On 29/06/2016 15:51, Chris Wilson wrote:
Since commit 2ed53a94d8cb ("drm/i915: On GPU reset, set the HWS
breadcrumb to the last seqno") once a hang is completed, the seqno is
advanced past all current requests. With this we know that if we wake up
on waiting for a request, if a hang has occurred and reset completed,
our request will be considered complete (i.e.
i915_gem_request_completed() returns true). Therefore we only need to
worry about the situation where a hang has occurred, but not yet reset,
where we may need to release our struct_mutex. Since we don't need to
detect the competed reset using the global gpu_error->reset_counter
s/competed/completed
anymore, we do not need to track the reset_counter epoch inside the
request.
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Cc: Arun Siluvery <arun.siluvery@xxxxxxxxxxxxxxx>
Cc: Mika Kuoppala <mika.kuoppala@xxxxxxxxx>
---
drivers/gpu/drm/i915/i915_drv.h | 1 -
drivers/gpu/drm/i915/i915_gem.c | 16 ++++++++--------
2 files changed, 8 insertions(+), 9 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index def011811421..485ab1148181 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2368,7 +2368,6 @@ struct drm_i915_gem_request {
/** On Which ring this request was generated */
struct drm_i915_private *i915;
struct intel_engine_cs *engine;
- unsigned reset_counter;
/** GEM sequence number associated with the previous request,
* when the HWS breadcrumb is equal to this the GPU is processing
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 51191b879747..1d9878258103 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1506,12 +1506,13 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
/* We need to check whether any gpu reset happened in between
* the request being submitted and now. If a reset has occurred,
- * the request is effectively complete (we either are in the
- * process of or have discarded the rendering and completely
- * reset the GPU. The results of the request are lost and we
- * are free to continue on with the original operation.
+ * the seqno will have been advance past ours and our request
+ * is complete. If we are in the process of handling a reset,
+ * the request is effectively complete as the rendering will
+ * be discarded, but we need to return in order to drop the
+ * struct_mutex.
*/
- if (req->reset_counter != i915_reset_counter(&dev_priv->gpu_error)) {
+ if (i915_reset_in_progress(&dev_priv->gpu_error)) {
ret = 0;
break;
}
@@ -1685,7 +1686,7 @@ i915_wait_request(struct drm_i915_gem_request *req)
return ret;
/* If the GPU hung, we want to keep the requests to find the guilty. */
- if (req->reset_counter == i915_reset_counter(&dev_priv->gpu_error))
+ if (!i915_reset_in_progress(&dev_priv->gpu_error))
__i915_gem_request_retire__upto(req);
return 0;
@@ -1746,7 +1747,7 @@ i915_gem_object_retire_request(struct drm_i915_gem_object *obj,
else if (obj->last_write_req == req)
i915_gem_object_retire__write(obj);
- if (req->reset_counter == i915_reset_counter(&req->i915->gpu_error))
+ if (!i915_reset_in_progress(&req->i915->gpu_error))
__i915_gem_request_retire__upto(req);
}
@@ -3021,7 +3022,6 @@ __i915_gem_request_alloc(struct intel_engine_cs *engine,
kref_init(&req->ref);
req->i915 = dev_priv;
req->engine = engine;
- req->reset_counter = reset_counter;
req->ctx = ctx;
i915_gem_context_reference(req->ctx);
this looks good to me,
Reviewed-by: Arun Siluvery <arun.siluvery@xxxxxxxxxxxxxxx>
regards
Arun
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
