drm/i915/guc: Move GuC wq_reserve_space to ring_reserve_space

Commit Message

[e-mail] Dec. 9, 2015, 1:04 a.m. UTC

From: Alex Dai <[e-mail]>

Split GuC work queue space reserve and submission and move the space
reserve to where ring space is reserved. The reason is that failure
in intel_logical_ring_advance_and_submit won't be handled. In the
case timeout happens, driver can return early in order to handle the
error.
---
 drivers/gpu/drm/i915/i915_guc_submission.c | 21 +++++++++------------
 drivers/gpu/drm/i915/intel_guc.h           |  1 +
 drivers/gpu/drm/i915/intel_lrc.c           | 10 +++++++++-
 3 files changed, 19 insertions(+), 13 deletions(-)

Comments

Chris Wilson Dec. 9, 2015, 9:05 a.m. UTC | #1

On Tue, Dec 08, 2015 at 05:04:50PM -0800, [e-mail] wrote:
> From: Alex Dai <[e-mail]>
> 
> Split GuC work queue space reserve and submission and move the space
> reserve to where ring space is reserved. The reason is that failure
> in intel_logical_ring_advance_and_submit won't be handled. In the
> case timeout happens, driver can return early in order to handle the
> error.

Not here. You want the request_alloc_extras callback.
-Chris

[e-mail] Dec. 9, 2015, 5:08 p.m. UTC | #2

On 12/09/2015 01:05 AM, Chris Wilson wrote:
> On Tue, Dec 08, 2015 at 05:04:50PM -0800, [e-mail] wrote:
> > From: Alex Dai <[e-mail]>
> >
> > Split GuC work queue space reserve and submission and move the space
> > reserve to where ring space is reserved. The reason is that failure
> > in intel_logical_ring_advance_and_submit won't be handled. In the
> > case timeout happens, driver can return early in order to handle the
> > error.
>
> Not here. You want the request_alloc_extras callback.

OK. That is even before touching the ring.

Thanks,
Alex

Patch

diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c
index 8a71458..264fdf7 100644
--- a/drivers/gpu/drm/i915/i915_guc_submission.c
+++ b/drivers/gpu/drm/i915/i915_guc_submission.c
@@ -472,25 +472,21 @@  static void guc_fini_ctx_desc(struct intel_guc *guc,
 			     sizeof(desc) * client->ctx_index);
 }
 
-/* Get valid workqueue item and return it back to offset */
-static int guc_get_workqueue_space(struct i915_guc_client *gc, u32 *offset)
+int i915_guc_wq_reserve_space(struct i915_guc_client *gc)
 {
 	struct guc_process_desc *desc;
 	void *base;
 	u32 size = sizeof(struct guc_wq_item);
 	int ret = -ETIMEDOUT, timeout_counter = 200;
 
+	if (!gc)
+		return 0;
+
 	base = kmap_atomic(i915_gem_object_get_page(gc->client_obj, 0));
 	desc = base + gc->proc_desc_offset;
 
 	while (timeout_counter-- > 0) {
 		if (CIRC_SPACE(gc->wq_tail, desc->head, gc->wq_size) >= size) {
-			*offset = gc->wq_tail;
-
-			/* advance the tail for next workqueue item */
-			gc->wq_tail += size;
-			gc->wq_tail &= gc->wq_size - 1;
-
 			/* this will break the loop */
 			timeout_counter = 0;
 			ret = 0;
@@ -512,11 +508,12 @@  static int guc_add_workqueue_item(struct i915_guc_client *gc,
 	struct guc_wq_item *wqi;
 	void *base;
 	u32 tail, wq_len, wq_off = 0;
-	int ret;
 
-	ret = guc_get_workqueue_space(gc, &wq_off);
-	if (ret)
-		return ret;
+	wq_off = gc->wq_tail;
+
+	/* advance the tail for next workqueue item */
+	gc->wq_tail += sizeof(struct guc_wq_item);
+	gc->wq_tail &= gc->wq_size - 1;
 
 	/* For now workqueue item is 4 DWs; workqueue buffer is 2 pages. So we
 	 * should not have the case where structure wqi is across page, neither
diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
index 0e048bf..59c8e21 100644
--- a/drivers/gpu/drm/i915/intel_guc.h
+++ b/drivers/gpu/drm/i915/intel_guc.h
@@ -123,5 +123,6 @@  int i915_guc_submit(struct i915_guc_client *client,
 		    struct drm_i915_gem_request *rq);
 void i915_guc_submission_disable(struct drm_device *dev);
 void i915_guc_submission_fini(struct drm_device *dev);
+int i915_guc_wq_reserve_space(struct i915_guc_client *client);
 
 #endif
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index f96fb51..25cbeab 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -844,6 +844,8 @@  int intel_logical_ring_begin(struct drm_i915_gem_request *req, int num_dwords)
 
 int intel_logical_ring_reserve_space(struct drm_i915_gem_request *request)
 {
+	int ret;
+
 	/*
 	 * The first call merely notes the reserve request and is common for
 	 * all back ends. The subsequent localised _begin() call actually
@@ -854,7 +856,13 @@  int intel_logical_ring_reserve_space(struct drm_i915_gem_request *request)
 	 */
 	intel_ring_reserved_space_reserve(request->ringbuf, MIN_SPACE_FOR_ADD_REQUEST);
 
-	return intel_logical_ring_begin(request, 0);
+	ret = intel_logical_ring_begin(request, 0);
+	if (ret)
+		return ret;
+
+	ret = i915_guc_wq_reserve_space(request->i915->guc.execbuf_client);
+
+	return ret;
 }
 
 /**