gpu: nvgpu: move snapshot_client memory handling to linux

We right now store dmabuf fd and dma_buf pointer for gk20a_cs_snapshot_client
But since dma_buf and all related APIs are linux specific, we need to remove
them from common code and move them to linux specific code

Add new linux specific structure gk20a_cs_snapshot_client_linux which includes
struct gk20a_cs_snapshot_client and linux specific dma_buf pointer

In gk20a_attach_cycle_stats_snapshot(), we first handle all dma_buf related
operations and then call gr_gk20a_css_attach()

Move gk20a_channel_free_cycle_stats_snapshot() to ioctl_channel.c
In gk20a_channel_free_cycle_stats_snapshot(), we call gr_gk20a_css_detach()
and then free up dma_buf in linux specific code

We also need to call gk20a_channel_free_cycle_stats_snapshot() while closing
the channel, so call it from linux specific nvgpu_channel_close_linux()

Jira NVGPU-397
Jira NVGPU-415

Change-Id: Ida27240541f6adf31f28d7d7ee4f51651c6d3de2
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1603908
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 8 insertions, 54 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/css_gr_gk20a.c b/drivers/gpu/nvgpu/gk20a/css_gr_gk20a.c
index e3896981..afba2496 100644
--- a/drivers/gpu/nvgpu/gk20a/css_gr_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/css_gr_gk20a.c
@@ -45,11 +45,6 @@
                 ((cl)->perfmon_start <= (pm) &&                 \
                 ((pm) - (cl)->perfmon_start) < (cl)->perfmon_count)
 
-/* the minimal size of client buffer */
-#define CSS_MIN_CLIENT_SNAPSHOT_SIZE                            \
-                (sizeof(struct gk20a_cs_snapshot_fifo) +        \
-                sizeof(struct gk20a_cs_snapshot_fifo_entry) * 256)
-
 /* address of fifo entry by offset */
 #define CSS_FIFO_ENTRY(fifo, offs)                              \
         ((struct gk20a_cs_snapshot_fifo_entry *)(((char *)(fifo)) + (offs)))
@@ -452,52 +447,16 @@ static int css_gr_free_client_data(struct gk20a *g,
                         ret = -EINVAL;
         }
 
-        if (client->dma_handler) {
-                if (client->snapshot)
-                        dma_buf_vunmap(client->dma_handler, client->snapshot);
-                dma_buf_put(client->dma_handler);
-        }
-
-        nvgpu_kfree(g, client);
-
         return ret;
 }
 
 static int css_gr_create_client_data(struct gk20a *g,
                         struct gk20a_cs_snapshot *data,
-                        u32 dmabuf_fd, u32 perfmon_count,
-                        struct gk20a_cs_snapshot_client **client)
+                        u32 perfmon_count,
+                        struct gk20a_cs_snapshot_client *cur)
 {
-        struct gk20a_cs_snapshot_client *cur;
         int ret = 0;
 
-        cur = nvgpu_kzalloc(g, sizeof(*cur));
-        if (!cur) {
-                ret = -ENOMEM;
-                goto failed;
-        }
-
-        cur->dmabuf_fd   = dmabuf_fd;
-        cur->dma_handler = dma_buf_get(cur->dmabuf_fd);
-        if (IS_ERR(cur->dma_handler)) {
-                ret = PTR_ERR(cur->dma_handler);
-                cur->dma_handler = NULL;
-                goto failed;
-        }
-
-        cur->snapshot = (struct gk20a_cs_snapshot_fifo *)
-                                        dma_buf_vmap(cur->dma_handler);
-        if (!cur->snapshot) {
-                ret = -ENOMEM;
-                goto failed;
-        }
-
-        cur->snapshot_size = cur->dma_handler->size;
-        if (cur->snapshot_size < CSS_MIN_CLIENT_SNAPSHOT_SIZE) {
-                ret = -ENOMEM;
-                goto failed;
-        }
-
         memset(cur->snapshot, 0, sizeof(*cur->snapshot));
         cur->snapshot->start = sizeof(*cur->snapshot);
         /* we should be ensure that can fit all fifo entries here */
@@ -523,12 +482,10 @@ static int css_gr_create_client_data(struct gk20a *g,
         }
 
         nvgpu_list_add_tail(&cur->list, &data->clients);
-        *client = cur;
 
         return 0;
 
 failed:
-        *client = NULL;
         if (cur)
                 css_gr_free_client_data(g, data, cur);
 
@@ -537,10 +494,9 @@ failed:
 
 
 int gr_gk20a_css_attach(struct channel_gk20a *ch,
-                        u32 dmabuf_fd,
                         u32 perfmon_count,
                         u32 *perfmon_start,
-                        struct gk20a_cs_snapshot_client **cs_client)
+                        struct gk20a_cs_snapshot_client *cs_client)
 {
         int ret = 0;
         struct gk20a *g = ch->g;
@@ -555,7 +511,6 @@ int gr_gk20a_css_attach(struct channel_gk20a *ch,
                 return -EINVAL;
 
         gr = &g->gr;
-        *cs_client = NULL;
 
         nvgpu_mutex_acquire(&gr->cs_lock);
 
@@ -564,18 +519,17 @@ int gr_gk20a_css_attach(struct channel_gk20a *ch,
                 goto failed;
 
         ret = css_gr_create_client_data(g, gr->cs_data,
-                                     dmabuf_fd,
                                      perfmon_count,
                                      cs_client);
         if (ret)
                 goto failed;
 
-        ret = g->ops.css.enable_snapshot(ch, *cs_client);
+        ret = g->ops.css.enable_snapshot(ch, cs_client);
         if (ret)
                 goto failed;
 
         if (perfmon_start)
-                *perfmon_start = (*cs_client)->perfmon_start;
+                *perfmon_start = cs_client->perfmon_start;
 
         nvgpu_mutex_release(&gr->cs_lock);
 
@@ -583,9 +537,9 @@ int gr_gk20a_css_attach(struct channel_gk20a *ch,
 
 failed:
         if (gr->cs_data) {
-                if (*cs_client) {
-                        css_gr_free_client_data(g, gr->cs_data, *cs_client);
-                        *cs_client = NULL;
+                if (cs_client) {
+                        css_gr_free_client_data(g, gr->cs_data, cs_client);
+                        cs_client = NULL;
                 }
 
                 if (nvgpu_list_empty(&gr->cs_data->clients))