diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c index 289fb08acaf50e5455e6171f7ac40ef9948e86e1..4eb1a76731b2251c9bb8a67b0cbbf8098c4cece1 100644 --- a/drivers/gpu/drm/i915/i915_gem_request.c +++ b/drivers/gpu/drm/i915/i915_gem_request.c @@ -1021,12 +1021,28 @@ static bool busywait_stop(unsigned long timeout, unsigned int cpu) return this_cpu != cpu; } -bool __i915_spin_request(const struct drm_i915_gem_request *req, - u32 seqno, int state, unsigned long timeout_us) +static bool __i915_spin_request(const struct drm_i915_gem_request *req, + u32 seqno, int state, unsigned long timeout_us) { struct intel_engine_cs *engine = req->engine; unsigned int irq, cpu; + GEM_BUG_ON(!seqno); + + /* + * Only wait for the request if we know it is likely to complete. + * + * We don't track the timestamps around requests, nor the average + * request length, so we do not have a good indicator that this + * request will complete within the timeout. What we do know is the + * order in which requests are executed by the engine and so we can + * tell if the request has started. If the request hasn't started yet, + * it is a fair assumption that it will not complete within our + * relatively short timeout. + */ + if (!i915_seqno_passed(intel_engine_get_seqno(engine), seqno - 1)) + return false; + /* When waiting for high frequency requests, e.g. during synchronous * rendering split between the CPU and GPU, the finite amount of time * required to set up the irq and wait upon it limits the response @@ -1040,8 +1056,7 @@ bool __i915_spin_request(const struct drm_i915_gem_request *req, irq = atomic_read(&engine->irq_count); timeout_us += local_clock_us(&cpu); do { - if (i915_seqno_passed(intel_engine_get_seqno(req->engine), - seqno)) + if (i915_seqno_passed(intel_engine_get_seqno(engine), seqno)) return seqno == i915_gem_request_global_seqno(req); /* Seqno are meant to be ordered *before* the interrupt. If @@ -1153,7 +1168,7 @@ long i915_wait_request(struct drm_i915_gem_request *req, GEM_BUG_ON(!i915_sw_fence_signaled(&req->submit)); /* Optimistic short spin before touching IRQs */ - if (i915_spin_request(req, state, 5)) + if (__i915_spin_request(req, wait.seqno, state, 5)) goto complete; set_current_state(state); @@ -1210,7 +1225,7 @@ long i915_wait_request(struct drm_i915_gem_request *req, continue; /* Only spin if we know the GPU is processing this request */ - if (i915_spin_request(req, state, 2)) + if (__i915_spin_request(req, wait.seqno, state, 2)) break; if (!intel_wait_check_request(&wait, req)) { diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h index 49a4c8994ff07afe721134bd26d50a0a88261a31..96eb52471dad4f2af0ba2dfa15863d8861f66896 100644 --- a/drivers/gpu/drm/i915/i915_gem_request.h +++ b/drivers/gpu/drm/i915/i915_gem_request.h @@ -312,26 +312,6 @@ static inline bool i915_seqno_passed(u32 seq1, u32 seq2) return (s32)(seq1 - seq2) >= 0; } -static inline bool -__i915_gem_request_started(const struct drm_i915_gem_request *req, u32 seqno) -{ - GEM_BUG_ON(!seqno); - return i915_seqno_passed(intel_engine_get_seqno(req->engine), - seqno - 1); -} - -static inline bool -i915_gem_request_started(const struct drm_i915_gem_request *req) -{ - u32 seqno; - - seqno = i915_gem_request_global_seqno(req); - if (!seqno) - return false; - - return __i915_gem_request_started(req, seqno); -} - static inline bool __i915_gem_request_completed(const struct drm_i915_gem_request *req, u32 seqno) { @@ -352,21 +332,6 @@ i915_gem_request_completed(const struct drm_i915_gem_request *req) return __i915_gem_request_completed(req, seqno); } -bool __i915_spin_request(const struct drm_i915_gem_request *request, - u32 seqno, int state, unsigned long timeout_us); -static inline bool i915_spin_request(const struct drm_i915_gem_request *request, - int state, unsigned long timeout_us) -{ - u32 seqno; - - seqno = i915_gem_request_global_seqno(request); - if (!seqno) - return 0; - - return (__i915_gem_request_started(request, seqno) && - __i915_spin_request(request, seqno, state, timeout_us)); -} - /* We treat requests as fences. This is not be to confused with our * "fence registers" but pipeline synchronisation objects ala GL_ARB_sync. * We use the fences to synchronize access from the CPU with activity on the