gpu: nvgpu: rework ecc structure and sysfs

- create common file common/ecc.c which include common functions for add
  ecc counters and remove counters.
- common code will create a list of all counter which make it easier to
  iterate all counters.
- Add chip specific file for adding ecc counters.
- add linux specific file os/linux/ecc_sysfs.c to export counters to
  sysfs.
- remove obsolete code
- MISRA violation for using snprintf is not solved, tracking with
  jira NVGPU-859

Jira NVGPUT-115

Change-Id: I1905c43c5c9b2b131199807533dee8e63ddc12f4
Signed-off-by: Richard Zhao <rizhao@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1763536
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 24 insertions, 24 deletions

diff --git a/drivers/gpu/nvgpu/gv11b/gr_gv11b.c b/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
index d3fe5f65..c2f47a20 100644
--- a/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
@@ -198,7 +198,7 @@ static int gr_gv11b_handle_l1_tag_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 l1_tag_corrected_err_count_delta +=
                         (is_l1_tag_ecc_corrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_l1_tag_ecc_corrected_err_count_total_s());
-                g->ecc.gr.sm_l1_tag_corrected_err_count.counters[tpc] +=
+                g->ecc.gr.sm_l1_tag_ecc_corrected_err_count[gpc][tpc].counter +=
                                                         l1_tag_corrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_l1_tag_ecc_corrected_err_count_r() + offset,
@@ -213,7 +213,7 @@ static int gr_gv11b_handle_l1_tag_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 l1_tag_uncorrected_err_count_delta +=
                         (is_l1_tag_ecc_uncorrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_l1_tag_ecc_uncorrected_err_count_total_s());
-                g->ecc.gr.sm_l1_tag_uncorrected_err_count.counters[tpc] +=
+                g->ecc.gr.sm_l1_tag_ecc_uncorrected_err_count[gpc][tpc].counter +=
                                                         l1_tag_uncorrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_l1_tag_ecc_uncorrected_err_count_r() + offset,
@@ -290,7 +290,7 @@ static int gr_gv11b_handle_lrf_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 lrf_corrected_err_count_delta +=
                         (is_lrf_ecc_corrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_lrf_ecc_corrected_err_count_total_s());
-                g->ecc.gr.sm_lrf_single_err_count.counters[tpc] +=
+                g->ecc.gr.sm_lrf_ecc_single_err_count[gpc][tpc].counter +=
                                                         lrf_corrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_lrf_ecc_corrected_err_count_r() + offset,
@@ -305,7 +305,7 @@ static int gr_gv11b_handle_lrf_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 lrf_uncorrected_err_count_delta +=
                         (is_lrf_ecc_uncorrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_lrf_ecc_uncorrected_err_count_total_s());
-                g->ecc.gr.sm_lrf_double_err_count.counters[tpc] +=
+                g->ecc.gr.sm_lrf_ecc_double_err_count[gpc][tpc].counter +=
                                                         lrf_uncorrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_lrf_ecc_uncorrected_err_count_r() + offset,
@@ -449,7 +449,7 @@ static int gr_gv11b_handle_cbu_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 cbu_corrected_err_count_delta +=
                         (is_cbu_ecc_corrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_cbu_ecc_corrected_err_count_total_s());
-                g->ecc.gr.sm_cbu_corrected_err_count.counters[tpc] +=
+                g->ecc.gr.sm_cbu_ecc_corrected_err_count[gpc][tpc].counter +=
                                                         cbu_corrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_cbu_ecc_corrected_err_count_r() + offset,
@@ -464,7 +464,7 @@ static int gr_gv11b_handle_cbu_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 cbu_uncorrected_err_count_delta +=
                         (is_cbu_ecc_uncorrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_cbu_ecc_uncorrected_err_count_total_s());
-                g->ecc.gr.sm_cbu_uncorrected_err_count.counters[tpc] +=
+                g->ecc.gr.sm_cbu_ecc_uncorrected_err_count[gpc][tpc].counter +=
                                                         cbu_uncorrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_cbu_ecc_uncorrected_err_count_r() + offset,
@@ -529,7 +529,7 @@ static int gr_gv11b_handle_l1_data_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 l1_data_corrected_err_count_delta +=
                         (is_l1_data_ecc_corrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_l1_data_ecc_corrected_err_count_total_s());
-                g->ecc.gr.sm_l1_data_corrected_err_count.counters[tpc] +=
+                g->ecc.gr.sm_l1_data_ecc_corrected_err_count[gpc][tpc].counter +=
                                                         l1_data_corrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_l1_data_ecc_corrected_err_count_r() + offset,
@@ -544,7 +544,7 @@ static int gr_gv11b_handle_l1_data_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 l1_data_uncorrected_err_count_delta +=
                         (is_l1_data_ecc_uncorrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_l1_data_ecc_uncorrected_err_count_total_s());
-                g->ecc.gr.sm_l1_data_uncorrected_err_count.counters[tpc] +=
+                g->ecc.gr.sm_l1_data_ecc_uncorrected_err_count[gpc][tpc].counter +=
                                                         l1_data_uncorrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_l1_data_ecc_uncorrected_err_count_r() + offset,
@@ -613,7 +613,7 @@ static int gr_gv11b_handle_icache_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 icache_corrected_err_count_delta +=
                         (is_icache_ecc_corrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_icache_ecc_corrected_err_count_total_s());
-                g->ecc.gr.sm_icache_corrected_err_count.counters[tpc] +=
+                g->ecc.gr.sm_icache_ecc_corrected_err_count[gpc][tpc].counter +=
                                                         icache_corrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_icache_ecc_corrected_err_count_r() + offset,
@@ -628,7 +628,7 @@ static int gr_gv11b_handle_icache_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 icache_uncorrected_err_count_delta +=
                         (is_icache_ecc_uncorrected_total_err_overflow <<
                          gr_pri_gpc0_tpc0_sm_icache_ecc_uncorrected_err_count_total_s());
-                g->ecc.gr.sm_icache_uncorrected_err_count.counters[tpc] +=
+                g->ecc.gr.sm_icache_ecc_uncorrected_err_count[gpc][tpc].counter +=
                                                         icache_uncorrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_tpc0_sm_icache_ecc_uncorrected_err_count_r() + offset,
@@ -717,7 +717,7 @@ int gr_gv11b_handle_gcc_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 gcc_l15_corrected_err_count_delta +=
                         (is_gcc_l15_ecc_corrected_total_err_overflow <<
                          gr_pri_gpc0_gcc_l15_ecc_corrected_err_count_total_s());
-                g->ecc.gr.gcc_l15_corrected_err_count.counters[gpc] +=
+                g->ecc.gr.gcc_l15_ecc_corrected_err_count[gpc].counter +=
                                                         gcc_l15_corrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_gcc_l15_ecc_corrected_err_count_r() + offset,
@@ -732,7 +732,7 @@ int gr_gv11b_handle_gcc_exception(struct gk20a *g, u32 gpc, u32 tpc,
                 gcc_l15_uncorrected_err_count_delta +=
                         (is_gcc_l15_ecc_uncorrected_total_err_overflow <<
                          gr_pri_gpc0_gcc_l15_ecc_uncorrected_err_count_total_s());
-                g->ecc.gr.gcc_l15_uncorrected_err_count.counters[gpc] +=
+                g->ecc.gr.gcc_l15_ecc_uncorrected_err_count[gpc].counter +=
                                                         gcc_l15_uncorrected_err_count_delta;
                 gk20a_writel(g,
                         gr_pri_gpc0_gcc_l15_ecc_uncorrected_err_count_r() + offset,
@@ -802,9 +802,9 @@ static int gr_gv11b_handle_gpcmmu_ecc_exception(struct gk20a *g, u32 gpc,
                 uncorrected_delta += (0x1UL << gr_gpc0_mmu_l1tlb_ecc_uncorrected_err_count_total_s());
 
 
-        g->ecc.gr.mmu_l1tlb_corrected_err_count.counters[gpc] +=
+        g->ecc.gr.mmu_l1tlb_ecc_corrected_err_count[gpc].counter +=
                                                         corrected_delta;
-        g->ecc.gr.mmu_l1tlb_uncorrected_err_count.counters[gpc] +=
+        g->ecc.gr.mmu_l1tlb_ecc_uncorrected_err_count[gpc].counter +=
                                                         uncorrected_delta;
         nvgpu_log(g, gpu_dbg_intr,
                         "mmu l1tlb gpc:%d ecc interrupt intr: 0x%x", gpc, hww_esr);
@@ -824,8 +824,8 @@ static int gr_gv11b_handle_gpcmmu_ecc_exception(struct gk20a *g, u32 gpc,
                 "ecc error address: 0x%x", ecc_addr);
         nvgpu_log(g, gpu_dbg_intr,
                 "ecc error count corrected: %d, uncorrected %d",
-                g->ecc.gr.mmu_l1tlb_corrected_err_count.counters[gpc],
-                g->ecc.gr.mmu_l1tlb_uncorrected_err_count.counters[gpc]);
+                g->ecc.gr.mmu_l1tlb_ecc_corrected_err_count[gpc].counter,
+                g->ecc.gr.mmu_l1tlb_ecc_uncorrected_err_count[gpc].counter);
 
         return ret;
 }
@@ -880,9 +880,9 @@ static int gr_gv11b_handle_gpccs_ecc_exception(struct gk20a *g, u32 gpc,
         gk20a_writel(g, gr_gpc0_gpccs_falcon_ecc_status_r() + offset,
                                 gr_gpc0_gpccs_falcon_ecc_status_reset_task_f());
 
-        g->ecc.gr.gpccs_corrected_err_count.counters[gpc] +=
+        g->ecc.gr.gpccs_ecc_corrected_err_count[gpc].counter +=
                                                         corrected_delta;
-        g->ecc.gr.gpccs_uncorrected_err_count.counters[gpc] +=
+        g->ecc.gr.gpccs_ecc_uncorrected_err_count[gpc].counter +=
                                                         uncorrected_delta;
         nvgpu_log(g, gpu_dbg_intr,
                         "gppcs gpc:%d ecc interrupt intr: 0x%x", gpc, hww_esr);
@@ -907,8 +907,8 @@ static int gr_gv11b_handle_gpccs_ecc_exception(struct gk20a *g, u32 gpc,
 
         nvgpu_log(g, gpu_dbg_intr,
                 "ecc error count corrected: %d, uncorrected %d",
-                g->ecc.gr.gpccs_corrected_err_count.counters[gpc],
-                g->ecc.gr.gpccs_uncorrected_err_count.counters[gpc]);
+                g->ecc.gr.gpccs_ecc_corrected_err_count[gpc].counter,
+                g->ecc.gr.gpccs_ecc_uncorrected_err_count[gpc].counter);
 
         return ret;
 }
@@ -2419,9 +2419,9 @@ static void gr_gv11b_handle_fecs_ecc_error(struct gk20a *g, u32 intr)
                 gk20a_writel(g, gr_fecs_falcon_ecc_status_r(),
                                 gr_fecs_falcon_ecc_status_reset_task_f());
 
-                g->ecc.gr.fecs_corrected_err_count.counters[0] +=
+                g->ecc.gr.fecs_ecc_corrected_err_count[0].counter +=
                                                         corrected_delta;
-                g->ecc.gr.fecs_uncorrected_err_count.counters[0] +=
+                g->ecc.gr.fecs_ecc_uncorrected_err_count[0].counter +=
                                                         uncorrected_delta;
 
                 nvgpu_log(g, gpu_dbg_intr,
@@ -2450,8 +2450,8 @@ static void gr_gv11b_handle_fecs_ecc_error(struct gk20a *g, u32 intr)
 
                 nvgpu_log(g, gpu_dbg_intr,
                         "ecc error count corrected: %d, uncorrected %d",
-                        g->ecc.gr.fecs_corrected_err_count.counters[0],
-                        g->ecc.gr.fecs_uncorrected_err_count.counters[0]);
+                        g->ecc.gr.fecs_ecc_corrected_err_count[0].counter,
+                        g->ecc.gr.fecs_ecc_uncorrected_err_count[0].counter);
         }
 }