From d63763749127b483ba042b2939ba39fd79792330 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 7 Feb 2018 21:05:44 +0000 Subject: drm/i915: Only allocate preempt context when required If we remove some hardcoded assumptions about the preempt context having a fixed id, reserved from use by normal user contexts, we may only allocate the i915_gem_context when required. Then the subsequent decisions on using preemption reduce to having the preempt context available. v2: Include an assert that we don't allocate the preempt context twice. Signed-off-by: Chris Wilson Cc: Michal Winiarski Cc: Tvrtko Ursulin Cc: Arkadiusz Hiler Cc: Mika Kuoppala Cc: Daniele Ceraolo Spurio Acked-by: Daniele Ceraolo Spurio Link: https://patchwork.freedesktop.org/patch/msgid/20180207210544.26351-3-chris@chris-wilson.co.uk Reviewed-by: Michel Thierry --- drivers/gpu/drm/i915/intel_engine_cs.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_engine_cs.c') diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 7eebfbb95e89..bf634432c9c6 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -631,7 +631,7 @@ int intel_engine_init_common(struct intel_engine_cs *engine) * Similarly the preempt context must always be available so that * we can interrupt the engine at any time. */ - if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) { + if (engine->i915->preempt_context) { ring = engine->context_pin(engine, engine->i915->preempt_context); if (IS_ERR(ring)) { @@ -656,7 +656,7 @@ int intel_engine_init_common(struct intel_engine_cs *engine) err_breadcrumbs: intel_engine_fini_breadcrumbs(engine); err_unpin_preempt: - if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) + if (engine->i915->preempt_context) engine->context_unpin(engine, engine->i915->preempt_context); err_unpin_kernel: engine->context_unpin(engine, engine->i915->kernel_context); @@ -686,7 +686,7 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) if (engine->default_state) i915_gem_object_put(engine->default_state); - if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) + if (engine->i915->preempt_context) engine->context_unpin(engine, engine->i915->preempt_context); engine->context_unpin(engine, engine->i915->kernel_context); } -- cgit v1.2.3 From 8e47b4b65bbd646b4d8be28e1ffff7a684f60fe6 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 8 Feb 2018 15:12:24 +0000 Subject: drm/i915: Remove redundant check on execlists interrupt Since commit 4a118ecbe99c ("drm/i915: Filter out spurious execlists context-switch interrupts") we probe execlists->active, and no longer have to peek at the execlist interrupt to determine if the tasklet still needs to be run to drain the ELSP. References: 4a118ecbe99c ("drm/i915: Filter out spurious execlists context-switch interrupts") Signed-off-by: Chris Wilson Cc: Michal Winiarski Cc: Tvrtko Ursulin Cc: Mika Kuoppala Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20180208151224.16285-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_engine_cs.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_engine_cs.c') diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index bf634432c9c6..0ad9184eba97 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1503,10 +1503,6 @@ bool intel_engine_is_idle(struct intel_engine_cs *engine) if (I915_SELFTEST_ONLY(engine->breadcrumbs.mock)) return true; - /* Interrupt/tasklet pending? */ - if (test_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted)) - return false; - /* Waiting to drain ELSP? */ if (READ_ONCE(engine->execlists.active)) return false; -- cgit v1.2.3 From 74d00d28a15c8452f65de0a9477b52d95639cc63 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 12 Feb 2018 09:39:28 +0000 Subject: drm/i915: Don't wake the device up to check if the engine is asleep If the entire device is powered off, we can safely assume that the engine is also asleep (and idle). Reported-by: Tvrtko Ursulin Fixes: a091d4ee931b ("drm/i915: Hold a wakeref for probing the ring registers") Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Mika Kuoppala Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20180212093928.6005-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_engine_cs.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/i915/intel_engine_cs.c') diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 0ad9184eba97..3efc589a7f37 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1464,7 +1464,9 @@ static bool ring_is_idle(struct intel_engine_cs *engine) struct drm_i915_private *dev_priv = engine->i915; bool idle = true; - intel_runtime_pm_get(dev_priv); + /* If the whole device is asleep, the engine must be idle */ + if (!intel_runtime_pm_get_if_in_use(dev_priv)) + return true; /* First check that no commands are left in the ring */ if ((I915_READ_HEAD(engine) & HEAD_ADDR) != -- cgit v1.2.3 From 3ceda3a4a856336c5c902a6cfbacff2c4effaa21 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 12 Feb 2018 10:24:15 +0000 Subject: drm/i915: Hold rpm wakeref for printing the engine's register state When dumping the engine, we print out the current register values. This requires the rpm wakeref. If the device is alseep, we can assume the engine is asleep (and the register state is uninteresting) so skip and only acquire the rpm wakeref if the device is already awake. Reported-by: Tvrtko Ursulin Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Mika Kuoppala Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20180212102415.24246-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_engine_cs.c | 162 ++++++++++++++++++--------------- 1 file changed, 91 insertions(+), 71 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_engine_cs.c') diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 3efc589a7f37..f3c5100d629e 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -691,7 +691,7 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) engine->context_unpin(engine, engine->i915->kernel_context); } -u64 intel_engine_get_active_head(struct intel_engine_cs *engine) +u64 intel_engine_get_active_head(const struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; u64 acthd; @@ -707,7 +707,7 @@ u64 intel_engine_get_active_head(struct intel_engine_cs *engine) return acthd; } -u64 intel_engine_get_last_batch_head(struct intel_engine_cs *engine) +u64 intel_engine_get_last_batch_head(const struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; u64 bbaddr; @@ -1705,73 +1705,20 @@ static void hexdump(struct drm_printer *m, const void *buf, size_t len) } } -void intel_engine_dump(struct intel_engine_cs *engine, - struct drm_printer *m, - const char *header, ...) +static void intel_engine_print_registers(const struct intel_engine_cs *engine, + struct drm_printer *m) { - struct intel_breadcrumbs * const b = &engine->breadcrumbs; - const struct intel_engine_execlists * const execlists = &engine->execlists; - struct i915_gpu_error * const error = &engine->i915->gpu_error; struct drm_i915_private *dev_priv = engine->i915; - struct drm_i915_gem_request *rq; - struct rb_node *rb; - char hdr[80]; + const struct intel_engine_execlists * const execlists = + &engine->execlists; u64 addr; - if (header) { - va_list ap; - - va_start(ap, header); - drm_vprintf(m, header, &ap); - va_end(ap); - } - - if (i915_terminally_wedged(&engine->i915->gpu_error)) - drm_printf(m, "*** WEDGED ***\n"); - - drm_printf(m, "\tcurrent seqno %x, last %x, hangcheck %x [%d ms], inflight %d\n", - intel_engine_get_seqno(engine), - intel_engine_last_submit(engine), - engine->hangcheck.seqno, - jiffies_to_msecs(jiffies - engine->hangcheck.action_timestamp), - engine->timeline->inflight_seqnos); - drm_printf(m, "\tReset count: %d (global %d)\n", - i915_reset_engine_count(error, engine), - i915_reset_count(error)); - - rcu_read_lock(); - - drm_printf(m, "\tRequests:\n"); - - rq = list_first_entry(&engine->timeline->requests, - struct drm_i915_gem_request, link); - if (&rq->link != &engine->timeline->requests) - print_request(m, rq, "\t\tfirst "); - - rq = list_last_entry(&engine->timeline->requests, - struct drm_i915_gem_request, link); - if (&rq->link != &engine->timeline->requests) - print_request(m, rq, "\t\tlast "); - - rq = i915_gem_find_active_request(engine); - if (rq) { - print_request(m, rq, "\t\tactive "); - drm_printf(m, - "\t\t[head %04x, postfix %04x, tail %04x, batch 0x%08x_%08x]\n", - rq->head, rq->postfix, rq->tail, - rq->batch ? upper_32_bits(rq->batch->node.start) : ~0u, - rq->batch ? lower_32_bits(rq->batch->node.start) : ~0u); - } - - drm_printf(m, "\tRING_START: 0x%08x [0x%08x]\n", - I915_READ(RING_START(engine->mmio_base)), - rq ? i915_ggtt_offset(rq->ring->vma) : 0); - drm_printf(m, "\tRING_HEAD: 0x%08x [0x%08x]\n", - I915_READ(RING_HEAD(engine->mmio_base)) & HEAD_ADDR, - rq ? rq->ring->head : 0); - drm_printf(m, "\tRING_TAIL: 0x%08x [0x%08x]\n", - I915_READ(RING_TAIL(engine->mmio_base)) & TAIL_ADDR, - rq ? rq->ring->tail : 0); + drm_printf(m, "\tRING_START: 0x%08x\n", + I915_READ(RING_START(engine->mmio_base))); + drm_printf(m, "\tRING_HEAD: 0x%08x\n", + I915_READ(RING_HEAD(engine->mmio_base)) & HEAD_ADDR); + drm_printf(m, "\tRING_TAIL: 0x%08x\n", + I915_READ(RING_TAIL(engine->mmio_base)) & TAIL_ADDR); drm_printf(m, "\tRING_CTL: 0x%08x%s\n", I915_READ(RING_CTL(engine->mmio_base)), I915_READ(RING_CTL(engine->mmio_base)) & (RING_WAIT | RING_WAIT_SEMAPHORE) ? " [waiting]" : ""); @@ -1780,6 +1727,11 @@ void intel_engine_dump(struct intel_engine_cs *engine, I915_READ(RING_MI_MODE(engine->mmio_base)), I915_READ(RING_MI_MODE(engine->mmio_base)) & (MODE_IDLE) ? " [idle]" : ""); } + + if (INTEL_GEN(dev_priv) >= 6) { + drm_printf(m, "\tRING_IMR: %08x\n", I915_READ_IMR(engine)); + } + if (HAS_LEGACY_SEMAPHORES(dev_priv)) { drm_printf(m, "\tSYNC_0: 0x%08x\n", I915_READ(RING_SYNC_0(engine->mmio_base))); @@ -1790,8 +1742,6 @@ void intel_engine_dump(struct intel_engine_cs *engine, I915_READ(RING_SYNC_2(engine->mmio_base))); } - rcu_read_unlock(); - addr = intel_engine_get_active_head(engine); drm_printf(m, "\tACTHD: 0x%08x_%08x\n", upper_32_bits(addr), lower_32_bits(addr)); @@ -1853,10 +1803,13 @@ void intel_engine_dump(struct intel_engine_cs *engine, rcu_read_lock(); for (idx = 0; idx < execlists_num_ports(execlists); idx++) { + struct drm_i915_gem_request *rq; unsigned int count; rq = port_unpack(&execlists->port[idx], &count); if (rq) { + char hdr[80]; + snprintf(hdr, sizeof(hdr), "\t\tELSP[%d] count=%d, rq: ", idx, count); @@ -1875,6 +1828,77 @@ void intel_engine_dump(struct intel_engine_cs *engine, drm_printf(m, "\tPP_DIR_DCLV: 0x%08x\n", I915_READ(RING_PP_DIR_DCLV(engine))); } +} + +void intel_engine_dump(struct intel_engine_cs *engine, + struct drm_printer *m, + const char *header, ...) +{ + struct intel_breadcrumbs * const b = &engine->breadcrumbs; + const struct intel_engine_execlists * const execlists = &engine->execlists; + struct i915_gpu_error * const error = &engine->i915->gpu_error; + struct drm_i915_gem_request *rq; + struct rb_node *rb; + + if (header) { + va_list ap; + + va_start(ap, header); + drm_vprintf(m, header, &ap); + va_end(ap); + } + + if (i915_terminally_wedged(&engine->i915->gpu_error)) + drm_printf(m, "*** WEDGED ***\n"); + + drm_printf(m, "\tcurrent seqno %x, last %x, hangcheck %x [%d ms], inflight %d\n", + intel_engine_get_seqno(engine), + intel_engine_last_submit(engine), + engine->hangcheck.seqno, + jiffies_to_msecs(jiffies - engine->hangcheck.action_timestamp), + engine->timeline->inflight_seqnos); + drm_printf(m, "\tReset count: %d (global %d)\n", + i915_reset_engine_count(error, engine), + i915_reset_count(error)); + + rcu_read_lock(); + + drm_printf(m, "\tRequests:\n"); + + rq = list_first_entry(&engine->timeline->requests, + struct drm_i915_gem_request, link); + if (&rq->link != &engine->timeline->requests) + print_request(m, rq, "\t\tfirst "); + + rq = list_last_entry(&engine->timeline->requests, + struct drm_i915_gem_request, link); + if (&rq->link != &engine->timeline->requests) + print_request(m, rq, "\t\tlast "); + + rq = i915_gem_find_active_request(engine); + if (rq) { + print_request(m, rq, "\t\tactive "); + drm_printf(m, + "\t\t[head %04x, postfix %04x, tail %04x, batch 0x%08x_%08x]\n", + rq->head, rq->postfix, rq->tail, + rq->batch ? upper_32_bits(rq->batch->node.start) : ~0u, + rq->batch ? lower_32_bits(rq->batch->node.start) : ~0u); + drm_printf(m, "\t\tring->start: 0x%08x\n", + i915_ggtt_offset(rq->ring->vma)); + drm_printf(m, "\t\tring->head: 0x%08x\n", + rq->ring->head); + drm_printf(m, "\t\tring->tail: 0x%08x\n", + rq->ring->tail); + } + + rcu_read_unlock(); + + if (intel_runtime_pm_get_if_in_use(engine->i915)) { + intel_engine_print_registers(engine, m); + intel_runtime_pm_put(engine->i915); + } else { + drm_printf(m, "\tDevice is asleep; skipping register dump\n"); + } spin_lock_irq(&engine->timeline->lock); list_for_each_entry(rq, &engine->timeline->requests, link) @@ -1897,10 +1921,6 @@ void intel_engine_dump(struct intel_engine_cs *engine, } spin_unlock_irq(&b->rb_lock); - if (INTEL_GEN(dev_priv) >= 6) { - drm_printf(m, "\tRING_IMR: %08x\n", I915_READ_IMR(engine)); - } - drm_printf(m, "IRQ? 0x%lx (breadcrumbs? %s) (execlists? %s)\n", engine->irq_posted, yesno(test_bit(ENGINE_IRQ_BREADCRUMB, -- cgit v1.2.3