From patchwork Wed Oct 27 10:52:09 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: =?utf-8?q?Thomas_Hellstr=C3=B6m?= X-Patchwork-Id: 12587029 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id E6B79C433F5 for ; Wed, 27 Oct 2021 10:53:11 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id B57B86103B for ; Wed, 27 Oct 2021 10:53:11 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.1 mail.kernel.org B57B86103B Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linux.intel.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 940AF6E59F; Wed, 27 Oct 2021 10:52:57 +0000 (UTC) Received: from mga18.intel.com (mga18.intel.com [134.134.136.126]) by gabe.freedesktop.org (Postfix) with ESMTPS id 2E55689AAE; Wed, 27 Oct 2021 10:52:55 +0000 (UTC) X-IronPort-AV: E=McAfee;i="6200,9189,10149"; a="217043357" X-IronPort-AV: E=Sophos;i="5.87,186,1631602800"; d="scan'208";a="217043357" Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by orsmga106.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 27 Oct 2021 03:52:29 -0700 X-IronPort-AV: E=Sophos;i="5.87,186,1631602800"; d="scan'208";a="572869976" Received: from djustese-mobl.ger.corp.intel.com (HELO thellstr-mobl1.intel.com) ([10.249.254.205]) by fmsmga003-auth.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 27 Oct 2021 03:52:27 -0700 From: =?utf-8?q?Thomas_Hellstr=C3=B6m?= To: intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org Cc: maarten.lankhorst@linux.intel.com, matthew.auld@intel.com, =?utf-8?q?Tho?= =?utf-8?q?mas_Hellstr=C3=B6m?= Subject: [PATCH v2 1/3] drm/i915: Introduce refcounted sg-tables Date: Wed, 27 Oct 2021 12:52:09 +0200 Message-Id: <20211027105211.485125-2-thomas.hellstrom@linux.intel.com> X-Mailer: git-send-email 2.31.1 In-Reply-To: <20211027105211.485125-1-thomas.hellstrom@linux.intel.com> References: <20211027105211.485125-1-thomas.hellstrom@linux.intel.com> MIME-Version: 1.0 X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" As we start to introduce asynchronous failsafe object migration, where we update the object state and then submit asynchronous commands we need to record what memory resources are actually used by various part of the command stream. Initially for three purposes: 1) Error capture. 2) Asynchronous migration error recovery. 3) Asynchronous vma bind. At the time where these happens, the object state may have been updated to be several migrations ahead and object sg-tables discarded. In order to make it possible to keep sg-tables with memory resource information for these operations, introduce refcounted sg-tables that aren't freed until the last user is done with them. The alternative would be to reference information sitting on the corresponding ttm_resources which typically have the same lifetime as these refcountes sg_tables, but that leads to other awkward constructs: Due to the design direction chosen for ttm resource managers that would lead to diamond-style inheritance, the LMEM resources may sometimes be prematurely freed, and finally the subclassed struct ttm_resource would have to bleed into the asynchronous vma bind code. Signed-off-by: Thomas Hellström --- drivers/gpu/drm/i915/gem/i915_gem_object.h | 4 +- .../gpu/drm/i915/gem/i915_gem_object_types.h | 3 +- drivers/gpu/drm/i915/gem/i915_gem_shmem.c | 16 +- drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 188 +++++++++++------- drivers/gpu/drm/i915/i915_scatterlist.c | 62 ++++-- drivers/gpu/drm/i915/i915_scatterlist.h | 76 ++++++- drivers/gpu/drm/i915/intel_region_ttm.c | 15 +- drivers/gpu/drm/i915/intel_region_ttm.h | 5 +- drivers/gpu/drm/i915/selftests/mock_region.c | 12 +- 9 files changed, 262 insertions(+), 119 deletions(-) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h index a5479ac7a4ad..c5ab364d4311 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h @@ -624,8 +624,8 @@ struct sg_table *shmem_alloc_st(struct drm_i915_private *i915, size_t size, struct intel_memory_region *mr, struct address_space *mapping, unsigned int max_segment); -void shmem_free_st(struct sg_table *st, struct address_space *mapping, - bool dirty, bool backup); +void shmem_free_st_table(struct sg_table *st, struct address_space *mapping, + bool dirty, bool backup); void __shmem_writeback(size_t size, struct address_space *mapping); #ifdef CONFIG_MMU_NOTIFIER diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h index a4b69a43b898..604ed5ad77f5 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h @@ -544,6 +544,7 @@ struct drm_i915_gem_object { */ struct list_head region_link; + struct i915_refct_sgt *rsgt; struct sg_table *pages; void *mapping; @@ -597,7 +598,7 @@ struct drm_i915_gem_object { } mm; struct { - struct sg_table *cached_io_st; + struct i915_refct_sgt *cached_io_rsgt; struct i915_gem_object_page_iter get_io_page; struct drm_i915_gem_object *backup; bool created:1; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c index 01f332d8dbde..67c6bee695c7 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c @@ -25,8 +25,8 @@ static void check_release_pagevec(struct pagevec *pvec) cond_resched(); } -void shmem_free_st(struct sg_table *st, struct address_space *mapping, - bool dirty, bool backup) +void shmem_free_st_table(struct sg_table *st, struct address_space *mapping, + bool dirty, bool backup) { struct sgt_iter sgt_iter; struct pagevec pvec; @@ -49,7 +49,6 @@ void shmem_free_st(struct sg_table *st, struct address_space *mapping, check_release_pagevec(&pvec); sg_free_table(st); - kfree(st); } struct sg_table *shmem_alloc_st(struct drm_i915_private *i915, @@ -171,7 +170,8 @@ struct sg_table *shmem_alloc_st(struct drm_i915_private *i915, err_sg: sg_mark_end(sg); if (sg != st->sgl) { - shmem_free_st(st, mapping, false, false); + shmem_free_st_table(st, mapping, false, false); + kfree(st); } else { mapping_clear_unevictable(mapping); sg_free_table(st); @@ -254,7 +254,8 @@ static int shmem_get_pages(struct drm_i915_gem_object *obj) return 0; err_pages: - shmem_free_st(st, mapping, false, false); + shmem_free_st_table(st, mapping, false, false); + kfree(st); /* * shmemfs first checks if there is enough memory to allocate the page * and reports ENOSPC should there be insufficient, along with the usual @@ -374,8 +375,9 @@ void i915_gem_object_put_pages_shmem(struct drm_i915_gem_object *obj, struct sg_ if (i915_gem_object_needs_bit17_swizzle(obj)) i915_gem_object_save_bit_17_swizzle(obj, pages); - shmem_free_st(pages, file_inode(obj->base.filp)->i_mapping, - obj->mm.dirty, obj->mm.madv == I915_MADV_WILLNEED); + shmem_free_st_table(pages, file_inode(obj->base.filp)->i_mapping, + obj->mm.dirty, obj->mm.madv == I915_MADV_WILLNEED); + kfree(pages); obj->mm.dirty = false; } diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c index 4fd2edb20dd9..6826e3859e18 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c @@ -34,7 +34,7 @@ * struct i915_ttm_tt - TTM page vector with additional private information * @ttm: The base TTM page vector. * @dev: The struct device used for dma mapping and unmapping. - * @cached_st: The cached scatter-gather table. + * @cached_rsgt: The cached scatter-gather table. * @is_shmem: Set if using shmem. * @filp: The shmem file, if using shmem backend. * @@ -47,7 +47,7 @@ struct i915_ttm_tt { struct ttm_tt ttm; struct device *dev; - struct sg_table *cached_st; + struct i915_refct_sgt cached_rsgt; bool is_shmem; struct file *filp; @@ -221,14 +221,10 @@ static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev, if (IS_ERR(st)) return PTR_ERR(st); - err = dma_map_sg_attrs(i915_tt->dev, - st->sgl, st->nents, - DMA_BIDIRECTIONAL, - DMA_ATTR_SKIP_CPU_SYNC); - if (err <= 0) { - err = -EINVAL; + err = dma_map_sgtable(i915_tt->dev, st, DMA_BIDIRECTIONAL, + DMA_ATTR_SKIP_CPU_SYNC); + if (err) goto err_free_st; - } i = 0; for_each_sgt_page(page, sgt_iter, st) @@ -237,11 +233,15 @@ static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev, if (ttm->page_flags & TTM_TT_FLAG_SWAPPED) ttm->page_flags &= ~TTM_TT_FLAG_SWAPPED; - i915_tt->cached_st = st; + i915_tt->cached_rsgt.table = *st; + kfree(st); + return 0; err_free_st: - shmem_free_st(st, filp->f_mapping, false, false); + shmem_free_st_table(st, filp->f_mapping, false, false); + kfree(st); + return err; } @@ -249,16 +249,29 @@ static void i915_ttm_tt_shmem_unpopulate(struct ttm_tt *ttm) { struct i915_ttm_tt *i915_tt = container_of(ttm, typeof(*i915_tt), ttm); bool backup = ttm->page_flags & TTM_TT_FLAG_SWAPPED; + struct sg_table *st = &i915_tt->cached_rsgt.table; - dma_unmap_sg(i915_tt->dev, i915_tt->cached_st->sgl, - i915_tt->cached_st->nents, - DMA_BIDIRECTIONAL); + if (st->sgl) + shmem_free_st_table(st, + file_inode(i915_tt->filp)->i_mapping, + backup, backup); +} - shmem_free_st(fetch_and_zero(&i915_tt->cached_st), - file_inode(i915_tt->filp)->i_mapping, - backup, backup); +static void i915_ttm_tt_release(struct kref *ref) +{ + struct i915_ttm_tt *i915_tt = + container_of(ref, typeof(*i915_tt), cached_rsgt.kref); + struct sg_table *st = &i915_tt->cached_rsgt.table; + + GEM_WARN_ON(st->sgl); + + kfree(i915_tt); } +static const struct i915_refct_sgt_ops tt_rsgt_ops = { + .release = i915_ttm_tt_release +}; + static struct ttm_tt *i915_ttm_tt_create(struct ttm_buffer_object *bo, uint32_t page_flags) { @@ -287,6 +300,9 @@ static struct ttm_tt *i915_ttm_tt_create(struct ttm_buffer_object *bo, if (ret) goto err_free; + i915_refct_sgt_init_ops(&i915_tt->cached_rsgt, bo->base.size, + &tt_rsgt_ops); + i915_tt->dev = obj->base.dev->dev; return &i915_tt->ttm; @@ -311,17 +327,15 @@ static int i915_ttm_tt_populate(struct ttm_device *bdev, static void i915_ttm_tt_unpopulate(struct ttm_device *bdev, struct ttm_tt *ttm) { struct i915_ttm_tt *i915_tt = container_of(ttm, typeof(*i915_tt), ttm); + struct sg_table *st = &i915_tt->cached_rsgt.table; + + if (st->sgl) + dma_unmap_sgtable(i915_tt->dev, st, DMA_BIDIRECTIONAL, 0); if (i915_tt->is_shmem) { i915_ttm_tt_shmem_unpopulate(ttm); } else { - if (i915_tt->cached_st) { - dma_unmap_sgtable(i915_tt->dev, i915_tt->cached_st, - DMA_BIDIRECTIONAL, 0); - sg_free_table(i915_tt->cached_st); - kfree(i915_tt->cached_st); - i915_tt->cached_st = NULL; - } + sg_free_table(st); ttm_pool_free(&bdev->pool, ttm); } } @@ -334,7 +348,7 @@ static void i915_ttm_tt_destroy(struct ttm_device *bdev, struct ttm_tt *ttm) fput(i915_tt->filp); ttm_tt_fini(ttm); - kfree(i915_tt); + i915_refct_sgt_put(&i915_tt->cached_rsgt); } static bool i915_ttm_eviction_valuable(struct ttm_buffer_object *bo, @@ -376,12 +390,12 @@ static int i915_ttm_move_notify(struct ttm_buffer_object *bo) return 0; } -static void i915_ttm_free_cached_io_st(struct drm_i915_gem_object *obj) +static void i915_ttm_free_cached_io_rsgt(struct drm_i915_gem_object *obj) { struct radix_tree_iter iter; void __rcu **slot; - if (!obj->ttm.cached_io_st) + if (!obj->ttm.cached_io_rsgt) return; rcu_read_lock(); @@ -389,9 +403,8 @@ static void i915_ttm_free_cached_io_st(struct drm_i915_gem_object *obj) radix_tree_delete(&obj->ttm.get_io_page.radix, iter.index); rcu_read_unlock(); - sg_free_table(obj->ttm.cached_io_st); - kfree(obj->ttm.cached_io_st); - obj->ttm.cached_io_st = NULL; + i915_refct_sgt_put(obj->ttm.cached_io_rsgt); + obj->ttm.cached_io_rsgt = NULL; } static void @@ -477,7 +490,7 @@ static int i915_ttm_purge(struct drm_i915_gem_object *obj) obj->write_domain = 0; obj->read_domains = 0; i915_ttm_adjust_gem_after_move(obj); - i915_ttm_free_cached_io_st(obj); + i915_ttm_free_cached_io_rsgt(obj); obj->mm.madv = __I915_MADV_PURGED; return 0; } @@ -532,7 +545,7 @@ static void i915_ttm_swap_notify(struct ttm_buffer_object *bo) int ret = i915_ttm_move_notify(bo); GEM_WARN_ON(ret); - GEM_WARN_ON(obj->ttm.cached_io_st); + GEM_WARN_ON(obj->ttm.cached_io_rsgt); if (!ret && obj->mm.madv != I915_MADV_WILLNEED) i915_ttm_purge(obj); } @@ -543,7 +556,7 @@ static void i915_ttm_delete_mem_notify(struct ttm_buffer_object *bo) if (likely(obj)) { __i915_gem_object_pages_fini(obj); - i915_ttm_free_cached_io_st(obj); + i915_ttm_free_cached_io_rsgt(obj); } } @@ -563,40 +576,35 @@ i915_ttm_region(struct ttm_device *bdev, int ttm_mem_type) ttm_mem_type - I915_PL_LMEM0); } -static struct sg_table *i915_ttm_tt_get_st(struct ttm_tt *ttm) +static struct i915_refct_sgt *i915_ttm_tt_get_st(struct ttm_tt *ttm) { struct i915_ttm_tt *i915_tt = container_of(ttm, typeof(*i915_tt), ttm); struct sg_table *st; int ret; - if (i915_tt->cached_st) - return i915_tt->cached_st; - - st = kzalloc(sizeof(*st), GFP_KERNEL); - if (!st) - return ERR_PTR(-ENOMEM); + if (i915_tt->cached_rsgt.table.sgl) + return i915_refct_sgt_get(&i915_tt->cached_rsgt); + st = &i915_tt->cached_rsgt.table; ret = sg_alloc_table_from_pages_segment(st, ttm->pages, ttm->num_pages, 0, (unsigned long)ttm->num_pages << PAGE_SHIFT, i915_sg_segment_size(), GFP_KERNEL); if (ret) { - kfree(st); + st->sgl = NULL; return ERR_PTR(ret); } ret = dma_map_sgtable(i915_tt->dev, st, DMA_BIDIRECTIONAL, 0); if (ret) { sg_free_table(st); - kfree(st); return ERR_PTR(ret); } - i915_tt->cached_st = st; - return st; + return i915_refct_sgt_get(&i915_tt->cached_rsgt); } -static struct sg_table * +static struct i915_refct_sgt * i915_ttm_resource_get_st(struct drm_i915_gem_object *obj, struct ttm_resource *res) { @@ -610,7 +618,21 @@ i915_ttm_resource_get_st(struct drm_i915_gem_object *obj, * the resulting st. Might make sense for GGTT. */ GEM_WARN_ON(!cpu_maps_iomem(res)); - return intel_region_ttm_resource_to_st(obj->mm.region, res); + if (bo->resource == res) { + if (!obj->ttm.cached_io_rsgt) { + struct i915_refct_sgt *rsgt; + + rsgt = intel_region_ttm_resource_to_rsgt(obj->mm.region, + res); + if (IS_ERR(rsgt)) + return rsgt; + + obj->ttm.cached_io_rsgt = rsgt; + } + return i915_refct_sgt_get(obj->ttm.cached_io_rsgt); + } + + return intel_region_ttm_resource_to_rsgt(obj->mm.region, res); } static int i915_ttm_accel_move(struct ttm_buffer_object *bo, @@ -621,10 +643,7 @@ static int i915_ttm_accel_move(struct ttm_buffer_object *bo, { struct drm_i915_private *i915 = container_of(bo->bdev, typeof(*i915), bdev); - struct ttm_resource_manager *src_man = - ttm_manager_type(bo->bdev, bo->resource->mem_type); struct drm_i915_gem_object *obj = i915_ttm_to_gem(bo); - struct sg_table *src_st; struct i915_request *rq; struct ttm_tt *src_ttm = bo->ttm; enum i915_cache_level src_level, dst_level; @@ -650,17 +669,22 @@ static int i915_ttm_accel_move(struct ttm_buffer_object *bo, } intel_engine_pm_put(i915->gt.migrate.context->engine); } else { - src_st = src_man->use_tt ? i915_ttm_tt_get_st(src_ttm) : - obj->ttm.cached_io_st; + struct i915_refct_sgt *src_rsgt = + i915_ttm_resource_get_st(obj, bo->resource); + + if (IS_ERR(src_rsgt)) + return PTR_ERR(src_rsgt); src_level = i915_ttm_cache_level(i915, bo->resource, src_ttm); intel_engine_pm_get(i915->gt.migrate.context->engine); ret = intel_context_migrate_copy(i915->gt.migrate.context, - NULL, src_st->sgl, src_level, + NULL, src_rsgt->table.sgl, + src_level, gpu_binds_iomem(bo->resource), dst_st->sgl, dst_level, gpu_binds_iomem(dst_mem), &rq); + i915_refct_sgt_put(src_rsgt); if (!ret && rq) { i915_request_wait(rq, 0, MAX_SCHEDULE_TIMEOUT); i915_request_put(rq); @@ -674,13 +698,14 @@ static int i915_ttm_accel_move(struct ttm_buffer_object *bo, static void __i915_ttm_move(struct ttm_buffer_object *bo, bool clear, struct ttm_resource *dst_mem, struct ttm_tt *dst_ttm, - struct sg_table *dst_st, + struct i915_refct_sgt *dst_rsgt, bool allow_accel) { int ret = -EINVAL; if (allow_accel) - ret = i915_ttm_accel_move(bo, clear, dst_mem, dst_ttm, dst_st); + ret = i915_ttm_accel_move(bo, clear, dst_mem, dst_ttm, + &dst_rsgt->table); if (ret) { struct drm_i915_gem_object *obj = i915_ttm_to_gem(bo); struct intel_memory_region *dst_reg, *src_reg; @@ -697,12 +722,13 @@ static void __i915_ttm_move(struct ttm_buffer_object *bo, bool clear, dst_iter = !cpu_maps_iomem(dst_mem) ? ttm_kmap_iter_tt_init(&_dst_iter.tt, dst_ttm) : ttm_kmap_iter_iomap_init(&_dst_iter.io, &dst_reg->iomap, - dst_st, dst_reg->region.start); + &dst_rsgt->table, + dst_reg->region.start); src_iter = !cpu_maps_iomem(bo->resource) ? ttm_kmap_iter_tt_init(&_src_iter.tt, bo->ttm) : ttm_kmap_iter_iomap_init(&_src_iter.io, &src_reg->iomap, - obj->ttm.cached_io_st, + &obj->ttm.cached_io_rsgt->table, src_reg->region.start); ttm_move_memcpy(clear, dst_mem->num_pages, dst_iter, src_iter); @@ -718,7 +744,7 @@ static int i915_ttm_move(struct ttm_buffer_object *bo, bool evict, struct ttm_resource_manager *dst_man = ttm_manager_type(bo->bdev, dst_mem->mem_type); struct ttm_tt *ttm = bo->ttm; - struct sg_table *dst_st; + struct i915_refct_sgt *dst_rsgt; bool clear; int ret; @@ -744,22 +770,24 @@ static int i915_ttm_move(struct ttm_buffer_object *bo, bool evict, return ret; } - dst_st = i915_ttm_resource_get_st(obj, dst_mem); - if (IS_ERR(dst_st)) - return PTR_ERR(dst_st); + dst_rsgt = i915_ttm_resource_get_st(obj, dst_mem); + if (IS_ERR(dst_rsgt)) + return PTR_ERR(dst_rsgt); clear = !cpu_maps_iomem(bo->resource) && (!ttm || !ttm_tt_is_populated(ttm)); if (!(clear && ttm && !(ttm->page_flags & TTM_TT_FLAG_ZERO_ALLOC))) - __i915_ttm_move(bo, clear, dst_mem, bo->ttm, dst_st, true); + __i915_ttm_move(bo, clear, dst_mem, bo->ttm, dst_rsgt, true); ttm_bo_move_sync_cleanup(bo, dst_mem); i915_ttm_adjust_domains_after_move(obj); - i915_ttm_free_cached_io_st(obj); + i915_ttm_free_cached_io_rsgt(obj); if (gpu_binds_iomem(dst_mem) || cpu_maps_iomem(dst_mem)) { - obj->ttm.cached_io_st = dst_st; - obj->ttm.get_io_page.sg_pos = dst_st->sgl; + obj->ttm.cached_io_rsgt = dst_rsgt; + obj->ttm.get_io_page.sg_pos = dst_rsgt->table.sgl; obj->ttm.get_io_page.sg_idx = 0; + } else { + i915_refct_sgt_put(dst_rsgt); } i915_ttm_adjust_lru(obj); @@ -825,7 +853,6 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj, .interruptible = true, .no_wait_gpu = false, }; - struct sg_table *st; int real_num_busy; int ret; @@ -862,12 +889,16 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj, } if (!i915_gem_object_has_pages(obj)) { - /* Object either has a page vector or is an iomem object */ - st = bo->ttm ? i915_ttm_tt_get_st(bo->ttm) : obj->ttm.cached_io_st; - if (IS_ERR(st)) - return PTR_ERR(st); + struct i915_refct_sgt *rsgt = + i915_ttm_resource_get_st(obj, bo->resource); + + if (IS_ERR(rsgt)) + return PTR_ERR(rsgt); - __i915_gem_object_set_pages(obj, st, i915_sg_dma_sizes(st->sgl)); + GEM_BUG_ON(obj->mm.rsgt); + obj->mm.rsgt = rsgt; + __i915_gem_object_set_pages(obj, &rsgt->table, + i915_sg_dma_sizes(rsgt->table.sgl)); } i915_ttm_adjust_lru(obj); @@ -941,6 +972,13 @@ static void i915_ttm_put_pages(struct drm_i915_gem_object *obj, * If the object is not destroyed next, The TTM eviction logic * and shrinkers will move it out if needed. */ + + if (obj->mm.rsgt) { + i915_refct_sgt_put(obj->mm.rsgt); + obj->mm.rsgt = NULL; + } + + i915_ttm_adjust_lru(obj); } static void i915_ttm_adjust_lru(struct drm_i915_gem_object *obj) @@ -1278,7 +1316,7 @@ int i915_gem_obj_copy_ttm(struct drm_i915_gem_object *dst, struct ttm_operation_ctx ctx = { .interruptible = intr, }; - struct sg_table *dst_st; + struct i915_refct_sgt *dst_rsgt; int ret; assert_object_held(dst); @@ -1293,11 +1331,11 @@ int i915_gem_obj_copy_ttm(struct drm_i915_gem_object *dst, if (ret) return ret; - dst_st = gpu_binds_iomem(dst_bo->resource) ? - dst->ttm.cached_io_st : i915_ttm_tt_get_st(dst_bo->ttm); - + dst_rsgt = i915_ttm_resource_get_st(dst, dst_bo->resource); __i915_ttm_move(src_bo, false, dst_bo->resource, dst_bo->ttm, - dst_st, allow_accel); + dst_rsgt, allow_accel); + + i915_refct_sgt_put(dst_rsgt); return 0; } diff --git a/drivers/gpu/drm/i915/i915_scatterlist.c b/drivers/gpu/drm/i915/i915_scatterlist.c index 4a6712dca838..8a510ee5d1ad 100644 --- a/drivers/gpu/drm/i915/i915_scatterlist.c +++ b/drivers/gpu/drm/i915/i915_scatterlist.c @@ -41,8 +41,32 @@ bool i915_sg_trim(struct sg_table *orig_st) return true; } +static void i915_refct_sgt_release(struct kref *ref) +{ + struct i915_refct_sgt *rsgt = + container_of(ref, typeof(*rsgt), kref); + + sg_free_table(&rsgt->table); + kfree(rsgt); +} + +static const struct i915_refct_sgt_ops rsgt_ops = { + .release = i915_refct_sgt_release +}; + +/** + * i915_refct_sgt_init - Initialize a struct i915_refct_sgt with default ops + * @rsgt: The struct i915_refct_sgt to initialize. + * size: The size of the underlying memory buffer. + */ +void i915_refct_sgt_init(struct i915_refct_sgt *rsgt, size_t size) +{ + i915_refct_sgt_init_ops(rsgt, size, &rsgt_ops); +} + /** - * i915_sg_from_mm_node - Create an sg_table from a struct drm_mm_node + * i915_rsgt_from_mm_node - Create a refcounted sg_table from a struct + * drm_mm_node * @node: The drm_mm_node. * @region_start: An offset to add to the dma addresses of the sg list. * @@ -50,25 +74,28 @@ bool i915_sg_trim(struct sg_table *orig_st) * taking a maximum segment length into account, splitting into segments * if necessary. * - * Return: A pointer to a kmalloced struct sg_table on success, negative + * Return: A pointer to a kmalloced struct i915_refct_sgt on success, negative * error code cast to an error pointer on failure. */ -struct sg_table *i915_sg_from_mm_node(const struct drm_mm_node *node, - u64 region_start) +struct i915_refct_sgt *i915_rsgt_from_mm_node(const struct drm_mm_node *node, + u64 region_start) { const u64 max_segment = SZ_1G; /* Do we have a limit on this? */ u64 segment_pages = max_segment >> PAGE_SHIFT; u64 block_size, offset, prev_end; + struct i915_refct_sgt *rsgt; struct sg_table *st; struct scatterlist *sg; - st = kmalloc(sizeof(*st), GFP_KERNEL); - if (!st) + rsgt = kmalloc(sizeof(*rsgt), GFP_KERNEL); + if (!rsgt) return ERR_PTR(-ENOMEM); + i915_refct_sgt_init(rsgt, node->size << PAGE_SHIFT); + st = &rsgt->table; if (sg_alloc_table(st, DIV_ROUND_UP(node->size, segment_pages), GFP_KERNEL)) { - kfree(st); + i915_refct_sgt_put(rsgt); return ERR_PTR(-ENOMEM); } @@ -104,11 +131,11 @@ struct sg_table *i915_sg_from_mm_node(const struct drm_mm_node *node, sg_mark_end(sg); i915_sg_trim(st); - return st; + return rsgt; } /** - * i915_sg_from_buddy_resource - Create an sg_table from a struct + * i915_rsgt_from_buddy_resource - Create a refcounted sg_table from a struct * i915_buddy_block list * @res: The struct i915_ttm_buddy_resource. * @region_start: An offset to add to the dma addresses of the sg list. @@ -117,11 +144,11 @@ struct sg_table *i915_sg_from_mm_node(const struct drm_mm_node *node, * taking a maximum segment length into account, splitting into segments * if necessary. * - * Return: A pointer to a kmalloced struct sg_table on success, negative + * Return: A pointer to a kmalloced struct i915_refct_sgts on success, negative * error code cast to an error pointer on failure. */ -struct sg_table *i915_sg_from_buddy_resource(struct ttm_resource *res, - u64 region_start) +struct i915_refct_sgt *i915_rsgt_from_buddy_resource(struct ttm_resource *res, + u64 region_start) { struct i915_ttm_buddy_resource *bman_res = to_ttm_buddy_resource(res); const u64 size = res->num_pages << PAGE_SHIFT; @@ -129,18 +156,21 @@ struct sg_table *i915_sg_from_buddy_resource(struct ttm_resource *res, struct i915_buddy_mm *mm = bman_res->mm; struct list_head *blocks = &bman_res->blocks; struct i915_buddy_block *block; + struct i915_refct_sgt *rsgt; struct scatterlist *sg; struct sg_table *st; resource_size_t prev_end; GEM_BUG_ON(list_empty(blocks)); - st = kmalloc(sizeof(*st), GFP_KERNEL); - if (!st) + rsgt = kmalloc(sizeof(*rsgt), GFP_KERNEL); + if (!rsgt) return ERR_PTR(-ENOMEM); + i915_refct_sgt_init(rsgt, size); + st = &rsgt->table; if (sg_alloc_table(st, res->num_pages, GFP_KERNEL)) { - kfree(st); + i915_refct_sgt_put(rsgt); return ERR_PTR(-ENOMEM); } @@ -181,7 +211,7 @@ struct sg_table *i915_sg_from_buddy_resource(struct ttm_resource *res, sg_mark_end(sg); i915_sg_trim(st); - return st; + return rsgt; } #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) diff --git a/drivers/gpu/drm/i915/i915_scatterlist.h b/drivers/gpu/drm/i915/i915_scatterlist.h index b8bd5925b03f..321fd4a9f777 100644 --- a/drivers/gpu/drm/i915/i915_scatterlist.h +++ b/drivers/gpu/drm/i915/i915_scatterlist.h @@ -144,10 +144,78 @@ static inline unsigned int i915_sg_segment_size(void) bool i915_sg_trim(struct sg_table *orig_st); -struct sg_table *i915_sg_from_mm_node(const struct drm_mm_node *node, - u64 region_start); +/** + * struct i915_refct_sgt_ops - Operations structure for struct i915_refct_sgt + */ +struct i915_refct_sgt_ops { + /** + * release() - Free the memory of the struct i915_refct_sgt + * @ref: struct kref that is embedded in the struct i915_refct_sgt + */ + void (*release)(struct kref *ref); +}; + +/** + * struct i915_refct_sgt - A refcounted scatter-gather table + * @kref: struct kref for refcounting + * @table: struct sg_table holding the scatter-gather table itself. Note that + * @table->sgl = NULL can be used to determine whether a scatter-gather table + * is present or not. + * @size: The size in bytes of the underlying memory buffer + * @ops: The operations structure. + */ +struct i915_refct_sgt { + struct kref kref; + struct sg_table table; + size_t size; + const struct i915_refct_sgt_ops *ops; +}; + +/** + * i915_refct_sgt_put - Put a refcounted sg-table + * @rsgt the struct i915_refct_sgt to put. + */ +static inline void i915_refct_sgt_put(struct i915_refct_sgt *rsgt) +{ + if (rsgt) + kref_put(&rsgt->kref, rsgt->ops->release); +} + +/** + * i915_refct_sgt_get - Get a refcounted sg-table + * @rsgt the struct i915_refct_sgt to get. + */ +static inline struct i915_refct_sgt * +i915_refct_sgt_get(struct i915_refct_sgt *rsgt) +{ + kref_get(&rsgt->kref); + return rsgt; +} + +/** + * i915_refct_sgt_init_ops - Initialize a refcounted sg-list with a custom + * operations structure + * @rsgt The struct i915_refct_sgt to initialize. + * @size: Size in bytes of the underlying memory buffer. + * @ops: A customized operations structure in case the refcounted sg-list + * is embedded into another structure. + */ +static inline void i915_refct_sgt_init_ops(struct i915_refct_sgt *rsgt, + size_t size, + const struct i915_refct_sgt_ops *ops) +{ + kref_init(&rsgt->kref); + rsgt->table.sgl = NULL; + rsgt->size = size; + rsgt->ops = ops; +} + +void i915_refct_sgt_init(struct i915_refct_sgt *rsgt, size_t size); + +struct i915_refct_sgt *i915_rsgt_from_mm_node(const struct drm_mm_node *node, + u64 region_start); -struct sg_table *i915_sg_from_buddy_resource(struct ttm_resource *res, - u64 region_start); +struct i915_refct_sgt *i915_rsgt_from_buddy_resource(struct ttm_resource *res, + u64 region_start); #endif diff --git a/drivers/gpu/drm/i915/intel_region_ttm.c b/drivers/gpu/drm/i915/intel_region_ttm.c index 98c7339bf8ba..2e901a27e259 100644 --- a/drivers/gpu/drm/i915/intel_region_ttm.c +++ b/drivers/gpu/drm/i915/intel_region_ttm.c @@ -115,8 +115,8 @@ void intel_region_ttm_fini(struct intel_memory_region *mem) } /** - * intel_region_ttm_resource_to_st - Convert an opaque TTM resource manager resource - * to an sg_table. + * intel_region_ttm_resource_to_rsgt - + * Convert an opaque TTM resource manager resource to a refcounted sg_table. * @mem: The memory region. * @res: The resource manager resource obtained from the TTM resource manager. * @@ -126,17 +126,18 @@ void intel_region_ttm_fini(struct intel_memory_region *mem) * * Return: A malloced sg_table on success, an error pointer on failure. */ -struct sg_table *intel_region_ttm_resource_to_st(struct intel_memory_region *mem, - struct ttm_resource *res) +struct i915_refct_sgt * +intel_region_ttm_resource_to_rsgt(struct intel_memory_region *mem, + struct ttm_resource *res) { if (mem->is_range_manager) { struct ttm_range_mgr_node *range_node = to_ttm_range_mgr_node(res); - return i915_sg_from_mm_node(&range_node->mm_nodes[0], - mem->region.start); + return i915_rsgt_from_mm_node(&range_node->mm_nodes[0], + mem->region.start); } else { - return i915_sg_from_buddy_resource(res, mem->region.start); + return i915_rsgt_from_buddy_resource(res, mem->region.start); } } diff --git a/drivers/gpu/drm/i915/intel_region_ttm.h b/drivers/gpu/drm/i915/intel_region_ttm.h index 6f44075920f2..7bbe2b46b504 100644 --- a/drivers/gpu/drm/i915/intel_region_ttm.h +++ b/drivers/gpu/drm/i915/intel_region_ttm.h @@ -22,8 +22,9 @@ int intel_region_ttm_init(struct intel_memory_region *mem); void intel_region_ttm_fini(struct intel_memory_region *mem); -struct sg_table *intel_region_ttm_resource_to_st(struct intel_memory_region *mem, - struct ttm_resource *res); +struct i915_refct_sgt * +intel_region_ttm_resource_to_rsgt(struct intel_memory_region *mem, + struct ttm_resource *res); void intel_region_ttm_resource_free(struct intel_memory_region *mem, struct ttm_resource *res); diff --git a/drivers/gpu/drm/i915/selftests/mock_region.c b/drivers/gpu/drm/i915/selftests/mock_region.c index 75793008c4ef..7ec5037eaa58 100644 --- a/drivers/gpu/drm/i915/selftests/mock_region.c +++ b/drivers/gpu/drm/i915/selftests/mock_region.c @@ -15,9 +15,9 @@ static void mock_region_put_pages(struct drm_i915_gem_object *obj, struct sg_table *pages) { + i915_refct_sgt_put(obj->mm.rsgt); + obj->mm.rsgt = NULL; intel_region_ttm_resource_free(obj->mm.region, obj->mm.res); - sg_free_table(pages); - kfree(pages); } static int mock_region_get_pages(struct drm_i915_gem_object *obj) @@ -36,12 +36,14 @@ static int mock_region_get_pages(struct drm_i915_gem_object *obj) if (IS_ERR(obj->mm.res)) return PTR_ERR(obj->mm.res); - pages = intel_region_ttm_resource_to_st(obj->mm.region, obj->mm.res); - if (IS_ERR(pages)) { - err = PTR_ERR(pages); + obj->mm.rsgt = intel_region_ttm_resource_to_rsgt(obj->mm.region, + obj->mm.res); + if (IS_ERR(obj->mm.rsgt)) { + err = PTR_ERR(obj->mm.rsgt); goto err_free_resource; } + pages = &obj->mm.rsgt->table; __i915_gem_object_set_pages(obj, pages, i915_sg_dma_sizes(pages->sgl)); return 0; From patchwork Wed Oct 27 10:52:10 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: =?utf-8?q?Thomas_Hellstr=C3=B6m?= X-Patchwork-Id: 12587039 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id C7122C433EF for ; Wed, 27 Oct 2021 10:53:14 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 9721B6103B for ; Wed, 27 Oct 2021 10:53:14 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.1 mail.kernel.org 9721B6103B Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linux.intel.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 90C236E881; Wed, 27 Oct 2021 10:52:58 +0000 (UTC) Received: from mga18.intel.com (mga18.intel.com [134.134.136.126]) by gabe.freedesktop.org (Postfix) with ESMTPS id 8529589AC3; Wed, 27 Oct 2021 10:52:55 +0000 (UTC) X-IronPort-AV: E=McAfee;i="6200,9189,10149"; a="217043361" X-IronPort-AV: E=Sophos;i="5.87,186,1631602800"; d="scan'208";a="217043361" Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by orsmga106.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 27 Oct 2021 03:52:31 -0700 X-IronPort-AV: E=Sophos;i="5.87,186,1631602800"; d="scan'208";a="572869990" Received: from djustese-mobl.ger.corp.intel.com (HELO thellstr-mobl1.intel.com) ([10.249.254.205]) by fmsmga003-auth.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 27 Oct 2021 03:52:29 -0700 From: =?utf-8?q?Thomas_Hellstr=C3=B6m?= To: intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org Cc: maarten.lankhorst@linux.intel.com, matthew.auld@intel.com, =?utf-8?q?Tho?= =?utf-8?q?mas_Hellstr=C3=B6m?= Subject: [PATCH v2 2/3] drm/i915: Update error capture code to avoid using the current vma state Date: Wed, 27 Oct 2021 12:52:10 +0200 Message-Id: <20211027105211.485125-3-thomas.hellstrom@linux.intel.com> X-Mailer: git-send-email 2.31.1 In-Reply-To: <20211027105211.485125-1-thomas.hellstrom@linux.intel.com> References: <20211027105211.485125-1-thomas.hellstrom@linux.intel.com> MIME-Version: 1.0 X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" With asynchronous migrations, the vma state may be several migrations ahead of the state that matches the request we're capturing. Address that by introducing an i915_vma_snapshot structure that can be used to snapshot relevant state at request submission. In order to make sure we access the correct memory, the snapshots take references on relevant sg-tables and memory regions. Also move the capture list allocation out of the fence signaling critical path and use the CONFIG_DRM_I915_CAPTURE_ERROR define to avoid compiling in members and functions used for error capture when they're not used. Finally, correct lockdep annotation would reveal that error capture is typically done in the fence signalling critical path. Alter the error capture memory allocation mode accordingly. Signed-off-by: Thomas Hellström Reported-by: kernel test robot --- drivers/gpu/drm/i915/Makefile | 1 + .../gpu/drm/i915/gem/i915_gem_execbuffer.c | 135 ++++++++++--- drivers/gpu/drm/i915/gt/intel_engine_cs.c | 8 +- .../drm/i915/gt/intel_execlists_submission.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 180 +++++++++++++----- drivers/gpu/drm/i915/i915_request.c | 63 ++++-- drivers/gpu/drm/i915/i915_request.h | 18 +- drivers/gpu/drm/i915/i915_vma_snapshot.c | 137 +++++++++++++ drivers/gpu/drm/i915/i915_vma_snapshot.h | 112 +++++++++++ 9 files changed, 558 insertions(+), 98 deletions(-) create mode 100644 drivers/gpu/drm/i915/i915_vma_snapshot.c create mode 100644 drivers/gpu/drm/i915/i915_vma_snapshot.h diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 467872cca027..2424c19cd0bc 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -173,6 +173,7 @@ i915-y += \ i915_trace_points.o \ i915_ttm_buddy_manager.o \ i915_vma.o \ + i915_vma_snapshot.o \ intel_wopcm.o # general-purpose microcontroller (GuC) support diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index ea5b7b2a4d70..301eb58bebd1 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -29,6 +29,7 @@ #include "i915_gem_ioctls.h" #include "i915_trace.h" #include "i915_user_extensions.h" +#include "i915_vma_snapshot.h" struct eb_vma { struct i915_vma *vma; @@ -307,11 +308,15 @@ struct i915_execbuffer { struct eb_fence *fences; unsigned long num_fences; +#ifdef CONFIG_DRM_I915_CAPTURE_ERROR + struct i915_capture_list *capture_lists[MAX_ENGINE_INSTANCE + 1]; +#endif }; static int eb_parse(struct i915_execbuffer *eb); static int eb_pin_engine(struct i915_execbuffer *eb, bool throttle); static void eb_unpin_engine(struct i915_execbuffer *eb); +static void eb_capture_release(struct i915_execbuffer *eb); static inline bool eb_use_cmdparser(const struct i915_execbuffer *eb) { @@ -1054,6 +1059,7 @@ static void eb_release_vmas(struct i915_execbuffer *eb, bool final) i915_vma_put(vma); } + eb_capture_release(eb); eb_unpin_engine(eb); } @@ -1891,36 +1897,113 @@ eb_find_first_request_added(struct i915_execbuffer *eb) return NULL; } -static int eb_move_to_gpu(struct i915_execbuffer *eb) +#ifdef CONFIG_DRM_I915_CAPTURE_ERROR + +/* Stage with GFP_KERNEL allocations before we enter the signaling critical path */ +static void eb_capture_stage(struct i915_execbuffer *eb) { const unsigned int count = eb->buffer_count; - unsigned int i = count; - int err = 0, j; + unsigned int i = count, j; + struct i915_vma_snapshot *vsnap; while (i--) { struct eb_vma *ev = &eb->vma[i]; struct i915_vma *vma = ev->vma; unsigned int flags = ev->flags; - struct drm_i915_gem_object *obj = vma->obj; - assert_vma_held(vma); + if (!(flags & EXEC_OBJECT_CAPTURE)) + continue; - if (flags & EXEC_OBJECT_CAPTURE) { + vsnap = i915_vma_snapshot_alloc(GFP_KERNEL); + if (!vsnap) + continue; + + i915_vma_snapshot_init(vsnap, vma, "user"); + for_each_batch_create_order(eb, j) { struct i915_capture_list *capture; - for_each_batch_create_order(eb, j) { - if (!eb->requests[j]) - break; + capture = kmalloc(sizeof(*capture), GFP_KERNEL); + if (!capture) + continue; - capture = kmalloc(sizeof(*capture), GFP_KERNEL); - if (capture) { - capture->next = - eb->requests[j]->capture_list; - capture->vma = vma; - eb->requests[j]->capture_list = capture; - } - } + capture->next = eb->capture_lists[j]; + capture->vma_snapshot = i915_vma_snapshot_get(vsnap); + eb->capture_lists[j] = capture; + } + i915_vma_snapshot_put(vsnap); + } +} + +/* Commit once we're in the critical path */ +static void eb_capture_commit(struct i915_execbuffer *eb) +{ + unsigned int j; + + for_each_batch_create_order(eb, j) { + struct i915_request *rq = eb->requests[j]; + + if (!rq) + break; + + rq->capture_list = eb->capture_lists[j]; + eb->capture_lists[j] = NULL; + } +} + +/* + * Release anything that didn't get committed due to errors. + * The capture_list will otherwise be freed at request retire. + */ +static void eb_capture_release(struct i915_execbuffer *eb) +{ + unsigned int j; + + for_each_batch_create_order(eb, j) { + if (eb->capture_lists[j]) { + i915_request_free_capture_list(eb->capture_lists[j]); + eb->capture_lists[j] = NULL; } + } +} + +static void eb_capture_list_clear(struct i915_execbuffer *eb) +{ + memset(eb->capture_lists, 0, sizeof(eb->capture_lists)); +} + +#else + +static void eb_capture_stage(struct i915_execbuffer *eb) +{ +} + +static void eb_capture_commit(struct i915_execbuffer *eb) +{ +} + +static void eb_capture_release(struct i915_execbuffer *eb) +{ +} + +static void eb_capture_list_clear(struct i915_execbuffer *eb) +{ +} + +#endif + +static int eb_move_to_gpu(struct i915_execbuffer *eb) +{ + const unsigned int count = eb->buffer_count; + unsigned int i = count; + int err = 0, j; + + while (i--) { + struct eb_vma *ev = &eb->vma[i]; + struct i915_vma *vma = ev->vma; + unsigned int flags = ev->flags; + struct drm_i915_gem_object *obj = vma->obj; + + assert_vma_held(vma); /* * If the GPU is not _reading_ through the CPU cache, we need @@ -2001,6 +2084,8 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb) /* Unconditionally flush any chipset caches (for streaming writes). */ intel_gt_chipset_flush(eb->gt); + eb_capture_commit(eb); + return 0; err_skip: @@ -3143,13 +3228,14 @@ eb_requests_create(struct i915_execbuffer *eb, struct dma_fence *in_fence, } /* - * Whilst this request exists, batch_obj will be on the - * active_list, and so will hold the active reference. Only when - * this request is retired will the batch_obj be moved onto - * the inactive_list and lose its active reference. Hence we do - * not need to explicitly hold another reference here. + * Not really on stack, but we don't want to call + * kfree on the batch_snapshot when we put it, so use the + * _onstack interface. */ - eb->requests[i]->batch = eb->batches[i]->vma; + if (eb->batches[i]->vma) + i915_vma_snapshot_init_onstack(&eb->requests[i]->batch_snapshot, + eb->batches[i]->vma, + "batch"); if (eb->batch_pool) { GEM_BUG_ON(intel_context_is_parallel(eb->context)); intel_gt_buffer_pool_mark_active(eb->batch_pool, @@ -3198,6 +3284,8 @@ i915_gem_do_execbuffer(struct drm_device *dev, eb.fences = NULL; eb.num_fences = 0; + eb_capture_list_clear(&eb); + memset(eb.requests, 0, sizeof(struct i915_request *) * ARRAY_SIZE(eb.requests)); eb.composite_fence = NULL; @@ -3284,6 +3372,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, } ww_acquire_done(&eb.ww.ctx); + eb_capture_stage(&eb); out_fence = eb_requests_create(&eb, in_fence, out_fence_fd); if (IS_ERR(out_fence)) { diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index ff6753ccb129..61e44185990a 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -1676,14 +1676,18 @@ static void intel_engine_print_registers(struct intel_engine_cs *engine, static void print_request_ring(struct drm_printer *m, struct i915_request *rq) { + struct i915_vma_snapshot *vsnap = &rq->batch_snapshot; void *ring; int size; + if (!i915_vma_snapshot_present(vsnap)) + vsnap = NULL; + drm_printf(m, "[head %04x, postfix %04x, tail %04x, batch 0x%08x_%08x]:\n", rq->head, rq->postfix, rq->tail, - rq->batch ? upper_32_bits(rq->batch->node.start) : ~0u, - rq->batch ? lower_32_bits(rq->batch->node.start) : ~0u); + vsnap ? upper_32_bits(vsnap->gtt_offset) : ~0u, + vsnap ? lower_32_bits(vsnap->gtt_offset) : ~0u); size = rq->tail - rq->head; if (rq->tail < rq->head) diff --git a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c index bedb80057046..620c7a262ad0 100644 --- a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c +++ b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c @@ -2186,7 +2186,7 @@ struct execlists_capture { static void execlists_capture_work(struct work_struct *work) { struct execlists_capture *cap = container_of(work, typeof(*cap), work); - const gfp_t gfp = GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN; + const gfp_t gfp = GFP_NOWAIT | __GFP_RETRY_MAYFAIL | __GFP_NOWARN; struct intel_engine_cs *engine = cap->rq->engine; struct intel_gt_coredump *gt = cap->error->gt; struct intel_engine_capture_vma *vma; diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 2a2d7643b551..397df5e473fd 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -48,8 +48,9 @@ #include "i915_gpu_error.h" #include "i915_memcpy.h" #include "i915_scatterlist.h" +#include "i915_vma_snapshot.h" -#define ALLOW_FAIL (GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN) +#define ALLOW_FAIL (GFP_NOWAIT | __GFP_RETRY_MAYFAIL | __GFP_NOWARN) #define ATOMIC_MAYFAIL (GFP_ATOMIC | __GFP_NOWARN) static void __sg_set_buf(struct scatterlist *sg, @@ -1009,8 +1010,7 @@ void __i915_gpu_coredump_free(struct kref *error_ref) static struct i915_vma_coredump * i915_vma_coredump_create(const struct intel_gt *gt, - const struct i915_vma *vma, - const char *name, + const struct i915_vma_snapshot *vsnap, struct i915_vma_compress *compress) { struct i915_ggtt *ggtt = gt->ggtt; @@ -1022,10 +1022,10 @@ i915_vma_coredump_create(const struct intel_gt *gt, might_sleep(); - if (!vma || !vma->pages || !compress) + if (!vsnap || !vsnap->pages || !compress) return NULL; - num_pages = min_t(u64, vma->size, vma->obj->base.size) >> PAGE_SHIFT; + num_pages = min_t(u64, vsnap->size, vsnap->obj_size) >> PAGE_SHIFT; num_pages = DIV_ROUND_UP(10 * num_pages, 8); /* worstcase zlib growth */ dst = kmalloc(sizeof(*dst) + num_pages * sizeof(u32 *), ALLOW_FAIL); if (!dst) @@ -1036,12 +1036,12 @@ i915_vma_coredump_create(const struct intel_gt *gt, return NULL; } - strcpy(dst->name, name); + strcpy(dst->name, vsnap->name); dst->next = NULL; - dst->gtt_offset = vma->node.start; - dst->gtt_size = vma->node.size; - dst->gtt_page_sizes = vma->page_sizes.gtt; + dst->gtt_offset = vsnap->gtt_offset; + dst->gtt_size = vsnap->gtt_size; + dst->gtt_page_sizes = vsnap->page_sizes; dst->num_pages = num_pages; dst->page_count = 0; dst->unused = 0; @@ -1051,7 +1051,7 @@ i915_vma_coredump_create(const struct intel_gt *gt, void __iomem *s; dma_addr_t dma; - for_each_sgt_daddr(dma, iter, vma->pages) { + for_each_sgt_daddr(dma, iter, vsnap->pages) { mutex_lock(&ggtt->error_mutex); ggtt->vm.insert_page(&ggtt->vm, dma, slot, I915_CACHE_NONE, 0); @@ -1069,11 +1069,11 @@ i915_vma_coredump_create(const struct intel_gt *gt, if (ret) break; } - } else if (__i915_gem_object_is_lmem(vma->obj)) { - struct intel_memory_region *mem = vma->obj->mm.region; + } else if (vsnap->mr && vsnap->mr->type != INTEL_MEMORY_SYSTEM) { + struct intel_memory_region *mem = vsnap->mr; dma_addr_t dma; - for_each_sgt_daddr(dma, iter, vma->pages) { + for_each_sgt_daddr(dma, iter, vsnap->pages) { void __iomem *s; s = io_mapping_map_wc(&mem->iomap, @@ -1089,7 +1089,7 @@ i915_vma_coredump_create(const struct intel_gt *gt, } else { struct page *page; - for_each_sgt_page(page, iter, vma->pages) { + for_each_sgt_page(page, iter, vsnap->pages) { void *s; drm_clflush_pages(&page, 1); @@ -1320,37 +1320,71 @@ static bool record_context(struct i915_gem_context_coredump *e, struct intel_engine_capture_vma { struct intel_engine_capture_vma *next; - struct i915_vma *vma; + struct i915_vma_snapshot *vsnap; char name[16]; + bool lockdep_cookie; }; static struct intel_engine_capture_vma * -capture_vma(struct intel_engine_capture_vma *next, - struct i915_vma *vma, - const char *name, - gfp_t gfp) +capture_vma_snapshot(struct intel_engine_capture_vma *next, + struct i915_vma_snapshot *vsnap, + gfp_t gfp) { struct intel_engine_capture_vma *c; - if (!vma) + if (!i915_vma_snapshot_present(vsnap)) return next; c = kmalloc(sizeof(*c), gfp); if (!c) return next; - if (!i915_active_acquire_if_busy(&vma->active)) { + if (!i915_vma_snapshot_resource_pin(vsnap, &c->lockdep_cookie)) { kfree(c); return next; } - strcpy(c->name, name); - c->vma = vma; /* reference held while active */ + strcpy(c->name, vsnap->name); + c->vsnap = vsnap; + i915_vma_snapshot_get(vsnap); c->next = next; return c; } +static struct intel_engine_capture_vma * +capture_vma(struct intel_engine_capture_vma *next, + struct i915_vma *vma, + const char *name, + gfp_t gfp) +{ + struct i915_vma_snapshot *vsnap; + + if (!vma) + return next; + + /* + * If the vma isn't pinned, then the vma should be snapshotted + * to a struct i915_vma_snapshot at command submission time. + * Not here. + */ + GEM_WARN_ON(!i915_vma_is_pinned(vma)); + if (!i915_vma_is_pinned(vma)) + return next; + + vsnap = i915_vma_snapshot_alloc(gfp); + if (!vsnap) + return next; + + i915_vma_snapshot_init(vsnap, vma, name); + next = capture_vma_snapshot(next, vsnap, gfp); + + /* FIXME: Replace on async unbind. */ + i915_vma_snapshot_put(vsnap); + + return next; +} + static struct intel_engine_capture_vma * capture_user(struct intel_engine_capture_vma *capture, const struct i915_request *rq, @@ -1359,7 +1393,7 @@ capture_user(struct intel_engine_capture_vma *capture, struct i915_capture_list *c; for (c = rq->capture_list; c; c = c->next) - capture = capture_vma(capture, c->vma, "user", gfp); + capture = capture_vma_snapshot(capture, c->vma_snapshot, gfp); return capture; } @@ -1373,6 +1407,36 @@ static void add_vma(struct intel_engine_coredump *ee, } } +static struct i915_vma_coredump * +create_vma_coredump(const struct intel_gt *gt, struct i915_vma *vma, + const char *name, struct i915_vma_compress *compress) +{ + struct i915_vma_coredump *ret = NULL; + struct i915_vma_snapshot tmp; + bool lockdep_cookie; + + if (!vma) + return NULL; + + i915_vma_snapshot_init_onstack(&tmp, vma, name); + if (i915_vma_snapshot_resource_pin(&tmp, &lockdep_cookie)) { + ret = i915_vma_coredump_create(gt, &tmp, compress); + i915_vma_snapshot_resource_unpin(&tmp, lockdep_cookie); + } + i915_vma_snapshot_put_onstack(&tmp); + + return ret; +} + +static void add_vma_coredump(struct intel_engine_coredump *ee, + const struct intel_gt *gt, + struct i915_vma *vma, + const char *name, + struct i915_vma_compress *compress) +{ + add_vma(ee, create_vma_coredump(gt, vma, name, compress)); +} + struct intel_engine_coredump * intel_engine_coredump_alloc(struct intel_engine_cs *engine, gfp_t gfp) { @@ -1406,7 +1470,7 @@ intel_engine_coredump_add_request(struct intel_engine_coredump *ee, * as the simplest method to avoid being overwritten * by userspace. */ - vma = capture_vma(vma, rq->batch, "batch", gfp); + vma = capture_vma_snapshot(vma, &rq->batch_snapshot, gfp); vma = capture_user(vma, rq, gfp); vma = capture_vma(vma, rq->ring->vma, "ring", gfp); vma = capture_vma(vma, rq->context->state, "HW context", gfp); @@ -1427,30 +1491,24 @@ intel_engine_coredump_add_vma(struct intel_engine_coredump *ee, while (capture) { struct intel_engine_capture_vma *this = capture; - struct i915_vma *vma = this->vma; + struct i915_vma_snapshot *vsnap = this->vsnap; add_vma(ee, i915_vma_coredump_create(engine->gt, - vma, this->name, - compress)); + vsnap, compress)); - i915_active_release(&vma->active); + i915_vma_snapshot_resource_unpin(vsnap, this->lockdep_cookie); + i915_vma_snapshot_put(vsnap); capture = this->next; kfree(this); } - add_vma(ee, - i915_vma_coredump_create(engine->gt, - engine->status_page.vma, - "HW Status", - compress)); + add_vma_coredump(ee, engine->gt, engine->status_page.vma, + "HW Status", compress); - add_vma(ee, - i915_vma_coredump_create(engine->gt, - engine->wa_ctx.vma, - "WA context", - compress)); + add_vma_coredump(ee, engine->gt, engine->wa_ctx.vma, + "WA context", compress); } static struct intel_engine_coredump * @@ -1486,17 +1544,25 @@ capture_engine(struct intel_engine_cs *engine, } } if (rq) - capture = intel_engine_coredump_add_request(ee, rq, - ATOMIC_MAYFAIL); + rq = i915_request_get_rcu(rq); + + if (!rq) + goto no_request_capture; + + capture = intel_engine_coredump_add_request(ee, rq, ATOMIC_MAYFAIL); if (!capture) { -no_request_capture: - kfree(ee); - return NULL; + i915_request_put(rq); + goto no_request_capture; } intel_engine_coredump_add_vma(ee, capture, compress); + i915_request_put(rq); return ee; + +no_request_capture: + kfree(ee); + return NULL; } static void @@ -1550,10 +1616,8 @@ gt_record_uc(struct intel_gt_coredump *gt, */ error_uc->guc_fw.path = kstrdup(uc->guc.fw.path, ALLOW_FAIL); error_uc->huc_fw.path = kstrdup(uc->huc.fw.path, ALLOW_FAIL); - error_uc->guc_log = - i915_vma_coredump_create(gt->_gt, - uc->guc.log.vma, "GuC log buffer", - compress); + error_uc->guc_log = create_vma_coredump(gt->_gt, uc->guc.log.vma, + "GuC log buffer", compress); return error_uc; } @@ -1839,8 +1903,8 @@ void i915_vma_capture_finish(struct intel_gt_coredump *gt, kfree(compress); } -struct i915_gpu_coredump * -i915_gpu_coredump(struct intel_gt *gt, intel_engine_mask_t engine_mask) +static struct i915_gpu_coredump * +__i915_gpu_coredump(struct intel_gt *gt, intel_engine_mask_t engine_mask) { struct drm_i915_private *i915 = gt->i915; struct i915_gpu_coredump *error; @@ -1881,6 +1945,22 @@ i915_gpu_coredump(struct intel_gt *gt, intel_engine_mask_t engine_mask) return error; } +struct i915_gpu_coredump * +i915_gpu_coredump(struct intel_gt *gt, intel_engine_mask_t engine_mask) +{ + static DEFINE_MUTEX(capture_mutex); + int ret = mutex_lock_interruptible(&capture_mutex); + struct i915_gpu_coredump *dump; + + if (ret) + return ERR_PTR(ret); + + dump = __i915_gpu_coredump(gt, engine_mask); + mutex_unlock(&capture_mutex); + + return dump; +} + void i915_error_state_store(struct i915_gpu_coredump *error) { struct drm_i915_private *i915; diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c index 820a1f38b271..24ec2a9beb2f 100644 --- a/drivers/gpu/drm/i915/i915_request.c +++ b/drivers/gpu/drm/i915/i915_request.c @@ -113,6 +113,10 @@ static void i915_fence_release(struct dma_fence *fence) GEM_BUG_ON(rq->guc_prio != GUC_PRIO_INIT && rq->guc_prio != GUC_PRIO_FINI); + i915_request_free_capture_list(fetch_and_zero(&rq->capture_list)); + if (i915_vma_snapshot_present(&rq->batch_snapshot)) + i915_vma_snapshot_put_onstack(&rq->batch_snapshot); + /* * The request is put onto a RCU freelist (i.e. the address * is immediately reused), mark the fences as being freed now. @@ -186,19 +190,6 @@ void i915_request_notify_execute_cb_imm(struct i915_request *rq) __notify_execute_cb(rq, irq_work_imm); } -static void free_capture_list(struct i915_request *request) -{ - struct i915_capture_list *capture; - - capture = fetch_and_zero(&request->capture_list); - while (capture) { - struct i915_capture_list *next = capture->next; - - kfree(capture); - capture = next; - } -} - static void __i915_request_fill(struct i915_request *rq, u8 val) { void *vaddr = rq->ring->vaddr; @@ -303,6 +294,37 @@ static void __rq_cancel_watchdog(struct i915_request *rq) i915_request_put(rq); } +#ifdef CONFIG_DRM_I915_CAPTURE_ERROR + +/** + * i915_request_free_capture_list - Free a capture list + * @capture: Pointer to the first list item or NULL + * + */ +void i915_request_free_capture_list(struct i915_capture_list *capture) +{ + while (capture) { + struct i915_capture_list *next = capture->next; + + i915_vma_snapshot_put(capture->vma_snapshot); + capture = next; + } +} + +#define assert_capture_list_is_null(_rq) GEM_BUG_ON((_rq)->capture_list) + +#define clear_capture_list(_rq) ((_rq)->capture_list = NULL) + +#else + +#define i915_request_free_capture_list(_a) do {} while (0) + +#define assert_capture_list_is_null(_a) do {} while (0) + +#define clear_capture_list(_rq) do {} while (0) + +#endif + bool i915_request_retire(struct i915_request *rq) { if (!__i915_request_is_complete(rq)) @@ -359,7 +381,6 @@ bool i915_request_retire(struct i915_request *rq) intel_context_exit(rq->context); intel_context_unpin(rq->context); - free_capture_list(rq); i915_sched_node_fini(&rq->sched); i915_request_put(rq); @@ -829,11 +850,18 @@ static void __i915_request_ctor(void *arg) i915_sw_fence_init(&rq->submit, submit_notify); i915_sw_fence_init(&rq->semaphore, semaphore_notify); - rq->capture_list = NULL; + clear_capture_list(rq); + rq->batch_snapshot.present = false; init_llist_head(&rq->execute_cb); } +#ifdef CONFIG_DRM_I915_SELFTEST +#define clear_batch_ptr(_rq) ((_rq)->batch = NULL) +#else +#define clear_batch_ptr(_a) do {} while (0) +#endif + struct i915_request * __i915_request_create(struct intel_context *ce, gfp_t gfp) { @@ -925,10 +953,11 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp) i915_sched_node_reinit(&rq->sched); /* No zalloc, everything must be cleared after use */ - rq->batch = NULL; + clear_batch_ptr(rq); __rq_init_watchdog(rq); - GEM_BUG_ON(rq->capture_list); + assert_capture_list_is_null(rq); GEM_BUG_ON(!llist_empty(&rq->execute_cb)); + GEM_BUG_ON(i915_vma_snapshot_present(&rq->batch_snapshot)); /* * Reserve space in the ring buffer for all the commands required to diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h index dc359242d1ae..f439bf968517 100644 --- a/drivers/gpu/drm/i915/i915_request.h +++ b/drivers/gpu/drm/i915/i915_request.h @@ -40,6 +40,7 @@ #include "i915_scheduler.h" #include "i915_selftest.h" #include "i915_sw_fence.h" +#include "i915_vma_snapshot.h" #include @@ -48,11 +49,15 @@ struct drm_i915_gem_object; struct drm_printer; struct i915_request; +#ifdef CONFIG_DRM_I915_CAPTURE_ERROR struct i915_capture_list { + struct i915_vma_snapshot *vma_snapshot; struct i915_capture_list *next; - struct i915_vma *vma; }; +void i915_request_free_capture_list(struct i915_capture_list *capture); +#endif + #define RQ_TRACE(rq, fmt, ...) do { \ const struct i915_request *rq__ = (rq); \ ENGINE_TRACE(rq__->engine, "fence %llx:%lld, current %d " fmt, \ @@ -289,10 +294,12 @@ struct i915_request { /** Preallocate space in the ring for the emitting the request */ u32 reserved_space; - /** Batch buffer related to this request if any (used for - * error state dump only). - */ - struct i915_vma *batch; + /** Batch buffer pointer for selftest internal use. */ + I915_SELFTEST_DECLARE(struct i915_vma *batch); + + struct i915_vma_snapshot batch_snapshot; + +#ifdef CONFIG_DRM_I915_CAPTURE_ERROR /** * Additional buffers requested by userspace to be captured upon * a GPU hang. The vma/obj on this list are protected by their @@ -300,6 +307,7 @@ struct i915_request { * on the active_list (of their final request). */ struct i915_capture_list *capture_list; +#endif /** Time at which this request was emitted, in jiffies. */ unsigned long emitted_jiffies; diff --git a/drivers/gpu/drm/i915/i915_vma_snapshot.c b/drivers/gpu/drm/i915/i915_vma_snapshot.c new file mode 100644 index 000000000000..44985d600f96 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_vma_snapshot.c @@ -0,0 +1,137 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2021 Intel Corporation + */ + +#include "i915_vma_snapshot.h" +#include "i915_vma_types.h" +#include "i915_vma.h" + +/** + * i915_vma_snapshot_init - Initialize a struct i915_vma_snapshot from + * a struct i915_vma. + * @vsnap: The i915_vma_snapshot to init. + * @vma: A struct i915_vma used to initialize @vsnap. + * @name: Name associated with the snapshot. The character pointer needs to + * stay alive over the lifitime of the shapsot + */ +void i915_vma_snapshot_init(struct i915_vma_snapshot *vsnap, + struct i915_vma *vma, + const char *name) +{ + if (!i915_vma_is_pinned(vma)) + assert_object_held(vma->obj); + + vsnap->name = name; + vsnap->size = vma->size; + vsnap->obj_size = vma->obj->base.size; + vsnap->gtt_offset = vma->node.start; + vsnap->gtt_size = vma->node.size; + vsnap->page_sizes = vma->page_sizes.gtt; + vsnap->pages = vma->pages; + vsnap->pages_rsgt = NULL; + vsnap->mr = NULL; + if (vma->obj->mm.rsgt) + vsnap->pages_rsgt = i915_refct_sgt_get(vma->obj->mm.rsgt); + if (vma->obj->mm.region) + vsnap->mr = intel_memory_region_get(vma->obj->mm.region); + kref_init(&vsnap->kref); + vsnap->vma_resource = &vma->active; + vsnap->onstack = false; + vsnap->present = true; +} + +/** + * i915_vma_snapshot_init_onstack - Initialize a struct i915_vma_snapshot from + * a struct i915_vma, but avoid kfreeing it on last put. + * @vsnap: The i915_vma_snapshot to init. + * @vma: A struct i915_vma used to initialize @vsnap. + * @name: Name associated with the snapshot. The character pointer needs to + * stay alive over the lifitime of the shapsot + */ +void i915_vma_snapshot_init_onstack(struct i915_vma_snapshot *vsnap, + struct i915_vma *vma, + const char *name) +{ + i915_vma_snapshot_init(vsnap, vma, name); + vsnap->onstack = true; +} + +static void vma_snapshot_release(struct kref *ref) +{ + struct i915_vma_snapshot *vsnap = + container_of(ref, typeof(*vsnap), kref); + + vsnap->present = false; + if (vsnap->mr) + intel_memory_region_put(vsnap->mr); + if (vsnap->pages_rsgt) + i915_refct_sgt_put(vsnap->pages_rsgt); + if (!vsnap->onstack) + kfree(vsnap); +} + +/** + * i915_vma_snapshot_put - Put an i915_vma_snapshot pointer reference + * @vsnap: The pointer reference + */ +void i915_vma_snapshot_put(struct i915_vma_snapshot *vsnap) +{ + kref_put(&vsnap->kref, vma_snapshot_release); +} + +/** + * i915_vma_snapshot_put_onstack - Put an onstcak i915_vma_snapshot pointer + * reference and varify that the structure is released + * @vsnap: The pointer reference + * + * This function is intended to be paired with a i915_vma_init_onstack() + * and should be called before exiting the scope that declared or + * freeing the structure that embedded @vsnap to verify that all references + * have been released. + */ +void i915_vma_snapshot_put_onstack(struct i915_vma_snapshot *vsnap) +{ + if (!kref_put(&vsnap->kref, vma_snapshot_release)) + GEM_BUG_ON(1); +} + +/** + * i915_vma_snapshot_resource_pin - Temporarily block the memory the + * vma snapshot is pointing to from being released. + * @vsnap: The vma snapshot. + * @lockdep_cookie: Pointer to bool needed for lockdep support. This needs + * to be passed to the paired i915_vma_snapshot_resource_unpin. + * + * This function will temporarily try to hold up a fence or similar structure + * and will therefore enter a fence signaling critical section. + * + * Return: true if we succeeded in blocking the memory from being released, + * false otherwise. + */ +bool i915_vma_snapshot_resource_pin(struct i915_vma_snapshot *vsnap, + bool *lockdep_cookie) +{ + bool pinned = i915_active_acquire_if_busy(vsnap->vma_resource); + + if (pinned) + *lockdep_cookie = dma_fence_begin_signalling(); + + return pinned; +} + +/** + * i915_vma_snapshot_resource_unpin - Unblock vma snapshot memory from + * being released. + * @vsnap: The vma snapshot. + * @lockdep_cookie: Cookie returned from matching i915_vma_resource_pin(). + * + * Might leave a fence signalling critical section and signal a fence. + */ +void i915_vma_snapshot_resource_unpin(struct i915_vma_snapshot *vsnap, + bool lockdep_cookie) +{ + dma_fence_end_signalling(lockdep_cookie); + + return i915_active_release(vsnap->vma_resource); +} diff --git a/drivers/gpu/drm/i915/i915_vma_snapshot.h b/drivers/gpu/drm/i915/i915_vma_snapshot.h new file mode 100644 index 000000000000..940581df4622 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_vma_snapshot.h @@ -0,0 +1,112 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2021 Intel Corporation + */ +#ifndef _I915_VMA_SNAPSHOT_H_ +#define _I915_VMA_SNAPSHOT_H_ + +#include +#include +#include + +struct i915_active; +struct i915_refct_sgt; +struct i915_vma; +struct intel_memory_region; +struct sg_table; + +/** + * DOC: Simple utilities for snapshotting GPU vma metadata, later used for + * error capture. Vi use a separate header for this to avoid issues due to + * recursive header includes. + */ + +/** + * struct i915_vma_snapshot - Snapshot of vma metadata. + * @size: The vma size in bytes. + * @obj_size: The size of the underlying object in bytes. + * @gtt_offset: The gtt offset the vma is bound to. + * @gtt_size: The size in bytes allocated for the vma in the GTT. + * @pages: The struct sg_table pointing to the pages bound. + * @pages_rsgt: The refcounted sg_table holding the reference for @pages if any. + * @mr: The memory region pointed for the pages bound. + * @kref: Reference for this structure. + * @vma_resource: FIXME: A means to keep the unbind fence from signaling. + * Temporarily while we have only sync unbinds, and still use the vma + * active, we use that. With async unbinding we need a signaling refcount + * for the unbind fence. + * @page_sizes: The vma GTT page sizes information. + * @onstack: Whether the structure shouldn't be freed on final put. + * @present: Whether the structure is present and initialized. + */ +struct i915_vma_snapshot { + const char *name; + size_t size; + size_t obj_size; + size_t gtt_offset; + size_t gtt_size; + struct sg_table *pages; + struct i915_refct_sgt *pages_rsgt; + struct intel_memory_region *mr; + struct kref kref; + struct i915_active *vma_resource; + u32 page_sizes; + bool onstack:1; + bool present:1; +}; + +void i915_vma_snapshot_init(struct i915_vma_snapshot *vsnap, + struct i915_vma *vma, + const char *name); + +void i915_vma_snapshot_init_onstack(struct i915_vma_snapshot *vsnap, + struct i915_vma *vma, + const char *name); + +void i915_vma_snapshot_put(struct i915_vma_snapshot *vsnap); + +void i915_vma_snapshot_put_onstack(struct i915_vma_snapshot *vsnap); + +bool i915_vma_snapshot_resource_pin(struct i915_vma_snapshot *vsnap, + bool *lockdep_cookie); + +void i915_vma_snapshot_resource_unpin(struct i915_vma_snapshot *vsnap, + bool lockdep_cookie); + +/** + * i915_vma_snapshot_alloc - Allocate a struct i915_vma_snapshot + * @gfp: Allocation mode. + * + * Return: A pointer to a struct i915_vma_snapshot if successful. + * NULL otherwise. + */ +static inline struct i915_vma_snapshot *i915_vma_snapshot_alloc(gfp_t gfp) +{ + return kmalloc(sizeof(struct i915_vma_snapshot), gfp); +} + +/** + * i915_vma_snapshot_get - Take a reference on a struct i915_vma_snapshot + * + * Return: A pointer to a struct i915_vma_snapshot. + */ +static inline struct i915_vma_snapshot * +i915_vma_snapshot_get(struct i915_vma_snapshot *vsnap) +{ + kref_get(&vsnap->kref); + return vsnap; +} + +/** + * i915_vma_snapshot_present - Whether a struct i915_vma_snapshot is + * present and initialized. + * + * Return: true if present and initialized; false otherwise. + */ +static inline bool +i915_vma_snapshot_present(const struct i915_vma_snapshot *vsnap) +{ + return vsnap && vsnap->present; +} + +#endif From patchwork Wed Oct 27 10:52:11 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: =?utf-8?q?Thomas_Hellstr=C3=B6m?= X-Patchwork-Id: 12587027 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 4BD7CC433EF for ; Wed, 27 Oct 2021 10:53:09 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 147B26103B for ; Wed, 27 Oct 2021 10:53:09 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.1 mail.kernel.org 147B26103B Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linux.intel.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id B4B1789C61; Wed, 27 Oct 2021 10:52:56 +0000 (UTC) Received: from mga18.intel.com (mga18.intel.com [134.134.136.126]) by gabe.freedesktop.org (Postfix) with ESMTPS id 8413789AAE; Wed, 27 Oct 2021 10:52:55 +0000 (UTC) X-IronPort-AV: E=McAfee;i="6200,9189,10149"; a="217043364" X-IronPort-AV: E=Sophos;i="5.87,186,1631602800"; d="scan'208";a="217043364" Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by orsmga106.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 27 Oct 2021 03:52:33 -0700 X-IronPort-AV: E=Sophos;i="5.87,186,1631602800"; d="scan'208";a="572869999" Received: from djustese-mobl.ger.corp.intel.com (HELO thellstr-mobl1.intel.com) ([10.249.254.205]) by fmsmga003-auth.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 27 Oct 2021 03:52:31 -0700 From: =?utf-8?q?Thomas_Hellstr=C3=B6m?= To: intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org Cc: maarten.lankhorst@linux.intel.com, matthew.auld@intel.com, =?utf-8?q?Tho?= =?utf-8?q?mas_Hellstr=C3=B6m?= Subject: [PATCH v2 3/3] drm/i915: Initial introduction of vma resources Date: Wed, 27 Oct 2021 12:52:11 +0200 Message-Id: <20211027105211.485125-4-thomas.hellstrom@linux.intel.com> X-Mailer: git-send-email 2.31.1 In-Reply-To: <20211027105211.485125-1-thomas.hellstrom@linux.intel.com> References: <20211027105211.485125-1-thomas.hellstrom@linux.intel.com> MIME-Version: 1.0 X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" From: Thomas Hellström The vma resource are needed for asynchronous bind management and are similar to TTM resources. They contain the data needed for asynchronous unbinding (typically the vm range, any backend private information and a means to do refcounting and to hold the unbinding for error capture). When a vma is bound, a vma resource is created and attached to the vma, and on async unbinding it is detached from the vma, and instead the vm records the fence marking unbind complete. This fence needs to be waited on before we can bind the same region again, so either the fence can be recorded for this particular range only, using an interval tree, or as a simpler approach, for the whole vm. The latter means no binding can take place on a vm until all detached vma resources scheduled for unbind are signaled. With an interval tree fence recording, the interval tree needs to be searched for fences to be signaled before binding can take place. But most of that is for later, this patch only introduces stub vma resources without unbind capability and the fences of which are waited for sync during unbinding. At this point we're interested in the hold capability as a POC for error capture. Note that the current sync waiting at unbind time is done uninterruptible, but that's OK since we're only ever waiting during error capture, and in that case there's very little gpu activity (if any) that can stall. v2: - Fix the mock gtt selftest to bind with vma resources. - Update a code comment. - Account for rebinding the same vma with different I915_VMA_*_BIND flags Signed-off-by: Thomas Hellström Reported-by: kernel test robot --- .../gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/i915_vma.c | 201 +++++++++++++++++- drivers/gpu/drm/i915/i915_vma.h | 20 +- drivers/gpu/drm/i915/i915_vma_snapshot.c | 14 +- drivers/gpu/drm/i915/i915_vma_snapshot.h | 2 +- drivers/gpu/drm/i915/i915_vma_types.h | 5 + drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 99 +++++---- 7 files changed, 284 insertions(+), 59 deletions(-) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index 301eb58bebd1..69915c00ce18 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -1376,7 +1376,7 @@ eb_relocate_entry(struct i915_execbuffer *eb, GRAPHICS_VER(eb->i915) == 6) { err = i915_vma_bind(target->vma, target->vma->obj->cache_level, - PIN_GLOBAL, NULL); + PIN_GLOBAL, NULL, NULL); if (err) return err; } diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c index 90546fa58fc1..9a8fcf89f7d8 100644 --- a/drivers/gpu/drm/i915/i915_vma.c +++ b/drivers/gpu/drm/i915/i915_vma.c @@ -38,8 +38,33 @@ #include "i915_trace.h" #include "i915_vma.h" +/** + * struct i915_vma_resource - Snapshotted unbind information. + * @unbind_fence: Fence to mark unbinding complete. Note that this fence + * is not considered published until unbind is scheduled, and as such it + * is illegal to access this fence before scheduled unbind other than + * for refcounting. + * @lock: The @unbind_fence lock. We're also using it to protect the weak + * pointer to the struct i915_vma, @vma during lookup and takedown. + * @vma: Weak back-pointer to the parent vma struct. This pointer is + * protected by @lock, and a reference on @vma needs to be taken + * using kref_get_unless_zero. + * @hold_count: Number of holders blocking the fence from finishing. + * The vma itself is keeping a hold, which is released when unbind + * is scheduled. + */ +struct i915_vma_resource { + struct dma_fence unbind_fence; + /* See above for description of the lock. */ + spinlock_t lock; + struct i915_vma *vma; + refcount_t hold_count; +}; + static struct kmem_cache *slab_vmas; +static struct dma_fence *i915_vma_resource_unbind(struct i915_vma_resource *vma_res); + struct i915_vma *i915_vma_alloc(void) { return kmem_cache_zalloc(slab_vmas, GFP_KERNEL); @@ -363,6 +388,8 @@ int i915_vma_wait_for_bind(struct i915_vma *vma) * @cache_level: mapping cache level * @flags: flags like global or local mapping * @work: preallocated worker for allocating and binding the PTE + * @vma_res: pointer to a preallocated vma resource. The resource is either + * consumed or freed. * * DMA addresses are taken from the scatter-gather table of this object (or of * this VMA in case of non-default GGTT views) and PTE entries set up. @@ -371,7 +398,8 @@ int i915_vma_wait_for_bind(struct i915_vma *vma) int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level, u32 flags, - struct i915_vma_work *work) + struct i915_vma_work *work, + struct i915_vma_resource *vma_res) { u32 bind_flags; u32 vma_flags; @@ -381,11 +409,15 @@ int i915_vma_bind(struct i915_vma *vma, if (GEM_DEBUG_WARN_ON(range_overflows(vma->node.start, vma->node.size, - vma->vm->total))) + vma->vm->total))) { + kfree(vma_res); return -ENODEV; + } - if (GEM_DEBUG_WARN_ON(!flags)) + if (GEM_DEBUG_WARN_ON(!flags)) { + kfree(vma_res); return -EINVAL; + } bind_flags = flags; bind_flags &= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND; @@ -394,11 +426,25 @@ int i915_vma_bind(struct i915_vma *vma, vma_flags &= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND; bind_flags &= ~vma_flags; - if (bind_flags == 0) + if (bind_flags == 0) { + kfree(vma_res); return 0; + } GEM_BUG_ON(!vma->pages); + if (!i915_vma_is_pinned(vma)) + lockdep_assert_held(&vma->vm->mutex); + + if ((vma->resource) || !vma_res) { + /* Rebinding with an additional I915_VMA_*_BIND */ + GEM_WARN_ON(!vma_flags); + kfree(vma_res); + } else { + lockdep_assert_held(&vma->vm->mutex); + i915_vma_resource_init(vma_res, vma); + vma->resource = vma_res; + } trace_i915_vma_bind(vma, bind_flags); if (work && bind_flags & vma->vm->bind_async_flags) { struct dma_fence *prev; @@ -870,6 +916,7 @@ int i915_vma_pin_ww(struct i915_vma *vma, struct i915_gem_ww_ctx *ww, u64 size, u64 alignment, u64 flags) { struct i915_vma_work *work = NULL; + struct i915_vma_resource *vma_res; intel_wakeref_t wakeref = 0; unsigned int bound; int err; @@ -923,6 +970,12 @@ int i915_vma_pin_ww(struct i915_vma *vma, struct i915_gem_ww_ctx *ww, } } + vma_res = i915_vma_resource_alloc(); + if (IS_ERR(vma_res)) { + err = PTR_ERR(vma_res); + goto err_fence; + } + /* * Differentiate between user/kernel vma inside the aliasing-ppgtt. * @@ -984,7 +1037,7 @@ int i915_vma_pin_ww(struct i915_vma *vma, struct i915_gem_ww_ctx *ww, GEM_BUG_ON(!vma->pages); err = i915_vma_bind(vma, vma->obj ? vma->obj->cache_level : 0, - flags, work); + flags, work, vma_res); if (err) goto err_remove; @@ -1014,6 +1067,7 @@ int i915_vma_pin_ww(struct i915_vma *vma, struct i915_gem_ww_ctx *ww, if (wakeref) intel_runtime_pm_put(&vma->vm->i915->runtime_pm, wakeref); vma_put_pages(vma); + return err; } @@ -1331,6 +1385,7 @@ void __i915_vma_evict(struct i915_vma *vma) int __i915_vma_unbind(struct i915_vma *vma) { + struct dma_fence *unbind_fence; int ret; lockdep_assert_held(&vma->vm->mutex); @@ -1356,6 +1411,16 @@ int __i915_vma_unbind(struct i915_vma *vma) __i915_vma_evict(vma); drm_mm_remove_node(&vma->node); /* pairs with i915_vma_release() */ + unbind_fence = i915_vma_resource_unbind(vma->resource); + + /* + * This uninterruptible wait under the vm mutex is currently + * only ever blocking while the vma is being captured from. + * With async unbinding, this wait here will be removed. + */ + dma_fence_wait(unbind_fence, false); + dma_fence_put(unbind_fence); + return 0; } @@ -1388,7 +1453,6 @@ int i915_vma_unbind(struct i915_vma *vma) err = __i915_vma_unbind(vma); mutex_unlock(&vm->mutex); - out_rpm: if (wakeref) intel_runtime_pm_put(&vm->i915->runtime_pm, wakeref); @@ -1411,6 +1475,131 @@ void i915_vma_make_purgeable(struct i915_vma *vma) i915_gem_object_make_purgeable(vma->obj); } +static const char *get_driver_name(struct dma_fence *fence) +{ + return "vma unbind fence"; +} + +static const char *get_timeline_name(struct dma_fence *fence) +{ + return "unbound"; +} + +static struct dma_fence_ops unbind_fence_ops = { + .get_driver_name = get_driver_name, + .get_timeline_name = get_timeline_name, +}; + +struct i915_vma_resource *i915_vma_resource_alloc(void) +{ + struct i915_vma_resource *vma_res = + kzalloc(sizeof(*vma_res), GFP_KERNEL); + + return vma_res ? vma_res : ERR_PTR(-ENOMEM); +} + +I915_SELFTEST_EXPORT void +i915_vma_resource_init(struct i915_vma_resource *vma_res, + struct i915_vma *vma) +{ + vma_res->vma = vma; + spin_lock_init(&vma_res->lock); + dma_fence_init(&vma_res->unbind_fence, &unbind_fence_ops, + &vma_res->lock, 0, 0); + refcount_set(&vma_res->hold_count, 1); +} + +static void __i915_vma_resource_unhold(struct i915_vma_resource *vma_res) +{ + if (refcount_dec_and_test(&vma_res->hold_count)) + dma_fence_signal(&vma_res->unbind_fence); +} + +/** + * i915_vma_resource_unhold - Unhold the signaling of the vma resource unbind + * fence. + * @vma_res: The vma resource. + * @lockdep_cookie: The lockdep cookie returned from i915_vma_resource_hold. + * + * The function may leave a dma_fence critical section. + */ +void i915_vma_resource_unhold(struct i915_vma_resource *vma_res, + bool lockdep_cookie) +{ + dma_fence_end_signalling(lockdep_cookie); + + if (IS_ENABLED(CONFIG_PROVE_LOCKING)) { + unsigned long irq_flags; + + /* Inefficient open-coded might_lock_irqsave() */ + spin_lock_irqsave(&vma_res->lock, irq_flags); + spin_unlock_irqrestore(&vma_res->lock, irq_flags); + } + + __i915_vma_resource_unhold(vma_res); +} + +/** + * i915_vma_resource_hold - Hold the signaling of the vma resource unbind fence. + * @vma_res: The vma resource. + * @lockdep_cookie: Pointer to a bool serving as a lockdep cooke that should + * be given as an argument to the pairing i915_vma_resource_unhold. + * + * If returning true, the function enters a dma_fence signalling critical + * section is not in one already. + * + * Return: true if holding successful, false if not. + */ +bool i915_vma_resource_hold(struct i915_vma_resource *vma_res, + bool *lockdep_cookie) +{ + bool held = refcount_inc_not_zero(&vma_res->hold_count); + + if (held) + *lockdep_cookie = dma_fence_begin_signalling(); + + return held; +} + +/** + * i915_vma_get_current_resource - Return the vma's current vma resource + * @vma: The vma referencing the resource. + * + * Return: A refcounted pointer to the vma's current vma resource. + */ +struct i915_vma_resource *i915_vma_get_current_resource(struct i915_vma *vma) +{ + GEM_BUG_ON(!vma->resource); + + dma_fence_get(&vma->resource->unbind_fence); + return vma->resource; +} + +/** + * i915_vma_resource_put - Release a reference to a struct i915_vma_resource + * @vma_res: The resource + */ +void i915_vma_resource_put(struct i915_vma_resource *vma_res) +{ + dma_fence_put(&vma_res->unbind_fence); +} + +static struct dma_fence * +i915_vma_resource_unbind(struct i915_vma_resource *vma_res) +{ + /* Reference is transferred to the returned dma_fence pointer */ + vma_res->vma->resource = NULL; + + spin_lock(&vma_res->lock); + /* Kill the weak reference under the spinlock. */ + vma_res->vma = NULL; + spin_unlock(&vma_res->lock); + + /* With async unbind, schedule it here. */ + __i915_vma_resource_unhold(vma_res); + return &vma_res->unbind_fence; +} + #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) #include "selftests/i915_vma.c" #endif diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h index 648dbe744c96..aa13d0d5bb91 100644 --- a/drivers/gpu/drm/i915/i915_vma.h +++ b/drivers/gpu/drm/i915/i915_vma.h @@ -206,7 +206,8 @@ struct i915_vma_work *i915_vma_work(void); int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level, u32 flags, - struct i915_vma_work *work); + struct i915_vma_work *work, + struct i915_vma_resource *vma_res); bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long color); bool i915_vma_misplaced(const struct i915_vma *vma, @@ -433,7 +434,24 @@ static inline int i915_vma_sync(struct i915_vma *vma) return i915_active_wait(&vma->active); } +bool i915_vma_resource_hold(struct i915_vma_resource *vma_res, + bool *lockdep_cookie); + +void i915_vma_resource_unhold(struct i915_vma_resource *vma_res, + bool lockdep_cookie); + +void i915_vma_resource_put(struct i915_vma_resource *vma_res); + +struct i915_vma_resource *i915_vma_get_current_resource(struct i915_vma *vma); + +struct i915_vma_resource *i915_vma_resource_alloc(void); + void i915_vma_module_exit(void); int i915_vma_module_init(void); +#ifdef CONFIG_DRM_I915_SELFTEST +void i915_vma_resource_init(struct i915_vma_resource *vma_res, + struct i915_vma *vma); +#endif + #endif diff --git a/drivers/gpu/drm/i915/i915_vma_snapshot.c b/drivers/gpu/drm/i915/i915_vma_snapshot.c index 44985d600f96..b4ee8220df85 100644 --- a/drivers/gpu/drm/i915/i915_vma_snapshot.c +++ b/drivers/gpu/drm/i915/i915_vma_snapshot.c @@ -36,7 +36,7 @@ void i915_vma_snapshot_init(struct i915_vma_snapshot *vsnap, if (vma->obj->mm.region) vsnap->mr = intel_memory_region_get(vma->obj->mm.region); kref_init(&vsnap->kref); - vsnap->vma_resource = &vma->active; + vsnap->vma_resource = i915_vma_get_current_resource(vma); vsnap->onstack = false; vsnap->present = true; } @@ -63,6 +63,7 @@ static void vma_snapshot_release(struct kref *ref) container_of(ref, typeof(*vsnap), kref); vsnap->present = false; + i915_vma_resource_put(vsnap->vma_resource); if (vsnap->mr) intel_memory_region_put(vsnap->mr); if (vsnap->pages_rsgt) @@ -112,12 +113,7 @@ void i915_vma_snapshot_put_onstack(struct i915_vma_snapshot *vsnap) bool i915_vma_snapshot_resource_pin(struct i915_vma_snapshot *vsnap, bool *lockdep_cookie) { - bool pinned = i915_active_acquire_if_busy(vsnap->vma_resource); - - if (pinned) - *lockdep_cookie = dma_fence_begin_signalling(); - - return pinned; + return i915_vma_resource_hold(vsnap->vma_resource, lockdep_cookie); } /** @@ -131,7 +127,5 @@ bool i915_vma_snapshot_resource_pin(struct i915_vma_snapshot *vsnap, void i915_vma_snapshot_resource_unpin(struct i915_vma_snapshot *vsnap, bool lockdep_cookie) { - dma_fence_end_signalling(lockdep_cookie); - - return i915_active_release(vsnap->vma_resource); + i915_vma_resource_unhold(vsnap->vma_resource, lockdep_cookie); } diff --git a/drivers/gpu/drm/i915/i915_vma_snapshot.h b/drivers/gpu/drm/i915/i915_vma_snapshot.h index 940581df4622..d083b6bf1b11 100644 --- a/drivers/gpu/drm/i915/i915_vma_snapshot.h +++ b/drivers/gpu/drm/i915/i915_vma_snapshot.h @@ -49,7 +49,7 @@ struct i915_vma_snapshot { struct i915_refct_sgt *pages_rsgt; struct intel_memory_region *mr; struct kref kref; - struct i915_active *vma_resource; + struct i915_vma_resource *vma_resource; u32 page_sizes; bool onstack:1; bool present:1; diff --git a/drivers/gpu/drm/i915/i915_vma_types.h b/drivers/gpu/drm/i915/i915_vma_types.h index 80e93bf00f2e..14d20ac5350c 100644 --- a/drivers/gpu/drm/i915/i915_vma_types.h +++ b/drivers/gpu/drm/i915/i915_vma_types.h @@ -95,6 +95,8 @@ enum i915_cache_level; * */ +struct i915_vma_resource; + struct intel_remapped_plane_info { /* in gtt pages */ u32 offset; @@ -284,6 +286,9 @@ struct i915_vma { struct list_head evict_link; struct list_head closed_link; + + /** The async vma resource. Protected by the vm_mutex */ + struct i915_vma_resource *resource; }; #endif diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c index 46f4236039a9..c76fb85dd3d9 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c @@ -1336,6 +1336,33 @@ static int igt_mock_drunk(void *arg) return exercise_mock(ggtt->vm.i915, drunk_hole); } +static int reserve_gtt_with_resource(struct i915_vma *vma, u64 offset) +{ + struct i915_address_space *vm = vma->vm; + struct i915_vma_resource *vma_res; + struct drm_i915_gem_object *obj = vma->obj; + int err; + + vma_res = i915_vma_resource_alloc(); + if (IS_ERR(vma_res)) + return PTR_ERR(vma_res); + + mutex_lock(&vm->mutex); + err = i915_gem_gtt_reserve(vm, &vma->node, obj->base.size, + offset, + obj->cache_level, + 0); + if (!err) { + i915_vma_resource_init(vma_res, vma); + vma->resource = vma_res; + } else { + kfree(vma_res); + } + mutex_unlock(&vm->mutex); + + return err; +} + static int igt_gtt_reserve(void *arg) { struct i915_ggtt *ggtt = arg; @@ -1370,20 +1397,13 @@ static int igt_gtt_reserve(void *arg) } list_add(&obj->st_link, &objects); - vma = i915_vma_instance(obj, &ggtt->vm, NULL); if (IS_ERR(vma)) { err = PTR_ERR(vma); goto out; } - mutex_lock(&ggtt->vm.mutex); - err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node, - obj->base.size, - total, - obj->cache_level, - 0); - mutex_unlock(&ggtt->vm.mutex); + err = reserve_gtt_with_resource(vma, total); if (err) { pr_err("i915_gem_gtt_reserve (pass 1) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1429,13 +1449,7 @@ static int igt_gtt_reserve(void *arg) goto out; } - mutex_lock(&ggtt->vm.mutex); - err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node, - obj->base.size, - total, - obj->cache_level, - 0); - mutex_unlock(&ggtt->vm.mutex); + err = reserve_gtt_with_resource(vma, total); if (err) { pr_err("i915_gem_gtt_reserve (pass 2) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1476,13 +1490,8 @@ static int igt_gtt_reserve(void *arg) 2 * I915_GTT_PAGE_SIZE, I915_GTT_MIN_ALIGNMENT); - mutex_lock(&ggtt->vm.mutex); - err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node, - obj->base.size, - offset, - obj->cache_level, - 0); - mutex_unlock(&ggtt->vm.mutex); + + err = reserve_gtt_with_resource(vma, offset); if (err) { pr_err("i915_gem_gtt_reserve (pass 3) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1509,6 +1518,31 @@ static int igt_gtt_reserve(void *arg) return err; } +static int insert_gtt_with_resource(struct i915_vma *vma) +{ + struct i915_address_space *vm = vma->vm; + struct i915_vma_resource *vma_res; + struct drm_i915_gem_object *obj = vma->obj; + int err; + + vma_res = i915_vma_resource_alloc(); + if (IS_ERR(vma_res)) + return PTR_ERR(vma_res); + + mutex_lock(&vm->mutex); + err = i915_gem_gtt_insert(vm, &vma->node, obj->base.size, 0, + obj->cache_level, 0, vm->total, 0); + if (!err) { + i915_vma_resource_init(vma_res, vma); + vma->resource = vma_res; + } else { + kfree(vma_res); + } + mutex_unlock(&vm->mutex); + + return err; +} + static int igt_gtt_insert(void *arg) { struct i915_ggtt *ggtt = arg; @@ -1593,12 +1627,7 @@ static int igt_gtt_insert(void *arg) goto out; } - mutex_lock(&ggtt->vm.mutex); - err = i915_gem_gtt_insert(&ggtt->vm, &vma->node, - obj->base.size, 0, obj->cache_level, - 0, ggtt->vm.total, - 0); - mutex_unlock(&ggtt->vm.mutex); + err = insert_gtt_with_resource(vma); if (err == -ENOSPC) { /* maxed out the GGTT space */ i915_gem_object_put(obj); @@ -1653,12 +1682,7 @@ static int igt_gtt_insert(void *arg) goto out; } - mutex_lock(&ggtt->vm.mutex); - err = i915_gem_gtt_insert(&ggtt->vm, &vma->node, - obj->base.size, 0, obj->cache_level, - 0, ggtt->vm.total, - 0); - mutex_unlock(&ggtt->vm.mutex); + err = insert_gtt_with_resource(vma); if (err) { pr_err("i915_gem_gtt_insert (pass 2) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err); @@ -1702,12 +1726,7 @@ static int igt_gtt_insert(void *arg) goto out; } - mutex_lock(&ggtt->vm.mutex); - err = i915_gem_gtt_insert(&ggtt->vm, &vma->node, - obj->base.size, 0, obj->cache_level, - 0, ggtt->vm.total, - 0); - mutex_unlock(&ggtt->vm.mutex); + err = insert_gtt_with_resource(vma); if (err) { pr_err("i915_gem_gtt_insert (pass 3) failed at %llu/%llu with err=%d\n", total, ggtt->vm.total, err);