[PATCH AUTOSEL 6.5 24/41] drm/amdgpu: Store CU info from all XCCs for GFX v9.4.3

[PATCH AUTOSEL 6.5 24/41] drm/amdgpu: Store CU info from all XCCs for GFX v9.4.3

[Sasha Levin]

From: Mukul Joshi <mukul.joshi@amd.com>

[ Upstream commit 97e3c6a853f2af9145daf0c6ca25bcdf55c759d4 ]

Currently, we store CU info only for a single XCC assuming
that it is the same for all XCCs. However, that may not be
true. As a result, store CU info for all XCCs. This info is
later used for CU masking.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c    |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h       |  3 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c       |  2 +-
 drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c        |  2 +-
 drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c        |  2 +-
 drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c         |  2 +-
 drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c         |  2 +-
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c         |  2 +-
 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c         |  4 +-
 drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c       | 76 +++++++++----------
 drivers/gpu/drm/amd/amdkfd/kfd_crat.c         |  3 +-
 drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c  |  8 +-
 drivers/gpu/drm/amd/amdkfd/kfd_topology.c     | 11 ++-
 .../gpu/drm/amd/include/kgd_kfd_interface.h   |  6 +-
 14 files changed, 60 insertions(+), 65 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index b4fcad0e62f7e..a7c8beff1647c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -492,7 +492,7 @@ void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev, struct kfd_cu_info *c
 	cu_info->cu_active_number = acu_info.number;
 	cu_info->cu_ao_mask = acu_info.ao_cu_mask;
 	memcpy(&cu_info->cu_bitmap[0], &acu_info.bitmap[0],
-	       sizeof(acu_info.bitmap));
+	       sizeof(cu_info->cu_bitmap));
 	cu_info->num_shader_engines = adev->gfx.config.max_shader_engines;
 	cu_info->num_shader_arrays_per_engine = adev->gfx.config.max_sh_per_se;
 	cu_info->num_cu_per_sh = adev->gfx.config.max_cu_per_sh;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index a4ff515ce8966..59ba03d387fcc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -43,6 +43,7 @@
 #define AMDGPU_GFX_LBPW_DISABLED_MODE		0x00000008L
 
 #define AMDGPU_MAX_GC_INSTANCES		8
+#define KGD_MAX_QUEUES			128
 
 #define AMDGPU_MAX_GFX_QUEUES KGD_MAX_QUEUES
 #define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES
@@ -254,7 +255,7 @@ struct amdgpu_cu_info {
 	uint32_t number;
 	uint32_t ao_cu_mask;
 	uint32_t ao_cu_bitmap[4][4];
-	uint32_t bitmap[4][4];
+	uint32_t bitmap[AMDGPU_MAX_GC_INSTANCES][4][4];
 };
 
 struct amdgpu_gfx_ras {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
index d4ca19ba5a289..f678bdd5f353d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -839,7 +839,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		memcpy(&dev_info->cu_ao_bitmap[0], &adev->gfx.cu_info.ao_cu_bitmap[0],
 		       sizeof(adev->gfx.cu_info.ao_cu_bitmap));
 		memcpy(&dev_info->cu_bitmap[0], &adev->gfx.cu_info.bitmap[0],
-		       sizeof(adev->gfx.cu_info.bitmap));
+		       sizeof(dev_info->cu_bitmap));
 		dev_info->vram_type = adev->gmc.vram_type;
 		dev_info->vram_bit_width = adev->gmc.vram_width;
 		dev_info->vce_harvest_config = adev->vce.harvest_config;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index 44af8022b89fa..f743bf2c92877 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -9448,7 +9448,7 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 				gfx_v10_0_set_user_wgp_inactive_bitmap_per_sh(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v10_0_get_cu_active_bitmap_per_sh(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index 0451533ddde41..a82cba884c48f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -6394,7 +6394,7 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
 			 *    SE6: {SH0,SH1} --> {bitmap[2][2], bitmap[2][3]}
 			 *    SE7: {SH0,SH1} --> {bitmap[3][2], bitmap[3][3]}
 			 */
-			cu_info->bitmap[i % 4][j + (i / 4) * 2] = bitmap;
+			cu_info->bitmap[0][i % 4][j + (i / 4) * 2] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
index da6caff78c22b..34f9211b26793 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
@@ -3577,7 +3577,7 @@ static void gfx_v6_0_get_cu_info(struct amdgpu_device *adev)
 				gfx_v6_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v6_0_get_cu_enabled(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 8c174c11eaee0..6feae2548e8ee 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -5122,7 +5122,7 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 				gfx_v7_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v7_0_get_cu_active_bitmap(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
 				if (bitmap & mask) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index 51c1745c83697..885ebd703260f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -7121,7 +7121,7 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev)
 				gfx_v8_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v8_0_get_cu_active_bitmap(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
 				if (bitmap & mask) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index 65577eca58f1c..e511e49d3023f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -1497,7 +1497,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
 			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
-				if (cu_info->bitmap[i][j] & mask) {
+				if (cu_info->bitmap[0][i][j] & mask) {
 					if (counter == pg_always_on_cu_num)
 						WREG32_SOC15(GC, 0, mmRLC_PG_ALWAYS_ON_CU_MASK, cu_bitmap);
 					if (counter < always_on_cu_num)
@@ -7234,7 +7234,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 			 *    SE6,SH0 --> bitmap[2][1]
 			 *    SE7,SH0 --> bitmap[3][1]
 			 */
-			cu_info->bitmap[i % 4][j + i / 4] = bitmap;
+			cu_info->bitmap[0][i % 4][j + i / 4] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
 				if (bitmap & mask) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index 4f883b94f98ef..84a74a6c6b2de 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -4228,7 +4228,7 @@ static void gfx_v9_4_3_set_gds_init(struct amdgpu_device *adev)
 }
 
 static void gfx_v9_4_3_set_user_cu_inactive_bitmap(struct amdgpu_device *adev,
-						 u32 bitmap)
+						 u32 bitmap, int xcc_id)
 {
 	u32 data;
 
@@ -4238,15 +4238,15 @@ static void gfx_v9_4_3_set_user_cu_inactive_bitmap(struct amdgpu_device *adev,
 	data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
 	data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
 
-	WREG32_SOC15(GC, GET_INST(GC, 0), regGC_USER_SHADER_ARRAY_CONFIG, data);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regGC_USER_SHADER_ARRAY_CONFIG, data);
 }
 
-static u32 gfx_v9_4_3_get_cu_active_bitmap(struct amdgpu_device *adev)
+static u32 gfx_v9_4_3_get_cu_active_bitmap(struct amdgpu_device *adev, int xcc_id)
 {
 	u32 data, mask;
 
-	data = RREG32_SOC15(GC, GET_INST(GC, 0), regCC_GC_SHADER_ARRAY_CONFIG);
-	data |= RREG32_SOC15(GC, GET_INST(GC, 0), regGC_USER_SHADER_ARRAY_CONFIG);
+	data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCC_GC_SHADER_ARRAY_CONFIG);
+	data |= RREG32_SOC15(GC, GET_INST(GC, xcc_id), regGC_USER_SHADER_ARRAY_CONFIG);
 
 	data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
 	data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
@@ -4259,7 +4259,7 @@ static u32 gfx_v9_4_3_get_cu_active_bitmap(struct amdgpu_device *adev)
 static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
 				 struct amdgpu_cu_info *cu_info)
 {
-	int i, j, k, counter, active_cu_number = 0;
+	int i, j, k, counter, xcc_id, active_cu_number = 0;
 	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0;
 	unsigned disable_masks[4 * 4];
 
@@ -4278,46 +4278,38 @@ static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
 				    adev->gfx.config.max_sh_per_se);
 
 	mutex_lock(&adev->grbm_idx_mutex);
-	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
-		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			mask = 1;
-			ao_bitmap = 0;
-			counter = 0;
-			gfx_v9_4_3_xcc_select_se_sh(adev, i, j, 0xffffffff, 0);
-			gfx_v9_4_3_set_user_cu_inactive_bitmap(
-				adev, disable_masks[i * adev->gfx.config.max_sh_per_se + j]);
-			bitmap = gfx_v9_4_3_get_cu_active_bitmap(adev);
-
-			/*
-			 * The bitmap(and ao_cu_bitmap) in cu_info structure is
-			 * 4x4 size array, and it's usually suitable for Vega
-			 * ASICs which has 4*2 SE/SH layout.
-			 * But for Arcturus, SE/SH layout is changed to 8*1.
-			 * To mostly reduce the impact, we make it compatible
-			 * with current bitmap array as below:
-			 *    SE4,SH0 --> bitmap[0][1]
-			 *    SE5,SH0 --> bitmap[1][1]
-			 *    SE6,SH0 --> bitmap[2][1]
-			 *    SE7,SH0 --> bitmap[3][1]
-			 */
-			cu_info->bitmap[i % 4][j + i / 4] = bitmap;
-
-			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
-				if (bitmap & mask) {
-					if (counter < adev->gfx.config.max_cu_per_sh)
-						ao_bitmap |= mask;
-					counter++;
+	for (xcc_id = 0; xcc_id < NUM_XCC(adev->gfx.xcc_mask); xcc_id++) {
+		for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+			for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+				mask = 1;
+				ao_bitmap = 0;
+				counter = 0;
+				gfx_v9_4_3_xcc_select_se_sh(adev, i, j, 0xffffffff, xcc_id);
+				gfx_v9_4_3_set_user_cu_inactive_bitmap(
+					adev,
+					disable_masks[i * adev->gfx.config.max_sh_per_se + j],
+					xcc_id);
+				bitmap = gfx_v9_4_3_get_cu_active_bitmap(adev, xcc_id);
+
+				cu_info->bitmap[xcc_id][i][j] = bitmap;
+
+				for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
+					if (bitmap & mask) {
+						if (counter < adev->gfx.config.max_cu_per_sh)
+							ao_bitmap |= mask;
+						counter++;
+					}
+					mask <<= 1;
 				}
-				mask <<= 1;
+				active_cu_number += counter;
+				if (i < 2 && j < 2)
+					ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8));
+				cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
 			}
-			active_cu_number += counter;
-			if (i < 2 && j < 2)
-				ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8));
-			cu_info->ao_cu_bitmap[i % 4][j + i / 4] = ao_bitmap;
 		}
+		gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+					    xcc_id);
 	}
-	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
-				    0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index f5a6f562e2a80..11b9837292536 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -2154,7 +2154,8 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 
 	amdgpu_amdkfd_get_cu_info(kdev->adev, &cu_info);
 	cu->num_simd_per_cu = cu_info.simd_per_cu;
-	cu->num_simd_cores = cu_info.simd_per_cu * cu_info.cu_active_number;
+	cu->num_simd_cores = cu_info.simd_per_cu *
+			(cu_info.cu_active_number / kdev->kfd->num_nodes);
 	cu->max_waves_simd = cu_info.max_waves_per_simd;
 
 	cu->wave_front_size = cu_info.wave_front_size;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
index 863cf060af484..35e05ee89eac5 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
@@ -104,11 +104,13 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 	bool wgp_mode_req = KFD_GC_VERSION(mm->dev) >= IP_VERSION(10, 0, 0);
 	uint32_t en_mask = wgp_mode_req ? 0x3 : 0x1;
 	int i, se, sh, cu, cu_bitmap_sh_mul, inc = wgp_mode_req ? 2 : 1;
+	uint32_t cu_active_per_node;
 
 	amdgpu_amdkfd_get_cu_info(mm->dev->adev, &cu_info);
 
-	if (cu_mask_count > cu_info.cu_active_number)
-		cu_mask_count = cu_info.cu_active_number;
+	cu_active_per_node = cu_info.cu_active_number / mm->dev->kfd->num_nodes;
+	if (cu_mask_count > cu_active_per_node)
+		cu_mask_count = cu_active_per_node;
 
 	/* Exceeding these bounds corrupts the stack and indicates a coding error.
 	 * Returning with no CU's enabled will hang the queue, which should be
@@ -141,7 +143,7 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 	for (se = 0; se < cu_info.num_shader_engines; se++)
 		for (sh = 0; sh < cu_info.num_shader_arrays_per_engine; sh++)
 			cu_per_sh[se][sh] = hweight32(
-				cu_info.cu_bitmap[se % 4][sh + (se / 4) * cu_bitmap_sh_mul]);
+				cu_info.cu_bitmap[0][se % 4][sh + (se / 4) * cu_bitmap_sh_mul]);
 
 	/* Symmetrically map cu_mask to all SEs & SHs:
 	 * se_mask programs up to 2 SH in the upper and lower 16 bits.
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index 4a17bb7c7b27d..ea67a353beb00 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -450,8 +450,7 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr,
 	sysfs_show_32bit_prop(buffer, offs, "cpu_cores_count",
 			      dev->node_props.cpu_cores_count);
 	sysfs_show_32bit_prop(buffer, offs, "simd_count",
-			      dev->gpu ? (dev->node_props.simd_count *
-					  NUM_XCC(dev->gpu->xcc_mask)) : 0);
+			      dev->gpu ? dev->node_props.simd_count : 0);
 	sysfs_show_32bit_prop(buffer, offs, "mem_banks_count",
 			      dev->node_props.mem_banks_count);
 	sysfs_show_32bit_prop(buffer, offs, "caches_count",
@@ -1658,7 +1657,7 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 	int i, j, k;
 	struct kfd_cache_properties *pcache = NULL;
 
-	cu_sibling_map_mask = cu_info->cu_bitmap[0][0];
+	cu_sibling_map_mask = cu_info->cu_bitmap[0][0][0];
 	cu_sibling_map_mask &=
 		((1 << pcache_info[cache_type].num_cu_shared) - 1);
 	first_active_cu = ffs(cu_sibling_map_mask);
@@ -1701,7 +1700,7 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 				pcache->sibling_map[k+3] = (uint8_t)((cu_sibling_map_mask >> 24) & 0xFF);
 				k += 4;
 
-				cu_sibling_map_mask = cu_info->cu_bitmap[i % 4][j + i / 4];
+				cu_sibling_map_mask = cu_info->cu_bitmap[0][i % 4][j + i / 4];
 				cu_sibling_map_mask &= ((1 << pcache_info[cache_type].num_cu_shared) - 1);
 			}
 		}
@@ -1762,8 +1761,8 @@ static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct
 					for (k = 0; k < pcu_info->num_cu_per_sh; k += pcache_info[ct].num_cu_shared) {
 
 						ret = fill_in_l1_pcache(&props_ext, pcache_info, pcu_info,
-										pcu_info->cu_bitmap[i % 4][j + i / 4], ct,
-										cu_processor_id, k);
+									pcu_info->cu_bitmap[0][i % 4][j + i / 4], ct,
+									cu_processor_id, k);
 
 						if (ret < 0)
 							break;
diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
index d0df3381539f0..74cc545085a02 100644
--- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
+++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
@@ -31,12 +31,12 @@
 #include <linux/types.h>
 #include <linux/bitmap.h>
 #include <linux/dma-fence.h>
+#include "amdgpu_irq.h"
+#include "amdgpu_gfx.h"
 
 struct pci_dev;
 struct amdgpu_device;
 
-#define KGD_MAX_QUEUES 128
-
 struct kfd_dev;
 struct kgd_mem;
 
@@ -68,7 +68,7 @@ struct kfd_cu_info {
 	uint32_t wave_front_size;
 	uint32_t max_scratch_slots_per_cu;
 	uint32_t lds_size;
-	uint32_t cu_bitmap[4][4];
+	uint32_t cu_bitmap[AMDGPU_MAX_GC_INSTANCES][4][4];
 };
 
 /* For getting GPU local memory information from KGD */
-- 
2.40.1

[PATCH AUTOSEL 6.5 25/41] drm/amdkfd: Update cache info reporting for GFX v9.4.3

[Sasha Levin]

From: Mukul Joshi <mukul.joshi@amd.com>

[ Upstream commit 0752e66e91fa86fa5481b04b22053363833ffb85 ]

Update cache info reporting in sysfs to report the correct
number of CUs and associated cache information based on
different spatial partitioning modes.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdkfd/kfd_crat.h     |  4 ++
 drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 82 +++++++++++++----------
 drivers/gpu/drm/amd/amdkfd/kfd_topology.h |  2 +-
 3 files changed, 51 insertions(+), 37 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.h b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h
index fc719389b5d65..4684711aa695a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h
@@ -79,6 +79,10 @@ struct crat_header {
 #define CRAT_SUBTYPE_IOLINK_AFFINITY		5
 #define CRAT_SUBTYPE_MAX			6
 
+/*
+ * Do not change the value of CRAT_SIBLINGMAP_SIZE from 32
+ * as it breaks the ABI.
+ */
 #define CRAT_SIBLINGMAP_SIZE	32
 
 /*
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index ea67a353beb00..5582191022106 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1650,14 +1650,17 @@ static int fill_in_l1_pcache(struct kfd_cache_properties **props_ext,
 static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 				struct kfd_gpu_cache_info *pcache_info,
 				struct kfd_cu_info *cu_info,
-				int cache_type, unsigned int cu_processor_id)
+				int cache_type, unsigned int cu_processor_id,
+				struct kfd_node *knode)
 {
 	unsigned int cu_sibling_map_mask;
 	int first_active_cu;
-	int i, j, k;
+	int i, j, k, xcc, start, end;
 	struct kfd_cache_properties *pcache = NULL;
 
-	cu_sibling_map_mask = cu_info->cu_bitmap[0][0][0];
+	start = ffs(knode->xcc_mask) - 1;
+	end = start + NUM_XCC(knode->xcc_mask);
+	cu_sibling_map_mask = cu_info->cu_bitmap[start][0][0];
 	cu_sibling_map_mask &=
 		((1 << pcache_info[cache_type].num_cu_shared) - 1);
 	first_active_cu = ffs(cu_sibling_map_mask);
@@ -1692,16 +1695,18 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 		cu_sibling_map_mask = cu_sibling_map_mask >> (first_active_cu - 1);
 		k = 0;
 
-		for (i = 0; i < cu_info->num_shader_engines; i++) {
-			for (j = 0; j < cu_info->num_shader_arrays_per_engine; j++) {
-				pcache->sibling_map[k] = (uint8_t)(cu_sibling_map_mask & 0xFF);
-				pcache->sibling_map[k+1] = (uint8_t)((cu_sibling_map_mask >> 8) & 0xFF);
-				pcache->sibling_map[k+2] = (uint8_t)((cu_sibling_map_mask >> 16) & 0xFF);
-				pcache->sibling_map[k+3] = (uint8_t)((cu_sibling_map_mask >> 24) & 0xFF);
-				k += 4;
-
-				cu_sibling_map_mask = cu_info->cu_bitmap[0][i % 4][j + i / 4];
-				cu_sibling_map_mask &= ((1 << pcache_info[cache_type].num_cu_shared) - 1);
+		for (xcc = start; xcc < end; xcc++) {
+			for (i = 0; i < cu_info->num_shader_engines; i++) {
+				for (j = 0; j < cu_info->num_shader_arrays_per_engine; j++) {
+					pcache->sibling_map[k] = (uint8_t)(cu_sibling_map_mask & 0xFF);
+					pcache->sibling_map[k+1] = (uint8_t)((cu_sibling_map_mask >> 8) & 0xFF);
+					pcache->sibling_map[k+2] = (uint8_t)((cu_sibling_map_mask >> 16) & 0xFF);
+					pcache->sibling_map[k+3] = (uint8_t)((cu_sibling_map_mask >> 24) & 0xFF);
+					k += 4;
+
+					cu_sibling_map_mask = cu_info->cu_bitmap[xcc][i % 4][j + i / 4];
+					cu_sibling_map_mask &= ((1 << pcache_info[cache_type].num_cu_shared) - 1);
+				}
 			}
 		}
 		pcache->sibling_map_size = k;
@@ -1719,7 +1724,7 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct kfd_node *kdev)
 {
 	struct kfd_gpu_cache_info *pcache_info = NULL;
-	int i, j, k;
+	int i, j, k, xcc, start, end;
 	int ct = 0;
 	unsigned int cu_processor_id;
 	int ret;
@@ -1753,37 +1758,42 @@ static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct
 	 *			then it will consider only one CU from
 	 *			the shared unit
 	 */
+	start = ffs(kdev->xcc_mask) - 1;
+	end = start + NUM_XCC(kdev->xcc_mask);
+
 	for (ct = 0; ct < num_of_cache_types; ct++) {
 		cu_processor_id = gpu_processor_id;
 		if (pcache_info[ct].cache_level == 1) {
-			for (i = 0; i < pcu_info->num_shader_engines; i++) {
-				for (j = 0; j < pcu_info->num_shader_arrays_per_engine; j++) {
-					for (k = 0; k < pcu_info->num_cu_per_sh; k += pcache_info[ct].num_cu_shared) {
-
-						ret = fill_in_l1_pcache(&props_ext, pcache_info, pcu_info,
-									pcu_info->cu_bitmap[0][i % 4][j + i / 4], ct,
-									cu_processor_id, k);
-
-						if (ret < 0)
-							break;
-
-						if (!ret) {
-							num_of_entries++;
-							list_add_tail(&props_ext->list, &dev->cache_props);
+			for (xcc = start; xcc < end; xcc++) {
+				for (i = 0; i < pcu_info->num_shader_engines; i++) {
+					for (j = 0; j < pcu_info->num_shader_arrays_per_engine; j++) {
+						for (k = 0; k < pcu_info->num_cu_per_sh; k += pcache_info[ct].num_cu_shared) {
+
+							ret = fill_in_l1_pcache(&props_ext, pcache_info, pcu_info,
+										pcu_info->cu_bitmap[xcc][i % 4][j + i / 4], ct,
+										cu_processor_id, k);
+
+							if (ret < 0)
+								break;
+
+							if (!ret) {
+								num_of_entries++;
+								list_add_tail(&props_ext->list, &dev->cache_props);
+							}
+
+							/* Move to next CU block */
+							num_cu_shared = ((k + pcache_info[ct].num_cu_shared) <=
+								pcu_info->num_cu_per_sh) ?
+								pcache_info[ct].num_cu_shared :
+								(pcu_info->num_cu_per_sh - k);
+							cu_processor_id += num_cu_shared;
 						}
-
-						/* Move to next CU block */
-						num_cu_shared = ((k + pcache_info[ct].num_cu_shared) <=
-							pcu_info->num_cu_per_sh) ?
-							pcache_info[ct].num_cu_shared :
-							(pcu_info->num_cu_per_sh - k);
-						cu_processor_id += num_cu_shared;
 					}
 				}
 			}
 		} else {
 			ret = fill_in_l2_l3_pcache(&props_ext, pcache_info,
-								pcu_info, ct, cu_processor_id);
+					pcu_info, ct, cu_processor_id, kdev);
 
 			if (ret < 0)
 				break;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
index cba2cd5ed9d19..46927263e014d 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
@@ -86,7 +86,7 @@ struct kfd_mem_properties {
 	struct attribute	attr;
 };
 
-#define CACHE_SIBLINGMAP_SIZE 64
+#define CACHE_SIBLINGMAP_SIZE 128
 
 struct kfd_cache_properties {
 	struct list_head	list;
-- 
2.40.1

[PATCH AUTOSEL 6.5 26/41] drm/amdkfd: Update CU masking for GFX 9.4.3

[Sasha Levin]

From: Mukul Joshi <mukul.joshi@amd.com>

[ Upstream commit fc6efed2c728c9c10b058512fc9c1613f870a8e8 ]

The CU mask passed from user-space will change based on
different spatial partitioning mode. As a result, update
CU masking code for GFX9.4.3 to work for all partitioning
modes.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c  | 28 ++++++++---
 drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h  |  2 +-
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c  |  2 +-
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c  |  2 +-
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c  |  2 +-
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c   | 46 ++++++++++++-------
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c   |  2 +-
 7 files changed, 56 insertions(+), 28 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
index 35e05ee89eac5..254f343f967a3 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
@@ -97,14 +97,16 @@ void free_mqd_hiq_sdma(struct mqd_manager *mm, void *mqd,
 
 void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 		const uint32_t *cu_mask, uint32_t cu_mask_count,
-		uint32_t *se_mask)
+		uint32_t *se_mask, uint32_t inst)
 {
 	struct kfd_cu_info cu_info;
 	uint32_t cu_per_sh[KFD_MAX_NUM_SE][KFD_MAX_NUM_SH_PER_SE] = {0};
 	bool wgp_mode_req = KFD_GC_VERSION(mm->dev) >= IP_VERSION(10, 0, 0);
 	uint32_t en_mask = wgp_mode_req ? 0x3 : 0x1;
-	int i, se, sh, cu, cu_bitmap_sh_mul, inc = wgp_mode_req ? 2 : 1;
+	int i, se, sh, cu, cu_bitmap_sh_mul, cu_inc = wgp_mode_req ? 2 : 1;
 	uint32_t cu_active_per_node;
+	int inc = cu_inc * NUM_XCC(mm->dev->xcc_mask);
+	int xcc_inst = inst + ffs(mm->dev->xcc_mask) - 1;
 
 	amdgpu_amdkfd_get_cu_info(mm->dev->adev, &cu_info);
 
@@ -143,7 +145,8 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 	for (se = 0; se < cu_info.num_shader_engines; se++)
 		for (sh = 0; sh < cu_info.num_shader_arrays_per_engine; sh++)
 			cu_per_sh[se][sh] = hweight32(
-				cu_info.cu_bitmap[0][se % 4][sh + (se / 4) * cu_bitmap_sh_mul]);
+				cu_info.cu_bitmap[xcc_inst][se % 4][sh + (se / 4) *
+				cu_bitmap_sh_mul]);
 
 	/* Symmetrically map cu_mask to all SEs & SHs:
 	 * se_mask programs up to 2 SH in the upper and lower 16 bits.
@@ -166,20 +169,33 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 	 * cu_mask[0] bit8 -> se_mask[0] bit1 (SE0,SH0,CU1)
 	 * ...
 	 *
+	 * For GFX 9.4.3, the following code only looks at a
+	 * subset of the cu_mask corresponding to the inst parameter.
+	 * If we have n XCCs under one GPU node
+	 * cu_mask[0] bit0 -> XCC0 se_mask[0] bit0 (XCC0,SE0,SH0,CU0)
+	 * cu_mask[0] bit1 -> XCC1 se_mask[0] bit0 (XCC1,SE0,SH0,CU0)
+	 * ..
+	 * cu_mask[0] bitn -> XCCn se_mask[0] bit0 (XCCn,SE0,SH0,CU0)
+	 * cu_mask[0] bit n+1 -> XCC0 se_mask[1] bit0 (XCC0,SE1,SH0,CU0)
+	 *
+	 * For example, if there are 6 XCCs under 1 KFD node, this code
+	 * running for each inst, will look at the bits as:
+	 * inst, inst + 6, inst + 12...
+	 *
 	 * First ensure all CUs are disabled, then enable user specified CUs.
 	 */
 	for (i = 0; i < cu_info.num_shader_engines; i++)
 		se_mask[i] = 0;
 
-	i = 0;
-	for (cu = 0; cu < 16; cu += inc) {
+	i = inst;
+	for (cu = 0; cu < 16; cu += cu_inc) {
 		for (sh = 0; sh < cu_info.num_shader_arrays_per_engine; sh++) {
 			for (se = 0; se < cu_info.num_shader_engines; se++) {
 				if (cu_per_sh[se][sh] > cu) {
 					if (cu_mask[i / 32] & (en_mask << (i % 32)))
 						se_mask[se] |= en_mask << (cu + sh * 16);
 					i += inc;
-					if (i == cu_mask_count)
+					if (i >= cu_mask_count)
 						return;
 				}
 			}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
index 23158db7da035..57bf5e513f4d1 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
@@ -138,7 +138,7 @@ void free_mqd_hiq_sdma(struct mqd_manager *mm, void *mqd,
 
 void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 		const uint32_t *cu_mask, uint32_t cu_mask_count,
-		uint32_t *se_mask);
+		uint32_t *se_mask, uint32_t inst);
 
 int kfd_hiq_load_mqd_kiq(struct mqd_manager *mm, void *mqd,
 		uint32_t pipe_id, uint32_t queue_id,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
index 65c9f01a1f86c..faa01ee0d1655 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
@@ -52,7 +52,7 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
 
 	m = get_mqd(mqd);
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
index 94c0fc2e57b7f..0fcb176601295 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
@@ -52,7 +52,7 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
 
 	m = get_mqd(mqd);
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
index 23b30783dce31..97f754949ca92 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
@@ -71,7 +71,7 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 	}
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
 
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
 	m->compute_static_thread_mgmt_se1 = se_mask[1];
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
index 601bb9f68048c..a76ae27c8a919 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
@@ -60,7 +60,7 @@ static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd)
 }
 
 static void update_cu_mask(struct mqd_manager *mm, void *mqd,
-			struct mqd_update_info *minfo)
+			struct mqd_update_info *minfo, uint32_t inst)
 {
 	struct v9_mqd *m;
 	uint32_t se_mask[KFD_MAX_NUM_SE] = {0};
@@ -69,27 +69,36 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, inst);
 
 	m = get_mqd(mqd);
+
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
 	m->compute_static_thread_mgmt_se1 = se_mask[1];
 	m->compute_static_thread_mgmt_se2 = se_mask[2];
 	m->compute_static_thread_mgmt_se3 = se_mask[3];
-	m->compute_static_thread_mgmt_se4 = se_mask[4];
-	m->compute_static_thread_mgmt_se5 = se_mask[5];
-	m->compute_static_thread_mgmt_se6 = se_mask[6];
-	m->compute_static_thread_mgmt_se7 = se_mask[7];
-
-	pr_debug("update cu mask to %#x %#x %#x %#x %#x %#x %#x %#x\n",
-		m->compute_static_thread_mgmt_se0,
-		m->compute_static_thread_mgmt_se1,
-		m->compute_static_thread_mgmt_se2,
-		m->compute_static_thread_mgmt_se3,
-		m->compute_static_thread_mgmt_se4,
-		m->compute_static_thread_mgmt_se5,
-		m->compute_static_thread_mgmt_se6,
-		m->compute_static_thread_mgmt_se7);
+	if (KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 3)) {
+		m->compute_static_thread_mgmt_se4 = se_mask[4];
+		m->compute_static_thread_mgmt_se5 = se_mask[5];
+		m->compute_static_thread_mgmt_se6 = se_mask[6];
+		m->compute_static_thread_mgmt_se7 = se_mask[7];
+
+		pr_debug("update cu mask to %#x %#x %#x %#x %#x %#x %#x %#x\n",
+			m->compute_static_thread_mgmt_se0,
+			m->compute_static_thread_mgmt_se1,
+			m->compute_static_thread_mgmt_se2,
+			m->compute_static_thread_mgmt_se3,
+			m->compute_static_thread_mgmt_se4,
+			m->compute_static_thread_mgmt_se5,
+			m->compute_static_thread_mgmt_se6,
+			m->compute_static_thread_mgmt_se7);
+	} else {
+		pr_debug("inst: %u, update cu mask to %#x %#x %#x %#x\n",
+			inst, m->compute_static_thread_mgmt_se0,
+			m->compute_static_thread_mgmt_se1,
+			m->compute_static_thread_mgmt_se2,
+			m->compute_static_thread_mgmt_se3);
+	}
 }
 
 static void set_priority(struct v9_mqd *m, struct queue_properties *q)
@@ -290,7 +299,8 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 	if (mm->dev->kfd->cwsr_enabled && q->ctx_save_restore_area_address)
 		m->cp_hqd_ctx_save_control = 0;
 
-	update_cu_mask(mm, mqd, minfo);
+	if (KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 3))
+		update_cu_mask(mm, mqd, minfo, 0);
 	set_priority(m, q);
 
 	q->is_active = QUEUE_IS_ACTIVE(*q);
@@ -654,6 +664,8 @@ static void update_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
 		m = get_mqd(mqd + size * xcc);
 		update_mqd(mm, m, q, minfo);
 
+		update_cu_mask(mm, mqd, minfo, xcc);
+
 		if (q->format == KFD_QUEUE_FORMAT_AQL) {
 			switch (xcc) {
 			case 0:
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
index d1e962da51dd3..2551a7529b5e0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
@@ -55,7 +55,7 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
 
 	m = get_mqd(mqd);
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
-- 
2.40.1

[PATCH AUTOSEL 6.5 27/41] drm/amd/display: Add dirty rect support for Replay

[Sasha Levin]

From: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>

[ Upstream commit 679fc891bf11845730b572fc44f8a0eb846aba29 ]

Dirty rect can be used with replay, so enable them to allow for more
powersaving.

Reviewed-by: Sun peng Li <sunpeng.li@amd.com>
Acked-by: Stylon Wang <stylon.wang@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index 3123ea2f4f30a..5c63a13f25193 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -8038,7 +8038,8 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 		bundle->surface_updates[planes_count].plane_info =
 			&bundle->plane_infos[planes_count];
 
-		if (acrtc_state->stream->link->psr_settings.psr_feature_enabled) {
+		if (acrtc_state->stream->link->psr_settings.psr_feature_enabled ||
+		    acrtc_state->stream->link->replay_settings.replay_feature_enabled) {
 			fill_dc_dirty_rects(plane, old_plane_state,
 					    new_plane_state, new_crtc_state,
 					    &bundle->flip_addrs[planes_count],
-- 
2.40.1

[PATCH AUTOSEL 6.5 28/41] drm/amd/display: Don't check registers, if using AUX BL control

[Sasha Levin]

From: Swapnil Patel <swapnil.patel@amd.com>

[ Upstream commit f5b2c10b57615828b531bb0ae56bd6325a41167e ]

[Why]
Currently the driver looks DCN registers to access if BL is on or not.
This check is not valid if we are using AUX based brightness control.
This causes driver to not send out "backlight off" command during power off
sequence as it already thinks it is off.

[How]
Only check DCN registers if we aren't using AUX based brightness control.

Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Acked-by: Stylon Wang <stylon.wang@amd.com>
Signed-off-by: Swapnil Patel <swapnil.patel@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c
index 6966420dfbac3..15fa19ee748cf 100644
--- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c
+++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c
@@ -964,7 +964,9 @@ void dce110_edp_backlight_control(
 		return;
 	}
 
-	if (link->panel_cntl) {
+	if (link->panel_cntl && !(link->dpcd_sink_ext_caps.bits.oled ||
+		link->dpcd_sink_ext_caps.bits.hdr_aux_backlight_control == 1 ||
+		link->dpcd_sink_ext_caps.bits.sdr_aux_backlight_control == 1)) {
 		bool is_backlight_on = link->panel_cntl->funcs->is_panel_backlight_on(link->panel_cntl);
 
 		if ((enable && is_backlight_on) || (!enable && !is_backlight_on)) {
-- 
2.40.1

[PATCH AUTOSEL 6.5 29/41] drm/amdgpu/soc21: don't remap HDP registers for SR-IOV

[Sasha Levin]

From: Alex Deucher <alexander.deucher@amd.com>

[ Upstream commit 1832403cd41ca6b19b24e9d64f79cb08d920ca44 ]

This matches the behavior for soc15 and nv.

Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Timmy Tsai <timmtsai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdgpu/soc21.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/soc21.c b/drivers/gpu/drm/amd/amdgpu/soc21.c
index e5e5d68a4d702..1a5ffbf884891 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc21.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc21.c
@@ -786,7 +786,7 @@ static int soc21_common_hw_init(void *handle)
 	 * for the purpose of expose those registers
 	 * to process space
 	 */
-	if (adev->nbio.funcs->remap_hdp_registers)
+	if (adev->nbio.funcs->remap_hdp_registers && !amdgpu_sriov_vf(adev))
 		adev->nbio.funcs->remap_hdp_registers(adev);
 	/* enable the doorbell aperture */
 	adev->nbio.funcs->enable_doorbell_aperture(adev, true);
-- 
2.40.1

[PATCH AUTOSEL 6.5 30/41] drm/amdgpu/nbio4.3: set proper rmmio_remap.reg_offset for SR-IOV

[Sasha Levin]

From: Alex Deucher <alexander.deucher@amd.com>

[ Upstream commit ab43213e7afd08ac68d4282060bacf309e70fd14 ]

Needed for HDP flush to work correctly.

Reviewed-by: Timmy Tsai <timmtsai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c b/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c
index d5ed9e0e1a5f1..e5b5b0f4940f4 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c
@@ -345,6 +345,9 @@ static void nbio_v4_3_init_registers(struct amdgpu_device *adev)
 		data &= ~RCC_DEV0_EPF2_STRAP2__STRAP_NO_SOFT_RESET_DEV0_F2_MASK;
 		WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF2_STRAP2, data);
 	}
+	if (amdgpu_sriov_vf(adev))
+		adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0,
+			regBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
 }
 
 static u32 nbio_v4_3_get_rom_offset(struct amdgpu_device *adev)
-- 
2.40.1

[PATCH AUTOSEL 6.5 31/41] drm/amdgpu: fallback to old RAS error message for aqua_vanjaram

[Sasha Levin]

From: Hawking Zhang <Hawking.Zhang@amd.com>

[ Upstream commit ffd6bde302061aeee405ab364403af30210f0b99 ]

So driver doesn't generate incorrect message until
the new format is settled down for aqua_vanjaram

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 8aaa427f8c0f6..7d5019a884024 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -1061,7 +1061,8 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
 	info->ce_count = obj->err_data.ce_count;
 
 	if (err_data.ce_count) {
-		if (adev->smuio.funcs &&
+		if (!adev->aid_mask &&
+		    adev->smuio.funcs &&
 		    adev->smuio.funcs->get_socket_id &&
 		    adev->smuio.funcs->get_die_id) {
 			dev_info(adev->dev, "socket: %d, die: %d "
@@ -1081,7 +1082,8 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
 		}
 	}
 	if (err_data.ue_count) {
-		if (adev->smuio.funcs &&
+		if (!adev->aid_mask &&
+		    adev->smuio.funcs &&
 		    adev->smuio.funcs->get_socket_id &&
 		    adev->smuio.funcs->get_die_id) {
 			dev_info(adev->dev, "socket: %d, die: %d "
-- 
2.40.1

[PATCH AUTOSEL 6.5 32/41] drm/amdkfd: Checkpoint and restore queues on GFX11

[Sasha Levin]

From: David Francis <David.Francis@amd.com>

[ Upstream commit 9296da8c40900b4dae3d973aa22be306e2a77671 ]

The code in kfd_mqd_manager_v11.c to support criu dump and
restore of queue state was missing.

Added it; should be equivalent to kfd_mqd_manager_v10.c.

CC: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: David Francis <David.Francis@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c  | 41 +++++++++++++++++++
 1 file changed, 41 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
index 97f754949ca92..352757f2d3202 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
@@ -321,6 +321,43 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
 	return 0;
 }
 
+static void checkpoint_mqd(struct mqd_manager *mm, void *mqd, void *mqd_dst, void *ctl_stack_dst)
+{
+	struct v11_compute_mqd *m;
+
+	m = get_mqd(mqd);
+
+	memcpy(mqd_dst, m, sizeof(struct v11_compute_mqd));
+}
+
+static void restore_mqd(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *qp,
+			const void *mqd_src,
+			const void *ctl_stack_src, const u32 ctl_stack_size)
+{
+	uint64_t addr;
+	struct v11_compute_mqd *m;
+
+	m = (struct v11_compute_mqd *) mqd_mem_obj->cpu_ptr;
+	addr = mqd_mem_obj->gpu_addr;
+
+	memcpy(m, mqd_src, sizeof(*m));
+
+	*mqd = m;
+	if (gart_addr)
+		*gart_addr = addr;
+
+	m->cp_hqd_pq_doorbell_control =
+		qp->doorbell_off <<
+			CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+	pr_debug("cp_hqd_pq_doorbell_control 0x%x\n",
+			m->cp_hqd_pq_doorbell_control);
+
+	qp->is_active = 0;
+}
+
+
 static void init_mqd_hiq(struct mqd_manager *mm, void **mqd,
 			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
 			struct queue_properties *q)
@@ -438,6 +475,8 @@ struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type,
 		mqd->mqd_size = sizeof(struct v11_compute_mqd);
 		mqd->get_wave_state = get_wave_state;
 		mqd->mqd_stride = kfd_mqd_stride;
+		mqd->checkpoint_mqd = checkpoint_mqd;
+		mqd->restore_mqd = restore_mqd;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
@@ -482,6 +521,8 @@ struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type,
 		mqd->update_mqd = update_mqd_sdma;
 		mqd->destroy_mqd = kfd_destroy_mqd_sdma;
 		mqd->is_occupied = kfd_is_occupied_sdma;
+		mqd->checkpoint_mqd = checkpoint_mqd;
+		mqd->restore_mqd = restore_mqd;
 		mqd->mqd_size = sizeof(struct v11_sdma_mqd);
 		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
-- 
2.40.1

[PATCH AUTOSEL 6.5 33/41] drm/amdgpu: Handle null atom context in VBIOS info ioctl

[Sasha Levin]

From: David Francis <David.Francis@amd.com>

[ Upstream commit 5e7e82254270c8cf8b107451c5de01cee2f135ae ]

On some APU systems, there is no atom context and so the
atom_context struct is null.

Add a check to the VBIOS_INFO branch of amdgpu_info_ioctl
to handle this case, returning all zeroes.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: David Francis <David.Francis@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
index f678bdd5f353d..b9fc7e2db5e59 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -940,12 +940,17 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 			struct atom_context *atom_context;
 
 			atom_context = adev->mode_info.atom_context;
-			memcpy(vbios_info.name, atom_context->name, sizeof(atom_context->name));
-			memcpy(vbios_info.vbios_pn, atom_context->vbios_pn, sizeof(atom_context->vbios_pn));
-			vbios_info.version = atom_context->version;
-			memcpy(vbios_info.vbios_ver_str, atom_context->vbios_ver_str,
-						sizeof(atom_context->vbios_ver_str));
-			memcpy(vbios_info.date, atom_context->date, sizeof(atom_context->date));
+			if (atom_context) {
+				memcpy(vbios_info.name, atom_context->name,
+				       sizeof(atom_context->name));
+				memcpy(vbios_info.vbios_pn, atom_context->vbios_pn,
+				       sizeof(atom_context->vbios_pn));
+				vbios_info.version = atom_context->version;
+				memcpy(vbios_info.vbios_ver_str, atom_context->vbios_ver_str,
+				       sizeof(atom_context->vbios_ver_str));
+				memcpy(vbios_info.date, atom_context->date,
+				       sizeof(atom_context->date));
+			}
 
 			return copy_to_user(out, &vbios_info,
 						min((size_t)size, sizeof(vbios_info))) ? -EFAULT : 0;
-- 
2.40.1