diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 3916 |
1 files changed, 3916 insertions, 0 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c new file mode 100644 index 000000000000..a447b70841c9 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -0,0 +1,3916 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include <linux/firmware.h> +#include "drmP.h" +#include "amdgpu.h" +#include "amdgpu_gfx.h" +#include "soc15.h" +#include "soc15d.h" + +#include "vega10/soc15ip.h" +#include "vega10/GC/gc_9_0_offset.h" +#include "vega10/GC/gc_9_0_sh_mask.h" +#include "vega10/vega10_enum.h" +#include "vega10/HDP/hdp_4_0_offset.h" + +#include "soc15_common.h" +#include "clearstate_gfx9.h" +#include "v9_structs.h" + +#define GFX9_NUM_GFX_RINGS 1 +#define GFX9_NUM_COMPUTE_RINGS 8 +#define GFX9_NUM_SE 4 +#define RLCG_UCODE_LOADING_START_ADDRESS 0x2000 + +MODULE_FIRMWARE("amdgpu/vega10_ce.bin"); +MODULE_FIRMWARE("amdgpu/vega10_pfp.bin"); +MODULE_FIRMWARE("amdgpu/vega10_me.bin"); +MODULE_FIRMWARE("amdgpu/vega10_mec.bin"); +MODULE_FIRMWARE("amdgpu/vega10_mec2.bin"); +MODULE_FIRMWARE("amdgpu/vega10_rlc.bin"); + +static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = +{ + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14)}, + {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15)} +}; + +static const u32 golden_settings_gc_9_0[] = +{ + SOC15_REG_OFFSET(GC, 0, mmDB_DEBUG2), 0xf00ffeff, 0x00000400, + SOC15_REG_OFFSET(GC, 0, mmPA_SC_BINNER_EVENT_CNTL_3), 0x00000003, 0x82400024, + SOC15_REG_OFFSET(GC, 0, mmPA_SC_ENHANCE), 0x3fffffff, 0x00000001, + SOC15_REG_OFFSET(GC, 0, mmPA_SC_LINE_STIPPLE_STATE), 0x0000ff0f, 0x00000000, + SOC15_REG_OFFSET(GC, 0, mmTA_CNTL_AUX), 0xfffffeef, 0x010b0000, + SOC15_REG_OFFSET(GC, 0, mmTCP_CHAN_STEER_HI), 0xffffffff, 0x4a2c0e68, + SOC15_REG_OFFSET(GC, 0, mmTCP_CHAN_STEER_LO), 0xffffffff, 0xb5d3f197, + SOC15_REG_OFFSET(GC, 0, mmVGT_GS_MAX_WAVE_ID), 0x00000fff, 0x000003ff +}; + +static const u32 golden_settings_gc_9_0_vg10[] = +{ + SOC15_REG_OFFSET(GC, 0, mmCB_HW_CONTROL), 0x0000f000, 0x00012107, + SOC15_REG_OFFSET(GC, 0, mmCB_HW_CONTROL_3), 0x30000000, 0x10000000, + SOC15_REG_OFFSET(GC, 0, mmGB_ADDR_CONFIG), 0xffff77ff, 0x2a114042, + SOC15_REG_OFFSET(GC, 0, mmGB_ADDR_CONFIG_READ), 0xffff77ff, 0x2a114042, + SOC15_REG_OFFSET(GC, 0, mmPA_SC_ENHANCE_1), 0x00008000, 0x00048000, + SOC15_REG_OFFSET(GC, 0, mmRMI_UTCL1_CNTL2), 0x00030000, 0x00020000, + SOC15_REG_OFFSET(GC, 0, mmTD_CNTL), 0x00001800, 0x00000800, + SOC15_REG_OFFSET(GC, 0, mmSPI_CONFIG_CNTL_1),0x0000000f, 0x00000007 +}; + +#define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042 + +static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev); +static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev); +static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev); +static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev); +static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, + struct amdgpu_cu_info *cu_info); +static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev); +static void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance); + +static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) +{ + switch (adev->asic_type) { + case CHIP_VEGA10: + amdgpu_program_register_sequence(adev, + golden_settings_gc_9_0, + (const u32)ARRAY_SIZE(golden_settings_gc_9_0)); + amdgpu_program_register_sequence(adev, + golden_settings_gc_9_0_vg10, + (const u32)ARRAY_SIZE(golden_settings_gc_9_0_vg10)); + break; + default: + break; + } +} + +static void gfx_v9_0_scratch_init(struct amdgpu_device *adev) +{ + adev->gfx.scratch.num_reg = 7; + adev->gfx.scratch.reg_base = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); + adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; +} + +static void gfx_v9_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel, + bool wc, uint32_t reg, uint32_t val) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, WRITE_DATA_ENGINE_SEL(eng_sel) | + WRITE_DATA_DST_SEL(0) | + (wc ? WR_CONFIRM : 0)); + amdgpu_ring_write(ring, reg); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, val); +} + +static void gfx_v9_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel, + int mem_space, int opt, uint32_t addr0, + uint32_t addr1, uint32_t ref, uint32_t mask, + uint32_t inv) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5)); + amdgpu_ring_write(ring, + /* memory (1) or register (0) */ + (WAIT_REG_MEM_MEM_SPACE(mem_space) | + WAIT_REG_MEM_OPERATION(opt) | /* wait */ + WAIT_REG_MEM_FUNCTION(3) | /* equal */ + WAIT_REG_MEM_ENGINE(eng_sel))); + + if (mem_space) + BUG_ON(addr0 & 0x3); /* Dword align */ + amdgpu_ring_write(ring, addr0); + amdgpu_ring_write(ring, addr1); + amdgpu_ring_write(ring, ref); + amdgpu_ring_write(ring, mask); + amdgpu_ring_write(ring, inv); /* poll interval */ +} + +static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t scratch; + uint32_t tmp = 0; + unsigned i; + int r; + + r = amdgpu_gfx_scratch_get(adev, &scratch); + if (r) { + DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + r = amdgpu_ring_alloc(ring, 3); + if (r) { + DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", + ring->idx, r); + amdgpu_gfx_scratch_free(adev, scratch); + return r; + } + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(ring, 0xDEADBEEF); + amdgpu_ring_commit(ring); + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i < adev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", + ring->idx, i); + } else { + DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n", + ring->idx, scratch, tmp); + r = -EINVAL; + } + amdgpu_gfx_scratch_free(adev, scratch); + return r; +} + +static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ib ib; + struct dma_fence *f = NULL; + uint32_t scratch; + uint32_t tmp = 0; + long r; + + r = amdgpu_gfx_scratch_get(adev, &scratch); + if (r) { + DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + memset(&ib, 0, sizeof(ib)); + r = amdgpu_ib_get(adev, NULL, 256, &ib); + if (r) { + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + goto err1; + } + ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1); + ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START)); + ib.ptr[2] = 0xDEADBEEF; + ib.length_dw = 3; + + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + if (r) + goto err2; + + r = dma_fence_wait_timeout(f, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out.\n"); + r = -ETIMEDOUT; + goto err2; + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); + goto err2; + } + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + r = 0; + } else { + DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n", + scratch, tmp); + r = -EINVAL; + } +err2: + amdgpu_ib_free(adev, &ib, NULL); + dma_fence_put(f); +err1: + amdgpu_gfx_scratch_free(adev, scratch); + return r; +} + +static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) +{ + const char *chip_name; + char fw_name[30]; + 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"); + + switch (adev->asic_type) { + case CHIP_VEGA10: + chip_name = "vega10"; + break; + default: + BUG(); + } + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); + err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + if (err) + goto out; + 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); + if (err) + goto out; + 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); + if (err) + goto out; + 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); + if (err) + goto out; + 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); + if (!err) { + 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; + } + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_PFP]; + info->ucode_id = AMDGPU_UCODE_ID_CP_PFP; + info->fw = adev->gfx.pfp_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_ME]; + info->ucode_id = AMDGPU_UCODE_ID_CP_ME; + info->fw = adev->gfx.me_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_CE]; + info->ucode_id = AMDGPU_UCODE_ID_CP_CE; + info->fw = adev->gfx.ce_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_G]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_G; + info->fw = adev->gfx.rlc_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1]; + info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1; + info->fw = adev->gfx.mec_fw; + header = (const struct common_firmware_header *)info->fw->data; + cp_hdr = (const struct gfx_firmware_header_v1_0 *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes) - le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1_JT]; + info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1_JT; + info->fw = adev->gfx.mec_fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); + + if (adev->gfx.mec2_fw) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2]; + info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2; + info->fw = adev->gfx.mec2_fw; + header = (const struct common_firmware_header *)info->fw->data; + cp_hdr = (const struct gfx_firmware_header_v1_0 *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes) - le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2_JT]; + info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2_JT; + info->fw = adev->gfx.mec2_fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); + } + + } + +out: + if (err) { + dev_err(adev->dev, + "gfx9: Failed to load firmware \"%s\"\n", + fw_name); + release_firmware(adev->gfx.pfp_fw); + adev->gfx.pfp_fw = NULL; + release_firmware(adev->gfx.me_fw); + adev->gfx.me_fw = NULL; + release_firmware(adev->gfx.ce_fw); + adev->gfx.ce_fw = NULL; + release_firmware(adev->gfx.rlc_fw); + adev->gfx.rlc_fw = NULL; + release_firmware(adev->gfx.mec_fw); + adev->gfx.mec_fw = NULL; + release_firmware(adev->gfx.mec2_fw); + adev->gfx.mec2_fw = NULL; + } + return err; +} + +static void gfx_v9_0_mec_fini(struct amdgpu_device *adev) +{ + int r; + + if (adev->gfx.mec.hpd_eop_obj) { + r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false); + if (unlikely(r != 0)) + dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r); + amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj); + amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); + + amdgpu_bo_unref(&adev->gfx.mec.hpd_eop_obj); + adev->gfx.mec.hpd_eop_obj = NULL; + } + if (adev->gfx.mec.mec_fw_obj) { + r = amdgpu_bo_reserve(adev->gfx.mec.mec_fw_obj, false); + if (unlikely(r != 0)) + dev_warn(adev->dev, "(%d) reserve mec firmware bo failed\n", r); + amdgpu_bo_unpin(adev->gfx.mec.mec_fw_obj); + amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj); + + amdgpu_bo_unref(&adev->gfx.mec.mec_fw_obj); + adev->gfx.mec.mec_fw_obj = NULL; + } +} + +#define MEC_HPD_SIZE 2048 + +static int gfx_v9_0_mec_init(struct amdgpu_device *adev) +{ + int r; + u32 *hpd; + const __le32 *fw_data; + unsigned fw_size; + u32 *fw; + + const struct gfx_firmware_header_v1_0 *mec_hdr; + + /* + * we assign only 1 pipe because all other pipes will + * be handled by KFD + */ + adev->gfx.mec.num_mec = 1; + adev->gfx.mec.num_pipe = 1; + adev->gfx.mec.num_queue = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe * 8; + + if (adev->gfx.mec.hpd_eop_obj == NULL) { + r = amdgpu_bo_create(adev, + adev->gfx.mec.num_queue * MEC_HPD_SIZE, + PAGE_SIZE, true, + AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, + &adev->gfx.mec.hpd_eop_obj); + if (r) { + dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); + return r; + } + } + + r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false); + if (unlikely(r != 0)) { + gfx_v9_0_mec_fini(adev); + return r; + } + r = amdgpu_bo_pin(adev->gfx.mec.hpd_eop_obj, AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.hpd_eop_gpu_addr); + if (r) { + dev_warn(adev->dev, "(%d) pin HDP EOP bo failed\n", r); + gfx_v9_0_mec_fini(adev); + return r; + } + r = amdgpu_bo_kmap(adev->gfx.mec.hpd_eop_obj, (void **)&hpd); + if (r) { + dev_warn(adev->dev, "(%d) map HDP EOP bo failed\n", r); + gfx_v9_0_mec_fini(adev); + return r; + } + + memset(hpd, 0, adev->gfx.mec.hpd_eop_obj->tbo.mem.size); + + amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj); + amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); + + mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; + + fw_data = (const __le32 *) + (adev->gfx.mec_fw->data + + le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes) / 4; + + if (adev->gfx.mec.mec_fw_obj == NULL) { + r = amdgpu_bo_create(adev, + mec_hdr->header.ucode_size_bytes, + PAGE_SIZE, true, + AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, + &adev->gfx.mec.mec_fw_obj); + if (r) { + dev_warn(adev->dev, "(%d) create mec firmware bo failed\n", r); + return r; + } + } + + r = amdgpu_bo_reserve(adev->gfx.mec.mec_fw_obj, false); + if (unlikely(r != 0)) { + gfx_v9_0_mec_fini(adev); + return r; + } + r = amdgpu_bo_pin(adev->gfx.mec.mec_fw_obj, AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.mec_fw_gpu_addr); + if (r) { + dev_warn(adev->dev, "(%d) pin mec firmware bo failed\n", r); + gfx_v9_0_mec_fini(adev); + return r; + } + r = amdgpu_bo_kmap(adev->gfx.mec.mec_fw_obj, (void **)&fw); + if (r) { + dev_warn(adev->dev, "(%d) map firmware bo failed\n", r); + gfx_v9_0_mec_fini(adev); + return r; + } + memcpy(fw, fw_data, fw_size); + + amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj); + amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj); + + + return 0; +} + +static void gfx_v9_0_kiq_fini(struct amdgpu_device *adev) +{ + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + + amdgpu_bo_free_kernel(&kiq->eop_obj, &kiq->eop_gpu_addr, NULL); +} + +static int gfx_v9_0_kiq_init(struct amdgpu_device *adev) +{ + int r; + u32 *hpd; + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + + r = amdgpu_bo_create_kernel(adev, MEC_HPD_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, &kiq->eop_obj, + &kiq->eop_gpu_addr, (void **)&hpd); + if (r) { + dev_warn(adev->dev, "failed to create KIQ bo (%d).\n", r); + return r; + } + + memset(hpd, 0, MEC_HPD_SIZE); + + r = amdgpu_bo_reserve(kiq->eop_obj, false); + if (unlikely(r != 0)) + dev_warn(adev->dev, "(%d) reserve kiq eop bo failed\n", r); + amdgpu_bo_kunmap(kiq->eop_obj); + amdgpu_bo_unreserve(kiq->eop_obj); + + return 0; +} + +static int gfx_v9_0_kiq_init_ring(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_irq_src *irq) +{ + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + int r = 0; + + r = amdgpu_wb_get(adev, &adev->virt.reg_val_offs); + if (r) + return r; + + ring->adev = NULL; + ring->ring_obj = NULL; + ring->use_doorbell = true; + ring->doorbell_index = AMDGPU_DOORBELL_KIQ; + if (adev->gfx.mec2_fw) { + ring->me = 2; + ring->pipe = 0; + } else { + ring->me = 1; + ring->pipe = 1; + } + + irq->data = ring; + ring->queue = 0; + ring->eop_gpu_addr = kiq->eop_gpu_addr; + sprintf(ring->name, "kiq %d.%d.%d", ring->me, ring->pipe, ring->queue); + r = amdgpu_ring_init(adev, ring, 1024, + irq, AMDGPU_CP_KIQ_IRQ_DRIVER0); + if (r) + dev_warn(adev->dev, "(%d) failed to init kiq ring\n", r); + + return r; +} +static void gfx_v9_0_kiq_free_ring(struct amdgpu_ring *ring, + struct amdgpu_irq_src *irq) +{ + amdgpu_wb_free(ring->adev, ring->adev->virt.reg_val_offs); + amdgpu_ring_fini(ring); + irq->data = NULL; +} + +/* create MQD for each compute queue */ +static int gfx_v9_0_compute_mqd_sw_init(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = NULL; + int r, i; + + /* create MQD for KIQ */ + ring = &adev->gfx.kiq.ring; + if (!ring->mqd_obj) { + r = amdgpu_bo_create_kernel(adev, sizeof(struct v9_mqd), PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj, + &ring->mqd_gpu_addr, (void **)&ring->mqd_ptr); + if (r) { + dev_warn(adev->dev, "failed to create ring mqd ob (%d)", r); + return r; + } + + /*TODO: prepare MQD backup */ + } + + /* create MQD for each KCQ */ + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + if (!ring->mqd_obj) { + r = amdgpu_bo_create_kernel(adev, sizeof(struct v9_mqd), PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj, + &ring->mqd_gpu_addr, (void **)&ring->mqd_ptr); + if (r) { + dev_warn(adev->dev, "failed to create ring mqd ob (%d)", r); + return r; + } + + /* TODO: prepare MQD backup */ + } + } + + return 0; +} + +static void gfx_v9_0_compute_mqd_sw_fini(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = NULL; + int i; + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + amdgpu_bo_free_kernel(&ring->mqd_obj, &ring->mqd_gpu_addr, (void **)&ring->mqd_ptr); + } + + ring = &adev->gfx.kiq.ring; + amdgpu_bo_free_kernel(&ring->mqd_obj, &ring->mqd_gpu_addr, (void **)&ring->mqd_ptr); +} + +static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address) +{ + WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_IND_INDEX), + (wave << SQ_IND_INDEX__WAVE_ID__SHIFT) | + (simd << SQ_IND_INDEX__SIMD_ID__SHIFT) | + (address << SQ_IND_INDEX__INDEX__SHIFT) | + (SQ_IND_INDEX__FORCE_READ_MASK)); + return RREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_IND_DATA)); +} + +static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t thread, + uint32_t regno, uint32_t num, uint32_t *out) +{ + WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_IND_INDEX), + (wave << SQ_IND_INDEX__WAVE_ID__SHIFT) | + (simd << SQ_IND_INDEX__SIMD_ID__SHIFT) | + (regno << SQ_IND_INDEX__INDEX__SHIFT) | + (thread << SQ_IND_INDEX__THREAD_ID__SHIFT) | + (SQ_IND_INDEX__FORCE_READ_MASK) | + (SQ_IND_INDEX__AUTO_INCR_MASK)); + while (num--) + *(out++) = RREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_IND_DATA)); +} + +static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields) +{ + /* type 1 wave data */ + dst[(*no_fields)++] = 1; + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_STATUS); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_PC_LO); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_PC_HI); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_EXEC_LO); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_EXEC_HI); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_HW_ID); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_INST_DW0); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_INST_DW1); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_GPR_ALLOC); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_LDS_ALLOC); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_TRAPSTS); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_IB_STS); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_IB_DBG0); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_M0); +} + +static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t start, + uint32_t size, uint32_t *dst) +{ + wave_read_regs( + adev, simd, wave, 0, + start + SQIND_WAVE_SGPRS_OFFSET, size, dst); +} + + +static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { + .get_gpu_clock_counter = &gfx_v9_0_get_gpu_clock_counter, + .select_se_sh = &gfx_v9_0_select_se_sh, + .read_wave_data = &gfx_v9_0_read_wave_data, + .read_wave_sgprs = &gfx_v9_0_read_wave_sgprs, +}; + +static void gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) +{ + u32 gb_addr_config; + + adev->gfx.funcs = &gfx_v9_0_gfx_funcs; + + switch (adev->asic_type) { + case CHIP_VEGA10: + adev->gfx.config.max_shader_engines = 4; + adev->gfx.config.max_tile_pipes = 8; //?? + adev->gfx.config.max_cu_per_sh = 16; + adev->gfx.config.max_sh_per_se = 1; + adev->gfx.config.max_backends_per_se = 4; + adev->gfx.config.max_texture_channel_caches = 16; + adev->gfx.config.max_gprs = 256; + adev->gfx.config.max_gs_threads = 32; + adev->gfx.config.max_hw_contexts = 8; + + adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; + adev->gfx.config.sc_prim_fifo_size_backend = 0x100; + adev->gfx.config.sc_hiz_tile_fifo_size = 0x30; + adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; + gb_addr_config = VEGA10_GB_ADDR_CONFIG_GOLDEN; + break; + default: + BUG(); + break; + } + + adev->gfx.config.gb_addr_config = gb_addr_config; + + adev->gfx.config.gb_addr_config_fields.num_pipes = 1 << + REG_GET_FIELD( + adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, + NUM_PIPES); + adev->gfx.config.gb_addr_config_fields.num_banks = 1 << + REG_GET_FIELD( + adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, + NUM_BANKS); + adev->gfx.config.gb_addr_config_fields.max_compress_frags = 1 << + REG_GET_FIELD( + adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, + MAX_COMPRESSED_FRAGS); + adev->gfx.config.gb_addr_config_fields.num_rb_per_se = 1 << + REG_GET_FIELD( + adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, + NUM_RB_PER_SE); + adev->gfx.config.gb_addr_config_fields.num_se = 1 << + REG_GET_FIELD( + adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, + NUM_SHADER_ENGINES); + adev->gfx.config.gb_addr_config_fields.pipe_interleave_size = 1 << (8 + + REG_GET_FIELD( + adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, + PIPE_INTERLEAVE_SIZE)); +} + +static int gfx_v9_0_ngg_create_buf(struct amdgpu_device *adev, + struct amdgpu_ngg_buf *ngg_buf, + int size_se, + int default_size_se) +{ + int r; + + if (size_se < 0) { + dev_err(adev->dev, "Buffer size is invalid: %d\n", size_se); + return -EINVAL; + } + size_se = size_se ? size_se : default_size_se; + + ngg_buf->size = size_se * GFX9_NUM_SE; + r = amdgpu_bo_create_kernel(adev, ngg_buf->size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &ngg_buf->bo, + &ngg_buf->gpu_addr, + NULL); + if (r) { + dev_err(adev->dev, "(%d) failed to create NGG buffer\n", r); + return r; + } + ngg_buf->bo_size = amdgpu_bo_size(ngg_buf->bo); + + return r; +} + +static int gfx_v9_0_ngg_fini(struct amdgpu_device *adev) +{ + int i; + + for (i = 0; i < NGG_BUF_MAX; i++) + amdgpu_bo_free_kernel(&adev->gfx.ngg.buf[i].bo, + &adev->gfx.ngg.buf[i].gpu_addr, + NULL); + + memset(&adev->gfx.ngg.buf[0], 0, + sizeof(struct amdgpu_ngg_buf) * NGG_BUF_MAX); + + adev->gfx.ngg.init = false; + + return 0; +} + +static int gfx_v9_0_ngg_init(struct amdgpu_device *adev) +{ + int r; + + if (!amdgpu_ngg || adev->gfx.ngg.init == true) + return 0; + + /* GDS reserve memory: 64 bytes alignment */ + adev->gfx.ngg.gds_reserve_size = ALIGN(5 * 4, 0x40); + adev->gds.mem.total_size -= adev->gfx.ngg.gds_reserve_size; + adev->gds.mem.gfx_partition_size -= adev->gfx.ngg.gds_reserve_size; + adev->gfx.ngg.gds_reserve_addr = amdgpu_gds_reg_offset[0].mem_base; + adev->gfx.ngg.gds_reserve_addr += adev->gds.mem.gfx_partition_size; + + /* Primitive Buffer */ + r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[PRIM], + amdgpu_prim_buf_per_se, + 64 * 1024); + if (r) { + dev_err(adev->dev, "Failed to create Primitive Buffer\n"); + goto err; + } + + /* Position Buffer */ + r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[POS], + amdgpu_pos_buf_per_se, + 256 * 1024); + if (r) { + dev_err(adev->dev, "Failed to create Position Buffer\n"); + goto err; + } + + /* Control Sideband */ + r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[CNTL], + amdgpu_cntl_sb_buf_per_se, + 256); + if (r) { + dev_err(adev->dev, "Failed to create Control Sideband Buffer\n"); + goto err; + } + + /* Parameter Cache, not created by default */ + if (amdgpu_param_buf_per_se <= 0) + goto out; + + r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[PARAM], + amdgpu_param_buf_per_se, + 512 * 1024); + if (r) { + dev_err(adev->dev, "Failed to create Parameter Cache\n"); + goto err; + } + +out: + adev->gfx.ngg.init = true; + return 0; +err: + gfx_v9_0_ngg_fini(adev); + return r; +} + +static int gfx_v9_0_ngg_en(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0]; + int r; + u32 data; + u32 size; + u32 base; + + if (!amdgpu_ngg) + return 0; + + /* Program buffer size */ + data = 0; + size = adev->gfx.ngg.buf[PRIM].size / 256; + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, size); + + size = adev->gfx.ngg.buf[POS].size / 256; + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, size); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_BUF_RESOURCE_1), data); + + data = 0; + size = adev->gfx.ngg.buf[CNTL].size / 256; + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, size); + + size = adev->gfx.ngg.buf[PARAM].size / 1024; + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, size); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_BUF_RESOURCE_2), data); + + /* Program buffer base address */ + base = lower_32_bits(adev->gfx.ngg.buf[PRIM].gpu_addr); + data = REG_SET_FIELD(0, WD_INDEX_BUF_BASE, BASE, base); + WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_INDEX_BUF_BASE), data); + + base = upper_32_bits(adev->gfx.ngg.buf[PRIM].gpu_addr); + data = REG_SET_FIELD(0, WD_INDEX_BUF_BASE_HI, BASE_HI, base); + WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_INDEX_BUF_BASE_HI), data); + + base = lower_32_bits(adev->gfx.ngg.buf[POS].gpu_addr); + data = REG_SET_FIELD(0, WD_POS_BUF_BASE, BASE, base); + WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_POS_BUF_BASE), data); + + base = upper_32_bits(adev->gfx.ngg.buf[POS].gpu_addr); + data = REG_SET_FIELD(0, WD_POS_BUF_BASE_HI, BASE_HI, base); + WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_POS_BUF_BASE_HI), data); + + base = lower_32_bits(adev->gfx.ngg.buf[CNTL].gpu_addr); + data = REG_SET_FIELD(0, WD_CNTL_SB_BUF_BASE, BASE, base); + WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_CNTL_SB_BUF_BASE), data); + + base = upper_32_bits(adev->gfx.ngg.buf[CNTL].gpu_addr); + data = REG_SET_FIELD(0, WD_CNTL_SB_BUF_BASE_HI, BASE_HI, base); + WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_CNTL_SB_BUF_BASE_HI), data); + + /* Clear GDS reserved memory */ + r = amdgpu_ring_alloc(ring, 17); + if (r) { + DRM_ERROR("amdgpu: NGG failed to lock ring %d (%d).\n", + ring->idx, r); + return r; + } + + gfx_v9_0_write_data_to_reg(ring, 0, false, + amdgpu_gds_reg_offset[0].mem_size, + (adev->gds.mem.total_size + + adev->gfx.ngg.gds_reserve_size) >> + AMDGPU_GDS_SHIFT); + + amdgpu_ring_write(ring, PACKET3(PACKET3_DMA_DATA, 5)); + amdgpu_ring_write(ring, (PACKET3_DMA_DATA_CP_SYNC | + PACKET3_DMA_DATA_SRC_SEL(2))); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, adev->gfx.ngg.gds_reserve_addr); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, adev->gfx.ngg.gds_reserve_size); + + + gfx_v9_0_write_data_to_reg(ring, 0, false, + amdgpu_gds_reg_offset[0].mem_size, 0); + + amdgpu_ring_commit(ring); + + return 0; +} + +static int gfx_v9_0_sw_init(void *handle) +{ + int i, r; + struct amdgpu_ring *ring; + struct amdgpu_kiq *kiq; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + /* KIQ event */ + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_GRBM_CP, 178, &adev->gfx.kiq.irq); + if (r) + return r; + + /* EOP Event */ + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_GRBM_CP, 181 |