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 | 2417 |
1 files changed, 1533 insertions, 884 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 044076ec1d03..0148d7ff34d9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -47,8 +47,10 @@ #include "amdgpu_ras.h" +#include "amdgpu_ring_mux.h" #include "gfx_v9_4.h" #include "gfx_v9_0.h" +#include "gfx_v9_0_cleaner_shader.h" #include "gfx_v9_4_2.h" #include "asic_reg/pwr/pwr_10_0_offset.h" @@ -56,6 +58,7 @@ #include "asic_reg/gc/gc_9_0_default.h" #define GFX9_NUM_GFX_RINGS 1 +#define GFX9_NUM_SW_GFX_RINGS 2 #define GFX9_MEC_HPD_SIZE 4096 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L #define RLC_SAVE_RESTORE_ADDR_STARTING_OFFSET 0x00000000L @@ -126,6 +129,8 @@ MODULE_FIRMWARE("amdgpu/green_sardine_rlc.bin"); MODULE_FIRMWARE("amdgpu/aldebaran_mec.bin"); MODULE_FIRMWARE("amdgpu/aldebaran_mec2.bin"); MODULE_FIRMWARE("amdgpu/aldebaran_rlc.bin"); +MODULE_FIRMWARE("amdgpu/aldebaran_sjt_mec.bin"); +MODULE_FIRMWARE("amdgpu/aldebaran_sjt_mec2.bin"); #define mmTCP_CHAN_STEER_0_ARCT 0x0b03 #define mmTCP_CHAN_STEER_0_ARCT_BASE_IDX 0 @@ -140,6 +145,167 @@ MODULE_FIRMWARE("amdgpu/aldebaran_rlc.bin"); #define mmTCP_CHAN_STEER_5_ARCT 0x0b0c #define mmTCP_CHAN_STEER_5_ARCT_BASE_IDX 0 +#define mmGOLDEN_TSC_COUNT_UPPER_Renoir 0x0025 +#define mmGOLDEN_TSC_COUNT_UPPER_Renoir_BASE_IDX 1 +#define mmGOLDEN_TSC_COUNT_LOWER_Renoir 0x0026 +#define mmGOLDEN_TSC_COUNT_LOWER_Renoir_BASE_IDX 1 + +static const struct amdgpu_hwip_reg_entry gc_reg_list_9[] = { + SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS2), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT1), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT2), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STALLED_STAT1), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STALLED_STAT1), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_BUSY_STAT), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_BUSY_STAT), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_BUSY_STAT), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_ERROR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_BASE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_BASE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_BASE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_BASE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_CMD_BUFSZ), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_CMD_BUFSZ), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_CMD_BUFSZ), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_CMD_BUFSZ), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_LO), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BUFSZ), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_LO), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BUFSZ), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_LO), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BUFSZ), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_LO), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BUFSZ), + SOC15_REG_ENTRY_STR(GC, 0, mmCPF_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmCPC_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmCPG_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmGDS_PROTECTION_FAULT), + SOC15_REG_ENTRY_STR(GC, 0, mmGDS_VM_PROTECTION_FAULT), + SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, mmPA_CL_CNTL_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmRMI_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmSQC_DCACHE_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmSQC_ICACHE_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmSQ_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmTCP_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmWD_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, mmVM_L2_PROTECTION_FAULT_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_DEBUG), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_INSTR_PNTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC1_INSTR_PNTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC2_INSTR_PNTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_INSTR_PNTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_INSTR_PNTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_STAT), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_COMMAND), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_MESSAGE), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_1), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_2), + SOC15_REG_ENTRY_STR(GC, 0, mmSMU_RLC_RESPONSE), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SAFE_MODE), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_SAFE_MODE), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_INT_STAT), + SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_GENERAL_6), + /* SE status registers */ + SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE0), + SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE1), + SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE2), + SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE3), + /* packet headers */ + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP) +}; + +static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_9[] = { + /* compute queue registers */ + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_VMID), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_ACTIVE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PERSISTENT_STATE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PIPE_PRIORITY), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUEUE_PRIORITY), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUANTUM), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_EVENTS), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_LO), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_OFFSET), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_SIZE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_WG_STATE_OFFSET), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_SIZE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_GDS_RESOURCE_STATE), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_ERROR), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR_MEM), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_LO), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_HI), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_GFX_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP) +}; + enum ta_ras_gfx_subblock { /*CPC*/ TA_RAS_BLOCK__GFX_CPC_INDEX_START = 0, @@ -734,75 +900,6 @@ static const u32 GFX_RLC_SRM_INDEX_CNTL_DATA_OFFSETS[] = mmRLC_SRM_INDEX_CNTL_DATA_7 - mmRLC_SRM_INDEX_CNTL_DATA_0, }; -static void gfx_v9_0_rlcg_w(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag) -{ - static void *scratch_reg0; - static void *scratch_reg1; - static void *scratch_reg2; - static void *scratch_reg3; - static void *spare_int; - static uint32_t grbm_cntl; - static uint32_t grbm_idx; - - scratch_reg0 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG0_BASE_IDX] + mmSCRATCH_REG0)*4; - scratch_reg1 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG1)*4; - scratch_reg2 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG2)*4; - scratch_reg3 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG3)*4; - spare_int = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmRLC_SPARE_INT_BASE_IDX] + mmRLC_SPARE_INT)*4; - - grbm_cntl = adev->reg_offset[GC_HWIP][0][mmGRBM_GFX_CNTL_BASE_IDX] + mmGRBM_GFX_CNTL; - grbm_idx = adev->reg_offset[GC_HWIP][0][mmGRBM_GFX_INDEX_BASE_IDX] + mmGRBM_GFX_INDEX; - - if (amdgpu_sriov_runtime(adev)) { - pr_err("shouldn't call rlcg write register during runtime\n"); - return; - } - - if (offset == grbm_cntl || offset == grbm_idx) { - if (offset == grbm_cntl) - writel(v, scratch_reg2); - else if (offset == grbm_idx) - writel(v, scratch_reg3); - - writel(v, ((void __iomem *)adev->rmmio) + (offset * 4)); - } else { - uint32_t i = 0; - uint32_t retries = 50000; - - writel(v, scratch_reg0); - writel(offset | 0x80000000, scratch_reg1); - writel(1, spare_int); - for (i = 0; i < retries; i++) { - u32 tmp; - - tmp = readl(scratch_reg1); - if (!(tmp & 0x80000000)) - break; - - udelay(10); - } - if (i >= retries) - pr_err("timeout: rlcg program reg:0x%05x failed !\n", offset); - } - -} - -static void gfx_v9_0_rlcg_wreg(struct amdgpu_device *adev, u32 offset, - u32 v, u32 acc_flags, u32 hwip) -{ - if ((acc_flags & AMDGPU_REGS_RLC) && - amdgpu_sriov_fullaccess(adev)) { - gfx_v9_0_rlcg_w(adev, offset, v, acc_flags); - - return; - } - - if (acc_flags & AMDGPU_REGS_NO_KIQ) - WREG32_NO_KIQ(offset, v); - else - WREG32(offset, v); -} - #define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042 #define VEGA12_GB_ADDR_CONFIG_GOLDEN 0x24104041 #define RAVEN_GB_ADDR_CONFIG_GOLDEN 0x24000042 @@ -815,17 +912,27 @@ 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_ring_emit_de_meta(struct amdgpu_ring *ring); +static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume, bool usegds); static u64 gfx_v9_0_ring_get_rptr_compute(struct amdgpu_ring *ring); -static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status); static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev, - void *inject_if); + void *inject_if, uint32_t instance_mask); static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev); +static void gfx_v9_0_update_spm_vmid_internal(struct amdgpu_device *adev, + unsigned int vmid); +static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id); +static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id); static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring, uint64_t queue_mask) { + struct amdgpu_device *adev = kiq_ring->adev; + u64 shader_mc_addr; + + /* Cleaner shader MC address */ + shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8; + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6)); amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) | @@ -835,8 +942,8 @@ static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring, lower_32_bits(queue_mask)); /* queue mask lo */ amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask)); /* queue mask hi */ - amdgpu_ring_write(kiq_ring, 0); /* gws mask lo */ - amdgpu_ring_write(kiq_ring, 0); /* gws mask hi */ + amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */ + amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */ amdgpu_ring_write(kiq_ring, 0); /* oac mask */ amdgpu_ring_write(kiq_ring, 0); /* gds heap base:0, gds heap size:0 */ } @@ -844,9 +951,8 @@ static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring, static void gfx_v9_0_kiq_map_queues(struct amdgpu_ring *kiq_ring, struct amdgpu_ring *ring) { - struct amdgpu_device *adev = kiq_ring->adev; uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj); - uint64_t wptr_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + uint64_t wptr_addr = ring->wptr_gpu_addr; uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); @@ -889,9 +995,10 @@ static void gfx_v9_0_kiq_unmap_queues(struct amdgpu_ring *kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); if (action == PREEMPT_QUEUES_NO_UNMAP) { - amdgpu_ring_write(kiq_ring, lower_32_bits(gpu_addr)); - amdgpu_ring_write(kiq_ring, upper_32_bits(gpu_addr)); - amdgpu_ring_write(kiq_ring, seq); + amdgpu_ring_write(kiq_ring, lower_32_bits(ring->wptr & ring->buf_mask)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + } else { amdgpu_ring_write(kiq_ring, 0); amdgpu_ring_write(kiq_ring, 0); @@ -933,12 +1040,47 @@ static void gfx_v9_0_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring, PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type)); } + +static void gfx_v9_0_kiq_reset_hw_queue(struct amdgpu_ring *kiq_ring, uint32_t queue_type, + uint32_t me_id, uint32_t pipe_id, uint32_t queue_id, + uint32_t xcc_id, uint32_t vmid) +{ + struct amdgpu_device *adev = kiq_ring->adev; + unsigned i; + + /* enter save mode */ + amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id); + mutex_lock(&adev->srbm_mutex); + soc15_grbm_select(adev, me_id, pipe_id, queue_id, 0, 0); + + if (queue_type == AMDGPU_RING_TYPE_COMPUTE) { + WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 0x2); + WREG32_SOC15(GC, 0, mmSPI_COMPUTE_QUEUE_RESET, 0x1); + /* wait till dequeue take effects */ + for (i = 0; i < adev->usec_timeout; i++) { + if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1)) + break; + udelay(1); + } + if (i >= adev->usec_timeout) + dev_err(adev->dev, "fail to wait on hqd deactive\n"); + } else { + dev_err(adev->dev, "reset queue_type(%d) not supported\n", queue_type); + } + + soc15_grbm_select(adev, 0, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + /* exit safe mode */ + amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id); +} + static const struct kiq_pm4_funcs gfx_v9_0_kiq_pm4_funcs = { .kiq_set_resources = gfx_v9_0_kiq_set_resources, .kiq_map_queues = gfx_v9_0_kiq_map_queues, .kiq_unmap_queues = gfx_v9_0_kiq_unmap_queues, .kiq_query_status = gfx_v9_0_kiq_query_status, .kiq_invalidate_tlbs = gfx_v9_0_kiq_invalidate_tlbs, + .kiq_reset_hw_queue = gfx_v9_0_kiq_reset_hw_queue, .set_resources_size = 8, .map_queues_size = 7, .unmap_queues_size = 6, @@ -948,13 +1090,13 @@ static const struct kiq_pm4_funcs gfx_v9_0_kiq_pm4_funcs = { static void gfx_v9_0_set_kiq_pm4_funcs(struct amdgpu_device *adev) { - adev->gfx.kiq.pmf = &gfx_v9_0_kiq_pm4_funcs; + adev->gfx.kiq[0].pmf = &gfx_v9_0_kiq_pm4_funcs; } static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_VEGA10: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): soc15_program_register_sequence(adev, golden_settings_gc_9_0, ARRAY_SIZE(golden_settings_gc_9_0)); @@ -962,7 +1104,7 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_0_vg10, ARRAY_SIZE(golden_settings_gc_9_0_vg10)); break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): soc15_program_register_sequence(adev, golden_settings_gc_9_2_1, ARRAY_SIZE(golden_settings_gc_9_2_1)); @@ -970,7 +1112,7 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_2_1_vg12, ARRAY_SIZE(golden_settings_gc_9_2_1_vg12)); break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): soc15_program_register_sequence(adev, golden_settings_gc_9_0, ARRAY_SIZE(golden_settings_gc_9_0)); @@ -978,12 +1120,13 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_0_vg20, ARRAY_SIZE(golden_settings_gc_9_0_vg20)); break; - case CHIP_ARCTURUS: + case IP_VERSION(9, 4, 1): soc15_program_register_sequence(adev, golden_settings_gc_9_4_1_arct, ARRAY_SIZE(golden_settings_gc_9_4_1_arct)); break; - case CHIP_RAVEN: + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): soc15_program_register_sequence(adev, golden_settings_gc_9_1, ARRAY_SIZE(golden_settings_gc_9_1)); if (adev->apu_flags & AMD_APU_IS_RAVEN2) @@ -995,12 +1138,12 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_1_rv1, ARRAY_SIZE(golden_settings_gc_9_1_rv1)); break; - case CHIP_RENOIR: + case IP_VERSION(9, 3, 0): soc15_program_register_sequence(adev, golden_settings_gc_9_1_rn, ARRAY_SIZE(golden_settings_gc_9_1_rn)); return; /* for renoir, don't need common goldensetting */ - case CHIP_ALDEBARAN: + case IP_VERSION(9, 4, 2): gfx_v9_4_2_init_golden_registers(adev, adev->smuio.funcs->get_die_id(adev)); break; @@ -1008,19 +1151,12 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) break; } - if ((adev->asic_type != CHIP_ARCTURUS) && - (adev->asic_type != CHIP_ALDEBARAN)) + if ((amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) && + (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2))) soc15_program_register_sequence(adev, golden_settings_gc_9_x_common, (const u32)ARRAY_SIZE(golden_settings_gc_9_x_common)); } -static void gfx_v9_0_scratch_init(struct amdgpu_device *adev) -{ - adev->gfx.scratch.num_reg = 8; - 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) { @@ -1058,22 +1194,18 @@ static void gfx_v9_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel, static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - uint32_t scratch; + uint32_t scratch = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); uint32_t tmp = 0; unsigned i; int r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) - return r; - WREG32(scratch, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); if (r) - goto error_free_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, scratch - PACKET3_SET_UCONFIG_REG_START); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); @@ -1086,9 +1218,6 @@ static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring) if (i >= adev->usec_timeout) r = -ETIMEDOUT; - -error_free_scratch: - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -1110,8 +1239,8 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) gpu_addr = adev->wb.gpu_addr + (index * 4); adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, 16, - AMDGPU_IB_POOL_DIRECT, &ib); + + r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib); if (r) goto err1; @@ -1141,7 +1270,7 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = -EINVAL; err2: - amdgpu_ib_free(adev, &ib, NULL); + amdgpu_ib_free(&ib, NULL); dma_fence_put(f); err1: amdgpu_device_wb_free(adev, index); @@ -1151,57 +1280,31 @@ err1: static void gfx_v9_0_free_microcode(struct amdgpu_device *adev) { - 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; + amdgpu_ucode_release(&adev->gfx.pfp_fw); + amdgpu_ucode_release(&adev->gfx.me_fw); + amdgpu_ucode_release(&adev->gfx.ce_fw); + amdgpu_ucode_release(&adev->gfx.rlc_fw); + amdgpu_ucode_release(&adev->gfx.mec_fw); + amdgpu_ucode_release(&adev->gfx.mec2_fw); kfree(adev->gfx.rlc.register_list_format); } -static void gfx_v9_0_init_rlc_ext_microcode(struct amdgpu_device *adev) -{ - const struct rlc_firmware_header_v2_1 *rlc_hdr; - - rlc_hdr = (const struct rlc_firmware_header_v2_1 *)adev->gfx.rlc_fw->data; - adev->gfx.rlc_srlc_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_ucode_ver); - adev->gfx.rlc_srlc_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_feature_ver); - adev->gfx.rlc.save_restore_list_cntl_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_cntl_size_bytes); - adev->gfx.rlc.save_restore_list_cntl = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_cntl_offset_bytes); - adev->gfx.rlc_srlg_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_ucode_ver); - adev->gfx.rlc_srlg_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_feature_ver); - adev->gfx.rlc.save_restore_list_gpm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_gpm_size_bytes); - adev->gfx.rlc.save_restore_list_gpm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_gpm_offset_bytes); - adev->gfx.rlc_srls_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_ucode_ver); - adev->gfx.rlc_srls_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_feature_ver); - adev->gfx.rlc.save_restore_list_srm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_srm_size_bytes); - adev->gfx.rlc.save_restore_list_srm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_srm_offset_bytes); - adev->gfx.rlc.reg_list_format_direct_reg_list_length = - le32_to_cpu(rlc_hdr->reg_list_format_direct_reg_list_length); -} - static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) { adev->gfx.me_fw_write_wait = false; adev->gfx.mec_fw_write_wait = false; - if ((adev->asic_type != CHIP_ARCTURUS) && + if ((amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) && + (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)) && ((adev->gfx.mec_fw_version < 0x000001a5) || - (adev->gfx.mec_feature_version < 46) || - (adev->gfx.pfp_fw_version < 0x000000b7) || - (adev->gfx.pfp_feature_version < 46))) + (adev->gfx.mec_feature_version < 46) || + (adev->gfx.pfp_fw_version < 0x000000b7) || + (adev->gfx.pfp_feature_version < 46))) DRM_WARN_ONCE("CP firmware version too old, please update!"); - switch (adev->asic_type) { - case CHIP_VEGA10: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): if ((adev->gfx.me_fw_version >= 0x0000009c) && (adev->gfx.me_feature_version >= 42) && (adev->gfx.pfp_fw_version >= 0x000000b1) && @@ -1212,7 +1315,7 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) (adev->gfx.mec_feature_version >= 42)) adev->gfx.mec_fw_write_wait = true; break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): if ((adev->gfx.me_fw_version >= 0x0000009c) && (adev->gfx.me_feature_version >= 44) && (adev->gfx.pfp_fw_version >= 0x000000b2) && @@ -1223,7 +1326,7 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) (adev->gfx.mec_feature_version >= 44)) adev->gfx.mec_fw_write_wait = true; break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): if ((adev->gfx.me_fw_version >= 0x0000009c) && (adev->gfx.me_feature_version >= 44) && (adev->gfx.pfp_fw_version >= 0x000000b2) && @@ -1234,7 +1337,8 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) (adev->gfx.mec_feature_version >= 44)) adev->gfx.mec_fw_write_wait = true; break; - case CHIP_RAVEN: + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 2, 2): if ((adev->gfx.me_fw_version >= 0x0000009c) && (adev->gfx.me_feature_version >= 42) && (adev->gfx.pfp_fw_version >= 0x000000b1) && @@ -1267,6 +1371,12 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = { { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 }, /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */ { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 }, + /* Apple MacBook Pro (15-inch, 2019) Radeon Pro Vega 20 4 GB */ + { 0x1002, 0x69af, 0x106b, 0x019a, 0xc0 }, + /* https://bbs.openkylin.top/t/topic/171497 */ + { 0x1002, 0x15d8, 0x19e5, 0x3e14, 0xc2 }, + /* HP 705G4 DM with R5 2400G */ + { 0x1002, 0x15dd, 0x103c, 0x8464, 0xd6 }, { 0, 0, 0, 0, 0 }, }; @@ -1295,17 +1405,28 @@ static bool is_raven_kicker(struct amdgpu_device *adev) return false; } +static bool check_if_enlarge_doorbell_range(struct amdgpu_device *adev) +{ + if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 3, 0)) && + (adev->gfx.me_fw_version >= 0x000000a5) && + (adev->gfx.me_feature_version >= 52)) + return true; + else + return false; +} + static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev) { if (gfx_v9_0_should_disable_gfxoff(adev->pdev)) adev->pm.pp_feature &= ~PP_GFXOFF_MASK; - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): break; - case CHIP_RAVEN: + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): if (!((adev->apu_flags & AMD_APU_IS_RAVEN2) || (adev->apu_flags & AMD_APU_IS_PICASSO)) && ((!is_raven_kicker(adev) && @@ -1319,7 +1440,7 @@ static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev) AMD_PG_SUPPORT_CP | AMD_PG_SUPPORT_RLC_SMU_HS; break; - case CHIP_RENOIR: + case IP_VERSION(9, 3, 0): if (adev->pm.pp_feature & PP_GFXOFF_MASK) adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | AMD_PG_SUPPORT_CP | @@ -1331,95 +1452,45 @@ static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev) } static int gfx_v9_0_init_cp_gfx_microcode(struct amdgpu_device *adev, - const char *chip_name) + 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; - 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); + err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_pfp.bin", chip_name); 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); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); - err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_me.bin", chip_name); 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); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); - err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_ce.bin", chip_name); 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); - - 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); - } + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE); 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; + amdgpu_ucode_release(&adev->gfx.pfp_fw); + amdgpu_ucode_release(&adev->gfx.me_fw); + amdgpu_ucode_release(&adev->gfx.ce_fw); } return err; } static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev, - const char *chip_name) + char *chip_name) { - char fw_name[30]; int err; - struct amdgpu_firmware_info *info = NULL; - const struct common_firmware_header *header = NULL; const struct rlc_firmware_header_v2_0 *rlc_hdr; - unsigned int *tmp = NULL; - unsigned int i = 0; uint16_t version_major; uint16_t version_minor; uint32_t smu_version; @@ -1435,270 +1506,114 @@ static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev, if (!strcmp(chip_name, "picasso") && (((adev->pdev->revision >= 0xC8) && (adev->pdev->revision <= 0xCF)) || ((adev->pdev->revision >= 0xD8) && (adev->pdev->revision <= 0xDF)))) - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc_am4.bin", chip_name); + err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_rlc_am4.bin", chip_name); else if (!strcmp(chip_name, "raven") && (amdgpu_pm_load_smu_firmware(adev, &smu_version) == 0) && (smu_version >= 0x41e2b)) /** *SMC is loaded by SBIOS on APU and it's able to get the SMU version directly. */ - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_kicker_rlc.bin", chip_name); + err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_kicker_rlc.bin", chip_name); else - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); - err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); + err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_rlc.bin", chip_name); if (err) goto out; - err = amdgpu_ucode_validate(adev->gfx.rlc_fw); - rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; + rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; version_major = le16_to_cpu(rlc_hdr->header.header_version_major); version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor); - if (version_major == 2 && version_minor == 1) - adev->gfx.rlc.is_rlc_v2_1 = true; - - adev->gfx.rlc_fw_version = le32_to_cpu(rlc_hdr->header.ucode_version); - adev->gfx.rlc_feature_version = le32_to_cpu(rlc_hdr->ucode_feature_version); - adev->gfx.rlc.save_and_restore_offset = - le32_to_cpu(rlc_hdr->save_and_restore_offset); - adev->gfx.rlc.clear_state_descriptor_offset = - le32_to_cpu(rlc_hdr->clear_state_descriptor_offset); - adev->gfx.rlc.avail_scratch_ram_locations = - le32_to_cpu(rlc_hdr->avail_scratch_ram_locations); - adev->gfx.rlc.reg_restore_list_size = - le32_to_cpu(rlc_hdr->reg_restore_list_size); - adev->gfx.rlc.reg_list_format_start = - le32_to_cpu(rlc_hdr->reg_list_format_start); - adev->gfx.rlc.reg_list_format_separate_start = - le32_to_cpu(rlc_hdr->reg_list_format_separate_start); - adev->gfx.rlc.starting_offsets_start = - le32_to_cpu(rlc_hdr->starting_offsets_start); - adev->gfx.rlc.reg_list_format_size_bytes = - le32_to_cpu(rlc_hdr->reg_list_format_size_bytes); - adev->gfx.rlc.reg_list_size_bytes = - le32_to_cpu(rlc_hdr->reg_list_size_bytes); - adev->gfx.rlc.register_list_format = - kmalloc(adev->gfx.rlc.reg_list_format_size_bytes + - adev->gfx.rlc.reg_list_size_bytes, GFP_KERNEL); - if (!adev->gfx.rlc.register_list_format) { - err = -ENOMEM; - goto out; - } - - tmp = (unsigned int *)((uintptr_t)rlc_hdr + - le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); - for (i = 0 ; i < (adev->gfx.rlc.reg_list_format_size_bytes >> 2); i++) - adev->gfx.rlc.register_list_format[i] = le32_to_cpu(tmp[i]); - - adev->gfx.rlc.register_restore = adev->gfx.rlc.register_list_format + i; - - tmp = (unsigned int *)((uintptr_t)rlc_hdr + - le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); - for (i = 0 ; i < (adev->gfx.rlc.reg_list_size_bytes >> 2); i++) - adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); - - if (adev->gfx.rlc.is_rlc_v2_1) - gfx_v9_0_init_rlc_ext_microcode(adev); - - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - 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); - - if (adev->gfx.rlc.is_rlc_v2_1 && - adev->gfx.rlc.save_restore_list_cntl_size_bytes && - adev->gfx.rlc.save_restore_list_gpm_size_bytes && - adev->gfx.rlc.save_restore_list_srm_size_bytes) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_cntl_size_bytes, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_gpm_size_bytes, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_srm_size_bytes, PAGE_SIZE); - } - } - + err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor); out: - if (err) { - dev_err(adev->dev, - "gfx9: Failed to load firmware \"%s\"\n", - fw_name); - release_firmware(adev->gfx.rlc_fw); - adev->gfx.rlc_fw = NULL; - } + if (err) + amdgpu_ucode_release(&adev->gfx.rlc_fw); + return err; } static bool gfx_v9_0_load_mec2_fw_bin_support(struct amdgpu_device *adev) { - if (adev->asic_type == CHIP_ALDEBARAN || - adev->asic_type == CHIP_ARCTURUS || - adev->asic_type == CHIP_RENOIR) + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) || + amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) || + amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 3, 0)) return false; return true; } static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev, - const char *chip_name) + 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; - 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 (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN)) + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_sjt_mec.bin", chip_name); + else + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec.bin", chip_name); 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); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT); if (gfx_v9_0_load_mec2_fw_bin_support(adev)) { - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); - err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN)) + err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_sjt_mec2.bin", chip_name); + else + err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec2.bin", chip_name); 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); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT); } else { err = 0; - adev->gfx.mec2_fw = NULL; + amdgpu_ucode_release(&adev->gfx.mec2_fw); } } else { adev->gfx.mec2_fw_version = adev->gfx.mec_fw_version; adev->gfx.mec2_feature_version = adev->gfx.mec_feature_version; } - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - 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); - - /* TODO: Determine if MEC2 JT FW loading can be removed - for all GFX V9 asic and above */ - if (gfx_v9_0_load_mec2_fw_bin_support(adev)) { - 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: gfx_v9_0_check_if_need_gfxoff(adev); gfx_v9_0_check_fw_write_wait(adev); - if (err) { - dev_err(adev->dev, - "gfx9: Failed to load firmware \"%s\"\n", - fw_name); - release_firmware(adev->gfx.mec_fw); - adev->gfx.mec_fw = NULL; - release_firmware(adev->gfx.mec2_fw); - adev->gfx.mec2_fw = NULL; - } + +out: + if (err) + amdgpu_ucode_release(&adev->gfx.mec_fw); return err; } static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) { - const char *chip_name; + char ucode_prefix[30]; int r; DRM_DEBUG("\n"); - - switch (adev->asic_type) { - case CHIP_VEGA10: - chip_name = "vega10"; - break; - case CHIP_VEGA12: - chip_name = "vega12"; - break; - case CHIP_VEGA20: - chip_name = "vega20"; - break; - case CHIP_RAVEN: - if (adev->apu_flags & AMD_APU_IS_RAVEN2) - chip_name = "raven2"; - else if (adev->apu_flags & AMD_APU_IS_PICASSO) - chip_name = "picasso"; - else - chip_name = "raven"; - break; - case CHIP_ARCTURUS: - chip_name = "arcturus"; - break; - case CHIP_RENOIR: - if (adev->apu_flags & AMD_APU_IS_RENOIR) - chip_name = "renoir"; - else - chip_name = "green_sardine"; - break; - case CHIP_ALDEBARAN: - chip_name = "aldebaran"; - break; - default: - BUG(); - } + amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix)); /* No CPG in Arcturus */ if (adev->gfx.num_gfx_rings) { - r = gfx_v9_0_init_cp_gfx_microcode(adev, chip_name); + r = gfx_v9_0_init_cp_gfx_microcode(adev, ucode_prefix); if (r) return r; } - r = gfx_v9_0_init_rlc_microcode(adev, chip_name); + r = gfx_v9_0_init_rlc_microcode(adev, ucode_prefix); if (r) return r; - r = gfx_v9_0_init_cp_compute_microcode(adev, chip_name); + r = gfx_v9_0_init_cp_compute_microcode(adev, ucode_prefix); if (r) return r; @@ -1733,45 +1648,18 @@ static u32 gfx_v9_0_get_csb_size(struct amdgpu_device *adev) return count; } -static void gfx_v9_0_get_csb_buffer(struct amdgpu_device *adev, - volatile u32 *buffer) +static void gfx_v9_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer) { - u32 count = 0, i; - const struct cs_section_def *sect = NULL; - const struct cs_extent_def *ext = NULL; + u32 count = 0; if (adev->gfx.rlc.cs_data == NULL) return; if (buffer == NULL) return; - buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0)); - buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE); - - buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1)); - buffer[count++] = cpu_to_le32(0x80000000); - buffer[count++] = cpu_to_le32(0x80000000); - - for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) { - for (ext = sect->section; ext->extent != NULL; ++ext) { - if (sect->id == SECT_CONTEXT) { - buffer[count++] = - cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count)); - buffer[count++] = cpu_to_le32(ext->reg_index - - PACKET3_SET_CONTEXT_REG_START); - for (i = 0; i < ext->reg_count; i++) - buffer[count++] = cpu_to_le32(ext->extent[i]); - } else { - return; - } - } - } - - buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0)); - buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE); - - buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0)); - buffer[count++] = cpu_to_le32(0); + count = amdgpu_gfx_csb_preamble_start(buffer); + count = amdgpu_gfx_csb_data_parser(adev, buffer, count); + amdgpu_gfx_csb_preamble_end(buffer, count); } static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) @@ -1784,7 +1672,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) if (adev->flags & AMD_IS_APU) always_on_cu_num = 4; - else if (adev->asic_type == CHIP_VEGA12) + else if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 2, 1)) always_on_cu_num = 8; else always_on_cu_num = 12; @@ -1795,10 +1683,10 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) mask = 1; cu_bitmap = 0; counter = 0; - gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0); for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) { - if (cu_info->bitmap[i][j] & mask) { + if (cu_info->bitmap[0][i][j] & mask) { if (counter == pg_always_on_cu_num) WREG32_SOC15(GC, 0, mmRLC_PG_ALWAYS_ON_CU_MASK, cu_bitmap); if (counter < always_on_cu_num) @@ -1814,7 +1702,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) cu_info->ao_cu_bitmap[i][j] = cu_bitmap; } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); } @@ -1836,7 +1724,7 @@ static void gfx_v9_0_init_lbpw(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); /* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/ - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff); /* set mmRLC_LB_PARAMS = 0x003F_1006 */ @@ -1885,7 +1773,7 @@ static void gfx_v9_4_init_lbpw(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); /* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/ - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff); /* set mmRLC_LB_PARAMS = 0x003F_1006 */ @@ -1929,6 +1817,21 @@ static int gfx_v9_0_cp_jump_table_num(struct amdgpu_device *adev) return 4; } +static void gfx_v9_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev) +{ + struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl; + + reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0]; + reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); + reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG1); + reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG2); + reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG3); + reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_CNTL); + reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX); + reg_access_ctrl->spare_int = SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT); + adev->gfx.rlc.rlcg_reg_access_supported = true; +} + static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) { const struct cs_section_def *cs_data; @@ -1953,21 +1856,6 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) return r; } - switch (adev->asic_type) { - case CHIP_RAVEN: - gfx_v9_0_init_lbpw(adev); - break; - case CHIP_VEGA20: - gfx_v9_4_init_lbpw(adev); - break; - default: - break; - } - - /* init spm vmid with 0xf */ - if (adev->gfx.rlc.funcs->update_spm_vmid) - adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf); - return 0; } @@ -1988,14 +1876,15 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev) const struct gfx_firmware_header_v1_0 *mec_hdr; - bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); + bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); /* take ownership of the relevant compute queues */ amdgpu_gfx_compute_queue_acquire(adev); mec_hpd_size = adev->gfx.num_compute_rings * GFX9_MEC_HPD_SIZE; if (mec_hpd_size) { r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT, &adev->gfx.mec.hpd_eop_obj, &adev->gfx.mec.hpd_eop_gpu_addr, (void **)&hpd); @@ -2062,7 +1951,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd, *(out++) = RREG32_SOC15(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) +static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields) { /* type 1 wave data */ dst[(*no_fields)++] = 1; @@ -2080,9 +1969,10 @@ static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u 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); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE); } -static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, +static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t start, uint32_t size, uint32_t *dst) { @@ -2091,7 +1981,7 @@ static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, start + SQIND_WAVE_SGPRS_OFFSET, size, dst); } -static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd, +static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t thread, uint32_t start, uint32_t size, uint32_t *dst) @@ -2102,9 +1992,9 @@ static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd, } static void gfx_v9_0_select_me_pipe_q(struct amdgpu_device *adev, - u32 me, u32 pipe, u32 q, u32 vm) + u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id) { - soc15_grbm_select(adev, me, pipe, q, vm); + soc15_grbm_select(adev, me, pipe, q, vm, 0); } static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { @@ -2116,12 +2006,16 @@ static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { .select_me_pipe_q = &gfx_v9_0_select_me_pipe_q, }; -static const struct amdgpu_gfx_ras_funcs gfx_v9_0_ras_funcs = { - .ras_late_init = amdgpu_gfx_ras_late_init, - .ras_fini = amdgpu_gfx_ras_fini, - .ras_error_inject = &gfx_v9_0_ras_error_inject, - .query_ras_error_count = &gfx_v9_0_query_ras_error_count, - .reset_ras_error_count = &gfx_v9_0_reset_ras_error_count, +const struct amdgpu_ras_block_hw_ops gfx_v9_0_ras_ops = { + .ras_error_inject = &gfx_v9_0_ras_error_inject, + .query_ras_error_count = &gfx_v9_0_query_ras_error_count, + .reset_ras_error_count = &gfx_v9_0_reset_ras_error_count, +}; + +static struct amdgpu_gfx_ras gfx_v9_0_ras = { + .ras_block = { + .hw_ops = &gfx_v9_0_ras_ops, + }, }; static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) @@ -2129,10 +2023,8 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) u32 gb_addr_config; int err; - adev->gfx.funcs = &gfx_v9_0_gfx_funcs; - - switch (adev->asic_type) { - case CHIP_VEGA10: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): 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; @@ -2140,7 +2032,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; gb_addr_config = VEGA10_GB_ADDR_CONFIG_GOLDEN; break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): 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; @@ -2149,8 +2041,8 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config = VEGA12_GB_ADDR_CONFIG_GOLDEN; DRM_INFO("fix gfx.config for vega12\n"); break; - case CHIP_VEGA20: - adev->gfx.ras_funcs = &gfx_v9_0_ras_funcs; + case IP_VERSION(9, 4, 0): + adev->gfx.ras = &gfx_v9_0_ras; 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; @@ -2164,7 +2056,8 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) if (err) return err; break; - case CHIP_RAVEN: + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): 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; @@ -2175,8 +2068,8 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) else gb_addr_config = RAVEN_GB_ADDR_CONFIG_GOLDEN; break; - case CHIP_ARCTURUS: - adev->gfx.ras_funcs = &gfx_v9_4_ras_funcs; + case IP_VERSION(9, 4, 1): + adev->gfx.ras = &gfx_v9_4_ras; 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; @@ -2186,7 +2079,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config &= ~0xf3e777ff; gb_addr_config |= 0x22014042; break; - case CHIP_RENOIR: + case IP_VERSION(9, 3, 0): 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; @@ -2196,8 +2089,8 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config &= ~0xf3e777ff; gb_addr_config |= 0x22010042; break; - case CHIP_ALDEBARAN: - adev->gfx.ras_funcs = &gfx_v9_4_2_ras_funcs; + case IP_VERSION(9, 4, 2): + adev->gfx.ras = &gfx_v9_4_2_ras; 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; @@ -2275,33 +2168,64 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1; ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr + (ring_id * GFX9_MEC_HPD_SIZE); + ring->vm_hub = AMDGPU_GFXHUB(0); sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue); irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP + ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec) + ring->pipe; hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ? - AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL; + AMDGPU_RING_PRIO_2 : AMDGPU_RING_PRIO_DEFAULT; /* type-2 packets are deprecated on MEC, use type-3 instead */ return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, hw_prio, NULL); } -static int gfx_v9_0_sw_init(void *handle) +static void gfx_v9_0_alloc_ip_dump(struct amdgpu_device *adev) +{ + uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9); + uint32_t *ptr; + uint32_t inst; + + ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL); + if (!ptr) { + DRM_ERROR("Failed to allocate memory for GFX IP Dump\n"); + adev->gfx.ip_dump_core = NULL; + } else { + adev->gfx.ip_dump_core = ptr; + } + + /* Allocate memory for compute queue registers for all the instances */ + reg_count = ARRAY_SIZE(gc_cp_reg_list_9); + inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec * + adev->gfx.mec.num_queue_per_pipe; + + ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL); + if (!ptr) { + DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n"); + adev->gfx.ip_dump_compute_queues = NULL; + } else { + adev->gfx.ip_dump_compute_queues = ptr; + } +} + +static int gfx_v9_0_sw_init(struct amdgpu_ip_block *ip_block) { int i, j, k, r, ring_id; + int xcc_id = 0; struct amdgpu_ring *ring; - struct amdgpu_kiq *kiq; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_ARCTURUS: - case CHIP_RENOIR: - case CHIP_ALDEBARAN: + struct amdgpu_device *adev = ip_block->adev; + unsigned int hw_prio; + + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 4, 1): + case IP_VERSION(9, 3, 0): + case IP_VERSION(9, 4, 2): adev->gfx.mec.num_mec = 2; break; default: @@ -2309,6 +2233,43 @@ static int gfx_v9_0_sw_init(void *handle) break; } + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 3, 0): + adev->gfx.cleaner_shader_ptr = gfx_9_4_2_cleaner_shader_hex; + adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_2_cleaner_shader_hex); + if (adev->gfx.me_fw_version >= 167 && + adev->gfx.pfp_fw_version >= 196 && + adev->gfx.mec_fw_version >= 474) { + adev->gfx.enable_cleaner_shader = true; + r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size); + if (r) { + adev->gfx.enable_cleaner_shader = false; + dev_err(adev->dev, "Failed to initialize cleaner shader\n"); + } + } + break; + case IP_VERSION(9, 4, 2): + adev->gfx.cleaner_shader_ptr = gfx_9_4_2_cleaner_shader_hex; + adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_2_cleaner_shader_hex); + if (adev->gfx.mec_fw_version >= 88) { + adev->gfx.enable_cleaner_shader = true; + r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size); + if (r) { + adev->gfx.enable_cleaner_shader = false; + dev_err(adev->dev, "Failed to initialize cleaner shader\n"); + } + } + break; + default: + adev->gfx.enable_cleaner_shader = false; + break; + } + adev->gfx.mec.num_pipe_per_mec = 4; adev->gfx.mec.num_queue_per_pipe = 8; @@ -2317,6 +2278,13 @@ static int gfx_v9_0_sw_init(void *handle) if (r) return r; + /* Bad opcode Event */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, + GFX_9_0__SRCID__CP_BAD_OPCODE_ERROR, + &adev->gfx.bad_op_irq); + if (r) + return r; + /* Privileged reg */ r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_REG_FAULT, &adev->gfx.priv_reg_irq); @@ -2343,18 +2311,14 @@ static int gfx_v9_0_sw_init(void *handle) adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE; - gfx_v9_0_scratch_init(adev); - - r = gfx_v9_0_init_microcode(adev); - if (r) { - DRM_ERROR("Failed to load gfx firmware!\n"); - return r; - } - - r = adev->gfx.rlc.funcs->init(adev); - if (r) { - DRM_ERROR("Failed to init rlc BOs!\n"); - return r; + if (adev->gfx.rlc.funcs) { + if (adev->gfx.rlc.funcs->init) { + r = adev->gfx.rlc.funcs->init(adev); + if (r) { + dev_err(adev->dev, "Failed to init rlc BOs!\n"); + return r; + } + } } r = gfx_v9_0_mec_init(adev); @@ -2373,6 +2337,10 @@ static int gfx_v9_0_sw_init(void *handle) sprintf(ring->name, "gfx_%d", i); ring->use_doorbell = true; ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; + + /* disable scheduler on the real ring */ + ring->no_scheduler = adev->gfx.mcbp; + ring->vm_hub = AMDGPU_GFXHUB(0); r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, AMDGPU_RING_PRIO_DEFAULT, NULL); @@ -2380,12 +2348,49 @@ static int gfx_v9_0_sw_init(void *handle) return r; } + /* set up the software rings */ + if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) { + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) { + ring = &adev->gfx.sw_gfx_ring[i]; + ring->ring_obj = NULL; + sprintf(ring->name, amdgpu_sw_ring_name(i)); + ring->use_doorbell = true; + ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; + ring->is_sw_ring = true; + hw_prio = amdgpu_sw_ring_priority(i); + ring->vm_hub = AMDGPU_GFXHUB(0); + r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, + AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, hw_prio, + NULL); + if (r) + return r; + ring->wptr = 0; + } + + /* init the muxer and add software rings */ + r = amdgpu_ring_mux_init(&adev->gfx.muxer, &adev->gfx.gfx_ring[0], + GFX9_NUM_SW_GFX_RINGS); + if (r) { + DRM_ERROR("amdgpu_ring_mux_init failed(%d)\n", r); + return r; + } + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) { + r = amdgpu_ring_mux_add_sw_ring(&adev->gfx.muxer, + &adev->gfx.sw_gfx_ring[i]); + if (r) { + DRM_ERROR("amdgpu_ring_mux_add_sw_ring failed(%d)\n", r); + return r; + } + } + } + /* set up the compute queues - allocate horizontally across pipes */ ring_id = 0; for (i = 0; i < adev->gfx.mec.num_mec; ++i) { for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) { for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) { - if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j)) + if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i, + k, j)) continue; r = gfx_v9_0_compute_ring_init(adev, @@ -2399,19 +2404,26 @@ static int gfx_v9_0_sw_init(void *handle) } } - r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE); + /* TODO: Add queue reset mask when FW fully supports it */ + adev->gfx.gfx_supported_reset = + amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]); + adev->gfx.compute_supported_reset = + amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]); + if (!amdgpu_sriov_vf(adev) && !adev->debug_disable_gpu_ring_reset) + adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE; + + r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE, 0); if (r) { DRM_ERROR("Failed to init KIQ BOs!\n"); return r; } - kiq = &adev->gfx.kiq; - r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq); + r = amdgpu_gfx_kiq_init_ring(adev, xcc_id); if (r) return r; /* create MQD for all compute queues as wel as KIQ for SRIOV case */ - r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation)); + r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation), 0); if (r) return r; @@ -2421,30 +2433,47 @@ static int gfx_v9_0_sw_init(void *handle) if (r) return r; + if (amdgpu_gfx_ras_sw_init(adev)) { + dev_err(adev->dev, "Failed to initialize gfx ras block!\n"); + return -EINVAL; + } + + gfx_v9_0_alloc_ip_dump(adev); + + r = amdgpu_gfx_sysfs_init(adev); + if (r) + return r; + return 0; } -static int gfx_v9_0_sw_fini(void *handle) +static int gfx_v9_0_sw_fini(struct amdgpu_ip_block *ip_block) { int i; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->ras_fini) - adev->gfx.ras_funcs->ras_fini(adev); + if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) { + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) + amdgpu_ring_fini(&adev->gfx.sw_gfx_ring[i]); + amdgpu_ring_mux_fini(&adev->gfx.muxer); + } for (i = 0; i < adev->gfx.num_gfx_rings; i++) amdgpu_ring_fini(&adev->gfx.gfx_ring[i]); for (i = 0; i < adev->gfx.num_compute_rings; i++) amdgpu_ring_fini(&adev->gfx.compute_ring[i]); - amdgpu_gfx_mqd_sw_fini(adev); - amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring); - amdgpu_gfx_kiq_fini(adev); + amdgpu_gfx_mqd_sw_fini(adev, 0); + amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring); + amdgpu_gfx_kiq_fini(adev, 0); + + amdgpu_gfx_cleaner_shader_sw_fini(adev); gfx_v9_0_mec_fini(adev); - amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj); + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); if (adev->flags & AMD_IS_APU) { amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, &adev->gfx.rlc.cp_table_gpu_addr, @@ -2452,6 +2481,11 @@ static int gfx_v9_0_sw_fini(void *handle) } gfx_v9_0_free_microcode(adev); + amdgpu_gfx_sysfs_fini(adev); + + kfree(adev->gfx.ip_dump_core); + kfree(adev->gfx.ip_dump_compute_queues); + return 0; } @@ -2462,7 +2496,7 @@ static void gfx_v9_0_tiling_mode_table_init(struct amdgpu_device *adev) } void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num, - u32 instance) + u32 instance, int xcc_id) { u32 data; @@ -2511,19 +2545,42 @@ static void gfx_v9_0_setup_rb(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0); data = gfx_v9_0_get_rb_active_bitmap(adev); active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) * rb_bitmap_width_per_sh); } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); adev->gfx.config.backend_enable_mask = active_rbs; adev->gfx.config.num_rbs = hweight32(active_rbs); } +static void gfx_v9_0_debug_trap_config_init(struct amdgpu_device *adev, + uint32_t first_vmid, + uint32_t last_vmid) +{ + uint32_t data; + uint32_t trap_config_vmid_mask = 0; + int i; + + /* Calculate trap config vmid mask */ + for (i = first_vmid; i < last_vmid; i++) + trap_config_vmid_mask |= (1 << i); + + data = REG_SET_FIELD(0, SPI_GDBG_TRAP_CONFIG, + VMID_SEL, trap_config_vmid_mask); + data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG, + TRAP_EN, 1); + WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data); + WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0); +} + #define DEFAULT_SH_MEM_BASES (0x6000) static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev) { @@ -2545,16 +2602,16 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev) mutex_lock(&adev->srbm_mutex); for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { - soc15_grbm_select(adev, 0, 0, 0, i); + soc15_grbm_select(adev, 0, 0, 0, i, 0); /* CP and shaders */ WREG32_SOC15_RLC(GC, 0, mmSH_MEM_CONFIG, sh_mem_config); WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, sh_mem_bases); } - soc15_grbm_select(adev, 0, 0, 0, 0); + soc15_grbm_select(adev, 0, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); /* Initialize all compute VMIDs to have no GDS, GWS, or OA - acccess. These should be enabled by FW for target VMIDs. */ + access. These should be enabled by FW for target VMIDs. */ for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_BASE, 2 * i, 0); WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_SIZE, 2 * i, 0); @@ -2585,13 +2642,16 @@ static void gfx_v9_0_init_sq_config(struct amdgpu_device *adev) { uint32_t tmp; - switch (adev->asic_type) { - case CHIP_ARCTURUS: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 4, 1): tmp = RREG32_SOC15(GC, 0, mmSQ_CONFIG); - tmp = REG_SET_FIELD(tmp, SQ_CONFIG, - DISABLE_BARRIER_WAITCNT, 1); + tmp = REG_SET_FIELD(tmp, SQ_CONFIG, DISABLE_BARRIER_WAITCNT, + !READ_ONCE(adev->barrier_has_auto_waitcnt)); WREG32_SOC15(GC, 0, mmSQ_CONFIG, tmp); break; + case IP_VERSION(9, 4, 2): + gfx_v9_4_2_init_sq(adev); + break; default: break; } @@ -2602,19 +2662,23 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev) u32 tmp; int i; - WREG32_FIELD15_RLC(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff); + if (!amdgpu_sriov_vf(adev) || + amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)) { + WREG32_FIELD15_RLC(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff); + } gfx_v9_0_tiling_mode_table_init(adev); - gfx_v9_0_setup_rb(adev); + if (adev->gfx.num_gfx_rings) + gfx_v9_0_setup_rb(adev); gfx_v9_0_get_cu_info(adev, &adev->gfx.cu_info); adev->gfx.config.db_debug2 = RREG32_SOC15(GC, 0, mmDB_DEBUG2); /* XXX SH_MEM regs */ /* where to put LDS, scratch, GPUVM in FSA64 space */ mutex_lock(&adev->srbm_mutex); - for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) { - soc15_grbm_select(adev, 0, 0, 0, i); + for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) { + soc15_grbm_select(adev, 0, 0, 0, i, 0); /* CP and shaders */ if (i == 0) { tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE, @@ -2636,7 +2700,7 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev) WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, tmp); } } - soc15_grbm_select(adev, 0, 0, 0, 0); + soc15_grbm_select(adev, 0, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); @@ -2653,15 +2717,15 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0); for (k = 0; k < adev->usec_timeout; k++) { if (RREG32_SOC15(GC, 0, mmRLC_SERDES_CU_MASTER_BUSY) == 0) break; udelay(1); } if (k == adev->usec_timeout) { - gfx_v9_0_select_se_sh(adev, 0xffffffff, - 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, + 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); DRM_INFO("Timeout wait for RLC serdes %u,%u\n", i, j); @@ -2669,7 +2733,7 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev) } } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK | @@ -2695,7 +2759,7 @@ static void gfx_v9_0_enable_gui_idle_interrupt(struct amdgpu_device *adev, tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE, enable ? 1 : 0); tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE, enable ? 1 : 0); tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, enable ? 1 : 0); - if(adev->gfx.num_gfx_rings) + if (adev->gfx.num_gfx_rings) tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE, enable ? 1 : 0); WREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0, tmp); @@ -2921,7 +2985,7 @@ static void gfx_v9_0_init_gfx_power_gating(struct amdgpu_device *adev) /* program GRBM_REG_SAVE_GFX_IDLE_THRESHOLD to 0x55f0 */ data |= (0x55f0 << RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD__SHIFT); WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_AUTO_PG_CTRL), data); - if (adev->asic_type != CHIP_RENOIR) + if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 3, 0)) pwr_10_0_gfxip_control_over_cgpg(adev, true); } } @@ -3033,7 +3097,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) * And it's needed by gfxoff feature. */ if (adev->gfx.rlc.is_rlc_v2_1) { - if (adev->asic_type == CHIP_VEGA12 || + if (amdgpu_ip_version(adev, GC_HWIP, 0) == + IP_VERSION(9, 2, 1) || (adev->apu_flags & AMD_APU_IS_RAVEN2)) gfx_v9_1_init_rlc_save_restore_list(adev); gfx_v9_0_enable_save_restore_machine(adev); @@ -3045,8 +3110,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) AMD_PG_SUPPORT_CP | AMD_PG_SUPPORT_GDS | AMD_PG_SUPPORT_RLC_SMU_HS)) { - WREG32(mmRLC_JUMP_TABLE_RESTORE, - adev->gfx.rlc.cp_table_gpu_addr >> 8); + WREG32_SOC15(GC, 0, mmRLC_JUMP_TABLE_RESTORE, + adev->gfx.rlc.cp_table_gpu_addr >> 8); gfx_v9_0_init_gfx_power_gating(adev); } } @@ -3146,14 +3211,17 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) return r; } - switch (adev->asic_type) { - case CHIP_RAVEN: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + gfx_v9_0_init_lbpw(adev); if (amdgpu_lbpw == 0) gfx_v9_0_enable_lbpw(adev, false); else gfx_v9_0_enable_lbpw(adev, true); break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): + gfx_v9_4_init_lbpw(adev); if (amdgpu_lbpw > 0) gfx_v9_0_enable_lbpw(adev, true); else @@ -3163,6 +3231,8 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) break; } + gfx_v9_0_update_spm_vmid_internal(adev, 0xf); + adev->gfx.rlc.funcs->start(adev); return 0; @@ -3172,6 +3242,15 @@ static void gfx_v9_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) { u32 tmp = RREG32_SOC15(GC, 0, mmCP_ME_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_INVALIDATE_ICACHE, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_INVALIDATE_ICACHE, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_INVALIDATE_ICACHE, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_PIPE0_RESET, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_PIPE1_RESET, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, enable ? 0 : 1); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1); @@ -3249,6 +3328,14 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev) gfx_v9_0_cp_gfx_enable(adev, true); + /* Now only limit the quirk on the APU gfx9 series and already + * confirmed that the APU gfx10/gfx11 needn't such update. + */ + if (adev->flags & AMD_IS_APU && + adev->in_s3 && !pm_resume_via_firmware()) { + DRM_INFO("Will skip the CSB packet resubmit\n"); + return 0; + } r = amdgpu_ring_alloc(ring, gfx_v9_0_get_csb_size(adev) + 4 + 3); if (r) { DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r); @@ -3326,12 +3413,12 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr)); - /* set the wb address wether it's enabled or not */ - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + /* set the wb address whether it's enabled or not */ + rptr_addr = ring->rptr_gpu_addr; WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK); - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr)); WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI, upper_32_bits(wptr_gpu_addr)); @@ -3363,7 +3450,6 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev) /* start the ring */ gfx_v9_0_cp_gfx_start(adev); - ring->sched.ready = true; return 0; } @@ -3374,8 +3460,16 @@ static void gfx_v9_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL, 0); } else { WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL, - (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK)); - adev->gfx.kiq.ring.sched.ready = false; + (CP_MEC_CNTL__MEC_INVALIDATE_ICACHE_MASK | + CP_MEC_CNTL__MEC_ME1_PIPE0_RESET_MASK | + CP_MEC_CNTL__MEC_ME1_PIPE1_RESET_MASK | + CP_MEC_CNTL__MEC_ME1_PIPE2_RESET_MASK | + CP_MEC_CNTL__MEC_ME1_PIPE3_RESET_MASK | + CP_MEC_CNTL__MEC_ME2_PIPE0_RESET_MASK | + CP_MEC_CNTL__MEC_ME2_PIPE1_RESET_MASK | + CP_MEC_CNTL__MEC_ME1_HALT_MASK | + CP_MEC_CNTL__MEC_ME2_HALT_MASK)); + adev->gfx.kiq[0].ring.sched.ready = false; } udelay(50); } @@ -3432,9 +3526,7 @@ static void gfx_v9_0_kiq_setting(struct amdgpu_ring *ring) tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS); tmp &= 0xffffff00; tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue); - WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp); - tmp |= 0x80; - WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp); + WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp | 0x80); } static void gfx_v9_0_mqd_set_priority(struct amdgpu_ring *ring, struct v9_mqd *mqd) @@ -3532,7 +3624,7 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, (order_base_2(ring->ring_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, - ((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8)); + (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1)); #ifdef __BIG_ENDIAN tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1); #endif @@ -3543,33 +3635,16 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) mqd->cp_hqd_pq_control = tmp; /* set the wb address whether it's enabled or not */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + wb_gpu_addr = ring->rptr_gpu_addr; mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wb_gpu_addr = ring->wptr_gpu_addr; mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; - tmp = 0; - /* enable the doorbell if requested */ - if (ring->use_doorbell) { - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_OFFSET, ring->doorbell_index); - - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_EN, 1); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_SOURCE, 0); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_HIT, 0); - } - - mqd->cp_hqd_pq_doorbell_control = tmp; - /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ ring->wptr = 0; mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR); @@ -3588,7 +3663,7 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) /* set static priority for a queue/ring */ gfx_v9_0_mqd_set_priority(ring, mqd); - mqd->cp_hqd_quantum = RREG32(mmCP_HQD_QUANTUM); + mqd->cp_hqd_quantum = RREG32_SOC15(GC, 0, mmCP_HQD_QUANTUM); /* map_queues packet doesn't need activate the queue, * so only kiq need set this field. @@ -3675,7 +3750,16 @@ static int gfx_v9_0_kiq_init_register(struct amdgpu_ring *ring) if (ring->use_doorbell) { WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_LOWER, (adev->doorbell_index.kiq * 2) << 2); - WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER, + /* If GC has entered CGPG, ringing doorbell > first page + * doesn't wakeup GC. Enlarge CP_MEC_DOORBELL_RANGE_UPPER to + * workaround this issue. And this change has to align with firmware + * update. + */ + if (check_if_enlarge_doorbell_range(adev)) + WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER, + (adev->doorbell.size - 4)); + else + WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER, (adev->doorbell_index.userqueue_end * 2) << 2); } @@ -3747,7 +3831,6 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; struct v9_mqd *mqd = ring->mqd_ptr; - int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS; struct v9_mqd *tmp_mqd; gfx_v9_0_kiq_setting(ring); @@ -3757,40 +3840,42 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) * driver need to re-init the mqd. * check mqd->cp_hqd_pq_control since this value should not be 0 */ - tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx]; + tmp_mqd = (struct v9_mqd *)adev->gfx.kiq[0].mqd_backup; if (amdgpu_in_reset(adev) && tmp_mqd->cp_hqd_pq_control){ /* for GPU_RESET case , reset MQD to a clean status */ - if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); + if (adev->gfx.kiq[0].mqd_backup) + memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct v9_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; amdgpu_ring_clear_ring(ring); mutex_lock(&adev->srbm_mutex); - soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0); gfx_v9_0_kiq_init_register(ring); - soc15_grbm_select(adev, 0, 0, 0, 0); + soc15_grbm_select(adev, 0, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } else { memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; + if (amdgpu_sriov_vf(adev) && adev->in_suspend) + amdgpu_ring_clear_ring(ring); mutex_lock(&adev->srbm_mutex); - soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0); gfx_v9_0_mqd_init(ring); gfx_v9_0_kiq_init_register(ring); - soc15_grbm_select(adev, 0, 0, 0, 0); + soc15_grbm_select(adev, 0, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); - if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); + if (adev->gfx.kiq[0].mqd_backup) + memcpy(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(struct v9_mqd_allocation)); } return 0; } -static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) +static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring, bool restore) { struct amdgpu_device *adev = ring->adev; struct v9_mqd *mqd = ring->mqd_ptr; @@ -3802,29 +3887,26 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) */ tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx]; - if (!tmp_mqd->cp_hqd_pq_control || - (!amdgpu_in_reset(adev) && !adev->in_suspend)) { + if (!restore && (!tmp_mqd->cp_hqd_pq_control || + (!amdgpu_in_reset(adev) && !adev->in_suspend))) { memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); - soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0); gfx_v9_0_mqd_init(ring); - soc15_grbm_select(adev, 0, 0, 0, 0); + soc15_grbm_select(adev, 0, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); - } else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */ - /* reset MQD to a clean status */ + } else { + /* restore MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); - /* reset ring buffer */ ring->wptr = 0; - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], 0); - amdgpu_ring_clear_ring(ring); - } else { + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0); amdgpu_ring_clear_ring(ring); } @@ -3833,54 +3915,23 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) static int gfx_v9_0_kiq_resume(struct amdgpu_device *adev) { - struct amdgpu_ring *ring; - int r; - - ring = &adev->gfx.kiq.ring; - - r = amdgpu_bo_reserve(ring->mqd_obj, false); - if (unlikely(r != 0)) - return r; - - r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr); - if (unlikely(r != 0)) - return r; - - gfx_v9_0_kiq_init_queue(ring); - amdgpu_bo_kunmap(ring->mqd_obj); - ring->mqd_ptr = NULL; - amdgpu_bo_unreserve(ring->mqd_obj); - ring->sched.ready = true; + gfx_v9_0_kiq_init_queue(&adev->gfx.kiq[0].ring); return 0; } static int gfx_v9_0_kcq_resume(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = NULL; - int r = 0, i; + int i, r; gfx_v9_0_cp_compute_enable(adev, true); for (i = 0; i < adev->gfx.num_compute_rings; i++) { - ring = &adev->gfx.compute_ring[i]; - - r = amdgpu_bo_reserve(ring->mqd_obj, false); - if (unlikely(r != 0)) - goto done; - r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr); - if (!r) { - r = gfx_v9_0_kcq_init_queue(ring); - amdgpu_bo_kunmap(ring->mqd_obj); - ring->mqd_ptr = NULL; - } - amdgpu_bo_unreserve(ring->mqd_obj); + r = gfx_v9_0_kcq_init_queue(&adev->gfx.compute_ring[i], false); if (r) - goto done; + return r; } - r = amdgpu_gfx_enable_kcq(adev); -done: - return r; + return amdgpu_gfx_enable_kcq(adev, 0); } static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) @@ -3904,6 +3955,10 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) return r; } + if (adev->gfx.num_gfx_rings) + gfx_v9_0_cp_gfx_enable(adev, false); + gfx_v9_0_cp_compute_enable(adev, false); + r = gfx_v9_0_kiq_resume(adev); if (r) return r; @@ -3939,8 +3994,8 @@ static void gfx_v9_0_init_tcp_config(struct amdgpu_device *adev) { u32 tmp; - if (adev->asic_type != CHIP_ARCTURUS && - adev->asic_type != CHIP_ALDEBARAN) + if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1) && + amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)) return; tmp = RREG32_SOC15(GC, 0, mmTCP_ADDR_CONFIG); @@ -3960,10 +4015,13 @@ static void gfx_v9_0_cp_enable(struct amdgpu_device *adev, bool enable) gfx_v9_0_cp_compute_enable(adev, enable); } -static int gfx_v9_0_hw_init(void *handle) +static int gfx_v9_0_hw_init(struct amdgpu_ip_block *ip_block) { int r; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; + + amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size, + adev->gfx.cleaner_shader_ptr); if (!amdgpu_sriov_vf(adev)) gfx_v9_0_init_golden_registers(adev); @@ -3980,24 +4038,27 @@ static int gfx_v9_0_hw_init(void *handle) if (r) return r; - if (adev->asic_type == CHIP_ALDEBARAN) + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) && + !amdgpu_sriov_vf(adev)) gfx_v9_4_2_set_power_brake_sequence(adev); return r; } -static int gfx_v9_0_hw_fini(void *handle) +static int gfx_v9_0_hw_fini(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; - amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0); + if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) + amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0); /* DF freeze and kcq disable will fail */ if (!amdgpu_ras_intr_triggered()) /* disable KCQ to avoid CPC touch memory not valid anymore */ - amdgpu_gfx_disable_kcq(adev); + amdgpu_gfx_disable_kcq(adev, 0); if (amdgpu_sriov_vf(adev)) { gfx_v9_0_cp_gfx_enable(adev, false); @@ -4015,19 +4076,20 @@ static int gfx_v9_0_hw_fini(void *handle) */ if (!amdgpu_in_reset(adev) && !adev->in_suspend) { mutex_lock(&adev->srbm_mutex); - soc15_grbm_select(adev, adev->gfx.kiq.ring.me, - adev->gfx.kiq.ring.pipe, - adev->gfx.kiq.ring.queue, 0); - gfx_v9_0_kiq_fini_register(&adev->gfx.kiq.ring); - soc15_grbm_select(adev, 0, 0, 0, 0); + soc15_grbm_select(adev, adev->gfx.kiq[0].ring.me, + adev->gfx.kiq[0].ring.pipe, + adev->gfx.kiq[0].ring.queue, 0, 0); + gfx_v9_0_kiq_fini_register(&adev->gfx.kiq[0].ring); + soc15_grbm_select(adev, 0, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } gfx_v9_0_cp_enable(adev, false); - /* Skip suspend with A+A reset */ - if (adev->gmc.xgmi.connected_to_cpu && amdgpu_in_reset(adev)) { - dev_dbg(adev->dev, "Device in reset. Skipping RLC halt\n"); + /* Skip stopping RLC with A+A reset or when RLC controls GFX clock */ + if ((adev->gmc.xgmi.connected_to_cpu && amdgpu_in_reset(adev)) || + (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(9, 4, 2))) { + dev_dbg(adev->dev, "Skipping RLC halt\n"); return 0; } @@ -4035,19 +4097,19 @@ static int gfx_v9_0_hw_fini(void *handle) return 0; } -static int gfx_v9_0_suspend(void *handle) +static int gfx_v9_0_suspend(struct amdgpu_ip_block *ip_block) { - return gfx_v9_0_hw_fini(handle); + return gfx_v9_0_hw_fini(ip_block); } -static int gfx_v9_0_resume(void *handle) +static int gfx_v9_0_resume(struct amdgpu_ip_block *ip_block) { - return gfx_v9_0_hw_init(handle); + return gfx_v9_0_hw_init(ip_block); } -static bool gfx_v9_0_is_idle(void *handle) +static bool gfx_v9_0_is_idle(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; if (REG_GET_FIELD(RREG32_SOC15(GC, 0, mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE)) @@ -4056,24 +4118,24 @@ static bool gfx_v9_0_is_idle(void *handle) return true; } -static int gfx_v9_0_wait_for_idle(void *handle) +static int gfx_v9_0_wait_for_idle(struct amdgpu_ip_block *ip_block) { unsigned i; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; for (i = 0; i < adev->usec_timeout; i++) { - if (gfx_v9_0_is_idle(handle)) + if (gfx_v9_0_is_idle(ip_block)) return 0; udelay(1); } return -ETIMEDOUT; } -static int gfx_v9_0_soft_reset(void *handle) +static int gfx_v9_0_soft_reset(struct amdgpu_ip_block *ip_block) { u32 grbm_soft_reset = 0; u32 tmp; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; /* GRBM_STATUS */ tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS); @@ -4112,19 +4174,17 @@ static int gfx_v9_0_soft_reset(void *handle) /* Disable MEC parsing/prefetching */ gfx_v9_0_cp_compute_enable(adev, false); - if (grbm_soft_reset) { - tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET); - tmp |= grbm_soft_reset; - dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp); - WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp); - tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET); + tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET); + tmp |= grbm_soft_reset; + dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp); + tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET); - udelay(50); + udelay(50); - tmp &= ~grbm_soft_reset; - WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp); - tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET); - } + tmp &= ~grbm_soft_reset; + WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp); + tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET); /* Wait a little for things to settle down */ udelay(50); @@ -4138,7 +4198,7 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) unsigned long flags; uint32_t seq, reg_val_offs = 0; uint64_t value = 0; - struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_kiq *kiq = &adev->gfx.kiq[0]; struct amdgpu_ring *ring = &kiq->ring; BUG_ON(!ring->funcs->emit_rreg); @@ -4208,19 +4268,40 @@ failed_kiq_read: static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev) { - uint64_t clock; + uint64_t clock, clock_lo, clock_hi, hi_check; - amdgpu_gfx_off_ctrl(adev, false); - mutex_lock(&adev->gfx.gpu_clock_mutex); - if (adev->asic_type == CHIP_VEGA10 && amdgpu_sriov_runtime(adev)) { - clock = gfx_v9_0_kiq_read_clock(adev); - } else { - WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1); - clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) | - ((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL); + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 3, 0): + preempt_disable(); + clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Renoir); + clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Renoir); + hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Renoir); + /* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over + * roughly every 42 seconds. + */ + if (hi_check != clock_hi) { + clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Renoir); + clock_hi = hi_check; + } + preempt_enable(); + clock = clock_lo | (clock_hi << 32ULL); + break; + default: + amdgpu_gfx_off_ctrl(adev, false); + mutex_lock(&adev->gfx.gpu_clock_mutex); + if (amdgpu_ip_version(adev, GC_HWIP, 0) == + IP_VERSION(9, 0, 1) && + amdgpu_sriov_runtime(adev)) { + clock = gfx_v9_0_kiq_read_clock(adev); + } else { + WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1); + clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) | + ((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL); + } + mutex_unlock(&adev->gfx.gpu_clock_mutex); + amdgpu_gfx_off_ctrl(adev, true); + break; } - mutex_unlock(&adev->gfx.gpu_clock_mutex); - amdgpu_gfx_off_ctrl(adev, true); return clock; } @@ -4562,7 +4643,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) if (!ring->sched.ready) return 0; - if (adev->asic_type == CHIP_ARCTURUS) { + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1)) { vgpr_init_shader_ptr = vgpr_init_compute_shader_arcturus; vgpr_init_shader_size = sizeof(vgpr_init_compute_shader_arcturus); vgpr_init_regs_ptr = vgpr_init_regs_arcturus; @@ -4702,21 +4783,24 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) } fail: - amdgpu_ib_free(adev, &ib, NULL); + amdgpu_ib_free(&ib, NULL); dma_fence_put(f); return r; } -static int gfx_v9_0_early_init(void *handle) +static int gfx_v9_0_early_init(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; + + adev->gfx.funcs = &gfx_v9_0_gfx_funcs; - if (adev->asic_type == CHIP_ARCTURUS || - adev->asic_type == CHIP_ALDEBARAN) + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) || + amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2)) adev->gfx.num_gfx_rings = 0; else adev->gfx.num_gfx_rings = GFX9_NUM_GFX_RINGS; + adev->gfx.xcc_mask = 1; adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev), AMDGPU_MAX_COMPUTE_RINGS); gfx_v9_0_set_kiq_pm4_funcs(adev); @@ -4725,12 +4809,15 @@ static int gfx_v9_0_early_init(void *handle) gfx_v9_0_set_gds_init(adev); gfx_v9_0_set_rlc_funcs(adev); - return 0; + /* init rlcg reg access ctrl */ + gfx_v9_0_init_rlcg_reg_access_ctrl(adev); + + return gfx_v9_0_init_microcode(adev); } -static int gfx_v9_0_ecc_late_init(void *handle) +static int gfx_v9_0_ecc_late_init(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; int r; /* @@ -4747,7 +4834,7 @@ static int gfx_v9_0_ecc_late_init(void *handle) } /* requires IBs so do in late init after IB pool is initialized */ - if (adev->asic_type == CHIP_ALDEBARAN) + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2)) r = gfx_v9_4_2_do_edc_gpr_workarounds(adev); else r = gfx_v9_0_do_edc_gpr_workarounds(adev); @@ -4755,23 +4842,16 @@ static int gfx_v9_0_ecc_late_init(void *handle) if (r) return r; - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->ras_late_init) { - r = adev->gfx.ras_funcs->ras_late_init(adev); - if (r) - return r; - } - - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->enable_watchdog_timer) - adev->gfx.ras_funcs->enable_watchdog_timer(adev); + if (adev->gfx.ras && + adev->gfx.ras->enable_watchdog_timer) + adev->gfx.ras->enable_watchdog_timer(adev); return 0; } -static int gfx_v9_0_late_init(void *handle) +static int gfx_v9_0_late_init(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; int r; r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0); @@ -4782,10 +4862,21 @@ static int gfx_v9_0_late_init(void *handle) if (r) return r; - r = gfx_v9_0_ecc_late_init(handle); + r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0); if (r) return r; + r = gfx_v9_0_ecc_late_init(ip_block); + if (r) + return r; + + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2)) + gfx_v9_4_2_debug_trap_config_init(adev, + adev->vm_manager.first_kfd_vmid, AMDGPU_NUM_VMID); + else + gfx_v9_0_debug_trap_config_init(adev, + adev->vm_manager.first_kfd_vmid, AMDGPU_NUM_VMID); + return 0; } @@ -4801,7 +4892,7 @@ static bool gfx_v9_0_is_rlc_enabled(struct amdgpu_device *adev) return true; } -static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev) +static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id) { uint32_t data; unsigned i; @@ -4818,7 +4909,7 @@ static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev) } } -static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev) +static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id) { uint32_t data; @@ -4829,7 +4920,7 @@ static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev) static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev, bool enable) { - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) && enable) { gfx_v9_0_enable_gfx_cg_power_gating(adev, true); @@ -4841,7 +4932,7 @@ static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev, gfx_v9_0_enable_gfx_pipeline_powergating(adev, false); } - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } static void gfx_v9_0_update_gfx_mg_power_gating(struct amdgpu_device *adev, @@ -4868,14 +4959,12 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev { uint32_t data, def; - amdgpu_gfx_rlc_enter_safe_mode(adev); - /* It is disabled by HW by default */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { /* 1 - RLC_CGTT_MGCG_OVERRIDE */ def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); - if (adev->asic_type != CHIP_VEGA12) + if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 2, 1)) data &= ~RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK; data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK | @@ -4909,7 +4998,7 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev /* 1 - MGCG_OVERRIDE */ def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); - if (adev->asic_type != CHIP_VEGA12) + if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 2, 1)) data |= RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK; data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK | @@ -4934,8 +5023,6 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev WREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL, data); } } - - amdgpu_gfx_rlc_exit_safe_mode(adev); } static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, @@ -4946,8 +5033,6 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, if (!adev->gfx.num_gfx_rings) return; - amdgpu_gfx_rlc_enter_safe_mode(adev); - /* Enable 3D CGCG/CGLS */ if (enable) { /* write cmd to clear cgcg/cgls ov */ @@ -4989,8 +5074,6 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, if (def != data) WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D, data); } - - amdgpu_gfx_rlc_exit_safe_mode(adev); } static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev, @@ -4998,8 +5081,6 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev { uint32_t def, data; - amdgpu_gfx_rlc_enter_safe_mode(adev); - if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) { def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); /* unset CGCG override */ @@ -5015,7 +5096,7 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev /* enable cgcg FSM(0x0000363F) */ def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL); - if (adev->asic_type == CHIP_ARCTURUS) + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1)) data = (0x2000 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK; else @@ -5041,13 +5122,12 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev if (def != data) WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, data); } - - amdgpu_gfx_rlc_exit_safe_mode(adev); } static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev, bool enable) { + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); if (enable) { /* CGCG/CGLS should be enabled after MGCG/MGLS * === MGCG + MGLS === @@ -5067,10 +5147,12 @@ static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev, /* === MGCG + MGLS === */ gfx_v9_0_update_medium_grain_clock_gating(adev, enable); } + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } -static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid) +static void gfx_v9_0_update_spm_vmid_internal(struct amdgpu_device *adev, + unsigned int vmid) { u32 reg, data; @@ -5078,7 +5160,7 @@ static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid) if (amdgpu_sriov_is_pp_one_vf(adev)) data = RREG32_NO_KIQ(reg); else - data = RREG32(reg); + data = RREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL); data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK; data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT; @@ -5089,6 +5171,15 @@ static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid) WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data); } +static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned int vmid) +{ + amdgpu_gfx_off_ctrl(adev, false); + + gfx_v9_0_update_spm_vmid_internal(adev, vmid); + + amdgpu_gfx_off_ctrl(adev, true); +} + static bool gfx_v9_0_check_rlcg_range(struct amdgpu_device *adev, uint32_t offset, struct soc15_reg_rlcg *entries, int arr_size) @@ -5131,21 +5222,21 @@ static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = { .reset = gfx_v9_0_rlc_reset, .start = gfx_v9_0_rlc_start, .update_spm_vmid = gfx_v9_0_update_spm_vmid, - .rlcg_wreg = gfx_v9_0_rlcg_wreg, .is_rlcg_access_range = gfx_v9_0_is_rlcg_access_range, }; -static int gfx_v9_0_set_powergating_state(void *handle, +static int gfx_v9_0_set_powergating_state(struct amdgpu_ip_block *ip_block, enum amd_powergating_state state) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; bool enable = (state == AMD_PG_STATE_GATE); - switch (adev->asic_type) { - case CHIP_RAVEN: - case CHIP_RENOIR: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 3, 0): if (!enable) - amdgpu_gfx_off_ctrl(adev, false); + amdgpu_gfx_off_ctrl_immediate(adev, false); if (adev->pg_flags & AMD_PG_SUPPORT_RLC_SMU_HS) { gfx_v9_0_enable_sck_slow_down_on_power_up(adev, true); @@ -5167,10 +5258,10 @@ static int gfx_v9_0_set_powergating_state(void *handle, gfx_v9_0_update_gfx_mg_power_gating(adev, enable); if (enable) - amdgpu_gfx_off_ctrl(adev, true); + amdgpu_gfx_off_ctrl_immediate(adev, true); break; - case CHIP_VEGA12: - amdgpu_gfx_off_ctrl(adev, enable); + case IP_VERSION(9, 2, 1): + amdgpu_gfx_off_ctrl_immediate(adev, enable); break; default: break; @@ -5179,22 +5270,23 @@ static int gfx_v9_0_set_powergating_state(void *handle, return 0; } -static int gfx_v9_0_set_clockgating_state(void *handle, +static int gfx_v9_0_set_clockgating_state(struct amdgpu_ip_block *ip_block, enum amd_clockgating_state state) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; if (amdgpu_sriov_vf(adev)) return 0; - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_ARCTURUS: - case CHIP_RENOIR: - case CHIP_ALDEBARAN: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 4, 1): + case IP_VERSION(9, 3, 0): + case IP_VERSION(9, 4, 2): gfx_v9_0_update_gfx_clock_gating(adev, state == AMD_CG_STATE_GATE); break; @@ -5204,9 +5296,9 @@ static int gfx_v9_0_set_clockgating_state(void *handle, return 0; } -static void gfx_v9_0_get_clockgating_state(void *handle, u32 *flags) +static void gfx_v9_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; int data; if (amdgpu_sriov_vf(adev)) @@ -5236,7 +5328,7 @@ static void gfx_v9_0_get_clockgating_state(void *handle, u32 *flags) if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK) *flags |= AMD_CG_SUPPORT_GFX_CP_LS | AMD_CG_SUPPORT_GFX_MGLS; - if (adev->asic_type != CHIP_ARCTURUS) { + if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) { /* AMD_CG_SUPPORT_GFX_3D_CGCG */ data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D)); if (data & RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK) @@ -5250,7 +5342,7 @@ static void gfx_v9_0_get_clockgating_state(void *handle, u32 *flags) static u64 gfx_v9_0_ring_get_rptr_gfx(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; /* gfx9 is 32bit rptr*/ + return *ring->rptr_cpu_addr; /* gfx9 is 32bit rptr*/ } static u64 gfx_v9_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) @@ -5260,7 +5352,7 @@ static u64 gfx_v9_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) { - wptr = atomic64_read((atomic64_t *)&adev->wb.wb[ring->wptr_offs]); + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); } else { wptr = RREG32_SOC15(GC, 0, mmCP_RB0_WPTR); wptr += (u64)RREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI) << 32; @@ -5275,7 +5367,7 @@ static void gfx_v9_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr); + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, ring->wptr); WDOORBELL64(ring->doorbell_index, ring->wptr); } else { WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); @@ -5327,11 +5419,18 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, control |= ib->length_dw | (vmid << 24); - if (amdgpu_sriov_vf(ring->adev) && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) { + if (ib->flags & AMDGPU_IB_FLAG_PREEMPT) { control |= INDIRECT_BUFFER_PRE_ENB(1); + if (flags & AMDGPU_IB_PREEMPTED) + control |= INDIRECT_BUFFER_PRE_RESUME(1); + if (!(ib->flags & AMDGPU_IB_FLAG_CE) && vmid) - gfx_v9_0_ring_emit_de_meta(ring); + gfx_v9_0_ring_emit_de_meta(ring, + (!amdgpu_sriov_vf(ring->adev) && + flags & AMDGPU_IB_PREEMPTED) ? + true : false, + job->gds_size > 0 && job->gds_base != 0); } amdgpu_ring_write(ring, header); @@ -5342,9 +5441,70 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, #endif lower_32_bits(ib->gpu_addr)); amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + amdgpu_ring_ib_on_emit_cntl(ring); amdgpu_ring_write(ring, control); } +static void gfx_v9_0_ring_patch_cntl(struct amdgpu_ring *ring, + unsigned offset) +{ + u32 control = ring->ring[offset]; + + control |= INDIRECT_BUFFER_PRE_RESUME(1); + ring->ring[offset] = control; +} + +static void gfx_v9_0_ring_patch_ce_meta(struct amdgpu_ring *ring, + unsigned offset) +{ + struct amdgpu_device *adev = ring->adev; + void *ce_payload_cpu_addr; + uint64_t payload_offset, payload_size; + + payload_size = sizeof(struct v9_ce_ib_state); + + payload_offset = offsetof(struct v9_gfx_meta_data, ce_payload); + ce_payload_cpu_addr = adev->virt.csa_cpu_addr + payload_offset; + + if (offset + (payload_size >> 2) <= ring->buf_mask + 1) { + memcpy((void *)&ring->ring[offset], ce_payload_cpu_addr, payload_size); + } else { + memcpy((void *)&ring->ring[offset], ce_payload_cpu_addr, + (ring->buf_mask + 1 - offset) << 2); + payload_size -= (ring->buf_mask + 1 - offset) << 2; + memcpy((void *)&ring->ring[0], + ce_payload_cpu_addr + ((ring->buf_mask + 1 - offset) << 2), + payload_size); + } +} + +static void gfx_v9_0_ring_patch_de_meta(struct amdgpu_ring *ring, + unsigned offset) +{ + struct amdgpu_device *adev = ring->adev; + void *de_payload_cpu_addr; + uint64_t payload_offset, payload_size; + + payload_size = sizeof(struct v9_de_ib_state); + + payload_offset = offsetof(struct v9_gfx_meta_data, de_payload); + de_payload_cpu_addr = adev->virt.csa_cpu_addr + payload_offset; + + ((struct v9_de_ib_state *)de_payload_cpu_addr)->ib_completion_status = + IB_COMPLETION_STATUS_PREEMPTED; + + if (offset + (payload_size >> 2) <= ring->buf_mask + 1) { + memcpy((void *)&ring->ring[offset], de_payload_cpu_addr, payload_size); + } else { + memcpy((void *)&ring->ring[offset], de_payload_cpu_addr, + (ring->buf_mask + 1 - offset) << 2); + payload_size -= (ring->buf_mask + 1 - offset) << 2; + memcpy((void *)&ring->ring[0], + de_payload_cpu_addr + ((ring->buf_mask + 1 - offset) << 2), + payload_size); + } +} + static void gfx_v9_0_ring_emit_ib_compute(struct amdgpu_ring *ring, struct amdgpu_job *job, struct amdgpu_ib *ib, @@ -5386,17 +5546,24 @@ static void gfx_v9_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT; bool int_sel = flags & AMDGPU_FENCE_FLAG_INT; bool writeback = flags & AMDGPU_FENCE_FLAG_TC_WB_ONLY; + bool exec = flags & AMDGPU_FENCE_FLAG_EXEC; + uint32_t dw2 = 0; /* RELEASE_MEM - flush caches, send int */ amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6)); - amdgpu_ring_write(ring, ((writeback ? (EOP_TC_WB_ACTION_EN | - EOP_TC_NC_ACTION_EN) : - (EOP_TCL1_ACTION_EN | - EOP_TC_ACTION_EN | - EOP_TC_WB_ACTION_EN | - EOP_TC_MD_ACTION_EN)) | - EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | - EVENT_INDEX(5))); + + if (writeback) { + dw2 = EOP_TC_NC_ACTION_EN; + } else { + dw2 = EOP_TCL1_ACTION_EN | EOP_TC_ACTION_EN | + EOP_TC_MD_ACTION_EN; + } + dw2 |= EOP_TC_WB_ACTION_EN | EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | + EVENT_INDEX(5); + if (exec) + dw2 |= EOP_EXEC; + + amdgpu_ring_write(ring, dw2); amdgpu_ring_write(ring, DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0)); /* @@ -5440,7 +5607,7 @@ static void gfx_v9_0_ring_emit_vm_flush(struct amdgpu_ring *ring, static u64 gfx_v9_0_ring_get_rptr_compute(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; /* gfx9 hardware is 32bit rptr */ + return *ring->rptr_cpu_addr; /* gfx9 hardware is 32bit rptr */ } static u64 gfx_v9_0_ring_get_wptr_compute(struct amdgpu_ring *ring) @@ -5449,7 +5616,7 @@ static u64 gfx_v9_0_ring_get_wptr_compute(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) - wptr = atomic64_read((atomic64_t *)&ring->adev->wb.wb[ring->wptr_offs]); + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); else BUG(); return wptr; @@ -5461,7 +5628,7 @@ static void gfx_v9_0_ring_set_wptr_compute(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) { - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr); + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, ring->wptr); WDOORBELL64(ring->doorbell_index, ring->wptr); } else{ BUG(); /* only DOORBELL method supported on gfx9 now */ @@ -5501,35 +5668,116 @@ static void gfx_v9_ring_emit_sb(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0); } -static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring) +static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume) { + struct amdgpu_device *adev = ring->adev; struct v9_ce_ib_state ce_payload = {0}; - uint64_t csa_addr; + uint64_t offset, ce_payload_gpu_addr; + void *ce_payload_cpu_addr; int cnt; cnt = (sizeof(ce_payload) >> 2) + 4 - 2; - csa_addr = amdgpu_csa_vaddr(ring->adev); + + offset = offsetof(struct v9_gfx_meta_data, ce_payload); + ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset; + ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset; amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt)); amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) | WRITE_DATA_DST_SEL(8) | WR_CONFIRM) | WRITE_DATA_CACHE_POLICY(0)); - amdgpu_ring_write(ring, lower_32_bits(csa_addr + offsetof(struct v9_gfx_meta_data, ce_payload))); - amdgpu_ring_write(ring, upper_32_bits(csa_addr + offsetof(struct v9_gfx_meta_data, ce_payload))); - amdgpu_ring_write_multiple(ring, (void *)&ce_payload, sizeof(ce_payload) >> 2); + amdgpu_ring_write(ring, lower_32_bits(ce_payload_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ce_payload_gpu_addr)); + + amdgpu_ring_ib_on_emit_ce(ring); + + if (resume) + amdgpu_ring_write_multiple(ring, ce_payload_cpu_addr, + sizeof(ce_payload) >> 2); + else + amdgpu_ring_write_multiple(ring, (void *)&ce_payload, + sizeof(ce_payload) >> 2); +} + +static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring) +{ + int i, r = 0; + struct amdgpu_device *adev = ring->adev; + struct amdgpu_kiq *kiq = &adev->gfx.kiq[0]; + struct amdgpu_ring *kiq_ring = &kiq->ring; + unsigned long flags; + + if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) + return -EINVAL; + + spin_lock_irqsave(&kiq->ring_lock, flags); + + if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) { + spin_unlock_irqrestore(&kiq->ring_lock, flags); + return -ENOMEM; + } + + /* assert preemption condition */ + amdgpu_ring_set_preempt_cond_exec(ring, false); + + ring->trail_seq += 1; + amdgpu_ring_alloc(ring, 13); + gfx_v9_0_ring_emit_fence(ring, ring->trail_fence_gpu_addr, + ring->trail_seq, AMDGPU_FENCE_FLAG_EXEC | AMDGPU_FENCE_FLAG_INT); + + /* assert IB preemption, emit the trailing fence */ + kiq->pmf->kiq_unmap_queues(kiq_ring, ring, PREEMPT_QUEUES_NO_UNMAP, + ring->trail_fence_gpu_addr, + ring->trail_seq); + + amdgpu_ring_commit(kiq_ring); + spin_unlock_irqrestore(&kiq->ring_lock, flags); + + /* poll the trailing fence */ + for (i = 0; i < adev->usec_timeout; i++) { + if (ring->trail_seq == + le32_to_cpu(*ring->trail_fence_cpu_addr)) + break; + udelay(1); + } + + if (i >= adev->usec_timeout) { + r = -EINVAL; + DRM_WARN("ring %d timeout to preempt ib\n", ring->idx); + } + + /*reset the CP_VMID_PREEMPT after trailing fence*/ + amdgpu_ring_emit_wreg(ring, + SOC15_REG_OFFSET(GC, 0, mmCP_VMID_PREEMPT), + 0x0); + amdgpu_ring_commit(ring); + + /* deassert preemption condition */ + amdgpu_ring_set_preempt_cond_exec(ring, true); + return r; } -static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) +static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume, bool usegds) { + struct amdgpu_device *adev = ring->adev; struct v9_de_ib_state de_payload = {0}; - uint64_t csa_addr, gds_addr; + uint64_t offset, gds_addr, de_payload_gpu_addr; + void *de_payload_cpu_addr; int cnt; - csa_addr = amdgpu_csa_vaddr(ring->adev); - gds_addr = csa_addr + 4096; - de_payload.gds_backup_addrlo = lower_32_bits(gds_addr); - de_payload.gds_backup_addrhi = upper_32_bits(gds_addr); + offset = offsetof(struct v9_gfx_meta_data, de_payload); + de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset; + de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset; + + gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) + + AMDGPU_CSA_SIZE - adev->gds.gds_size, + PAGE_SIZE); + + if (usegds) { + de_payload.gds_backup_addrlo = lower_32_bits(gds_addr); + de_payload.gds_backup_addrhi = upper_32_bits(gds_addr); + } cnt = (sizeof(de_payload) >> 2) + 4 - 2; amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt)); @@ -5537,9 +5785,16 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) WRITE_DATA_DST_SEL(8) | WR_CONFIRM) | WRITE_DATA_CACHE_POLICY(0)); - amdgpu_ring_write(ring, lower_32_bits(csa_addr + offsetof(struct v9_gfx_meta_data, de_payload))); - amdgpu_ring_write(ring, upper_32_bits(csa_addr + offsetof(struct v9_gfx_meta_data, de_payload))); - amdgpu_ring_write_multiple(ring, (void *)&de_payload, sizeof(de_payload) >> 2); + amdgpu_ring_write(ring, lower_32_bits(de_payload_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(de_payload_gpu_addr)); + + amdgpu_ring_ib_on_emit_de(ring); + if (resume) + amdgpu_ring_write_multiple(ring, de_payload_cpu_addr, + sizeof(de_payload) >> 2); + else + amdgpu_ring_write_multiple(ring, (void *)&de_payload, + sizeof(de_payload) >> 2); } static void gfx_v9_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start, @@ -5555,8 +5810,9 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) { uint32_t dw2 = 0; - if (amdgpu_sriov_vf(ring->adev)) - gfx_v9_0_ring_emit_ce_meta(ring); + gfx_v9_0_ring_emit_ce_meta(ring, + (!amdgpu_sriov_vf(ring->adev) && + flags & AMDGPU_IB_PREEMPTED) ? true : false); dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */ if (flags & AMDGPU_HAVE_CTX_SWITCH) { @@ -5583,31 +5839,21 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) amdgpu_ring_write(ring, 0); } -static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring) +static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring, + uint64_t addr) { unsigned ret; amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3)); - amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr)); - amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr)); - amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */ + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + /* discard following DWs if *cond_exec_gpu_addr==0 */ + amdgpu_ring_write(ring, 0); ret = ring->wptr & ring->buf_mask; - amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */ + /* patch dummy value later */ + amdgpu_ring_write(ring, 0); return ret; } -static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset) -{ - unsigned cur; - BUG_ON(offset > ring->buf_mask); - BUG_ON(ring->ring[offset] != 0x55aa55aa); - - cur = (ring->wptr & ring->buf_mask) - 1; - if (likely(cur > offset)) - ring->ring[offset] = cur - offset; - else - ring->ring[offset] = (ring->ring_size>>2) - offset + cur; -} - static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t reg_val_offs) { @@ -5680,7 +5926,9 @@ static void gfx_v9_0_ring_soft_recovery(struct amdgpu_ring *ring, unsigned vmid) value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01); value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1); value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); WREG32_SOC15(GC, 0, mmSQ_CMD, value); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } static void gfx_v9_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev, @@ -5735,33 +5983,111 @@ static void gfx_v9_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, switch (state) { case AMDGPU_IRQ_STATE_DISABLE: - mec_int_cntl = RREG32(mec_int_cntl_reg); + mec_int_cntl = RREG32_SOC15_IP(GC,mec_int_cntl_reg); mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, TIME_STAMP_INT_ENABLE, 0); - WREG32(mec_int_cntl_reg, mec_int_cntl); + WREG32_SOC15_IP(GC, mec_int_cntl_reg, mec_int_cntl); break; case AMDGPU_IRQ_STATE_ENABLE: - mec_int_cntl = RREG32(mec_int_cntl_reg); + mec_int_cntl = RREG32_SOC15_IP(GC, mec_int_cntl_reg); mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, TIME_STAMP_INT_ENABLE, 1); - WREG32(mec_int_cntl_reg, mec_int_cntl); + WREG32_SOC15_IP(GC, mec_int_cntl_reg, mec_int_cntl); break; default: break; } } +static u32 gfx_v9_0_get_cpc_int_cntl(struct amdgpu_device *adev, + int me, int pipe) +{ + /* + * amdgpu controls only the first MEC. That's why this function only + * handles the setting of interrupts for this specific MEC. All other + * pipes' interrupts are set by amdkfd. + */ + if (me != 1) + return 0; + + switch (pipe) { + case 0: + return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL); + case 1: + return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE1_INT_CNTL); + case 2: + return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE2_INT_CNTL); + case 3: + return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE3_INT_CNTL); + default: + return 0; + } +} + static int gfx_v9_0_set_priv_reg_fault_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, unsigned type, enum amdgpu_interrupt_state state) { + u32 cp_int_cntl_reg, cp_int_cntl; + int i, j; + switch (state) { case AMDGPU_IRQ_STATE_DISABLE: case AMDGPU_IRQ_STATE_ENABLE: WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0, PRIV_REG_INT_ENABLE, state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + for (i = 0; i < adev->gfx.mec.num_mec; i++) { + for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) { + /* MECs start at 1 */ + cp_int_cntl_reg = gfx_v9_0_get_cpc_int_cntl(adev, i + 1, j); + + if (cp_int_cntl_reg) { + cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg); + cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL, + PRIV_REG_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl); + } + } + } + break; + default: + break; + } + + return 0; +} + +static int gfx_v9_0_set_bad_op_fault_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + u32 cp_int_cntl_reg, cp_int_cntl; + int i, j; + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + case AMDGPU_IRQ_STATE_ENABLE: + WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0, + OPCODE_ERROR_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + for (i = 0; i < adev->gfx.mec.num_mec; i++) { + for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) { + /* MECs start at 1 */ + cp_int_cntl_reg = gfx_v9_0_get_cpc_int_cntl(adev, i + 1, j); + + if (cp_int_cntl_reg) { + cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg); + cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL, + OPCODE_ERROR_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl); + } + } + } break; default: break; @@ -5882,7 +6208,15 @@ static int gfx_v9_0_eop_irq(struct amdgpu_device *adev, switch (me_id) { case 0: - amdgpu_fence_process(&adev->gfx.gfx_ring[0]); + if (adev->gfx.num_gfx_rings) { + if (!adev->gfx.mcbp) { + amdgpu_fence_process(&adev->gfx.gfx_ring[0]); + } else if (!amdgpu_mcbp_handle_trailing_fence_irq(&adev->gfx.muxer)) { + /* Fence signals are handled on the software rings*/ + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) + amdgpu_fence_process(&adev->gfx.sw_gfx_ring[i]); + } + } break; case 1: case 2: @@ -5935,6 +6269,15 @@ static int gfx_v9_0_priv_reg_irq(struct amdgpu_device *adev, return 0; } +static int gfx_v9_0_bad_op_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + DRM_ERROR("Illegal opcode in command stream\n"); + gfx_v9_0_fault(adev, entry); + return 0; +} + static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -6387,7 +6730,7 @@ static const struct soc15_ras_field_entry gfx_v9_0_ras_fields[] = { }; static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev, - void *inject_if) + void *inject_if, uint32_t instance_mask) { struct ras_inject_if *info = (struct ras_inject_if *)inject_if; int ret; @@ -6426,13 +6769,13 @@ static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev, block_info.value = info->value; mutex_lock(&adev->grbm_idx_mutex); - ret = psp_ras_trigger_error(&adev->psp, &block_info); + ret = psp_ras_trigger_error(&adev->psp, &block_info, instance_mask); mutex_unlock(&adev->grbm_idx_mutex); return ret; } -static const char *vml2_mems[] = { +static const char * const vml2_mems[] = { "UTC_VML2_BANK_CACHE_0_BIGK_MEM0", "UTC_VML2_BANK_CACHE_0_BIGK_MEM1", "UTC_VML2_BANK_CACHE_0_4K_MEM0", @@ -6451,7 +6794,7 @@ static const char *vml2_mems[] = { "UTC_VML2_BANK_CACHE_3_4K_MEM1", }; -static const char *vml2_walker_mems[] = { +static const char * const vml2_walker_mems[] = { "UTC_VML2_CACHE_PDE0_MEM0", "UTC_VML2_CACHE_PDE0_MEM1", "UTC_VML2_CACHE_PDE1_MEM0", @@ -6461,7 +6804,7 @@ static const char *vml2_walker_mems[] = { "UTC_VML2_RDIF_LOG_FIFO", }; -static const char *atc_l2_cache_2m_mems[] = { +static const char * const atc_l2_cache_2m_mems[] = { "UTC_ATCL2_CACHE_2M_BANK0_WAY0_MEM", "UTC_ATCL2_CACHE_2M_BANK0_WAY1_MEM", "UTC_ATCL2_CACHE_2M_BANK1_WAY0_MEM", @@ -6654,7 +6997,7 @@ static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev) for (i = 0; i < ARRAY_SIZE(gfx_v9_0_edc_counter_regs); i++) { for (j = 0; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) { for (k = 0; k < gfx_v9_0_edc_counter_regs[i].instance; k++) { - gfx_v9_0_select_se_sh(adev, j, 0x0, k); + amdgpu_gfx_select_se_sh(adev, j, 0x0, k, 0); RREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_0_edc_counter_regs[i])); } } @@ -6697,7 +7040,7 @@ static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, mmATC_L2_CACHE_4K_EDC_INDEX, 255); } -static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -6706,7 +7049,7 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, uint32_t reg_value; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) - return -EINVAL; + return; err_data->ue_count = 0; err_data->ce_count = 0; @@ -6716,7 +7059,7 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, for (i = 0; i < ARRAY_SIZE(gfx_v9_0_edc_counter_regs); i++) { for (j = 0; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) { for (k = 0; k < gfx_v9_0_edc_counter_regs[i].instance; k++) { - gfx_v9_0_select_se_sh(adev, j, 0, k); + amdgpu_gfx_select_se_sh(adev, j, 0, k, 0); reg_value = RREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_0_edc_counter_regs[i])); if (reg_value) @@ -6731,12 +7074,10 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, err_data->ce_count += sec_count; err_data->ue_count += ded_count; - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); gfx_v9_0_query_utc_edc_status(adev, err_data); - - return 0; } static void gfx_v9_0_emit_mem_sync(struct amdgpu_ring *ring) @@ -6817,6 +7158,230 @@ static void gfx_v9_0_emit_wave_limit(struct amdgpu_ring *ring, bool enable) } } +static void gfx_v9_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop) +{ + /* Header itself is a NOP packet */ + if (num_nop == 1) { + amdgpu_ring_write(ring, ring->funcs->nop); + return; + } + + /* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/ + amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe))); + + /* Header is at index 0, followed by num_nops - 1 NOP packet's */ + amdgpu_ring_insert_nop(ring, num_nop - 1); +} + +static int gfx_v9_0_reset_kcq(struct amdgpu_ring *ring, + unsigned int vmid, + struct amdgpu_fence *timedout_fence) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_kiq *kiq = &adev->gfx.kiq[0]; + struct amdgpu_ring *kiq_ring = &kiq->ring; + unsigned long flags; + int i, r; + + if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) + return -EINVAL; + + amdgpu_ring_reset_helper_begin(ring, timedout_fence); + + spin_lock_irqsave(&kiq->ring_lock, flags); + + if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) { + spin_unlock_irqrestore(&kiq->ring_lock, flags); + return -ENOMEM; + } + + kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES, + 0, 0); + amdgpu_ring_commit(kiq_ring); + + spin_unlock_irqrestore(&kiq->ring_lock, flags); + + r = amdgpu_ring_test_ring(kiq_ring); + if (r) + return r; + + /* make sure dequeue is complete*/ + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); + mutex_lock(&adev->srbm_mutex); + soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0); + for (i = 0; i < adev->usec_timeout; i++) { + if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1)) + break; + udelay(1); + } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + soc15_grbm_select(adev, 0, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); + if (r) { + dev_err(adev->dev, "fail to wait on hqd deactive\n"); + return r; + } + + r = gfx_v9_0_kcq_init_queue(ring, true); + if (r) { + dev_err(adev->dev, "fail to init kcq\n"); + return r; + } + spin_lock_irqsave(&kiq->ring_lock, flags); + r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size); + if (r) { + spin_unlock_irqrestore(&kiq->ring_lock, flags); + return -ENOMEM; + } + kiq->pmf->kiq_map_queues(kiq_ring, ring); + amdgpu_ring_commit(kiq_ring); + r = amdgpu_ring_test_ring(kiq_ring); + spin_unlock_irqrestore(&kiq->ring_lock, flags); + if (r) { + DRM_ERROR("fail to remap queue\n"); + return r; + } + return amdgpu_ring_reset_helper_end(ring, timedout_fence); +} + +static void gfx_v9_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p) +{ + struct amdgpu_device *adev = ip_block->adev; + uint32_t i, j, k, reg, index = 0; + uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9); + + if (!adev->gfx.ip_dump_core) + return; + + for (i = 0; i < reg_count; i++) + drm_printf(p, "%-50s \t 0x%08x\n", + gc_reg_list_9[i].reg_name, + adev->gfx.ip_dump_core[i]); + + /* print compute queue registers for all instances */ + if (!adev->gfx.ip_dump_compute_queues) + return; + + reg_count = ARRAY_SIZE(gc_cp_reg_list_9); + drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n", + adev->gfx.mec.num_mec, + adev->gfx.mec.num_pipe_per_mec, + adev->gfx.mec.num_queue_per_pipe); + + for (i = 0; i < adev->gfx.mec.num_mec; i++) { + for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) { + for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) { + drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k); + for (reg = 0; reg < reg_count; reg++) { + if (i && gc_cp_reg_list_9[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP) + drm_printf(p, "%-50s \t 0x%08x\n", + "mmCP_MEC_ME2_HEADER_DUMP", + adev->gfx.ip_dump_compute_queues[index + reg]); + else + drm_printf(p, "%-50s \t 0x%08x\n", + gc_cp_reg_list_9[reg].reg_name, + adev->gfx.ip_dump_compute_queues[index + reg]); + } + index += reg_count; + } + } + } + +} + +static void gfx_v9_ip_dump(struct amdgpu_ip_block *ip_block) +{ + struct amdgpu_device *adev = ip_block->adev; + uint32_t i, j, k, reg, index = 0; + uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9); + + if (!adev->gfx.ip_dump_core || !adev->gfx.num_gfx_rings) + return; + + amdgpu_gfx_off_ctrl(adev, false); + for (i = 0; i < reg_count; i++) + adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_9[i])); + amdgpu_gfx_off_ctrl(adev, true); + + /* dump compute queue registers for all instances */ + if (!adev->gfx.ip_dump_compute_queues) + return; + + reg_count = ARRAY_SIZE(gc_cp_reg_list_9); + amdgpu_gfx_off_ctrl(adev, false); + mutex_lock(&adev->srbm_mutex); + for (i = 0; i < adev->gfx.mec.num_mec; i++) { + for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) { + for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) { + /* ME0 is for GFX so start from 1 for CP */ + soc15_grbm_select(adev, 1 + i, j, k, 0, 0); + + for (reg = 0; reg < reg_count; reg++) { + if (i && gc_cp_reg_list_9[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP) + adev->gfx.ip_dump_compute_queues[index + reg] = + RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME2_HEADER_DUMP)); + else + adev->gfx.ip_dump_compute_queues[index + reg] = + RREG32(SOC15_REG_ENTRY_OFFSET( + gc_cp_reg_list_9[reg])); + } + index += reg_count; + } + } + } + soc15_grbm_select(adev, 0, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + amdgpu_gfx_off_ctrl(adev, true); + +} + +static void gfx_v9_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + /* Emit the cleaner shader */ + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2)) + amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0)); + else + amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER_9_0, 0)); + + amdgpu_ring_write(ring, 0); /* RESERVED field, programmed to zero */ +} + +static void gfx_v9_0_ring_begin_use_compute(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ip_block *gfx_block = + amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX); + + amdgpu_gfx_enforce_isolation_ring_begin_use(ring); + + /* Raven and PCO APUs seem to have stability issues + * with compute and gfxoff and gfx pg. Disable gfx pg during + * submission and allow again afterwards. + */ + if (gfx_block && amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0)) + gfx_v9_0_set_powergating_state(gfx_block, AMD_PG_STATE_UNGATE); +} + +static void gfx_v9_0_ring_end_use_compute(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ip_block *gfx_block = + amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX); + + /* Raven and PCO APUs seem to have stability issues + * with compute and gfxoff and gfx pg. Disable gfx pg during + * submission and allow again afterwards. + */ + if (gfx_block && amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0)) + gfx_v9_0_set_powergating_state(gfx_block, AMD_PG_STATE_GATE); + + amdgpu_gfx_enforce_isolation_ring_end_use(ring); +} + static const struct amd_ip_funcs gfx_v9_0_ip_funcs = { .name = "gfx_v9_0", .early_init = gfx_v9_0_early_init, @@ -6833,6 +7398,8 @@ static const struct amd_ip_funcs gfx_v9_0_ip_funcs = { .set_clockgating_state = gfx_v9_0_set_clockgating_state, .set_powergating_state = gfx_v9_0_set_powergating_state, .get_clockgating_state = gfx_v9_0_get_clockgating_state, + .dump_ip_state = gfx_v9_ip_dump, + .print_ip_state = gfx_v9_ip_print, }; static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { @@ -6840,7 +7407,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .align_mask = 0xff, .nop = PACKET3(PACKET3_NOP, 0x3FFF), .support_64bit_ptrs = true, - .vmhub = AMDGPU_GFXHUB_0, + .secure_submission_supported = true, .get_rptr = gfx_v9_0_ring_get_rptr_gfx, .get_wptr = gfx_v9_0_ring_get_wptr_gfx, .set_wptr = gfx_v9_0_ring_set_wptr_gfx, @@ -6864,7 +7431,65 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { 5 + /* HDP_INVL */ 8 + 8 + /* FENCE x2 */ 2 + /* SWITCH_BUFFER */ - 7, /* gfx_v9_0_emit_mem_sync */ + 7 + /* gfx_v9_0_emit_mem_sync */ + 2, /* gfx_v9_0_ring_emit_cleaner_shader */ + .emit_ib_size = 4, /* gfx_v9_0_ring_emit_ib_gfx */ + .emit_ib = gfx_v9_0_ring_emit_ib_gfx, + .emit_fence = gfx_v9_0_ring_emit_fence, + .emit_pipeline_sync = gfx_v9_0_ring_emit_pipeline_sync, + .emit_vm_flush = gfx_v9_0_ring_emit_vm_flush, + .emit_gds_switch = gfx_v9_0_ring_emit_gds_switch, + .emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush, + .test_ring = gfx_v9_0_ring_test_ring, + .insert_nop = gfx_v9_ring_insert_nop, + .pad_ib = amdgpu_ring_generic_pad_ib, + .emit_switch_buffer = gfx_v9_ring_emit_sb, + .emit_cntxcntl = gfx_v9_ring_emit_cntxcntl, + .init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec, + .preempt_ib = gfx_v9_0_ring_preempt_ib, + .emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl, + .emit_wreg = gfx_v9_0_ring_emit_wreg, + .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, + .soft_recovery = gfx_v9_0_ring_soft_recovery, + .emit_mem_sync = gfx_v9_0_emit_mem_sync, + .emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader, + .begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use, + .end_use = amdgpu_gfx_enforce_isolation_ring_end_use, +}; + +static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = { + .type = AMDGPU_RING_TYPE_GFX, + .align_mask = 0xff, + .nop = PACKET3(PACKET3_NOP, 0x3FFF), + .support_64bit_ptrs = true, + .secure_submission_supported = true, + .get_rptr = amdgpu_sw_ring_get_rptr_gfx, + .get_wptr = amdgpu_sw_ring_get_wptr_gfx, + .set_wptr = amdgpu_sw_ring_set_wptr_gfx, + .emit_frame_size = /* totally 242 maximum if 16 IBs */ + 5 + /* COND_EXEC */ + 7 + /* PIPELINE_SYNC */ + SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + + 2 + /* VM_FLUSH */ + 8 + /* FENCE for VM_FLUSH */ + 20 + /* GDS switch */ + 4 + /* double SWITCH_BUFFER, + * the first COND_EXEC jump to the place just + * prior to this double SWITCH_BUFFER + */ + 5 + /* COND_EXEC */ + 7 + /* HDP_flush */ + 4 + /* VGT_flush */ + 14 + /* CE_META */ + 31 + /* DE_META */ + 3 + /* CNTX_CTRL */ + 5 + /* HDP_INVL */ + 8 + 8 + /* FENCE x2 */ + 2 + /* SWITCH_BUFFER */ + 7 + /* gfx_v9_0_emit_mem_sync */ + 2, /* gfx_v9_0_ring_emit_cleaner_shader */ .emit_ib_size = 4, /* gfx_v9_0_ring_emit_ib_gfx */ .emit_ib = gfx_v9_0_ring_emit_ib_gfx, .emit_fence = gfx_v9_0_ring_emit_fence, @@ -6874,18 +7499,23 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush, .test_ring = gfx_v9_0_ring_test_ring, .test_ib = gfx_v9_0_ring_test_ib, - .insert_nop = amdgpu_ring_insert_nop, + .insert_nop = gfx_v9_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .emit_switch_buffer = gfx_v9_ring_emit_sb, .emit_cntxcntl = gfx_v9_ring_emit_cntxcntl, .init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec, - .patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec, .emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl, .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, .soft_recovery = gfx_v9_0_ring_soft_recovery, .emit_mem_sync = gfx_v9_0_emit_mem_sync, + .patch_cntl = gfx_v9_0_ring_patch_cntl, + .patch_de = gfx_v9_0_ring_patch_de_meta, + .patch_ce = gfx_v9_0_ring_patch_ce_meta, + .emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader, + .begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use, + .end_use = amdgpu_gfx_enforce_isolation_ring_end_use, }; static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { @@ -6893,7 +7523,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { .align_mask = 0xff, .nop = PACKET3(PACKET3_NOP, 0x3FFF), .support_64bit_ptrs = true, - .vmhub = AMDGPU_GFXHUB_0, .get_rptr = gfx_v9_0_ring_get_rptr_compute, .get_wptr = gfx_v9_0_ring_get_wptr_compute, .set_wptr = gfx_v9_0_ring_set_wptr_compute, @@ -6904,11 +7533,11 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { 7 + /* gfx_v9_0_ring_emit_pipeline_sync */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + - 2 + /* gfx_v9_0_ring_emit_vm_flush */ 8 + 8 + 8 + /* gfx_v9_0_ring_emit_fence x3 for user fence, vm fence */ 7 + /* gfx_v9_0_emit_mem_sync */ 5 + /* gfx_v9_0_emit_wave_limit for updating mmSPI_WCL_PIPE_PERCENT_GFX register */ - 15, /* for updating 3 mmSPI_WCL_PIPE_PERCENT_CS registers */ + 15 + /* for updating 3 mmSPI_WCL_PIPE_PERCENT_CS registers */ + 2, /* gfx_v9_0_ring_emit_cleaner_shader */ .emit_ib_size = 7, /* gfx_v9_0_ring_emit_ib_compute */ .emit_ib = gfx_v9_0_ring_emit_ib_compute, .emit_fence = gfx_v9_0_ring_emit_fence, @@ -6918,13 +7547,18 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { .emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush, .test_ring = gfx_v9_0_ring_test_ring, .test_ib = gfx_v9_0_ring_test_ib, - .insert_nop = amdgpu_ring_insert_nop, + .insert_nop = gfx_v9_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, + .soft_recovery = gfx_v9_0_ring_soft_recovery, .emit_mem_sync = gfx_v9_0_emit_mem_sync, .emit_wave_limit = gfx_v9_0_emit_wave_limit, + .reset = gfx_v9_0_reset_kcq, + .emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader, + .begin_use = gfx_v9_0_ring_begin_use_compute, + .end_use = gfx_v9_0_ring_end_use_compute, }; static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { @@ -6932,7 +7566,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { .align_mask = 0xff, .nop = PACKET3(PACKET3_NOP, 0x3FFF), .support_64bit_ptrs = true, - .vmhub = AMDGPU_GFXHUB_0, .get_rptr = gfx_v9_0_ring_get_rptr_compute, .get_wptr = gfx_v9_0_ring_get_wptr_compute, .set_wptr = gfx_v9_0_ring_set_wptr_compute, @@ -6943,7 +7576,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { 7 + /* gfx_v9_0_ring_emit_pipeline_sync */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + - 2 + /* gfx_v9_0_ring_emit_vm_flush */ 8 + 8 + 8, /* gfx_v9_0_ring_emit_fence_kiq x3 for user fence, vm fence */ .emit_ib_size = 7, /* gfx_v9_0_ring_emit_ib_compute */ .emit_fence = gfx_v9_0_ring_emit_fence_kiq, @@ -6954,17 +7586,23 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, + .emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush, }; static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev) { int i; - adev->gfx.kiq.ring.funcs = &gfx_v9_0_ring_funcs_kiq; + adev->gfx.kiq[0].ring.funcs = &gfx_v9_0_ring_funcs_kiq; for (i = 0; i < adev->gfx.num_gfx_rings; i++) adev->gfx.gfx_ring[i].funcs = &gfx_v9_0_ring_funcs_gfx; + if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) { + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) + adev->gfx.sw_gfx_ring[i].funcs = &gfx_v9_0_sw_ring_funcs_gfx; + } + for (i = 0; i < adev->gfx.num_compute_rings; i++) adev->gfx.compute_ring[i].funcs = &gfx_v9_0_ring_funcs_compute; } @@ -6979,6 +7617,11 @@ static const struct amdgpu_irq_src_funcs gfx_v9_0_priv_reg_irq_funcs = { .process = gfx_v9_0_priv_reg_irq, }; +static const struct amdgpu_irq_src_funcs gfx_v9_0_bad_op_irq_funcs = { + .set = gfx_v9_0_set_bad_op_fault_state, + .process = gfx_v9_0_bad_op_irq, +}; + static const struct amdgpu_irq_src_funcs gfx_v9_0_priv_inst_irq_funcs = { .set = gfx_v9_0_set_priv_inst_fault_state, .process = gfx_v9_0_priv_inst_irq, @@ -6998,6 +7641,9 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev) adev->gfx.priv_reg_irq.num_types = 1; adev->gfx.priv_reg_irq.funcs = &gfx_v9_0_priv_reg_irq_funcs; + adev->gfx.bad_op_irq.num_types = 1; + adev->gfx.bad_op_irq.funcs = &gfx_v9_0_bad_op_irq_funcs; + adev->gfx.priv_inst_irq.num_types = 1; adev->gfx.priv_inst_irq.funcs = &gfx_v9_0_priv_inst_irq_funcs; @@ -7007,14 +7653,15 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev) static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_ARCTURUS: - case CHIP_RENOIR: - case CHIP_ALDEBARAN: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 4, 1): + case IP_VERSION(9, 3, 0): + case IP_VERSION(9, 4, 2): adev->gfx.rlc.funcs = &gfx_v9_0_rlc_funcs; break; default: @@ -7025,17 +7672,18 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev) static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev) { /* init asci gds info */ - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): adev->gds.gds_size = 0x10000; break; - case CHIP_RAVEN: - case CHIP_ARCTURUS: + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 4, 1): adev->gds.gds_size = 0x1000; break; - case CHIP_ALDEBARAN: + case IP_VERSION(9, 4, 2): /* aldebaran removed all the GDS internal memory, * only support GWS opcode in kernel, like barrier * semaphore.etc */ @@ -7046,24 +7694,25 @@ static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev) break; } - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA20: + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 4, 0): adev->gds.gds_compute_max_wave_id = 0x7ff; break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): adev->gds.gds_compute_max_wave_id = 0x27f; break; - case CHIP_RAVEN: + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): if (adev->apu_flags & AMD_APU_IS_RAVEN2) adev->gds.gds_compute_max_wave_id = 0x77; /* raven2 */ else adev->gds.gds_compute_max_wave_id = 0x15f; /* raven1 */ break; - case CHIP_ARCTURUS: + case IP_VERSION(9, 4, 1): adev->gds.gds_compute_max_wave_id = 0xfff; break; - case CHIP_ALDEBARAN: + case IP_VERSION(9, 4, 2): /* deprecated for Aldebaran, no usage at all */ adev->gds.gds_compute_max_wave_id = 0; break; @@ -7133,7 +7782,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, mask = 1; ao_bitmap = 0; counter = 0; - gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0); gfx_v9_0_set_user_cu_inactive_bitmap( adev, disable_masks[i * adev->gfx.config.max_sh_per_se + j]); bitmap = gfx_v9_0_get_cu_active_bitmap(adev); @@ -7150,7 +7799,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, * SE6,SH0 --> bitmap[2][1] * SE7,SH0 --> bitmap[3][1] */ - cu_info->bitmap[i % 4][j + i / 4] = bitmap; + cu_info->bitmap[0][i % 4][j + i / 4] = bitmap; for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) { if (bitmap & mask) { @@ -7166,7 +7815,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, cu_info->ao_cu_bitmap[i % 4][j + i / 4] = ao_bitmap; } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); cu_info->number = active_cu_number; |