/************************************************************************** * * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA * All Rights Reserved. * * 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, sub license, 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 (including the * next paragraph) 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 NON-INFRINGEMENT. IN NO EVENT SHALL * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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. * **************************************************************************/ /* * Authors: Thomas Hellstrom */ #define pr_fmt(fmt) "[TTM] " fmt #include #include #include #include #include #include #include #include #include #include #include #define TTM_ASSERT_LOCKED(param) #define TTM_DEBUG(fmt, arg...) #define TTM_BO_HASH_ORDER 13 static inline int ttm_mem_type_from_place(const struct ttm_place *place, uint32_t *mem_type) { int i; for (i = 0; i <= TTM_PL_PRIV5; i++) if (place->flags & (1 << i)) { *mem_type = i; return 0; } return -EINVAL; } static inline uint32_t ttm_bo_type_flags(unsigned type) { return 1 << (type); } static void ttm_bo_release_list(struct kref *list_kref) { struct ttm_buffer_object *bo = container_of(list_kref, struct ttm_buffer_object, list_kref); struct ttm_bo_device *bdev = bo->bdev; size_t acc_size = bo->acc_size; BUG_ON(atomic_read(&bo->list_kref.refcount)); BUG_ON(atomic_read(&bo->kref.refcount)); BUG_ON(atomic_read(&bo->cpu_writers)); BUG_ON(bo->mem.mm_node != NULL); BUG_ON(!list_empty(&bo->lru)); BUG_ON(!list_empty(&bo->ddestroy)); if (bo->ttm) ttm_tt_destroy(bo->ttm); atomic_dec(&bo->glob->bo_count); if (bo->resv == &bo->ttm_resv) reservation_object_fini(&bo->ttm_resv); mutex_destroy(&bo->wu_mutex); if (bo->destroy) bo->destroy(bo); else { kfree(bo); } } void ttm_bo_add_to_lru(struct ttm_buffer_object *bo) { struct ttm_bo_device *bdev = bo->bdev; struct ttm_mem_type_manager *man; lockdep_assert_held(&bo->resv->lock.base); if (!(bo->mem.placement & TTM_PL_FLAG_NO_EVICT)) { BUG_ON(!list_empty(&bo->lru)); man = &bdev->man[bo->mem.mem_type]; list_add_tail(&bo->lru, &man->lru); kref_get(&bo->list_kref); if (bo->ttm != NULL) { list_add_tail(&bo->swap, &bo->glob->swap_lru); kref_get(&bo->list_kref); } } } EXPORT_SYMBOL(ttm_bo_add_to_lru); int ttm_bo_del_from_lru(struct ttm_buffer_object *bo) { int put_count = 0; if (!list_empty(&bo->swap)) { list_del_init(&bo->swap); ++put_count; } if (!list_empty(&bo->lru)) { list_del_init(&bo->lru); ++put_count; } /* * TODO: Add a driver hook to delete from * driver-specific LRU's here. */ return put_count; } static void ttm_bo_ref_bug(struct kref *list_kref) { BUG(); } void ttm_bo_list_ref_sub(struct ttm_buffer_object *bo, int count, bool never_free) { kref_sub(&bo->list_kref, count, (never_free) ? ttm_bo_ref_bug : ttm_bo_release_list); } void ttm_bo_del_sub_from_lru(struct ttm_buffer_object *bo) { int put_count; spin_lock(&bo->glob->lru_lock); put_count = ttm_bo_del_from_lru(bo); spin_unlock(&bo->glob->lru_lock); ttm_bo_list_ref_sub(bo, put_count, true); } EXPORT_SYMBOL(ttm_bo_del_sub_from_lru); /* * Call bo->mutex locked. */ static int ttm_bo_add_ttm(struct ttm_buffer_object *bo, bool zero_alloc) { struct ttm_bo_device *bdev = bo->bdev; struct ttm_bo_global *glob = bo->glob; int ret = 0; uint32_t page_flags = 0; TTM_ASSERT_LOCKED(&bo->mutex); bo->ttm = NULL; if (bdev->need_dma32) page_flags |= TTM_PAGE_FLAG_DMA32; switch (bo->type) { case ttm_bo_type_device: if (zero_alloc) page_flags |= TTM_PAGE_FLAG_ZERO_ALLOC; case ttm_bo_type_kernel: bo->ttm = bdev->driver->ttm_tt_create(bdev, bo->num_pages << PAGE_SHIFT, page_flags, glob->dummy_read_page); if (unlikely(bo->ttm == NULL)) ret = -ENOMEM; break; case ttm_bo_type_sg: bo->ttm = bdev->driver->ttm_tt_create(bdev, bo->num_pages << PAGE_SHIFT, page_flags | TTM_PAGE_FLAG_SG, glob->dummy_read_page); if (unlikely(bo->ttm == NULL)) { ret = -ENOMEM; break; } bo->ttm->sg = bo->sg; break; default: pr_err("Illegal buffer object type\n"); ret = -EINVAL; break; } return ret; } static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo, struct ttm_mem_reg *mem, bool evict, bool interruptible, bool no_wait_gpu) { struct ttm_bo_device *bdev = bo->bdev; bool old_is_pci = ttm_mem_reg_is_pci(bdev, &bo->mem); bool new_is_pci = ttm_mem_reg_is_pci(bdev, mem); struct ttm_mem_type_manager *old_man = &bdev->man[bo->mem.mem_type]; struct ttm_mem_type_manager *new_man = &bdev->man[mem->mem_type]; int ret = 0; if (old_is_pci || new_is_pci || ((mem->placement & bo->mem.placement & TTM_PL_MASK_CACHING) == 0)) { ret = ttm_mem_io_lock(old_man, true); if (unlikely(ret != 0)) goto out_err; ttm_bo_unmap_virtual_locked(bo); ttm_mem_io_unlock(old_man); } /* * Create and bind a ttm if required. */ if (!(new_man->flags & TTM_MEMTYPE_FLAG_FIXED)) { if (bo->ttm == NULL) { bool zero = !(old_man->flags & TTM_MEMTYPE_FLAG_FIXED); ret = ttm_bo_add_ttm(bo, zero); if (ret) goto out_err; } ret = ttm_tt_set_placement_caching(bo->ttm, mem->placement); if (ret) goto out_err; if (mem->mem_type != TTM_PL_SYSTEM) { ret = ttm_tt_bind(bo->ttm, mem); if (ret) goto out_err; } if (bo->mem.mem_type == TTM_PL_SYSTEM) { if (bdev->driver->move_notify) bdev->driver->move_notify(bo, mem); bo->mem = *mem; mem->mm_node = NULL; goto moved; } } if (bdev->driver->move_notify) bdev->driver->move_notify(bo, mem); if (!(old_man->flags & TTM_MEMTYPE_FLAG_FIXED) && !(new_man->flags & TTM_MEMTYPE_FLAG_FIXED)) ret = ttm_bo_move_ttm(bo, evict, no_wait_gpu, mem); else if (bdev->driver->move) ret = bdev->driver->move(bo, evict, interruptible, no_wait_gpu, mem); else ret = ttm_bo_move_memcpy(bo, evict, no_wait_gpu, mem); if (ret) { if (bdev->driver->move_notify) { struct ttm_mem_reg tmp_mem = *mem; *mem = bo->mem; bo->mem = tmp_mem; bdev->driver->move_notify(bo, mem); bo->mem = *mem; *mem = tmp_mem; } goto out_err; } moved: if (bo->evicted) { if (bdev->driver->invalidate_caches) { ret = bdev->driver->invalidate_caches(bdev, bo->mem.placement); if (ret) pr_err("Can not flush read caches\n"); } bo->evicted = false; } if (bo->mem.mm_node) { bo->offset = (bo->mem.start << PAGE_SHIFT) + bdev->man[bo->mem.mem_type].gpu_offset; bo->cur_placement = bo->mem.placement; } else bo->offset = 0; return 0; out_err: new_man = &bdev->man[bo->mem.mem_type]; if ((new_man->flags & TTM_MEMTYPE_FLAG_FIXED) && bo->ttm) { ttm_tt_unbind(bo->ttm); ttm_tt_destroy(bo->ttm); bo->ttm = NULL; } return ret; } /** * Call bo::reserved. * Will release GPU memory type usage on destruction. * This is the place to put in driver specific hooks to release * driver private resources. * Will release the bo::reserved lock. */ static void ttm_bo_cleanup_memtype_use(struct ttm_buffer_object *bo) { if (bo->bdev->driver->move_notify) bo->bdev->driver->move_notify(bo, NULL); if (bo->ttm) { ttm_tt_unbind(bo->ttm); ttm_tt_destroy(bo->ttm); bo->ttm = NULL; } ttm_bo_mem_put(bo, &bo->mem); ww_mutex_unlock (&bo->resv->lock); } static void ttm_bo_flush_all_fences(struct ttm_buffer_object *bo) { struct reservation_object_list *fobj; struct fence *fence; int i; fobj = reservation_object_get_list(bo->resv); fence = reservation_object_get_excl(bo->resv); if (fence && !fence->ops->signaled) fence_enable_sw_signaling(fence); for (i = 0; fobj && i < fobj->shared_count; ++i) { fence = rcu_dereference_protected(fobj->shared[i], reservation_object_held(bo->resv)); if (!fence->ops->signaled) fence_enable_sw_signaling(fence); } } static void ttm_bo_cleanup_refs_or_queue(struct ttm_buffer_object *bo) { struct ttm_bo_device *bdev = bo->bdev; struct ttm_bo_global *glob = bo->glob; int put_count; int ret; spin_lock(&glob->lru_lock); ret = __ttm_bo_reserve(bo, false, true, false, NULL); if (!ret) { if (!ttm_bo_wait(bo, false, false, true)) { put_count = ttm_bo_del_from_lru(bo); spin_unlock(&glob->lru_lock); ttm_bo_cleanup_memtype_use(bo); ttm_bo_list_ref_sub(bo, put_count, true); return; } else ttm_bo_flush_all_fences(bo); /* * Make NO_EVICT bos immediately available to * shrinkers, now that they are queued for * destruction. */ if (bo->mem.placement & TTM_PL_FLAG_NO_EVICT) { bo->mem.placement &= ~TTM_PL_FLAG_NO_EVICT; ttm_bo_add_to_lru(bo); } __ttm_bo_unreserve(bo); } kref_get(&bo->list_kref); list_add_tail(&bo->ddestroy, &bdev->ddestroy); spin_unlock(&glob->lru_lock); // schedule_delayed_work(&bdev->wq, // ((HZ / 100) < 1) ? 1 : HZ / 100); } /** * function ttm_bo_cleanup_refs_and_unlock * If bo idle, remove from delayed- and lru lists, and unref. * If not idle, do nothing. * * Must be called with lru_lock and reservation held, this function * will drop both before returning. * * @interruptible Any sleeps should occur interruptibly. * @no_wait_gpu Never wait for gpu. Return -EBUSY instead. */ static int ttm_bo_cleanup_refs_and_unlock(struct ttm_buffer_object *bo, bool interruptible, bool no_wait_gpu) { struct ttm_bo_global *glob = bo->glob; int put_count; int ret; ret = ttm_bo_wait(bo, false, false, true); if (ret && !no_wait_gpu) { long lret; ww_mutex_unlock(&bo->resv->lock); spin_unlock(&glob->lru_lock); lret = reservation_object_wait_timeout_rcu(bo->resv, true, interruptible, 30 * HZ); if (lret < 0) return lret; else if (lret == 0) return -EBUSY; spin_lock(&glob->lru_lock); ret = __ttm_bo_reserve(bo, false, true, false, NULL); /* * We raced, and lost, someone else holds the reservation now, * and is probably busy in ttm_bo_cleanup_memtype_use. * * Even if it's not the case, because we finished waiting any * delayed destruction would succeed, so just return success * here. */ if (ret) { spin_unlock(&glob->lru_lock); return 0; } /* * remove sync_obj with ttm_bo_wait, the wait should be * finished, and no new wait object should have been added. */ ret = ttm_bo_wait(bo, false, false, true); WARN_ON(ret); } if (ret || unlikely(list_empty(&bo->ddestroy))) { __ttm_bo_unreserve(bo); spin_unlock(&glob->lru_lock); return ret; } put_count = ttm_bo_del_from_lru(bo); list_del_init(&bo->ddestroy); ++put_count; spin_unlock(&glob->lru_lock); ttm_bo_cleanup_memtype_use(bo); ttm_bo_list_ref_sub(bo, put_count, true); return 0; } /** * Traverse the delayed list, and call ttm_bo_cleanup_refs on all * encountered buffers. */ static int ttm_bo_delayed_delete(struct ttm_bo_device *bdev, bool remove_all) { struct ttm_bo_global *glob = bdev->glob; struct ttm_buffer_object *entry = NULL; int ret = 0; spin_lock(&glob->lru_lock); if (list_empty(&bdev->ddestroy)) goto out_unlock; entry = list_first_entry(&bdev->ddestroy, struct ttm_buffer_object, ddestroy); kref_get(&entry->list_kref); for (;;) { struct ttm_buffer_object *nentry = NULL; if (entry->ddestroy.next != &bdev->ddestroy) { nentry = list_first_entry(&entry->ddestroy, struct ttm_buffer_object, ddestroy); kref_get(&nentry->list_kref); } ret = __ttm_bo_reserve(entry, false, true, false, NULL); if (remove_all && ret) { spin_unlock(&glob->lru_lock); ret = __ttm_bo_reserve(entry, false, false, false, NULL); spin_lock(&glob->lru_lock); } if (!ret) ret = ttm_bo_cleanup_refs_and_unlock(entry, false, !remove_all); else spin_unlock(&glob->lru_lock); kref_put(&entry->list_kref, ttm_bo_release_list); entry = nentry; if (ret || !entry) goto out; spin_lock(&glob->lru_lock); if (list_empty(&entry->ddestroy)) break; } out_unlock: spin_unlock(&glob->lru_lock); out: if (entry) kref_put(&entry->list_kref, ttm_bo_release_list); return ret; } static void ttm_bo_delayed_workqueue(struct work_struct *work) { struct ttm_bo_device *bdev = container_of(work, struct ttm_bo_device, wq.work); if (ttm_bo_delayed_delete(bdev, false)) { schedule_delayed_work(&bdev->wq, ((HZ / 100) < 1) ? 1 : HZ / 100); } } static void ttm_bo_release(struct kref *kref) { struct ttm_buffer_object *bo = container_of(kref, struct ttm_buffer_object, kref); struct ttm_bo_device *bdev = bo->bdev; struct ttm_mem_type_manager *man = &bdev->man[bo->mem.mem_type]; drm_vma_offset_remove(&bdev->vma_manager, &bo->vma_node); ttm_mem_io_lock(man, false); ttm_mem_io_free_vm(bo); ttm_mem_io_unlock(man); ttm_bo_cleanup_refs_or_queue(bo); kref_put(&bo->list_kref, ttm_bo_release_list); } void ttm_bo_unref(struct ttm_buffer_object **p_bo) { struct ttm_buffer_object *bo = *p_bo; *p_bo = NULL; kref_put(&bo->kref, ttm_bo_release); } EXPORT_SYMBOL(ttm_bo_unref); void ttm_bo_mem_put(struct ttm_buffer_object *bo, struct ttm_mem_reg *mem) { struct ttm_mem_type_manager *man = &bo->bdev->man[mem->mem_type]; if (mem->mm_node) (*man->func->put_node)(man, mem); } EXPORT_SYMBOL(ttm_bo_mem_put); /** * Repeatedly evict memory from the LRU for @mem_type until we create enough * space, or we've evicted everything and there isn't enough space. */ static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo, uint32_t mem_type, const struct ttm_place *place, struct ttm_mem_reg *mem, bool interruptible, bool no_wait_gpu) { struct ttm_bo_device *bdev = bo->bdev; struct ttm_mem_type_manager *man = &bdev->man[mem_type]; int ret; do { ret = (*man->func->get_node)(man, bo, place, mem); if (unlikely(ret != 0)) return ret; if (mem->mm_node) break; // ret = ttm_mem_evict_first(bdev, mem_type, // interruptible, no_wait_gpu); // if (unlikely(ret != 0)) // return ret; } while (1); if (mem->mm_node == NULL) return -ENOMEM; mem->mem_type = mem_type; return 0; } static uint32_t ttm_bo_select_caching(struct ttm_mem_type_manager *man, uint32_t cur_placement, uint32_t proposed_placement) { uint32_t caching = proposed_placement & TTM_PL_MASK_CACHING; uint32_t result = proposed_placement & ~TTM_PL_MASK_CACHING; /** * Keep current caching if possible. */ if ((cur_placement & caching) != 0) result |= (cur_placement & caching); else if ((man->default_caching & caching) != 0) result |= man->default_caching; else if ((TTM_PL_FLAG_CACHED & caching) != 0) result |= TTM_PL_FLAG_CACHED; else if ((TTM_PL_FLAG_WC & caching) != 0) result |= TTM_PL_FLAG_WC; else if ((TTM_PL_FLAG_UNCACHED & caching) != 0) result |= TTM_PL_FLAG_UNCACHED; return result; } static bool ttm_bo_mt_compatible(struct ttm_mem_type_manager *man, uint32_t mem_type, const struct ttm_place *place, uint32_t *masked_placement) { uint32_t cur_flags = ttm_bo_type_flags(mem_type); if ((cur_flags & place->flags & TTM_PL_MASK_MEM) == 0) return false; if ((place->flags & man->available_caching) == 0) return false; cur_flags |= (place->flags & man->available_caching); *masked_placement = cur_flags; return true; } /** * Creates space for memory region @mem according to its type. * * This function first searches for free space in compatible memory types in * the priority order defined by the driver. If free space isn't found, then * ttm_bo_mem_force_space is attempted in priority order to evict and find * space. */ int ttm_bo_mem_space(struct ttm_buffer_object *bo, struct ttm_placement *placement, struct ttm_mem_reg *mem, bool interruptible, bool no_wait_gpu) { struct ttm_bo_device *bdev = bo->bdev; struct ttm_mem_type_manager *man; uint32_t mem_type = TTM_PL_SYSTEM; uint32_t cur_flags = 0; bool type_found = false; bool type_ok = false; bool has_erestartsys = false; int i, ret; mem->mm_node = NULL; for (i = 0; i < placement->num_placement; ++i) { const struct ttm_place *place = &placement->placement[i]; ret = ttm_mem_type_from_place(place, &mem_type); if (ret) return ret; man = &bdev->man[mem_type]; if (!man->has_type || !man->use_type) continue; type_ok = ttm_bo_mt_compatible(man, mem_type, place, &cur_flags); if (!type_ok) continue; type_found = true; cur_flags = ttm_bo_select_caching(man, bo->mem.placement, cur_flags); /* * Use the access and other non-mapping-related flag bits from * the memory placement flags to the current flags */ ttm_flag_masked(&cur_flags, place->flags, ~TTM_PL_MASK_MEMTYPE); if (mem_type == TTM_PL_SYSTEM) break; ret = (*man->func->get_node)(man, bo, place, mem); if (unlikely(ret)) return ret; if (mem->mm_node) break; } if ((type_ok && (mem_type == TTM_PL_SYSTEM)) || mem->mm_node) { mem->mem_type = mem_type; mem->placement = cur_flags; return 0; } for (i = 0; i < placement->num_busy_placement; ++i) { const struct ttm_place *place = &placement->busy_placement[i]; ret = ttm_mem_type_from_place(place, &mem_type); if (ret) return ret; man = &bdev->man[mem_type]; if (!man->has_type || !man->use_type) continue; if (!ttm_bo_mt_compatible(man, mem_type, place, &cur_flags)) continue; type_found = true; cur_flags = ttm_bo_select_caching(man, bo->mem.placement, cur_flags); /* * Use the access and other non-mapping-related flag bits from * the memory placement flags to the current flags */ ttm_flag_masked(&cur_flags, place->flags, ~TTM_PL_MASK_MEMTYPE); if (mem_type == TTM_PL_SYSTEM) { mem->mem_type = mem_type; mem->placement = cur_flags; mem->mm_node = NULL; return 0; } ret = ttm_bo_mem_force_space(bo, mem_type, place, mem, interruptible, no_wait_gpu); if (ret == 0 && mem->mm_node) { mem->placement = cur_flags; return 0; } if (ret == -ERESTARTSYS) has_erestartsys = true; } if (!type_found) { printk(KERN_ERR TTM_PFX "No compatible memory type found.\n"); return -EINVAL; } return (has_erestartsys) ? -ERESTARTSYS : -ENOMEM; } EXPORT_SYMBOL(ttm_bo_mem_space); static int ttm_bo_move_buffer(struct ttm_buffer_object *bo, struct ttm_placement *placement, bool interruptible, bool no_wait_gpu) { int ret = 0; struct ttm_mem_reg mem; lockdep_assert_held(&bo->resv->lock.base); /* * FIXME: It's possible to pipeline buffer moves. * Have the driver move function wait for idle when necessary, * instead of doing it here. */ ret = ttm_bo_wait(bo, false, interruptible, no_wait_gpu); if (ret) return ret; mem.num_pages = bo->num_pages; mem.size = mem.num_pages << PAGE_SHIFT; mem.page_alignment = bo->mem.page_alignment; mem.bus.io_reserved_vm = false; mem.bus.io_reserved_count = 0; /* * Determine where to move the buffer. */ ret = ttm_bo_mem_space(bo, placement, &mem, interruptible, no_wait_gpu); if (ret) goto out_unlock; ret = ttm_bo_handle_move_mem(bo, &mem, false, interruptible, no_wait_gpu); out_unlock: if (ret && mem.mm_node) ttm_bo_mem_put(bo, &mem); return ret; } bool ttm_bo_mem_compat(struct ttm_placement *placement, struct ttm_mem_reg *mem, uint32_t *new_flags) { int i; for (i = 0; i < placement->num_placement; i++) { const struct ttm_place *heap = &placement->placement[i]; if (mem->mm_node && (mem->start < heap->fpfn || (heap->lpfn != 0 && (mem->start + mem->num_pages) > heap->lpfn))) continue; *new_flags = heap->flags; if ((*new_flags & mem->placement & TTM_PL_MASK_CACHING) && (*new_flags & mem->placement & TTM_PL_MASK_MEM)) return true; } for (i = 0; i < placement->num_busy_placement; i++) { const struct ttm_place *heap = &placement->busy_placement[i]; if (mem->mm_node && (mem->start < heap->fpfn || (heap->lpfn != 0 && (mem->start + mem->num_pages) > heap->lpfn))) continue; *new_flags = heap->flags; if ((*new_flags & mem->placement & TTM_PL_MASK_CACHING) && (*new_flags & mem->placement & TTM_PL_MASK_MEM)) return true; } return false; } EXPORT_SYMBOL(ttm_bo_mem_compat); int ttm_bo_validate(struct ttm_buffer_object *bo, struct ttm_placement *placement, bool interruptible, bool no_wait_gpu) { int ret; uint32_t new_flags; lockdep_assert_held(&bo->resv->lock.base); /* * Check whether we need to move buffer. */ if (!ttm_bo_mem_compat(placement, &bo->mem, &new_flags)) { ret = ttm_bo_move_buffer(bo, placement, interruptible, no_wait_gpu); if (ret) return ret; } else { /* * Use the access and other non-mapping-related flag bits from * the compatible memory placement flags to the active flags */ ttm_flag_masked(&bo->mem.placement, new_flags, ~TTM_PL_MASK_MEMTYPE); } /* * We might need to add a TTM. */ if (bo->mem.mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) { ret = ttm_bo_add_ttm(bo, true); if (ret) return ret; } return 0; } EXPORT_SYMBOL(ttm_bo_validate); int ttm_bo_init(struct ttm_bo_device *bdev, struct ttm_buffer_object *bo, unsigned long size, enum ttm_bo_type type, struct ttm_placement *placement, uint32_t page_alignment, bool interruptible, struct file *persistent_swap_storage, size_t acc_size, struct sg_table *sg, struct reservation_object *resv, void (*destroy) (struct ttm_buffer_object *)) { int ret = 0; unsigned long num_pages; bool locked; num_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; if (num_pages == 0) { pr_err("Illegal buffer object size\n"); if (destroy) (*destroy)(bo); else kfree(bo); return -EINVAL; } bo->destroy = destroy; kref_init(&bo->kref); kref_init(&bo->list_kref); atomic_set(&bo->cpu_writers, 0); INIT_LIST_HEAD(&bo->lru); INIT_LIST_HEAD(&bo->ddestroy); INIT_LIST_HEAD(&bo->swap); INIT_LIST_HEAD(&bo->io_reserve_lru); mutex_init(&bo->wu_mutex); bo->bdev = bdev; bo->glob = bdev->glob; bo->type = type; bo->num_pages = num_pages; bo->mem.size = num_pages << PAGE_SHIFT; bo->mem.mem_type = TTM_PL_SYSTEM; bo->mem.num_pages = bo->num_pages; bo->mem.mm_node = NULL; bo->mem.page_alignment = page_alignment; bo->mem.bus.io_reserved_vm = false; bo->mem.bus.io_reserved_count = 0; bo->priv_flags = 0; bo->mem.placement = (TTM_PL_FLAG_SYSTEM | TTM_PL_FLAG_CACHED); bo->persistent_swap_storage = persistent_swap_storage; bo->acc_size = acc_size; bo->sg = sg; if (resv) { bo->resv = resv; lockdep_assert_held(&bo->resv->lock.base); } else { bo->resv = &bo->ttm_resv; reservation_object_init(&bo->ttm_resv); } atomic_inc(&bo->glob->bo_count); drm_vma_node_reset(&bo->vma_node); /* * For ttm_bo_type_device buffers, allocate * address space from the device. */ if (bo->type == ttm_bo_type_device || bo->type == ttm_bo_type_sg) ret = drm_vma_offset_add(&bdev->vma_manager, &bo->vma_node, bo->mem.num_pages); /* passed reservation objects should already be locked, * since otherwise lockdep will be angered in radeon. */ if (!resv) { locked = ww_mutex_trylock(&bo->resv->lock); WARN_ON(!locked); } if (likely(!ret)) ret = ttm_bo_validate(bo, placement, interruptible, false); if (!resv) ttm_bo_unreserve(bo); if (unlikely(ret)) ttm_bo_unref(&bo); return ret; } EXPORT_SYMBOL(ttm_bo_init); size_t ttm_bo_acc_size(struct ttm_bo_device *bdev, unsigned long bo_size, unsigned struct_size) { unsigned npages = (PAGE_ALIGN(bo_size)) >> PAGE_SHIFT; size_t size = 0; size += ttm_round_pot(struct_size); size += PAGE_ALIGN(npages * sizeof(void *)); size += ttm_round_pot(sizeof(struct ttm_tt)); return size; } EXPORT_SYMBOL(ttm_bo_acc_size); size_t ttm_bo_dma_acc_size(struct ttm_bo_device *bdev, unsigned long bo_size, unsigned struct_size) { unsigned npages = (PAGE_ALIGN(bo_size)) >> PAGE_SHIFT; size_t size = 0; size += ttm_round_pot(struct_size); size += PAGE_ALIGN(npages * sizeof(void *)); size += PAGE_ALIGN(npages * sizeof(dma_addr_t)); size += ttm_round_pot(sizeof(struct ttm_dma_tt)); return size; } EXPORT_SYMBOL(ttm_bo_dma_acc_size); int ttm_bo_create(struct ttm_bo_device *bdev, unsigned long size, enum ttm_bo_type type, struct ttm_placement *placement, uint32_t page_alignment, bool interruptible, struct file *persistent_swap_storage, struct ttm_buffer_object **p_bo) { struct ttm_buffer_object *bo; size_t acc_size; int ret; bo = kzalloc(sizeof(*bo), GFP_KERNEL); if (unlikely(bo == NULL)) return -ENOMEM; acc_size = ttm_bo_acc_size(bdev, size, sizeof(struct ttm_buffer_object)); ret = ttm_bo_init(bdev, bo, size, type, placement, page_alignment, interruptible, persistent_swap_storage, acc_size, NULL, NULL, NULL); if (likely(ret == 0)) *p_bo = bo; return ret; } EXPORT_SYMBOL(ttm_bo_create); static int ttm_bo_force_list_clean(struct ttm_bo_device *bdev, unsigned mem_type, bool allow_errors) { struct ttm_mem_type_manager *man = &bdev->man[mem_type]; struct ttm_bo_global *glob = bdev->glob; int ret; /* * Can't use standard list traversal since we're unlocking. */ spin_lock(&glob->lru_lock); while (!list_empty(&man->lru)) { spin_unlock(&glob->lru_lock); ret = ttm_mem_evict_first(bdev, mem_type, NULL, false, false); if (ret) { if (allow_errors) { return ret; } else { pr_err("Cleanup eviction failed\n"); } } spin_lock(&glob->lru_lock); } spin_unlock(&glob->lru_lock); return 0; } int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type, unsigned long p_size) { int ret = -EINVAL; struct ttm_mem_type_manager *man; BUG_ON(type >= TTM_NUM_MEM_TYPES); man = &bdev->man[type]; BUG_ON(man->has_type); man->io_reserve_fastpath = true; man->use_io_reserve_lru = false; mutex_init(&man->io_reserve_mutex); INIT_LIST_HEAD(&man->io_reserve_lru); ret = bdev->driver->init_mem_type(bdev, type, man); if (ret) return ret; man->bdev = bdev; ret = 0; if (type != TTM_PL_SYSTEM) { ret = (*man->func->init)(man, p_size); if (ret) return ret; } man->has_type = true; man->use_type = true; man->size = p_size; INIT_LIST_HEAD(&man->lru); return 0; } EXPORT_SYMBOL(ttm_bo_init_mm); void ttm_bo_global_release(struct drm_global_reference *ref) { struct ttm_bo_global *glob = ref->object; } EXPORT_SYMBOL(ttm_bo_global_release); int ttm_bo_global_init(struct drm_global_reference *ref) { struct ttm_bo_global_ref *bo_ref = container_of(ref, struct ttm_bo_global_ref, ref); struct ttm_bo_global *glob = ref->object; int ret; mutex_init(&glob->device_list_mutex); spin_lock_init(&glob->lru_lock); glob->mem_glob = bo_ref->mem_glob; glob->dummy_read_page = alloc_page(__GFP_ZERO | GFP_DMA32); if (unlikely(glob->dummy_read_page == NULL)) { ret = -ENOMEM; goto out_no_drp; } INIT_LIST_HEAD(&glob->swap_lru); INIT_LIST_HEAD(&glob->device_list); atomic_set(&glob->bo_count, 0); return 0; out_no_drp: kfree(glob); return ret; } EXPORT_SYMBOL(ttm_bo_global_init); int ttm_bo_device_init(struct ttm_bo_device *bdev, struct ttm_bo_global *glob, struct ttm_bo_driver *driver, struct address_space *mapping, uint64_t file_page_offset, bool need_dma32) { int ret = -EINVAL; bdev->driver = driver; memset(bdev->man, 0, sizeof(bdev->man)); /* * Initialize the system memory buffer type. * Other types need to be driver / IOCTL initialized. */ ret = ttm_bo_init_mm(bdev, TTM_PL_SYSTEM, 0); if (unlikely(ret != 0)) goto out_no_sys; drm_vma_offset_manager_init(&bdev->vma_manager, file_page_offset, 0x10000000); INIT_DELAYED_WORK(&bdev->wq, ttm_bo_delayed_workqueue); INIT_LIST_HEAD(&bdev->ddestroy); bdev->dev_mapping = mapping; bdev->glob = glob; bdev->need_dma32 = need_dma32; bdev->val_seq = 0; mutex_lock(&glob->device_list_mutex); list_add_tail(&bdev->device_list, &glob->device_list); mutex_unlock(&glob->device_list_mutex); return 0; out_no_sys: return ret; } EXPORT_SYMBOL(ttm_bo_device_init); /* * buffer object vm functions. */ bool ttm_mem_reg_is_pci(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem) { struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type]; if (!(man->flags & TTM_MEMTYPE_FLAG_FIXED)) { if (mem->mem_type == TTM_PL_SYSTEM) return false; if (man->flags & TTM_MEMTYPE_FLAG_CMA) return false; if (mem->placement & TTM_PL_FLAG_CACHED) return false; } return true; } void ttm_bo_unmap_virtual_locked(struct ttm_buffer_object *bo) { struct ttm_bo_device *bdev = bo->bdev; drm_vma_node_unmap(&bo->vma_node, bdev->dev_mapping); ttm_mem_io_free_vm(bo); } void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo) { struct ttm_bo_device *bdev = bo->bdev; struct ttm_mem_type_manager *man = &bdev->man[bo->mem.mem_type]; ttm_mem_io_lock(man, false); ttm_bo_unmap_virtual_locked(bo); ttm_mem_io_unlock(man); } EXPORT_SYMBOL(ttm_bo_unmap_virtual); int ttm_bo_wait(struct ttm_buffer_object *bo, bool lazy, bool interruptible, bool no_wait) { struct reservation_object_list *fobj; struct reservation_object *resv; struct fence *excl; long timeout = 15 * HZ; int i; resv = bo->resv; fobj = reservation_object_get_list(resv); excl = reservation_object_get_excl(resv); if (excl) { if (!fence_is_signaled(excl)) { if (no_wait) return -EBUSY; timeout = fence_wait_timeout(excl, interruptible, timeout); } } for (i = 0; fobj && timeout > 0 && i < fobj->shared_count; ++i) { struct fence *fence; fence = rcu_dereference_protected(fobj->shared[i], reservation_object_held(resv)); if (!fence_is_signaled(fence)) { if (no_wait) return -EBUSY; timeout = fence_wait_timeout(fence, interruptible, timeout); } } if (timeout < 0) return timeout; if (timeout == 0) return -EBUSY; reservation_object_add_excl_fence(resv, NULL); clear_bit(TTM_BO_PRIV_FLAG_MOVING, &bo->priv_flags); return 0; } EXPORT_SYMBOL(ttm_bo_wait); int ttm_bo_synccpu_write_grab(struct ttm_buffer_object *bo, bool no_wait) { int ret = 0; /* * Using ttm_bo_reserve makes sure the lru lists are updated. */ ret = ttm_bo_reserve(bo, true, no_wait, false, NULL); if (unlikely(ret != 0)) return ret; ret = ttm_bo_wait(bo, false, true, no_wait); if (likely(ret == 0)) atomic_inc(&bo->cpu_writers); ttm_bo_unreserve(bo); return ret; } EXPORT_SYMBOL(ttm_bo_synccpu_write_grab); void ttm_bo_synccpu_write_release(struct ttm_buffer_object *bo) { atomic_dec(&bo->cpu_writers); } int ttm_bo_wait_unreserved(struct ttm_buffer_object *bo) { int ret; /* * In the absense of a wait_unlocked API, * Use the bo::wu_mutex to avoid triggering livelocks due to * concurrent use of this function. Note that this use of * bo::wu_mutex can go away if we change locking order to * mmap_sem -> bo::reserve. */ ret = mutex_lock_interruptible(&bo->wu_mutex); if (unlikely(ret != 0)) return -ERESTARTSYS; if (!ww_mutex_is_locked(&bo->resv->lock)) goto out_unlock; ret = __ttm_bo_reserve(bo, true, false, false, NULL); if (unlikely(ret != 0)) goto out_unlock; __ttm_bo_unreserve(bo); out_unlock: mutex_unlock(&bo->wu_mutex); return ret; }