; - mmio_lock = NULL; - if (exclusive_mmio_access(i915)) - mmio_lock = &engine->uncore->lock; - - if (unlikely(mmio_lock)) - spin_lock_irqsave(mmio_lock, flags); - - val = ENGINE_READ_FW(engine, RING_CTL); - if (val == 0) /* powerwell off => engine idle */ - goto skip; - - if (val & RING_WAIT) - add_sample(&pmu->sample[I915_SAMPLE_WAIT], period_ns); - if (val & RING_WAIT_SEMAPHORE) - add_sample(&pmu->sample[I915_SAMPLE_SEMA], period_ns); - - /* No need to sample when busy stats are supported. */ - if (intel_engine_supports_stats(engine)) - goto skip; - - /* - * While waiting on a semaphore or event, MI_MODE reports the - * ring as idle. However, previously using the seqno, and with - * execlists sampling, we account for the ring waiting as the - * engine being busy. Therefore, we record the sample as being - * busy if either waiting or !idle. - */ - busy = val & (RING_WAIT_SEMAPHORE | RING_WAIT); - if (!busy) { - val = ENGINE_READ_FW(engine, RING_MI_MODE); - busy = !(val & MODE_IDLE); + if (exclusive_mmio_access(i915)) { + spin_lock_irqsave(&engine->uncore->lock, flags); + engine_sample(engine, period_ns); + spin_unlock_irqrestore(&engine->uncore->lock, flags); + } else { + engine_sample(engine, period_ns); } - if (busy) - add_sample(&pmu->sample[I915_SAMPLE_BUSY], period_ns); -skip: - if (unlikely(mmio_lock)) - spin_unlock_irqrestore(mmio_lock, flags); intel_engine_pm_put_async(engine); } } @@ -441,8 +441,10 @@ static u64 count_interrupts(struct drm_i915_private *i915) static void i915_pmu_event_destroy(struct perf_event *event) { - WARN_ON(event->parent); - module_put(THIS_MODULE); + struct drm_i915_private *i915 = + container_of(event->pmu, typeof(*i915), pmu.base); + + drm_WARN_ON(&i915->drm, event->parent); } static int @@ -534,10 +536,8 @@ static int i915_pmu_event_init(struct perf_event *event) if (ret) return ret; - if (!event->parent) { - __module_get(THIS_MODULE); + if (!event->parent) event->destroy = i915_pmu_event_destroy; - } return 0; } @@ -561,7 +561,10 @@ static u64 __i915_pmu_event_read(struct perf_event *event) /* Do nothing */ } else if (sample == I915_SAMPLE_BUSY && intel_engine_supports_stats(engine)) { - val = ktime_to_ns(intel_engine_get_busy_time(engine)); + ktime_t unused; + + val = ktime_to_ns(intel_engine_get_busy_time(engine, + &unused)); } else { val = engine->pmu.sample[sample].cur; } @@ -1058,8 +1061,10 @@ static int i915_pmu_register_cpuhp_state(struct i915_pmu *pmu) static void i915_pmu_unregister_cpuhp_state(struct i915_pmu *pmu) { - WARN_ON(pmu->cpuhp.slot == CPUHP_INVALID); - WARN_ON(cpuhp_state_remove_instance(pmu->cpuhp.slot, &pmu->cpuhp.node)); + struct drm_i915_private *i915 = container_of(pmu, typeof(*i915), pmu); + + 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; } @@ -1121,6 +1126,7 @@ void i915_pmu_register(struct drm_i915_private *i915) if (!pmu->base.attr_groups) goto err_attr; + pmu->base.module = THIS_MODULE; pmu->base.task_ctx_nr = perf_invalid_context; pmu->base.event_init = i915_pmu_event_init; pmu->base.add = i915_pmu_event_add; |