gpu: nvgpu: add remove_gr_sysfs gpu op

Add remove_gr_sys() op to gpu_ops to reverse steps
done in create_gr_sysfs().

Make gv11b_tegra_remove() specific to gv11b instead
to properly remove sysfs nodes. This also helps in
having gv11b specific remove steps.

Also, update platform remove function of dGPU i.e.
nvgpu_pci_tegra_remove() to remove sysfs nodes. This
adds parity with iGPU platform remove.

Bug 1987855

Change-Id: Ibbaffac5c24346709347f86444a951461894354d
Signed-off-by: Nitin Kumbhar <nkumbhar@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1735987
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

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

@@ -52,6 +52,11 @@ static int nvgpu_pci_tegra_probe(struct device *dev)
 
 static int nvgpu_pci_tegra_remove(struct device *dev)
 {
+	struct gk20a *g = get_gk20a(dev);
+
+	if (g->ops.gr.remove_gr_sysfs)
+		g->ops.gr.remove_gr_sysfs(g);
+
 	return 0;
 }
 

drivers/gpu/nvgpu/common/linux/platform_gp10b_tegra.c

@@ -67,8 +67,6 @@ static struct {
 	{"gpu", GPCCLK_INIT_RATE},
 	{"gpu_sys", 204000000} };
 
-static void gr_gp10b_remove_sysfs(struct device *dev);
-
 /*
  * gp10b_tegra_get_clocks()
  *
@@ -174,9 +172,12 @@ static int gp10b_tegra_late_probe(struct device *dev)
 	return 0;
 }
 
-int gp10b_tegra_remove(struct device *dev)
+static int gp10b_tegra_remove(struct device *dev)
 {
-	gr_gp10b_remove_sysfs(dev);
+	struct gk20a *g = get_gk20a(dev);
+
+	if (g->ops.gr.remove_gr_sysfs)
+		g->ops.gr.remove_gr_sysfs(g);
 
 	/* deinitialise tegra specific scaling quirks */
 	gp10b_tegra_scale_exit(dev);
@@ -792,9 +793,9 @@ void gr_gp10b_create_sysfs(struct gk20a *g)
 		dev_err(dev, "Failed to create sysfs attributes!\n");
 }
 
-static void gr_gp10b_remove_sysfs(struct device *dev)
+void gr_gp10b_remove_sysfs(struct gk20a *g)
 {
-	struct gk20a *g = get_gk20a(dev);
+	struct device *dev = dev_from_gk20a(g);
 
 	if (!g->ecc.gr.sm_lrf_single_err_count.counters)
 		return;

drivers/gpu/nvgpu/common/linux/platform_gp10b_tegra.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2017-2018, NVIDIA CORPORATION. All rights reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms and conditions of the GNU General Public License,
@@ -43,6 +43,4 @@ void gp10b_ecc_stat_remove(struct device *dev,
 				struct gk20a_ecc_stat *ecc_stat,
 				struct device_attribute *dev_attr_array);
 
-int gp10b_tegra_remove(struct device *dev);
-
 #endif

drivers/gpu/nvgpu/common/linux/platform_gv11b_tegra.c

@@ -44,7 +44,15 @@
 #include "platform_gk20a_tegra.h"
 #include "gv11b/gr_gv11b.h"
 
-static void gr_gv11b_remove_sysfs(struct device *dev);
+static void gv11b_tegra_scale_exit(struct device *dev)
+{
+	struct gk20a_platform *platform = gk20a_get_platform(dev);
+	struct gk20a_scale_profile *profile = platform->g->scale_profile;
+
+	if (profile)
+		tegra_bwmgr_unregister(
+			(struct tegra_bwmgr_client *)profile->private_data);
+}
 
 static int gv11b_tegra_probe(struct device *dev)
 {
@@ -82,11 +90,19 @@ static int gv11b_tegra_late_probe(struct device *dev)
 	return 0;
 }
 
+
 static int gv11b_tegra_remove(struct device *dev)
 {
-	gp10b_tegra_remove(dev);
+	struct gk20a *g = get_gk20a(dev);
 
-	gr_gv11b_remove_sysfs(dev);
+	if (g->ops.gr.remove_gr_sysfs)
+		g->ops.gr.remove_gr_sysfs(g);
+
+	gv11b_tegra_scale_exit(dev);
+
+#ifdef CONFIG_TEGRA_GK20A_NVHOST
+	nvgpu_free_nvhost_dev(get_gk20a(dev));
+#endif
 
 	return 0;
 }
@@ -498,12 +514,13 @@ void gr_gv11b_create_sysfs(struct gk20a *g)
 		dev_err(dev, "Failed to create gv11b sysfs attributes!\n");
 }
 
-static void gr_gv11b_remove_sysfs(struct device *dev)
+void gr_gv11b_remove_sysfs(struct gk20a *g)
 {
-	struct gk20a *g = get_gk20a(dev);
+	struct device *dev = dev_from_gk20a(g);
 
 	if (!g->ecc.gr.sm_l1_tag_corrected_err_count.counters)
 		return;
+	gr_gp10b_remove_sysfs(g);
 
 	gr_gp10b_ecc_stat_remove(dev,
 			0,

drivers/gpu/nvgpu/gk20a/gk20a.h

@@ -367,6 +367,7 @@ struct gpu_ops {
 		void (*enable_gpc_exceptions)(struct gk20a *g);
 		void (*enable_exceptions)(struct gk20a *g);
 		void (*create_gr_sysfs)(struct gk20a *g);
+		void (*remove_gr_sysfs)(struct gk20a *g);
 		u32 (*get_lrf_tex_ltc_dram_override)(struct gk20a *g);
 		int (*record_sm_error_state)(struct gk20a *g, u32 gpc, u32 tpc,
 				u32 sm, struct channel_gk20a *fault_ch);

drivers/gpu/nvgpu/gp106/hal_gp106.c

@@ -381,7 +381,10 @@ static const struct gpu_ops gp106_ops = {
 		.update_boosted_ctx = NULL,
 		.set_bes_crop_debug3 = gr_gp10b_set_bes_crop_debug3,
 		.set_bes_crop_debug4 = gr_gp10b_set_bes_crop_debug4,
+#ifdef CONFIG_SYSFS
 		.create_gr_sysfs = NULL,
+		.remove_gr_sysfs = NULL,
+#endif
 		.set_ctxsw_preemption_mode = gr_gp106_set_ctxsw_preemption_mode,
 		.load_ctxsw_ucode = gr_gm20b_load_ctxsw_ucode,
 		.fecs_host_int_enable = gr_gk20a_fecs_host_int_enable,

drivers/gpu/nvgpu/gp10b/gr_gp10b.h

@@ -60,6 +60,7 @@ int gr_gp10b_init_fs_state(struct gk20a *g);
 int gr_gp10b_alloc_buffer(struct vm_gk20a *vm, size_t size,
 			struct nvgpu_mem *mem);
 void gr_gp10b_create_sysfs(struct gk20a *g);
+void gr_gp10b_remove_sysfs(struct gk20a *g);
 int gr_gp10b_handle_fecs_error(struct gk20a *g,
 			struct channel_gk20a *__ch,
 			struct gr_gk20a_isr_data *isr_data);

drivers/gpu/nvgpu/gp10b/hal_gp10b.c

@@ -341,6 +341,7 @@ static const struct gpu_ops gp10b_ops = {
 		.set_bes_crop_debug3 = gr_gp10b_set_bes_crop_debug3,
 #ifdef CONFIG_SYSFS
 		.create_gr_sysfs = gr_gp10b_create_sysfs,
+		.remove_gr_sysfs = gr_gp10b_remove_sysfs,
 #endif
 		.set_ctxsw_preemption_mode = gr_gp10b_set_ctxsw_preemption_mode,
 		.init_ctxsw_hdr_data = gr_gp10b_init_ctxsw_hdr_data,

drivers/gpu/nvgpu/gv100/hal_gv100.c

@@ -413,6 +413,7 @@ static const struct gpu_ops gv100_ops = {
 		.set_bes_crop_debug4 = gr_gp10b_set_bes_crop_debug4,
 #ifdef CONFIG_SYSFS
 		.create_gr_sysfs = gr_gv11b_create_sysfs,
+		.remove_gr_sysfs = gr_gv11b_remove_sysfs,
 #endif
 		.set_ctxsw_preemption_mode = gr_gp10b_set_ctxsw_preemption_mode,
 		.is_etpc_addr = gv11b_gr_pri_is_etpc_addr,

drivers/gpu/nvgpu/gv11b/gr_gv11b.h

@@ -88,6 +88,7 @@ int gr_gv11b_alloc_buffer(struct vm_gk20a *vm, size_t size,
 void gr_gv11b_program_zcull_mapping(struct gk20a *g, u32 zcull_num_entries,
 					u32 *zcull_map_tiles);
 void gr_gv11b_create_sysfs(struct gk20a *g);
+void gr_gv11b_remove_sysfs(struct gk20a *g);
 
 bool gr_gv11b_is_valid_class(struct gk20a *g, u32 class_num);
 bool gr_gv11b_is_valid_gfx_class(struct gk20a *g, u32 class_num);

drivers/gpu/nvgpu/gv11b/hal_gv11b.c

@@ -370,6 +370,7 @@ static const struct gpu_ops gv11b_ops = {
 		.set_bes_crop_debug4 = gr_gp10b_set_bes_crop_debug4,
 #ifdef CONFIG_SYSFS
 		.create_gr_sysfs = gr_gv11b_create_sysfs,
+		.remove_gr_sysfs = gr_gv11b_remove_sysfs,
 #endif
 		.set_ctxsw_preemption_mode = gr_gv11b_set_ctxsw_preemption_mode,
 		.is_etpc_addr = gv11b_gr_pri_is_etpc_addr,

drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c

@@ -216,6 +216,7 @@ static const struct gpu_ops vgpu_gp10b_ops = {
 		.set_bes_crop_debug4 = gr_gp10b_set_bes_crop_debug4,
 #ifdef CONFIG_SYSFS
 		.create_gr_sysfs = gr_gp10b_create_sysfs,
+		.remove_gr_sysfs = gr_gp10b_remove_sysfs,
 #endif
 		.set_ctxsw_preemption_mode =
 					vgpu_gr_gp10b_set_ctxsw_preemption_mode,

drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c

@@ -232,6 +232,7 @@ static const struct gpu_ops vgpu_gv11b_ops = {
 		.set_bes_crop_debug4 = gr_gp10b_set_bes_crop_debug4,
 #ifdef CONFIG_SYSFS
 		.create_gr_sysfs = gr_gv11b_create_sysfs,
+		.remove_gr_sysfs = gr_gv11b_remove_sysfs,
 #endif
 		.set_ctxsw_preemption_mode = vgpu_gr_gp10b_set_ctxsw_preemption_mode,
 		.is_etpc_addr = gv11b_gr_pri_is_etpc_addr,