gpu: nvgpu: initialize gr struct in poweron path

struct nvgpu_gr is right now initialized during probe and from OS
specific code. To support multiple instances of graphics engine,
nvgpu needs to initialize nvgpu_gr after number of engine instances
have been enumerated in poweron path.
Hence move nvgpu_gr_alloc() to poweron path and after gr manager has
been initialized.

Some of the members of nvgpu_gr are initialized in probe path and they
too are in OS specific code. Move them to common code in
nvgpu_gr_alloc()

Add field fecs_feature_override_ecc_val to struct gk20a to store the
override flag read from device tree. This flag is later copied to
nvgpu_gr in poweron path.

Update tpc_pg_mask_store() to check for g->gr being NULL before
accessing golden image pointer.
Update tpc_fs_mask_store() to return error if g->gr is not initialized.
This path needs nvgpu_gr struct initialized. Also fix the incorrect
NULL pointer check in tpc_fs_mask_store() which breaks the write path
to this sysfs.

Jira NVGPU-5648

Change-Id: Ifa2f66f3663dc2f7c8891cb03b25e997e148ab06
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2397259
Reviewed-by: automaticguardword <automaticguardword@nvidia.com>
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-cert <svc-mobile-cert@nvidia.com>
Reviewed-by: Lakshmanan M <lm@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

drivers/gpu/nvgpu/common/gr/gr.c

@@ -35,6 +35,7 @@
 #endif
 #include <nvgpu/netlist.h>
 #include <nvgpu/gr/gr_falcon.h>
+#include <nvgpu/gr/gr_utils.h>
 #include <nvgpu/gr/ctx.h>
 #include <nvgpu/gr/hwpm_map.h>
 #include <nvgpu/gr/obj_ctx.h>
@@ -159,11 +160,6 @@ u32 nvgpu_gr_rop_offset(struct gk20a *g, u32 rop)
 	return rop_offset;
 }
 
-void nvgpu_gr_init(struct gk20a *g)
-{
-	(void)nvgpu_cond_init(&g->gr->init_wq);
-}
-
 static void disable_gr_interrupts(struct gk20a *g)
 {
 	/** Disable gr intr */
@@ -840,8 +836,14 @@ int nvgpu_gr_alloc(struct gk20a *g)
 	if (gr == NULL) {
 		return -ENOMEM;
 	}
+
 	g->gr = gr;
 
+	nvgpu_cond_init(&gr->init_wq);
+#ifdef CONFIG_NVGPU_NON_FUSA
+	nvgpu_gr_override_ecc_val(g, g->fecs_feature_override_ecc_val);
+#endif
+
 	return 0;
 }
 
@@ -950,15 +952,17 @@ ctxsw_already_enabled:
 
 void nvgpu_gr_remove_support(struct gk20a *g)
 {
-	if (g->gr->remove_support != NULL) {
+	if (g->gr != NULL && g->gr->remove_support != NULL) {
 		g->gr->remove_support(g);
 	}
 }
 
 void nvgpu_gr_sw_ready(struct gk20a *g, bool enable)
 {
+	if (g->gr != NULL) {
 		g->gr->sw_ready = enable;
 	}
+}
 
 #ifdef CONFIG_NVGPU_HAL_NON_FUSA
 /* Wait until GR is initialized */

drivers/gpu/nvgpu/common/init/nvgpu_init.c

@@ -40,6 +40,7 @@
 #include <nvgpu/nvhost.h>
 #include <nvgpu/fb.h>
 #include <nvgpu/device.h>
+#include <nvgpu/gr/gr.h>
 #include <nvgpu/pm_reservation.h>
 
 #ifdef CONFIG_NVGPU_LS_PMU
@@ -649,6 +650,7 @@ int nvgpu_finalize_poweron(struct gk20a *g)
 #endif
 		NVGPU_INIT_TABLE_ENTRY(g->ops.grmgr.init_gr_manager, NO_FLAG),
 		/* prepare portion of sw required for enable hw */
+		NVGPU_INIT_TABLE_ENTRY(&nvgpu_gr_alloc, NO_FLAG),
 		NVGPU_INIT_TABLE_ENTRY(g->ops.gr.gr_prepare_sw, NO_FLAG),
 		NVGPU_INIT_TABLE_ENTRY(g->ops.gr.gr_enable_hw, NO_FLAG),
 		NVGPU_INIT_TABLE_ENTRY(g->ops.acr.acr_construct_execute,

drivers/gpu/nvgpu/common/vgpu/init/init_vgpu.c

@@ -215,6 +215,12 @@ int vgpu_finalize_poweron_common(struct gk20a *g)
 		return err;
 	}
 
+	err = nvgpu_gr_alloc(g);
+	if (err != 0) {
+		nvgpu_err(g, "couldn't allocate gr memory");
+		return err;
+	}
+
 	err = vgpu_init_gr_support(g);
 	if (err != 0) {
 		nvgpu_err(g, "failed to init gk20a gr");

drivers/gpu/nvgpu/include/nvgpu/gk20a.h

@@ -947,6 +947,7 @@ struct gk20a {
 
 #ifdef CONFIG_NVGPU_NON_FUSA
 	u32 tpc_fs_mask_user;
+	u32 fecs_feature_override_ecc_val;
 #endif
 
 	u32 tpc_pg_mask;

drivers/gpu/nvgpu/include/nvgpu/gr/gr.h

@@ -140,17 +140,6 @@ int nvgpu_gr_alloc(struct gk20a *g);
  */
 void nvgpu_gr_free(struct gk20a *g);
 
-/**
- * @brief Initialize GR struct fields
- *
- * @param g [in]	Pointer to GPU driver struct.
- *
- * Calling this function ensures that various GR struct fields are
- * initialized before they are referenced by other units or before
- * GR initialization sequence is executed.
- */
-void nvgpu_gr_init(struct gk20a *g);
-
 /**
  * @brief Initialize the s/w required to enable h/w.
  *

drivers/gpu/nvgpu/os/linux/driver_common.c

@@ -109,10 +109,8 @@ static void nvgpu_init_vars(struct gk20a *g)
 	nvgpu_set_enabled(g, NVGPU_HAS_SYNCPOINTS, platform->has_syncpoints);
 }
 
-static void nvgpu_init_gr_vars(struct gk20a *g)
+static void nvgpu_init_max_comptag(struct gk20a *g)
 {
-	nvgpu_gr_init(g);
-
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 0, 0)
 	nvgpu_log_info(g, "total ram pages : %lu", totalram_pages());
 #else
@@ -264,7 +262,7 @@ int nvgpu_probe(struct gk20a *g,
 	int err = 0;
 
 	nvgpu_init_vars(g);
-	nvgpu_init_gr_vars(g);
+	nvgpu_init_max_comptag(g);
 	nvgpu_init_timeout(g);
 	nvgpu_init_timeslice(g);
 	nvgpu_init_pm_vars(g);

drivers/gpu/nvgpu/os/linux/module.c

@@ -1095,12 +1095,6 @@ static int gk20a_init_support(struct platform_device *pdev)
 	if (err)
 		goto fail_sim;
 
-	err = nvgpu_gr_alloc(g);
-	if (err != 0) {
-		nvgpu_err(g, "couldn't allocate gr memory");
-		goto fail_sim;
-	}
-
 	nvgpu_init_usermode_support(g);
 	return 0;
 
@@ -1568,7 +1562,7 @@ static int nvgpu_read_fuse_overrides(struct gk20a *g)
 			g->tpc_fs_mask_user = ~value;
 			break;
 		case GP10B_FUSE_OPT_ECC_EN:
-			nvgpu_gr_override_ecc_val(g, value);
+			g->fecs_feature_override_ecc_val = value;
 			break;
 		case GV11B_FUSE_OPT_TPC_DISABLE:
 			if (platform->set_tpc_pg_mask != NULL)

drivers/gpu/nvgpu/os/linux/pci.c

@@ -381,12 +381,6 @@ static int nvgpu_pci_init_support(struct pci_dev *pdev)
 	if (err)
 		goto fail_sim;
 
-	err = nvgpu_gr_alloc(g);
-	if (err != 0) {
-		nvgpu_err(g, "couldn't allocate gr memory");
-		goto fail_sim;
-	}
-
 	return 0;
 
  fail_sim:

drivers/gpu/nvgpu/os/linux/sysfs.c

@@ -846,8 +846,7 @@ static ssize_t tpc_pg_mask_store(struct device *dev,
 {
 	struct gk20a *g = get_gk20a(dev);
 	unsigned long val = 0;
-	struct nvgpu_gr_obj_ctx_golden_image *gr_golden_image =
+	struct nvgpu_gr_obj_ctx_golden_image *gr_golden_image = NULL;
-					nvgpu_gr_get_golden_image_ptr(g);
 
 	nvgpu_mutex_acquire(&g->tpc_pg_lock);
 
@@ -862,6 +861,10 @@ static ssize_t tpc_pg_mask_store(struct device *dev,
 		goto exit;
 	}
 
+	if (g->gr != NULL) {
+		gr_golden_image = nvgpu_gr_get_golden_image_ptr(g);
+	}
+
 	if (gr_golden_image &&
 			nvgpu_gr_obj_ctx_get_golden_image_size(gr_golden_image)
 			!= 0) {
@@ -892,17 +895,23 @@ static ssize_t tpc_fs_mask_store(struct device *dev,
 {
 #ifdef CONFIG_NVGPU_TEGRA_FUSE
 	struct gk20a *g = get_gk20a(dev);
-	struct nvgpu_gr_config *gr_config = nvgpu_gr_get_config_ptr(g);
+	struct nvgpu_gr_config *gr_config;
-	struct nvgpu_gr_obj_ctx_golden_image *gr_golden_image =
+	struct nvgpu_gr_obj_ctx_golden_image *gr_golden_image;
-					nvgpu_gr_get_golden_image_ptr(g);
+	struct nvgpu_gr_falcon *gr_falcon;
-	struct nvgpu_gr_falcon *gr_falcon =
-					nvgpu_gr_get_falcon_ptr(g);
 	unsigned long val = 0;
 
 	if (kstrtoul(buf, 10, &val) < 0)
 		return -EINVAL;
 
-	if (nvgpu_gr_config_get_gpc_tpc_mask_base(gr_config) != NULL)
+	if (g->gr == NULL) {
+		return -ENODEV;
+	}
+
+	gr_config = nvgpu_gr_get_config_ptr(g);
+	gr_golden_image = nvgpu_gr_get_golden_image_ptr(g);
+	gr_falcon = nvgpu_gr_get_falcon_ptr(g);
+
+	if (nvgpu_gr_config_get_gpc_tpc_mask_base(gr_config) == NULL)
 		return -ENODEV;
 
 	if (val && val != nvgpu_gr_config_get_gpc_tpc_mask(gr_config, 0) &&

drivers/gpu/nvgpu/os/linux/vgpu/vgpu_linux.c

@@ -161,12 +161,6 @@ static int vgpu_init_support(struct platform_device *pdev)
 		SZ_4K / sizeof(g->dbg_regops_tmp_buf[0]);
 #endif
 
-	err = nvgpu_gr_alloc(g);
-	if (err != 0) {
-		nvgpu_err(g, "couldn't allocate gr memory");
-		goto fail;
-	}
-
 	g->remove_support = vgpu_remove_support;
 	return 0;
 
@@ -480,7 +474,6 @@ int vgpu_probe(struct platform_device *pdev)
 	nvgpu_atomic_set(&gk20a->timeouts_disabled_refcount, 0);
 
 	vgpu_create_sysfs(dev);
-	nvgpu_gr_init(gk20a);
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 0, 0)
 	nvgpu_log_info(gk20a, "total ram pages : %lu", totalram_pages());

libs/dgpu/libnvgpu-drv-dgpu_safe.export

@@ -484,7 +484,6 @@ nvgpu_gr_global_ctx_desc_free
 nvgpu_gr_global_ctx_init_local_golden_image
 nvgpu_gr_global_ctx_load_local_golden_image
 nvgpu_gr_global_ctx_set_size
-nvgpu_gr_init
 nvgpu_gr_init_support
 nvgpu_gr_intr_init_support
 nvgpu_gr_intr_remove_support

libs/igpu/libnvgpu-drv-igpu_safe.export

@@ -498,7 +498,6 @@ nvgpu_gr_global_ctx_desc_free
 nvgpu_gr_global_ctx_init_local_golden_image
 nvgpu_gr_global_ctx_load_local_golden_image
 nvgpu_gr_global_ctx_set_size
-nvgpu_gr_init
 nvgpu_gr_init_support
 nvgpu_gr_intr_init_support
 nvgpu_gr_intr_remove_support

userspace/units/gr/nvgpu-gr.c

@@ -104,8 +104,6 @@ int test_gr_init_support(struct unit_module *m, struct gk20a *g, void *args)
 {
 	int err;
 
-	nvgpu_gr_init(g);
-
 	g->ops.ecc.ecc_init_support(g);
 	g->ops.ltc.init_ltc_support(g);
 	g->ops.mm.init_mm_support(g);