ppgtt: ring initialization fixes
Subsystems (2)
cover ppgtt: ring initialization fixes
With this series I can get reliable ring initialization on my gen7 and ppgtt=2. No page faults nor rings running wild across address space. Prerequisite for testing dynamic page table series on gen7.
Chris Wilson (2):
drm/i915: Detect page faults during hangcheck
drm/i915: Reorder hw init to avoid executing with invalid context/mm
state
Mika Kuoppala (2):
drm/i915: Introduce ring->start_ring() drm/i915: Stop rings before cleaning up on reset
drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/i915_gem.c | 66 ++++++++++++++++++++++++--------- drivers/gpu/drm/i915/i915_irq.c | 5 +++ drivers/gpu/drm/i915/intel_lrc.c | 20 ++++++++++ drivers/gpu/drm/i915/intel_lrc.h | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 51 ++++++++++++++++++------- drivers/gpu/drm/i915/intel_ringbuffer.h | 3 ++ drivers/gpu/drm/i915/intel_uncore.c | 2 + 8 files changed, 118 insertions(+), 31 deletions(-)
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 1/4 drm/i915: Detect page faults during hangcheck
From: Chris Wilson <[email protected]>
On Sandybridge+, the GPU provides the ERROR register for detecting page faults. Hook this up to our hangcheck so that we can dump the error state soon after such an event occurs. This would be better inside an interrupt handler, but it serves a purpose here as it detects that our initial context setup is invalid...
--- drivers/gpu/drm/i915/i915_irq.c | 5 +++++ drivers/gpu/drm/i915/intel_uncore.c | 2 ++ 2 files changed, 7 insertions(+)
drivers/gpu/drm/i915/i915_irq.c +5 -0
@@ -2929,6 +2929,11 @@ static void i915_hangcheck_elapsed(struct work_struct *work)if (!i915.enable_hangcheck)return ; +if (INTEL_INFO (dev_priv)->gen >=6 &&I915_READ (ERROR_GEN6 )) { +i915_handle_error (dev, false,"GPU reported a page fault" ); +I915_WRITE (ERROR_GEN6 ,0 ); + } +for_each_ring (ring, dev_priv, i) {u64 acthd;u32 seqno;
drivers/gpu/drm/i915/intel_uncore.c +2 -0
@@ -359,6 +359,8 @@ static void __intel_uncore_early_sanitize(struct drm_device *dev,if (IS_GEN6 (dev) ||IS_GEN7 (dev))__raw_i915_write32 (dev_priv,GTFIFODBG ,__raw_i915_read32 (dev_priv,GTFIFODBG )); +if (INTEL_INFO (dev)->gen >=6 ) +__raw_i915_write32 (dev_priv,ERROR_GEN6 ,0 );intel_uncore_forcewake_reset (dev, restore_forcewake); }
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 2/4 drm/i915: Introduce ring->start_ring()
In reply to Mika KuoppalaThis allows us to fill ringbuffer without engine being active. The next patch will take advantage of this.
--- drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/i915_gem.c | 39 ++++++++++++++++++------- drivers/gpu/drm/i915/intel_lrc.c | 20 +++++++++++++ drivers/gpu/drm/i915/intel_lrc.h | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 51 ++++++++++++++++++++++++--------- drivers/gpu/drm/i915/intel_ringbuffer.h | 3 ++ 6 files changed, 90 insertions(+), 25 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +1 -0
@@ -1823,6 +1823,7 @@ struct drm_i915_private {int (*init_rings)(struct drm_device *dev);void (*cleanup_ring)(struct intel_engine_cs *ring);void (*stop_ring)(struct intel_engine_cs *ring); +void (*start_ring)(struct intel_engine_cs *ring); } gt;uint32_t request_uniq;
drivers/gpu/drm/i915/i915_gem.c +28 -11
@@ -2694,6 +2694,8 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, struct drm_i915_gem_request, list); + request->ringbuf->last_retired_head = request->postfix; +i915_gem_free_request (request); }
@@ -2722,6 +2724,28 @@ void i915_gem_restore_fences(struct drm_device *dev) } } +static void +i915_gem_stop_ringbuffers (struct drm_device *dev) +{ +struct drm_i915_private *dev_priv = dev->dev_private; +struct intel_engine_cs *ring; +int i; + +for_each_ring (ring, dev_priv, i) + dev_priv->gt.stop_ring (ring); +} + +static void +i915_gem_start_ringbuffers (struct drm_device *dev) +{ +struct drm_i915_private *dev_priv = dev->dev_private; +struct intel_engine_cs *ring; +int i; + +for_each_ring (ring, dev_priv, i) + dev_priv->gt.start_ring (ring); +} +void i915_gem_reset (struct drm_device *dev) {struct drm_i915_private *dev_priv = dev->dev_private;
@@ -4648,17 +4672,6 @@ void i915_gem_vma_destroy(struct i915_vma *vma)kfree (vma); } -static void -i915_gem_stop_ringbuffers (struct drm_device *dev) -{ -struct drm_i915_private *dev_priv = dev->dev_private; -struct intel_engine_cs *ring; -int i; - -for_each_ring (ring, dev_priv, i) - dev_priv->gt.stop_ring (ring); -} -int i915_gem_suspend (struct drm_device *dev) {
@@ -4891,6 +4904,8 @@ i915_gem_init_hw(struct drm_device *dev) goto out; } +i915_gem_start_ringbuffers (dev); +for (i =0 ; i <NUM_L3_SLICES (dev); i++)i915_gem_l3_remap (&dev_priv->ring[RCS ], i);
@@ -4936,11 +4951,13 @@ int i915_gem_init(struct drm_device *dev) dev_priv->gt.init_rings = i915_gem_init_rings; dev_priv->gt.cleanup_ring = intel_cleanup_ring_buffer; dev_priv->gt.stop_ring = intel_stop_ring_buffer; + dev_priv->gt.start_ring = intel_start_ring_buffer; }else { dev_priv->gt.do_execbuf = intel_execlists_submission; dev_priv->gt.init_rings = intel_logical_rings_init; dev_priv->gt.cleanup_ring = intel_logical_ring_cleanup; dev_priv->gt.stop_ring = intel_logical_ring_stop; + dev_priv->gt.start_ring = intel_logical_ring_start; }/* This is just a security blanket to placate dragons.
drivers/gpu/drm/i915/intel_lrc.c +20 -0
@@ -756,6 +756,8 @@ void intel_logical_ring_stop(struct intel_engine_cs *ring)if (!intel_ring_initialized (ring))return ; +intel_ring_stop (ring); + ret =intel_ring_idle (ring);if (ret && !i915_reset_in_progress (&to_i915 (ring->dev)->gpu_error)) DRM_ERROR("failed to quiesce %s whilst cleaning up: %d\n",
@@ -811,6 +813,22 @@ intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf,execlists_context_queue (ring, ctx, ringbuf->tail, request); } +void intel_logical_ring_start (struct intel_engine_cs *ring) +{ +struct intel_context *ctx = ring->default_context; +struct intel_ringbuffer *ringbuf; + +if (WARN_ON (!ctx)) +return ; + + ringbuf = ctx->engine[ring->id].ringbuf; +if (WARN_ON (!ringbuf)) +return ; + +intel_ring_start (ring); +intel_logical_ring_advance_and_submit (ringbuf, ctx,NULL ); +} +static int intel_lr_context_pin (struct intel_engine_cs *ring,struct intel_context *ctx) {
@@ -1121,6 +1139,8 @@ static int gen8_init_common_ring(struct intel_engine_cs *ring) ring->next_context_status_buffer =0 ;DRM_DEBUG_DRIVER ("Execlists enabled for %s\n" , ring->name); +intel_ring_stop (ring); +memset (&ring->hangcheck,0 ,sizeof (ring->hangcheck));return 0 ;
drivers/gpu/drm/i915/intel_lrc.h +1 -0
@@ -37,6 +37,7 @@/* Logical Rings */ void intel_logical_ring_stop (struct intel_engine_cs *ring); +void intel_logical_ring_start (struct intel_engine_cs *ring);void intel_logical_ring_cleanup (struct intel_engine_cs *ring);int intel_logical_rings_init (struct drm_device *dev);
drivers/gpu/drm/i915/intel_ringbuffer.c +37 -14
110 lines not on this page · raw message
drivers/gpu/drm/i915/intel_ringbuffer.h +3 -0
@@ -387,6 +387,7 @@ void intel_destroy_ringbuffer_obj(struct intel_ringbuffer *ringbuf);int intel_alloc_ringbuffer_obj (struct drm_device *dev,struct intel_ringbuffer *ringbuf); +void intel_start_ring_buffer (struct intel_engine_cs *ring);void intel_stop_ring_buffer (struct intel_engine_cs *ring);void intel_cleanup_ring_buffer (struct intel_engine_cs *ring);
@@ -408,6 +409,8 @@ int __intel_ring_space(int head, int tail, int size);void intel_ring_update_space (struct intel_ringbuffer *ringbuf);int intel_ring_space (struct intel_ringbuffer *ringbuf);bool intel_ring_stopped (struct intel_engine_cs *ring); +void intel_ring_start (struct intel_engine_cs *ring); +void intel_ring_stop (struct intel_engine_cs *ring);void __intel_ring_advance (struct intel_engine_cs *ring);int __must_check intel_ring_idle (struct intel_engine_cs *ring);
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 2/4] drm/i915: Introduce ring->start_ring()
Quoted 2 lines · Mika Kuoppala
This allows us to fill ringbuffer without engine being active. The next patch will take advantage of this.
Oh, I thought you were just worring about the reset case. Couldn't you just borrow some of my code to initialise the rings correctly? -Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 3/4 drm/i915: Reorder hw init to avoid executing with invalid context/mm state
In reply to Mika KuoppalaFrom: Chris Wilson <[email protected]>
Currently we initialise the rings, add the first context switch to the ring and execute our golden state then enable (aliasing or full) ppgtt. However, as we enable ppgtt using direct MMIO but load the PD using MI_LRI, we end up executing the context switch and golden render state with an invalid PD generating page faults. To solve this issue, first do the ppgtt PD setup, then set the default context and write the commands to run the render state into the ring, before we activate the ring. This allows us to be sure that the register state is valid before we begin execution.
This was spotted when writing the seqno/request conversion, but only with the ERROR capture did I realise that it was a necessity now.
RFC: cleanup the error handling in i915_gem_init_hw.
v2: added intel_ring_reset
v3: adapt to ring->start_ring
--- drivers/gpu/drm/i915/i915_gem.c | 29 +++++++++++++++++++++-------- 1 file changed, 21 insertions(+), 8 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +21 -8
@@ -4856,11 +4856,25 @@ cleanup_render_ring:return ret; } +static int i915_gem_init_ring_hw (struct drm_device *dev) +{ +struct drm_i915_private *dev_priv = dev->dev_private; +struct intel_engine_cs *ring; +int i, ret; + +for_each_ring (ring, dev_priv, i) { + ret = ring->init_hw (ring); +if (ret) +return ret; + } + +return 0 ; +} +int i915_gem_init_hw (struct drm_device *dev) {struct drm_i915_private *dev_priv = dev->dev_private; -struct intel_engine_cs *ring;int ret, i;if (INTEL_INFO (dev)->gen <6 && !intel_enable_gtt ())
@@ -4898,13 +4912,7 @@ i915_gem_init_hw(struct drm_device *dev)*/ init_unused_rings (dev); -for_each_ring (ring, dev_priv, i) { - ret = ring->init_hw (ring); -if (ret) - goto out; - } - -i915_gem_start_ringbuffers (dev); +i915_gem_init_ring_hw (dev);for (i =0 ; i <NUM_L3_SLICES (dev); i++)i915_gem_l3_remap (&dev_priv->ring[RCS ], i);
@@ -4915,6 +4923,9 @@ i915_gem_init_hw(struct drm_device *dev)i915_gem_cleanup_ringbuffer (dev); } +for (i =0 ; i <NUM_L3_SLICES (dev); i++) +i915_gem_l3_remap (&dev_priv->ring[RCS ], i); + ret =i915_gem_context_enable (dev_priv);if (ret && ret != -EIO ) {DRM_ERROR ("Context enable failed %d\n" , ret);
@@ -4923,6 +4934,8 @@ i915_gem_init_hw(struct drm_device *dev) goto out; } +i915_gem_start_ringbuffers (dev); + out:intel_uncore_forcewake_put (dev_priv,FORCEWAKE_ALL );return ret;
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 4/4 drm/i915: Stop rings before cleaning up on reset
In reply to Mika KuoppalaWe want to ensure that the hardware is stopped before we analyse and cleanup the request queue.
--- drivers/gpu/drm/i915/i915_gem.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/gpu/drm/i915/i915_gem.c +2 -0
@@ -2752,6 +2752,8 @@ void i915_gem_reset(struct drm_device *dev)struct intel_engine_cs *ring;int i; +i915_gem_stop_ringbuffers (dev); +/* * Before we free the objects from the requests, we need to inspect * them for finding the guilty party. As the requests only borrow
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 4/4] drm/i915: Stop rings before cleaning up on reset
Task id: 5989 -------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV 268/268 268/268 ILK 303/303 303/303 SNB 283/283 283/283 IVB -1 343/343 342/343 BYT 287/287 287/287 HSW 363/363 363/363 BDW 308/308 308/308
-------------------------------------Detailed-------------------------------------
Platform Test drm-intel-nightly Series Applied
*IVB igt_gem_userptr_blits_minor-unsync-normal PASS(2) DMESG_WARN(1)PASS(1)
Note: You need to pay more attention to line start with '*'
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx