gpu: nvgpu: support skipping buffer refcounting in submit

In job submission path, we always take refcount on all
the mapped buffers to safeguard against case where user
space releases the buffer early

But in case user space itself is doing proper buffer
management, kernel need not take refcounts on all the
buffers - which is also a overhead in submit path

Hence, provide a new submit flag
NVGPU_SUBMIT_GPFIFO_FLAGS_SKIP_BUFFER_REFCOUNTING to
optionally skip taking refcounts on all the buffers

Also, if we do not take refcounts, then no need to drop
any refcounts in gk20a_channel_update() as well

Bug 1698667
Bug 200141116

Change-Id: I81bb7a03240300b691c70bcec04ea1badd5934f4
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/824718
(cherry picked from commit 8c8978fa303ec4e6db0233becdbdcbad4a248173)
Reviewed-on: http://git-master/r/835801
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>

1 files changed, 16 insertions, 8 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/channel_gk20a.c b/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
index 0b84b7da..9bacb5c9 100644
--- a/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
@@ -1726,20 +1726,24 @@ fail_unlock:
 
 static int gk20a_channel_add_job(struct channel_gk20a *c,
                                  struct gk20a_fence *pre_fence,
-                                 struct gk20a_fence *post_fence)
+                                 struct gk20a_fence *post_fence,
+                                 bool skip_buffer_refcounting)
 {
         struct vm_gk20a *vm = c->vm;
         struct channel_gk20a_job *job = NULL;
         struct mapped_buffer_node **mapped_buffers = NULL;
-        int err = 0, num_mapped_buffers;
+        int err = 0, num_mapped_buffers = 0;
 
         /* job needs reference to this vm (released in channel_update) */
         gk20a_vm_get(vm);
 
-        err = gk20a_vm_get_buffers(vm, &mapped_buffers, &num_mapped_buffers);
-        if (err) {
-                gk20a_vm_put(vm);
-                return err;
+        if (!skip_buffer_refcounting) {
+                err = gk20a_vm_get_buffers(vm, &mapped_buffers,
+                                        &num_mapped_buffers);
+                if (err) {
+                        gk20a_vm_put(vm);
+                        return err;
+                }
         }
 
         job = kzalloc(sizeof(*job), GFP_KERNEL);
@@ -1795,7 +1799,8 @@ void gk20a_channel_update(struct channel_gk20a *c, int nr_completed)
                 if (c->sync)
                         c->sync->signal_timeline(c->sync);
 
-                gk20a_vm_put_buffers(vm, job->mapped_buffers,
+                if (job->num_mapped_buffers)
+                        gk20a_vm_put_buffers(vm, job->mapped_buffers,
                                 job->num_mapped_buffers);
 
                 /* Close the fences (this will unref the semaphores and release
@@ -1858,6 +1863,8 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
         const int extra_entries = 2;
         bool need_wfi = !(flags & NVGPU_SUBMIT_GPFIFO_FLAGS_SUPPRESS_WFI);
         struct nvgpu_gpfifo *gpfifo_mem = c->gpfifo.mem.cpu_va;
+        bool skip_buffer_refcounting = (flags &
+                        NVGPU_SUBMIT_GPFIFO_FLAGS_SKIP_BUFFER_REFCOUNTING);
 
         if (c->has_timedout)
                 return -ETIMEDOUT;
@@ -2106,7 +2113,8 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
                 *fence_out = gk20a_fence_get(post_fence);
 
         /* TODO! Check for errors... */
-        gk20a_channel_add_job(c, pre_fence, post_fence);
+        gk20a_channel_add_job(c, pre_fence, post_fence,
+                                skip_buffer_refcounting);
 
         c->cmds_pending = true;
         gk20a_bar1_writel(g,