[Intel-gfx] [PATCH 2/3] drm/i915: Store the reset counter when constructing a request
Daniel Vetter
daniel at ffwll.ch
Thu Dec 3 00:59:59 PST 2015
On Tue, Dec 01, 2015 at 11:05:34AM +0000, Chris Wilson wrote:
> As the request is only valid during the same global reset epoch, we can
> record the current reset_counter when constructing the request and reuse
> it when waiting upon that request in future. This removes a very hairy
> atomic check serialised by the struct_mutex at the time of waiting and
> allows us to transfer those waits to a central dispatcher for all
> waiters and all requests.
Please add
"This also allows us to clean up the check_wedge handling a bit, since we
don't have to be super-careful any more with running into a zombie gpu.
There's now only two places we check for a wedge gpu: request_alloc (to
catch command submission) and in the inner loop of wait_request (no point
bailing out if we don't have to wait anyway)."
With that added Reviewed-by: Daniel Vetter <daniel.vetter at ffwll.ch>
>
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_drv.h | 2 +-
> drivers/gpu/drm/i915/i915_gem.c | 40 +++++++++++----------------------
> drivers/gpu/drm/i915/intel_display.c | 7 +-----
> drivers/gpu/drm/i915/intel_lrc.c | 7 ------
> drivers/gpu/drm/i915/intel_ringbuffer.c | 6 -----
> 5 files changed, 15 insertions(+), 47 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 2a4cfa06c28d..ee7677343056 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -2181,6 +2181,7 @@ struct drm_i915_gem_request {
> /** On Which ring this request was generated */
> struct drm_i915_private *i915;
> struct intel_engine_cs *ring;
> + unsigned reset_counter;
>
> /** GEM sequence number associated with the previous request,
> * when the HWS breadcrumb is equal to this the GPU is processing
> @@ -3049,7 +3050,6 @@ void __i915_add_request(struct drm_i915_gem_request *req,
> #define i915_add_request_no_flush(req) \
> __i915_add_request(req, NULL, false)
> int __i915_wait_request(struct drm_i915_gem_request *req,
> - unsigned reset_counter,
> bool interruptible,
> s64 *timeout,
> struct intel_rps_client *rps);
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index bff245de8ade..68d4bab93fd8 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1223,7 +1223,6 @@ static int __i915_spin_request(struct drm_i915_gem_request *req, int state)
> /**
> * __i915_wait_request - wait until execution of request has finished
> * @req: duh!
> - * @reset_counter: reset sequence associated with the given request
> * @interruptible: do an interruptible wait (normally yes)
> * @timeout: in - how long to wait (NULL forever); out - how much time remaining
> *
> @@ -1238,7 +1237,6 @@ static int __i915_spin_request(struct drm_i915_gem_request *req, int state)
> * errno with remaining time filled in timeout argument.
> */
> int __i915_wait_request(struct drm_i915_gem_request *req,
> - unsigned reset_counter,
> bool interruptible,
> s64 *timeout,
> struct intel_rps_client *rps)
> @@ -1297,7 +1295,7 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
>
> /* We need to check whether any gpu reset happened in between
> * the caller grabbing the seqno and now ... */
> - if (reset_counter != i915_reset_counter(&dev_priv->gpu_error)) {
> + if (req->reset_counter != i915_reset_counter(&dev_priv->gpu_error)) {
> /* ... but upgrade the -EAGAIN to an -EIO if the gpu
> * is truely gone. */
> ret = i915_gem_check_wedge(&dev_priv->gpu_error, interruptible);
> @@ -1470,13 +1468,7 @@ i915_wait_request(struct drm_i915_gem_request *req)
>
> BUG_ON(!mutex_is_locked(&dev->struct_mutex));
>
> - ret = i915_gem_check_wedge(&dev_priv->gpu_error, interruptible);
> - if (ret)
> - return ret;
> -
> - ret = __i915_wait_request(req,
> - i915_reset_counter(&dev_priv->gpu_error),
> - interruptible, NULL, NULL);
> + ret = __i915_wait_request(req, interruptible, NULL, NULL);
> if (ret)
> return ret;
>
> @@ -1551,7 +1543,6 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
> struct drm_device *dev = obj->base.dev;
> struct drm_i915_private *dev_priv = dev->dev_private;
> struct drm_i915_gem_request *requests[I915_NUM_RINGS];
> - unsigned reset_counter;
> int ret, i, n = 0;
>
> BUG_ON(!mutex_is_locked(&dev->struct_mutex));
> @@ -1560,12 +1551,6 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
> if (!obj->active)
> return 0;
>
> - ret = i915_gem_check_wedge(&dev_priv->gpu_error, true);
> - if (ret)
> - return ret;
> -
> - reset_counter = i915_reset_counter(&dev_priv->gpu_error);
> -
> if (readonly) {
> struct drm_i915_gem_request *req;
>
> @@ -1587,9 +1572,9 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
> }
>
> mutex_unlock(&dev->struct_mutex);
> + ret = 0;
> for (i = 0; ret == 0 && i < n; i++)
> - ret = __i915_wait_request(requests[i], reset_counter, true,
> - NULL, rps);
> + ret = __i915_wait_request(requests[i], true, NULL, rps);
> mutex_lock(&dev->struct_mutex);
>
> for (i = 0; i < n; i++) {
> @@ -2693,6 +2678,7 @@ int i915_gem_request_alloc(struct intel_engine_cs *ring,
> struct drm_i915_gem_request **req_out)
> {
> struct drm_i915_private *dev_priv = to_i915(ring->dev);
> + unsigned reset_counter = i915_reset_counter(&dev_priv->gpu_error);
> struct drm_i915_gem_request *req;
> int ret;
>
> @@ -2701,6 +2687,11 @@ int i915_gem_request_alloc(struct intel_engine_cs *ring,
>
> *req_out = NULL;
>
> + ret = i915_gem_check_wedge(&dev_priv->gpu_error,
> + dev_priv->mm.interruptible);
> + if (ret)
> + return ret;
> +
> req = kmem_cache_zalloc(dev_priv->requests, GFP_KERNEL);
> if (req == NULL)
> return -ENOMEM;
> @@ -2712,6 +2703,7 @@ int i915_gem_request_alloc(struct intel_engine_cs *ring,
> kref_init(&req->ref);
> req->i915 = dev_priv;
> req->ring = ring;
> + req->reset_counter = reset_counter;
> req->ctx = ctx;
> i915_gem_context_reference(req->ctx);
>
> @@ -3072,11 +3064,9 @@ retire:
> int
> 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[I915_NUM_RINGS];
> - unsigned reset_counter;
> int i, n = 0;
> int ret;
>
> @@ -3110,7 +3100,6 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
> }
>
> drm_gem_object_unreference(&obj->base);
> - reset_counter = i915_reset_counter(&dev_priv->gpu_error);
>
> for (i = 0; i < I915_NUM_RINGS; i++) {
> if (obj->last_read_req[i] == NULL)
> @@ -3123,7 +3112,7 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
>
> for (i = 0; i < n; i++) {
> if (ret == 0)
> - ret = __i915_wait_request(req[i], reset_counter, true,
> + ret = __i915_wait_request(req[i], true,
> args->timeout_ns > 0 ? &args->timeout_ns : NULL,
> file->driver_priv);
> i915_gem_request_unreference__unlocked(req[i]);
> @@ -3155,7 +3144,6 @@ __i915_gem_object_sync(struct drm_i915_gem_object *obj,
> if (!i915_semaphore_is_enabled(obj->base.dev)) {
> struct drm_i915_private *i915 = to_i915(obj->base.dev);
> ret = __i915_wait_request(from_req,
> - i915_reset_counter(&i915->gpu_error),
> i915->mm.interruptible,
> NULL,
> &i915->rps.semaphores);
> @@ -4084,7 +4072,6 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
> struct drm_i915_file_private *file_priv = file->driver_priv;
> unsigned long recent_enough = jiffies - DRM_I915_THROTTLE_JIFFIES;
> struct drm_i915_gem_request *request, *target = NULL;
> - unsigned reset_counter;
> int ret;
>
> ret = i915_gem_wait_for_error(&dev_priv->gpu_error);
> @@ -4109,7 +4096,6 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
>
> target = request;
> }
> - reset_counter = i915_reset_counter(&dev_priv->gpu_error);
> if (target)
> i915_gem_request_reference(target);
> spin_unlock(&file_priv->mm.lock);
> @@ -4117,7 +4103,7 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
> if (target == NULL)
> return 0;
>
> - ret = __i915_wait_request(target, reset_counter, true, NULL, NULL);
> + ret = __i915_wait_request(target, true, NULL, NULL);
> if (ret == 0)
> queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work, 0);
>
> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> index 511ead08ccd8..4447e73b54db 100644
> --- a/drivers/gpu/drm/i915/intel_display.c
> +++ b/drivers/gpu/drm/i915/intel_display.c
> @@ -11313,7 +11313,6 @@ static void intel_mmio_flip_work_func(struct work_struct *work)
>
> if (mmio_flip->req) {
> WARN_ON(__i915_wait_request(mmio_flip->req,
> - mmio_flip->crtc->reset_counter,
> false, NULL,
> &mmio_flip->i915->rps.mmioflips));
> i915_gem_request_unreference__unlocked(mmio_flip->req);
> @@ -13305,9 +13304,6 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
>
> ret = drm_atomic_helper_prepare_planes(dev, state);
> if (!ret && !async && !i915_reset_in_progress(&dev_priv->gpu_error)) {
> - u32 reset_counter;
> -
> - reset_counter = i915_reset_counter(&dev_priv->gpu_error);
> mutex_unlock(&dev->struct_mutex);
>
> for_each_plane_in_state(state, plane, plane_state, i) {
> @@ -13318,8 +13314,7 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
> continue;
>
> ret = __i915_wait_request(intel_plane_state->wait_req,
> - reset_counter, true,
> - NULL, NULL);
> + true, NULL, NULL);
>
> /* Swallow -EIO errors to allow updates during hw lockup. */
> if (ret == -EIO)
> diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
> index 4ebafab53f30..fe71c768dbc4 100644
> --- a/drivers/gpu/drm/i915/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/intel_lrc.c
> @@ -819,16 +819,9 @@ static int logical_ring_prepare(struct drm_i915_gem_request *req, int bytes)
> */
> int intel_logical_ring_begin(struct drm_i915_gem_request *req, int num_dwords)
> {
> - struct drm_i915_private *dev_priv;
> int ret;
>
> WARN_ON(req == NULL);
> - dev_priv = req->ring->dev->dev_private;
> -
> - ret = i915_gem_check_wedge(&dev_priv->gpu_error,
> - dev_priv->mm.interruptible);
> - if (ret)
> - return ret;
>
> ret = logical_ring_prepare(req, num_dwords * sizeof(uint32_t));
> if (ret)
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
> index 8970267d27bb..913526c0264f 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.c
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
> @@ -2286,7 +2286,6 @@ int intel_ring_idle(struct intel_engine_cs *ring)
>
> /* Make sure we do not trigger any retires */
> return __i915_wait_request(req,
> - i915_reset_counter(&to_i915(ring->dev)->gpu_error),
> to_i915(ring->dev)->mm.interruptible,
> NULL, NULL);
> }
> @@ -2417,11 +2416,6 @@ int intel_ring_begin(struct drm_i915_gem_request *req,
> ring = req->ring;
> dev_priv = ring->dev->dev_private;
>
> - ret = i915_gem_check_wedge(&dev_priv->gpu_error,
> - dev_priv->mm.interruptible);
> - if (ret)
> - return ret;
> -
> ret = __intel_ring_prepare(ring, num_dwords * sizeof(uint32_t));
> if (ret)
> return ret;
> --
> 2.6.2
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
More information about the Intel-gfx
mailing list