Re: [PATCH 40/50] drm/i915/bdw: Handle context switch events

[Daniel Vetter <daniel@ffwll.ch>]

On Fri, May 09, 2014 at 01:09:10PM +0100, oscar.mateo@intel.com wrote:
> From: Thomas Daniel <thomas.daniel@intel.com>
> 
> Handle all context status events in the context status buffer on every
> context switch interrupt. We only remove work from the execlist queue
> after a context status buffer reports that it has completed and we only
> attempt to schedule new contexts on interrupt when a previously submitted
> context completes (unless no contexts are queued, which means the GPU is
> free).
> 
> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com>
> 
> v2: Unreferencing the context when we are freeing the request might free
> the backing bo, which requires the struct_mutex to be grabbed, so defer
> unreferencing and freeing to a bottom half.
> 
> v3:
> - Ack the interrupt inmediately, before trying to handle it (fix for
> missing interrupts by Bob Beckett <robert.beckett@intel.com>).

This interrupt handling change is interesting since it might explain our
irq handling woes on gen5+ with the two-level GT interrupt handling
scheme. Can you please roll this out as a prep patch for all the existing
gt interrupt sources we handle already for gen5+?

Thanks, Daniel

> - Update the Context Status Buffer Read Pointer, just in case (spotted
> by Damien Lespiau).
> 
> Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
> ---
>  drivers/gpu/drm/i915/i915_drv.h         |   3 +
>  drivers/gpu/drm/i915/i915_irq.c         |  38 +++++++-----
>  drivers/gpu/drm/i915/intel_lrc.c        | 102 +++++++++++++++++++++++++++++++-
>  drivers/gpu/drm/i915/intel_ringbuffer.c |   1 +
>  drivers/gpu/drm/i915/intel_ringbuffer.h |   1 +
>  5 files changed, 129 insertions(+), 16 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index f2aae6a..07b8bdc 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -1748,6 +1748,8 @@ struct drm_i915_gem_request {
>  
>  	/** execlist queue entry for this request */
>  	struct list_head execlist_link;
> +	/** Struct to handle this request in the bottom half of an interrupt */
> +	struct work_struct work;
>  };
>  
>  struct drm_i915_file_private {
> @@ -2449,6 +2451,7 @@ static inline u32 intel_get_lr_contextid(struct drm_i915_gem_object *ctx_obj)
>  int gen8_switch_context_queue(struct intel_engine *ring,
>  			      struct i915_hw_context *to,
>  			      u32 tail);
> +void gen8_handle_context_events(struct intel_engine *ring);
>  
>  /* i915_gem_evict.c */
>  int __must_check i915_gem_evict_something(struct drm_device *dev,
> diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
> index a28cf6c..fbffead 100644
> --- a/drivers/gpu/drm/i915/i915_irq.c
> +++ b/drivers/gpu/drm/i915/i915_irq.c
> @@ -1300,6 +1300,7 @@ static irqreturn_t gen8_gt_irq_handler(struct drm_device *dev,
>  				       struct drm_i915_private *dev_priv,
>  				       u32 master_ctl)
>  {
> +	struct intel_engine *ring;
>  	u32 rcs, bcs, vcs, vecs;
>  	uint32_t tmp = 0;
>  	irqreturn_t ret = IRQ_NONE;
> @@ -1307,16 +1308,22 @@ static irqreturn_t gen8_gt_irq_handler(struct drm_device *dev,
>  	if (master_ctl & (GEN8_GT_RCS_IRQ | GEN8_GT_BCS_IRQ)) {
>  		tmp = I915_READ(GEN8_GT_IIR(0));
>  		if (tmp) {
> +			I915_WRITE(GEN8_GT_IIR(0), tmp);
>  			ret = IRQ_HANDLED;
> +
>  			rcs = tmp >> GEN8_RCS_IRQ_SHIFT;
> -			bcs = tmp >> GEN8_BCS_IRQ_SHIFT;
> +			ring = &dev_priv->ring[RCS];
>  			if (rcs & GT_RENDER_USER_INTERRUPT)
> -				notify_ring(dev, &dev_priv->ring[RCS]);
> +				notify_ring(dev, ring);
> +			if (rcs & GEN8_GT_CONTEXT_SWITCH_INTERRUPT)
> +				gen8_handle_context_events(ring);
> +
> +			bcs = tmp >> GEN8_BCS_IRQ_SHIFT;
> +			ring = &dev_priv->ring[BCS];
>  			if (bcs & GT_RENDER_USER_INTERRUPT)
> -				notify_ring(dev, &dev_priv->ring[BCS]);
> -			if ((rcs | bcs) & GEN8_GT_CONTEXT_SWITCH_INTERRUPT)
> -			     DRM_DEBUG_DRIVER("TODO: Context switch\n");
> -			I915_WRITE(GEN8_GT_IIR(0), tmp);
> +				notify_ring(dev, ring);
> +			if (bcs & GEN8_GT_CONTEXT_SWITCH_INTERRUPT)
> +				gen8_handle_context_events(ring);
>  		} else
>  			DRM_ERROR("The master control interrupt lied (GT0)!\n");
>  	}
> @@ -1324,18 +1331,20 @@ static irqreturn_t gen8_gt_irq_handler(struct drm_device *dev,
>  	if (master_ctl & (GEN8_GT_VCS1_IRQ | GEN8_GT_VCS2_IRQ)) {
>  		tmp = I915_READ(GEN8_GT_IIR(1));
>  		if (tmp) {
> +			I915_WRITE(GEN8_GT_IIR(1), tmp);
>  			ret = IRQ_HANDLED;
>  			vcs = tmp >> GEN8_VCS1_IRQ_SHIFT;
> +			ring = &dev_priv->ring[VCS];
>  			if (vcs & GT_RENDER_USER_INTERRUPT)
> -				notify_ring(dev, &dev_priv->ring[VCS]);
> +				notify_ring(dev, ring);
>  			if (vcs & GEN8_GT_CONTEXT_SWITCH_INTERRUPT)
> -			     DRM_DEBUG_DRIVER("TODO: Context switch\n");
> +			     gen8_handle_context_events(ring);
>  			vcs = tmp >> GEN8_VCS2_IRQ_SHIFT;
> +			ring = &dev_priv->ring[VCS2];
>  			if (vcs & GT_RENDER_USER_INTERRUPT)
> -				notify_ring(dev, &dev_priv->ring[VCS2]);
> +				notify_ring(dev, ring);
>  			if (vcs & GEN8_GT_CONTEXT_SWITCH_INTERRUPT)
> -			     DRM_DEBUG_DRIVER("TODO: Context switch\n");
> -			I915_WRITE(GEN8_GT_IIR(1), tmp);
> +			     gen8_handle_context_events(ring);
>  		} else
>  			DRM_ERROR("The master control interrupt lied (GT1)!\n");
>  	}
> @@ -1343,13 +1352,14 @@ static irqreturn_t gen8_gt_irq_handler(struct drm_device *dev,
>  	if (master_ctl & GEN8_GT_VECS_IRQ) {
>  		tmp = I915_READ(GEN8_GT_IIR(3));
>  		if (tmp) {
> +			I915_WRITE(GEN8_GT_IIR(3), tmp);
>  			ret = IRQ_HANDLED;
>  			vecs = tmp >> GEN8_VECS_IRQ_SHIFT;
> +			ring = &dev_priv->ring[VECS];
>  			if (vecs & GT_RENDER_USER_INTERRUPT)
> -				notify_ring(dev, &dev_priv->ring[VECS]);
> +				notify_ring(dev, ring);
>  			if (vecs & GEN8_GT_CONTEXT_SWITCH_INTERRUPT)
> -			     DRM_DEBUG_DRIVER("TODO: Context switch\n");
> -			I915_WRITE(GEN8_GT_IIR(3), tmp);
> +				gen8_handle_context_events(ring);
>  		} else
>  			DRM_ERROR("The master control interrupt lied (GT3)!\n");
>  	}
> diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
> index 6da7db9..1ff493a 100644
> --- a/drivers/gpu/drm/i915/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/intel_lrc.c
> @@ -49,6 +49,22 @@
>  #define RING_ELSP(ring)			((ring)->mmio_base+0x230)
>  #define RING_EXECLIST_STATUS(ring)	((ring)->mmio_base+0x234)
>  #define RING_CONTEXT_CONTROL(ring)	((ring)->mmio_base+0x244)
> +#define RING_CONTEXT_STATUS_BUF(ring)	((ring)->mmio_base+0x370)
> +#define RING_CONTEXT_STATUS_PTR(ring)	((ring)->mmio_base+0x3a0)
> +
> +#define RING_EXECLIST_QFULL		(1 << 0x2)
> +#define RING_EXECLIST1_VALID		(1 << 0x3)
> +#define RING_EXECLIST0_VALID		(1 << 0x4)
> +#define RING_EXECLIST_ACTIVE_STATUS	(3 << 0xE)
> +#define RING_EXECLIST1_ACTIVE		(1 << 0x11)
> +#define RING_EXECLIST0_ACTIVE		(1 << 0x12)
> +
> +#define GEN8_CTX_STATUS_IDLE_ACTIVE	(1 << 0)
> +#define GEN8_CTX_STATUS_PREEMPTED	(1 << 1)
> +#define GEN8_CTX_STATUS_ELEMENT_SWITCH	(1 << 2)
> +#define GEN8_CTX_STATUS_ACTIVE_IDLE	(1 << 3)
> +#define GEN8_CTX_STATUS_COMPLETE	(1 << 4)
> +#define GEN8_CTX_STATUS_LITE_RESTORE	(1 << 15)
>  
>  #define CTX_LRI_HEADER_0		0x01
>  #define CTX_CONTEXT_CONTROL		0x02
> @@ -203,6 +219,9 @@ static void gen8_switch_context_unqueue(struct intel_engine *ring)
>  {
>  	struct drm_i915_gem_request *req0 = NULL, *req1 = NULL;
>  	struct drm_i915_gem_request *cursor = NULL, *tmp = NULL;
> +	struct drm_i915_private *dev_priv = ring->dev->dev_private;
> +
> +	assert_spin_locked(&ring->execlist_lock);
>  
>  	if (list_empty(&ring->execlist_queue))
>  		return;
> @@ -215,8 +234,7 @@ static void gen8_switch_context_unqueue(struct intel_engine *ring)
>  			/* Same ctx: ignore first request, as second request
>  			 * will update tail past first request's workload */
>  			list_del(&req0->execlist_link);
> -			i915_gem_context_unreference(req0->ctx);
> -			kfree(req0);
> +			queue_work(dev_priv->wq, &req0->work);
>  			req0 = cursor;
>  		} else {
>  			req1 = cursor;
> @@ -228,6 +246,85 @@ static void gen8_switch_context_unqueue(struct intel_engine *ring)
>  			req1? req1->ctx : NULL, req1? req1->tail : 0));
>  }
>  
> +static bool check_remove_request(struct intel_engine *ring, u32 request_id)
> +{
> +	struct drm_i915_private *dev_priv = ring->dev->dev_private;
> +	struct drm_i915_gem_request *head_req;
> +
> +	assert_spin_locked(&ring->execlist_lock);
> +
> +	head_req = list_first_entry_or_null(&ring->execlist_queue,
> +			struct drm_i915_gem_request, execlist_link);
> +	if (head_req != NULL) {
> +		struct drm_i915_gem_object *ctx_obj =
> +				head_req->ctx->engine[ring->id].obj;
> +		if (intel_get_lr_contextid(ctx_obj) == request_id) {
> +			list_del(&head_req->execlist_link);
> +			queue_work(dev_priv->wq, &head_req->work);
> +			return true;
> +		}
> +	}
> +
> +	return false;
> +}
> +
> +void gen8_handle_context_events(struct intel_engine *ring)
> +{
> +	struct drm_i915_private *dev_priv = ring->dev->dev_private;
> +	u32 status_pointer;
> +	u8 read_pointer;
> +	u8 write_pointer;
> +	u32 status;
> +	u32 status_id;
> +	u32 submit_contexts = 0;
> +
> +	status_pointer = I915_READ(RING_CONTEXT_STATUS_PTR(ring));
> +
> +	read_pointer = ring->next_context_status_buffer;
> +	write_pointer = status_pointer & 0x07;
> +	if (read_pointer > write_pointer)
> +		write_pointer += 6;
> +
> +	spin_lock(&ring->execlist_lock);
> +
> +	while (read_pointer < write_pointer) {
> +		read_pointer++;
> +		status = I915_READ(RING_CONTEXT_STATUS_BUF(ring) +
> +				(read_pointer % 6) * 8);
> +		status_id = I915_READ(RING_CONTEXT_STATUS_BUF(ring) +
> +				(read_pointer % 6) * 8 + 4);
> +
> +		if (status & GEN8_CTX_STATUS_COMPLETE) {
> +			if (check_remove_request(ring, status_id))
> +				submit_contexts++;
> +		}
> +	}
> +
> +	if (submit_contexts != 0)
> +		gen8_switch_context_unqueue(ring);
> +
> +	spin_unlock(&ring->execlist_lock);
> +
> +	WARN(submit_contexts > 2, "More than two context complete events?\n");
> +	ring->next_context_status_buffer = write_pointer % 6;
> +
> +	I915_WRITE(RING_CONTEXT_STATUS_PTR(ring),
> +			((u32)ring->next_context_status_buffer & 0x07) << 8);
> +}
> +
> +static void free_request_task(struct work_struct *work)
> +{
> +	struct drm_i915_gem_request *req =
> +			container_of(work, struct drm_i915_gem_request, work);
> +	struct drm_device *dev = req->ring->dev;
> +
> +	mutex_lock(&dev->struct_mutex);
> +	i915_gem_context_unreference(req->ctx);
> +	mutex_unlock(&dev->struct_mutex);
> +
> +	kfree(req);
> +}
> +
>  int gen8_switch_context_queue(struct intel_engine *ring,
>  			      struct i915_hw_context *to,
>  			      u32 tail)
> @@ -243,6 +340,7 @@ int gen8_switch_context_queue(struct intel_engine *ring,
>  	req->ctx = to;
>  	i915_gem_context_reference(req->ctx);
>  	req->tail = tail;
> +	INIT_WORK(&req->work, free_request_task);
>  
>  	spin_lock_irqsave(&ring->execlist_lock, flags);
>  
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
> index 35ced7c..9cd6ee8 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.c
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
> @@ -1573,6 +1573,7 @@ static int intel_init_ring(struct drm_device *dev,
>  		if (ring->status_page.page_addr == NULL)
>  			return -ENOMEM;
>  		ring->status_page.obj = obj;
> +		ring->next_context_status_buffer = 0;
>  	} else if (I915_NEED_GFX_HWS(dev)) {
>  		ret = init_status_page(ring);
>  		if (ret)
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
> index daf91de..f3ae547 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.h
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
> @@ -178,6 +178,7 @@ struct intel_engine {
>  
>  	spinlock_t execlist_lock;
>  	struct list_head execlist_queue;
> +	u8 next_context_status_buffer;
>  
>  	struct i915_hw_context *default_context;
>  	struct i915_hw_context *last_context;
> -- 
> 1.9.0
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch