From 0c46f8a5e112c08c172ee2c692832e1753ffbcce Mon Sep 17 00:00:00 2001 From: Deepak Nibade Date: Tue, 20 Feb 2018 23:49:37 -0800 Subject: gpu: nvgpu: support user fence updates Add support for user fence updates i.e. increments added by user space in pushbuffer directly Add a submit IOCTL flag NVGPU_SUBMIT_GPFIFO_FLAGS_USER_FENCE_UPDATE to indicate if User has added increments in pushbuffer If yes, number_of_increment value is received in fence.value from User If User is adding increments in the pushbuffer then we don't need to do any job tracking in the kernel So fail the submit if we evaluate need_job_tracking to true and FLAGS_USER_FENCE_UPDATE is set User is responsible for ensuring all pre-requisites for a fast submit and to prevent kernel job tracking Since user space adds increments in the pushbuffer, just handle the threshold book keeping in kernel. Bug 200326065 Jira NVGPU-179 Change-Id: Ic0f0b1aa69e3389a4c3305fb6a559c5113719e0f Signed-off-by: Deepak Nibade Reviewed-on: https://git-master.nvidia.com/r/1661854 Reviewed-by: svc-mobile-coverity GVS: Gerrit_Virtual_Submit Reviewed-by: Vijayakumar Subbu Reviewed-by: Konsta Holtta Reviewed-by: mobile promotions Tested-by: mobile promotions --- drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c | 15 +++++++++++++++ drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.h | 3 +++ 2 files changed, 18 insertions(+) (limited to 'drivers/gpu/nvgpu/gk20a') diff --git a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c index e965a329..18d61faa 100644 --- a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c +++ b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c @@ -301,6 +301,13 @@ static u64 gk20a_channel_syncpt_address(struct gk20a_channel_sync *s) return sp->syncpt_buf.gpu_va; } +static u32 gk20a_channel_add_user_incrs(struct gk20a_channel_sync *s, u32 val) +{ + struct gk20a_channel_syncpt *sp = + container_of(s, struct gk20a_channel_syncpt, ops); + return nvgpu_nvhost_syncpt_incr_max_ext(sp->nvhost_dev, sp->id, val); +} + static void gk20a_channel_syncpt_destroy(struct gk20a_channel_sync *s) { struct gk20a_channel_syncpt *sp = @@ -353,6 +360,7 @@ gk20a_channel_syncpt_create(struct channel_gk20a *c) sp->ops.signal_timeline = gk20a_channel_syncpt_signal_timeline; sp->ops.syncpt_id = gk20a_channel_syncpt_id; sp->ops.syncpt_address = gk20a_channel_syncpt_address; + sp->ops.add_user_incrs = gk20a_channel_add_user_incrs; sp->ops.destroy = gk20a_channel_syncpt_destroy; return &sp->ops; @@ -878,6 +886,12 @@ static u64 gk20a_channel_semaphore_syncpt_address(struct gk20a_channel_sync *s) return 0; } +static u32 gk20a_channel_semaphore_add_user_incrs(struct gk20a_channel_sync *s, + u32 val) +{ + return 0; +} + static void gk20a_channel_semaphore_destroy(struct gk20a_channel_sync *s) { struct gk20a_channel_semaphore *sema = @@ -930,6 +944,7 @@ gk20a_channel_semaphore_create(struct channel_gk20a *c) sema->ops.signal_timeline = gk20a_channel_semaphore_signal_timeline; sema->ops.syncpt_id = gk20a_channel_semaphore_syncpt_id; sema->ops.syncpt_address = gk20a_channel_semaphore_syncpt_address; + sema->ops.add_user_incrs = gk20a_channel_semaphore_add_user_incrs; sema->ops.destroy = gk20a_channel_semaphore_destroy; return &sema->ops; diff --git a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.h b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.h index fe1d8526..c80ebd38 100644 --- a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.h +++ b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.h @@ -105,6 +105,9 @@ struct gk20a_channel_sync { /* Returns the sync point address of sync point or 0 if not supported */ u64 (*syncpt_address)(struct gk20a_channel_sync *s); + /* Handle user added increments in the push buffer */ + u32 (*add_user_incrs)(struct gk20a_channel_sync *s, u32 val); + /* Free the resources allocated by gk20a_channel_sync_create. */ void (*destroy)(struct gk20a_channel_sync *s); }; -- cgit v1.2.2