Am 22.08.2018 um 10:57 schrieb Christian König:
> Am 22.08.2018 um 10:49 schrieb Chunming Zhou:
>> VK_KHR_timeline_semaphore:
>> This extension introduces a new type of semaphore that has an integer
>> payload
>> identifying a point in a timeline. Such timeline semaphores support the
>> following operations:
>> Â Â * Host query - A host operation that allows querying the payload
>> of the
>> Â Â Â Â timeline semaphore.
>> Â Â * Host wait - A host operation that allows a blocking wait for a
>> Â Â Â Â timeline semaphore to reach a specified value.
>> Â Â * Device wait - A device operation that allows waiting for a
>> Â Â Â Â timeline semaphore to reach a specified value.
>> Â Â * Device signal - A device operation that allows advancing the
>> Â Â Â Â timeline semaphore to a specified value.
>>
>> Since it's a timeline, that means the front time point(PT) always is
>> signaled before the late PT.
>> a. signal PT design:
>> Signal PT fence N depends on PT[N-1] fence and signal opertion fence,
>> when PT[N] fence is signaled,
>> the timeline will increase to value of PT[N].
>> b. wait PT design:
>> Wait PT fence is signaled by reaching timeline point value, when
>> timeline is increasing, will compare
>> wait PTs value with new timeline value, if PT value is lower than
>> timeline value, then wait PT will be
>> signaled, otherwise keep in list. semaphore wait operation can wait
>> on any point of timeline,
>> so need a RB tree to order them. And wait PT could ahead of signal
>> PT, we need a sumission fence to
>> perform that.
>>
>> TODO:
>> CPU query and wait on timeline semaphore.
>
> Please drop DRM_SYNCOBJ_CREATE_TYPE_NORMAL, that isn't used in the
> whole implementation.
>
> Additional to that I would split up the change to
> drm_syncobj_find_fence() in a separate patch.
And please drop the submission_fence implementation and instead use
wait_event() for that.
Implementing this as a fence has the possiblity that somebody is abusing
it for wait before signal.
Christian.
>
> Christian.
>
>>
>> Change-Id: I9f09aae225e268442c30451badac40406f24262c
>> Signed-off-by: Chunming Zhou <david1.zhou at amd.com>
>> Cc: Christian Konig <christian.koenig at amd.com>
>> Cc: Dave Airlie <airlied at redhat.com>
>> Cc: Daniel Rakos <Daniel.Rakos at amd.com>
>> ---
>> Â drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c |Â Â 7 +-
>>  drivers/gpu/drm/drm_syncobj.c         | 385
>> ++++++++++++++++++++++++++++++++-
>>  drivers/gpu/drm/v3d/v3d_gem.c         |  4 +-
>>  drivers/gpu/drm/vc4/vc4_gem.c         |  2 +-
>>  include/drm/drm_syncobj.h             | 45 +++-
>>  include/uapi/drm/drm.h                |  3 +-
>> Â 6 files changed, 435 insertions(+), 11 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> index d42d1c8f78f6..463cc8960723 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> @@ -1105,7 +1105,7 @@ static int
>> amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p,
>> Â {
>> Â Â Â Â Â int r;
>> Â Â Â Â Â struct dma_fence *fence;
>> -Â Â Â r = drm_syncobj_find_fence(p->filp, handle, &fence);
>> +Â Â Â r = drm_syncobj_find_fence(p->filp, handle, &fence, 0);
>> Â Â Â Â Â if (r)
>> Â Â Â Â Â Â Â Â Â return r;
>> Â @@ -1193,8 +1193,9 @@ static void
>> amdgpu_cs_post_dependencies(struct amdgpu_cs_parser *p)
>> Â {
>> Â Â Â Â Â int i;
>> Â -Â Â Â for (i = 0; i < p->num_post_dep_syncobjs; ++i)
>> -Â Â Â Â Â Â Â drm_syncobj_replace_fence(p->post_dep_syncobjs[i], p->fence);
>> +Â Â Â for (i = 0; i < p->num_post_dep_syncobjs; ++i) {
>> +Â Â Â Â Â Â Â drm_syncobj_signal_fence(p->post_dep_syncobjs[i], p->fence, 0);
>> +Â Â Â }
>> Â }
>> Â Â static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
>> diff --git a/drivers/gpu/drm/drm_syncobj.c
>> b/drivers/gpu/drm/drm_syncobj.c
>> index 70af32d0def1..3709f36c901e 100644
>> --- a/drivers/gpu/drm/drm_syncobj.c
>> +++ b/drivers/gpu/drm/drm_syncobj.c
>> @@ -187,6 +187,191 @@ void drm_syncobj_replace_fence(struct
>> drm_syncobj *syncobj,
>> Â }
>> Â EXPORT_SYMBOL(drm_syncobj_replace_fence);
>> Â +static void drm_syncobj_timeline_signal_submission_fences(struct
>> drm_syncobj *syncobj)
>> +{
>> +Â Â Â struct rb_node *node = NULL;
>> +Â Â Â struct drm_syncobj_wait_pt *wait_pt = NULL;
>> +
>> +Â Â Â spin_lock(&syncobj->lock);
>> +Â Â Â for(node = rb_first(&syncobj->syncobj_timeline.wait_pt_tree);
>> +Â Â Â Â Â Â Â node != NULL; node = rb_next(node)) {
>> +Â Â Â Â Â Â Â wait_pt = rb_entry(node, struct drm_syncobj_wait_pt, node);
>> +Â Â Â Â Â Â Â if (wait_pt->value <= syncobj->syncobj_timeline.signal_point) {
>> +Â Â Â Â Â Â Â Â Â Â Â if (wait_pt->submission_fence)
>> + dma_fence_signal(&wait_pt->submission_fence->base);
>> +Â Â Â Â Â Â Â } else {
>> +Â Â Â Â Â Â Â Â Â Â Â /* the loop is from left to right, the later entry value is
>> +Â Â Â Â Â Â Â Â Â Â Â Â * bigger, so don't need to check any more */
>> +Â Â Â Â Â Â Â Â Â Â Â break;
>> +Â Â Â Â Â Â Â }
>> +Â Â Â }
>> +Â Â Â spin_unlock(&syncobj->lock);
>> +}
>> +
>> +static void drm_syncobj_timeline_signal_wait_pts(struct drm_syncobj
>> *syncobj)
>> +{
>> +Â Â Â struct rb_node *node = NULL;
>> +Â Â Â struct drm_syncobj_wait_pt *wait_pt = NULL;
>> +
>> +Â Â Â spin_lock(&syncobj->lock);
>> +Â Â Â for(node = rb_first(&syncobj->syncobj_timeline.wait_pt_tree);
>> +Â Â Â Â Â Â Â node != NULL; ) {
>> +Â Â Â Â Â Â Â wait_pt = rb_entry(node, struct drm_syncobj_wait_pt, node);
>> +Â Â Â Â Â Â Â node = rb_next(node);
>> +Â Â Â Â Â Â Â if (wait_pt->value <= syncobj->syncobj_timeline.timeline) {
>> +Â Â Â Â Â Â Â Â Â Â Â dma_fence_signal(&wait_pt->base.base);
>> +Â Â Â Â Â Â Â Â Â Â Â rb_erase(&wait_pt->node,
>> + &syncobj->syncobj_timeline.wait_pt_tree);
>> +Â Â Â Â Â Â Â Â Â Â Â RB_CLEAR_NODE(&wait_pt->node);
>> +Â Â Â Â Â Â Â Â Â Â Â if (wait_pt->submission_fence)
>> + dma_fence_put(&wait_pt->submission_fence->base);
>> +Â Â Â Â Â Â Â Â Â Â Â /* kfree(wait_pt) is excuted by fence put */
>> +Â Â Â Â Â Â Â Â Â Â Â dma_fence_put(&wait_pt->base.base);
>> +Â Â Â Â Â Â Â } else {
>> +Â Â Â Â Â Â Â Â Â Â Â /* the loop is from left to right, the later entry value is
>> +Â Â Â Â Â Â Â Â Â Â Â Â * bigger, so don't need to check any more */
>> +Â Â Â Â Â Â Â Â Â Â Â break;
>> +Â Â Â Â Â Â Â }
>> +Â Â Â }
>> +Â Â Â spin_unlock(&syncobj->lock);
>> +}
>> +
>> +
>> +static void pt_fence_cb(struct drm_syncobj_signal_pt *signal_pt)
>> +{
>> +Â Â Â struct dma_fence *fence = NULL;
>> +Â Â Â struct drm_syncobj *syncobj;
>> +
>> +Â Â Â fence = signal_pt->signal_fence;
>> +Â Â Â signal_pt->signal_fence = NULL;
>> +Â Â Â dma_fence_put(fence);
>> +Â Â Â fence = signal_pt->pre_pt_base;
>> +Â Â Â signal_pt->pre_pt_base = NULL;
>> +Â Â Â dma_fence_put(fence);
>> +
>> +Â Â Â syncobj = signal_pt->syncobj;
>> +Â Â Â spin_lock(&syncobj->lock);
>> +Â Â Â list_del(&signal_pt->list);
>> +Â Â Â syncobj->syncobj_timeline.timeline = signal_pt->value;
>> +Â Â Â spin_unlock(&syncobj->lock);
>> +   /* kfree(signal_pt) will be executed by below fence put */
>> +Â Â Â dma_fence_put(&signal_pt->base.base);
>> +Â Â Â drm_syncobj_timeline_signal_wait_pts(syncobj);
>> +}
>> +static void pt_signal_fence_func(struct dma_fence *fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence_cb *cb)
>> +{
>> +Â Â Â struct drm_syncobj_signal_pt *signal_pt =
>> +Â Â Â Â Â Â Â container_of(cb, struct drm_syncobj_signal_pt, signal_cb);
>> +
>> +Â Â Â if (signal_pt->pre_pt_base &&
>> +Â Â Â Â Â Â Â !dma_fence_is_signaled(signal_pt->pre_pt_base))
>> +Â Â Â Â Â Â Â return;
>> +
>> +Â Â Â pt_fence_cb(signal_pt);
>> +}
>> +static void pt_pre_fence_func(struct dma_fence *fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence_cb *cb)
>> +{
>> +Â Â Â struct drm_syncobj_signal_pt *signal_pt =
>> +Â Â Â Â Â Â Â container_of(cb, struct drm_syncobj_signal_pt, pre_pt_cb);
>> +
>> +Â Â Â if (signal_pt->signal_fence &&
>> +Â Â Â Â Â Â Â !dma_fence_is_signaled(signal_pt->pre_pt_base))
>> +Â Â Â Â Â Â Â return;
>> +
>> +Â Â Â pt_fence_cb(signal_pt);
>> +}
>> +
>> +static int drm_syncobj_timeline_signal_fence(struct drm_syncobj
>> *syncobj,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence *fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â u64 point)
>> +{
>> +Â Â Â struct drm_syncobj_signal_pt *signal_pt =
>> +Â Â Â Â Â Â Â kzalloc(sizeof(struct drm_syncobj_signal_pt), GFP_KERNEL);
>> +Â Â Â struct drm_syncobj_signal_pt *tail_pt;
>> +Â Â Â struct dma_fence *tail_pt_fence = NULL;
>> +Â Â Â int ret = 0;
>> +
>> +Â Â Â if (!signal_pt)
>> +Â Â Â Â Â Â Â return -ENOMEM;
>> +Â Â Â if (syncobj->syncobj_timeline.signal_point >= point) {
>> +Â Â Â Â Â Â Â DRM_WARN("A later signal is ready!");
>> +Â Â Â Â Â Â Â goto out;
>> +Â Â Â }
>> +Â Â Â if (fence)
>> +Â Â Â Â Â Â Â dma_fence_get(fence);
>> +Â Â Â spin_lock(&syncobj->lock);
>> +Â Â Â spin_lock_init(&signal_pt->base.lock);
>> +Â Â Â dma_fence_init(&signal_pt->base.base,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â &drm_syncobj_stub_fence_ops,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â &signal_pt->base.lock,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â syncobj->syncobj_timeline.timeline_context, point);
>> +Â Â Â signal_pt->signal_fence =
>> +Â Â Â Â Â Â Â rcu_dereference_protected(fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â lockdep_is_held(&fence->lock));
>> +Â Â Â if (!list_empty(&syncobj->syncobj_timeline.signal_pt_list)) {
>> +Â Â Â Â Â Â Â tail_pt =
>> list_last_entry(&syncobj->syncobj_timeline.signal_pt_list,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct drm_syncobj_signal_pt, list);
>> +Â Â Â Â Â Â Â tail_pt_fence = &tail_pt->base.base;
>> +Â Â Â Â Â Â Â if (dma_fence_is_signaled(tail_pt_fence))
>> +Â Â Â Â Â Â Â Â Â Â Â tail_pt_fence = NULL;
>> +Â Â Â }
>> +Â Â Â if (tail_pt_fence)
>> +Â Â Â Â Â Â Â signal_pt->pre_pt_base =
>> + dma_fence_get(rcu_dereference_protected(tail_pt_fence,
>> + lockdep_is_held(&tail_pt_fence->lock)));
>> +
>> +Â Â Â signal_pt->value = point;
>> +Â Â Â syncobj->syncobj_timeline.signal_point = point;
>> +Â Â Â signal_pt->syncobj = syncobj;
>> +Â Â Â INIT_LIST_HEAD(&signal_pt->list);
>> +Â Â Â list_add_tail(&signal_pt->list,
>> &syncobj->syncobj_timeline.signal_pt_list);
>> +Â Â Â spin_unlock(&syncobj->lock);
>> +Â Â Â drm_syncobj_timeline_signal_submission_fences(syncobj);
>> +Â Â Â /**
>> +Â Â Â Â * Every pt is depending on signal fence and previous pt fence, add
>> +Â Â Â Â * callbacks to them
>> +Â Â Â Â */
>> +Â Â Â if (!dma_fence_is_signaled(signal_pt->signal_fence))
>> +Â Â Â Â Â Â Â dma_fence_add_callback(signal_pt->signal_fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &signal_pt->signal_cb,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â pt_signal_fence_func);
>> +Â Â Â else
>> +Â Â Â Â Â Â Â pt_signal_fence_func(signal_pt->signal_fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &signal_pt->signal_cb);
>> +Â Â Â if (signal_pt->pre_pt_base &&
>> !dma_fence_is_signaled(signal_pt->pre_pt_base))
>> +Â Â Â Â Â Â Â dma_fence_add_callback(signal_pt->pre_pt_base,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &signal_pt->pre_pt_cb,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â pt_pre_fence_func);
>> +Â Â Â else
>> +Â Â Â Â Â Â Â pt_pre_fence_func(signal_pt->pre_pt_base,
>> &signal_pt->pre_pt_cb);
>> +
>> +
>> +Â Â Â return 0;
>> +out:
>> +Â Â Â kfree(signal_pt);
>> +Â Â Â return ret;
>> +}
>> +
>> +/**
>> + * drm_syncobj_signal_fence - place fence into a sync object.
>> + * @syncobj: Sync object to replace fence in
>> + * @fence: fence to install in sync file.
>> + *
>> + * This places the fence into normal or timeline sync object
>> + */
>> +void drm_syncobj_signal_fence(struct drm_syncobj *syncobj,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence *fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â u64 point)
>> +{
>> +Â Â Â if (syncobj->type == DRM_SYNCOBJ_TYPE_NORMAL)
>> +Â Â Â Â Â Â Â drm_syncobj_replace_fence(syncobj, fence);
>> +Â Â Â else if (syncobj->type == DRM_SYNCOBJ_TYPE_TIMELINE)
>> +Â Â Â Â Â Â Â drm_syncobj_timeline_signal_fence(syncobj, fence, point);
>> +}
>> +EXPORT_SYMBOL(drm_syncobj_signal_fence);
>> +
>> Â static int drm_syncobj_assign_null_handle(struct drm_syncobj *syncobj)
>> Â {
>> Â Â Â Â Â struct drm_syncobj_stub_fence *fence;
>> @@ -205,12 +390,150 @@ static int
>> drm_syncobj_assign_null_handle(struct drm_syncobj *syncobj)
>> Â Â Â Â Â Â return 0;
>> Â }
>> +static struct drm_syncobj_wait_pt *
>> +drm_syncobj_timeline_lookup_wait_pt(struct drm_syncobj *syncobj, u64
>> point)
>> +{
>> +Â Â Â struct rb_node *node =
>> syncobj->syncobj_timeline.wait_pt_tree.rb_node;
>> +Â Â Â struct drm_syncobj_wait_pt *wait_pt = NULL;
>> +
>> +
>> +Â Â Â spin_lock(&syncobj->lock);
>> +Â Â Â while(node) {
>> +Â Â Â Â Â Â Â int result = point - wait_pt->value;
>> +
>> +Â Â Â Â Â Â Â wait_pt = rb_entry(node, struct drm_syncobj_wait_pt, node);
>> +Â Â Â Â Â Â Â if (result < 0)
>> +Â Â Â Â Â Â Â Â Â Â Â node = node->rb_left;
>> +Â Â Â Â Â Â Â else if (result > 0)
>> +Â Â Â Â Â Â Â Â Â Â Â node = node->rb_right;
>> +Â Â Â Â Â Â Â else
>> +Â Â Â Â Â Â Â Â Â Â Â break;
>> +Â Â Â }
>> +Â Â Â spin_unlock(&syncobj->lock);
>> +
>> +Â Â Â return wait_pt;
>> +}
>> +
>> +static struct drm_syncobj_wait_pt *
>> +drm_syncobj_timeline_create_wait_pt(struct drm_syncobj *syncobj, u64
>> point)
>> +{
>> +Â Â Â struct drm_syncobj_wait_pt *wait_pt;
>> +Â Â Â struct rb_node **new =
>> &(syncobj->syncobj_timeline.wait_pt_tree.rb_node), *parent = NULL;
>> +
>> +Â Â Â wait_pt = kzalloc(sizeof(*wait_pt), GFP_KERNEL);
>> +Â Â Â if (!wait_pt)
>> +Â Â Â Â Â Â Â return NULL;
>> +Â Â Â spin_lock_init(&wait_pt->base.lock);
>> +Â Â Â dma_fence_init(&wait_pt->base.base,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â &drm_syncobj_stub_fence_ops,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â &wait_pt->base.lock,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â syncobj->syncobj_timeline.timeline_context, point);
>> +Â Â Â wait_pt->submission_fence = NULL;
>> +Â Â Â /* check if the pt is already sumbitted, if yes, then don't need to
>> +Â Â Â Â * create submission fence, otherwise create it */
>> +Â Â Â if (point > syncobj->syncobj_timeline.signal_point) {
>> +Â Â Â Â Â Â Â wait_pt->submission_fence =
>> +Â Â Â Â Â Â Â Â Â Â Â kzalloc(sizeof(struct drm_syncobj_stub_fence),
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â GFP_KERNEL);
>> +Â Â Â Â Â Â Â if (!wait_pt->submission_fence) {
>> +Â Â Â Â Â Â Â Â Â Â Â dma_fence_put(&wait_pt->base.base);
>> +Â Â Â Â Â Â Â Â Â Â Â return NULL;
>> +Â Â Â Â Â Â Â }
>> + spin_lock_init(&wait_pt->submission_fence->lock);
>> + dma_fence_init(&wait_pt->submission_fence->base,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &drm_syncobj_stub_fence_ops,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &wait_pt->submission_fence->lock,
>> + syncobj->syncobj_timeline.submission_context,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â point);
>> +Â Â Â }
>> +Â Â Â wait_pt->value = point;
>> +
>> +Â Â Â /* wait pt must be in an order, so that we can easily lookup and
>> signal
>> +Â Â Â Â * it */
>> +Â Â Â spin_lock(&syncobj->lock);
>> +Â Â Â if (point <= syncobj->syncobj_timeline.timeline)
>> +Â Â Â Â Â Â Â dma_fence_signal(&wait_pt->base.base);
>> +Â Â Â if (wait_pt->submission_fence &&
>> +Â Â Â Â Â Â Â point <= syncobj->syncobj_timeline.signal_point)
>> + dma_fence_signal(&wait_pt->submission_fence->base);
>> +Â Â Â while(*new) {
>> +Â Â Â Â Â Â Â struct drm_syncobj_wait_pt *this =
>> +Â Â Â Â Â Â Â Â Â Â Â rb_entry(*new, struct drm_syncobj_wait_pt, node);
>> +Â Â Â Â Â Â Â int result = wait_pt->value - this->value;
>> +
>> +Â Â Â Â Â Â Â parent = *new;
>> +Â Â Â Â Â Â Â if (result < 0)
>> +Â Â Â Â Â Â Â Â Â Â Â new = &((*new)->rb_left);
>> +Â Â Â Â Â Â Â else if (result > 0)
>> +Â Â Â Â Â Â Â Â Â Â Â new = &((*new)->rb_right);
>> +Â Â Â Â Â Â Â else
>> +Â Â Â Â Â Â Â Â Â Â Â goto exist;
>> +Â Â Â }
>> +
>> +Â Â Â rb_link_node(&wait_pt->node, parent, new);
>> +Â Â Â rb_insert_color(&wait_pt->node,
>> &syncobj->syncobj_timeline.wait_pt_tree);
>> +Â Â Â spin_unlock(&syncobj->lock);
>> +Â Â Â return wait_pt;
>> +exist:
>> +Â Â Â spin_unlock(&syncobj->lock);
>> +Â Â Â if (wait_pt->submission_fence)
>> + dma_fence_put(&wait_pt->submission_fence->base);
>> +Â Â Â dma_fence_put(&wait_pt->base.base);
>> +Â Â Â wait_pt = drm_syncobj_timeline_lookup_wait_pt(syncobj, point);
>> +Â Â Â return wait_pt;
>> +}
>> +
>> +static struct dma_fence *
>> +drm_syncobj_timeline_point_get(struct drm_syncobj *syncobj, u64
>> point, u64 flag)
>> +{
>> +Â Â Â struct drm_syncobj_wait_pt *wait_pt;
>> +
>> +Â Â Â /* already signaled, simply return a signaled stub fence */
>> +Â Â Â if (point <= syncobj->syncobj_timeline.timeline) {
>> +Â Â Â Â Â Â Â struct drm_syncobj_stub_fence *fence;
>> +
>> +Â Â Â Â Â Â Â fence = kzalloc(sizeof(*fence), GFP_KERNEL);
>> +Â Â Â Â Â Â Â if (fence == NULL)
>> +Â Â Â Â Â Â Â Â Â Â Â return NULL;
>> +
>> +Â Â Â Â Â Â Â spin_lock_init(&fence->lock);
>> +Â Â Â Â Â Â Â dma_fence_init(&fence->base, &drm_syncobj_stub_fence_ops,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &fence->lock, 0, 0);
>> +Â Â Â Â Â Â Â dma_fence_signal(&fence->base);
>> +Â Â Â Â Â Â Â return &fence->base;
>> +Â Â Â }
>> +
>> +Â Â Â /* check if the wait pt exists */
>> +Â Â Â wait_pt = drm_syncobj_timeline_lookup_wait_pt(syncobj, point);
>> +Â Â Â if (!wait_pt) {
>> +Â Â Â Â Â Â Â /* This is a new wait pt, so create it */
>> +Â Â Â Â Â Â Â wait_pt = drm_syncobj_timeline_create_wait_pt(syncobj, point);
>> +Â Â Â Â Â Â Â if (!wait_pt)
>> +Â Â Â Â Â Â Â Â Â Â Â return NULL;
>> +Â Â Â }
>> +Â Â Â if (wait_pt) {
>> +Â Â Â Â Â Â Â struct dma_fence *fence;
>> +Â Â Â Â Â Â Â if (wait_pt->submission_fence) {
>> +Â Â Â Â Â Â Â Â Â Â Â if (flag & DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT)
>> + dma_fence_wait(&wait_pt->submission_fence->base, true);
>> +Â Â Â Â Â Â Â Â Â Â Â else if
>> (!dma_fence_is_signaled(&wait_pt->submission_fence->base))
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â return NULL;
>> +Â Â Â Â Â Â Â }
>> +Â Â Â Â Â Â Â rcu_read_lock();
>> +Â Â Â Â Â Â Â fence = dma_fence_get_rcu(&wait_pt->base.base);
>> +Â Â Â Â Â Â Â rcu_read_unlock();
>> +Â Â Â Â Â Â Â return fence;
>> +Â Â Â }
>> +Â Â Â return NULL;
>> +}
>> +
>> Â Â /**
>> Â Â * drm_syncobj_find_fence - lookup and reference the fence in a
>> sync object
>> Â Â * @file_private: drm file private pointer
>> Â Â * @handle: sync object handle to lookup.
>> Â Â * @fence: out parameter for the fence
>> + * @point: timeline point
>> Â Â *
>> Â Â * This is just a convenience function that combines
>> drm_syncobj_find() and
>> Â Â * drm_syncobj_fence_get().
>> @@ -221,7 +544,8 @@ static int drm_syncobj_assign_null_handle(struct
>> drm_syncobj *syncobj)
>> Â Â */
>> Â int drm_syncobj_find_fence(struct drm_file *file_private,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â u32 handle,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence **fence)
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence **fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â u64 point)
>> Â {
>> Â Â Â Â Â struct drm_syncobj *syncobj = drm_syncobj_find(file_private,
>> handle);
>> Â Â Â Â Â int ret = 0;
>> @@ -229,7 +553,15 @@ int drm_syncobj_find_fence(struct drm_file
>> *file_private,
>> Â Â Â Â Â if (!syncobj)
>> Â Â Â Â Â Â Â Â Â return -ENOENT;
>> Â -Â Â Â *fence = drm_syncobj_fence_get(syncobj);
>> +Â Â Â if (syncobj->type == DRM_SYNCOBJ_TYPE_NORMAL) {
>> +Â Â Â Â Â Â Â *fence = drm_syncobj_fence_get(syncobj);
>> +Â Â Â }else if (syncobj->type == DRM_SYNCOBJ_TYPE_TIMELINE) {
>> +Â Â Â Â Â Â Â *fence = drm_syncobj_timeline_point_get(syncobj, point,
>> + DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT);
>> +Â Â Â } else {
>> +Â Â Â Â Â Â Â DRM_ERROR("invalid syncobj type\n");
>> +Â Â Â Â Â Â Â return -EINVAL;
>> +Â Â Â }
>> Â Â Â Â Â if (!*fence) {
>> Â Â Â Â Â Â Â Â Â ret = -EINVAL;
>> Â Â Â Â Â }
>> @@ -238,6 +570,35 @@ int drm_syncobj_find_fence(struct drm_file
>> *file_private,
>> Â }
>> Â EXPORT_SYMBOL(drm_syncobj_find_fence);
>> Â +static void drm_syncobj_timeline_fini(struct drm_syncobj *syncobj,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct drm_syncobj_timeline *syncobj_timeline)
>> +{
>> +Â Â Â struct rb_node *node = NULL;
>> +Â Â Â struct drm_syncobj_wait_pt *wait_pt = NULL;
>> +Â Â Â struct drm_syncobj_signal_pt *signal_pt = NULL, *tmp;
>> +
>> +Â Â Â spin_lock(&syncobj->lock);
>> +Â Â Â for(node = rb_first(&syncobj_timeline->wait_pt_tree);
>> +Â Â Â Â Â Â Â node != NULL; ) {
>> +Â Â Â Â Â Â Â wait_pt = rb_entry(node, struct drm_syncobj_wait_pt, node);
>> +Â Â Â Â Â Â Â node = rb_next(node);
>> +Â Â Â Â Â Â Â rb_erase(&wait_pt->node,
>> +Â Â Â Â Â Â Â Â Â Â Â Â &syncobj_timeline->wait_pt_tree);
>> +Â Â Â Â Â Â Â RB_CLEAR_NODE(&wait_pt->node);
>> +Â Â Â Â Â Â Â if (wait_pt->submission_fence)
>> + dma_fence_put(&wait_pt->submission_fence->base);
>> +Â Â Â Â Â Â Â /* kfree(wait_pt) is excuted by fence put */
>> +Â Â Â Â Â Â Â dma_fence_put(&wait_pt->base.base);
>> +Â Â Â }
>> +Â Â Â list_for_each_entry_safe(signal_pt, tmp,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &syncobj_timeline->signal_pt_list, list) {
>> +Â Â Â Â Â Â Â list_del(&signal_pt->list);
>> +Â Â Â Â Â Â Â dma_fence_put(signal_pt->signal_fence);
>> +Â Â Â Â Â Â Â dma_fence_put(signal_pt->pre_pt_base);
>> +Â Â Â Â Â Â Â dma_fence_put(&signal_pt->base.base);
>> +Â Â Â }
>> +Â Â Â spin_unlock(&syncobj->lock);
>> +}
>> Â /**
>> Â Â * drm_syncobj_free - free a sync object.
>> Â Â * @kref: kref to free.
>> @@ -250,10 +611,22 @@ void drm_syncobj_free(struct kref *kref)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct drm_syncobj,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â refcount);
>> Â Â Â Â Â drm_syncobj_replace_fence(syncobj, NULL);
>> +Â Â Â drm_syncobj_timeline_fini(syncobj, &syncobj->syncobj_timeline);
>> Â Â Â Â Â kfree(syncobj);
>> Â }
>> Â EXPORT_SYMBOL(drm_syncobj_free);
>> Â +static void drm_syncobj_timeline_init(struct drm_syncobj_timeline
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â *syncobj_timeline)
>> +{
>> +Â Â Â syncobj_timeline->timeline_context = dma_fence_context_alloc(1);
>> +Â Â Â syncobj_timeline->submission_context = dma_fence_context_alloc(1);
>> +Â Â Â syncobj_timeline->timeline = 0;
>> +Â Â Â syncobj_timeline->signal_point = 0;
>> +
>> +Â Â Â syncobj_timeline->wait_pt_tree = RB_ROOT;
>> +Â Â Â INIT_LIST_HEAD(&syncobj_timeline->signal_pt_list);
>> +}
>> Â /**
>> Â Â * drm_syncobj_create - create a new syncobj
>> Â Â * @out_syncobj: returned syncobj
>> @@ -279,6 +652,12 @@ int drm_syncobj_create(struct drm_syncobj
>> **out_syncobj, uint32_t flags,
>> Â Â Â Â Â kref_init(&syncobj->refcount);
>> Â Â Â Â Â INIT_LIST_HEAD(&syncobj->cb_list);
>> Â Â Â Â Â spin_lock_init(&syncobj->lock);
>> +Â Â Â if (flags & DRM_SYNCOBJ_CREATE_TYPE_TIMELINE) {
>> +Â Â Â Â Â Â Â syncobj->type = DRM_SYNCOBJ_TYPE_TIMELINE;
>> + drm_syncobj_timeline_init(&syncobj->syncobj_timeline);
>> +Â Â Â } else {
>> +Â Â Â Â Â Â Â syncobj->type = DRM_SYNCOBJ_TYPE_NORMAL;
>> +Â Â Â }
>> Â Â Â Â Â Â if (flags & DRM_SYNCOBJ_CREATE_SIGNALED) {
>> Â Â Â Â Â Â Â Â Â ret = drm_syncobj_assign_null_handle(syncobj);
>> @@ -491,7 +870,7 @@ static int drm_syncobj_export_sync_file(struct
>> drm_file *file_private,
>> Â Â Â Â Â if (fd < 0)
>> Â Â Â Â Â Â Â Â Â return fd;
>> Â -Â Â Â ret = drm_syncobj_find_fence(file_private, handle, &fence);
>> +Â Â Â ret = drm_syncobj_find_fence(file_private, handle, &fence, 0);
>> Â Â Â Â Â if (ret)
>> Â Â Â Â Â Â Â Â Â goto err_put_fd;
>> Â diff --git a/drivers/gpu/drm/v3d/v3d_gem.c
>> b/drivers/gpu/drm/v3d/v3d_gem.c
>> index 9029590267aa..f24c3eccb4d5 100644
>> --- a/drivers/gpu/drm/v3d/v3d_gem.c
>> +++ b/drivers/gpu/drm/v3d/v3d_gem.c
>> @@ -521,12 +521,12 @@ v3d_submit_cl_ioctl(struct drm_device *dev,
>> void *data,
>> Â Â Â Â Â kref_init(&exec->refcount);
>> Â Â Â Â Â Â ret = drm_syncobj_find_fence(file_priv, args->in_sync_bcl,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &exec->bin.in_fence);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &exec->bin.in_fence, 0);
>> Â Â Â Â Â if (ret == -EINVAL)
>> Â Â Â Â Â Â Â Â Â goto fail;
>> Â Â Â Â Â Â ret = drm_syncobj_find_fence(file_priv, args->in_sync_rcl,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &exec->render.in_fence);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &exec->render.in_fence, 0);
>> Â Â Â Â Â if (ret == -EINVAL)
>> Â Â Â Â Â Â Â Â Â goto fail;
>> Â diff --git a/drivers/gpu/drm/vc4/vc4_gem.c
>> b/drivers/gpu/drm/vc4/vc4_gem.c
>> index 7910b9acedd6..f7b4971342e8 100644
>> --- a/drivers/gpu/drm/vc4/vc4_gem.c
>> +++ b/drivers/gpu/drm/vc4/vc4_gem.c
>> @@ -1173,7 +1173,7 @@ vc4_submit_cl_ioctl(struct drm_device *dev,
>> void *data,
>> Â Â Â Â Â Â if (args->in_sync) {
>> Â Â Â Â Â Â Â Â Â ret = drm_syncobj_find_fence(file_priv, args->in_sync,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &in_fence);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â &in_fence, 0);
>> Â Â Â Â Â Â Â Â Â if (ret)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â goto fail;
>> Â diff --git a/include/drm/drm_syncobj.h b/include/drm/drm_syncobj.h
>> index b04c492ddbb5..b6e193c6cc9a 100644
>> --- a/include/drm/drm_syncobj.h
>> +++ b/include/drm/drm_syncobj.h
>> @@ -54,6 +54,41 @@ const struct dma_fence_ops
>> drm_syncobj_stub_fence_ops = {
>> Â Â Â Â Â .release = NULL,
>> Â };
>> Â +enum drm_syncobj_type {
>> +Â Â Â DRM_SYNCOBJ_TYPE_NORMAL,
>> +Â Â Â DRM_SYNCOBJ_TYPE_TIMELINE
>> +};
>> +
>> +struct drm_syncobj;
>> +struct drm_syncobj_wait_pt {
>> +Â Â Â struct drm_syncobj_stub_fence base;
>> +Â Â Â struct drm_syncobj_stub_fence *submission_fence;
>> +Â Â Â u64Â Â Â value;
>> +   struct rb_node  node;
>> +};
>> +struct drm_syncobj_signal_pt {
>> +Â Â Â struct drm_syncobj_stub_fence base;
>> +Â Â Â struct dma_fence *signal_fence;
>> +Â Â Â struct dma_fence *pre_pt_base;
>> +Â Â Â struct dma_fence_cb signal_cb;
>> +Â Â Â struct dma_fence_cb pre_pt_cb;
>> +Â Â Â struct drm_syncobj *syncobj;
>> +Â Â Â u64Â Â Â value;
>> +Â Â Â struct list_head list;
>> +};
>> +struct drm_syncobj_timeline {
>> +Â Â Â u64 timeline_context;
>> +Â Â Â u64 submission_context;
>> +Â Â Â /**
>> +Â Â Â Â * @timeline: syncobj timeline
>> +Â Â Â Â */
>> +Â Â Â u64 timeline;
>> +Â Â Â u64 signal_point;
>> +
>> +
>> +Â Â Â struct rb_root wait_pt_tree;
>> +Â Â Â struct list_head signal_pt_list;
>> +};
>> Â /**
>> Â Â * struct drm_syncobj - sync object.
>> Â Â *
>> @@ -64,6 +99,11 @@ struct drm_syncobj {
>> Â Â Â Â Â Â * @refcount: Reference count of this object.
>> Â Â Â Â Â Â */
>> Â Â Â Â Â struct kref refcount;
>> +Â Â Â /**
>> +Â Â Â Â * @type: indicate syncobj type
>> +Â Â Â Â */
>> +Â Â Â enum drm_syncobj_type type;
>> +Â Â Â struct drm_syncobj_timeline syncobj_timeline;
>> Â Â Â Â Â /**
>> Â Â Â Â Â Â * @fence:
>> Â Â Â Â Â Â * NULL or a pointer to the fence bound to this object.
>> @@ -162,9 +202,12 @@ void drm_syncobj_remove_callback(struct
>> drm_syncobj *syncobj,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct drm_syncobj_cb *cb);
>> Â void drm_syncobj_replace_fence(struct drm_syncobj *syncobj,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence *fence);
>> +void drm_syncobj_signal_fence(struct drm_syncobj *syncobj,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence *fence,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â u64 point);
>> Â int drm_syncobj_find_fence(struct drm_file *file_private,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â u32 handle,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence **fence);
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence **fence, u64 point);
>> Â void drm_syncobj_free(struct kref *kref);
>> Â int drm_syncobj_create(struct drm_syncobj **out_syncobj, uint32_t
>> flags,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct dma_fence *fence);
>> diff --git a/include/uapi/drm/drm.h b/include/uapi/drm/drm.h
>> index 300f336633f2..71e6cd1c88f8 100644
>> --- a/include/uapi/drm/drm.h
>> +++ b/include/uapi/drm/drm.h
>> @@ -717,6 +717,8 @@ struct drm_prime_handle {
>> Â struct drm_syncobj_create {
>> Â Â Â Â Â __u32 handle;
>> Â #define DRM_SYNCOBJ_CREATE_SIGNALED (1 << 0)
>> +#define DRM_SYNCOBJ_CREATE_TYPE_NORMAL (1 << 1)
>> +#define DRM_SYNCOBJ_CREATE_TYPE_TIMELINE (1 << 2)
>> Â Â Â Â Â __u32 flags;
>> Â };
>> Â @@ -734,7 +736,6 @@ struct drm_syncobj_handle {
>> Â Â Â Â Â __s32 fd;
>> Â Â Â Â Â __u32 pad;
>> Â };
>> -
>> Â #define DRM_SYNCOBJ_WAIT_FLAGS_WAIT_ALL (1 << 0)
>> Â #define DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT (1 << 1)
>> Â struct drm_syncobj_wait {
>
