From 724f49f6ebbd5d93c6edbc58832c947ec6ab5c83 Mon Sep 17 00:00:00 2001
From: Tejal Kudav <tkudav@nvidia.com>
Date: Wed, 5 Oct 2022 11:10:48 +0000
Subject: [PATCH] gpu: nvgpu: Remove dependency on DGPU CONFIG

The error injection code was enabled only when CONFIG_NVGPU_DGPU = n
so that the dGPUs do not attempt any error injection callback
function registration. But, this introduced dependency on DGPU
config when needs to be explicitly set to n for error injection to
be enabled.
Remove the dependency by moving the error injection callback
registration and deregistration to a HAL which is enabled only
on GA10b.

Bug 3819160

Change-Id: I4f4eb99189b1af3502d719536a91cc5e5d866bce
Signed-off-by: Tejal Kudav <tkudav@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2787202
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
---
 drivers/gpu/nvgpu/Makefile.linux.configs      |  2 -
 drivers/gpu/nvgpu/common/cic/mon/mon_init.c   | 43 ++++++++++---------
 drivers/gpu/nvgpu/hal/init/hal_ga10b.c        |  4 ++
 .../gpu/nvgpu/include/nvgpu/gops/cic_mon.h    | 18 ++++++++
 4 files changed, 45 insertions(+), 22 deletions(-)

diff --git a/drivers/gpu/nvgpu/Makefile.linux.configs b/drivers/gpu/nvgpu/Makefile.linux.configs
index 4c15d36d9..0f3c38fc2 100644
--- a/drivers/gpu/nvgpu/Makefile.linux.configs
+++ b/drivers/gpu/nvgpu/Makefile.linux.configs
@@ -199,11 +199,9 @@ endif
 
 ifeq ($(CONFIG_TEGRA_HSIERRRPTINJ),y)
 ifeq ($(CONFIG_NVGPU_ENABLE_MISC_EC),y)
-ifeq ($(CONFIG_NVGPU_DGPU),n)
 CONFIG_NVGPU_FSI_ERR_INJECTION := y
 endif
 endif
-endif
 
 ifeq ($(CONFIG_GK20A_PMU),y)
 ccflags-y += -DCONFIG_GK20A_PMU
diff --git a/drivers/gpu/nvgpu/common/cic/mon/mon_init.c b/drivers/gpu/nvgpu/common/cic/mon/mon_init.c
index 05553559d..a3560871f 100644
--- a/drivers/gpu/nvgpu/common/cic/mon/mon_init.c
+++ b/drivers/gpu/nvgpu/common/cic/mon/mon_init.c
@@ -46,16 +46,18 @@ int nvgpu_cic_mon_setup(struct gk20a *g)
 	g->cic_mon = cic_mon;
 
 #ifdef CONFIG_NVGPU_FSI_ERR_INJECTION
-	err = nvgpu_cic_mon_reg_errinj_cb(g);
-	if (err != 0) {
-		nvgpu_err(g,
-			"Err inj callback registration failed: %d",
-			err);
-		/* Continue CIC init despite err inj utility
-		 * registration failure, as the err inj support
-		 * is meant only for debug purposes.
-		 */
-		err = 0;
+	if (g->ops.cic_mon.reg_errinj_cb != NULL) {
+		err = g->ops.cic_mon.reg_errinj_cb(g);
+		if (err != 0) {
+			nvgpu_err(g,
+				"Err inj callback registration failed: %d",
+				err);
+			/* Continue CIC init despite err inj utility
+			 * registration failure, as the err inj support
+			 * is meant only for debug purposes.
+			 */
+			err = 0;
+		}
 	}
 #endif
 
@@ -100,16 +102,17 @@ int nvgpu_cic_mon_remove(struct gk20a *g)
 	struct nvgpu_cic_mon *cic_mon;
 
 #ifdef CONFIG_NVGPU_FSI_ERR_INJECTION
-	int err = nvgpu_cic_mon_dereg_errinj_cb();
-
-	if (err != 0) {
-		nvgpu_err(g,
-			"Err inj callback de-registration failed: %d",
-			err);
-		/* Continue CIC remove despite err inj utility
-		 * de-registration failure, as the err inj support
-		 * is meant only for debug purposes.
-		 */
+	if (g->ops.cic_mon.dereg_errinj_cb != NULL) {
+		int err = g->ops.cic_mon.dereg_errinj_cb();
+		if (err != 0) {
+			nvgpu_err(g,
+				"Err inj callback de-registration failed: %d",
+				err);
+			/* Continue CIC remove despite err inj utility
+			 * de-registration failure, as the err inj support
+			 * is meant only for debug purposes.
+			 */
+		}
 	}
 #endif
 
diff --git a/drivers/gpu/nvgpu/hal/init/hal_ga10b.c b/drivers/gpu/nvgpu/hal/init/hal_ga10b.c
index 4f3a8d333..92c5417dd 100644
--- a/drivers/gpu/nvgpu/hal/init/hal_ga10b.c
+++ b/drivers/gpu/nvgpu/hal/init/hal_ga10b.c
@@ -1757,6 +1757,10 @@ static const struct gops_mssnvlink ga10b_ops_mssnvlink = {
 #endif
 
 static const struct gops_cic_mon ga10b_ops_cic_mon = {
+#ifdef CONFIG_NVGPU_FSI_ERR_INJECTION
+	.reg_errinj_cb = nvgpu_cic_mon_reg_errinj_cb,
+	.dereg_errinj_cb = nvgpu_cic_mon_dereg_errinj_cb,
+#endif
 	.init = ga10b_cic_mon_init,
 	.report_err = nvgpu_cic_mon_report_err_safety_services
 };
diff --git a/drivers/gpu/nvgpu/include/nvgpu/gops/cic_mon.h b/drivers/gpu/nvgpu/include/nvgpu/gops/cic_mon.h
index 1387a9b73..71154a41d 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/gops/cic_mon.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/gops/cic_mon.h
@@ -58,6 +58,24 @@ struct gops_cic_mon {
 	 * @return 0 in case of success, < 0 in case of failure.
 	 */
 	int (*report_err)(struct gk20a *g, u32 err_id);
+
+	/**
+	 * @brief Register error injection callback with error
+	 *        injection utility.
+	 *
+	 * @param g [in]		Pointer to GPU driver struct.
+	 *
+	 * @return 0 in case of success, < 0 in case of failure.
+	 */
+	int (*reg_errinj_cb)(struct gk20a *g);
+
+	/**
+	 * @brief Unregister error injection callback from error
+	 *        injection utility.
+	 *
+	 * @return 0 in case of success, < 0 in case of failure.
+	 */
+	int (*dereg_errinj_cb)(void);
 };
 
 #endif/*NVGPU_GOPS_CIC_MON_H*/