From 72525b3f333de54fa0c42ef87f27861e41478f1e Mon Sep 17 00:00:00 2001 From: David Herrmann Date: Wed, 24 Jul 2013 21:08:53 +0200 Subject: [PATCH] drm/ttm: convert to unified vma offset manager Use the new vma-manager infrastructure. This doesn't change any implementation details as the vma-offset-manager is nearly copied 1-to-1 from TTM. The vm_lock is moved into the offset manager so we can drop it from TTM. During lookup, we use the vma locking helpers to take a reference to the found object. In all other scenarios, locking stays the same as before. We always guarantee that drm_vma_offset_remove() is called only during destruction. Hence, helpers like drm_vma_node_offset_addr() are always safe as long as the node has a valid offset. This also drops the addr_space_offset member as it is a copy of vm_start in vma_node objects. Use the accessor functions instead. v4: - remove vm_lock - use drm_vma_offset_lock_lookup() to protect lookup (instead of vm_lock) Cc: Dave Airlie Cc: Ben Skeggs Cc: Maarten Lankhorst Cc: Martin Peres Cc: Alex Deucher Cc: Thomas Hellstrom Signed-off-by: David Herrmann Reviewed-by: Daniel Vetter Signed-off-by: Dave Airlie --- drivers/gpu/drm/ast/ast_main.c | 2 +- drivers/gpu/drm/cirrus/cirrus_main.c | 2 +- drivers/gpu/drm/mgag200/mgag200_main.c | 2 +- drivers/gpu/drm/nouveau/nouveau_display.c | 2 +- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 +- drivers/gpu/drm/qxl/qxl_object.h | 2 +- drivers/gpu/drm/qxl/qxl_release.c | 2 +- drivers/gpu/drm/radeon/radeon_object.h | 5 +- drivers/gpu/drm/ttm/ttm_bo.c | 89 ++++------------------- drivers/gpu/drm/ttm/ttm_bo_util.c | 3 +- drivers/gpu/drm/ttm/ttm_bo_vm.c | 81 ++++++++------------- drivers/gpu/drm/vmwgfx/vmwgfx_resource.c | 4 +- include/drm/ttm/ttm_bo_api.h | 15 +--- include/drm/ttm/ttm_bo_driver.h | 10 +-- 14 files changed, 66 insertions(+), 155 deletions(-) diff --git a/drivers/gpu/drm/ast/ast_main.c b/drivers/gpu/drm/ast/ast_main.c index f60fd7bd1183..c195dc2abc09 100644 --- a/drivers/gpu/drm/ast/ast_main.c +++ b/drivers/gpu/drm/ast/ast_main.c @@ -487,7 +487,7 @@ void ast_gem_free_object(struct drm_gem_object *obj) static inline u64 ast_bo_mmap_offset(struct ast_bo *bo) { - return bo->bo.addr_space_offset; + return drm_vma_node_offset_addr(&bo->bo.vma_node); } int ast_dumb_mmap_offset(struct drm_file *file, diff --git a/drivers/gpu/drm/cirrus/cirrus_main.c b/drivers/gpu/drm/cirrus/cirrus_main.c index 35cbae827771..3a7a0efe3675 100644 --- a/drivers/gpu/drm/cirrus/cirrus_main.c +++ b/drivers/gpu/drm/cirrus/cirrus_main.c @@ -294,7 +294,7 @@ void cirrus_gem_free_object(struct drm_gem_object *obj) static inline u64 cirrus_bo_mmap_offset(struct cirrus_bo *bo) { - return bo->bo.addr_space_offset; + return drm_vma_node_offset_addr(&bo->bo.vma_node); } int diff --git a/drivers/gpu/drm/mgag200/mgag200_main.c b/drivers/gpu/drm/mgag200/mgag200_main.c index 9fa5685baee0..1a75ea395b33 100644 --- a/drivers/gpu/drm/mgag200/mgag200_main.c +++ b/drivers/gpu/drm/mgag200/mgag200_main.c @@ -349,7 +349,7 @@ void mgag200_gem_free_object(struct drm_gem_object *obj) static inline u64 mgag200_bo_mmap_offset(struct mgag200_bo *bo) { - return bo->bo.addr_space_offset; + return drm_vma_node_offset_addr(&bo->bo.vma_node); } int diff --git a/drivers/gpu/drm/nouveau/nouveau_display.c b/drivers/gpu/drm/nouveau/nouveau_display.c index 708b2d1c0037..7a8caa126db6 100644 --- a/drivers/gpu/drm/nouveau/nouveau_display.c +++ b/drivers/gpu/drm/nouveau/nouveau_display.c @@ -697,7 +697,7 @@ nouveau_display_dumb_map_offset(struct drm_file *file_priv, gem = drm_gem_object_lookup(dev, file_priv, handle); if (gem) { struct nouveau_bo *bo = gem->driver_private; - *poffset = bo->bo.addr_space_offset; + *poffset = drm_vma_node_offset_addr(&bo->bo.vma_node); drm_gem_object_unreference_unlocked(gem); return 0; } diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c index e72d09c068a8..86597ebf8c98 100644 --- a/drivers/gpu/drm/nouveau/nouveau_gem.c +++ b/drivers/gpu/drm/nouveau/nouveau_gem.c @@ -226,7 +226,7 @@ nouveau_gem_info(struct drm_file *file_priv, struct drm_gem_object *gem, } rep->size = nvbo->bo.mem.num_pages << PAGE_SHIFT; - rep->map_handle = nvbo->bo.addr_space_offset; + rep->map_handle = drm_vma_node_offset_addr(&nvbo->bo.vma_node); rep->tile_mode = nvbo->tile_mode; rep->tile_flags = nvbo->tile_flags; return 0; diff --git a/drivers/gpu/drm/qxl/qxl_object.h b/drivers/gpu/drm/qxl/qxl_object.h index ee7ad79ce781..af10165adb0d 100644 --- a/drivers/gpu/drm/qxl/qxl_object.h +++ b/drivers/gpu/drm/qxl/qxl_object.h @@ -59,7 +59,7 @@ static inline unsigned long qxl_bo_size(struct qxl_bo *bo) static inline u64 qxl_bo_mmap_offset(struct qxl_bo *bo) { - return bo->tbo.addr_space_offset; + return drm_vma_node_offset_addr(&bo->tbo.vma_node); } static inline int qxl_bo_wait(struct qxl_bo *bo, u32 *mem_type, diff --git a/drivers/gpu/drm/qxl/qxl_release.c b/drivers/gpu/drm/qxl/qxl_release.c index b443d6751d5f..1a648e1da6a6 100644 --- a/drivers/gpu/drm/qxl/qxl_release.c +++ b/drivers/gpu/drm/qxl/qxl_release.c @@ -87,7 +87,7 @@ qxl_release_free(struct qxl_device *qdev, for (i = 0 ; i < release->bo_count; ++i) { QXL_INFO(qdev, "release %llx\n", - release->bos[i]->tbo.addr_space_offset + drm_vma_node_offset_addr(&release->bos[i]->tbo.vma_node) - DRM_FILE_OFFSET); qxl_fence_remove_release(&release->bos[i]->fence, release->id); qxl_bo_unref(&release->bos[i]); diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h index 49c82c480013..209b11150263 100644 --- a/drivers/gpu/drm/radeon/radeon_object.h +++ b/drivers/gpu/drm/radeon/radeon_object.h @@ -113,13 +113,10 @@ static inline unsigned radeon_bo_gpu_page_alignment(struct radeon_bo *bo) * @bo: radeon object for which we query the offset * * Returns mmap offset of the object. - * - * Note: addr_space_offset is constant after ttm bo init thus isn't protected - * by any lock. */ static inline u64 radeon_bo_mmap_offset(struct radeon_bo *bo) { - return bo->tbo.addr_space_offset; + return drm_vma_node_offset_addr(&bo->tbo.vma_node); } extern int radeon_bo_wait(struct radeon_bo *bo, u32 *mem_type, diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index cb9dd674670c..3dc08b612292 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -615,13 +615,7 @@ static void ttm_bo_release(struct kref *kref) struct ttm_bo_device *bdev = bo->bdev; struct ttm_mem_type_manager *man = &bdev->man[bo->mem.mem_type]; - write_lock(&bdev->vm_lock); - if (likely(bo->vm_node != NULL)) { - rb_erase(&bo->vm_rb, &bdev->addr_space_rb); - drm_mm_put_block(bo->vm_node); - bo->vm_node = NULL; - } - write_unlock(&bdev->vm_lock); + 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); @@ -1129,6 +1123,7 @@ int ttm_bo_init(struct ttm_bo_device *bdev, bo->resv = &bo->ttm_resv; reservation_object_init(bo->resv); atomic_inc(&bo->glob->bo_count); + drm_vma_node_reset(&bo->vma_node); ret = ttm_bo_check_placement(bo, placement); @@ -1424,10 +1419,7 @@ int ttm_bo_device_release(struct ttm_bo_device *bdev) TTM_DEBUG("Swap list was clean\n"); spin_unlock(&glob->lru_lock); - BUG_ON(!drm_mm_clean(&bdev->addr_space_mm)); - write_lock(&bdev->vm_lock); - drm_mm_takedown(&bdev->addr_space_mm); - write_unlock(&bdev->vm_lock); + drm_vma_offset_manager_destroy(&bdev->vma_manager); return ret; } @@ -1441,7 +1433,6 @@ int ttm_bo_device_init(struct ttm_bo_device *bdev, { int ret = -EINVAL; - rwlock_init(&bdev->vm_lock); bdev->driver = driver; memset(bdev->man, 0, sizeof(bdev->man)); @@ -1454,9 +1445,8 @@ int ttm_bo_device_init(struct ttm_bo_device *bdev, if (unlikely(ret != 0)) goto out_no_sys; - bdev->addr_space_rb = RB_ROOT; - drm_mm_init(&bdev->addr_space_mm, file_page_offset, 0x10000000); - + 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 = NULL; @@ -1498,12 +1488,17 @@ bool ttm_mem_reg_is_pci(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem) void ttm_bo_unmap_virtual_locked(struct ttm_buffer_object *bo) { struct ttm_bo_device *bdev = bo->bdev; - loff_t offset = (loff_t) bo->addr_space_offset; - loff_t holelen = ((loff_t) bo->mem.num_pages) << PAGE_SHIFT; + loff_t offset, holelen; if (!bdev->dev_mapping) return; - unmap_mapping_range(bdev->dev_mapping, offset, holelen, 1); + + if (drm_vma_node_has_offset(&bo->vma_node)) { + offset = (loff_t) drm_vma_node_offset_addr(&bo->vma_node); + holelen = ((loff_t) bo->mem.num_pages) << PAGE_SHIFT; + + unmap_mapping_range(bdev->dev_mapping, offset, holelen, 1); + } ttm_mem_io_free_vm(bo); } @@ -1520,31 +1515,6 @@ void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo) EXPORT_SYMBOL(ttm_bo_unmap_virtual); -static void ttm_bo_vm_insert_rb(struct ttm_buffer_object *bo) -{ - struct ttm_bo_device *bdev = bo->bdev; - struct rb_node **cur = &bdev->addr_space_rb.rb_node; - struct rb_node *parent = NULL; - struct ttm_buffer_object *cur_bo; - unsigned long offset = bo->vm_node->start; - unsigned long cur_offset; - - while (*cur) { - parent = *cur; - cur_bo = rb_entry(parent, struct ttm_buffer_object, vm_rb); - cur_offset = cur_bo->vm_node->start; - if (offset < cur_offset) - cur = &parent->rb_left; - else if (offset > cur_offset) - cur = &parent->rb_right; - else - BUG(); - } - - rb_link_node(&bo->vm_rb, parent, cur); - rb_insert_color(&bo->vm_rb, &bdev->addr_space_rb); -} - /** * ttm_bo_setup_vm: * @@ -1559,38 +1529,9 @@ static void ttm_bo_vm_insert_rb(struct ttm_buffer_object *bo) static int ttm_bo_setup_vm(struct ttm_buffer_object *bo) { struct ttm_bo_device *bdev = bo->bdev; - int ret; - -retry_pre_get: - ret = drm_mm_pre_get(&bdev->addr_space_mm); - if (unlikely(ret != 0)) - return ret; - - write_lock(&bdev->vm_lock); - bo->vm_node = drm_mm_search_free(&bdev->addr_space_mm, - bo->mem.num_pages, 0, 0); - - if (unlikely(bo->vm_node == NULL)) { - ret = -ENOMEM; - goto out_unlock; - } - bo->vm_node = drm_mm_get_block_atomic(bo->vm_node, - bo->mem.num_pages, 0); - - if (unlikely(bo->vm_node == NULL)) { - write_unlock(&bdev->vm_lock); - goto retry_pre_get; - } - - ttm_bo_vm_insert_rb(bo); - write_unlock(&bdev->vm_lock); - bo->addr_space_offset = ((uint64_t) bo->vm_node->start) << PAGE_SHIFT; - - return 0; -out_unlock: - write_unlock(&bdev->vm_lock); - return ret; + return drm_vma_offset_add(&bdev->vma_manager, &bo->vma_node, + bo->mem.num_pages); } int ttm_bo_wait(struct ttm_buffer_object *bo, diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c index 319cf4127c5b..7cc904d3a4d1 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_util.c +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c @@ -30,6 +30,7 @@ #include #include +#include #include #include #include @@ -450,7 +451,7 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo, INIT_LIST_HEAD(&fbo->lru); INIT_LIST_HEAD(&fbo->swap); INIT_LIST_HEAD(&fbo->io_reserve_lru); - fbo->vm_node = NULL; + drm_vma_node_reset(&fbo->vma_node); atomic_set(&fbo->cpu_writers, 0); spin_lock(&bdev->fence_lock); diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c index 3df9f16b041c..8c0e2c020215 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_vm.c +++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c @@ -33,6 +33,7 @@ #include #include #include +#include #include #include #include @@ -40,37 +41,6 @@ #define TTM_BO_VM_NUM_PREFAULT 16 -static struct ttm_buffer_object *ttm_bo_vm_lookup_rb(struct ttm_bo_device *bdev, - unsigned long page_start, - unsigned long num_pages) -{ - struct rb_node *cur = bdev->addr_space_rb.rb_node; - unsigned long cur_offset; - struct ttm_buffer_object *bo; - struct ttm_buffer_object *best_bo = NULL; - - while (likely(cur != NULL)) { - bo = rb_entry(cur, struct ttm_buffer_object, vm_rb); - cur_offset = bo->vm_node->start; - if (page_start >= cur_offset) { - cur = cur->rb_right; - best_bo = bo; - if (page_start == cur_offset) - break; - } else - cur = cur->rb_left; - } - - if (unlikely(best_bo == NULL)) - return NULL; - - if (unlikely((best_bo->vm_node->start + best_bo->num_pages) < - (page_start + num_pages))) - return NULL; - - return best_bo; -} - static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) { struct ttm_buffer_object *bo = (struct ttm_buffer_object *) @@ -146,9 +116,9 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) } page_offset = ((address - vma->vm_start) >> PAGE_SHIFT) + - bo->vm_node->start - vma->vm_pgoff; + drm_vma_node_start(&bo->vma_node) - vma->vm_pgoff; page_last = vma_pages(vma) + - bo->vm_node->start - vma->vm_pgoff; + drm_vma_node_start(&bo->vma_node) - vma->vm_pgoff; if (unlikely(page_offset >= bo->num_pages)) { retval = VM_FAULT_SIGBUS; @@ -249,6 +219,30 @@ static const struct vm_operations_struct ttm_bo_vm_ops = { .close = ttm_bo_vm_close }; +static struct ttm_buffer_object *ttm_bo_vm_lookup(struct ttm_bo_device *bdev, + unsigned long offset, + unsigned long pages) +{ + struct drm_vma_offset_node *node; + struct ttm_buffer_object *bo = NULL; + + drm_vma_offset_lock_lookup(&bdev->vma_manager); + + node = drm_vma_offset_lookup_locked(&bdev->vma_manager, offset, pages); + if (likely(node)) { + bo = container_of(node, struct ttm_buffer_object, vma_node); + if (!kref_get_unless_zero(&bo->kref)) + bo = NULL; + } + + drm_vma_offset_unlock_lookup(&bdev->vma_manager); + + if (!bo) + pr_err("Could not find buffer object to map\n"); + + return bo; +} + int ttm_bo_mmap(struct file *filp, struct vm_area_struct *vma, struct ttm_bo_device *bdev) { @@ -256,17 +250,9 @@ int ttm_bo_mmap(struct file *filp, struct vm_area_struct *vma, struct ttm_buffer_object *bo; int ret; - read_lock(&bdev->vm_lock); - bo = ttm_bo_vm_lookup_rb(bdev, vma->vm_pgoff, - vma_pages(vma)); - if (likely(bo != NULL) && !kref_get_unless_zero(&bo->kref)) - bo = NULL; - read_unlock(&bdev->vm_lock); - - if (unlikely(bo == NULL)) { - pr_err("Could not find buffer object to map\n"); + bo = ttm_bo_vm_lookup(bdev, vma->vm_pgoff, vma_pages(vma)); + if (unlikely(!bo)) return -EINVAL; - } driver = bo->bdev->driver; if (unlikely(!driver->verify_access)) { @@ -324,12 +310,7 @@ ssize_t ttm_bo_io(struct ttm_bo_device *bdev, struct file *filp, bool no_wait = false; bool dummy; - read_lock(&bdev->vm_lock); - bo = ttm_bo_vm_lookup_rb(bdev, dev_offset, 1); - if (likely(bo != NULL)) - ttm_bo_reference(bo); - read_unlock(&bdev->vm_lock); - + bo = ttm_bo_vm_lookup(bdev, dev_offset, 1); if (unlikely(bo == NULL)) return -EFAULT; @@ -343,7 +324,7 @@ ssize_t ttm_bo_io(struct ttm_bo_device *bdev, struct file *filp, if (unlikely(ret != 0)) goto out_unref; - kmap_offset = dev_offset - bo->vm_node->start; + kmap_offset = dev_offset - drm_vma_node_start(&bo->vma_node); if (unlikely(kmap_offset >= bo->num_pages)) { ret = -EFBIG; goto out_unref; diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c index 7953d1f90b63..0e67cf41065d 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c @@ -500,7 +500,7 @@ int vmw_dmabuf_alloc_ioctl(struct drm_device *dev, void *data, goto out_no_dmabuf; rep->handle = handle; - rep->map_handle = dma_buf->base.addr_space_offset; + rep->map_handle = drm_vma_node_offset_addr(&dma_buf->base.vma_node); rep->cur_gmr_id = handle; rep->cur_gmr_offset = 0; @@ -834,7 +834,7 @@ int vmw_dumb_map_offset(struct drm_file *file_priv, if (ret != 0) return -EINVAL; - *offset = out_buf->base.addr_space_offset; + *offset = drm_vma_node_offset_addr(&out_buf->base.vma_node); vmw_dmabuf_unreference(&out_buf); return 0; } diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h index 8a6aa56ece52..751eaffbf0d5 100644 --- a/include/drm/ttm/ttm_bo_api.h +++ b/include/drm/ttm/ttm_bo_api.h @@ -32,12 +32,12 @@ #define _TTM_BO_API_H_ #include +#include #include #include #include #include #include -#include #include #include @@ -145,7 +145,6 @@ struct ttm_tt; * @type: The bo type. * @destroy: Destruction function. If NULL, kfree is used. * @num_pages: Actual number of pages. - * @addr_space_offset: Address space offset. * @acc_size: Accounted size for this object. * @kref: Reference count of this buffer object. When this refcount reaches * zero, the object is put on the delayed delete list. @@ -166,8 +165,7 @@ struct ttm_tt; * @swap: List head for swap LRU list. * @sync_obj: Pointer to a synchronization object. * @priv_flags: Flags describing buffer object internal state. - * @vm_rb: Rb node for the vm rb tree. - * @vm_node: Address space manager node. + * @vma_node: Address space manager node. * @offset: The current GPU offset, which can have different meanings * depending on the memory type. For SYSTEM type memory, it should be 0. * @cur_placement: Hint of current placement. @@ -194,7 +192,6 @@ struct ttm_buffer_object { enum ttm_bo_type type; void (*destroy) (struct ttm_buffer_object *); unsigned long num_pages; - uint64_t addr_space_offset; size_t acc_size; /** @@ -238,13 +235,7 @@ struct ttm_buffer_object { void *sync_obj; unsigned long priv_flags; - /** - * Members protected by the bdev::vm_lock - */ - - struct rb_node vm_rb; - struct drm_mm_node *vm_node; - + struct drm_vma_offset_node vma_node; /** * Special members that are protected by the reserve lock diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h index 984fc2d571a1..8639c85d61c4 100644 --- a/include/drm/ttm/ttm_bo_driver.h +++ b/include/drm/ttm/ttm_bo_driver.h @@ -36,6 +36,7 @@ #include #include #include +#include #include #include #include @@ -519,7 +520,7 @@ struct ttm_bo_global { * @man: An array of mem_type_managers. * @fence_lock: Protects the synchronizing members on *all* bos belonging * to this device. - * @addr_space_mm: Range manager for the device address space. + * @vma_manager: Address space manager * lru_lock: Spinlock that protects the buffer+device lru lists and * ddestroy lists. * @val_seq: Current validation sequence. @@ -537,14 +538,13 @@ struct ttm_bo_device { struct list_head device_list; struct ttm_bo_global *glob; struct ttm_bo_driver *driver; - rwlock_t vm_lock; struct ttm_mem_type_manager man[TTM_NUM_MEM_TYPES]; spinlock_t fence_lock; + /* - * Protected by the vm lock. + * Protected by internal locks. */ - struct rb_root addr_space_rb; - struct drm_mm addr_space_mm; + struct drm_vma_offset_manager vma_manager; /* * Protected by the global:lru lock. -- 2.39.5