All of lore.kernel.org
 help / color / mirror / Atom feed
From: Chris Wilson <chris@chris-wilson.co.uk>
To: intel-gfx@lists.freedesktop.org
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Subject: [PATCH 2/6] drm/i915: Separate out the seqno-barrier from engine->get_seqno
Date: Fri, 15 Jan 2016 14:35:41 +0000	[thread overview]
Message-ID: <1452868545-19586-3-git-send-email-chris@chris-wilson.co.uk> (raw)
In-Reply-To: <1452868545-19586-1-git-send-email-chris@chris-wilson.co.uk>

In order to simplify future patches, extract the
lazy_coherency optimisation our of the engine->get_seqno() vfunc into
its own callback.

v2: Rename the barrier to engine->irq_seqno_barrier to try and better
reflect that the barrier is only required after the user interrupt before
reading the seqno (to ensure that the seqno update lands in time as we
do not have strict seqno-irq ordering on all platforms).

Reviewed-by: Dave Gordon <david.s.gordon@intel.com> [#v2]

v3: Comments for hangcheck paranoia. Mika wanted to keep the extra
barrier inside the hangcheck, just in case. I can argue that it doesn't
provide a barrier against anything, but the side-effects of applying the
barrier may prevent a false declaration of a hung GPU.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c     |  6 +++---
 drivers/gpu/drm/i915/i915_drv.h         | 12 +++++++----
 drivers/gpu/drm/i915/i915_gpu_error.c   |  2 +-
 drivers/gpu/drm/i915/i915_irq.c         | 14 +++++++++++--
 drivers/gpu/drm/i915/i915_trace.h       |  2 +-
 drivers/gpu/drm/i915/intel_lrc.c        | 18 ++++++-----------
 drivers/gpu/drm/i915/intel_ringbuffer.c | 36 +++++++++++++++++----------------
 drivers/gpu/drm/i915/intel_ringbuffer.h |  4 ++--
 8 files changed, 52 insertions(+), 42 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index e3377abc0d4d..b421b53ca128 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -600,7 +600,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)
 					   ring->name,
 					   i915_gem_request_get_seqno(work->flip_queued_req),
 					   dev_priv->next_seqno,
-					   ring->get_seqno(ring, true),
+					   ring->get_seqno(ring),
 					   i915_gem_request_completed(work->flip_queued_req, true));
 			} else
 				seq_printf(m, "Flip not associated with any ring\n");
@@ -732,7 +732,7 @@ static void i915_ring_seqno_info(struct seq_file *m,
 {
 	if (ring->get_seqno) {
 		seq_printf(m, "Current sequence (%s): %x\n",
-			   ring->name, ring->get_seqno(ring, false));
+			   ring->name, ring->get_seqno(ring));
 	}
 }
 
@@ -1342,8 +1342,8 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused)
 	intel_runtime_pm_get(dev_priv);
 
 	for_each_ring(ring, dev_priv, i) {
-		seqno[i] = ring->get_seqno(ring, false);
 		acthd[i] = intel_ring_get_active_head(ring);
+		seqno[i] = ring->get_seqno(ring);
 	}
 
 	i915_get_extra_instdone(dev, instdone);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index eb7bb97f7316..0cdaed66b786 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2961,15 +2961,19 @@ i915_seqno_passed(uint32_t seq1, uint32_t seq2)
 static inline bool i915_gem_request_started(struct drm_i915_gem_request *req,
 					   bool lazy_coherency)
 {
-	u32 seqno = req->ring->get_seqno(req->ring, lazy_coherency);
-	return i915_seqno_passed(seqno, req->previous_seqno);
+	if (!lazy_coherency && req->ring->irq_seqno_barrier)
+		req->ring->irq_seqno_barrier(req->ring);
+	return i915_seqno_passed(req->ring->get_seqno(req->ring),
+				 req->previous_seqno);
 }
 
 static inline bool i915_gem_request_completed(struct drm_i915_gem_request *req,
 					      bool lazy_coherency)
 {
-	u32 seqno = req->ring->get_seqno(req->ring, lazy_coherency);
-	return i915_seqno_passed(seqno, req->seqno);
+	if (!lazy_coherency && req->ring->irq_seqno_barrier)
+		req->ring->irq_seqno_barrier(req->ring);
+	return i915_seqno_passed(req->ring->get_seqno(req->ring),
+				 req->seqno);
 }
 
 int __must_check i915_gem_get_seqno(struct drm_device *dev, u32 *seqno);
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 06ca4082735b..b978febf67a0 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -902,8 +902,8 @@ static void i915_record_ring_state(struct drm_device *dev,
 
 	ering->waiting = waitqueue_active(&ring->irq_queue);
 	ering->instpm = I915_READ(RING_INSTPM(ring->mmio_base));
-	ering->seqno = ring->get_seqno(ring, false);
 	ering->acthd = intel_ring_get_active_head(ring);
+	ering->seqno = ring->get_seqno(ring);
 	ering->start = I915_READ_START(ring);
 	ering->head = I915_READ_HEAD(ring);
 	ering->tail = I915_READ_TAIL(ring);
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 25a89373df63..07bc2cdd6252 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -2937,7 +2937,7 @@ static int semaphore_passed(struct intel_engine_cs *ring)
 	if (signaller->hangcheck.deadlock >= I915_NUM_RINGS)
 		return -1;
 
-	if (i915_seqno_passed(signaller->get_seqno(signaller, false), seqno))
+	if (i915_seqno_passed(signaller->get_seqno(signaller), seqno))
 		return 1;
 
 	/* cursory check for an unkickable deadlock */
@@ -3101,8 +3101,18 @@ static void i915_hangcheck_elapsed(struct work_struct *work)
 
 		semaphore_clear_deadlocks(dev_priv);
 
-		seqno = ring->get_seqno(ring, false);
+		/* We don't strictly need an irq-barrier here, as we are not
+		 * serving an interrupt request, be paranoid in case the
+		 * barrier has side-effects (such as preventing a broken
+		 * cacheline snoop) and so be sure that we can see the seqno
+		 * advance. If the seqno should stick, due to a stale
+		 * cacheline, we would erroneously declare the GPU hung.
+		 */
+		if (ring->irq_seqno_barrier)
+			ring->irq_seqno_barrier(ring);
+
 		acthd = intel_ring_get_active_head(ring);
+		seqno = ring->get_seqno(ring);
 
 		if (ring->hangcheck.seqno == seqno) {
 			if (ring_idle(ring, seqno)) {
diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index 52b2d409945d..cfb5f78a6e84 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -573,7 +573,7 @@ TRACE_EVENT(i915_gem_request_notify,
 	    TP_fast_assign(
 			   __entry->dev = ring->dev->primary->index;
 			   __entry->ring = ring->id;
-			   __entry->seqno = ring->get_seqno(ring, false);
+			   __entry->seqno = ring->get_seqno(ring);
 			   ),
 
 	    TP_printk("dev=%u, ring=%u, seqno=%u",
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index f5d89c845ede..3ed4ab7f571e 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -1781,7 +1781,7 @@ static int gen8_emit_flush_render(struct drm_i915_gem_request *request,
 	return 0;
 }
 
-static u32 gen8_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency)
+static u32 gen8_get_seqno(struct intel_engine_cs *ring)
 {
 	return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
 }
@@ -1791,9 +1791,8 @@ static void gen8_set_seqno(struct intel_engine_cs *ring, u32 seqno)
 	intel_write_status_page(ring, I915_GEM_HWS_INDEX, seqno);
 }
 
-static u32 bxt_a_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency)
+static void bxt_seqno_barrier(struct intel_engine_cs *ring)
 {
-
 	/*
 	 * On BXT A steppings there is a HW coherency issue whereby the
 	 * MI_STORE_DATA_IMM storing the completed request's seqno
@@ -1804,11 +1803,7 @@ static u32 bxt_a_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency)
 	 * bxt_a_set_seqno(), where we also do a clflush after the write. So
 	 * this clflush in practice becomes an invalidate operation.
 	 */
-
-	if (!lazy_coherency)
-		intel_flush_status_page(ring, I915_GEM_HWS_INDEX);
-
-	return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
+	intel_flush_status_page(ring, I915_GEM_HWS_INDEX);
 }
 
 static void bxt_a_set_seqno(struct intel_engine_cs *ring, u32 seqno)
@@ -1954,12 +1949,11 @@ logical_ring_default_vfuncs(struct drm_device *dev,
 	ring->irq_get = gen8_logical_ring_get_irq;
 	ring->irq_put = gen8_logical_ring_put_irq;
 	ring->emit_bb_start = gen8_emit_bb_start;
+	ring->get_seqno = gen8_get_seqno;
+	ring->set_seqno = gen8_set_seqno;
 	if (IS_BXT_REVID(dev, 0, BXT_REVID_A1)) {
-		ring->get_seqno = bxt_a_get_seqno;
+		ring->irq_seqno_barrier = bxt_seqno_barrier;
 		ring->set_seqno = bxt_a_set_seqno;
-	} else {
-		ring->get_seqno = gen8_get_seqno;
-		ring->set_seqno = gen8_set_seqno;
 	}
 }
 
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 935add1422ae..13ef3e956901 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -1487,8 +1487,8 @@ pc_render_add_request(struct drm_i915_gem_request *req)
 	return 0;
 }
 
-static u32
-gen6_ring_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency)
+static void
+gen6_seqno_barrier(struct intel_engine_cs *ring)
 {
 	/* Workaround to force correct ordering between irq and seqno writes on
 	 * ivb (and maybe also on snb) by reading from a CS register (like
@@ -1502,18 +1502,14 @@ gen6_ring_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency)
 	 * a delay after every batch i.e. much more frequent than a delay
 	 * when waiting for the interrupt (with the same net latency).
 	 */
-	if (!lazy_coherency) {
-		struct drm_i915_private *dev_priv = ring->dev->dev_private;
-		POSTING_READ_FW(RING_ACTHD(ring->mmio_base));
-
-		intel_flush_status_page(ring, I915_GEM_HWS_INDEX);
-	}
+	struct drm_i915_private *dev_priv = to_i915(ring->dev);
+	POSTING_READ_FW(RING_ACTHD(ring->mmio_base));
 
-	return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
+	intel_flush_status_page(ring, I915_GEM_HWS_INDEX);
 }
 
 static u32
-ring_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency)
+ring_get_seqno(struct intel_engine_cs *ring)
 {
 	return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
 }
@@ -1525,7 +1521,7 @@ ring_set_seqno(struct intel_engine_cs *ring, u32 seqno)
 }
 
 static u32
-pc_render_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency)
+pc_render_get_seqno(struct intel_engine_cs *ring)
 {
 	return ring->scratch.cpu_page[0];
 }
@@ -2717,7 +2713,8 @@ int intel_init_render_ring_buffer(struct drm_device *dev)
 		ring->irq_get = gen8_ring_get_irq;
 		ring->irq_put = gen8_ring_put_irq;
 		ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
-		ring->get_seqno = gen6_ring_get_seqno;
+		ring->irq_seqno_barrier = gen6_seqno_barrier;
+		ring->get_seqno = ring_get_seqno;
 		ring->set_seqno = ring_set_seqno;
 		if (i915_semaphore_is_enabled(dev)) {
 			WARN_ON(!dev_priv->semaphore_obj);
@@ -2734,7 +2731,8 @@ int intel_init_render_ring_buffer(struct drm_device *dev)
 		ring->irq_get = gen6_ring_get_irq;
 		ring->irq_put = gen6_ring_put_irq;
 		ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
-		ring->get_seqno = gen6_ring_get_seqno;
+		ring->irq_seqno_barrier = gen6_seqno_barrier;
+		ring->get_seqno = ring_get_seqno;
 		ring->set_seqno = ring_set_seqno;
 		if (i915_semaphore_is_enabled(dev)) {
 			ring->semaphore.sync_to = gen6_ring_sync;
@@ -2848,7 +2846,8 @@ int intel_init_bsd_ring_buffer(struct drm_device *dev)
 			ring->write_tail = gen6_bsd_ring_write_tail;
 		ring->flush = gen6_bsd_ring_flush;
 		ring->add_request = gen6_add_request;
-		ring->get_seqno = gen6_ring_get_seqno;
+		ring->irq_seqno_barrier = gen6_seqno_barrier;
+		ring->get_seqno = ring_get_seqno;
 		ring->set_seqno = ring_set_seqno;
 		if (INTEL_INFO(dev)->gen >= 8) {
 			ring->irq_enable_mask =
@@ -2920,7 +2919,8 @@ int intel_init_bsd2_ring_buffer(struct drm_device *dev)
 	ring->mmio_base = GEN8_BSD2_RING_BASE;
 	ring->flush = gen6_bsd_ring_flush;
 	ring->add_request = gen6_add_request;
-	ring->get_seqno = gen6_ring_get_seqno;
+	ring->irq_seqno_barrier = gen6_seqno_barrier;
+	ring->get_seqno = ring_get_seqno;
 	ring->set_seqno = ring_set_seqno;
 	ring->irq_enable_mask =
 			GT_RENDER_USER_INTERRUPT << GEN8_VCS2_IRQ_SHIFT;
@@ -2950,7 +2950,8 @@ int intel_init_blt_ring_buffer(struct drm_device *dev)
 	ring->write_tail = ring_write_tail;
 	ring->flush = gen6_ring_flush;
 	ring->add_request = gen6_add_request;
-	ring->get_seqno = gen6_ring_get_seqno;
+	ring->irq_seqno_barrier = gen6_seqno_barrier;
+	ring->get_seqno = ring_get_seqno;
 	ring->set_seqno = ring_set_seqno;
 	if (INTEL_INFO(dev)->gen >= 8) {
 		ring->irq_enable_mask =
@@ -3007,7 +3008,8 @@ int intel_init_vebox_ring_buffer(struct drm_device *dev)
 	ring->write_tail = ring_write_tail;
 	ring->flush = gen6_ring_flush;
 	ring->add_request = gen6_add_request;
-	ring->get_seqno = gen6_ring_get_seqno;
+	ring->irq_seqno_barrier = gen6_seqno_barrier;
+	ring->get_seqno = ring_get_seqno;
 	ring->set_seqno = ring_set_seqno;
 
 	if (INTEL_INFO(dev)->gen >= 8) {
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 7349d9258191..8fb02b21e75d 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -193,8 +193,8 @@ struct  intel_engine_cs {
 	 * seen value is good enough. Note that the seqno will always be
 	 * monotonic, even if not coherent.
 	 */
-	u32		(*get_seqno)(struct intel_engine_cs *ring,
-				     bool lazy_coherency);
+	void		(*irq_seqno_barrier)(struct intel_engine_cs *ring);
+	u32		(*get_seqno)(struct intel_engine_cs *ring);
 	void		(*set_seqno)(struct intel_engine_cs *ring,
 				     u32 seqno);
 	int		(*dispatch_execbuffer)(struct drm_i915_gem_request *req,
-- 
2.7.0.rc3

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

  parent reply	other threads:[~2016-01-15 14:35 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-01-15 14:35 "missed-interrupt" syndrome on Broadwell+ Chris Wilson
2016-01-15 14:35 ` [PATCH 1/6] drm/i915: Remove forcewake dance from seqno/irq barrier on legacy gen6+ Chris Wilson
2016-01-15 17:55   ` Mika Kuoppala
2016-01-15 14:35 ` Chris Wilson [this message]
2016-01-15 14:35 ` [PATCH 3/6] drm/i915: Broadwell execlists needs exactly the same seqno w/a as legacy Chris Wilson
2016-01-15 14:35 ` [PATCH 4/6] drm/i915: Harden detection of missed interrupts Chris Wilson
2016-01-15 14:35 ` [PATCH 5/6] drm/i915: Use simplest form for flushing the single cacheline in the HWS Chris Wilson
2016-01-15 14:35 ` [PATCH 6/6] drm/i915: Replace manual barrier() with READ_ONCE() in HWS accessor Chris Wilson
2016-01-15 15:20 ` ✗ Fi.CI.BAT: warning for series starting with [1/6] drm/i915: Remove forcewake dance from seqno/irq barrier on legacy gen6+ Patchwork
2016-01-16  9:46 ` [PATCH v2 1/6] " Chris Wilson
2016-01-16  9:46   ` [PATCH v2 2/6] drm/i915: Separate out the seqno-barrier from engine->get_seqno Chris Wilson
2016-01-16  9:46   ` [PATCH v2 3/6] drm/i915: Use ordered seqno write interrupt generation on gen8+ execlists Chris Wilson
2016-01-18  8:58     ` [PATCH] magic-clflush-fix Chris Wilson
2016-01-18  9:02       ` Chris Wilson
2016-01-18  9:02     ` [PATCH] drm/i915: Use ordered seqno write interrupt generation on gen8+ execlists Chris Wilson
2016-01-16  9:46   ` [PATCH v2 4/6] drm/i915: Harden detection of missed interrupts Chris Wilson
2016-01-18 13:07     ` Mika Kuoppala
2016-01-18 15:35       ` Chris Wilson
2016-01-16  9:46   ` [PATCH v2 5/6] drm/i915: Use simplest form for flushing the single cacheline in the HWS Chris Wilson
2016-01-16  9:46   ` [PATCH v2 6/6] drm/i915: Replace manual barrier() with READ_ONCE() in HWS accessor Chris Wilson
2016-01-16 10:01 ` ✗ Fi.CI.BAT: failure for series starting with [1/6] drm/i915: Remove forcewake dance from seqno/irq barrier on legacy gen6+ (rev6) Patchwork
2016-01-18  9:30 ` ✗ Fi.CI.BAT: failure for series starting with [1/6] drm/i915: Remove forcewake dance from seqno/irq barrier on legacy gen6+ (rev8) Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1452868545-19586-3-git-send-email-chris@chris-wilson.co.uk \
    --to=chris@chris-wilson.co.uk \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=mika.kuoppala@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.