diff options
Diffstat (limited to 'drivers/gpu/drm/ttm')
| -rw-r--r-- | drivers/gpu/drm/ttm/ttm_bo.c | 78 | ||||
| -rw-r--r-- | drivers/gpu/drm/ttm/ttm_bo_manager.c | 5 | ||||
| -rw-r--r-- | drivers/gpu/drm/ttm/ttm_bo_util.c | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/ttm/ttm_bo_vm.c | 7 | ||||
| -rw-r--r-- | drivers/gpu/drm/ttm/ttm_page_alloc.c | 45 | ||||
| -rw-r--r-- | drivers/gpu/drm/ttm/ttm_page_alloc_dma.c | 42 |
6 files changed, 126 insertions, 53 deletions
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index 22b57020790d..d79607a1187c 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -70,6 +70,7 @@ static inline int ttm_mem_type_from_place(const struct ttm_place *place, static void ttm_mem_type_debug(struct ttm_bo_device *bdev, int mem_type) { struct ttm_mem_type_manager *man = &bdev->man[mem_type]; + struct drm_printer p = drm_debug_printer(TTM_PFX); pr_err(" has_type: %d\n", man->has_type); pr_err(" use_type: %d\n", man->use_type); @@ -79,7 +80,7 @@ static void ttm_mem_type_debug(struct ttm_bo_device *bdev, int mem_type) pr_err(" available_caching: 0x%08X\n", man->available_caching); pr_err(" default_caching: 0x%08X\n", man->default_caching); if (mem_type != TTM_PL_SYSTEM) - (*man->func->debug)(man, TTM_PFX); + (*man->func->debug)(man, &p); } static void ttm_bo_mem_space_debug(struct ttm_buffer_object *bo, @@ -108,8 +109,8 @@ static ssize_t ttm_bo_global_show(struct kobject *kobj, struct ttm_bo_global *glob = container_of(kobj, struct ttm_bo_global, kobj); - return snprintf(buffer, PAGE_SIZE, "%lu\n", - (unsigned long) atomic_read(&glob->bo_count)); + return snprintf(buffer, PAGE_SIZE, "%d\n", + atomic_read(&glob->bo_count)); } static struct attribute *ttm_bo_global_attrs[] = { @@ -394,14 +395,33 @@ static void ttm_bo_cleanup_memtype_use(struct ttm_buffer_object *bo) ww_mutex_unlock (&bo->resv->lock); } +static int ttm_bo_individualize_resv(struct ttm_buffer_object *bo) +{ + int r; + + if (bo->resv == &bo->ttm_resv) + return 0; + + reservation_object_init(&bo->ttm_resv); + BUG_ON(!reservation_object_trylock(&bo->ttm_resv)); + + r = reservation_object_copy_fences(&bo->ttm_resv, bo->resv); + if (r) { + reservation_object_unlock(&bo->ttm_resv); + reservation_object_fini(&bo->ttm_resv); + } + + return r; +} + static void ttm_bo_flush_all_fences(struct ttm_buffer_object *bo) { struct reservation_object_list *fobj; struct dma_fence *fence; int i; - fobj = reservation_object_get_list(bo->resv); - fence = reservation_object_get_excl(bo->resv); + fobj = reservation_object_get_list(&bo->ttm_resv); + fence = reservation_object_get_excl(&bo->ttm_resv); if (fence && !fence->ops->signaled) dma_fence_enable_sw_signaling(fence); @@ -420,18 +440,30 @@ static void ttm_bo_cleanup_refs_or_queue(struct ttm_buffer_object *bo) struct ttm_bo_global *glob = bo->glob; int ret; + ret = ttm_bo_individualize_resv(bo); + if (ret) { + /* Last resort, if we fail to allocate memory for the + * fences block for the BO to become idle + */ + reservation_object_wait_timeout_rcu(bo->resv, true, false, + 30 * HZ); + spin_lock(&glob->lru_lock); + goto error; + } + spin_lock(&glob->lru_lock); ret = __ttm_bo_reserve(bo, false, true, NULL); - if (!ret) { - if (!ttm_bo_wait(bo, false, true)) { + if (reservation_object_test_signaled_rcu(&bo->ttm_resv, true)) { ttm_bo_del_from_lru(bo); spin_unlock(&glob->lru_lock); + if (bo->resv != &bo->ttm_resv) + reservation_object_unlock(&bo->ttm_resv); ttm_bo_cleanup_memtype_use(bo); - return; - } else - ttm_bo_flush_all_fences(bo); + } + + ttm_bo_flush_all_fences(bo); /* * Make NO_EVICT bos immediately available to @@ -445,7 +477,10 @@ static void ttm_bo_cleanup_refs_or_queue(struct ttm_buffer_object *bo) __ttm_bo_unreserve(bo); } + if (bo->resv != &bo->ttm_resv) + reservation_object_unlock(&bo->ttm_resv); +error: kref_get(&bo->list_kref); list_add_tail(&bo->ddestroy, &bdev->ddestroy); spin_unlock(&glob->lru_lock); @@ -471,17 +506,25 @@ static int ttm_bo_cleanup_refs_and_unlock(struct ttm_buffer_object *bo, bool no_wait_gpu) { struct ttm_bo_global *glob = bo->glob; + struct reservation_object *resv; int ret; - ret = ttm_bo_wait(bo, false, true); + if (unlikely(list_empty(&bo->ddestroy))) + resv = bo->resv; + else + resv = &bo->ttm_resv; + + if (reservation_object_test_signaled_rcu(resv, true)) + ret = 0; + else + ret = -EBUSY; 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, + lret = reservation_object_wait_timeout_rcu(resv, true, interruptible, 30 * HZ); @@ -505,13 +548,6 @@ static int ttm_bo_cleanup_refs_and_unlock(struct ttm_buffer_object *bo, 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, true); - WARN_ON(ret); } if (ret || unlikely(list_empty(&bo->ddestroy))) { @@ -521,6 +557,8 @@ static int ttm_bo_cleanup_refs_and_unlock(struct ttm_buffer_object *bo, } ttm_bo_del_from_lru(bo); + if (!list_empty(&bo->ddestroy) && (bo->resv != &bo->ttm_resv)) + reservation_object_fini(&bo->ttm_resv); list_del_init(&bo->ddestroy); kref_put(&bo->list_kref, ttm_bo_ref_bug); diff --git a/drivers/gpu/drm/ttm/ttm_bo_manager.c b/drivers/gpu/drm/ttm/ttm_bo_manager.c index 90a6c0b03afc..a7c232dc39cb 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_manager.c +++ b/drivers/gpu/drm/ttm/ttm_bo_manager.c @@ -136,13 +136,12 @@ static int ttm_bo_man_takedown(struct ttm_mem_type_manager *man) } static void ttm_bo_man_debug(struct ttm_mem_type_manager *man, - const char *prefix) + struct drm_printer *printer) { struct ttm_range_manager *rman = (struct ttm_range_manager *) man->priv; - struct drm_printer p = drm_debug_printer(prefix); spin_lock(&rman->lock); - drm_mm_print(&rman->mm, &p); + drm_mm_print(&rman->mm, printer); spin_unlock(&rman->lock); } diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c index d0459b392e5e..78cb99be7146 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_util.c +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c @@ -469,6 +469,7 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo, * TODO: Explicit member copy would probably be better here. */ + atomic_inc(&bo->glob->bo_count); INIT_LIST_HEAD(&fbo->ddestroy); INIT_LIST_HEAD(&fbo->lru); INIT_LIST_HEAD(&fbo->swap); @@ -586,7 +587,6 @@ int ttm_bo_kmap(struct ttm_buffer_object *bo, unsigned long offset, size; int ret; - BUG_ON(!list_empty(&bo->swap)); map->virtual = NULL; map->bo = bo; if (num_pages > bo->num_pages) diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c index a01e5c90fd87..c8ebb757e36b 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_vm.c +++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c @@ -39,6 +39,7 @@ #include <linux/rbtree.h> #include <linux/module.h> #include <linux/uaccess.h> +#include <linux/mem_encrypt.h> #define TTM_BO_VM_NUM_PREFAULT 16 @@ -230,9 +231,11 @@ static int ttm_bo_vm_fault(struct vm_fault *vmf) * first page. */ for (i = 0; i < TTM_BO_VM_NUM_PREFAULT; ++i) { - if (bo->mem.bus.is_iomem) + if (bo->mem.bus.is_iomem) { + /* Iomem should not be marked encrypted */ + cvma.vm_page_prot = pgprot_decrypted(cvma.vm_page_prot); pfn = bdev->driver->io_mem_pfn(bo, page_offset); - else { + } else { page = ttm->pages[page_offset]; if (unlikely(!page && i == 0)) { retval = VM_FAULT_OOM; diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index eeddc1e48409..052e1f102113 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -615,7 +615,7 @@ static void ttm_page_pool_fill_locked(struct ttm_page_pool *pool, } else { pr_err("Failed to fill pool (%p)\n", pool); /* If we have any pages left put them to the pool. */ - list_for_each_entry(p, &pool->list, lru) { + list_for_each_entry(p, &new_pages, lru) { ++cpages; } list_splice(&new_pages, &pool->list); @@ -920,6 +920,49 @@ void ttm_pool_unpopulate(struct ttm_tt *ttm) } EXPORT_SYMBOL(ttm_pool_unpopulate); +#if defined(CONFIG_SWIOTLB) || defined(CONFIG_INTEL_IOMMU) +int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt) +{ + unsigned i; + int r; + + r = ttm_pool_populate(&tt->ttm); + if (r) + return r; + + for (i = 0; i < tt->ttm.num_pages; i++) { + tt->dma_address[i] = dma_map_page(dev, tt->ttm.pages[i], + 0, PAGE_SIZE, + DMA_BIDIRECTIONAL); + if (dma_mapping_error(dev, tt->dma_address[i])) { + while (i--) { + dma_unmap_page(dev, tt->dma_address[i], + PAGE_SIZE, DMA_BIDIRECTIONAL); + tt->dma_address[i] = 0; + } + ttm_pool_unpopulate(&tt->ttm); + return -EFAULT; + } + } + return 0; +} +EXPORT_SYMBOL(ttm_populate_and_map_pages); + +void ttm_unmap_and_unpopulate_pages(struct device *dev, struct ttm_dma_tt *tt) +{ + unsigned i; + + for (i = 0; i < tt->ttm.num_pages; i++) { + if (tt->dma_address[i]) { + dma_unmap_page(dev, tt->dma_address[i], + PAGE_SIZE, DMA_BIDIRECTIONAL); + } + } + ttm_pool_unpopulate(&tt->ttm); +} +EXPORT_SYMBOL(ttm_unmap_and_unpopulate_pages); +#endif + int ttm_page_alloc_debugfs(struct seq_file *m, void *data) { struct ttm_page_pool *p; diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c index 90ddbdca93bd..06bc14b55e66 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c @@ -60,22 +60,15 @@ #define NUM_PAGES_TO_ALLOC (PAGE_SIZE/sizeof(struct page *)) #define SMALL_ALLOCATION 4 #define FREE_ALL_PAGES (~0U) -/* times are in msecs */ -#define IS_UNDEFINED (0) -#define IS_WC (1<<1) -#define IS_UC (1<<2) -#define IS_CACHED (1<<3) -#define IS_DMA32 (1<<4) enum pool_type { - POOL_IS_UNDEFINED, - POOL_IS_WC = IS_WC, - POOL_IS_UC = IS_UC, - POOL_IS_CACHED = IS_CACHED, - POOL_IS_WC_DMA32 = IS_WC | IS_DMA32, - POOL_IS_UC_DMA32 = IS_UC | IS_DMA32, - POOL_IS_CACHED_DMA32 = IS_CACHED | IS_DMA32, + IS_UNDEFINED = 0, + IS_WC = 1 << 1, + IS_UC = 1 << 2, + IS_CACHED = 1 << 3, + IS_DMA32 = 1 << 4 }; + /* * The pool structure. There are usually six pools: * - generic (not restricted to DMA32): @@ -86,11 +79,9 @@ enum pool_type { * The other ones can be shrunk by the shrinker API if neccessary. * @pools: The 'struct device->dma_pools' link. * @type: Type of the pool - * @lock: Protects the inuse_list and free_list from concurrnet access. Must be + * @lock: Protects the free_list from concurrnet access. Must be * used with irqsave/irqrestore variants because pool allocator maybe called * from delayed work. - * @inuse_list: Pool of pages that are in use. The order is very important and - * it is in the order that the TTM pages that are put back are in. * @free_list: Pool of pages that are free to be used. No order requirements. * @dev: The device that is associated with these pools. * @size: Size used during DMA allocation. @@ -107,7 +98,6 @@ struct dma_pool { struct list_head pools; /* The 'struct device->dma_pools link */ enum pool_type type; spinlock_t lock; - struct list_head inuse_list; struct list_head free_list; struct device *dev; unsigned size; @@ -609,7 +599,6 @@ static struct dma_pool *ttm_dma_pool_init(struct device *dev, gfp_t flags, sec_pool->pool = pool; INIT_LIST_HEAD(&pool->free_list); - INIT_LIST_HEAD(&pool->inuse_list); INIT_LIST_HEAD(&pool->pools); spin_lock_init(&pool->lock); pool->dev = dev; @@ -882,22 +871,23 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev) struct dma_pool *pool; enum pool_type type; unsigned i; - gfp_t gfp_flags; int ret; if (ttm->state != tt_unpopulated) return 0; type = ttm_to_type(ttm->page_flags, ttm->caching_state); - if (ttm->page_flags & TTM_PAGE_FLAG_DMA32) - gfp_flags = GFP_USER | GFP_DMA32; - else - gfp_flags = GFP_HIGHUSER; - if (ttm->page_flags & TTM_PAGE_FLAG_ZERO_ALLOC) - gfp_flags |= __GFP_ZERO; - pool = ttm_dma_find_pool(dev, type); if (!pool) { + gfp_t gfp_flags; + + if (ttm->page_flags & TTM_PAGE_FLAG_DMA32) + gfp_flags = GFP_USER | GFP_DMA32; + else + gfp_flags = GFP_HIGHUSER; + if (ttm->page_flags & TTM_PAGE_FLAG_ZERO_ALLOC) + gfp_flags |= __GFP_ZERO; + pool = ttm_dma_pool_init(dev, gfp_flags, type); if (IS_ERR_OR_NULL(pool)) { return -ENOMEM; |

