Author | Tokens | Token Proportion | Commits | Commit Proportion |
---|---|---|---|---|
Ben Goz | 1241 | 48.67% | 4 | 8.33% |
Felix Kuhling | 647 | 25.37% | 10 | 20.83% |
Oak Zeng | 299 | 11.73% | 4 | 8.33% |
Jay Cornwall | 91 | 3.57% | 1 | 2.08% |
Yong Zhao | 57 | 2.24% | 12 | 25.00% |
Yair Shachar | 43 | 1.69% | 2 | 4.17% |
Shaoyun Liu | 32 | 1.25% | 1 | 2.08% |
Oded Gabbay | 30 | 1.18% | 3 | 6.25% |
Joseph Greathouse | 27 | 1.06% | 1 | 2.08% |
Alexey Skidanov | 24 | 0.94% | 1 | 2.08% |
Kent Russell | 20 | 0.78% | 4 | 8.33% |
Philip Yang | 16 | 0.63% | 1 | 2.08% |
Amber Lin | 12 | 0.47% | 1 | 2.08% |
Andres Rodriguez | 5 | 0.20% | 1 | 2.08% |
Dan Carpenter | 4 | 0.16% | 1 | 2.08% |
Dave Airlie | 2 | 0.08% | 1 | 2.08% |
Total | 2550 | 48 |
/* * Copyright 2014 Advanced Micro Devices, Inc. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in * all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. * */ #include <linux/slab.h> #include <linux/list.h> #include "kfd_device_queue_manager.h" #include "kfd_priv.h" #include "kfd_kernel_queue.h" #include "amdgpu_amdkfd.h" static inline struct process_queue_node *get_queue_by_qid( struct process_queue_manager *pqm, unsigned int qid) { struct process_queue_node *pqn; list_for_each_entry(pqn, &pqm->queues, process_queue_list) { if ((pqn->q && pqn->q->properties.queue_id == qid) || (pqn->kq && pqn->kq->queue->properties.queue_id == qid)) return pqn; } return NULL; } static int find_available_queue_slot(struct process_queue_manager *pqm, unsigned int *qid) { unsigned long found; found = find_first_zero_bit(pqm->queue_slot_bitmap, KFD_MAX_NUM_OF_QUEUES_PER_PROCESS); pr_debug("The new slot id %lu\n", found); if (found >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) { pr_info("Cannot open more queues for process with pasid 0x%x\n", pqm->process->pasid); return -ENOMEM; } set_bit(found, pqm->queue_slot_bitmap); *qid = found; return 0; } void kfd_process_dequeue_from_device(struct kfd_process_device *pdd) { struct kfd_dev *dev = pdd->dev; if (pdd->already_dequeued) return; dev->dqm->ops.process_termination(dev->dqm, &pdd->qpd); pdd->already_dequeued = true; } int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid, void *gws) { struct kfd_dev *dev = NULL; struct process_queue_node *pqn; struct kfd_process_device *pdd; struct kgd_mem *mem = NULL; int ret; pqn = get_queue_by_qid(pqm, qid); if (!pqn) { pr_err("Queue id does not match any known queue\n"); return -EINVAL; } if (pqn->q) dev = pqn->q->device; if (WARN_ON(!dev)) return -ENODEV; pdd = kfd_get_process_device_data(dev, pqm->process); if (!pdd) { pr_err("Process device data doesn't exist\n"); return -EINVAL; } /* Only allow one queue per process can have GWS assigned */ if (gws && pdd->qpd.num_gws) return -EBUSY; if (!gws && pdd->qpd.num_gws == 0) return -EINVAL; if (gws) ret = amdgpu_amdkfd_add_gws_to_process(pdd->process->kgd_process_info, gws, &mem); else ret = amdgpu_amdkfd_remove_gws_from_process(pdd->process->kgd_process_info, pqn->q->gws); if (unlikely(ret)) return ret; pqn->q->gws = mem; pdd->qpd.num_gws = gws ? amdgpu_amdkfd_get_num_gws(dev->kgd) : 0; return pqn->q->device->dqm->ops.update_queue(pqn->q->device->dqm, pqn->q); } void kfd_process_dequeue_from_all_devices(struct kfd_process *p) { struct kfd_process_device *pdd; list_for_each_entry(pdd, &p->per_device_data, per_device_list) kfd_process_dequeue_from_device(pdd); } int pqm_init(struct process_queue_manager *pqm, struct kfd_process *p) { INIT_LIST_HEAD(&pqm->queues); pqm->queue_slot_bitmap = kzalloc(DIV_ROUND_UP(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS, BITS_PER_BYTE), GFP_KERNEL); if (!pqm->queue_slot_bitmap) return -ENOMEM; pqm->process = p; return 0; } void pqm_uninit(struct process_queue_manager *pqm) { struct process_queue_node *pqn, *next; list_for_each_entry_safe(pqn, next, &pqm->queues, process_queue_list) { if (pqn->q && pqn->q->gws) amdgpu_amdkfd_remove_gws_from_process(pqm->process->kgd_process_info, pqn->q->gws); uninit_queue(pqn->q); list_del(&pqn->process_queue_list); kfree(pqn); } kfree(pqm->queue_slot_bitmap); pqm->queue_slot_bitmap = NULL; } static int init_user_queue(struct process_queue_manager *pqm, struct kfd_dev *dev, struct queue **q, struct queue_properties *q_properties, struct file *f, unsigned int qid) { int retval; /* Doorbell initialized in user space*/ q_properties->doorbell_ptr = NULL; /* let DQM handle it*/ q_properties->vmid = 0; q_properties->queue_id = qid; retval = init_queue(q, q_properties); if (retval != 0) return retval; (*q)->device = dev; (*q)->process = pqm->process; pr_debug("PQM After init queue"); return retval; } int pqm_create_queue(struct process_queue_manager *pqm, struct kfd_dev *dev, struct file *f, struct queue_properties *properties, unsigned int *qid, uint32_t *p_doorbell_offset_in_process) { int retval; struct kfd_process_device *pdd; struct queue *q; struct process_queue_node *pqn; struct kernel_queue *kq; enum kfd_queue_type type = properties->type; unsigned int max_queues = 127; /* HWS limit */ q = NULL; kq = NULL; pdd = kfd_get_process_device_data(dev, pqm->process); if (!pdd) { pr_err("Process device data doesn't exist\n"); return -1; } /* * for debug process, verify that it is within the static queues limit * currently limit is set to half of the total avail HQD slots * If we are just about to create DIQ, the is_debug flag is not set yet * Hence we also check the type as well */ if ((pdd->qpd.is_debug) || (type == KFD_QUEUE_TYPE_DIQ)) max_queues = dev->device_info->max_no_of_hqd/2; if (pdd->qpd.queue_count >= max_queues) return -ENOSPC; retval = find_available_queue_slot(pqm, qid); if (retval != 0) return retval; if (list_empty(&pdd->qpd.queues_list) && list_empty(&pdd->qpd.priv_queue_list)) dev->dqm->ops.register_process(dev->dqm, &pdd->qpd); pqn = kzalloc(sizeof(*pqn), GFP_KERNEL); if (!pqn) { retval = -ENOMEM; goto err_allocate_pqn; } switch (type) { case KFD_QUEUE_TYPE_SDMA: case KFD_QUEUE_TYPE_SDMA_XGMI: /* SDMA queues are always allocated statically no matter * which scheduler mode is used. We also do not need to * check whether a SDMA queue can be allocated here, because * allocate_sdma_queue() in create_queue() has the * corresponding check logic. */ retval = init_user_queue(pqm, dev, &q, properties, f, *qid); if (retval != 0) goto err_create_queue; pqn->q = q; pqn->kq = NULL; retval = dev->dqm->ops.create_queue(dev->dqm, q, &pdd->qpd); print_queue(q); break; case KFD_QUEUE_TYPE_COMPUTE: /* check if there is over subscription */ if ((dev->dqm->sched_policy == KFD_SCHED_POLICY_HWS_NO_OVERSUBSCRIPTION) && ((dev->dqm->processes_count >= dev->vm_info.vmid_num_kfd) || (dev->dqm->active_queue_count >= get_cp_queues_num(dev->dqm)))) { pr_debug("Over-subscription is not allowed when amdkfd.sched_policy == 1\n"); retval = -EPERM; goto err_create_queue; } retval = init_user_queue(pqm, dev, &q, properties, f, *qid); if (retval != 0) goto err_create_queue; pqn->q = q; pqn->kq = NULL; retval = dev->dqm->ops.create_queue(dev->dqm, q, &pdd->qpd); print_queue(q); break; case KFD_QUEUE_TYPE_DIQ: kq = kernel_queue_init(dev, KFD_QUEUE_TYPE_DIQ); if (!kq) { retval = -ENOMEM; goto err_create_queue; } kq->queue->properties.queue_id = *qid; pqn->kq = kq; pqn->q = NULL; retval = dev->dqm->ops.create_kernel_queue(dev->dqm, kq, &pdd->qpd); break; default: WARN(1, "Invalid queue type %d", type); retval = -EINVAL; } if (retval != 0) { pr_err("Pasid 0x%x DQM create queue type %d failed. ret %d\n", pqm->process->pasid, type, retval); goto err_create_queue; } if (q && p_doorbell_offset_in_process) /* Return the doorbell offset within the doorbell page * to the caller so it can be passed up to user mode * (in bytes). * There are always 1024 doorbells per process, so in case * of 8-byte doorbells, there are two doorbell pages per * process. */ *p_doorbell_offset_in_process = (q->properties.doorbell_off * sizeof(uint32_t)) & (kfd_doorbell_process_slice(dev) - 1); pr_debug("PQM After DQM create queue\n"); list_add(&pqn->process_queue_list, &pqm->queues); if (q) { pr_debug("PQM done creating queue\n"); kfd_procfs_add_queue(q); print_queue_properties(&q->properties); } return retval; err_create_queue: uninit_queue(q); if (kq) kernel_queue_uninit(kq, false); kfree(pqn); err_allocate_pqn: /* check if queues list is empty unregister process from device */ clear_bit(*qid, pqm->queue_slot_bitmap); if (list_empty(&pdd->qpd.queues_list) && list_empty(&pdd->qpd.priv_queue_list)) dev->dqm->ops.unregister_process(dev->dqm, &pdd->qpd); return retval; } int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) { struct process_queue_node *pqn; struct kfd_process_device *pdd; struct device_queue_manager *dqm; struct kfd_dev *dev; int retval; dqm = NULL; retval = 0; pqn = get_queue_by_qid(pqm, qid); if (!pqn) { pr_err("Queue id does not match any known queue\n"); return -EINVAL; } dev = NULL; if (pqn->kq) dev = pqn->kq->dev; if (pqn->q) dev = pqn->q->device; if (WARN_ON(!dev)) return -ENODEV; pdd = kfd_get_process_device_data(dev, pqm->process); if (!pdd) { pr_err("Process device data doesn't exist\n"); return -1; } if (pqn->kq) { /* destroy kernel queue (DIQ) */ dqm = pqn->kq->dev->dqm; dqm->ops.destroy_kernel_queue(dqm, pqn->kq, &pdd->qpd); kernel_queue_uninit(pqn->kq, false); } if (pqn->q) { kfd_procfs_del_queue(pqn->q); dqm = pqn->q->device->dqm; retval = dqm->ops.destroy_queue(dqm, &pdd->qpd, pqn->q); if (retval) { pr_err("Pasid 0x%x destroy queue %d failed, ret %d\n", pqm->process->pasid, pqn->q->properties.queue_id, retval); if (retval != -ETIME) goto err_destroy_queue; } if (pqn->q->gws) { amdgpu_amdkfd_remove_gws_from_process(pqm->process->kgd_process_info, pqn->q->gws); pdd->qpd.num_gws = 0; } kfree(pqn->q->properties.cu_mask); pqn->q->properties.cu_mask = NULL; uninit_queue(pqn->q); } list_del(&pqn->process_queue_list); kfree(pqn); clear_bit(qid, pqm->queue_slot_bitmap); if (list_empty(&pdd->qpd.queues_list) && list_empty(&pdd->qpd.priv_queue_list)) dqm->ops.unregister_process(dqm, &pdd->qpd); err_destroy_queue: return retval; } int pqm_update_queue(struct process_queue_manager *pqm, unsigned int qid, struct queue_properties *p) { int retval; struct process_queue_node *pqn; pqn = get_queue_by_qid(pqm, qid); if (!pqn) { pr_debug("No queue %d exists for update operation\n", qid); return -EFAULT; } pqn->q->properties.queue_address = p->queue_address; pqn->q->properties.queue_size = p->queue_size; pqn->q->properties.queue_percent = p->queue_percent; pqn->q->properties.priority = p->priority; retval = pqn->q->device->dqm->ops.update_queue(pqn->q->device->dqm, pqn->q); if (retval != 0) return retval; return 0; } int pqm_set_cu_mask(struct process_queue_manager *pqm, unsigned int qid, struct queue_properties *p) { int retval; struct process_queue_node *pqn; pqn = get_queue_by_qid(pqm, qid); if (!pqn) { pr_debug("No queue %d exists for update operation\n", qid); return -EFAULT; } /* Free the old CU mask memory if it is already allocated, then * allocate memory for the new CU mask. */ kfree(pqn->q->properties.cu_mask); pqn->q->properties.cu_mask_count = p->cu_mask_count; pqn->q->properties.cu_mask = p->cu_mask; retval = pqn->q->device->dqm->ops.update_queue(pqn->q->device->dqm, pqn->q); if (retval != 0) return retval; return 0; } struct kernel_queue *pqm_get_kernel_queue( struct process_queue_manager *pqm, unsigned int qid) { struct process_queue_node *pqn; pqn = get_queue_by_qid(pqm, qid); if (pqn && pqn->kq) return pqn->kq; return NULL; } int pqm_get_wave_state(struct process_queue_manager *pqm, unsigned int qid, void __user *ctl_stack, u32 *ctl_stack_used_size, u32 *save_area_used_size) { struct process_queue_node *pqn; pqn = get_queue_by_qid(pqm, qid); if (!pqn) { pr_debug("amdkfd: No queue %d exists for operation\n", qid); return -EFAULT; } return pqn->q->device->dqm->ops.get_wave_state(pqn->q->device->dqm, pqn->q, ctl_stack, ctl_stack_used_size, save_area_used_size); } #if defined(CONFIG_DEBUG_FS) int pqm_debugfs_mqds(struct seq_file *m, void *data) { struct process_queue_manager *pqm = data; struct process_queue_node *pqn; struct queue *q; enum KFD_MQD_TYPE mqd_type; struct mqd_manager *mqd_mgr; int r = 0; list_for_each_entry(pqn, &pqm->queues, process_queue_list) { if (pqn->q) { q = pqn->q; switch (q->properties.type) { case KFD_QUEUE_TYPE_SDMA: case KFD_QUEUE_TYPE_SDMA_XGMI: seq_printf(m, " SDMA queue on device %x\n", q->device->id); mqd_type = KFD_MQD_TYPE_SDMA; break; case KFD_QUEUE_TYPE_COMPUTE: seq_printf(m, " Compute queue on device %x\n", q->device->id); mqd_type = KFD_MQD_TYPE_CP; break; default: seq_printf(m, " Bad user queue type %d on device %x\n", q->properties.type, q->device->id); continue; } mqd_mgr = q->device->dqm->mqd_mgrs[mqd_type]; } else if (pqn->kq) { q = pqn->kq->queue; mqd_mgr = pqn->kq->mqd_mgr; switch (q->properties.type) { case KFD_QUEUE_TYPE_DIQ: seq_printf(m, " DIQ on device %x\n", pqn->kq->dev->id); break; default: seq_printf(m, " Bad kernel queue type %d on device %x\n", q->properties.type, pqn->kq->dev->id); continue; } } else { seq_printf(m, " Weird: Queue node with neither kernel nor user queue\n"); continue; } r = mqd_mgr->debugfs_show_mqd(m, q->mqd); if (r != 0) break; } return r; } #endif
Information contained on this website is for historical information purposes only and does not indicate or represent copyright ownership.
Created with Cregit http://github.com/cregit/cregit
Version 2.0-RC1