drm/i915: Separate out the seqno-barrier from engine->get_seqno

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>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1460195877-20520-2-git-send-email-chris@chris-wilson.co.uk

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index ebbf4e40068423e44bf2f670e16f407b193b454b..919c05ba9932c041df1d19775c4ddae6d0c979fa 100644 (file)
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -598,7 +598,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)
                                           engine->name,
                                           i915_gem_request_get_seqno(work->flip_queued_req),
                                           dev_priv->next_seqno,
-                                          engine->get_seqno(engine, true),
+                                          engine->get_seqno(engine),
                                           i915_gem_request_completed(work->flip_queued_req, true));
                        } else
                                seq_printf(m, "Flip not associated with any ring\n");
@@ -730,7 +730,7 @@ static void i915_ring_seqno_info(struct seq_file *m,
 {
        if (engine->get_seqno) {
                seq_printf(m, "Current sequence (%s): %x\n",
-                          engine->name, engine->get_seqno(engine, false));
+                          engine->name, engine->get_seqno(engine));
        }
 }
 
@@ -1346,8 +1346,8 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused)
        intel_runtime_pm_get(dev_priv);
 
        for_each_engine_id(engine, dev_priv, id) {
-               seqno[id] = engine->get_seqno(engine, false);
                acthd[id] = intel_ring_get_active_head(engine);
+               seqno[id] = engine->get_seqno(engine);
        }
 
        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 a93e5dd4fa9acbdd78573fb120827ca3977b0d8d..542401659013e4f312e970fc9ec4633db5053828 100644 (file)
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3017,15 +3017,19 @@ 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->engine->get_seqno(req->engine, lazy_coherency);
-       return i915_seqno_passed(seqno, req->previous_seqno);
+       if (!lazy_coherency && req->engine->irq_seqno_barrier)
+               req->engine->irq_seqno_barrier(req->engine);
+       return i915_seqno_passed(req->engine->get_seqno(req->engine),
+                                req->previous_seqno);
 }
 
 static inline bool i915_gem_request_completed(struct drm_i915_gem_request *req,
                                              bool lazy_coherency)
 {
-       u32 seqno = req->engine->get_seqno(req->engine, lazy_coherency);
-       return i915_seqno_passed(seqno, req->seqno);
+       if (!lazy_coherency && req->engine->irq_seqno_barrier)
+               req->engine->irq_seqno_barrier(req->engine);
+       return i915_seqno_passed(req->engine->get_seqno(req->engine),
+                                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 ce77713a555d8b5d2751ffce9f82f9cafdf20acc..89725c9efc2580c2adc061365456b3624bfbffdb 100644 (file)
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -931,8 +931,8 @@ static void i915_record_ring_state(struct drm_device *dev,
 
        ering->waiting = waitqueue_active(&engine->irq_queue);
        ering->instpm = I915_READ(RING_INSTPM(engine->mmio_base));
-       ering->seqno = engine->get_seqno(engine, false);
        ering->acthd = intel_ring_get_active_head(engine);
+       ering->seqno = engine->get_seqno(engine);
        ering->last_seqno = engine->last_submitted_seqno;
        ering->start = I915_READ_START(engine);
        ering->head = I915_READ_HEAD(engine);

diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index c30a12ef6dafdaa1261f2f440bf32a0ffbb4b92f..3b946e1c76146d3a049abbdc705254f793513e7d 100644 (file)
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -2941,7 +2941,7 @@ static int semaphore_passed(struct intel_engine_cs *engine)
        if (signaller->hangcheck.deadlock >= I915_NUM_ENGINES)
                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 */
@@ -3100,8 +3100,18 @@ static void i915_hangcheck_elapsed(struct work_struct *work)
 
                semaphore_clear_deadlocks(dev_priv);
 
-               seqno = engine->get_seqno(engine, 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 (engine->irq_seqno_barrier)
+                       engine->irq_seqno_barrier(engine);
+
                acthd = intel_ring_get_active_head(engine);
+               seqno = engine->get_seqno(engine);
 
                if (engine->hangcheck.seqno == seqno) {
                        if (ring_idle(engine, seqno)) {

diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index afdd8aefb5b75fe8b5d96791b6310ac3f2612e73..dc0def210097eeae4311b2777b0776473d69cfc3 100644 (file)
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -562,7 +562,7 @@ TRACE_EVENT(i915_gem_request_notify,
            TP_fast_assign(
                           __entry->dev = engine->dev->primary->index;
                           __entry->ring = engine->id;
-                          __entry->seqno = engine->get_seqno(engine, false);
+                          __entry->seqno = engine->get_seqno(engine);
                           ),
 
            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 67d48d759b46397bd24604242f986145e1d8b7cd..f209ecfdcb5c9a408e18e8a699e50fabd05bb4a7 100644 (file)
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -1852,7 +1852,7 @@ static int gen8_emit_flush_render(struct drm_i915_gem_request *request,
        return 0;
 }
 
-static u32 gen8_get_seqno(struct intel_engine_cs *engine, bool lazy_coherency)
+static u32 gen8_get_seqno(struct intel_engine_cs *engine)
 {
        return intel_read_status_page(engine, I915_GEM_HWS_INDEX);
 }
@@ -1862,10 +1862,8 @@ static void gen8_set_seqno(struct intel_engine_cs *engine, u32 seqno)
        intel_write_status_page(engine, I915_GEM_HWS_INDEX, seqno);
 }
 
-static u32 bxt_a_get_seqno(struct intel_engine_cs *engine,
-                          bool lazy_coherency)
+static void bxt_a_seqno_barrier(struct intel_engine_cs *engine)
 {
-
        /*
         * On BXT A steppings there is a HW coherency issue whereby the
         * MI_STORE_DATA_IMM storing the completed request's seqno
@@ -1876,11 +1874,7 @@ static u32 bxt_a_get_seqno(struct intel_engine_cs *engine,
         * 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(engine, I915_GEM_HWS_INDEX);
-
-       return intel_read_status_page(engine, I915_GEM_HWS_INDEX);
+       intel_flush_status_page(engine, I915_GEM_HWS_INDEX);
 }
 
 static void bxt_a_set_seqno(struct intel_engine_cs *engine, u32 seqno)
@@ -2058,12 +2052,11 @@ logical_ring_default_vfuncs(struct drm_device *dev,
        engine->irq_get = gen8_logical_ring_get_irq;
        engine->irq_put = gen8_logical_ring_put_irq;
        engine->emit_bb_start = gen8_emit_bb_start;
+       engine->get_seqno = gen8_get_seqno;
+       engine->set_seqno = gen8_set_seqno;
        if (IS_BXT_REVID(dev, 0, BXT_REVID_A1)) {
-               engine->get_seqno = bxt_a_get_seqno;
+               engine->irq_seqno_barrier = bxt_a_seqno_barrier;
                engine->set_seqno = bxt_a_set_seqno;
-       } else {
-               engine->get_seqno = gen8_get_seqno;
-               engine->set_seqno = gen8_set_seqno;
        }
 }
 

diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 69cc3bc20495f1f4ae8798a84b7bd4dac504cae7..e144f4f301bfe0def96c105be9d02ccf8a4e0e94 100644 (file)
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -1568,8 +1568,8 @@ pc_render_add_request(struct drm_i915_gem_request *req)
        return 0;
 }
 
-static u32
-gen6_ring_get_seqno(struct intel_engine_cs *engine, bool lazy_coherency)
+static void
+gen6_seqno_barrier(struct intel_engine_cs *engine)
 {
        /* Workaround to force correct ordering between irq and seqno writes on
         * ivb (and maybe also on snb) by reading from a CS register (like
@@ -1583,16 +1583,12 @@ gen6_ring_get_seqno(struct intel_engine_cs *engine, bool lazy_coherency)
         * 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 = engine->dev->dev_private;
-               POSTING_READ_FW(RING_ACTHD(engine->mmio_base));
-       }
-
-       return intel_read_status_page(engine, I915_GEM_HWS_INDEX);
+       struct drm_i915_private *dev_priv = engine->dev->dev_private;
+       POSTING_READ_FW(RING_ACTHD(engine->mmio_base));
 }
 
 static u32
-ring_get_seqno(struct intel_engine_cs *engine, bool lazy_coherency)
+ring_get_seqno(struct intel_engine_cs *engine)
 {
        return intel_read_status_page(engine, I915_GEM_HWS_INDEX);
 }
@@ -1604,7 +1600,7 @@ ring_set_seqno(struct intel_engine_cs *engine, u32 seqno)
 }
 
 static u32
-pc_render_get_seqno(struct intel_engine_cs *engine, bool lazy_coherency)
+pc_render_get_seqno(struct intel_engine_cs *engine)
 {
        return engine->scratch.cpu_page[0];
 }
@@ -2828,7 +2824,8 @@ int intel_init_render_ring_buffer(struct drm_device *dev)
                engine->irq_get = gen8_ring_get_irq;
                engine->irq_put = gen8_ring_put_irq;
                engine->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
-               engine->get_seqno = gen6_ring_get_seqno;
+               engine->irq_seqno_barrier = gen6_seqno_barrier;
+               engine->get_seqno = ring_get_seqno;
                engine->set_seqno = ring_set_seqno;
                if (i915_semaphore_is_enabled(dev)) {
                        WARN_ON(!dev_priv->semaphore_obj);
@@ -2845,7 +2842,8 @@ int intel_init_render_ring_buffer(struct drm_device *dev)
                engine->irq_get = gen6_ring_get_irq;
                engine->irq_put = gen6_ring_put_irq;
                engine->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
-               engine->get_seqno = gen6_ring_get_seqno;
+               engine->irq_seqno_barrier = gen6_seqno_barrier;
+               engine->get_seqno = ring_get_seqno;
                engine->set_seqno = ring_set_seqno;
                if (i915_semaphore_is_enabled(dev)) {
                        engine->semaphore.sync_to = gen6_ring_sync;
@@ -2960,7 +2958,8 @@ int intel_init_bsd_ring_buffer(struct drm_device *dev)
                        engine->write_tail = gen6_bsd_ring_write_tail;
                engine->flush = gen6_bsd_ring_flush;
                engine->add_request = gen6_add_request;
-               engine->get_seqno = gen6_ring_get_seqno;
+               engine->irq_seqno_barrier = gen6_seqno_barrier;
+               engine->get_seqno = ring_get_seqno;
                engine->set_seqno = ring_set_seqno;
                if (INTEL_INFO(dev)->gen >= 8) {
                        engine->irq_enable_mask =
@@ -3033,7 +3032,8 @@ int intel_init_bsd2_ring_buffer(struct drm_device *dev)
        engine->mmio_base = GEN8_BSD2_RING_BASE;
        engine->flush = gen6_bsd_ring_flush;
        engine->add_request = gen6_add_request;
-       engine->get_seqno = gen6_ring_get_seqno;
+       engine->irq_seqno_barrier = gen6_seqno_barrier;
+       engine->get_seqno = ring_get_seqno;
        engine->set_seqno = ring_set_seqno;
        engine->irq_enable_mask =
                        GT_RENDER_USER_INTERRUPT << GEN8_VCS2_IRQ_SHIFT;
@@ -3064,7 +3064,8 @@ int intel_init_blt_ring_buffer(struct drm_device *dev)
        engine->write_tail = ring_write_tail;
        engine->flush = gen6_ring_flush;
        engine->add_request = gen6_add_request;
-       engine->get_seqno = gen6_ring_get_seqno;
+       engine->irq_seqno_barrier = gen6_seqno_barrier;
+       engine->get_seqno = ring_get_seqno;
        engine->set_seqno = ring_set_seqno;
        if (INTEL_INFO(dev)->gen >= 8) {
                engine->irq_enable_mask =
@@ -3122,7 +3123,8 @@ int intel_init_vebox_ring_buffer(struct drm_device *dev)
        engine->write_tail = ring_write_tail;
        engine->flush = gen6_ring_flush;
        engine->add_request = gen6_add_request;
-       engine->get_seqno = gen6_ring_get_seqno;
+       engine->irq_seqno_barrier = gen6_seqno_barrier;
+       engine->get_seqno = ring_get_seqno;
        engine->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 98eadfa7911682fda6b4f98fb85d44d67d50a0de..3f04906a081fbe7025788b2c2d93c739bac2bf73 100644 (file)
--- 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,