diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 694 |
1 files changed, 676 insertions, 18 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c index 20ea6cb01edf..408e5600bb61 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c @@ -37,6 +37,7 @@ #include "gc/gc_9_4_3_sh_mask.h" #include "gfx_v9_4_3.h" +#include "gfx_v9_4_3_cleaner_shader.h" #include "amdgpu_xcp.h" #include "amdgpu_aca.h" @@ -63,6 +64,98 @@ MODULE_FIRMWARE("amdgpu/gc_9_4_4_rlc.bin"); #define NORMALIZE_XCC_REG_OFFSET(offset) \ (offset & 0xFFFF) +static const struct amdgpu_hwip_reg_entry gc_reg_list_9_4_3[] = { + SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2), + SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT1), + SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT2), + SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STALLED_STAT1), + SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STALLED_STAT1), + SOC15_REG_ENTRY_STR(GC, 0, regCP_BUSY_STAT), + SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT), + SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT), + SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_ERROR), + SOC15_REG_ENTRY_STR(GC, 0, regCPF_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regCPC_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regCPG_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regGDS_PROTECTION_FAULT), + SOC15_REG_ENTRY_STR(GC, 0, regGDS_VM_PROTECTION_FAULT), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regRMI_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regSQC_DCACHE_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regSQC_ICACHE_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regSQ_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regTCP_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regWD_UTCL1_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, regVM_L2_PROTECTION_FAULT_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regCP_DEBUG), + SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC1_INSTR_PNTR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC2_INSTR_PNTR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_STAT), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_COMMAND), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_MESSAGE), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_ARGUMENT_1), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_ARGUMENT_2), + SOC15_REG_ENTRY_STR(GC, 0, regSMU_RLC_RESPONSE), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_SAFE_MODE), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_SAFE_MODE), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_INT_STAT), + SOC15_REG_ENTRY_STR(GC, 0, regRLC_GPM_GENERAL_6), + /* cp header registers */ + SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP), + SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME2_HEADER_DUMP), + /* SE status registers */ + SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE0), + SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE1), + SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE2), + SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE3) +}; + +static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_9_4_3[] = { + /* compute queue registers */ + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_VMID), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ACTIVE), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PERSISTENT_STATE), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PIPE_PRIORITY), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUEUE_PRIORITY), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUANTUM), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_REQUEST), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_EVENTS), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_LO), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_CONTROL), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_OFFSET), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_SIZE), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_WG_STATE_OFFSET), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_SIZE), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GDS_RESOURCE_STATE), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ERROR), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR_MEM), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_LO), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GFX_STATUS), +}; + struct amdgpu_gfx_ras gfx_v9_4_3_ras; static void gfx_v9_4_3_set_ring_funcs(struct amdgpu_device *adev); @@ -71,10 +164,18 @@ static void gfx_v9_4_3_set_gds_init(struct amdgpu_device *adev); static void gfx_v9_4_3_set_rlc_funcs(struct amdgpu_device *adev); static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev, struct amdgpu_cu_info *cu_info); +static void gfx_v9_4_3_xcc_set_safe_mode(struct amdgpu_device *adev, int xcc_id); +static void gfx_v9_4_3_xcc_unset_safe_mode(struct amdgpu_device *adev, int xcc_id); static void gfx_v9_4_3_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) | @@ -84,8 +185,8 @@ static void gfx_v9_4_3_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 */ } @@ -182,12 +283,46 @@ static void gfx_v9_4_3_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring, PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type)); } +static void gfx_v9_4_3_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, xcc_id); + + if (queue_type == AMDGPU_RING_TYPE_COMPUTE) { + WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 0x2); + WREG32_SOC15(GC, GET_INST(GC, xcc_id), regSPI_COMPUTE_QUEUE_RESET, 0x1); + /* wait till dequeue take effects */ + for (i = 0; i < adev->usec_timeout; i++) { + if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_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\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_4_3_kiq_pm4_funcs = { .kiq_set_resources = gfx_v9_4_3_kiq_set_resources, .kiq_map_queues = gfx_v9_4_3_kiq_map_queues, .kiq_unmap_queues = gfx_v9_4_3_kiq_unmap_queues, .kiq_query_status = gfx_v9_4_3_kiq_query_status, .kiq_invalidate_tlbs = gfx_v9_4_3_kiq_invalidate_tlbs, + .kiq_reset_hw_queue = gfx_v9_4_3_kiq_reset_hw_queue, .set_resources_size = 8, .map_queues_size = 7, .unmap_queues_size = 6, @@ -885,11 +1020,59 @@ static int gfx_v9_4_3_compute_ring_init(struct amdgpu_device *adev, int ring_id, hw_prio, NULL); } +static void gfx_v9_4_3_alloc_ip_dump(struct amdgpu_device *adev) +{ + uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3); + uint32_t *ptr, num_xcc, inst; + + num_xcc = NUM_XCC(adev->gfx.xcc_mask); + + ptr = kcalloc(reg_count * num_xcc, 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_4_3); + 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 * num_xcc, 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_4_3_sw_init(void *handle) { int i, j, k, r, ring_id, xcc_id, num_xcc; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 4, 3): + case IP_VERSION(9, 4, 4): + adev->gfx.cleaner_shader_ptr = gfx_9_4_3_cleaner_shader_hex; + adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_3_cleaner_shader_hex); + if (adev->gfx.mec_fw_version >= 153) { + 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_mec = 2; adev->gfx.mec.num_pipe_per_mec = 4; adev->gfx.mec.num_queue_per_pipe = 8; @@ -901,6 +1084,13 @@ static int gfx_v9_4_3_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); @@ -976,10 +1166,19 @@ static int gfx_v9_4_3_sw_init(void *handle) return r; - if (!amdgpu_sriov_vf(adev)) + if (!amdgpu_sriov_vf(adev)) { r = amdgpu_gfx_sysfs_init(adev); + if (r) + return r; + } - return r; + gfx_v9_4_3_alloc_ip_dump(adev); + + r = amdgpu_gfx_sysfs_isolation_shader_init(adev); + if (r) + return r; + + return 0; } static int gfx_v9_4_3_sw_fini(void *handle) @@ -997,11 +1196,17 @@ static int gfx_v9_4_3_sw_fini(void *handle) amdgpu_gfx_kiq_fini(adev, i); } + amdgpu_gfx_cleaner_shader_sw_fini(adev); + gfx_v9_4_3_mec_fini(adev); amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj); gfx_v9_4_3_free_microcode(adev); if (!amdgpu_sriov_vf(adev)) amdgpu_gfx_sysfs_fini(adev); + amdgpu_gfx_sysfs_isolation_shader_fini(adev); + + kfree(adev->gfx.ip_dump_core); + kfree(adev->gfx.ip_dump_compute_queues); return 0; } @@ -1910,7 +2115,7 @@ static int gfx_v9_4_3_xcc_kiq_init_queue(struct amdgpu_ring *ring, int xcc_id) return 0; } -static int gfx_v9_4_3_xcc_kcq_init_queue(struct amdgpu_ring *ring, int xcc_id) +static int gfx_v9_4_3_xcc_kcq_init_queue(struct amdgpu_ring *ring, int xcc_id, bool restore) { struct amdgpu_device *adev = ring->adev; struct v9_mqd *mqd = ring->mqd_ptr; @@ -1922,8 +2127,8 @@ static int gfx_v9_4_3_xcc_kcq_init_queue(struct amdgpu_ring *ring, int xcc_id) */ 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; @@ -2008,7 +2213,7 @@ static int gfx_v9_4_3_xcc_kcq_resume(struct amdgpu_device *adev, int xcc_id) goto done; r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr); if (!r) { - r = gfx_v9_4_3_xcc_kcq_init_queue(ring, xcc_id); + r = gfx_v9_4_3_xcc_kcq_init_queue(ring, xcc_id, false); amdgpu_bo_kunmap(ring->mqd_obj); ring->mqd_ptr = NULL; } @@ -2139,6 +2344,9 @@ static int gfx_v9_4_3_hw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size, + adev->gfx.cleaner_shader_ptr); + if (!amdgpu_sriov_vf(adev)) gfx_v9_4_3_init_golden_registers(adev); @@ -2162,6 +2370,7 @@ static int gfx_v9_4_3_hw_fini(void *handle) 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); num_xcc = NUM_XCC(adev->gfx.xcc_mask); for (i = 0; i < num_xcc; i++) { @@ -2327,6 +2536,10 @@ static int gfx_v9_4_3_late_init(void *handle) if (r) return r; + r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0); + if (r) + return r; + if (adev->gfx.ras && adev->gfx.ras->enable_watchdog_timer) adev->gfx.ras->enable_watchdog_timer(adev); @@ -2833,6 +3046,24 @@ static void gfx_v9_4_3_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring, ref, mask); } +static void gfx_v9_4_3_ring_soft_recovery(struct amdgpu_ring *ring, + unsigned vmid) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t value = 0; + + if (!adev->debug_exp_resets) + return; + + value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03); + 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, ring->xcc_id); + WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regSQ_CMD, value); + amdgpu_gfx_rlc_exit_safe_mode(adev, ring->xcc_id); +} + static void gfx_v9_4_3_xcc_set_compute_eop_interrupt_state( struct amdgpu_device *adev, int me, int pipe, enum amdgpu_interrupt_state state, int xcc_id) @@ -2886,21 +3117,103 @@ static void gfx_v9_4_3_xcc_set_compute_eop_interrupt_state( } } +static u32 gfx_v9_4_3_get_cpc_int_cntl(struct amdgpu_device *adev, + int xcc_id, 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, GET_INST(GC, xcc_id), regCP_ME1_PIPE0_INT_CNTL); + case 1: + return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE1_INT_CNTL); + case 2: + return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE2_INT_CNTL); + case 3: + return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE3_INT_CNTL); + default: + return 0; + } +} + static int gfx_v9_4_3_set_priv_reg_fault_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, unsigned type, enum amdgpu_interrupt_state state) { - int i, num_xcc; + u32 mec_int_cntl_reg, mec_int_cntl; + int i, j, k, num_xcc; num_xcc = NUM_XCC(adev->gfx.xcc_mask); switch (state) { case AMDGPU_IRQ_STATE_DISABLE: case AMDGPU_IRQ_STATE_ENABLE: - for (i = 0; i < num_xcc; i++) + for (i = 0; i < num_xcc; i++) { WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0, - PRIV_REG_INT_ENABLE, - state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + PRIV_REG_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + for (j = 0; j < adev->gfx.mec.num_mec; j++) { + for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) { + /* MECs start at 1 */ + mec_int_cntl_reg = gfx_v9_4_3_get_cpc_int_cntl(adev, i, j + 1, k); + + if (mec_int_cntl_reg) { + mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, i); + mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, + PRIV_REG_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? + 1 : 0); + WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, i); + } + } + } + } + break; + default: + break; + } + + return 0; +} + +static int gfx_v9_4_3_set_bad_op_fault_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + u32 mec_int_cntl_reg, mec_int_cntl; + int i, j, k, num_xcc; + + num_xcc = NUM_XCC(adev->gfx.xcc_mask); + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + case AMDGPU_IRQ_STATE_ENABLE: + for (i = 0; i < num_xcc; i++) { + WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0, + OPCODE_ERROR_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + for (j = 0; j < adev->gfx.mec.num_mec; j++) { + for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) { + /* MECs start at 1 */ + mec_int_cntl_reg = gfx_v9_4_3_get_cpc_int_cntl(adev, i, j + 1, k); + + if (mec_int_cntl_reg) { + mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, i); + mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, + OPCODE_ERROR_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? + 1 : 0); + WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, i); + } + } + } + } break; default: break; @@ -3061,6 +3374,15 @@ static int gfx_v9_4_3_priv_reg_irq(struct amdgpu_device *adev, return 0; } +static int gfx_v9_4_3_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_4_3_fault(adev, entry); + return 0; +} + static int gfx_v9_4_3_priv_inst_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -3147,6 +3469,183 @@ static void gfx_v9_4_3_emit_wave_limit(struct amdgpu_ring *ring, bool enable) } } +static int gfx_v9_4_3_unmap_done(struct amdgpu_device *adev, uint32_t me, + uint32_t pipe, uint32_t queue, + uint32_t xcc_id) +{ + int i, r; + /* make sure dequeue is complete*/ + gfx_v9_4_3_xcc_set_safe_mode(adev, xcc_id); + mutex_lock(&adev->srbm_mutex); + soc15_grbm_select(adev, me, pipe, queue, 0, GET_INST(GC, xcc_id)); + for (i = 0; i < adev->usec_timeout; i++) { + if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1)) + break; + udelay(1); + } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + else + r = 0; + soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id)); + mutex_unlock(&adev->srbm_mutex); + gfx_v9_4_3_xcc_unset_safe_mode(adev, xcc_id); + + return r; + +} + +static bool gfx_v9_4_3_pipe_reset_support(struct amdgpu_device *adev) +{ + /*TODO: Need check gfx9.4.4 mec fw whether supports pipe reset as well.*/ + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) && + adev->gfx.mec_fw_version >= 0x0000009b) + return true; + else + dev_warn_once(adev->dev, "Please use the latest MEC version to see whether support pipe reset\n"); + + return false; +} + +static int gfx_v9_4_3_reset_hw_pipe(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t reset_pipe, clean_pipe; + int r; + + if (!gfx_v9_4_3_pipe_reset_support(adev)) + return -EINVAL; + + gfx_v9_4_3_xcc_set_safe_mode(adev, ring->xcc_id); + mutex_lock(&adev->srbm_mutex); + + reset_pipe = RREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL); + clean_pipe = reset_pipe; + + if (ring->me == 1) { + switch (ring->pipe) { + case 0: + reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL, + MEC_ME1_PIPE0_RESET, 1); + break; + case 1: + reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL, + MEC_ME1_PIPE1_RESET, 1); + break; + case 2: + reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL, + MEC_ME1_PIPE2_RESET, 1); + break; + case 3: + reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL, + MEC_ME1_PIPE3_RESET, 1); + break; + default: + break; + } + } else { + if (ring->pipe) + reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL, + MEC_ME2_PIPE1_RESET, 1); + else + reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL, + MEC_ME2_PIPE0_RESET, 1); + } + + WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL, reset_pipe); + WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL, clean_pipe); + mutex_unlock(&adev->srbm_mutex); + gfx_v9_4_3_xcc_unset_safe_mode(adev, ring->xcc_id); + + r = gfx_v9_4_3_unmap_done(adev, ring->me, ring->pipe, ring->queue, ring->xcc_id); + return r; +} + +static int gfx_v9_4_3_reset_kcq(struct amdgpu_ring *ring, + unsigned int vmid) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_kiq *kiq = &adev->gfx.kiq[ring->xcc_id]; + struct amdgpu_ring *kiq_ring = &kiq->ring; + unsigned long flags; + int r; + + if (!adev->debug_exp_resets) + return -EINVAL; + + if (amdgpu_sriov_vf(adev)) + return -EINVAL; + + 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; + } + + 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) { + dev_err(adev->dev, "kiq ring test failed after ring: %s queue reset\n", + ring->name); + goto pipe_reset; + } + + r = gfx_v9_4_3_unmap_done(adev, ring->me, ring->pipe, ring->queue, ring->xcc_id); + if (r) + dev_err(adev->dev, "fail to wait on hqd deactive and will try pipe reset\n"); + +pipe_reset: + if(r) { + r = gfx_v9_4_3_reset_hw_pipe(ring); + dev_info(adev->dev, "ring: %s pipe reset :%s\n", ring->name, + r ? "failed" : "successfully"); + if (r) + return r; + } + + r = amdgpu_bo_reserve(ring->mqd_obj, false); + if (unlikely(r != 0)){ + dev_err(adev->dev, "fail to resv mqd_obj\n"); + return r; + } + r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr); + if (!r) { + r = gfx_v9_4_3_xcc_kcq_init_queue(ring, ring->xcc_id, true); + amdgpu_bo_kunmap(ring->mqd_obj); + ring->mqd_ptr = NULL; + } + amdgpu_bo_unreserve(ring->mqd_obj); + if (r) { + dev_err(adev->dev, "fail to unresv mqd_obj\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); + spin_unlock_irqrestore(&kiq->ring_lock, flags); + + r = amdgpu_ring_test_ring(kiq_ring); + if (r) { + dev_err(adev->dev, "fail to remap queue\n"); + return r; + } + return amdgpu_ring_test_ring(ring); +} + enum amdgpu_gfx_cp_ras_mem_id { AMDGPU_GFX_CP_MEM1 = 1, AMDGPU_GFX_CP_MEM2, @@ -3959,8 +4458,8 @@ static void gfx_v9_4_3_inst_query_ras_err_count(struct amdgpu_device *adev, /* the caller should make sure initialize value of * err_data->ue_count and err_data->ce_count */ - amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, NULL, ue_count); - amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, NULL, ce_count); + amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count); + amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, ce_count); } static void gfx_v9_4_3_inst_reset_ras_err_count(struct amdgpu_device *adev, @@ -4062,6 +4561,151 @@ static void gfx_v9_4_3_enable_watchdog_timer(struct amdgpu_device *adev) amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_enable_watchdog_timer); } +static void gfx_v9_4_3_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop) +{ + int i; + + /* 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 */ + for (i = 1; i < num_nop; i++) + amdgpu_ring_write(ring, ring->funcs->nop); +} + +static void gfx_v9_4_3_ip_print(void *handle, struct drm_printer *p) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + uint32_t i, j, k; + uint32_t xcc_id, xcc_offset, inst_offset; + uint32_t num_xcc, reg, num_inst; + uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3); + + if (!adev->gfx.ip_dump_core) + return; + + num_xcc = NUM_XCC(adev->gfx.xcc_mask); + drm_printf(p, "Number of Instances:%d\n", num_xcc); + for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) { + xcc_offset = xcc_id * reg_count; + drm_printf(p, "\nInstance id:%d\n", xcc_id); + for (i = 0; i < reg_count; i++) + drm_printf(p, "%-50s \t 0x%08x\n", + gc_reg_list_9_4_3[i].reg_name, + adev->gfx.ip_dump_core[xcc_offset + i]); + } + + /* print compute queue registers for all instances */ + if (!adev->gfx.ip_dump_compute_queues) + return; + + num_inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec * + adev->gfx.mec.num_queue_per_pipe; + + reg_count = ARRAY_SIZE(gc_cp_reg_list_9_4_3); + drm_printf(p, "\nnum_xcc: %d num_mec: %d num_pipe: %d num_queue: %d\n", + num_xcc, + adev->gfx.mec.num_mec, + adev->gfx.mec.num_pipe_per_mec, + adev->gfx.mec.num_queue_per_pipe); + + for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) { + xcc_offset = xcc_id * reg_count * num_inst; + inst_offset = 0; + 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, + "\nxcc:%d mec:%d, pipe:%d, queue:%d\n", + xcc_id, i, j, k); + for (reg = 0; reg < reg_count; reg++) { + drm_printf(p, + "%-50s \t 0x%08x\n", + gc_cp_reg_list_9_4_3[reg].reg_name, + adev->gfx.ip_dump_compute_queues + [xcc_offset + inst_offset + + reg]); + } + inst_offset += reg_count; + } + } + } + } +} + +static void gfx_v9_4_3_ip_dump(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + uint32_t i, j, k; + uint32_t num_xcc, reg, num_inst; + uint32_t xcc_id, xcc_offset, inst_offset; + uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3); + + if (!adev->gfx.ip_dump_core) + return; + + num_xcc = NUM_XCC(adev->gfx.xcc_mask); + + amdgpu_gfx_off_ctrl(adev, false); + for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) { + xcc_offset = xcc_id * reg_count; + for (i = 0; i < reg_count; i++) + adev->gfx.ip_dump_core[xcc_offset + i] = + RREG32(SOC15_REG_ENTRY_OFFSET_INST(gc_reg_list_9_4_3[i], + GET_INST(GC, xcc_id))); + } + amdgpu_gfx_off_ctrl(adev, true); + + /* dump compute queue registers for all instances */ + if (!adev->gfx.ip_dump_compute_queues) + return; + + num_inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec * + adev->gfx.mec.num_queue_per_pipe; + reg_count = ARRAY_SIZE(gc_cp_reg_list_9_4_3); + amdgpu_gfx_off_ctrl(adev, false); + mutex_lock(&adev->srbm_mutex); + for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) { + xcc_offset = xcc_id * reg_count * num_inst; + inst_offset = 0; + 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, + GET_INST(GC, xcc_id)); + + for (reg = 0; reg < reg_count; reg++) { + adev->gfx.ip_dump_compute_queues + [xcc_offset + + inst_offset + reg] = + RREG32(SOC15_REG_ENTRY_OFFSET_INST( + gc_cp_reg_list_9_4_3[reg], + GET_INST(GC, xcc_id))); + } + inst_offset += 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_4_3_ring_emit_cleaner_shader(struct amdgpu_ring *ring) +{ + /* Emit the cleaner shader */ + amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0)); + amdgpu_ring_write(ring, 0); /* RESERVED field, programmed to zero */ +} + static const struct amd_ip_funcs gfx_v9_4_3_ip_funcs = { .name = "gfx_v9_4_3", .early_init = gfx_v9_4_3_early_init, @@ -4078,8 +4722,8 @@ static const struct amd_ip_funcs gfx_v9_4_3_ip_funcs = { .set_clockgating_state = gfx_v9_4_3_set_clockgating_state, .set_powergating_state = gfx_v9_4_3_set_powergating_state, .get_clockgating_state = gfx_v9_4_3_get_clockgating_state, - .dump_ip_state = NULL, - .print_ip_state = NULL, + .dump_ip_state = gfx_v9_4_3_ip_dump, + .print_ip_state = gfx_v9_4_3_ip_print, }; static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_compute = { @@ -4101,7 +4745,8 @@ static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_compute = { 8 + 8 + 8 + /* gfx_v9_4_3_ring_emit_fence x3 for user fence, vm fence */ 7 + /* gfx_v9_4_3_emit_mem_sync */ 5 + /* gfx_v9_4_3_emit_wave_limit for updating regSPI_WCL_PIPE_PERCENT_GFX register */ - 15, /* for updating 3 regSPI_WCL_PIPE_PERCENT_CS registers */ + 15 + /* for updating 3 regSPI_WCL_PIPE_PERCENT_CS registers */ + 2, /* gfx_v9_4_3_ring_emit_cleaner_shader */ .emit_ib_size = 7, /* gfx_v9_4_3_ring_emit_ib_compute */ .emit_ib = gfx_v9_4_3_ring_emit_ib_compute, .emit_fence = gfx_v9_4_3_ring_emit_fence, @@ -4111,13 +4756,18 @@ static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_compute = { .emit_hdp_flush = gfx_v9_4_3_ring_emit_hdp_flush, .test_ring = gfx_v9_4_3_ring_test_ring, .test_ib = gfx_v9_4_3_ring_test_ib, - .insert_nop = amdgpu_ring_insert_nop, + .insert_nop = gfx_v9_4_3_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .emit_wreg = gfx_v9_4_3_ring_emit_wreg, .emit_reg_wait = gfx_v9_4_3_ring_emit_reg_wait, .emit_reg_write_reg_wait = gfx_v9_4_3_ring_emit_reg_write_reg_wait, + .soft_recovery = gfx_v9_4_3_ring_soft_recovery, .emit_mem_sync = gfx_v9_4_3_emit_mem_sync, .emit_wave_limit = gfx_v9_4_3_emit_wave_limit, + .reset = gfx_v9_4_3_reset_kcq, + .emit_cleaner_shader = gfx_v9_4_3_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_4_3_ring_funcs_kiq = { @@ -4172,6 +4822,11 @@ static const struct amdgpu_irq_src_funcs gfx_v9_4_3_priv_reg_irq_funcs = { .process = gfx_v9_4_3_priv_reg_irq, }; +static const struct amdgpu_irq_src_funcs gfx_v9_4_3_bad_op_irq_funcs = { + .set = gfx_v9_4_3_set_bad_op_fault_state, + .process = gfx_v9_4_3_bad_op_irq, +}; + static const struct amdgpu_irq_src_funcs gfx_v9_4_3_priv_inst_irq_funcs = { .set = gfx_v9_4_3_set_priv_inst_fault_state, .process = gfx_v9_4_3_priv_inst_irq, @@ -4185,6 +4840,9 @@ static void gfx_v9_4_3_set_irq_funcs(struct amdgpu_device *adev) adev->gfx.priv_reg_irq.num_types = 1; adev->gfx.priv_reg_irq.funcs = &gfx_v9_4_3_priv_reg_irq_funcs; + adev->gfx.bad_op_irq.num_types = 1; + adev->gfx.bad_op_irq.funcs = &gfx_v9_4_3_bad_op_irq_funcs; + adev->gfx.priv_inst_irq.num_types = 1; adev->gfx.priv_inst_irq.funcs = &gfx_v9_4_3_priv_inst_irq_funcs; } |