1 files changed, 607 insertions, 279 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c b/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c
index d0fc62784e82..f9e2edf5260b 100644
--- a/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c
+++ b/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c
@@ -24,16 +24,16 @@
 #include "soc15.h"
 
 #include "soc15_common.h"
+#include "amdgpu_reg_state.h"
 #include "amdgpu_xcp.h"
 #include "gfx_v9_4_3.h"
 #include "gfxhub_v1_2.h"
 #include "sdma_v4_4_2.h"
+#include "amdgpu_ip.h"
 
 #define XCP_INST_MASK(num_inst, xcp_id)                                        \
 	(num_inst ? GENMASK(num_inst - 1, 0) << (xcp_id * num_inst) : 0)
 
-#define AMDGPU_XCP_OPS_KFD	(1 << 0)
-
 void aqua_vanjaram_doorbell_index_init(struct amdgpu_device *adev)
 {
 	int i;
@@ -61,222 +61,6 @@ void aqua_vanjaram_doorbell_index_init(struct amdgpu_device *adev)
 	adev->doorbell_index.max_assignment = AMDGPU_DOORBELL_LAYOUT1_MAX_ASSIGNMENT << 1;
 }
 
-static void aqua_vanjaram_set_xcp_id(struct amdgpu_device *adev,
-			     uint32_t inst_idx, struct amdgpu_ring *ring)
-{
-	int xcp_id;
-	enum AMDGPU_XCP_IP_BLOCK ip_blk;
-	uint32_t inst_mask;
-
-	ring->xcp_id = AMDGPU_XCP_NO_PARTITION;
-	if (adev->xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
-		return;
-
-	inst_mask = 1 << inst_idx;
-
-	switch (ring->funcs->type) {
-	case AMDGPU_HW_IP_GFX:
-	case AMDGPU_RING_TYPE_COMPUTE:
-	case AMDGPU_RING_TYPE_KIQ:
-		ip_blk = AMDGPU_XCP_GFX;
-		break;
-	case AMDGPU_RING_TYPE_SDMA:
-		ip_blk = AMDGPU_XCP_SDMA;
-		break;
-	case AMDGPU_RING_TYPE_VCN_ENC:
-	case AMDGPU_RING_TYPE_VCN_JPEG:
-		ip_blk = AMDGPU_XCP_VCN;
-		if (adev->xcp_mgr->mode == AMDGPU_CPX_PARTITION_MODE)
-			inst_mask = 1 << (inst_idx * 2);
-		break;
-	default:
-		DRM_ERROR("Not support ring type %d!", ring->funcs->type);
-		return;
-	}
-
-	for (xcp_id = 0; xcp_id < adev->xcp_mgr->num_xcps; xcp_id++) {
-		if (adev->xcp_mgr->xcp[xcp_id].ip[ip_blk].inst_mask & inst_mask) {
-			ring->xcp_id = xcp_id;
-			break;
-		}
-	}
-}
-
-static void aqua_vanjaram_xcp_gpu_sched_update(
-		struct amdgpu_device *adev,
-		struct amdgpu_ring *ring,
-		unsigned int sel_xcp_id)
-{
-	unsigned int *num_gpu_sched;
-
-	num_gpu_sched = &adev->xcp_mgr->xcp[sel_xcp_id]
-			.gpu_sched[ring->funcs->type][ring->hw_prio].num_scheds;
-	adev->xcp_mgr->xcp[sel_xcp_id].gpu_sched[ring->funcs->type][ring->hw_prio]
-			.sched[(*num_gpu_sched)++] = &ring->sched;
-	DRM_DEBUG("%s :[%d] gpu_sched[%d][%d] = %d", ring->name,
-			sel_xcp_id, ring->funcs->type,
-			ring->hw_prio, *num_gpu_sched);
-}
-
-static int aqua_vanjaram_xcp_sched_list_update(
-		struct amdgpu_device *adev)
-{
-	struct amdgpu_ring *ring;
-	int i;
-
-	for (i = 0; i < MAX_XCP; i++) {
-		atomic_set(&adev->xcp_mgr->xcp[i].ref_cnt, 0);
-		memset(adev->xcp_mgr->xcp[i].gpu_sched, 0, sizeof(adev->xcp_mgr->xcp->gpu_sched));
-	}
-
-	if (adev->xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
-		return 0;
-
-	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
-		ring = adev->rings[i];
-		if (!ring || !ring->sched.ready || ring->no_scheduler)
-			continue;
-
-		aqua_vanjaram_xcp_gpu_sched_update(adev, ring, ring->xcp_id);
-
-		/* VCN is shared by two partitions under CPX MODE */
-		if ((ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC ||
-			ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) &&
-			adev->xcp_mgr->mode == AMDGPU_CPX_PARTITION_MODE)
-			aqua_vanjaram_xcp_gpu_sched_update(adev, ring, ring->xcp_id + 1);
-	}
-
-	return 0;
-}
-
-static int aqua_vanjaram_update_partition_sched_list(struct amdgpu_device *adev)
-{
-	int i;
-
-	for (i = 0; i < adev->num_rings; i++) {
-		struct amdgpu_ring *ring = adev->rings[i];
-
-		if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE ||
-			ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
-			aqua_vanjaram_set_xcp_id(adev, ring->xcc_id, ring);
-		else
-			aqua_vanjaram_set_xcp_id(adev, ring->me, ring);
-	}
-
-	return aqua_vanjaram_xcp_sched_list_update(adev);
-}
-
-static int aqua_vanjaram_select_scheds(
-		struct amdgpu_device *adev,
-		u32 hw_ip,
-		u32 hw_prio,
-		struct amdgpu_fpriv *fpriv,
-		unsigned int *num_scheds,
-		struct drm_gpu_scheduler ***scheds)
-{
-	u32 sel_xcp_id;
-	int i;
-
-	if (fpriv->xcp_id == AMDGPU_XCP_NO_PARTITION) {
-		u32 least_ref_cnt = ~0;
-
-		fpriv->xcp_id = 0;
-		for (i = 0; i < adev->xcp_mgr->num_xcps; i++) {
-			u32 total_ref_cnt;
-
-			total_ref_cnt = atomic_read(&adev->xcp_mgr->xcp[i].ref_cnt);
-			if (total_ref_cnt < least_ref_cnt) {
-				fpriv->xcp_id = i;
-				least_ref_cnt = total_ref_cnt;
-			}
-		}
-	}
-	sel_xcp_id = fpriv->xcp_id;
-
-	if (adev->xcp_mgr->xcp[sel_xcp_id].gpu_sched[hw_ip][hw_prio].num_scheds) {
-		*num_scheds = adev->xcp_mgr->xcp[fpriv->xcp_id].gpu_sched[hw_ip][hw_prio].num_scheds;
-		*scheds = adev->xcp_mgr->xcp[fpriv->xcp_id].gpu_sched[hw_ip][hw_prio].sched;
-		atomic_inc(&adev->xcp_mgr->xcp[sel_xcp_id].ref_cnt);
-		DRM_DEBUG("Selected partition #%d", sel_xcp_id);
-	} else {
-		DRM_ERROR("Failed to schedule partition #%d.", sel_xcp_id);
-		return -ENOENT;
-	}
-
-	return 0;
-}
-
-static int8_t aqua_vanjaram_logical_to_dev_inst(struct amdgpu_device *adev,
-					 enum amd_hw_ip_block_type block,
-					 int8_t inst)
-{
-	int8_t dev_inst;
-
-	switch (block) {
-	case GC_HWIP:
-	case SDMA0_HWIP:
-	/* Both JPEG and VCN as JPEG is only alias of VCN */
-	case VCN_HWIP:
-		dev_inst = adev->ip_map.dev_inst[block][inst];
-		break;
-	default:
-		/* For rest of the IPs, no look up required.
-		 * Assume 'logical instance == physical instance' for all configs. */
-		dev_inst = inst;
-		break;
-	}
-
-	return dev_inst;
-}
-
-static uint32_t aqua_vanjaram_logical_to_dev_mask(struct amdgpu_device *adev,
-					 enum amd_hw_ip_block_type block,
-					 uint32_t mask)
-{
-	uint32_t dev_mask = 0;
-	int8_t log_inst, dev_inst;
-
-	while (mask) {
-		log_inst = ffs(mask) - 1;
-		dev_inst = aqua_vanjaram_logical_to_dev_inst(adev, block, log_inst);
-		dev_mask |= (1 << dev_inst);
-		mask &= ~(1 << log_inst);
-	}
-
-	return dev_mask;
-}
-
-static void aqua_vanjaram_populate_ip_map(struct amdgpu_device *adev,
-					  enum amd_hw_ip_block_type ip_block,
-					  uint32_t inst_mask)
-{
-	int l = 0, i;
-
-	while (inst_mask) {
-		i = ffs(inst_mask) - 1;
-		adev->ip_map.dev_inst[ip_block][l++] = i;
-		inst_mask &= ~(1 << i);
-	}
-	for (; l < HWIP_MAX_INSTANCE; l++)
-		adev->ip_map.dev_inst[ip_block][l] = -1;
-}
-
-void aqua_vanjaram_ip_map_init(struct amdgpu_device *adev)
-{
-	u32 ip_map[][2] = {
-		{ GC_HWIP, adev->gfx.xcc_mask },
-		{ SDMA0_HWIP, adev->sdma.sdma_mask },
-		{ VCN_HWIP, adev->vcn.inst_mask },
-	};
-	int i;
-
-	for (i = 0; i < ARRAY_SIZE(ip_map); ++i)
-		aqua_vanjaram_populate_ip_map(adev, ip_map[i][0], ip_map[i][1]);
-
-	adev->ip_map.logical_to_dev_inst = aqua_vanjaram_logical_to_dev_inst;
-	adev->ip_map.logical_to_dev_mask = aqua_vanjaram_logical_to_dev_mask;
-}
-
 /* Fixed pattern for smn addressing on different AIDs:
  *   bit[34]: indicate cross AID access
  *   bit[33:32]: indicate target AID id
@@ -296,13 +80,59 @@ u64 aqua_vanjaram_encode_ext_smn_addressing(int ext_id)
 	return ext_offset;
 }
 
+static enum amdgpu_gfx_partition
+__aqua_vanjaram_calc_xcp_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_xcc, num_xcc_per_xcp = 0, mode = 0;
+
+	num_xcc = NUM_XCC(xcp_mgr->adev->gfx.xcc_mask);
+	if (adev->gfx.funcs->get_xccs_per_xcp)
+		num_xcc_per_xcp = adev->gfx.funcs->get_xccs_per_xcp(adev);
+	if ((num_xcc_per_xcp) && (num_xcc % num_xcc_per_xcp == 0))
+		mode = num_xcc / num_xcc_per_xcp;
+
+	if (num_xcc_per_xcp == 1)
+		return AMDGPU_CPX_PARTITION_MODE;
+
+	switch (mode) {
+	case 1:
+		return AMDGPU_SPX_PARTITION_MODE;
+	case 2:
+		return AMDGPU_DPX_PARTITION_MODE;
+	case 3:
+		return AMDGPU_TPX_PARTITION_MODE;
+	case 4:
+		return AMDGPU_QPX_PARTITION_MODE;
+	default:
+		return AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+	}
+
+	return AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+}
+
 static int aqua_vanjaram_query_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr)
 {
-	enum amdgpu_gfx_partition mode = AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+	enum amdgpu_gfx_partition derv_mode,
+		mode = AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
 	struct amdgpu_device *adev = xcp_mgr->adev;
 
-	if (adev->nbio.funcs->get_compute_partition_mode)
+	derv_mode = __aqua_vanjaram_calc_xcp_mode(xcp_mgr);
+
+	if (amdgpu_sriov_vf(adev))
+		return derv_mode;
+
+	if (adev->nbio.funcs->get_compute_partition_mode) {
 		mode = adev->nbio.funcs->get_compute_partition_mode(adev);
+		if (mode != derv_mode) {
+			dev_warn(
+				adev->dev,
+				"Mismatch in compute partition mode - reported : %d derived : %d",
+				mode, derv_mode);
+			if (derv_mode == AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE)
+				amdgpu_device_bus_status_check(adev);
+		}
+	}
 
 	return mode;
 }
@@ -339,38 +169,31 @@ static int __aqua_vanjaram_get_xcp_ip_info(struct amdgpu_xcp_mgr *xcp_mgr, int x
 				    struct amdgpu_xcp_ip *ip)
 {
 	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_sdma, num_vcn, num_shared_vcn, num_xcp;
 	int num_xcc_xcp, num_sdma_xcp, num_vcn_xcp;
-	int num_sdma, num_vcn;
 
 	num_sdma = adev->sdma.num_instances;
 	num_vcn = adev->vcn.num_vcn_inst;
+	num_shared_vcn = 1;
+
+	num_xcc_xcp = adev->gfx.num_xcc_per_xcp;
+	num_xcp = NUM_XCC(adev->gfx.xcc_mask) / num_xcc_xcp;
 
 	switch (xcp_mgr->mode) {
 	case AMDGPU_SPX_PARTITION_MODE:
-		num_sdma_xcp = num_sdma;
-		num_vcn_xcp = num_vcn;
-		break;
 	case AMDGPU_DPX_PARTITION_MODE:
-		num_sdma_xcp = num_sdma / 2;
-		num_vcn_xcp = num_vcn / 2;
-		break;
 	case AMDGPU_TPX_PARTITION_MODE:
-		num_sdma_xcp = num_sdma / 3;
-		num_vcn_xcp = num_vcn / 3;
-		break;
 	case AMDGPU_QPX_PARTITION_MODE:
-		num_sdma_xcp = num_sdma / 4;
-		num_vcn_xcp = num_vcn / 4;
-		break;
 	case AMDGPU_CPX_PARTITION_MODE:
-		num_sdma_xcp = 2;
-		num_vcn_xcp = num_vcn ? 1 : 0;
+		num_sdma_xcp = DIV_ROUND_UP(num_sdma, num_xcp);
+		num_vcn_xcp = DIV_ROUND_UP(num_vcn, num_xcp);
 		break;
 	default:
 		return -EINVAL;
 	}
 
-	num_xcc_xcp = adev->gfx.num_xcc_per_xcp;
+	if (num_vcn && num_xcp > num_vcn)
+		num_shared_vcn = num_xcp / num_vcn;
 
 	switch (ip_id) {
 	case AMDGPU_XCP_GFXHUB:
@@ -386,7 +209,8 @@ static int __aqua_vanjaram_get_xcp_ip_info(struct amdgpu_xcp_mgr *xcp_mgr, int x
 		ip->ip_funcs = &sdma_v4_4_2_xcp_funcs;
 		break;
 	case AMDGPU_XCP_VCN:
-		ip->inst_mask = XCP_INST_MASK(num_vcn_xcp, xcp_id);
+		ip->inst_mask =
+			XCP_INST_MASK(num_vcn_xcp, xcp_id / num_shared_vcn);
 		/* TODO : Assign IP funcs */
 		break;
 	default:
@@ -398,6 +222,94 @@ static int __aqua_vanjaram_get_xcp_ip_info(struct amdgpu_xcp_mgr *xcp_mgr, int x
 	return 0;
 }
 
+static int __aqua_vanjaram_get_px_mode_info(struct amdgpu_xcp_mgr *xcp_mgr,
+					    int px_mode, int *num_xcp,
+					    uint16_t *nps_modes)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	uint32_t gc_ver = amdgpu_ip_version(adev, GC_HWIP, 0);
+
+	if (!num_xcp || !nps_modes || !(xcp_mgr->supp_xcp_modes & BIT(px_mode)))
+		return -EINVAL;
+
+	switch (px_mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		*num_xcp = 1;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE);
+		break;
+	case AMDGPU_DPX_PARTITION_MODE:
+		*num_xcp = 2;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	case AMDGPU_TPX_PARTITION_MODE:
+		*num_xcp = 3;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		break;
+	case AMDGPU_QPX_PARTITION_MODE:
+		*num_xcp = 4;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		if (gc_ver == IP_VERSION(9, 5, 0))
+			*nps_modes |= BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	case AMDGPU_CPX_PARTITION_MODE:
+		*num_xcp = NUM_XCC(adev->gfx.xcc_mask);
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		if (gc_ver == IP_VERSION(9, 5, 0))
+			*nps_modes |= BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int aqua_vanjaram_get_xcp_res_info(struct amdgpu_xcp_mgr *xcp_mgr,
+					  int mode,
+					  struct amdgpu_xcp_cfg *xcp_cfg)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int max_res[AMDGPU_XCP_RES_MAX] = {};
+	bool res_lt_xcp;
+	int num_xcp, i, r;
+	u16 nps_modes;
+
+	if (!(xcp_mgr->supp_xcp_modes & BIT(mode)))
+		return -EINVAL;
+
+	max_res[AMDGPU_XCP_RES_XCC] = NUM_XCC(adev->gfx.xcc_mask);
+	max_res[AMDGPU_XCP_RES_DMA] = adev->sdma.num_instances;
+	max_res[AMDGPU_XCP_RES_DEC] = adev->vcn.num_vcn_inst;
+	max_res[AMDGPU_XCP_RES_JPEG] = adev->jpeg.num_jpeg_inst;
+
+	r = __aqua_vanjaram_get_px_mode_info(xcp_mgr, mode, &num_xcp, &nps_modes);
+	if (r)
+		return r;
+
+	xcp_cfg->compatible_nps_modes =
+		(adev->gmc.supported_nps_modes & nps_modes);
+	xcp_cfg->num_res = ARRAY_SIZE(max_res);
+
+	for (i = 0; i < xcp_cfg->num_res; i++) {
+		res_lt_xcp = max_res[i] < num_xcp;
+		xcp_cfg->xcp_res[i].id = i;
+		xcp_cfg->xcp_res[i].num_inst =
+			res_lt_xcp ? 1 : max_res[i] / num_xcp;
+		xcp_cfg->xcp_res[i].num_inst =
+			i == AMDGPU_XCP_RES_JPEG ?
+			xcp_cfg->xcp_res[i].num_inst *
+			adev->jpeg.num_jpeg_rings : xcp_cfg->xcp_res[i].num_inst;
+		xcp_cfg->xcp_res[i].num_shared =
+			res_lt_xcp ? num_xcp / max_res[i] : 1;
+	}
+
+	return 0;
+}
+
 static enum amdgpu_gfx_partition
 __aqua_vanjaram_get_auto_mode(struct amdgpu_xcp_mgr *xcp_mgr)
 {
@@ -414,7 +326,7 @@ __aqua_vanjaram_get_auto_mode(struct amdgpu_xcp_mgr *xcp_mgr)
 
 	if (adev->gmc.num_mem_partitions == num_xcc / 2)
 		return (adev->flags & AMD_IS_APU) ? AMDGPU_TPX_PARTITION_MODE :
-						    AMDGPU_QPX_PARTITION_MODE;
+						    AMDGPU_CPX_PARTITION_MODE;
 
 	if (adev->gmc.num_mem_partitions == 2 && !(adev->flags & AMD_IS_APU))
 		return AMDGPU_DPX_PARTITION_MODE;
@@ -426,27 +338,31 @@ static bool __aqua_vanjaram_is_valid_mode(struct amdgpu_xcp_mgr *xcp_mgr,
 					  enum amdgpu_gfx_partition mode)
 {
 	struct amdgpu_device *adev = xcp_mgr->adev;
-	int num_xcc, num_xccs_per_xcp;
+	int num_xcc, num_xccs_per_xcp, r;
+	int num_xcp, nps_mode;
+	u16 supp_nps_modes;
+	bool comp_mode;
+
+	nps_mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	r = __aqua_vanjaram_get_px_mode_info(xcp_mgr, mode, &num_xcp,
+					       &supp_nps_modes);
+	if (r)
+		return false;
 
+	comp_mode = !!(BIT(nps_mode) & supp_nps_modes);
 	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
 	switch (mode) {
 	case AMDGPU_SPX_PARTITION_MODE:
-		return adev->gmc.num_mem_partitions == 1 && num_xcc > 0;
+		return comp_mode && num_xcc > 0;
 	case AMDGPU_DPX_PARTITION_MODE:
-		return adev->gmc.num_mem_partitions != 8 && (num_xcc % 4) == 0;
+		return comp_mode && (num_xcc % 4) == 0;
 	case AMDGPU_TPX_PARTITION_MODE:
-		return (adev->gmc.num_mem_partitions == 1 ||
-			adev->gmc.num_mem_partitions == 3) &&
-		       ((num_xcc % 3) == 0);
+		return comp_mode && ((num_xcc % 3) == 0);
 	case AMDGPU_QPX_PARTITION_MODE:
 		num_xccs_per_xcp = num_xcc / 4;
-		return (adev->gmc.num_mem_partitions == 1 ||
-			adev->gmc.num_mem_partitions == 4) &&
-		       (num_xccs_per_xcp >= 2);
+		return comp_mode && (num_xccs_per_xcp >= 2);
 	case AMDGPU_CPX_PARTITION_MODE:
-		return ((num_xcc > 1) &&
-		       (adev->gmc.num_mem_partitions == 1 || adev->gmc.num_mem_partitions == 4) &&
-		       (num_xcc % adev->gmc.num_mem_partitions) == 0);
+		return comp_mode && (num_xcc > 1);
 	default:
 		return false;
 	}
@@ -454,31 +370,16 @@ static bool __aqua_vanjaram_is_valid_mode(struct amdgpu_xcp_mgr *xcp_mgr,
 	return false;
 }
 
-static int __aqua_vanjaram_pre_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags)
+static void __aqua_vanjaram_update_available_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr)
 {
-	/* TODO:
-	 * Stop user queues and threads, and make sure GPU is empty of work.
-	 */
-
-	if (flags & AMDGPU_XCP_OPS_KFD)
-		amdgpu_amdkfd_device_fini_sw(xcp_mgr->adev);
-
-	return 0;
-}
+	int mode;
 
-static int __aqua_vanjaram_post_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags)
-{
-	int ret = 0;
+	xcp_mgr->avail_xcp_modes = 0;
 
-	if (flags & AMDGPU_XCP_OPS_KFD) {
-		amdgpu_amdkfd_device_probe(xcp_mgr->adev);
-		amdgpu_amdkfd_device_init(xcp_mgr->adev);
-		/* If KFD init failed, return failure */
-		if (!xcp_mgr->adev->kfd.init_complete)
-			ret = -EIO;
+	for_each_inst(mode, xcp_mgr->supp_xcp_modes) {
+		if (__aqua_vanjaram_is_valid_mode(xcp_mgr, mode))
+			xcp_mgr->avail_xcp_modes |= BIT(mode);
 	}
-
-	return ret;
 }
 
 static int aqua_vanjaram_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
@@ -493,6 +394,12 @@ static int aqua_vanjaram_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
 
 	if (mode == AMDGPU_AUTO_COMPUTE_PARTITION_MODE) {
 		mode = __aqua_vanjaram_get_auto_mode(xcp_mgr);
+		if (mode == AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE) {
+			dev_err(adev->dev,
+				"Invalid config, no compatible compute partition mode found, available memory partitions: %d",
+				adev->gmc.num_mem_partitions);
+			return -EINVAL;
+		}
 	} else if (!__aqua_vanjaram_is_valid_mode(xcp_mgr, mode)) {
 		dev_err(adev->dev,
 			"Invalid compute partition mode requested, requested: %s, available memory partitions: %d",
@@ -500,7 +407,8 @@ static int aqua_vanjaram_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
 		return -EINVAL;
 	}
 
-	if (adev->kfd.init_complete)
+	if (adev->kfd.init_complete && !amdgpu_in_reset(adev) &&
+		!adev->in_suspend)
 		flags |= AMDGPU_XCP_OPS_KFD;
 
 	if (flags & AMDGPU_XCP_OPS_KFD) {
@@ -509,7 +417,7 @@ static int aqua_vanjaram_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
 			goto out;
 	}
 
-	ret = __aqua_vanjaram_pre_partition_switch(xcp_mgr, flags);
+	ret = amdgpu_xcp_pre_partition_switch(xcp_mgr, flags);
 	if (ret)
 		goto unlock;
 
@@ -522,7 +430,9 @@ static int aqua_vanjaram_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
 	*num_xcps = num_xcc / num_xcc_per_xcp;
 	amdgpu_xcp_init(xcp_mgr, *num_xcps, mode);
 
-	ret = __aqua_vanjaram_post_partition_switch(xcp_mgr, flags);
+	ret = amdgpu_xcp_post_partition_switch(xcp_mgr, flags);
+	if (!ret)
+		__aqua_vanjaram_update_available_partition_mode(xcp_mgr);
 unlock:
 	if (flags & AMDGPU_XCP_OPS_KFD)
 		amdgpu_amdkfd_unlock_kfd(adev);
@@ -601,20 +511,23 @@ struct amdgpu_xcp_mgr_funcs aqua_vanjaram_xcp_funcs = {
 	.switch_partition_mode = &aqua_vanjaram_switch_partition_mode,
 	.query_partition_mode = &aqua_vanjaram_query_partition_mode,
 	.get_ip_details = &aqua_vanjaram_get_xcp_ip_details,
+	.get_xcp_res_info = &aqua_vanjaram_get_xcp_res_info,
 	.get_xcp_mem_id = &aqua_vanjaram_get_xcp_mem_id,
-	.select_scheds = &aqua_vanjaram_select_scheds,
-	.update_partition_sched_list = &aqua_vanjaram_update_partition_sched_list
 };
 
 static int aqua_vanjaram_xcp_mgr_init(struct amdgpu_device *adev)
 {
 	int ret;
 
+	if (amdgpu_sriov_vf(adev))
+		aqua_vanjaram_xcp_funcs.switch_partition_mode = NULL;
+
 	ret = amdgpu_xcp_mgr_init(adev, AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE, 1,
 				  &aqua_vanjaram_xcp_funcs);
 	if (ret)
 		return ret;
 
+	amdgpu_xcp_update_supported_modes(adev->xcp_mgr);
 	/* TODO: Default memory node affinity init */
 
 	return ret;
@@ -622,7 +535,7 @@ static int aqua_vanjaram_xcp_mgr_init(struct amdgpu_device *adev)
 
 int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev)
 {
-	u32 mask, inst_mask = adev->sdma.sdma_mask;
+	u32 mask, avail_inst, inst_mask = adev->sdma.sdma_mask;
 	int ret, i;
 
 	/* generally 1 AID supports 4 instances */
@@ -634,7 +547,9 @@ int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev)
 
 	for (mask = (1 << adev->sdma.num_inst_per_aid) - 1; inst_mask;
 	     inst_mask >>= adev->sdma.num_inst_per_aid, ++i) {
-		if ((inst_mask & mask) == mask)
+		avail_inst = inst_mask & mask;
+		if (avail_inst == mask || avail_inst == 0x3 ||
+		    avail_inst == 0xc)
 			adev->aid_mask |= (1 << i);
 	}
 
@@ -652,7 +567,420 @@ int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev)
 	if (ret)
 		return ret;
 
-	aqua_vanjaram_ip_map_init(adev);
+	amdgpu_ip_map_init(adev);
 
 	return 0;
 }
+
+static void aqua_read_smn(struct amdgpu_device *adev,
+			  struct amdgpu_smn_reg_data *regdata,
+			  uint64_t smn_addr)
+{
+	regdata->addr = smn_addr;
+	regdata->value = RREG32_PCIE(smn_addr);
+}
+
+struct aqua_reg_list {
+	uint64_t start_addr;
+	uint32_t num_regs;
+	uint32_t incrx;
+};
+
+#define DW_ADDR_INCR	4
+
+static void aqua_read_smn_ext(struct amdgpu_device *adev,
+			      struct amdgpu_smn_reg_data *regdata,
+			      uint64_t smn_addr, int i)
+{
+	regdata->addr =
+		smn_addr + adev->asic_funcs->encode_ext_smn_addressing(i);
+	regdata->value = RREG32_PCIE_EXT(regdata->addr);
+}
+
+#define smnreg_0x1A340218	0x1A340218
+#define smnreg_0x1A3402E4	0x1A3402E4
+#define smnreg_0x1A340294	0x1A340294
+#define smreg_0x1A380088	0x1A380088
+
+#define NUM_PCIE_SMN_REGS	14
+
+static struct aqua_reg_list pcie_reg_addrs[] = {
+	{ smnreg_0x1A340218, 1, 0 },
+	{ smnreg_0x1A3402E4, 1, 0 },
+	{ smnreg_0x1A340294, 6, DW_ADDR_INCR },
+	{ smreg_0x1A380088, 6, DW_ADDR_INCR },
+};
+
+static ssize_t aqua_vanjaram_read_pcie_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_pcie_v1_0 *pcie_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_pcie_v1_0 *pcie_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	struct pci_dev *us_pdev, *ds_pdev;
+	int aer_cap, r, n;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	pcie_reg_state = (struct amdgpu_reg_state_pcie_v1_0 *)buf;
+
+	szbuf = sizeof(*pcie_reg_state) +
+		amdgpu_reginst_size(1, sizeof(*pcie_regs), NUM_PCIE_SMN_REGS);
+	/* Only one instance of pcie regs */
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	pcie_regs = (struct amdgpu_regs_pcie_v1_0 *)((uint8_t *)buf +
+						     sizeof(*pcie_reg_state));
+	pcie_regs->inst_header.instance = 0;
+	pcie_regs->inst_header.state = AMDGPU_INST_S_OK;
+	pcie_regs->inst_header.num_smn_regs = NUM_PCIE_SMN_REGS;
+
+	reg_data = pcie_regs->smn_reg_values;
+
+	for (r = 0; r < ARRAY_SIZE(pcie_reg_addrs); r++) {
+		start_addr = pcie_reg_addrs[r].start_addr;
+		incrx = pcie_reg_addrs[r].incrx;
+		num_regs = pcie_reg_addrs[r].num_regs;
+		for (n = 0; n < num_regs; n++) {
+			aqua_read_smn(adev, reg_data, start_addr + n * incrx);
+			++reg_data;
+		}
+	}
+
+	ds_pdev = pci_upstream_bridge(adev->pdev);
+	us_pdev = pci_upstream_bridge(ds_pdev);
+
+	pcie_capability_read_word(us_pdev, PCI_EXP_DEVSTA,
+				  &pcie_regs->device_status);
+	pcie_capability_read_word(us_pdev, PCI_EXP_LNKSTA,
+				  &pcie_regs->link_status);
+
+	aer_cap = pci_find_ext_capability(us_pdev, PCI_EXT_CAP_ID_ERR);
+	if (aer_cap) {
+		pci_read_config_dword(us_pdev, aer_cap + PCI_ERR_COR_STATUS,
+				      &pcie_regs->pcie_corr_err_status);
+		pci_read_config_dword(us_pdev, aer_cap + PCI_ERR_UNCOR_STATUS,
+				      &pcie_regs->pcie_uncorr_err_status);
+	}
+
+	pci_read_config_dword(us_pdev, PCI_PRIMARY_BUS,
+			      &pcie_regs->sub_bus_number_latency);
+
+	pcie_reg_state->common_header.structure_size = szbuf;
+	pcie_reg_state->common_header.format_revision = 1;
+	pcie_reg_state->common_header.content_revision = 0;
+	pcie_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_PCIE;
+	pcie_reg_state->common_header.num_instances = 1;
+
+	return pcie_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x11A00050	0x11A00050
+#define smnreg_0x11A00180	0x11A00180
+#define smnreg_0x11A00070	0x11A00070
+#define smnreg_0x11A00200	0x11A00200
+#define smnreg_0x11A0020C	0x11A0020C
+#define smnreg_0x11A00210	0x11A00210
+#define smnreg_0x11A00108	0x11A00108
+
+#define XGMI_LINK_REG(smnreg, l) ((smnreg) | (l << 20))
+
+#define NUM_XGMI_SMN_REGS 25
+
+static struct aqua_reg_list xgmi_reg_addrs[] = {
+	{ smnreg_0x11A00050, 1, 0 },
+	{ smnreg_0x11A00180, 16, DW_ADDR_INCR },
+	{ smnreg_0x11A00070, 4, DW_ADDR_INCR },
+	{ smnreg_0x11A00200, 1, 0 },
+	{ smnreg_0x11A0020C, 1, 0 },
+	{ smnreg_0x11A00210, 1, 0 },
+	{ smnreg_0x11A00108, 1, 0 },
+};
+
+static ssize_t aqua_vanjaram_read_xgmi_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_xgmi_v1_0 *xgmi_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_xgmi_v1_0 *xgmi_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_xgmi_instances = 8;
+	int inst = 0, i, j, r, n;
+	const int xgmi_inst = 2;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	xgmi_reg_state = (struct amdgpu_reg_state_xgmi_v1_0 *)buf;
+
+	szbuf = sizeof(*xgmi_reg_state) +
+		amdgpu_reginst_size(max_xgmi_instances, sizeof(*xgmi_regs),
+				    NUM_XGMI_SMN_REGS);
+	/* Only one instance of pcie regs */
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &xgmi_reg_state->xgmi_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		for (j = 0; j < xgmi_inst; ++j) {
+			xgmi_regs = (struct amdgpu_regs_xgmi_v1_0 *)p;
+			xgmi_regs->inst_header.instance = inst++;
+
+			xgmi_regs->inst_header.state = AMDGPU_INST_S_OK;
+			xgmi_regs->inst_header.num_smn_regs = NUM_XGMI_SMN_REGS;
+
+			reg_data = xgmi_regs->smn_reg_values;
+
+			for (r = 0; r < ARRAY_SIZE(xgmi_reg_addrs); r++) {
+				start_addr = xgmi_reg_addrs[r].start_addr;
+				incrx = xgmi_reg_addrs[r].incrx;
+				num_regs = xgmi_reg_addrs[r].num_regs;
+
+				for (n = 0; n < num_regs; n++) {
+					aqua_read_smn_ext(
+						adev, reg_data,
+						XGMI_LINK_REG(start_addr, j) +
+							n * incrx,
+						i);
+					++reg_data;
+				}
+			}
+			p = reg_data;
+		}
+	}
+
+	xgmi_reg_state->common_header.structure_size = szbuf;
+	xgmi_reg_state->common_header.format_revision = 1;
+	xgmi_reg_state->common_header.content_revision = 0;
+	xgmi_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_XGMI;
+	xgmi_reg_state->common_header.num_instances = max_xgmi_instances;
+
+	return xgmi_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x11C00070	0x11C00070
+#define smnreg_0x11C00210	0x11C00210
+
+static struct aqua_reg_list wafl_reg_addrs[] = {
+	{ smnreg_0x11C00070, 4, DW_ADDR_INCR },
+	{ smnreg_0x11C00210, 1, 0 },
+};
+
+#define WAFL_LINK_REG(smnreg, l) ((smnreg) | (l << 20))
+
+#define NUM_WAFL_SMN_REGS 5
+
+static ssize_t aqua_vanjaram_read_wafl_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_wafl_v1_0 *wafl_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_wafl_v1_0 *wafl_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_wafl_instances = 8;
+	int inst = 0, i, j, r, n;
+	const int wafl_inst = 2;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	wafl_reg_state = (struct amdgpu_reg_state_wafl_v1_0 *)buf;
+
+	szbuf = sizeof(*wafl_reg_state) +
+		amdgpu_reginst_size(max_wafl_instances, sizeof(*wafl_regs),
+				    NUM_WAFL_SMN_REGS);
+
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &wafl_reg_state->wafl_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		for (j = 0; j < wafl_inst; ++j) {
+			wafl_regs = (struct amdgpu_regs_wafl_v1_0 *)p;
+			wafl_regs->inst_header.instance = inst++;
+
+			wafl_regs->inst_header.state = AMDGPU_INST_S_OK;
+			wafl_regs->inst_header.num_smn_regs = NUM_WAFL_SMN_REGS;
+
+			reg_data = wafl_regs->smn_reg_values;
+
+			for (r = 0; r < ARRAY_SIZE(wafl_reg_addrs); r++) {
+				start_addr = wafl_reg_addrs[r].start_addr;
+				incrx = wafl_reg_addrs[r].incrx;
+				num_regs = wafl_reg_addrs[r].num_regs;
+				for (n = 0; n < num_regs; n++) {
+					aqua_read_smn_ext(
+						adev, reg_data,
+						WAFL_LINK_REG(start_addr, j) +
+							n * incrx,
+						i);
+					++reg_data;
+				}
+			}
+			p = reg_data;
+		}
+	}
+
+	wafl_reg_state->common_header.structure_size = szbuf;
+	wafl_reg_state->common_header.format_revision = 1;
+	wafl_reg_state->common_header.content_revision = 0;
+	wafl_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_WAFL;
+	wafl_reg_state->common_header.num_instances = max_wafl_instances;
+
+	return wafl_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x1B311060 0x1B311060
+#define smnreg_0x1B411060 0x1B411060
+#define smnreg_0x1B511060 0x1B511060
+#define smnreg_0x1B611060 0x1B611060
+
+#define smnreg_0x1C307120 0x1C307120
+#define smnreg_0x1C317120 0x1C317120
+
+#define smnreg_0x1C320830 0x1C320830
+#define smnreg_0x1C380830 0x1C380830
+#define smnreg_0x1C3D0830 0x1C3D0830
+#define smnreg_0x1C420830 0x1C420830
+
+#define smnreg_0x1C320100 0x1C320100
+#define smnreg_0x1C380100 0x1C380100
+#define smnreg_0x1C3D0100 0x1C3D0100
+#define smnreg_0x1C420100 0x1C420100
+
+#define smnreg_0x1B310500 0x1B310500
+#define smnreg_0x1C300400 0x1C300400
+
+#define USR_CAKE_INCR 0x11000
+#define USR_LINK_INCR 0x100000
+#define USR_CP_INCR 0x10000
+
+#define NUM_USR_SMN_REGS	20
+
+struct aqua_reg_list usr_reg_addrs[] = {
+	{ smnreg_0x1B311060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B411060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B511060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B611060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1C307120, 2, DW_ADDR_INCR },
+	{ smnreg_0x1C317120, 2, DW_ADDR_INCR },
+};
+
+#define NUM_USR1_SMN_REGS	46
+struct aqua_reg_list usr1_reg_addrs[] = {
+	{ smnreg_0x1C320830, 6, USR_CAKE_INCR },
+	{ smnreg_0x1C380830, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C3D0830, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C420830, 4, USR_CAKE_INCR },
+	{ smnreg_0x1C320100, 6, USR_CAKE_INCR },
+	{ smnreg_0x1C380100, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C3D0100, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C420100, 4, USR_CAKE_INCR },
+	{ smnreg_0x1B310500, 4, USR_LINK_INCR },
+	{ smnreg_0x1C300400, 2, USR_CP_INCR },
+};
+
+static ssize_t aqua_vanjaram_read_usr_state(struct amdgpu_device *adev,
+					    void *buf, size_t max_size,
+					    int reg_state)
+{
+	uint32_t start_addr, incrx, num_regs, szbuf, num_smn;
+	struct amdgpu_reg_state_usr_v1_0 *usr_reg_state;
+	struct amdgpu_regs_usr_v1_0 *usr_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_usr_instances = 4;
+	struct aqua_reg_list *reg_addrs;
+	int inst = 0, i, n, r, arr_size;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	switch (reg_state) {
+	case AMDGPU_REG_STATE_TYPE_USR:
+		arr_size = ARRAY_SIZE(usr_reg_addrs);
+		reg_addrs = usr_reg_addrs;
+		num_smn = NUM_USR_SMN_REGS;
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR_1:
+		arr_size = ARRAY_SIZE(usr1_reg_addrs);
+		reg_addrs = usr1_reg_addrs;
+		num_smn = NUM_USR1_SMN_REGS;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	usr_reg_state = (struct amdgpu_reg_state_usr_v1_0 *)buf;
+
+	szbuf = sizeof(*usr_reg_state) + amdgpu_reginst_size(max_usr_instances,
+							     sizeof(*usr_regs),
+							     num_smn);
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &usr_reg_state->usr_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		usr_regs = (struct amdgpu_regs_usr_v1_0 *)p;
+		usr_regs->inst_header.instance = inst++;
+		usr_regs->inst_header.state = AMDGPU_INST_S_OK;
+		usr_regs->inst_header.num_smn_regs = num_smn;
+		reg_data = usr_regs->smn_reg_values;
+
+		for (r = 0; r < arr_size; r++) {
+			start_addr = reg_addrs[r].start_addr;
+			incrx = reg_addrs[r].incrx;
+			num_regs = reg_addrs[r].num_regs;
+			for (n = 0; n < num_regs; n++) {
+				aqua_read_smn_ext(adev, reg_data,
+						  start_addr + n * incrx, i);
+				reg_data++;
+			}
+		}
+		p = reg_data;
+	}
+
+	usr_reg_state->common_header.structure_size = szbuf;
+	usr_reg_state->common_header.format_revision = 1;
+	usr_reg_state->common_header.content_revision = 0;
+	usr_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_USR;
+	usr_reg_state->common_header.num_instances = max_usr_instances;
+
+	return usr_reg_state->common_header.structure_size;
+}
+
+ssize_t aqua_vanjaram_get_reg_state(struct amdgpu_device *adev,
+				    enum amdgpu_reg_state reg_state, void *buf,
+				    size_t max_size)
+{
+	ssize_t size;
+
+	switch (reg_state) {
+	case AMDGPU_REG_STATE_TYPE_PCIE:
+		size = aqua_vanjaram_read_pcie_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_XGMI:
+		size = aqua_vanjaram_read_xgmi_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_WAFL:
+		size = aqua_vanjaram_read_wafl_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR:
+		size = aqua_vanjaram_read_usr_state(adev, buf, max_size,
+						    AMDGPU_REG_STATE_TYPE_USR);
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR_1:
+		size = aqua_vanjaram_read_usr_state(
+			adev, buf, max_size, AMDGPU_REG_STATE_TYPE_USR_1);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return size;
+}