drm/amdgpu: implement initialization part on VCN2.0 for SRIOV

something need to do for VCN2.0 enablement on SRIOV:
1)use one dec ring and one enc ring
2)allocate MM table for MMSCH usage
3)implement SRIOV version vcn_start which orgnize vcn programing
with patcket format and implement start mmsch for to run those
packet
4)doorbell is changed for SRIOV
Singed-off-by: darlington Opara <darlington.opara@amd.com>
Signed-off-by: Jinage Zhao <jiange.zhao@amd.com>
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c

@@ -29,6 +29,7 @@
 #include "soc15d.h"
 #include "amdgpu_pm.h"
 #include "amdgpu_psp.h"
+#include "mmsch_v2_0.h"
 
 #include "vcn/vcn_2_0_0_offset.h"
 #include "vcn/vcn_2_0_0_sh_mask.h"
@@ -54,7 +55,7 @@ static int vcn_v2_0_set_powergating_state(void *handle,
 				enum amd_powergating_state state);
 static int vcn_v2_0_pause_dpg_mode(struct amdgpu_device *adev,
 				int inst_idx, struct dpg_pause_state *new_state);
-
+static int vcn_v2_0_start_sriov(struct amdgpu_device *adev);
 /**
  * vcn_v2_0_early_init - set function pointers
  *
@@ -67,7 +68,10 @@ static int vcn_v2_0_early_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	adev->vcn.num_vcn_inst = 1;
-	adev->vcn.num_enc_rings = 2;
+	if (amdgpu_sriov_vf(adev))
+		adev->vcn.num_enc_rings = 1;
+	else
+		adev->vcn.num_enc_rings = 2;
 
 	vcn_v2_0_set_dec_ring_funcs(adev);
 	vcn_v2_0_set_enc_ring_funcs(adev);
@@ -154,7 +158,10 @@ static int vcn_v2_0_sw_init(void *handle)
 	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
 		ring = &adev->vcn.inst->ring_enc[i];
 		ring->use_doorbell = true;
-		ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + i;
+		if (!amdgpu_sriov_vf(adev))
+			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + i;
+		else
+			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1 + i;
 		sprintf(ring->name, "vcn_enc%d", i);
 		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0);
 		if (r)
@@ -163,6 +170,10 @@ static int vcn_v2_0_sw_init(void *handle)
 
 	adev->vcn.pause_dpg_mode = vcn_v2_0_pause_dpg_mode;
 
+	r = amdgpu_virt_alloc_mm_table(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -178,6 +189,8 @@ static int vcn_v2_0_sw_fini(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	amdgpu_virt_free_mm_table(adev);
+
 	r = amdgpu_vcn_suspend(adev);
 	if (r)
 		return r;
@@ -203,6 +216,9 @@ static int vcn_v2_0_hw_init(void *handle)
 	adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
 					     ring->doorbell_index, 0);
 
+	if (amdgpu_sriov_vf(adev))
+		vcn_v2_0_start_sriov(adev);
+
 	r = amdgpu_ring_test_helper(ring);
 	if (r)
 		goto done;
@@ -1680,6 +1696,215 @@ static int vcn_v2_0_set_powergating_state(void *handle,
 	return ret;
 }
 
+static int vcn_v2_0_start_mmsch(struct amdgpu_device *adev,
+				struct amdgpu_mm_table *table)
+{
+	uint32_t data = 0, loop;
+	uint64_t addr = table->gpu_addr;
+	struct mmsch_v2_0_init_header *header;
+	uint32_t size;
+	int i;
+
+	header = (struct mmsch_v2_0_init_header *)table->cpu_addr;
+	size = header->header_size + header->vcn_table_size;
+
+	/* 1, write to vce_mmsch_vf_ctx_addr_lo/hi register with GPU mc addr
+	 * of memory descriptor location
+	 */
+	WREG32_SOC15(UVD, 0, mmMMSCH_VF_CTX_ADDR_LO, lower_32_bits(addr));
+	WREG32_SOC15(UVD, 0, mmMMSCH_VF_CTX_ADDR_HI, upper_32_bits(addr));
+
+	/* 2, update vmid of descriptor */
+	data = RREG32_SOC15(UVD, 0, mmMMSCH_VF_VMID);
+	data &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK;
+	/* use domain0 for MM scheduler */
+	data |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT);
+	WREG32_SOC15(UVD, 0, mmMMSCH_VF_VMID, data);
+
+	/* 3, notify mmsch about the size of this descriptor */
+	WREG32_SOC15(UVD, 0, mmMMSCH_VF_CTX_SIZE, size);
+
+	/* 4, set resp to zero */
+	WREG32_SOC15(UVD, 0, mmMMSCH_VF_MAILBOX_RESP, 0);
+
+	adev->vcn.inst->ring_dec.wptr = 0;
+	adev->vcn.inst->ring_dec.wptr_old = 0;
+	vcn_v2_0_dec_ring_set_wptr(&adev->vcn.inst->ring_dec);
+
+	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+		adev->vcn.inst->ring_enc[i].wptr = 0;
+		adev->vcn.inst->ring_enc[i].wptr_old = 0;
+		vcn_v2_0_enc_ring_set_wptr(&adev->vcn.inst->ring_enc[i]);
+	}
+
+	/* 5, kick off the initialization and wait until
+	 * VCE_MMSCH_VF_MAILBOX_RESP becomes non-zero
+	 */
+	WREG32_SOC15(UVD, 0, mmMMSCH_VF_MAILBOX_HOST, 0x10000001);
+
+	data = RREG32_SOC15(UVD, 0, mmMMSCH_VF_MAILBOX_RESP);
+	loop = 1000;
+	while ((data & 0x10000002) != 0x10000002) {
+		udelay(10);
+		data = RREG32_SOC15(UVD, 0, mmMMSCH_VF_MAILBOX_RESP);
+		loop--;
+		if (!loop)
+			break;
+	}
+
+	if (!loop) {
+		DRM_ERROR("failed to init MMSCH, " \
+			"mmMMSCH_VF_MAILBOX_RESP = 0x%08x\n", data);
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+static int vcn_v2_0_start_sriov(struct amdgpu_device *adev)
+{
+	int r;
+	uint32_t tmp;
+	struct amdgpu_ring *ring;
+	uint32_t offset, size;
+	uint32_t table_size = 0;
+	struct mmsch_v2_0_cmd_direct_write direct_wt = { {0} };
+	struct mmsch_v2_0_cmd_direct_read_modify_write direct_rd_mod_wt = { {0} };
+	struct mmsch_v2_0_cmd_direct_polling direct_poll = { {0} };
+	struct mmsch_v2_0_cmd_end end = { {0} };
+	struct mmsch_v2_0_init_header *header;
+	uint32_t *init_table = adev->virt.mm_table.cpu_addr;
+	uint8_t i = 0;
+
+	header = (struct mmsch_v2_0_init_header *)init_table;
+	direct_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_WRITE;
+	direct_rd_mod_wt.cmd_header.command_type =
+		MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE;
+	direct_poll.cmd_header.command_type =
+		MMSCH_COMMAND__DIRECT_REG_POLLING;
+	end.cmd_header.command_type = MMSCH_COMMAND__END;
+
+	if (header->vcn_table_offset == 0 && header->vcn_table_size == 0) {
+		header->version = MMSCH_VERSION;
+		header->header_size = sizeof(struct mmsch_v2_0_init_header) >> 2;
+
+		header->vcn_table_offset = header->header_size;
+
+		init_table += header->vcn_table_offset;
+
+		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+
+		MMSCH_V2_0_INSERT_DIRECT_RD_MOD_WT(
+			SOC15_REG_OFFSET(UVD, i, mmUVD_STATUS),
+			0xFFFFFFFF, 0x00000004);
+
+		/* mc resume*/
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+			tmp = AMDGPU_UCODE_ID_VCN;
+			MMSCH_V2_0_INSERT_DIRECT_WT(
+				SOC15_REG_OFFSET(UVD, i,
+					mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				adev->firmware.ucode[tmp].tmr_mc_addr_lo);
+			MMSCH_V2_0_INSERT_DIRECT_WT(
+				SOC15_REG_OFFSET(UVD, i,
+					mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				adev->firmware.ucode[tmp].tmr_mc_addr_hi);
+			offset = 0;
+		} else {
+			MMSCH_V2_0_INSERT_DIRECT_WT(
+				SOC15_REG_OFFSET(UVD, i,
+					mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				lower_32_bits(adev->vcn.inst->gpu_addr));
+			MMSCH_V2_0_INSERT_DIRECT_WT(
+				SOC15_REG_OFFSET(UVD, i,
+					mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				upper_32_bits(adev->vcn.inst->gpu_addr));
+			offset = size;
+		}
+
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET0),
+			0);
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE0),
+			size);
+
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i,
+				mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst->gpu_addr + offset));
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i,
+				mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst->gpu_addr + offset));
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET1),
+			0);
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE1),
+			AMDGPU_VCN_STACK_SIZE);
+
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i,
+				mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst->gpu_addr + offset +
+				AMDGPU_VCN_STACK_SIZE));
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i,
+				mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst->gpu_addr + offset +
+				AMDGPU_VCN_STACK_SIZE));
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET2),
+			0);
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE2),
+			AMDGPU_VCN_CONTEXT_SIZE);
+
+		for (r = 0; r < adev->vcn.num_enc_rings; ++r) {
+			ring = &adev->vcn.inst->ring_enc[r];
+			ring->wptr = 0;
+			MMSCH_V2_0_INSERT_DIRECT_WT(
+				SOC15_REG_OFFSET(UVD, i, mmUVD_RB_BASE_LO),
+				lower_32_bits(ring->gpu_addr));
+			MMSCH_V2_0_INSERT_DIRECT_WT(
+				SOC15_REG_OFFSET(UVD, i, mmUVD_RB_BASE_HI),
+				upper_32_bits(ring->gpu_addr));
+			MMSCH_V2_0_INSERT_DIRECT_WT(
+				SOC15_REG_OFFSET(UVD, i, mmUVD_RB_SIZE),
+				ring->ring_size / 4);
+		}
+
+		ring = &adev->vcn.inst->ring_dec;
+		ring->wptr = 0;
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i,
+				mmUVD_LMI_RBC_RB_64BIT_BAR_LOW),
+			lower_32_bits(ring->gpu_addr));
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i,
+				mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH),
+			upper_32_bits(ring->gpu_addr));
+		/* force RBC into idle state */
+		tmp = order_base_2(ring->ring_size);
+		tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, tmp);
+		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
+		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
+		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
+		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
+		MMSCH_V2_0_INSERT_DIRECT_WT(
+			SOC15_REG_OFFSET(UVD, i, mmUVD_RBC_RB_CNTL), tmp);
+
+		/* add end packet */
+		tmp = sizeof(struct mmsch_v2_0_cmd_end);
+		memcpy((void *)init_table, &end, tmp);
+		table_size += (tmp / 4);
+		header->vcn_table_size = table_size;
+
+	}
+	return vcn_v2_0_start_mmsch(adev, &adev->virt.mm_table);
+}
+
 static const struct amd_ip_funcs vcn_v2_0_ip_funcs = {
 	.name = "vcn_v2_0",
 	.early_init = vcn_v2_0_early_init,