gpu: nvgpu: create sync_fence only if needed

Currently, we create sync_fence (from nvhost_sync_create_fence())
for every submit
But not all submits request for a sync_fence.

Also, nvhost_sync_create_fence() API takes about 1/3rd of the total
submit path.

Hence to optimize, we can allocate sync_fence
only when user explicitly asks for it using
(NVGPU_SUBMIT_GPFIFO_FLAGS_FENCE_GET &&
NVGPU_SUBMIT_GPFIFO_FLAGS_SYNC_FENCE)

Also, in CDE path from gk20a_prepare_compressible_read(),
we reuse existing fence stored in "state" and that can
result into not returning sync_fence_fd when user asked
for it
Hence, force allocation of sync_fence when job submission
comes from CDE path

Bug 200141116

Change-Id: Ia921701bf0e2432d6b8a5e8b7d91160e7f52db1e
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/812845
(cherry picked from commit 5fd47015eeed00352cc8473eff969a66c94fee98)
Reviewed-on: http://git-master/r/837662
Reviewed-by: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Sachin Nikam <snikam@nvidia.com>

1 files changed, 19 insertions, 13 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
index 95647774..c0c8ec6d 100644
--- a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
@@ -166,7 +166,8 @@ static int __gk20a_channel_syncpt_incr(struct gk20a_channel_sync *s,
                                        bool wfi_cmd,
                                        bool register_irq,
                                        struct priv_cmd_entry **entry,
-                                       struct gk20a_fence **fence)
+                                       struct gk20a_fence **fence,
+                                       bool need_sync_fence)
 {
         u32 thresh;
         int incr_cmd_size;
@@ -239,7 +240,7 @@ static int __gk20a_channel_syncpt_incr(struct gk20a_channel_sync *s,
         }
 
         *fence = gk20a_fence_from_syncpt(sp->host1x_pdev, sp->id, thresh,
-                                         wfi_cmd);
+                                         wfi_cmd, need_sync_fence);
         *entry = incr_cmd;
         return 0;
 }
@@ -251,33 +252,35 @@ static int gk20a_channel_syncpt_incr_wfi(struct gk20a_channel_sync *s,
         return __gk20a_channel_syncpt_incr(s,
                         true /* wfi */,
                         false /* no irq handler */,
-                        entry, fence);
+                        entry, fence, true);
 }
 
 static int gk20a_channel_syncpt_incr(struct gk20a_channel_sync *s,
                               struct priv_cmd_entry **entry,
-                              struct gk20a_fence **fence)
+                              struct gk20a_fence **fence,
+                              bool need_sync_fence)
 {
         /* Don't put wfi cmd to this one since we're not returning
          * a fence to user space. */
         return __gk20a_channel_syncpt_incr(s,
                         false /* no wfi */,
                         true /* register irq */,
-                        entry, fence);
+                        entry, fence, need_sync_fence);
 }
 
 static int gk20a_channel_syncpt_incr_user(struct gk20a_channel_sync *s,
                                    int wait_fence_fd,
                                    struct priv_cmd_entry **entry,
                                    struct gk20a_fence **fence,
-                                   bool wfi)
+                                   bool wfi,
+                                   bool need_sync_fence)
 {
         /* Need to do 'wfi + host incr' since we return the fence
          * to user space. */
         return __gk20a_channel_syncpt_incr(s,
                         wfi,
                         true /* register irq */,
-                        entry, fence);
+                        entry, fence, need_sync_fence);
 }
 
 static void gk20a_channel_syncpt_set_min_eq_max(struct gk20a_channel_sync *s)
@@ -513,7 +516,8 @@ static int __gk20a_channel_semaphore_incr(
                 struct gk20a_channel_sync *s, bool wfi_cmd,
                 struct sync_fence *dependency,
                 struct priv_cmd_entry **entry,
-                struct gk20a_fence **fence)
+                struct gk20a_fence **fence,
+                bool need_sync_fence)
 {
         u64 va;
         int incr_cmd_size;
@@ -560,18 +564,19 @@ static int gk20a_channel_semaphore_incr_wfi(
         return __gk20a_channel_semaphore_incr(s,
                         true /* wfi */,
                         NULL,
-                        entry, fence);
+                        entry, fence, true);
 }
 
 static int gk20a_channel_semaphore_incr(
                 struct gk20a_channel_sync *s,
                 struct priv_cmd_entry **entry,
-                struct gk20a_fence **fence)
+                struct gk20a_fence **fence,
+                bool need_sync_fence)
 {
         /* Don't put wfi cmd to this one since we're not returning
          * a fence to user space. */
         return __gk20a_channel_semaphore_incr(s, false /* no wfi */,
-                                              NULL, entry, fence);
+                                      NULL, entry, fence, need_sync_fence);
 }
 
 static int gk20a_channel_semaphore_incr_user(
@@ -579,7 +584,8 @@ static int gk20a_channel_semaphore_incr_user(
                 int wait_fence_fd,
                 struct priv_cmd_entry **entry,
                 struct gk20a_fence **fence,
-                bool wfi)
+                bool wfi,
+                bool need_sync_fence)
 {
 #ifdef CONFIG_SYNC
         struct sync_fence *dependency = NULL;
@@ -592,7 +598,7 @@ static int gk20a_channel_semaphore_incr_user(
         }
 
         err = __gk20a_channel_semaphore_incr(s, wfi, dependency,
-                                             entry, fence);
+                                             entry, fence, need_sync_fence);
         if (err) {
                 if (dependency)
                         sync_fence_put(dependency);