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 | 1178 |
1 files changed, 872 insertions, 306 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 458faf657042..0148d7ff34d9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -50,6 +50,7 @@ #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" @@ -149,6 +150,162 @@ MODULE_FIRMWARE("amdgpu/aldebaran_sjt_mec2.bin"); #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, @@ -764,10 +921,18 @@ static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev, 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) | @@ -777,8 +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 */ } @@ -875,12 +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, @@ -895,7 +1095,7 @@ static void gfx_v9_0_set_kiq_pm4_funcs(struct amdgpu_device *adev) static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) { - switch (adev->ip_versions[GC_HWIP][0]) { + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { case IP_VERSION(9, 0, 1): soc15_program_register_sequence(adev, golden_settings_gc_9_0, @@ -951,8 +1151,8 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) break; } - if ((adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) && - (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 2))) + 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)); } @@ -1039,8 +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; @@ -1070,7 +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); @@ -1095,14 +1295,15 @@ 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->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) && + 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->ip_versions[GC_HWIP][0]) { + 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) && @@ -1172,6 +1373,10 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = { { 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 }, }; @@ -1202,7 +1407,7 @@ static bool is_raven_kicker(struct amdgpu_device *adev) static bool check_if_enlarge_doorbell_range(struct amdgpu_device *adev) { - if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 3, 0)) && + 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; @@ -1215,7 +1420,7 @@ 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->ip_versions[GC_HWIP][0]) { + 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): @@ -1249,23 +1454,25 @@ 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, char *chip_name) { - char fw_name[30]; int err; - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); - err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name); + err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_pfp.bin", chip_name); if (err) goto out; amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); - err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name); + err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_me.bin", chip_name); if (err) goto out; amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); - err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name); + err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_ce.bin", chip_name); if (err) goto out; amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE); @@ -1282,7 +1489,6 @@ out: static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev, char *chip_name) { - char fw_name[30]; int err; const struct rlc_firmware_header_v2_0 *rlc_hdr; uint16_t version_major; @@ -1300,20 +1506,25 @@ 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 = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name); + err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_rlc.bin", chip_name); if (err) goto out; - 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); err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor); @@ -1326,9 +1537,9 @@ out: static bool gfx_v9_0_load_mec2_fw_bin_support(struct amdgpu_device *adev) { - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) || - adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) || - adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 3, 0)) + 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; @@ -1337,28 +1548,31 @@ static bool gfx_v9_0_load_mec2_fw_bin_support(struct amdgpu_device *adev) static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev, char *chip_name) { - char fw_name[30]; int err; if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN)) - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sjt_mec.bin", chip_name); + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_sjt_mec.bin", chip_name); else - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); - - err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name); + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec.bin", chip_name); if (err) goto out; + 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)) { if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN)) - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sjt_mec2.bin", chip_name); + err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_sjt_mec2.bin", chip_name); else - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); - - /* ignore failures to load */ - err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name); + err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec2.bin", chip_name); if (!err) { amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2); amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT); @@ -1434,45 +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) @@ -1485,7 +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->ip_versions[GC_HWIP][0] == IP_VERSION(9, 2, 1)) + 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; @@ -1499,7 +1686,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) 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) @@ -1836,7 +2023,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) u32 gb_addr_config; int err; - switch (adev->ip_versions[GC_HWIP][0]) { + 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; @@ -1994,15 +2181,43 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, 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; + struct amdgpu_device *adev = ip_block->adev; unsigned int hw_prio; - switch (adev->ip_versions[GC_HWIP][0]) { + 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): @@ -2018,6 +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; @@ -2026,6 +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); @@ -2080,7 +2339,7 @@ static int gfx_v9_0_sw_init(void *handle) ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; /* disable scheduler on the real ring */ - ring->no_scheduler = true; + 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, @@ -2090,7 +2349,7 @@ static int gfx_v9_0_sw_init(void *handle) } /* set up the software rings */ - if (adev->gfx.num_gfx_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; @@ -2145,14 +2404,21 @@ static int gfx_v9_0_sw_init(void *handle) } } + /* 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[0]; - r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq, 0); + r = amdgpu_gfx_kiq_init_ring(adev, xcc_id); if (r) return r; @@ -2172,16 +2438,22 @@ static int gfx_v9_0_sw_init(void *handle) 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.num_gfx_rings) { + 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); @@ -2196,6 +2468,8 @@ static int gfx_v9_0_sw_fini(void *handle) 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_free_kernel(&adev->gfx.rlc.clear_state_obj, &adev->gfx.rlc.clear_state_gpu_addr, @@ -2207,6 +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; } @@ -2363,13 +2642,16 @@ static void gfx_v9_0_init_sq_config(struct amdgpu_device *adev) { uint32_t tmp; - switch (adev->ip_versions[GC_HWIP][0]) { + 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, !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; } @@ -2380,7 +2662,10 @@ 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); @@ -2474,7 +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); @@ -2700,7 +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->ip_versions[GC_HWIP][0] != IP_VERSION(9, 3, 0)) + if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 3, 0)) pwr_10_0_gfxip_control_over_cgpg(adev, true); } } @@ -2812,7 +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->ip_versions[GC_HWIP][0] == IP_VERSION(9, 2, 1) || + 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); @@ -2925,7 +3211,7 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) return r; } - switch (adev->ip_versions[GC_HWIP][0]) { + 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); @@ -2956,6 +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); @@ -3033,6 +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); @@ -3110,7 +3413,7 @@ 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 */ + /* 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); @@ -3157,7 +3460,15 @@ 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)); + (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); @@ -3215,9 +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) @@ -3566,7 +3875,7 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) 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; @@ -3578,8 +3887,8 @@ 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; @@ -3606,55 +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[0].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)) { - amdgpu_bo_unreserve(ring->mqd_obj); - 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); + 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, 0); -done: - return r; + return amdgpu_gfx_enable_kcq(adev, 0); } static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) @@ -3678,6 +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; @@ -3713,8 +3994,8 @@ static void gfx_v9_0_init_tcp_config(struct amdgpu_device *adev) { u32 tmp; - if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1) && - adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 2)) + 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); @@ -3734,10 +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); @@ -3754,20 +4038,22 @@ static int gfx_v9_0_hw_init(void *handle) if (r) return r; - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + 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; 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()) @@ -3802,7 +4088,7 @@ static int gfx_v9_0_hw_fini(void *handle) /* Skip stopping RLC with A+A reset or when RLC controls GFX clock */ if ((adev->gmc.xgmi.connected_to_cpu && amdgpu_in_reset(adev)) || - (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(9, 4, 2))) { + (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(9, 4, 2))) { dev_dbg(adev->dev, "Skipping RLC halt\n"); return 0; } @@ -3811,19 +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)) @@ -3832,24 +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); @@ -3888,19 +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); @@ -3986,7 +4270,7 @@ static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev) { uint64_t clock, clock_lo, clock_hi, hi_check; - switch (adev->ip_versions[GC_HWIP][0]) { + 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); @@ -4005,7 +4289,9 @@ static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev) default: amdgpu_gfx_off_ctrl(adev, false); mutex_lock(&adev->gfx.gpu_clock_mutex); - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 0, 1) && amdgpu_sriov_runtime(adev)) { + 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); @@ -4357,7 +4643,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) if (!ring->sched.ready) return 0; - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1)) { + 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; @@ -4497,20 +4783,20 @@ 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->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) || - adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + 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; @@ -4529,9 +4815,9 @@ static int gfx_v9_0_early_init(void *handle) 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; /* @@ -4548,7 +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->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + 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); @@ -4563,9 +4849,9 @@ static int gfx_v9_0_ecc_late_init(void *handle) 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); @@ -4576,11 +4862,15 @@ 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; - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + 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 @@ -4669,14 +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, 0); - /* 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->ip_versions[GC_HWIP][0] != IP_VERSION(9, 2, 1)) + 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 | @@ -4710,7 +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->ip_versions[GC_HWIP][0] != IP_VERSION(9, 2, 1)) + 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 | @@ -4735,8 +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, 0); } static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, @@ -4747,8 +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, 0); - /* Enable 3D CGCG/CGLS */ if (enable) { /* write cmd to clear cgcg/cgls ov */ @@ -4790,8 +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, 0); } static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev, @@ -4799,8 +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, 0); - if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) { def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); /* unset CGCG override */ @@ -4816,7 +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->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1)) + 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 @@ -4842,13 +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, 0); } 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 === @@ -4868,6 +5147,7 @@ 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; } @@ -4891,7 +5171,7 @@ static void gfx_v9_0_update_spm_vmid_internal(struct amdgpu_device *adev, WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data); } -static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned int vmid) +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); @@ -4945,18 +5225,18 @@ static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = { .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->ip_versions[GC_HWIP][0]) { + 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); @@ -4978,10 +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 IP_VERSION(9, 2, 1): - amdgpu_gfx_off_ctrl(adev, enable); + amdgpu_gfx_off_ctrl_immediate(adev, enable); break; default: break; @@ -4990,15 +5270,15 @@ 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->ip_versions[GC_HWIP][0]) { + 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): @@ -5016,9 +5296,9 @@ static int gfx_v9_0_set_clockgating_state(void *handle, return 0; } -static void gfx_v9_0_get_clockgating_state(void *handle, u64 *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)) @@ -5048,7 +5328,7 @@ static void gfx_v9_0_get_clockgating_state(void *handle, u64 *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->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) { + 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) @@ -5183,16 +5463,8 @@ static void gfx_v9_0_ring_patch_ce_meta(struct amdgpu_ring *ring, payload_size = sizeof(struct v9_ce_ib_state); - if (ring->is_mes_queue) { - payload_offset = offsetof(struct amdgpu_mes_ctx_meta_data, - gfx[0].gfx_meta_data) + - offsetof(struct v9_gfx_meta_data, ce_payload); - ce_payload_cpu_addr = - amdgpu_mes_ctx_get_offs_cpu_addr(ring, payload_offset); - } else { - payload_offset = offsetof(struct v9_gfx_meta_data, ce_payload); - ce_payload_cpu_addr = adev->virt.csa_cpu_addr + payload_offset; - } + 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); @@ -5215,16 +5487,8 @@ static void gfx_v9_0_ring_patch_de_meta(struct amdgpu_ring *ring, payload_size = sizeof(struct v9_de_ib_state); - if (ring->is_mes_queue) { - payload_offset = offsetof(struct amdgpu_mes_ctx_meta_data, - gfx[0].gfx_meta_data) + - offsetof(struct v9_gfx_meta_data, de_payload); - de_payload_cpu_addr = - amdgpu_mes_ctx_get_offs_cpu_addr(ring, payload_offset); - } else { - payload_offset = offsetof(struct v9_gfx_meta_data, de_payload); - de_payload_cpu_addr = adev->virt.csa_cpu_addr + payload_offset; - } + 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; @@ -5414,19 +5678,9 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume) cnt = (sizeof(ce_payload) >> 2) + 4 - 2; - if (ring->is_mes_queue) { - offset = offsetof(struct amdgpu_mes_ctx_meta_data, - gfx[0].gfx_meta_data) + - offsetof(struct v9_gfx_meta_data, ce_payload); - ce_payload_gpu_addr = - amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); - ce_payload_cpu_addr = - amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); - } else { - 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; - } + 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) | @@ -5512,28 +5766,13 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume, bo void *de_payload_cpu_addr; int cnt; - if (ring->is_mes_queue) { - offset = offsetof(struct amdgpu_mes_ctx_meta_data, - gfx[0].gfx_meta_data) + - offsetof(struct v9_gfx_meta_data, de_payload); - de_payload_gpu_addr = - amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); - de_payload_cpu_addr = - amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); - - offset = offsetof(struct amdgpu_mes_ctx_meta_data, - gfx[0].gds_backup) + - offsetof(struct v9_gfx_meta_data, de_payload); - gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); - } else { - 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; + 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); - } + 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); @@ -5600,31 +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 - 1) & ring->buf_mask; - 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) { @@ -5697,7 +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, @@ -5768,17 +5999,95 @@ static void gfx_v9_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, } } +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; @@ -5899,11 +6208,14 @@ static int gfx_v9_0_eop_irq(struct amdgpu_device *adev, switch (me_id) { case 0: - if (adev->gfx.num_gfx_rings && - !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]); + 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: @@ -5957,6 +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) @@ -6454,7 +6775,7 @@ static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev, 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", @@ -6473,7 +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", @@ -6483,7 +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", @@ -6837,6 +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, @@ -6853,6 +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 = { @@ -6884,7 +7431,8 @@ 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, @@ -6893,12 +7441,11 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .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 = 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, .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, @@ -6906,6 +7453,9 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .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 = { @@ -6938,7 +7488,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_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, @@ -6948,12 +7499,11 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_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_sw_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, @@ -6963,6 +7513,9 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = { .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 = { @@ -6980,11 +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, @@ -6994,13 +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 = { @@ -7018,7 +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, @@ -7029,6 +7586,7 @@ 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) @@ -7040,7 +7598,7 @@ static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev) 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.num_gfx_rings) { + 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; } @@ -7059,6 +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, @@ -7078,6 +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; @@ -7087,7 +7653,7 @@ 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->ip_versions[GC_HWIP][0]) { + 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): @@ -7106,7 +7672,7 @@ 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->ip_versions[GC_HWIP][0]) { + 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): @@ -7128,7 +7694,7 @@ static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev) break; } - switch (adev->ip_versions[GC_HWIP][0]) { + 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; @@ -7233,7 +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) { |