gpu: nvgpu: wait ACK for FECS watchdog timeout

On Volta, nvgpu needs to wait for explicit ACK from CTXSW while
setting FECS watchdog timeoout

This is manual port of the fixes 4d7e5026e38528b88a4a168eca9a8b180475b368
and ad89436b03428a42e43042b6a849c15843fdebc4 on dev-main since clean
cherry-pick is not possible due to huge file and structure differences.

Bug 200603566

Change-Id: Icba69998ab45eee5fdf2a29e1ac1067589301be6
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2371708
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 32 insertions, 10 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
index 636d5714..7bcf5281 100644
--- a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
@@ -541,17 +541,12 @@ int gr_gk20a_ctx_wait_ucode(struct gk20a *g, u32 mailbox_id,
         return 0;
 }
 
-/* The following is a less brittle way to call gr_gk20a_submit_fecs_method(...)
- * We should replace most, if not all, fecs method calls to this instead. */
-int gr_gk20a_submit_fecs_method_op(struct gk20a *g,
+int gr_gk20a_submit_fecs_method_op_locked(struct gk20a *g,
                                    struct fecs_method_op_gk20a op,
                                    bool sleepduringwait)
 {
-        struct gr_gk20a *gr = &g->gr;
         int ret;
 
-        nvgpu_mutex_acquire(&gr->fecs_mutex);
-
         if (op.mailbox.id != 0) {
                 gk20a_writel(g, gr_fecs_ctxsw_mailbox_r(op.mailbox.id),
                              op.mailbox.data);
@@ -579,6 +574,22 @@ int gr_gk20a_submit_fecs_method_op(struct gk20a *g,
                         op.method.data, op.method.addr);
         }
 
+        return ret;
+}
+
+/* The following is a less brittle way to call gr_gk20a_submit_fecs_method(...)
+ * We should replace most, if not all, fecs method calls to this instead. */
+int gr_gk20a_submit_fecs_method_op(struct gk20a *g,
+                                   struct fecs_method_op_gk20a op,
+                                   bool sleepduringwait)
+{
+        struct gr_gk20a *gr = &g->gr;
+        int ret;
+
+        nvgpu_mutex_acquire(&gr->fecs_mutex);
+
+        ret = gr_gk20a_submit_fecs_method_op_locked(g, op, sleepduringwait);
+
         nvgpu_mutex_release(&gr->fecs_mutex);
 
         return ret;
@@ -2486,6 +2497,16 @@ int gr_gk20a_load_ctxsw_ucode(struct gk20a *g)
         return 0;
 }
 
+int gr_gk20a_set_fecs_watchdog_timeout(struct gk20a *g)
+{
+        gk20a_writel(g, gr_fecs_ctxsw_mailbox_clear_r(0), 0xffffffff);
+        gk20a_writel(g, gr_fecs_method_data_r(), 0x7fffffff);
+        gk20a_writel(g, gr_fecs_method_push_r(),
+                gr_fecs_method_push_adr_set_watchdog_timeout_f());
+
+        return 0;
+}
+
 static int gr_gk20a_wait_ctxsw_ready(struct gk20a *g)
 {
         u32 ret;
@@ -2507,10 +2528,11 @@ static int gr_gk20a_wait_ctxsw_ready(struct gk20a *g)
                         gr_fecs_current_ctx_valid_false_f());
         }
 
-        gk20a_writel(g, gr_fecs_ctxsw_mailbox_clear_r(0), 0xffffffff);
-        gk20a_writel(g, gr_fecs_method_data_r(), 0x7fffffff);
-        gk20a_writel(g, gr_fecs_method_push_r(),
-                     gr_fecs_method_push_adr_set_watchdog_timeout_f());
+        ret = g->ops.gr.set_fecs_watchdog_timeout(g);
+        if (ret) {
+                nvgpu_err(g, "fail to set watchdog timeout");
+                return ret;
+        }
 
         nvgpu_log_fn(g, "done");
         return 0;