From 9d77db3c730780c8ef5ccd4b66403ff5675cfe4e Mon Sep 17 00:00:00 2001 From: hc <hc@nodka.com> Date: Mon, 13 May 2024 10:30:14 +0000 Subject: [PATCH] modify sin led gpio --- kernel/drivers/gpu/drm/ttm/ttm_bo.c | 1353 +++++++++++++++++++++++++--------------------------------- 1 files changed, 592 insertions(+), 761 deletions(-) diff --git a/kernel/drivers/gpu/drm/ttm/ttm_bo.c b/kernel/drivers/gpu/drm/ttm/ttm_bo.c index 81e0766..f673292 100644 --- a/kernel/drivers/gpu/drm/ttm/ttm_bo.c +++ b/kernel/drivers/gpu/drm/ttm/ttm_bo.c @@ -41,9 +41,17 @@ #include <linux/file.h> #include <linux/module.h> #include <linux/atomic.h> -#include <linux/reservation.h> +#include <linux/dma-resv.h> static void ttm_bo_global_kobj_release(struct kobject *kobj); + +/** + * ttm_global_mutex - protecting the global BO state + */ +DEFINE_MUTEX(ttm_global_mutex); +unsigned ttm_bo_glob_use_count; +struct ttm_bo_global ttm_bo_glob; +EXPORT_SYMBOL(ttm_bo_glob); static struct attribute ttm_bo_count = { .name = "bo_count", @@ -56,51 +64,22 @@ kfree(bo); } -static inline int ttm_mem_type_from_place(const struct ttm_place *place, - uint32_t *mem_type) -{ - int pos; - - pos = ffs(place->flags & TTM_PL_MASK_MEM); - if (unlikely(!pos)) - return -EINVAL; - - *mem_type = pos - 1; - return 0; -} - -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); - pr_err(" flags: 0x%08X\n", man->flags); - pr_err(" gpu_offset: 0x%08llX\n", man->gpu_offset); - pr_err(" size: %llu\n", man->size); - 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, &p); -} - static void ttm_bo_mem_space_debug(struct ttm_buffer_object *bo, struct ttm_placement *placement) { - int i, ret, mem_type; + struct drm_printer p = drm_debug_printer(TTM_PFX); + struct ttm_resource_manager *man; + int i, mem_type; - pr_err("No space for %p (%lu pages, %luK, %luM)\n", - bo, bo->mem.num_pages, bo->mem.size >> 10, - bo->mem.size >> 20); + drm_printf(&p, "No space for %p (%lu pages, %luK, %luM)\n", + bo, bo->mem.num_pages, bo->mem.size >> 10, + bo->mem.size >> 20); for (i = 0; i < placement->num_placement; i++) { - ret = ttm_mem_type_from_place(&placement->placement[i], - &mem_type); - if (ret) - return; - pr_err(" placement[%d]=0x%08X (%d)\n", - i, placement->placement[i].flags, mem_type); - ttm_mem_type_debug(bo->bdev, mem_type); + mem_type = placement->placement[i].mem_type; + drm_printf(&p, " placement[%d]=0x%08X (%d)\n", + i, placement->placement[i].flags, mem_type); + man = ttm_manager_type(bo->bdev, mem_type); + ttm_resource_manager_debug(man, &p); } } @@ -130,137 +109,162 @@ .default_attrs = ttm_bo_global_attrs }; - -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(kref_read(&bo->list_kref)); - BUG_ON(kref_read(&bo->kref)); - 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)); - ttm_tt_destroy(bo->ttm); - atomic_dec(&bo->bdev->glob->bo_count); - dma_fence_put(bo->moving); - reservation_object_fini(&bo->ttm_resv); - mutex_destroy(&bo->wu_mutex); - bo->destroy(bo); - ttm_mem_global_free(bdev->glob->mem_glob, acc_size); -} - -void ttm_bo_add_to_lru(struct ttm_buffer_object *bo) +static void ttm_bo_add_mem_to_lru(struct ttm_buffer_object *bo, + struct ttm_resource *mem) { struct ttm_bo_device *bdev = bo->bdev; - struct ttm_mem_type_manager *man; + struct ttm_resource_manager *man; - reservation_object_assert_held(bo->resv); + if (!list_empty(&bo->lru)) + return; - if (!(bo->mem.placement & TTM_PL_FLAG_NO_EVICT)) { - BUG_ON(!list_empty(&bo->lru)); + if (mem->placement & TTM_PL_FLAG_NO_EVICT) + return; - man = &bdev->man[bo->mem.mem_type]; - list_add_tail(&bo->lru, &man->lru[bo->priority]); - kref_get(&bo->list_kref); + man = ttm_manager_type(bdev, mem->mem_type); + list_add_tail(&bo->lru, &man->lru[bo->priority]); - if (bo->ttm && !(bo->ttm->page_flags & - (TTM_PAGE_FLAG_SG | TTM_PAGE_FLAG_SWAPPED))) { - list_add_tail(&bo->swap, - &bdev->glob->swap_lru[bo->priority]); - kref_get(&bo->list_kref); - } + if (man->use_tt && bo->ttm && + !(bo->ttm->page_flags & (TTM_PAGE_FLAG_SG | + TTM_PAGE_FLAG_SWAPPED))) { + list_add_tail(&bo->swap, &ttm_bo_glob.swap_lru[bo->priority]); } } -EXPORT_SYMBOL(ttm_bo_add_to_lru); -static void ttm_bo_ref_bug(struct kref *list_kref) +static void ttm_bo_del_from_lru(struct ttm_buffer_object *bo) { - BUG(); -} + struct ttm_bo_device *bdev = bo->bdev; + bool notify = false; -void ttm_bo_del_from_lru(struct ttm_buffer_object *bo) -{ if (!list_empty(&bo->swap)) { list_del_init(&bo->swap); - kref_put(&bo->list_kref, ttm_bo_ref_bug); + notify = true; } if (!list_empty(&bo->lru)) { list_del_init(&bo->lru); - kref_put(&bo->list_kref, ttm_bo_ref_bug); + notify = true; } - /* - * TODO: Add a driver hook to delete from - * driver-specific LRU's here. - */ + if (notify && bdev->driver->del_from_lru_notify) + bdev->driver->del_from_lru_notify(bo); } -void ttm_bo_del_sub_from_lru(struct ttm_buffer_object *bo) +static void ttm_bo_bulk_move_set_pos(struct ttm_lru_bulk_move_pos *pos, + struct ttm_buffer_object *bo) { - struct ttm_bo_global *glob = bo->bdev->glob; - - spin_lock(&glob->lru_lock); - ttm_bo_del_from_lru(bo); - spin_unlock(&glob->lru_lock); + if (!pos->first) + pos->first = bo; + pos->last = bo; } -EXPORT_SYMBOL(ttm_bo_del_sub_from_lru); -void ttm_bo_move_to_lru_tail(struct ttm_buffer_object *bo) +void ttm_bo_move_to_lru_tail(struct ttm_buffer_object *bo, + struct ttm_lru_bulk_move *bulk) { - reservation_object_assert_held(bo->resv); + dma_resv_assert_held(bo->base.resv); ttm_bo_del_from_lru(bo); - ttm_bo_add_to_lru(bo); + ttm_bo_add_mem_to_lru(bo, &bo->mem); + + if (bulk && !(bo->mem.placement & TTM_PL_FLAG_NO_EVICT)) { + switch (bo->mem.mem_type) { + case TTM_PL_TT: + ttm_bo_bulk_move_set_pos(&bulk->tt[bo->priority], bo); + break; + + case TTM_PL_VRAM: + ttm_bo_bulk_move_set_pos(&bulk->vram[bo->priority], bo); + break; + } + if (bo->ttm && !(bo->ttm->page_flags & + (TTM_PAGE_FLAG_SG | TTM_PAGE_FLAG_SWAPPED))) + ttm_bo_bulk_move_set_pos(&bulk->swap[bo->priority], bo); + } } EXPORT_SYMBOL(ttm_bo_move_to_lru_tail); +void ttm_bo_bulk_move_lru_tail(struct ttm_lru_bulk_move *bulk) +{ + unsigned i; + + for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) { + struct ttm_lru_bulk_move_pos *pos = &bulk->tt[i]; + struct ttm_resource_manager *man; + + if (!pos->first) + continue; + + dma_resv_assert_held(pos->first->base.resv); + dma_resv_assert_held(pos->last->base.resv); + + man = ttm_manager_type(pos->first->bdev, TTM_PL_TT); + list_bulk_move_tail(&man->lru[i], &pos->first->lru, + &pos->last->lru); + } + + for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) { + struct ttm_lru_bulk_move_pos *pos = &bulk->vram[i]; + struct ttm_resource_manager *man; + + if (!pos->first) + continue; + + dma_resv_assert_held(pos->first->base.resv); + dma_resv_assert_held(pos->last->base.resv); + + man = ttm_manager_type(pos->first->bdev, TTM_PL_VRAM); + list_bulk_move_tail(&man->lru[i], &pos->first->lru, + &pos->last->lru); + } + + for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) { + struct ttm_lru_bulk_move_pos *pos = &bulk->swap[i]; + struct list_head *lru; + + if (!pos->first) + continue; + + dma_resv_assert_held(pos->first->base.resv); + dma_resv_assert_held(pos->last->base.resv); + + lru = &ttm_bo_glob.swap_lru[i]; + list_bulk_move_tail(lru, &pos->first->swap, &pos->last->swap); + } +} +EXPORT_SYMBOL(ttm_bo_bulk_move_lru_tail); + static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo, - struct ttm_mem_reg *mem, bool evict, + struct ttm_resource *mem, bool evict, struct ttm_operation_ctx *ctx) { 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; + struct ttm_resource_manager *old_man = ttm_manager_type(bdev, bo->mem.mem_type); + struct ttm_resource_manager *new_man = ttm_manager_type(bdev, mem->mem_type); + int ret; - 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); - } + ttm_bo_unmap_virtual(bo); /* * 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_tt_create(bo, zero); - if (ret) - goto out_err; - } + if (new_man->use_tt) { + /* Zero init the new TTM structure if the old location should + * have used one as well. + */ + ret = ttm_tt_create(bo, old_man->use_tt); + 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, ctx); + ret = ttm_tt_populate(bdev, bo->ttm, ctx); + if (ret) + goto out_err; + + ret = ttm_bo_tt_bind(bo, mem); if (ret) goto out_err; } @@ -269,7 +273,6 @@ if (bdev->driver->move_notify) bdev->driver->move_notify(bo, evict, mem); bo->mem = *mem; - mem->mm_node = NULL; goto moved; } } @@ -277,8 +280,7 @@ if (bdev->driver->move_notify) bdev->driver->move_notify(bo, evict, mem); - if (!(old_man->flags & TTM_MEMTYPE_FLAG_FIXED) && - !(new_man->flags & TTM_MEMTYPE_FLAG_FIXED)) + if (old_man->use_tt && new_man->use_tt) ret = ttm_bo_move_ttm(bo, ctx, mem); else if (bdev->driver->move) ret = bdev->driver->move(bo, evict, ctx, mem); @@ -296,30 +298,13 @@ } 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; - else - bo->offset = 0; - ctx->bytes_moved += bo->num_pages << PAGE_SHIFT; return 0; out_err: - new_man = &bdev->man[bo->mem.mem_type]; - if (new_man->flags & TTM_MEMTYPE_FLAG_FIXED) { - ttm_tt_destroy(bo->ttm); - bo->ttm = NULL; - } + new_man = ttm_manager_type(bdev, bo->mem.mem_type); + if (!new_man->use_tt) + ttm_bo_tt_destroy(bo); return ret; } @@ -337,108 +322,63 @@ if (bo->bdev->driver->move_notify) bo->bdev->driver->move_notify(bo, false, NULL); - ttm_tt_destroy(bo->ttm); - bo->ttm = NULL; - ttm_bo_mem_put(bo, &bo->mem); + ttm_bo_tt_destroy(bo); + ttm_resource_free(bo, &bo->mem); } static int ttm_bo_individualize_resv(struct ttm_buffer_object *bo) { int r; - if (bo->resv == &bo->ttm_resv) + if (bo->base.resv == &bo->base._resv) return 0; - BUG_ON(!reservation_object_trylock(&bo->ttm_resv)); + BUG_ON(!dma_resv_trylock(&bo->base._resv)); - r = reservation_object_copy_fences(&bo->ttm_resv, bo->resv); + r = dma_resv_copy_fences(&bo->base._resv, bo->base.resv); + dma_resv_unlock(&bo->base._resv); if (r) - reservation_object_unlock(&bo->ttm_resv); + return r; + + if (bo->type != ttm_bo_type_sg) { + /* This works because the BO is about to be destroyed and nobody + * reference it any more. The only tricky case is the trylock on + * the resv object while holding the lru_lock. + */ + spin_lock(&ttm_bo_glob.lru_lock); + bo->base.resv = &bo->base._resv; + spin_unlock(&ttm_bo_glob.lru_lock); + } return r; } static void ttm_bo_flush_all_fences(struct ttm_buffer_object *bo) { - struct reservation_object_list *fobj; + struct dma_resv *resv = &bo->base._resv; + struct dma_resv_list *fobj; struct dma_fence *fence; int i; - fobj = reservation_object_get_list(&bo->ttm_resv); - fence = reservation_object_get_excl(&bo->ttm_resv); + rcu_read_lock(); + fobj = rcu_dereference(resv->fence); + fence = rcu_dereference(resv->fence_excl); if (fence && !fence->ops->signaled) dma_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)); + fence = rcu_dereference(fobj->shared[i]); if (!fence->ops->signaled) dma_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 = bdev->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 = reservation_object_trylock(bo->resv) ? 0 : -EBUSY; - if (!ret) { - 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); - reservation_object_unlock(bo->resv); - return; - } - - 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); - } - - reservation_object_unlock(bo->resv); - } - 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); - - schedule_delayed_work(&bdev->wq, - ((HZ / 100) < 1) ? 1 : HZ / 100); + rcu_read_unlock(); } /** * function ttm_bo_cleanup_refs - * If bo idle, remove from delayed- and lru lists, and unref. - * If not idle, do nothing. + * If bo idle, remove from lru lists, and unref. + * If not idle, block if possible. * * Must be called with lru_lock and reservation held, this function * will drop the lru lock and optionally the reservation lock before returning. @@ -452,16 +392,10 @@ bool interruptible, bool no_wait_gpu, bool unlock_resv) { - struct ttm_bo_global *glob = bo->bdev->glob; - struct reservation_object *resv; + struct dma_resv *resv = &bo->base._resv; int ret; - if (unlikely(list_empty(&bo->ddestroy))) - resv = bo->resv; - else - resv = &bo->ttm_resv; - - if (reservation_object_test_signaled_rcu(resv, true)) + if (dma_resv_test_signaled_rcu(resv, true)) ret = 0; else ret = -EBUSY; @@ -470,20 +404,19 @@ long lret; if (unlock_resv) - reservation_object_unlock(bo->resv); - spin_unlock(&glob->lru_lock); + dma_resv_unlock(bo->base.resv); + spin_unlock(&ttm_bo_glob.lru_lock); - lret = reservation_object_wait_timeout_rcu(resv, true, - interruptible, - 30 * HZ); + lret = dma_resv_wait_timeout_rcu(resv, true, interruptible, + 30 * HZ); if (lret < 0) return lret; else if (lret == 0) return -EBUSY; - spin_lock(&glob->lru_lock); - if (unlock_resv && !reservation_object_trylock(bo->resv)) { + spin_lock(&ttm_bo_glob.lru_lock); + if (unlock_resv && !dma_resv_trylock(bo->base.resv)) { /* * We raced, and lost, someone else holds the reservation now, * and is probably busy in ttm_bo_cleanup_memtype_use. @@ -492,7 +425,7 @@ * delayed destruction would succeed, so just return success * here. */ - spin_unlock(&glob->lru_lock); + spin_unlock(&ttm_bo_glob.lru_lock); return 0; } ret = 0; @@ -500,20 +433,20 @@ if (ret || unlikely(list_empty(&bo->ddestroy))) { if (unlock_resv) - reservation_object_unlock(bo->resv); - spin_unlock(&glob->lru_lock); + dma_resv_unlock(bo->base.resv); + spin_unlock(&ttm_bo_glob.lru_lock); return ret; } ttm_bo_del_from_lru(bo); list_del_init(&bo->ddestroy); - kref_put(&bo->list_kref, ttm_bo_ref_bug); - - spin_unlock(&glob->lru_lock); + spin_unlock(&ttm_bo_glob.lru_lock); ttm_bo_cleanup_memtype_use(bo); if (unlock_resv) - reservation_object_unlock(bo->resv); + dma_resv_unlock(bo->base.resv); + + ttm_bo_put(bo); return 0; } @@ -524,7 +457,7 @@ */ static bool ttm_bo_delayed_delete(struct ttm_bo_device *bdev, bool remove_all) { - struct ttm_bo_global *glob = bdev->glob; + struct ttm_bo_global *glob = &ttm_bo_glob; struct list_head removed; bool empty; @@ -536,23 +469,24 @@ bo = list_first_entry(&bdev->ddestroy, struct ttm_buffer_object, ddestroy); - kref_get(&bo->list_kref); list_move_tail(&bo->ddestroy, &removed); + if (!ttm_bo_get_unless_zero(bo)) + continue; - if (remove_all || bo->resv != &bo->ttm_resv) { + if (remove_all || bo->base.resv != &bo->base._resv) { spin_unlock(&glob->lru_lock); - reservation_object_lock(bo->resv, NULL); + dma_resv_lock(bo->base.resv, NULL); spin_lock(&glob->lru_lock); ttm_bo_cleanup_refs(bo, false, !remove_all, true); - } else if (reservation_object_trylock(bo->resv)) { + } else if (dma_resv_trylock(bo->base.resv)) { ttm_bo_cleanup_refs(bo, false, !remove_all, true); } else { spin_unlock(&glob->lru_lock); } - kref_put(&bo->list_kref, ttm_bo_release_list); + ttm_bo_put(bo); spin_lock(&glob->lru_lock); } list_splice_tail(&removed, &bdev->ddestroy); @@ -577,14 +511,68 @@ 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]; + size_t acc_size = bo->acc_size; + int ret; - 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); + if (!bo->deleted) { + 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 + */ + dma_resv_wait_timeout_rcu(bo->base.resv, true, false, + 30 * HZ); + } + + if (bo->bdev->driver->release_notify) + bo->bdev->driver->release_notify(bo); + + drm_vma_offset_remove(bdev->vma_manager, &bo->base.vma_node); + ttm_mem_io_free(bdev, &bo->mem); + } + + if (!dma_resv_test_signaled_rcu(bo->base.resv, true) || + !dma_resv_trylock(bo->base.resv)) { + /* The BO is not idle, resurrect it for delayed destroy */ + ttm_bo_flush_all_fences(bo); + bo->deleted = true; + + spin_lock(&ttm_bo_glob.lru_lock); + + /* + * 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_del_from_lru(bo); + ttm_bo_add_mem_to_lru(bo, &bo->mem); + } + + kref_init(&bo->kref); + list_add_tail(&bo->ddestroy, &bdev->ddestroy); + spin_unlock(&ttm_bo_glob.lru_lock); + + schedule_delayed_work(&bdev->wq, + ((HZ / 100) < 1) ? 1 : HZ / 100); + return; + } + + spin_lock(&ttm_bo_glob.lru_lock); + ttm_bo_del_from_lru(bo); + list_del(&bo->ddestroy); + spin_unlock(&ttm_bo_glob.lru_lock); + + ttm_bo_cleanup_memtype_use(bo); + dma_resv_unlock(bo->base.resv); + + atomic_dec(&ttm_bo_glob.bo_count); + dma_fence_put(bo->moving); + if (!ttm_bo_uses_embedded_gem_object(bo)) + dma_resv_fini(&bo->base._resv); + bo->destroy(bo); + ttm_mem_global_free(&ttm_mem_glob, acc_size); } void ttm_bo_put(struct ttm_buffer_object *bo) @@ -592,15 +580,6 @@ kref_put(&bo->kref, ttm_bo_release); } EXPORT_SYMBOL(ttm_bo_put); - -void ttm_bo_unref(struct ttm_buffer_object **p_bo) -{ - struct ttm_buffer_object *bo = *p_bo; - - *p_bo = NULL; - ttm_bo_put(bo); -} -EXPORT_SYMBOL(ttm_bo_unref); int ttm_bo_lock_delayed_workqueue(struct ttm_bo_device *bdev) { @@ -620,28 +599,27 @@ struct ttm_operation_ctx *ctx) { struct ttm_bo_device *bdev = bo->bdev; - struct ttm_mem_reg evict_mem; + struct ttm_resource evict_mem; struct ttm_placement placement; int ret = 0; - reservation_object_assert_held(bo->resv); + dma_resv_assert_held(bo->base.resv); placement.num_placement = 0; placement.num_busy_placement = 0; bdev->driver->evict_flags(bo, &placement); if (!placement.num_placement && !placement.num_busy_placement) { - ret = ttm_bo_pipeline_gutting(bo); - if (ret) - return ret; + ttm_bo_wait(bo, false, false); + ttm_bo_cleanup_memtype_use(bo); return ttm_tt_create(bo, false); } evict_mem = bo->mem; evict_mem.mm_node = NULL; - evict_mem.bus.io_reserved_vm = false; - evict_mem.bus.io_reserved_count = 0; + evict_mem.bus.offset = 0; + evict_mem.bus.addr = NULL; ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx); if (ret) { @@ -657,10 +635,8 @@ if (unlikely(ret)) { if (ret != -ERESTARTSYS) pr_err("Buffer eviction failed\n"); - ttm_bo_mem_put(bo, &evict_mem); - goto out; + ttm_resource_free(bo, &evict_mem); } - bo->evicted = true; out: return ret; } @@ -690,46 +666,95 @@ * b. Otherwise, trylock it. */ static bool ttm_bo_evict_swapout_allowable(struct ttm_buffer_object *bo, - struct ttm_operation_ctx *ctx, bool *locked) + struct ttm_operation_ctx *ctx, bool *locked, bool *busy) { bool ret = false; - *locked = false; - if (bo->resv == ctx->resv) { - reservation_object_assert_held(bo->resv); - if (ctx->flags & TTM_OPT_FLAG_ALLOW_RES_EVICT - || !list_empty(&bo->ddestroy)) + if (bo->base.resv == ctx->resv) { + dma_resv_assert_held(bo->base.resv); + if (ctx->flags & TTM_OPT_FLAG_ALLOW_RES_EVICT) ret = true; + *locked = false; + if (busy) + *busy = false; } else { - *locked = reservation_object_trylock(bo->resv); - ret = *locked; + ret = dma_resv_trylock(bo->base.resv); + *locked = ret; + if (busy) + *busy = !ret; } return ret; } -static int ttm_mem_evict_first(struct ttm_bo_device *bdev, - uint32_t mem_type, - const struct ttm_place *place, - struct ttm_operation_ctx *ctx) +/** + * ttm_mem_evict_wait_busy - wait for a busy BO to become available + * + * @busy_bo: BO which couldn't be locked with trylock + * @ctx: operation context + * @ticket: acquire ticket + * + * Try to lock a busy buffer object to avoid failing eviction. + */ +static int ttm_mem_evict_wait_busy(struct ttm_buffer_object *busy_bo, + struct ttm_operation_ctx *ctx, + struct ww_acquire_ctx *ticket) { - struct ttm_bo_global *glob = bdev->glob; - struct ttm_mem_type_manager *man = &bdev->man[mem_type]; - struct ttm_buffer_object *bo = NULL; + int r; + + if (!busy_bo || !ticket) + return -EBUSY; + + if (ctx->interruptible) + r = dma_resv_lock_interruptible(busy_bo->base.resv, + ticket); + else + r = dma_resv_lock(busy_bo->base.resv, ticket); + + /* + * TODO: It would be better to keep the BO locked until allocation is at + * least tried one more time, but that would mean a much larger rework + * of TTM. + */ + if (!r) + dma_resv_unlock(busy_bo->base.resv); + + return r == -EDEADLK ? -EBUSY : r; +} + +int ttm_mem_evict_first(struct ttm_bo_device *bdev, + struct ttm_resource_manager *man, + const struct ttm_place *place, + struct ttm_operation_ctx *ctx, + struct ww_acquire_ctx *ticket) +{ + struct ttm_buffer_object *bo = NULL, *busy_bo = NULL; bool locked = false; unsigned i; int ret; - spin_lock(&glob->lru_lock); + spin_lock(&ttm_bo_glob.lru_lock); for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) { list_for_each_entry(bo, &man->lru[i], lru) { - if (!ttm_bo_evict_swapout_allowable(bo, ctx, &locked)) + bool busy; + + if (!ttm_bo_evict_swapout_allowable(bo, ctx, &locked, + &busy)) { + if (busy && !busy_bo && ticket != + dma_resv_locking_ctx(bo->base.resv)) + busy_bo = bo; continue; + } if (place && !bdev->driver->eviction_valuable(bo, place)) { if (locked) - reservation_object_unlock(bo->resv); + dma_resv_unlock(bo->base.resv); + continue; + } + if (!ttm_bo_get_unless_zero(bo)) { + if (locked) + dma_resv_unlock(bo->base.resv); continue; } break; @@ -743,50 +768,41 @@ } if (!bo) { - spin_unlock(&glob->lru_lock); - return -EBUSY; - } - - kref_get(&bo->list_kref); - - if (!list_empty(&bo->ddestroy)) { - ret = ttm_bo_cleanup_refs(bo, ctx->interruptible, - ctx->no_wait_gpu, locked); - kref_put(&bo->list_kref, ttm_bo_release_list); + if (busy_bo && !ttm_bo_get_unless_zero(busy_bo)) + busy_bo = NULL; + spin_unlock(&ttm_bo_glob.lru_lock); + ret = ttm_mem_evict_wait_busy(busy_bo, ctx, ticket); + if (busy_bo) + ttm_bo_put(busy_bo); return ret; } - ttm_bo_del_from_lru(bo); - spin_unlock(&glob->lru_lock); - - ret = ttm_bo_evict(bo, ctx); - if (locked) { - ttm_bo_unreserve(bo); - } else { - spin_lock(&glob->lru_lock); - ttm_bo_add_to_lru(bo); - spin_unlock(&glob->lru_lock); + if (bo->deleted) { + ret = ttm_bo_cleanup_refs(bo, ctx->interruptible, + ctx->no_wait_gpu, locked); + ttm_bo_put(bo); + return ret; } - kref_put(&bo->list_kref, ttm_bo_release_list); + spin_unlock(&ttm_bo_glob.lru_lock); + + ret = ttm_bo_evict(bo, ctx); + if (locked) + ttm_bo_unreserve(bo); + else + ttm_bo_move_to_lru_tail_unlocked(bo); + + ttm_bo_put(bo); return ret; } - -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); /** * Add the last move fence to the BO and reserve a new shared slot. */ static int ttm_bo_add_move_fence(struct ttm_buffer_object *bo, - struct ttm_mem_type_manager *man, - struct ttm_mem_reg *mem) + struct ttm_resource_manager *man, + struct ttm_resource *mem, + bool no_wait_gpu) { struct dma_fence *fence; int ret; @@ -795,17 +811,24 @@ fence = dma_fence_get(man->move); spin_unlock(&man->move_lock); - if (fence) { - reservation_object_add_shared_fence(bo->resv, fence); + if (!fence) + return 0; - ret = reservation_object_reserve_shared(bo->resv); - if (unlikely(ret)) - return ret; - - dma_fence_put(bo->moving); - bo->moving = fence; + if (no_wait_gpu) { + dma_fence_put(fence); + return -EBUSY; } + dma_resv_add_shared_fence(bo->base.resv, fence); + + ret = dma_resv_reserve_shared(bo->base.resv, 1); + if (unlikely(ret)) { + dma_fence_put(fence); + return ret; + } + + dma_fence_put(bo->moving); + bo->moving = fence; return 0; } @@ -814,30 +837,32 @@ * 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, - struct ttm_operation_ctx *ctx) + const struct ttm_place *place, + struct ttm_resource *mem, + struct ttm_operation_ctx *ctx) { struct ttm_bo_device *bdev = bo->bdev; - struct ttm_mem_type_manager *man = &bdev->man[mem_type]; + struct ttm_resource_manager *man = ttm_manager_type(bdev, mem->mem_type); + struct ww_acquire_ctx *ticket; int ret; + ticket = dma_resv_locking_ctx(bo->base.resv); do { - ret = (*man->func->get_node)(man, bo, place, mem); - if (unlikely(ret != 0)) - return ret; - if (mem->mm_node) + ret = ttm_resource_alloc(bo, place, mem); + if (likely(!ret)) break; - ret = ttm_mem_evict_first(bdev, mem_type, place, ctx); + if (unlikely(ret != -ENOSPC)) + return ret; + ret = ttm_mem_evict_first(bdev, man, place, ctx, + ticket); if (unlikely(ret != 0)) return ret; } while (1); - mem->mem_type = mem_type; - return ttm_bo_add_move_fence(bo, man, mem); + + return ttm_bo_add_move_fence(bo, man, mem, ctx->no_wait_gpu); } -static uint32_t ttm_bo_select_caching(struct ttm_mem_type_manager *man, +static uint32_t ttm_bo_select_caching(struct ttm_resource_manager *man, uint32_t cur_placement, uint32_t proposed_placement) { @@ -850,8 +875,6 @@ 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) @@ -862,23 +885,43 @@ 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) +/** + * ttm_bo_mem_placement - check if placement is compatible + * @bo: BO to find memory for + * @place: where to search + * @mem: the memory object to fill in + * @ctx: operation context + * + * Check if placement is compatible and fill in mem structure. + * Returns -EBUSY if placement won't work or negative error code. + * 0 when placement can be used. + */ +static int ttm_bo_mem_placement(struct ttm_buffer_object *bo, + const struct ttm_place *place, + struct ttm_resource *mem, + struct ttm_operation_ctx *ctx) { - uint32_t cur_flags = ttm_bo_type_flags(mem_type); + struct ttm_bo_device *bdev = bo->bdev; + struct ttm_resource_manager *man; + uint32_t cur_flags = 0; - if ((cur_flags & place->flags & TTM_PL_MASK_MEM) == 0) - return false; + man = ttm_manager_type(bdev, place->mem_type); + if (!man || !ttm_resource_manager_used(man)) + return -EBUSY; - if ((place->flags & man->available_caching) == 0) - return false; + cur_flags = ttm_bo_select_caching(man, bo->mem.placement, + place->flags); + cur_flags |= place->flags & ~TTM_PL_MASK_CACHING; - cur_flags |= (place->flags & man->available_caching); + mem->mem_type = place->mem_type; + mem->placement = cur_flags; - *masked_placement = cur_flags; - return true; + spin_lock(&ttm_bo_glob.lru_lock); + ttm_bo_del_from_lru(bo); + ttm_bo_add_mem_to_lru(bo, mem); + spin_unlock(&ttm_bo_glob.lru_lock); + + return 0; } /** @@ -891,116 +934,72 @@ */ int ttm_bo_mem_space(struct ttm_buffer_object *bo, struct ttm_placement *placement, - struct ttm_mem_reg *mem, + struct ttm_resource *mem, struct ttm_operation_ctx *ctx) { 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; - ret = reservation_object_reserve_shared(bo->resv); + ret = dma_resv_reserve_shared(bo->base.resv, 1); if (unlikely(ret)) return ret; - mem->mm_node = NULL; for (i = 0; i < placement->num_placement; ++i) { const struct ttm_place *place = &placement->placement[i]; + struct ttm_resource_manager *man; - ret = ttm_mem_type_from_place(place, &mem_type); + ret = ttm_bo_mem_placement(bo, place, mem, ctx); 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); + ret = ttm_resource_alloc(bo, place, mem); + if (ret == -ENOSPC) + continue; if (unlikely(ret)) - return ret; + goto error; - if (mem->mm_node) { - ret = ttm_bo_add_move_fence(bo, man, mem); - if (unlikely(ret)) { - (*man->func->put_node)(man, mem); - return ret; - } - break; + man = ttm_manager_type(bdev, mem->mem_type); + ret = ttm_bo_add_move_fence(bo, man, mem, ctx->no_wait_gpu); + if (unlikely(ret)) { + ttm_resource_free(bo, mem); + if (ret == -EBUSY) + continue; + + goto error; } - } - - 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); + ret = ttm_bo_mem_placement(bo, place, mem, ctx); 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; + ret = ttm_bo_mem_force_space(bo, place, mem, ctx); + if (likely(!ret)) return 0; - } - ret = ttm_bo_mem_force_space(bo, mem_type, place, mem, ctx); - if (ret == 0 && mem->mm_node) { - mem->placement = cur_flags; - return 0; - } - if (ret == -ERESTARTSYS) - has_erestartsys = true; + if (ret && ret != -EBUSY) + goto error; } + ret = -ENOMEM; if (!type_found) { pr_err(TTM_PFX "No compatible memory type found\n"); - return -EINVAL; + ret = -EINVAL; } - return (has_erestartsys) ? -ERESTARTSYS : -ENOMEM; +error: + if (bo->mem.mem_type == TTM_PL_SYSTEM && !list_empty(&bo->lru)) { + ttm_bo_move_to_lru_tail_unlocked(bo); + } + + return ret; } EXPORT_SYMBOL(ttm_bo_mem_space); @@ -1009,15 +1008,17 @@ struct ttm_operation_ctx *ctx) { int ret = 0; - struct ttm_mem_reg mem; + struct ttm_resource mem; - reservation_object_assert_held(bo->resv); + dma_resv_assert_held(bo->base.resv); 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; + mem.bus.offset = 0; + mem.bus.addr = NULL; + mem.mm_node = NULL; + /* * Determine where to move the buffer. */ @@ -1026,14 +1027,14 @@ goto out_unlock; ret = ttm_bo_handle_move_mem(bo, &mem, false, ctx); out_unlock: - if (ret && mem.mm_node) - ttm_bo_mem_put(bo, &mem); + if (ret) + ttm_resource_free(bo, &mem); return ret; } static bool ttm_bo_places_compat(const struct ttm_place *places, unsigned num_placement, - struct ttm_mem_reg *mem, + struct ttm_resource *mem, uint32_t *new_flags) { unsigned i; @@ -1041,13 +1042,13 @@ for (i = 0; i < num_placement; i++) { const struct ttm_place *heap = &places[i]; - if (mem->mm_node && (mem->start < heap->fpfn || + if ((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) && + (mem->mem_type == heap->mem_type) && (!(*new_flags & TTM_PL_FLAG_CONTIGUOUS) || (mem->placement & TTM_PL_FLAG_CONTIGUOUS))) return true; @@ -1056,7 +1057,7 @@ } bool ttm_bo_mem_compat(struct ttm_placement *placement, - struct ttm_mem_reg *mem, + struct ttm_resource *mem, uint32_t *new_flags) { if (ttm_bo_places_compat(placement->placement, placement->num_placement, @@ -1081,7 +1082,19 @@ int ret; uint32_t new_flags; - reservation_object_assert_held(bo->resv); + dma_resv_assert_held(bo->base.resv); + + /* + * Remove the backing store if no placement is given. + */ + if (!placement->num_placement && !placement->num_busy_placement) { + ret = ttm_bo_pipeline_gutting(bo); + if (ret) + return ret; + + return ttm_tt_create(bo, false); + } + /* * Check whether we need to move buffer. */ @@ -1090,17 +1103,13 @@ 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); + bo->mem.placement &= TTM_PL_MASK_CACHING; + bo->mem.placement |= new_flags & ~TTM_PL_MASK_CACHING; } /* * We might need to add a TTM. */ - if (bo->mem.mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) { + if (bo->mem.mem_type == TTM_PL_SYSTEM) { ret = ttm_tt_create(bo, true); if (ret) return ret; @@ -1118,12 +1127,12 @@ struct ttm_operation_ctx *ctx, size_t acc_size, struct sg_table *sg, - struct reservation_object *resv, + struct dma_resv *resv, void (*destroy) (struct ttm_buffer_object *)) { + struct ttm_mem_global *mem_glob = &ttm_mem_glob; int ret = 0; unsigned long num_pages; - struct ttm_mem_global *mem_glob = bdev->glob->mem_glob; bool locked; ret = ttm_mem_global_alloc(mem_glob, acc_size, ctx); @@ -1149,13 +1158,9 @@ bo->destroy = destroy ? destroy : ttm_bo_default_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->type = type; bo->num_pages = num_pages; @@ -1164,21 +1169,27 @@ 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->mem.bus.offset = 0; + bo->mem.bus.addr = NULL; bo->moving = NULL; - bo->mem.placement = (TTM_PL_FLAG_SYSTEM | TTM_PL_FLAG_CACHED); + bo->mem.placement = TTM_PL_FLAG_CACHED; bo->acc_size = acc_size; bo->sg = sg; if (resv) { - bo->resv = resv; - reservation_object_assert_held(bo->resv); + bo->base.resv = resv; + dma_resv_assert_held(bo->base.resv); } else { - bo->resv = &bo->ttm_resv; + bo->base.resv = &bo->base._resv; } - reservation_object_init(&bo->ttm_resv); - atomic_inc(&bo->bdev->glob->bo_count); - drm_vma_node_reset(&bo->vma_node); + if (!ttm_bo_uses_embedded_gem_object(bo)) { + /* + * bo.gem is not initialized, so we have to setup the + * struct elements we want use regardless. + */ + dma_resv_init(&bo->base._resv); + drm_vma_node_reset(&bo->base.vma_node); + } + atomic_inc(&ttm_bo_glob.bo_count); /* * For ttm_bo_type_device buffers, allocate @@ -1186,14 +1197,14 @@ */ if (bo->type == ttm_bo_type_device || bo->type == ttm_bo_type_sg) - ret = drm_vma_offset_add(&bdev->vma_manager, &bo->vma_node, + ret = drm_vma_offset_add(bdev->vma_manager, &bo->base.vma_node, bo->mem.num_pages); /* passed reservation objects should already be locked, * since otherwise lockdep will be angered in radeon. */ if (!resv) { - locked = reservation_object_trylock(bo->resv); + locked = dma_resv_trylock(bo->base.resv); WARN_ON(!locked); } @@ -1208,11 +1219,7 @@ return ret; } - if (resv && !(bo->mem.placement & TTM_PL_FLAG_NO_EVICT)) { - spin_lock(&bdev->glob->lru_lock); - ttm_bo_add_to_lru(bo); - spin_unlock(&bdev->glob->lru_lock); - } + ttm_bo_move_to_lru_tail_unlocked(bo); return ret; } @@ -1227,7 +1234,7 @@ bool interruptible, size_t acc_size, struct sg_table *sg, - struct reservation_object *resv, + struct dma_resv *resv, void (*destroy) (struct ttm_buffer_object *)) { struct ttm_operation_ctx ctx = { interruptible, false }; @@ -1246,9 +1253,9 @@ } EXPORT_SYMBOL(ttm_bo_init); -size_t ttm_bo_acc_size(struct ttm_bo_device *bdev, - unsigned long bo_size, - unsigned struct_size) +static 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; @@ -1258,7 +1265,6 @@ 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, @@ -1301,143 +1307,23 @@ } EXPORT_SYMBOL(ttm_bo_create); -static int ttm_bo_force_list_clean(struct ttm_bo_device *bdev, - unsigned mem_type) -{ - struct ttm_operation_ctx ctx = { - .interruptible = false, - .no_wait_gpu = false, - .flags = TTM_OPT_FLAG_FORCE_ALLOC - }; - struct ttm_mem_type_manager *man = &bdev->man[mem_type]; - struct ttm_bo_global *glob = bdev->glob; - struct dma_fence *fence; - int ret; - unsigned i; - - /* - * Can't use standard list traversal since we're unlocking. - */ - - spin_lock(&glob->lru_lock); - for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) { - while (!list_empty(&man->lru[i])) { - spin_unlock(&glob->lru_lock); - ret = ttm_mem_evict_first(bdev, mem_type, NULL, &ctx); - if (ret) - return ret; - spin_lock(&glob->lru_lock); - } - } - spin_unlock(&glob->lru_lock); - - spin_lock(&man->move_lock); - fence = dma_fence_get(man->move); - spin_unlock(&man->move_lock); - - if (fence) { - ret = dma_fence_wait(fence, false); - dma_fence_put(fence); - if (ret) - return ret; - } - - return 0; -} - -int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type) -{ - struct ttm_mem_type_manager *man; - int ret = -EINVAL; - - if (mem_type >= TTM_NUM_MEM_TYPES) { - pr_err("Illegal memory type %d\n", mem_type); - return ret; - } - man = &bdev->man[mem_type]; - - if (!man->has_type) { - pr_err("Trying to take down uninitialized memory manager type %u\n", - mem_type); - return ret; - } - - man->use_type = false; - man->has_type = false; - - ret = 0; - if (mem_type > 0) { - ret = ttm_bo_force_list_clean(bdev, mem_type); - if (ret) { - pr_err("Cleanup eviction failed\n"); - return ret; - } - - ret = (*man->func->takedown)(man); - } - - dma_fence_put(man->move); - man->move = NULL; - - return ret; -} -EXPORT_SYMBOL(ttm_bo_clean_mm); - int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type) { - struct ttm_mem_type_manager *man = &bdev->man[mem_type]; + struct ttm_resource_manager *man = ttm_manager_type(bdev, mem_type); if (mem_type == 0 || mem_type >= TTM_NUM_MEM_TYPES) { pr_err("Illegal memory manager memory type %u\n", mem_type); return -EINVAL; } - if (!man->has_type) { + if (!man) { pr_err("Memory type %u has not been initialized\n", mem_type); return 0; } - return ttm_bo_force_list_clean(bdev, mem_type); + return ttm_resource_manager_force_list_clean(bdev, man); } EXPORT_SYMBOL(ttm_bo_evict_mm); - -int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type, - unsigned long p_size) -{ - int ret; - struct ttm_mem_type_manager *man; - unsigned i; - - 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); - spin_lock_init(&man->move_lock); - INIT_LIST_HEAD(&man->io_reserve_lru); - - ret = bdev->driver->init_mem_type(bdev, type, man); - if (ret) - return ret; - man->bdev = bdev; - - 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; - - for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) - INIT_LIST_HEAD(&man->lru[i]); - man->move = NULL; - - return 0; -} -EXPORT_SYMBOL(ttm_bo_init_mm); static void ttm_bo_global_kobj_release(struct kobject *kobj) { @@ -1447,32 +1333,42 @@ __free_page(glob->dummy_read_page); } -void ttm_bo_global_release(struct drm_global_reference *ref) +static void ttm_bo_global_release(void) { - struct ttm_bo_global *glob = ref->object; + struct ttm_bo_global *glob = &ttm_bo_glob; + + mutex_lock(&ttm_global_mutex); + if (--ttm_bo_glob_use_count > 0) + goto out; kobject_del(&glob->kobj); kobject_put(&glob->kobj); + ttm_mem_global_release(&ttm_mem_glob); + memset(glob, 0, sizeof(*glob)); +out: + mutex_unlock(&ttm_global_mutex); } -EXPORT_SYMBOL(ttm_bo_global_release); -int ttm_bo_global_init(struct drm_global_reference *ref) +static int ttm_bo_global_init(void) { - 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; + struct ttm_bo_global *glob = &ttm_bo_glob; + int ret = 0; unsigned i; - mutex_init(&glob->device_list_mutex); + mutex_lock(&ttm_global_mutex); + if (++ttm_bo_glob_use_count > 1) + goto out; + + ret = ttm_mem_global_init(&ttm_mem_glob); + if (ret) + goto out; + spin_lock_init(&glob->lru_lock); - glob->mem_glob = bo_ref->mem_glob; - glob->mem_glob->bo_glob = glob; glob->dummy_read_page = alloc_page(__GFP_ZERO | GFP_DMA32); if (unlikely(glob->dummy_read_page == NULL)) { ret = -ENOMEM; - goto out_no_drp; + goto out; } for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) @@ -1484,37 +1380,25 @@ &glob->kobj, &ttm_bo_glob_kobj_type, ttm_get_kobj(), "buffer_objects"); if (unlikely(ret != 0)) kobject_put(&glob->kobj); - return ret; -out_no_drp: - kfree(glob); +out: + mutex_unlock(&ttm_global_mutex); return ret; } -EXPORT_SYMBOL(ttm_bo_global_init); - int ttm_bo_device_release(struct ttm_bo_device *bdev) { + struct ttm_bo_global *glob = &ttm_bo_glob; int ret = 0; - unsigned i = TTM_NUM_MEM_TYPES; - struct ttm_mem_type_manager *man; - struct ttm_bo_global *glob = bdev->glob; + unsigned i; + struct ttm_resource_manager *man; - while (i--) { - man = &bdev->man[i]; - if (man->has_type) { - man->use_type = false; - if ((i != TTM_PL_SYSTEM) && ttm_bo_clean_mm(bdev, i)) { - ret = -EBUSY; - pr_err("DRM memory manager type %d is not clean\n", - i); - } - man->has_type = false; - } - } + man = ttm_manager_type(bdev, TTM_PL_SYSTEM); + ttm_resource_manager_set_used(man, false); + ttm_set_driver_manager(bdev, TTM_PL_SYSTEM, NULL); - mutex_lock(&glob->device_list_mutex); + mutex_lock(&ttm_global_mutex); list_del(&bdev->device_list); - mutex_unlock(&glob->device_list_mutex); + mutex_unlock(&ttm_global_mutex); cancel_delayed_work_sync(&bdev->wq); @@ -1523,51 +1407,62 @@ spin_lock(&glob->lru_lock); for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) - if (list_empty(&bdev->man[0].lru[0])) + if (list_empty(&man->lru[0])) pr_debug("Swap list %d was clean\n", i); spin_unlock(&glob->lru_lock); - drm_vma_offset_manager_destroy(&bdev->vma_manager); + if (!ret) + ttm_bo_global_release(); return ret; } EXPORT_SYMBOL(ttm_bo_device_release); -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) +static void ttm_bo_init_sysman(struct ttm_bo_device *bdev) { - int ret = -EINVAL; - - bdev->driver = driver; - - memset(bdev->man, 0, sizeof(bdev->man)); + struct ttm_resource_manager *man = &bdev->sysman; /* * 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; + man->use_tt = true; - drm_vma_offset_manager_init(&bdev->vma_manager, file_page_offset, - 0x10000000); + ttm_resource_manager_init(man, 0); + ttm_set_driver_manager(bdev, TTM_PL_SYSTEM, man); + ttm_resource_manager_set_used(man, true); +} + +int ttm_bo_device_init(struct ttm_bo_device *bdev, + struct ttm_bo_driver *driver, + struct address_space *mapping, + struct drm_vma_offset_manager *vma_manager, + bool need_dma32) +{ + struct ttm_bo_global *glob = &ttm_bo_glob; + int ret; + + if (WARN_ON(vma_manager == NULL)) + return -EINVAL; + + ret = ttm_bo_global_init(); + if (ret) + return ret; + + bdev->driver = driver; + + ttm_bo_init_sysman(bdev); + + bdev->vma_manager = vma_manager; 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; - mutex_lock(&glob->device_list_mutex); + mutex_lock(&ttm_global_mutex); list_add_tail(&bdev->device_list, &glob->device_list); - mutex_unlock(&glob->device_list_mutex); + mutex_unlock(&ttm_global_mutex); return 0; -out_no_sys: - return ret; } EXPORT_SYMBOL(ttm_bo_device_init); @@ -1575,42 +1470,13 @@ * 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); + drm_vma_node_unmap(&bo->base.vma_node, bdev->dev_mapping); + ttm_mem_io_free(bdev, &bo->mem); } - - EXPORT_SYMBOL(ttm_bo_unmap_virtual); int ttm_bo_wait(struct ttm_buffer_object *bo, @@ -1619,13 +1485,13 @@ long timeout = 15 * HZ; if (no_wait) { - if (reservation_object_test_signaled_rcu(bo->resv, true)) + if (dma_resv_test_signaled_rcu(bo->base.resv, true)) return 0; else return -EBUSY; } - timeout = reservation_object_wait_timeout_rcu(bo->resv, true, + timeout = dma_resv_wait_timeout_rcu(bo->base.resv, true, interruptible, timeout); if (timeout < 0) return timeout; @@ -1633,35 +1499,10 @@ if (timeout == 0) return -EBUSY; - reservation_object_add_excl_fence(bo->resv, NULL); + dma_resv_add_excl_fence(bo->base.resv, NULL); 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, NULL); - if (unlikely(ret != 0)) - return ret; - ret = ttm_bo_wait(bo, 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); -} -EXPORT_SYMBOL(ttm_bo_synccpu_write_release); /** * A buffer object shrink method that tries to swap out the first @@ -1677,10 +1518,18 @@ spin_lock(&glob->lru_lock); for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) { list_for_each_entry(bo, &glob->swap_lru[i], swap) { - if (ttm_bo_evict_swapout_allowable(bo, ctx, &locked)) { - ret = 0; - break; + if (!ttm_bo_evict_swapout_allowable(bo, ctx, &locked, + NULL)) + continue; + + if (!ttm_bo_get_unless_zero(bo)) { + if (locked) + dma_resv_unlock(bo->base.resv); + continue; } + + ret = 0; + break; } if (!ret) break; @@ -1691,11 +1540,9 @@ return ret; } - kref_get(&bo->list_kref); - - if (!list_empty(&bo->ddestroy)) { + if (bo->deleted) { ret = ttm_bo_cleanup_refs(bo, false, false, locked); - kref_put(&bo->list_kref, ttm_bo_release_list); + ttm_bo_put(bo); return ret; } @@ -1709,11 +1556,11 @@ if (bo->mem.mem_type != TTM_PL_SYSTEM || bo->ttm->caching_state != tt_cached) { struct ttm_operation_ctx ctx = { false, false }; - struct ttm_mem_reg evict_mem; + struct ttm_resource evict_mem; evict_mem = bo->mem; evict_mem.mm_node = NULL; - evict_mem.placement = TTM_PL_FLAG_SYSTEM | TTM_PL_FLAG_CACHED; + evict_mem.placement = TTM_PL_FLAG_CACHED; evict_mem.mem_type = TTM_PL_SYSTEM; ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, &ctx); @@ -1739,7 +1586,7 @@ if (bo->bdev->driver->swap_notify) bo->bdev->driver->swap_notify(bo); - ret = ttm_tt_swapout(bo->ttm, bo->persistent_swap_storage); + ret = ttm_tt_swapout(bo->bdev, bo->ttm, bo->persistent_swap_storage); out: /** @@ -1748,54 +1595,38 @@ * already swapped buffer. */ if (locked) - reservation_object_unlock(bo->resv); - kref_put(&bo->list_kref, ttm_bo_release_list); + dma_resv_unlock(bo->base.resv); + ttm_bo_put(bo); return ret; } EXPORT_SYMBOL(ttm_bo_swapout); -void ttm_bo_swapout_all(struct ttm_bo_device *bdev) +void ttm_bo_swapout_all(void) { struct ttm_operation_ctx ctx = { .interruptible = false, .no_wait_gpu = false }; - while (ttm_bo_swapout(bdev->glob, &ctx) == 0) - ; + while (ttm_bo_swapout(&ttm_bo_glob, &ctx) == 0); } EXPORT_SYMBOL(ttm_bo_swapout_all); -/** - * ttm_bo_wait_unreserved - interruptible wait for a buffer object to become - * unreserved - * - * @bo: Pointer to buffer - */ -int ttm_bo_wait_unreserved(struct ttm_buffer_object *bo) +void ttm_bo_tt_destroy(struct ttm_buffer_object *bo) { - int ret; + if (bo->ttm == NULL) + return; - /* - * 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 = reservation_object_lock_interruptible(bo->resv, NULL); - if (ret == -EINTR) - ret = -ERESTARTSYS; - if (unlikely(ret != 0)) - goto out_unlock; - reservation_object_unlock(bo->resv); + ttm_tt_destroy(bo->bdev, bo->ttm); + bo->ttm = NULL; +} -out_unlock: - mutex_unlock(&bo->wu_mutex); - return ret; +int ttm_bo_tt_bind(struct ttm_buffer_object *bo, struct ttm_resource *mem) +{ + return bo->bdev->driver->ttm_tt_bind(bo->bdev, bo->ttm, mem); +} + +void ttm_bo_tt_unbind(struct ttm_buffer_object *bo) +{ + bo->bdev->driver->ttm_tt_unbind(bo->bdev, bo->ttm); } -- Gitblit v1.6.2