gpu: nvgpu: add gops.fifo.set_error_notifier

RM Server overrides it for handling stall interrupts.

Jira VQRM-3058

Change-Id: I8b14f073e952d19c808cb693958626b8d8aee8ca
Signed-off-by: Richard Zhao <rizhao@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1679709
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

12 files changed, 39 insertions, 21 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
index 79aec7a2..576a7f81 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
@@ -1393,7 +1393,7 @@ void gk20a_fifo_set_ctx_mmu_error_ch(struct gk20a *g,
 {
         nvgpu_err(g,
                 "channel %d generated a mmu fault", refch->chid);
-        nvgpu_set_error_notifier(refch,
+        g->ops.fifo.set_error_notifier(refch,
                                 NVGPU_ERR_NOTIFIER_FIFO_ERROR_MMU_ERR_FLT);
 }
 
@@ -1938,7 +1938,8 @@ int gk20a_fifo_force_reset_ch(struct channel_gk20a *ch,
                 nvgpu_list_for_each_entry(ch_tsg, &tsg->ch_list,
                                 channel_gk20a, ch_entry) {
                         if (gk20a_channel_get(ch_tsg)) {
-                                nvgpu_set_error_notifier(ch_tsg, err_code);
+                                g->ops.fifo.set_error_notifier(ch_tsg,
+                                                                err_code);
                                 gk20a_channel_put(ch_tsg);
                         }
                 }
@@ -1946,7 +1947,7 @@ int gk20a_fifo_force_reset_ch(struct channel_gk20a *ch,
                 nvgpu_rwsem_up_read(&tsg->ch_list_lock);
                 gk20a_fifo_recover_tsg(g, ch->tsgid, verbose);
         } else {
-                nvgpu_set_error_notifier(ch, err_code);
+                g->ops.fifo.set_error_notifier(ch, err_code);
                 gk20a_fifo_recover_ch(g, ch->chid, verbose);
         }
 
@@ -2108,7 +2109,7 @@ bool gk20a_fifo_check_ch_ctxsw_timeout(struct channel_gk20a *ch,
                 *verbose = ch->timeout_debug_dump;
                 *ms = ch->timeout_accumulated_ms;
                 if (recover)
-                        nvgpu_set_error_notifier(ch,
+                        ch->g->ops.fifo.set_error_notifier(ch,
                                         NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT);
 
                 gk20a_channel_put(ch);
@@ -2172,7 +2173,7 @@ bool gk20a_fifo_check_tsg_ctxsw_timeout(struct tsg_gk20a *tsg,
                 nvgpu_list_for_each_entry(ch, &tsg->ch_list,
                                 channel_gk20a, ch_entry) {
                         if (gk20a_channel_get(ch)) {
-                                nvgpu_set_error_notifier(ch,
+                                ch->g->ops.fifo.set_error_notifier(ch,
                                         NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT);
                                 *verbose |= ch->timeout_debug_dump;
                                 gk20a_channel_put(ch);
@@ -2487,7 +2488,7 @@ static void gk20a_fifo_pbdma_fault_rc(struct gk20a *g,
                 struct channel_gk20a *ch = &f->channel[id];
 
                 if (gk20a_channel_get(ch)) {
-                        nvgpu_set_error_notifier(ch, error_notifier);
+                        g->ops.fifo.set_error_notifier(ch, error_notifier);
                         gk20a_fifo_recover_ch(g, id, true);
                         gk20a_channel_put(ch);
                 }
@@ -2500,7 +2501,7 @@ static void gk20a_fifo_pbdma_fault_rc(struct gk20a *g,
                 nvgpu_list_for_each_entry(ch, &tsg->ch_list,
                                 channel_gk20a, ch_entry) {
                         if (gk20a_channel_get(ch)) {
-                                nvgpu_set_error_notifier(ch,
+                                g->ops.fifo.set_error_notifier(ch,
                                         error_notifier);
                                 gk20a_channel_put(ch);
                         }
@@ -2662,7 +2663,7 @@ void __locked_fifo_preempt_timeout_rc(struct gk20a *g, u32 id,
                                 channel_gk20a, ch_entry) {
                         if (!gk20a_channel_get(ch))
                                 continue;
-                        nvgpu_set_error_notifier(ch,
+                        g->ops.fifo.set_error_notifier(ch,
                                 NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT);
                         gk20a_channel_put(ch);
                 }
@@ -2675,7 +2676,7 @@ void __locked_fifo_preempt_timeout_rc(struct gk20a *g, u32 id,
                         "preempt channel %d timeout", id);
 
                 if (gk20a_channel_get(ch)) {
-                        nvgpu_set_error_notifier(ch,
+                        g->ops.fifo.set_error_notifier(ch,
                                         NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT);
                         gk20a_fifo_recover_ch(g, id, true);
                         gk20a_channel_put(ch);
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.h b/drivers/gpu/nvgpu/gk20a/gk20a.h
index c7653133..e8cfaf13 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.h
@@ -635,6 +635,7 @@ struct gpu_ops {
                                                 bool *verbose, u32 *ms);
                 int (*channel_suspend)(struct gk20a *g);
                 int (*channel_resume)(struct gk20a *g);
+                void (*set_error_notifier)(struct channel_gk20a *ch, u32 error);
 #ifdef CONFIG_TEGRA_GK20A_NVHOST
                 int (*alloc_syncpt_buf)(struct channel_gk20a *c,
                                 u32 syncpt_id, struct nvgpu_mem *syncpt_buf);
diff --git a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
index cf63546b..e06d190e 100644
--- a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
@@ -5154,14 +5154,14 @@ void gk20a_gr_set_error_notifier(struct gk20a *g,
                         nvgpu_list_for_each_entry(ch_tsg, &tsg->ch_list,
                                         channel_gk20a, ch_entry) {
                                 if (gk20a_channel_get(ch_tsg)) {
-                                        nvgpu_set_error_notifier(ch_tsg,
+                                        g->ops.fifo.set_error_notifier(ch_tsg,
                                                          error_notifier);
                                         gk20a_channel_put(ch_tsg);
                                 }
                         }
                         nvgpu_rwsem_up_read(&tsg->ch_list_lock);
                 } else {
-                        nvgpu_set_error_notifier(ch, error_notifier);
+                        g->ops.fifo.set_error_notifier(ch, error_notifier);
                 }
         }
 }
diff --git a/drivers/gpu/nvgpu/gm20b/hal_gm20b.c b/drivers/gpu/nvgpu/gm20b/hal_gm20b.c
index b41e07f9..0aac68ff 100644
--- a/drivers/gpu/nvgpu/gm20b/hal_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/hal_gm20b.c
@@ -60,6 +60,7 @@
 #include <nvgpu/bug.h>
 #include <nvgpu/enabled.h>
 #include <nvgpu/bus.h>
+#include <nvgpu/error_notifier.h>
 
 #include <nvgpu/hw/gm20b/hw_proj_gm20b.h>
 #include <nvgpu/hw/gm20b/hw_fuse_gm20b.h>
@@ -443,6 +444,7 @@ static const struct gpu_ops gm20b_ops = {
                 .check_ch_ctxsw_timeout = gk20a_fifo_check_ch_ctxsw_timeout,
                 .channel_suspend = gk20a_channel_suspend,
                 .channel_resume = gk20a_channel_resume,
+                .set_error_notifier = nvgpu_set_error_notifier,
 #ifdef CONFIG_TEGRA_GK20A_NVHOST
                 .alloc_syncpt_buf = gk20a_fifo_alloc_syncpt_buf,
                 .free_syncpt_buf = gk20a_fifo_free_syncpt_buf,
diff --git a/drivers/gpu/nvgpu/gp106/hal_gp106.c b/drivers/gpu/nvgpu/gp106/hal_gp106.c
index 2f3c1a5e..e77316d6 100644
--- a/drivers/gpu/nvgpu/gp106/hal_gp106.c
+++ b/drivers/gpu/nvgpu/gp106/hal_gp106.c
@@ -89,6 +89,7 @@
 #include <nvgpu/bus.h>
 #include <nvgpu/enabled.h>
 #include <nvgpu/ctxsw_trace.h>
+#include <nvgpu/error_notifier.h>
 
 #include <nvgpu/hw/gp106/hw_proj_gp106.h>
 #include <nvgpu/hw/gp106/hw_fifo_gp106.h>
@@ -504,6 +505,7 @@ static const struct gpu_ops gp106_ops = {
                 .check_ch_ctxsw_timeout = gk20a_fifo_check_ch_ctxsw_timeout,
                 .channel_suspend = gk20a_channel_suspend,
                 .channel_resume = gk20a_channel_resume,
+                .set_error_notifier = nvgpu_set_error_notifier,
 #ifdef CONFIG_TEGRA_GK20A_NVHOST
                 .alloc_syncpt_buf = gk20a_fifo_alloc_syncpt_buf,
                 .free_syncpt_buf = gk20a_fifo_free_syncpt_buf,
diff --git a/drivers/gpu/nvgpu/gp10b/hal_gp10b.c b/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
index 9238c4d7..78633e9b 100644
--- a/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
+++ b/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
@@ -71,6 +71,7 @@
 #include <nvgpu/enabled.h>
 #include <nvgpu/bus.h>
 #include <nvgpu/ctxsw_trace.h>
+#include <nvgpu/error_notifier.h>
 
 #include <nvgpu/hw/gp10b/hw_proj_gp10b.h>
 #include <nvgpu/hw/gp10b/hw_fuse_gp10b.h>
@@ -475,6 +476,7 @@ static const struct gpu_ops gp10b_ops = {
                 .check_ch_ctxsw_timeout = gk20a_fifo_check_ch_ctxsw_timeout,
                 .channel_suspend = gk20a_channel_suspend,
                 .channel_resume = gk20a_channel_resume,
+                .set_error_notifier = nvgpu_set_error_notifier,
 #ifdef CONFIG_TEGRA_GK20A_NVHOST
                 .alloc_syncpt_buf = gk20a_fifo_alloc_syncpt_buf,
                 .free_syncpt_buf = gk20a_fifo_free_syncpt_buf,
diff --git a/drivers/gpu/nvgpu/gv100/hal_gv100.c b/drivers/gpu/nvgpu/gv100/hal_gv100.c
index 2f2d278e..45325469 100644
--- a/drivers/gpu/nvgpu/gv100/hal_gv100.c
+++ b/drivers/gpu/nvgpu/gv100/hal_gv100.c
@@ -109,6 +109,7 @@
 #include <nvgpu/debug.h>
 #include <nvgpu/enabled.h>
 #include <nvgpu/ctxsw_trace.h>
+#include <nvgpu/error_notifier.h>
 
 #include <nvgpu/hw/gv100/hw_proj_gv100.h>
 #include <nvgpu/hw/gv100/hw_fifo_gv100.h>
@@ -516,6 +517,7 @@ static const struct gpu_ops gv100_ops = {
                 .check_ch_ctxsw_timeout = gk20a_fifo_check_ch_ctxsw_timeout,
                 .channel_suspend = gk20a_channel_suspend,
                 .channel_resume = gk20a_channel_resume,
+                .set_error_notifier = nvgpu_set_error_notifier,
 #ifdef CONFIG_TEGRA_GK20A_NVHOST
                 .alloc_syncpt_buf = gv11b_fifo_alloc_syncpt_buf,
                 .free_syncpt_buf = gv11b_fifo_free_syncpt_buf,
diff --git a/drivers/gpu/nvgpu/gv11b/hal_gv11b.c b/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
index 67b3adbf..60effbf2 100644
--- a/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
@@ -85,6 +85,7 @@
 #include <nvgpu/debug.h>
 #include <nvgpu/enabled.h>
 #include <nvgpu/ctxsw_trace.h>
+#include <nvgpu/error_notifier.h>
 
 #include <nvgpu/hw/gv11b/hw_proj_gv11b.h>
 #include <nvgpu/hw/gv11b/hw_fifo_gv11b.h>
@@ -533,6 +534,7 @@ static const struct gpu_ops gv11b_ops = {
                 .check_ch_ctxsw_timeout = gk20a_fifo_check_ch_ctxsw_timeout,
                 .channel_suspend = gk20a_channel_suspend,
                 .channel_resume = gk20a_channel_resume,
+                .set_error_notifier = nvgpu_set_error_notifier,
 #ifdef CONFIG_TEGRA_GK20A_NVHOST
                 .alloc_syncpt_buf = gv11b_fifo_alloc_syncpt_buf,
                 .free_syncpt_buf = gv11b_fifo_free_syncpt_buf,
diff --git a/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c b/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
index 6406c0da..c8a8ce49 100644
--- a/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
@@ -643,7 +643,8 @@ int vgpu_fifo_force_reset_ch(struct channel_gk20a *ch,
                 nvgpu_list_for_each_entry(ch_tsg, &tsg->ch_list,
                                 channel_gk20a, ch_entry) {
                         if (gk20a_channel_get(ch_tsg)) {
-                                nvgpu_set_error_notifier(ch_tsg, err_code);
+                                g->ops.fifo.set_error_notifier(ch_tsg,
+                                                                err_code);
                                 ch_tsg->has_timedout = true;
                                 gk20a_channel_put(ch_tsg);
                         }
@@ -651,7 +652,7 @@ int vgpu_fifo_force_reset_ch(struct channel_gk20a *ch,
 
                 nvgpu_rwsem_up_read(&tsg->ch_list_lock);
         } else {
-                nvgpu_set_error_notifier(ch, err_code);
+                g->ops.fifo.set_error_notifier(ch, err_code);
                 ch->has_timedout = true;
         }
 
@@ -726,10 +727,11 @@ int vgpu_fifo_isr(struct gk20a *g, struct tegra_vgpu_fifo_intr_info *info)
 
         switch (info->type) {
         case TEGRA_VGPU_FIFO_INTR_PBDMA:
-                nvgpu_set_error_notifier(ch, NVGPU_ERR_NOTIFIER_PBDMA_ERROR);
+                g->ops.fifo.set_error_notifier(ch,
+                                                NVGPU_ERR_NOTIFIER_PBDMA_ERROR);
                 break;
         case TEGRA_VGPU_FIFO_INTR_CTXSW_TIMEOUT:
-                nvgpu_set_error_notifier(ch,
+                g->ops.fifo.set_error_notifier(ch,
                                         NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT);
                 break;
         case TEGRA_VGPU_FIFO_INTR_MMU_FAULT:
diff --git a/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c b/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c
index ab08e09d..8b9f04dd 100644
--- a/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c
+++ b/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c
@@ -65,6 +65,7 @@
 
 #include <nvgpu/enabled.h>
 #include <nvgpu/vgpu/vgpu.h>
+#include <nvgpu/error_notifier.h>
 
 #include <nvgpu/hw/gp10b/hw_fuse_gp10b.h>
 #include <nvgpu/hw/gp10b/hw_fifo_gp10b.h>
@@ -349,6 +350,7 @@ static const struct gpu_ops vgpu_gp10b_ops = {
                 .check_ch_ctxsw_timeout = gk20a_fifo_check_ch_ctxsw_timeout,
                 .channel_suspend = gk20a_channel_suspend,
                 .channel_resume = gk20a_channel_resume,
+                .set_error_notifier = nvgpu_set_error_notifier,
 #ifdef CONFIG_TEGRA_GK20A_NVHOST
                 .alloc_syncpt_buf = gk20a_fifo_alloc_syncpt_buf,
                 .free_syncpt_buf = gk20a_fifo_free_syncpt_buf,
diff --git a/drivers/gpu/nvgpu/vgpu/gr_vgpu.c b/drivers/gpu/nvgpu/vgpu/gr_vgpu.c
index 1c72ca76..a6cb3828 100644
--- a/drivers/gpu/nvgpu/vgpu/gr_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/gr_vgpu.c
@@ -926,30 +926,30 @@ int vgpu_gr_isr(struct gk20a *g, struct tegra_vgpu_gr_intr_info *info)
                 nvgpu_cond_broadcast_interruptible(&ch->semaphore_wq);
                 break;
         case TEGRA_VGPU_GR_INTR_SEMAPHORE_TIMEOUT:
-                nvgpu_set_error_notifier(ch,
+                g->ops.fifo.set_error_notifier(ch,
                                 NVGPU_ERR_NOTIFIER_GR_SEMAPHORE_TIMEOUT);
                 break;
         case TEGRA_VGPU_GR_INTR_ILLEGAL_NOTIFY:
-                nvgpu_set_error_notifier(ch,
+                g->ops.fifo.set_error_notifier(ch,
                                         NVGPU_ERR_NOTIFIER_GR_ILLEGAL_NOTIFY);
         case TEGRA_VGPU_GR_INTR_ILLEGAL_METHOD:
                 break;
         case TEGRA_VGPU_GR_INTR_ILLEGAL_CLASS:
-                nvgpu_set_error_notifier(ch,
+                g->ops.fifo.set_error_notifier(ch,
                                         NVGPU_ERR_NOTIFIER_GR_ERROR_SW_NOTIFY);
                 break;
         case TEGRA_VGPU_GR_INTR_FECS_ERROR:
                 break;
         case TEGRA_VGPU_GR_INTR_CLASS_ERROR:
-                nvgpu_set_error_notifier(ch,
+                g->ops.fifo.set_error_notifier(ch,
                                         NVGPU_ERR_NOTIFIER_GR_ERROR_SW_NOTIFY);
                 break;
         case TEGRA_VGPU_GR_INTR_FIRMWARE_METHOD:
-                nvgpu_set_error_notifier(ch,
+                g->ops.fifo.set_error_notifier(ch,
                                 NVGPU_ERR_NOTIFIER_GR_ERROR_SW_NOTIFY);
                 break;
         case TEGRA_VGPU_GR_INTR_EXCEPTION:
-                nvgpu_set_error_notifier(ch,
+                g->ops.fifo.set_error_notifier(ch,
                                 NVGPU_ERR_NOTIFIER_GR_ERROR_SW_NOTIFY);
                 break;
         case TEGRA_VGPU_GR_INTR_SM_EXCEPTION:
diff --git a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c
index 5e3af9d7..32edbd21 100644
--- a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c
+++ b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c
@@ -23,6 +23,7 @@
 #include <gk20a/gk20a.h>
 #include <gv11b/hal_gv11b.h>
 #include <nvgpu/vgpu/vgpu.h>
+#include <nvgpu/error_notifier.h>
 
 #include "vgpu/fifo_vgpu.h"
 #include "vgpu/gr_vgpu.h"
@@ -392,6 +393,7 @@ static const struct gpu_ops vgpu_gv11b_ops = {
                 .check_ch_ctxsw_timeout = gk20a_fifo_check_ch_ctxsw_timeout,
                 .channel_suspend = gk20a_channel_suspend,
                 .channel_resume = gk20a_channel_resume,
+                .set_error_notifier = nvgpu_set_error_notifier,
 #ifdef CONFIG_TEGRA_GK20A_NVHOST
                 .alloc_syncpt_buf = vgpu_gv11b_fifo_alloc_syncpt_buf,
                 .free_syncpt_buf = gv11b_fifo_free_syncpt_buf,