gpu: nvgpu: correct calculation of sm_id for .record_sm_error_state

Starting with Volta, one TPC could have more than 1 SMs. So
.record_sm_error_state needs to have sm number as parameter.
Logic tpc id should be read from gr_gpc0_gpm_pd_sm_id_r.

Let the function return logical sm_id. RM server will need it to nofify
client.

Jira EVLR-2643
Bug 200405202

Change-Id: Iffaff05b89b1c5058616b8a6bf50dd73bd4e52f6
Signed-off-by: Richard Zhao <rizhao@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1742165
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 2 insertions, 2 deletions

diff --git a/drivers/gpu/nvgpu/gm20b/gr_gm20b.c b/drivers/gpu/nvgpu/gm20b/gr_gm20b.c
index 331c3af9..261c3054 100644
--- a/drivers/gpu/nvgpu/gm20b/gr_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/gr_gm20b.c
@@ -1278,7 +1278,7 @@ void gr_gm20b_get_access_map(struct gk20a *g,
         *num_entries = ARRAY_SIZE(wl_addr_gm20b);
 }
 
-int gm20b_gr_record_sm_error_state(struct gk20a *g, u32 gpc, u32 tpc,
+int gm20b_gr_record_sm_error_state(struct gk20a *g, u32 gpc, u32 tpc, u32 sm,
                                 struct channel_gk20a *fault_ch)
 {
         int sm_id;
@@ -1306,7 +1306,7 @@ int gm20b_gr_record_sm_error_state(struct gk20a *g, u32 gpc, u32 tpc,
 
         nvgpu_mutex_release(&g->dbg_sessions_lock);
 
-        return 0;
+        return sm_id;
 }
 
 int gm20b_gr_update_sm_error_state(struct gk20a *g,