diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c | 67 |
1 files changed, 54 insertions, 13 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c index d3e2b6a599a4..1695dd78ede8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c @@ -48,12 +48,11 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd, struct v10_compute_mqd *m; uint32_t se_mask[4] = {0}; /* 4 is the max # of SEs */ - if (!minfo || (minfo->update_flag != UPDATE_FLAG_CU_MASK) || - !minfo->cu_mask.ptr) + if (!minfo || !minfo->cu_mask.ptr) return; mqd_symmetrically_map_cu_mask(mm, - minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask); + minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0); m = get_mqd(mqd); m->compute_static_thread_mgmt_se0 = se_mask[0]; @@ -74,7 +73,7 @@ static void set_priority(struct v10_compute_mqd *m, struct queue_properties *q) m->cp_hqd_queue_priority = q->priority; } -static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd, +static struct kfd_mem_obj *allocate_mqd(struct kfd_node *kfd, struct queue_properties *q) { struct kfd_mem_obj *mqd_mem_obj; @@ -108,6 +107,8 @@ 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); @@ -117,12 +118,17 @@ static void init_mqd(struct mqd_manager *mm, void **mqd, 1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT | 1 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT; + /* Set cp_hqd_hq_scheduler0 bit 14 to 1 to have the CP set up the + * DISPATCH_PTR. This is required for the kfd debugger + */ + m->cp_hqd_hq_scheduler0 = 1 << 14; + if (q->format == KFD_QUEUE_FORMAT_AQL) { m->cp_hqd_aql_control = 1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT; } - if (mm->dev->cwsr_enabled) { + if (mm->dev->kfd->cwsr_enabled) { m->cp_hqd_persistent_state |= (1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT); m->cp_hqd_ctx_save_base_addr_lo = @@ -151,7 +157,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, r = mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id, (uint32_t __user *)p->write_ptr, - wptr_shift, 0, mms); + wptr_shift, 0, mms, 0); return r; } @@ -163,9 +169,10 @@ 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 |= ffs(q->queue_size / sizeof(unsigned int)) - 1 - 1; + pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control); m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); @@ -210,7 +217,7 @@ static void update_mqd(struct mqd_manager *mm, void *mqd, m->cp_hqd_pq_doorbell_control |= 1 << CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_BIF_DROP__SHIFT; } - if (mm->dev->cwsr_enabled) + if (mm->dev->kfd->cwsr_enabled) m->cp_hqd_ctx_save_control = 0; update_cu_mask(mm, mqd, minfo); @@ -219,19 +226,21 @@ static void update_mqd(struct mqd_manager *mm, void *mqd, q->is_active = QUEUE_IS_ACTIVE(*q); } -static uint32_t read_doorbell_id(void *mqd) +static bool check_preemption_failed(struct mqd_manager *mm, void *mqd) { struct v10_compute_mqd *m = (struct v10_compute_mqd *)mqd; - return m->queue_doorbell_id0; + return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0); } static int get_wave_state(struct mqd_manager *mm, void *mqd, + struct queue_properties *q, void __user *ctl_stack, u32 *ctl_stack_used_size, u32 *save_area_used_size) { struct v10_compute_mqd *m; + struct kfd_context_save_area_header header; m = get_mqd(mqd); @@ -250,6 +259,15 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd, * accessible to user mode */ + header.wave_state.control_stack_size = *ctl_stack_used_size; + header.wave_state.wave_state_size = *save_area_used_size; + + header.wave_state.wave_state_offset = m->cp_hqd_wg_state_offset; + header.wave_state.control_stack_offset = m->cp_hqd_cntl_stack_offset; + + if (copy_to_user(ctl_stack, &header, sizeof(header.wave_state))) + return -EFAULT; + return 0; } @@ -303,6 +321,26 @@ static void init_mqd_hiq(struct mqd_manager *mm, void **mqd, 1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT; } +static int destroy_hiq_mqd(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, unsigned int timeout, + uint32_t pipe_id, uint32_t queue_id) +{ + int err; + struct v10_compute_mqd *m; + u32 doorbell_off; + + m = get_mqd(mqd); + + doorbell_off = m->cp_hqd_pq_doorbell_control >> + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT; + + err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, 0); + if (err) + pr_debug("Destroy HIQ MQD failed: %d\n", err); + + return err; +} + static void init_mqd_sdma(struct mqd_manager *mm, void **mqd, struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr, struct queue_properties *q) @@ -405,7 +443,7 @@ static int debugfs_show_mqd_sdma(struct seq_file *m, void *data) #endif struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type, - struct kfd_dev *dev) + struct kfd_node *dev) { struct mqd_manager *mqd; @@ -432,6 +470,7 @@ struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type, mqd->get_wave_state = get_wave_state; mqd->checkpoint_mqd = checkpoint_mqd; mqd->restore_mqd = restore_mqd; + mqd->mqd_stride = kfd_mqd_stride; #if defined(CONFIG_DEBUG_FS) mqd->debugfs_show_mqd = debugfs_show_mqd; #endif @@ -444,13 +483,14 @@ struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type, mqd->free_mqd = free_mqd_hiq_sdma; mqd->load_mqd = kfd_hiq_load_mqd_kiq; mqd->update_mqd = update_mqd; - mqd->destroy_mqd = kfd_destroy_mqd_cp; + mqd->destroy_mqd = destroy_hiq_mqd; mqd->is_occupied = kfd_is_occupied_cp; mqd->mqd_size = sizeof(struct v10_compute_mqd); + mqd->mqd_stride = kfd_mqd_stride; #if defined(CONFIG_DEBUG_FS) mqd->debugfs_show_mqd = debugfs_show_mqd; #endif - mqd->read_doorbell_id = read_doorbell_id; + mqd->check_preemption_failed = check_preemption_failed; pr_debug("%s@%i\n", __func__, __LINE__); break; case KFD_MQD_TYPE_DIQ: @@ -478,6 +518,7 @@ struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type, mqd->checkpoint_mqd = checkpoint_mqd_sdma; mqd->restore_mqd = restore_mqd_sdma; mqd->mqd_size = sizeof(struct v10_sdma_mqd); + mqd->mqd_stride = kfd_mqd_stride; #if defined(CONFIG_DEBUG_FS) mqd->debugfs_show_mqd = debugfs_show_mqd_sdma; #endif |
