diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c')
-rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c | 85 |
1 files changed, 74 insertions, 11 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c index 697b6d530d12..97933d2a3803 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c @@ -77,7 +77,9 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd, m->compute_static_thread_mgmt_se1 = se_mask[1]; m->compute_static_thread_mgmt_se2 = se_mask[2]; m->compute_static_thread_mgmt_se3 = se_mask[3]; - if (KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 3)) { + if (KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 3) && + KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 4) && + KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 5, 0)) { m->compute_static_thread_mgmt_se4 = se_mask[4]; m->compute_static_thread_mgmt_se5 = se_mask[5]; m->compute_static_thread_mgmt_se6 = se_mask[6]; @@ -181,6 +183,9 @@ static void init_mqd(struct mqd_manager *mm, void **mqd, m->cp_hqd_persistent_state = CP_HQD_PERSISTENT_STATE__PRELOAD_REQ_MASK | 0x53 << CP_HQD_PERSISTENT_STATE__PRELOAD_SIZE__SHIFT; + m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT; + m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__UNORD_DISPATCH_MASK; + m->cp_mqd_control = 1 << CP_MQD_CONTROL__PRIV_STATE__SHIFT; m->cp_mqd_base_addr_lo = lower_32_bits(addr); @@ -243,7 +248,7 @@ static void update_mqd(struct mqd_manager *mm, void *mqd, m = get_mqd(mqd); - m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT; + m->cp_hqd_pq_control &= ~CP_HQD_PQ_CONTROL__QUEUE_SIZE_MASK; m->cp_hqd_pq_control |= order_base_2(q->queue_size / 4) - 1; pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control); @@ -299,7 +304,9 @@ static void update_mqd(struct mqd_manager *mm, void *mqd, if (mm->dev->kfd->cwsr_enabled && q->ctx_save_restore_area_address) m->cp_hqd_ctx_save_control = 0; - if (KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 3)) + if (KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 3) && + KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 4) && + KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 5, 0)) update_cu_mask(mm, mqd, minfo, 0); set_priority(m, q); @@ -316,11 +323,14 @@ static void update_mqd(struct mqd_manager *mm, void *mqd, } -static uint32_t read_doorbell_id(void *mqd) +static bool check_preemption_failed(struct mqd_manager *mm, void *mqd) { struct v9_mqd *m = (struct v9_mqd *)mqd; + uint32_t doorbell_id = m->queue_doorbell_id0; + + m->queue_doorbell_id0 = 0; - return m->queue_doorbell_id0; + return kfd_check_hiq_mqd_doorbell_id(mm->dev, doorbell_id, 0); } static int get_wave_state(struct mqd_manager *mm, void *mqd, @@ -485,6 +495,10 @@ static void update_mqd_sdma(struct mqd_manager *mm, void *mqd, m->sdma_engine_id = q->sdma_engine_id; m->sdma_queue_id = q->sdma_queue_id; m->sdmax_rlcx_dummy_reg = SDMA_RLC_DUMMY_DEFAULT; + /* Allow context switch so we don't cross-process starve with a massive + * command buffer of long-running SDMA commands + */ + m->sdmax_rlcx_ib_cntl |= SDMA0_GFX_IB_CNTL__SWITCH_INSIDE_IB_MASK; q->is_active = QUEUE_IS_ACTIVE(*q); } @@ -544,6 +558,9 @@ static void init_mqd_hiq_v9_4_3(struct mqd_manager *mm, void **mqd, m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK | 1 << CP_HQD_PQ_CONTROL__PRIV_STATE__SHIFT | 1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT; + if (amdgpu_sriov_multi_vf_mode(mm->dev->adev)) + m->cp_hqd_pq_doorbell_control |= 1 << + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_MODE__SHIFT; m->cp_mqd_stride_size = kfd_hiq_mqd_stride(mm->dev); if (xcc == 0) { /* Set no_update_rptr = 0 in Master XCC */ @@ -607,6 +624,25 @@ static int destroy_hiq_mqd_v9_4_3(struct mqd_manager *mm, void *mqd, return err; } +static bool check_preemption_failed_v9_4_3(struct mqd_manager *mm, void *mqd) +{ + uint64_t hiq_mqd_size = kfd_hiq_mqd_stride(mm->dev); + uint32_t xcc_mask = mm->dev->xcc_mask; + int inst = 0, xcc_id; + struct v9_mqd *m; + bool ret = false; + + for_each_inst(xcc_id, xcc_mask) { + m = get_mqd(mqd + hiq_mqd_size * inst); + ret |= kfd_check_hiq_mqd_doorbell_id(mm->dev, + m->queue_doorbell_id0, inst); + m->queue_doorbell_id0 = 0; + ++inst; + } + + return ret; +} + static void get_xcc_mqd(struct kfd_mem_obj *mqd_mem_obj, struct kfd_mem_obj *xcc_mqd_mem_obj, uint64_t offset) @@ -635,7 +671,9 @@ static void init_mqd_v9_4_3(struct mqd_manager *mm, void **mqd, get_xcc_mqd(mqd_mem_obj, &xcc_mqd_mem_obj, offset*xcc); init_mqd(mm, (void **)&m, &xcc_mqd_mem_obj, &xcc_gart_addr, q); - + if (amdgpu_sriov_multi_vf_mode(mm->dev->adev)) + m->cp_hqd_pq_doorbell_control |= 1 << + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_MODE__SHIFT; m->cp_mqd_stride_size = offset; /* @@ -695,7 +733,10 @@ static void update_mqd_v9_4_3(struct mqd_manager *mm, void *mqd, m = get_mqd(mqd + size * xcc); update_mqd(mm, m, q, minfo); - update_cu_mask(mm, mqd, minfo, xcc); + if (amdgpu_sriov_multi_vf_mode(mm->dev->adev)) + m->cp_hqd_pq_doorbell_control |= 1 << + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_MODE__SHIFT; + update_cu_mask(mm, m, minfo, xcc); if (q->format == KFD_QUEUE_FORMAT_AQL) { switch (xcc) { @@ -717,6 +758,21 @@ static void update_mqd_v9_4_3(struct mqd_manager *mm, void *mqd, } } +static void restore_mqd_v9_4_3(struct mqd_manager *mm, void **mqd, + struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr, + struct queue_properties *qp, + const void *mqd_src, + const void *ctl_stack_src, u32 ctl_stack_size) +{ + restore_mqd(mm, mqd, mqd_mem_obj, gart_addr, qp, mqd_src, ctl_stack_src, ctl_stack_size); + if (amdgpu_sriov_multi_vf_mode(mm->dev->adev)) { + struct v9_mqd *m; + + m = (struct v9_mqd *) mqd_mem_obj->cpu_ptr; + m->cp_hqd_pq_doorbell_control |= 1 << + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_MODE__SHIFT; + } +} static int destroy_mqd_v9_4_3(struct mqd_manager *mm, void *mqd, enum kfd_preempt_type type, unsigned int timeout, uint32_t pipe_id, uint32_t queue_id) @@ -851,22 +907,25 @@ struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type, mqd->is_occupied = kfd_is_occupied_cp; mqd->get_checkpoint_info = get_checkpoint_info; mqd->checkpoint_mqd = checkpoint_mqd; - mqd->restore_mqd = restore_mqd; mqd->mqd_size = sizeof(struct v9_mqd); mqd->mqd_stride = mqd_stride_v9; #if defined(CONFIG_DEBUG_FS) mqd->debugfs_show_mqd = debugfs_show_mqd; #endif - if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3)) { + if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3) || + KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 4) || + KFD_GC_VERSION(dev) == IP_VERSION(9, 5, 0)) { mqd->init_mqd = init_mqd_v9_4_3; mqd->load_mqd = load_mqd_v9_4_3; mqd->update_mqd = update_mqd_v9_4_3; + mqd->restore_mqd = restore_mqd_v9_4_3; mqd->destroy_mqd = destroy_mqd_v9_4_3; mqd->get_wave_state = get_wave_state_v9_4_3; } else { mqd->init_mqd = init_mqd; mqd->load_mqd = load_mqd; mqd->update_mqd = update_mqd; + mqd->restore_mqd = restore_mqd; mqd->destroy_mqd = kfd_destroy_mqd_cp; mqd->get_wave_state = get_wave_state; } @@ -881,15 +940,19 @@ struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type, #if defined(CONFIG_DEBUG_FS) mqd->debugfs_show_mqd = debugfs_show_mqd; #endif - mqd->read_doorbell_id = read_doorbell_id; - if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3)) { + mqd->check_preemption_failed = check_preemption_failed; + if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3) || + KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 4) || + KFD_GC_VERSION(dev) == IP_VERSION(9, 5, 0)) { mqd->init_mqd = init_mqd_hiq_v9_4_3; mqd->load_mqd = hiq_load_mqd_kiq_v9_4_3; mqd->destroy_mqd = destroy_hiq_mqd_v9_4_3; + mqd->check_preemption_failed = check_preemption_failed_v9_4_3; } else { mqd->init_mqd = init_mqd_hiq; mqd->load_mqd = kfd_hiq_load_mqd_kiq; mqd->destroy_mqd = destroy_hiq_mqd; + mqd->check_preemption_failed = check_preemption_failed; } break; case KFD_MQD_TYPE_DIQ: |