diff mbox

[v3,06/28] drm/i915: Ensure requests stick around during waits

Message ID 1416854990-1920-7-git-send-email-John.C.Harrison@Intel.com (mailing list archive)
State New, archived
Headers show

Commit Message

John Harrison Nov. 24, 2014, 6:49 p.m. UTC
From: John Harrison <John.C.Harrison@Intel.com>

Added reference counting of the request structure around __wait_seqno() calls.
This is a precursor to updating the wait code itself to take the request rather
than a seqno. At that point, it would be a Bad Idea for a request object to be
retired and freed while the wait code is still using it.

v3:

Note that even though the mutex lock is held during a call to i915_wait_seqno(),
it is still necessary to explicitly bump the reference count. It appears that
the shrinker can asynchronously retire items even though the mutex is locked.

For: VIZ-4377
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Reviewed-by: Thomas Daniel <Thomas.Daniel@intel.com>
---
 drivers/gpu/drm/i915/i915_gem.c |   28 +++++++++++++++++++++++-----
 1 file changed, 23 insertions(+), 5 deletions(-)

Comments

John Harrison Nov. 26, 2014, 12:27 p.m. UTC | #1
NB: The v3 update was to fold in a new patch for asynchronous shrinker 
shenanigans. Unfortunately, the new patch was not actually valid at this 
point in the patch series, thus this patch now breaks the build until a 
later patch in the series fixes it up again! I'll post out an updated v4 
set with the order re-worked so that it still builds and runs at each 
step along the way...


On 24/11/2014 18:49, John.C.Harrison@Intel.com wrote:
> From: John Harrison <John.C.Harrison@Intel.com>
>
> Added reference counting of the request structure around __wait_seqno() calls.
> This is a precursor to updating the wait code itself to take the request rather
> than a seqno. At that point, it would be a Bad Idea for a request object to be
> retired and freed while the wait code is still using it.
>
> v3:
>
> Note that even though the mutex lock is held during a call to i915_wait_seqno(),
> it is still necessary to explicitly bump the reference count. It appears that
> the shrinker can asynchronously retire items even though the mutex is locked.
>
> For: VIZ-4377
> Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
> Reviewed-by: Thomas Daniel <Thomas.Daniel@intel.com>
> ---
>   drivers/gpu/drm/i915/i915_gem.c |   28 +++++++++++++++++++++++-----
>   1 file changed, 23 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 9ac9157..71303a1 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1333,8 +1333,11 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno)
>   		return ret;
>   
>   	reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
> -	return __i915_wait_seqno(ring, seqno, reset_counter, interruptible,
> -				 NULL, NULL);
> +	i915_gem_request_reference(req);
> +	ret = __i915_wait_seqno(ring, seqno, reset_counter, interruptible,
> +				NULL, NULL);
> +	i915_gem_request_unreference(req);
> +	return ret;
>   }
>   
>   static int
> @@ -1417,10 +1420,12 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
>   		return ret;
>   
>   	reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
> +	i915_gem_request_reference(req);
>   	mutex_unlock(&dev->struct_mutex);
>   	ret = __i915_wait_seqno(ring, seqno, reset_counter, true, NULL,
>   				file_priv);
>   	mutex_lock(&dev->struct_mutex);
> +	i915_gem_request_unreference(req);
>   	if (ret)
>   		return ret;
>   
> @@ -2920,6 +2925,7 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
>   	struct drm_i915_private *dev_priv = dev->dev_private;
>   	struct drm_i915_gem_wait *args = data;
>   	struct drm_i915_gem_object *obj;
> +	struct drm_i915_gem_request *req;
>   	struct intel_engine_cs *ring = NULL;
>   	unsigned reset_counter;
>   	u32 seqno = 0;
> @@ -2946,7 +2952,8 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
>   	if (!obj->active || !obj->last_read_req)
>   		goto out;
>   
> -	seqno = i915_gem_request_get_seqno(obj->last_read_req);
> +	req = obj->last_read_req;
> +	seqno = i915_gem_request_get_seqno(req);
>   	WARN_ON(seqno == 0);
>   	ring = obj->ring;
>   
> @@ -2960,10 +2967,15 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
>   
>   	drm_gem_object_unreference(&obj->base);
>   	reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
> +	i915_gem_request_reference(req);
>   	mutex_unlock(&dev->struct_mutex);
>   
> -	return __i915_wait_seqno(ring, seqno, reset_counter, true,
> -				 &args->timeout_ns, file->driver_priv);
> +	ret = __i915_wait_seqno(ring, seqno, reset_counter, true, &args->timeout_ns,
> +			   file->driver_priv);
> +	mutex_lock(&dev->struct_mutex);
> +	i915_gem_request_unreference(req);
> +	mutex_unlock(&dev->struct_mutex);
> +	return ret;
>   
>   out:
>   	drm_gem_object_unreference(&obj->base);
> @@ -4122,6 +4134,8 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
>   		target = request;
>   	}
>   	reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
> +	if (target)
> +		i915_gem_request_reference(target);
>   	spin_unlock(&file_priv->mm.lock);
>   
>   	if (target == NULL)
> @@ -4133,6 +4147,10 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
>   	if (ret == 0)
>   		queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work, 0);
>   
> +	mutex_lock(&dev->struct_mutex);
> +	i915_gem_request_unreference(target);
> +	mutex_unlock(&dev->struct_mutex);
> +
>   	return ret;
>   }
>
Daniel Vetter Nov. 26, 2014, 1:14 p.m. UTC | #2
On Wed, Nov 26, 2014 at 12:27:07PM +0000, John Harrison wrote:
> NB: The v3 update was to fold in a new patch for asynchronous shrinker
> shenanigans. Unfortunately, the new patch was not actually valid at this
> point in the patch series, thus this patch now breaks the build until a
> later patch in the series fixes it up again! I'll post out an updated v4 set
> with the order re-worked so that it still builds and runs at each step along
> the way...

Please don't, I already have your series partially merged. In general
never repost the entire series while discussions are ongoing, it makes a
big mess out of the threads on the m-l.
-Daniel
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 9ac9157..71303a1 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1333,8 +1333,11 @@  i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno)
 		return ret;
 
 	reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
-	return __i915_wait_seqno(ring, seqno, reset_counter, interruptible,
-				 NULL, NULL);
+	i915_gem_request_reference(req);
+	ret = __i915_wait_seqno(ring, seqno, reset_counter, interruptible,
+				NULL, NULL);
+	i915_gem_request_unreference(req);
+	return ret;
 }
 
 static int
@@ -1417,10 +1420,12 @@  i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
 		return ret;
 
 	reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
+	i915_gem_request_reference(req);
 	mutex_unlock(&dev->struct_mutex);
 	ret = __i915_wait_seqno(ring, seqno, reset_counter, true, NULL,
 				file_priv);
 	mutex_lock(&dev->struct_mutex);
+	i915_gem_request_unreference(req);
 	if (ret)
 		return ret;
 
@@ -2920,6 +2925,7 @@  i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_gem_wait *args = data;
 	struct drm_i915_gem_object *obj;
+	struct drm_i915_gem_request *req;
 	struct intel_engine_cs *ring = NULL;
 	unsigned reset_counter;
 	u32 seqno = 0;
@@ -2946,7 +2952,8 @@  i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 	if (!obj->active || !obj->last_read_req)
 		goto out;
 
-	seqno = i915_gem_request_get_seqno(obj->last_read_req);
+	req = obj->last_read_req;
+	seqno = i915_gem_request_get_seqno(req);
 	WARN_ON(seqno == 0);
 	ring = obj->ring;
 
@@ -2960,10 +2967,15 @@  i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 
 	drm_gem_object_unreference(&obj->base);
 	reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
+	i915_gem_request_reference(req);
 	mutex_unlock(&dev->struct_mutex);
 
-	return __i915_wait_seqno(ring, seqno, reset_counter, true,
-				 &args->timeout_ns, file->driver_priv);
+	ret = __i915_wait_seqno(ring, seqno, reset_counter, true, &args->timeout_ns,
+			   file->driver_priv);
+	mutex_lock(&dev->struct_mutex);
+	i915_gem_request_unreference(req);
+	mutex_unlock(&dev->struct_mutex);
+	return ret;
 
 out:
 	drm_gem_object_unreference(&obj->base);
@@ -4122,6 +4134,8 @@  i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
 		target = request;
 	}
 	reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
+	if (target)
+		i915_gem_request_reference(target);
 	spin_unlock(&file_priv->mm.lock);
 
 	if (target == NULL)
@@ -4133,6 +4147,10 @@  i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
 	if (ret == 0)
 		queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work, 0);
 
+	mutex_lock(&dev->struct_mutex);
+	i915_gem_request_unreference(target);
+	mutex_unlock(&dev->struct_mutex);
+
 	return ret;
 }