Merge drm/drm-next into drm-intel-next

Backmerge to sync up with drm-intel-gt-next and drm-misc-next.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>

15 files changed, 156 insertions, 72 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index 86fb7ac7982a..f76b7aee5c0a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -2087,7 +2087,8 @@ 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_crat.h b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h
index 387a8ef49385..74c2d7a0d628 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_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index b166f30f083e..8a6cb41444a4 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -1677,8 +1677,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 			dqm->dev->kfd2kgd->build_grace_period_packet_info(
 					dqm->dev->adev,	dqm->wait_times,
 					grace_period, &reg_offset,
-					&dqm->wait_times,
-					ffs(dqm->dev->xcc_mask) - 1);
+					&dqm->wait_times);
 	}
 
 	dqm_unlock(dqm);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
index c2e0b79dcc6d..7b38537c7c99 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
@@ -162,6 +162,7 @@ void __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd,
 		return NULL;
 
 	*doorbell_off = amdgpu_doorbell_index_on_bar(kfd->adev, kfd->doorbells, inx);
+	inx *= 2;
 
 	pr_debug("Get kernel queue doorbell\n"
 			"     doorbell offset   == 0x%08X\n"
@@ -176,6 +177,7 @@ void kfd_release_kernel_doorbell(struct kfd_dev *kfd, u32 __iomem *db_addr)
 	unsigned int inx;
 
 	inx = (unsigned int)(db_addr - kfd->doorbell_kernel_ptr);
+	inx /= 2;
 
 	mutex_lock(&kfd->doorbell_mutex);
 	__clear_bit(inx, kfd->doorbell_bitmap);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
index d01bb57733b3..447829c22295 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
@@ -97,18 +97,22 @@ 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);
 
-	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 +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[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.
@@ -164,20 +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 23158db7da03..57bf5e513f4d 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 ee1d32d957f2..1a4a69943c71 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 83699392c808..8b7fed913526 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 0bbf0edbabd4..15277f1d5cf0 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];
@@ -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)
@@ -458,6 +495,8 @@ 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
@@ -502,6 +541,8 @@ 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)
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 e23d32f35607..42d881809dc7 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);
@@ -676,6 +686,8 @@ 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 657c37822980..3e1a574d4ea6 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];
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
index 8ce6f5200905..1a03173e2313 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
@@ -299,8 +299,7 @@ static int pm_set_grace_period_v9(struct packet_manager *pm,
 			pm->dqm->wait_times,
 			grace_period,
 			&reg_offset,
-			&reg_data,
-			0);
+			&reg_data);
 
 	if (grace_period == USE_DEFAULT_GRACE_PERIOD)
 		reg_data = pm->dqm->wait_times;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 3d9ce44d88da..fa24e1852493 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -1466,8 +1466,7 @@ void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type);
 
 static inline bool kfd_flush_tlb_after_unmap(struct kfd_dev *dev)
 {
-	return KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3) ||
-	       KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2) ||
+	return KFD_GC_VERSION(dev) > IP_VERSION(9, 4, 2) ||
 	       (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 1) && dev->sdma_fw_version >= 18) ||
 	       KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 0);
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index ff98fded9534..c8c75ff7cea8 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",
@@ -1597,14 +1596,17 @@ 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];
+	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);
@@ -1639,16 +1641,18 @@ 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[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;
@@ -1666,7 +1670,7 @@ 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;
@@ -1700,37 +1704,42 @@ 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) {
+			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[i % 4][j + i / 4], ct,
+							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 < 0)
+								break;
 
-						if (!ret) {
-							num_of_entries++;
-							list_add_tail(&props_ext->list, &dev->cache_props);
-						}
+							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 dea32a9e5506..27386ce9a021 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
@@ -89,7 +89,7 @@ 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;