drm/amdgpu: merge amdgpu_family.h into amd_shared.h (v2)
[cascardo/linux.git] / drivers / gpu / drm / amd / amdgpu / gfx_v8_0.c
index 1c7c992..17df5c2 100644 (file)
@@ -587,6 +587,7 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
        int err;
        struct amdgpu_firmware_info *info = NULL;
        const struct common_firmware_header *header = NULL;
+       const struct gfx_firmware_header_v1_0 *cp_hdr;
 
        DRM_DEBUG("\n");
 
@@ -611,6 +612,9 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
        err = amdgpu_ucode_validate(adev->gfx.pfp_fw);
        if (err)
                goto out;
+       cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
+       adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
+       adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
        snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
        err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
@@ -619,6 +623,9 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
        err = amdgpu_ucode_validate(adev->gfx.me_fw);
        if (err)
                goto out;
+       cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
+       adev->gfx.me_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
+       adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
        snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
        err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
@@ -627,12 +634,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
        err = amdgpu_ucode_validate(adev->gfx.ce_fw);
        if (err)
                goto out;
+       cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
+       adev->gfx.ce_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
+       adev->gfx.ce_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
        snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
        err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev);
        if (err)
                goto out;
        err = amdgpu_ucode_validate(adev->gfx.rlc_fw);
+       cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.rlc_fw->data;
+       adev->gfx.rlc_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
+       adev->gfx.rlc_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
        snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
        err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
@@ -641,6 +654,9 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
        err = amdgpu_ucode_validate(adev->gfx.mec_fw);
        if (err)
                goto out;
+       cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
+       adev->gfx.mec_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
+       adev->gfx.mec_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
        snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
        err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
@@ -648,6 +664,12 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
                err = amdgpu_ucode_validate(adev->gfx.mec2_fw);
                if (err)
                        goto out;
+               cp_hdr = (const struct gfx_firmware_header_v1_0 *)
+                                               adev->gfx.mec2_fw->data;
+               adev->gfx.mec2_fw_version = le32_to_cpu(
+                                               cp_hdr->header.ucode_version);
+               adev->gfx.mec2_feature_version = le32_to_cpu(
+                                               cp_hdr->ucode_feature_version);
        } else {
                err = 0;
                adev->gfx.mec2_fw = NULL;
@@ -1983,6 +2005,7 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev)
                adev->gfx.config.max_shader_engines = 1;
                adev->gfx.config.max_tile_pipes = 2;
                adev->gfx.config.max_sh_per_se = 1;
+               adev->gfx.config.max_backends_per_se = 2;
 
                switch (adev->pdev->revision) {
                case 0xc4:
@@ -1991,7 +2014,6 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev)
                case 0xcc:
                        /* B10 */
                        adev->gfx.config.max_cu_per_sh = 8;
-                       adev->gfx.config.max_backends_per_se = 2;
                        break;
                case 0xc5:
                case 0x81:
@@ -2000,14 +2022,12 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev)
                case 0xcd:
                        /* B8 */
                        adev->gfx.config.max_cu_per_sh = 6;
-                       adev->gfx.config.max_backends_per_se = 2;
                        break;
                case 0xc6:
                case 0xca:
                case 0xce:
                        /* B6 */
                        adev->gfx.config.max_cu_per_sh = 6;
-                       adev->gfx.config.max_backends_per_se = 2;
                        break;
                case 0xc7:
                case 0x87:
@@ -2015,7 +2035,6 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev)
                default:
                        /* B4 */
                        adev->gfx.config.max_cu_per_sh = 4;
-                       adev->gfx.config.max_backends_per_se = 1;
                        break;
                }
 
@@ -2059,7 +2078,7 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev)
 
        adev->gfx.config.num_tile_pipes = adev->gfx.config.max_tile_pipes;
        adev->gfx.config.mem_max_burst_length_bytes = 256;
-       if (adev->flags & AMDGPU_IS_APU) {
+       if (adev->flags & AMD_IS_APU) {
                /* Get memory bank mapping mode. */
                tmp = RREG32(mmMC_FUS_DRAM0_BANK_ADDR_MAPPING);
                dimm00_addr_map = REG_GET_FIELD(tmp, MC_FUS_DRAM0_BANK_ADDR_MAPPING, DIMM0ADDRMAP);
@@ -2275,7 +2294,6 @@ static int gfx_v8_0_rlc_load_microcode(struct amdgpu_device *adev)
 
        hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
        amdgpu_ucode_print_rlc_hdr(&hdr->header);
-       adev->gfx.rlc_fw_version = le32_to_cpu(hdr->header.ucode_version);
 
        fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
                           le32_to_cpu(hdr->header.ucode_array_offset_bytes));
@@ -2361,12 +2379,6 @@ static int gfx_v8_0_cp_gfx_load_microcode(struct amdgpu_device *adev)
        amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header);
        amdgpu_ucode_print_gfx_hdr(&ce_hdr->header);
        amdgpu_ucode_print_gfx_hdr(&me_hdr->header);
-       adev->gfx.pfp_fw_version = le32_to_cpu(pfp_hdr->header.ucode_version);
-       adev->gfx.ce_fw_version = le32_to_cpu(ce_hdr->header.ucode_version);
-       adev->gfx.me_fw_version = le32_to_cpu(me_hdr->header.ucode_version);
-       adev->gfx.me_feature_version = le32_to_cpu(me_hdr->ucode_feature_version);
-       adev->gfx.ce_feature_version = le32_to_cpu(ce_hdr->ucode_feature_version);
-       adev->gfx.pfp_feature_version = le32_to_cpu(pfp_hdr->ucode_feature_version);
 
        gfx_v8_0_cp_gfx_enable(adev, false);
 
@@ -2622,7 +2634,6 @@ static int gfx_v8_0_cp_compute_load_microcode(struct amdgpu_device *adev)
 
        mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
        amdgpu_ucode_print_gfx_hdr(&mec_hdr->header);
-       adev->gfx.mec_fw_version = le32_to_cpu(mec_hdr->header.ucode_version);
 
        fw_data = (const __le32 *)
                (adev->gfx.mec_fw->data +
@@ -2641,7 +2652,6 @@ static int gfx_v8_0_cp_compute_load_microcode(struct amdgpu_device *adev)
 
                mec2_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data;
                amdgpu_ucode_print_gfx_hdr(&mec2_hdr->header);
-               adev->gfx.mec2_fw_version = le32_to_cpu(mec2_hdr->header.ucode_version);
 
                fw_data = (const __le32 *)
                        (adev->gfx.mec2_fw->data +
@@ -3125,7 +3135,7 @@ static int gfx_v8_0_cp_compute_resume(struct amdgpu_device *adev)
                                WREG32(mmCP_MEC_DOORBELL_RANGE_LOWER,
                                       AMDGPU_DOORBELL_KIQ << 2);
                                WREG32(mmCP_MEC_DOORBELL_RANGE_UPPER,
-                                               0x7FFFF << 2);
+                                      AMDGPU_DOORBELL_MEC_RING7 << 2);
                        }
                        tmp = RREG32(mmCP_HQD_PQ_DOORBELL_CONTROL);
                        tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
@@ -3753,7 +3763,7 @@ static void gfx_v8_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
        amdgpu_ring_write(ring, 0x20); /* poll interval */
 }
 
-static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring,
+static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
                                  struct amdgpu_ib *ib)
 {
        bool need_ctx_switch = ring->current_ctx != ib->ctx;
@@ -3761,15 +3771,10 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring,
        u32 next_rptr = ring->wptr + 5;
 
        /* drop the CE preamble IB for the same context */
-       if ((ring->type == AMDGPU_RING_TYPE_GFX) &&
-           (ib->flags & AMDGPU_IB_FLAG_PREAMBLE) &&
-           !need_ctx_switch)
+       if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch)
                return;
 
-       if (ring->type == AMDGPU_RING_TYPE_COMPUTE)
-               control |= INDIRECT_BUFFER_VALID;
-
-       if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX)
+       if (need_ctx_switch)
                next_rptr += 2;
 
        next_rptr += 4;
@@ -3780,7 +3785,7 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring,
        amdgpu_ring_write(ring, next_rptr);
 
        /* insert SWITCH_BUFFER packet before first IB in the ring frame */
-       if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) {
+       if (need_ctx_switch) {
                amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
                amdgpu_ring_write(ring, 0);
        }
@@ -3803,6 +3808,36 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring,
        amdgpu_ring_write(ring, control);
 }
 
+static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
+                                 struct amdgpu_ib *ib)
+{
+       u32 header, control = 0;
+       u32 next_rptr = ring->wptr + 5;
+
+       control |= INDIRECT_BUFFER_VALID;
+
+       next_rptr += 4;
+       amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+       amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
+       amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
+       amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff);
+       amdgpu_ring_write(ring, next_rptr);
+
+       header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
+
+       control |= ib->length_dw |
+                          (ib->vm ? (ib->vm->ids[ring->idx].id << 24) : 0);
+
+       amdgpu_ring_write(ring, header);
+       amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+                                         (2 << 0) |
+#endif
+                                         (ib->gpu_addr & 0xFFFFFFFC));
+       amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF);
+       amdgpu_ring_write(ring, control);
+}
+
 static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
                                         u64 seq, unsigned flags)
 {
@@ -4224,7 +4259,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
        .get_wptr = gfx_v8_0_ring_get_wptr_gfx,
        .set_wptr = gfx_v8_0_ring_set_wptr_gfx,
        .parse_cs = NULL,
-       .emit_ib = gfx_v8_0_ring_emit_ib,
+       .emit_ib = gfx_v8_0_ring_emit_ib_gfx,
        .emit_fence = gfx_v8_0_ring_emit_fence_gfx,
        .emit_semaphore = gfx_v8_0_ring_emit_semaphore,
        .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush,
@@ -4240,7 +4275,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
        .get_wptr = gfx_v8_0_ring_get_wptr_compute,
        .set_wptr = gfx_v8_0_ring_set_wptr_compute,
        .parse_cs = NULL,
-       .emit_ib = gfx_v8_0_ring_emit_ib,
+       .emit_ib = gfx_v8_0_ring_emit_ib_compute,
        .emit_fence = gfx_v8_0_ring_emit_fence_compute,
        .emit_semaphore = gfx_v8_0_ring_emit_semaphore,
        .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush,