gpu: nvgpu: Allocate channel table with vmalloc

Channel table can be bigger than one page, so allocate it with
vmalloc.

Also add a free for tsg table, which did not exist before, and
remove per-channel remove_channel callback which was never used.

JIRA DNVGPU-50

Change-Id: I3ee84b65d94881df52bf0618bf4c5f2e85758223
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/1129244
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Ken Adams <kadams@nvidia.com>
GVS: Gerrit_Virtual_Submit

3 files changed, 11 insertions, 20 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/channel_gk20a.h b/drivers/gpu/nvgpu/gk20a/channel_gk20a.h
index fbf1ad9e..1e9341ae 100644
--- a/drivers/gpu/nvgpu/gk20a/channel_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/channel_gk20a.h
@@ -156,7 +156,6 @@ struct channel_gk20a {
                 struct mutex fence_lock;
         } last_submit;
 
-        void (*remove_support)(struct channel_gk20a *);
 #if defined(CONFIG_GK20A_CYCLE_STATS)
         struct {
         void *cyclestate_buffer;
diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
index d8c07c89..e69fb288 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
@@ -167,14 +167,8 @@ static void gk20a_remove_fifo_support(struct fifo_gk20a *f)
 
         gk20a_dbg_fn("");
 
-        if (f->channel) {
-                int c;
-                for (c = 0; c < f->num_channels; c++) {
-                        if (f->channel[c].remove_support)
-                                f->channel[c].remove_support(f->channel+c);
-                }
-                kfree(f->channel);
-        }
+        vfree(f->channel);
+        vfree(f->tsg);
         gk20a_gmmu_unmap_free(&g->mm.bar1.vm, &f->userd);
 
         engine_info = f->engine_info + ENGINE_GR_GK20A;
@@ -502,10 +496,8 @@ static int gk20a_init_fifo_setup_sw(struct gk20a *g)
 
         gk20a_dbg(gpu_dbg_map, "userd bar1 va = 0x%llx", f->userd.gpu_va);
 
-        f->channel = kzalloc(f->num_channels * sizeof(*f->channel),
-                                GFP_KERNEL);
-        f->tsg = kzalloc(f->num_channels * sizeof(*f->tsg),
-                                GFP_KERNEL);
+        f->channel = vzalloc(f->num_channels * sizeof(*f->channel));
+        f->tsg = vzalloc(f->num_channels * sizeof(*f->tsg));
         f->pbdma_map = kzalloc(f->num_pbdma * sizeof(*f->pbdma_map),
                                 GFP_KERNEL);
         f->engine_info = kzalloc(f->max_engines * sizeof(*f->engine_info),
@@ -555,8 +547,10 @@ clean_up:
         gk20a_dbg_fn("fail");
         gk20a_gmmu_unmap_free(&g->mm.bar1.vm, &f->userd);
 
-        kfree(f->channel);
+        vfree(f->channel);
         f->channel = NULL;
+        vfree(f->tsg);
+        f->tsg = NULL;
         kfree(f->pbdma_map);
         f->pbdma_map = NULL;
         kfree(f->engine_info);
diff --git a/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c b/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
index 294581b9..dc82d57b 100644
--- a/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
@@ -267,10 +267,8 @@ static int vgpu_init_fifo_setup_sw(struct gk20a *g)
 
         gk20a_dbg(gpu_dbg_map, "userd bar1 va = 0x%llx", f->userd.gpu_va);
 
-        f->channel = kzalloc(f->num_channels * sizeof(*f->channel),
-                                GFP_KERNEL);
-        f->tsg = kzalloc(f->num_channels * sizeof(*f->tsg),
-                                GFP_KERNEL);
+        f->channel = vzalloc(f->num_channels * sizeof(*f->channel));
+        f->tsg = vzalloc(f->num_channels * sizeof(*f->tsg));
         f->engine_info = kzalloc(f->max_engines * sizeof(*f->engine_info),
                                 GFP_KERNEL);
 
@@ -315,9 +313,9 @@ clean_up:
 
         memset(&f->userd, 0, sizeof(f->userd));
 
-        kfree(f->channel);
+        vfree(f->channel);
         f->channel = NULL;
-        kfree(f->tsg);
+        vfree(f->tsg);
         f->tsg = NULL;
         kfree(f->engine_info);
         f->engine_info = NULL;