From patchwork Tue Jun 29 11:37:11 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: 12349719 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-16.8 required=3.0 tests=BAYES_00, HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER,INCLUDES_PATCH, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 96610C11F66 for ; Tue, 29 Jun 2021 11:37:39 +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 2668661DC2 for ; Tue, 29 Jun 2021 11:37:39 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 2668661DC2 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=dri-devel-bounces@lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 2F6936E865; Tue, 29 Jun 2021 11:37:34 +0000 (UTC) Received: from mga06.intel.com (mga06.intel.com [134.134.136.31]) by gabe.freedesktop.org (Postfix) with ESMTPS id B74636E866; Tue, 29 Jun 2021 11:37:32 +0000 (UTC) X-IronPort-AV: E=McAfee;i="6200,9189,10029"; a="269259976" X-IronPort-AV: E=Sophos;i="5.83,308,1616482800"; d="scan'208";a="269259976" Received: from orsmga005.jf.intel.com ([10.7.209.41]) by orsmga104.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 29 Jun 2021 04:37:32 -0700 X-IronPort-AV: E=Sophos;i="5.83,308,1616482800"; d="scan'208";a="625600016" Received: from ettammin-mobl1.ger.corp.intel.com (HELO thellst-mobl1.intel.com) ([10.249.254.141]) by orsmga005-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 29 Jun 2021 04:37:30 -0700 From: =?utf-8?q?Thomas_Hellstr=C3=B6m?= To: intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org Subject: [PATCH v4 1/3] drm/i915/gem: Implement object migration Date: Tue, 29 Jun 2021 13:37:11 +0200 Message-Id: <20210629113713.154654-2-thomas.hellstrom@linux.intel.com> X-Mailer: git-send-email 2.31.1 In-Reply-To: <20210629113713.154654-1-thomas.hellstrom@linux.intel.com> References: <20210629113713.154654-1-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: , Cc: =?utf-8?q?Thomas_Hellstr=C3=B6m?= , "Michael J . Ruhl" , matthew.auld@intel.com, kernel test robot Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" Introduce an interface to migrate objects between regions. This is primarily intended to migrate objects to LMEM for display and to SYSTEM for dma-buf, but might be reused in one form or another for performance-based migration. v2: - Verify that the memory region given as an id really exists. (Reported by Matthew Auld) - Call i915_gem_object_{init,release}_memory_region() when switching region to handle also switching region lists. (Reported by Matthew Auld) v3: - Fix i915_gem_object_can_migrate() to return true if object is already in the correct region, even if the object ops doesn't have a migrate() callback. - Update typo in commit message. - Fix kerneldoc of i915_gem_object_wait_migration(). v4: - Improve documentation (Suggested by Mattew Auld and Michael Ruhl) - Always assume TTM migration hits a TTM move and unsets the pages through move_notify. (Reported by Matthew Auld) - Add a dma_fence_might_wait() annotation to i915_gem_object_wait_migration() (Suggested by Daniel Vetter) Reported-by: kernel test robot Signed-off-by: Thomas Hellström Reviewed-by: Michael J. Ruhl Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/gem/i915_gem_object.c | 112 ++++++++++++++++++ drivers/gpu/drm/i915/gem/i915_gem_object.h | 12 ++ .../gpu/drm/i915/gem/i915_gem_object_types.h | 9 ++ drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 77 ++++++++++-- drivers/gpu/drm/i915/gem/i915_gem_wait.c | 19 +++ 5 files changed, 217 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c index 07e8ff9a8aae..225b77fb4314 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c @@ -513,6 +513,118 @@ bool i915_gem_object_has_iomem(const struct drm_i915_gem_object *obj) return obj->mem_flags & I915_BO_FLAG_IOMEM; } +/** + * i915_gem_object_can_migrate - Whether an object likely can be migrated + * + * @obj: The object to migrate + * @id: The region intended to migrate to + * + * Check whether the object backend supports migration to the + * given region. Note that pinning may affect the ability to migrate as + * returned by this function. + * + * This function is primarily intended as a helper for checking the + * possibility to migrate objects and might be slightly less permissive + * than i915_gem_object_migrate() when it comes to objects with the + * I915_BO_ALLOC_USER flag set. + * + * Return: true if migration is possible, false otherwise. + */ +bool i915_gem_object_can_migrate(struct drm_i915_gem_object *obj, + enum intel_region_id id) +{ + struct drm_i915_private *i915 = to_i915(obj->base.dev); + unsigned int num_allowed = obj->mm.n_placements; + struct intel_memory_region *mr; + unsigned int i; + + GEM_BUG_ON(id >= INTEL_REGION_UNKNOWN); + GEM_BUG_ON(obj->mm.madv != I915_MADV_WILLNEED); + + mr = i915->mm.regions[id]; + if (!mr) + return false; + + if (obj->mm.region == mr) + return true; + + if (!i915_gem_object_evictable(obj)) + return false; + + if (!obj->ops->migrate) + return false; + + if (!(obj->flags & I915_BO_ALLOC_USER)) + return true; + + if (num_allowed == 0) + return false; + + for (i = 0; i < num_allowed; ++i) { + if (mr == obj->mm.placements[i]) + return true; + } + + return false; +} + +/** + * i915_gem_object_migrate - Migrate an object to the desired region id + * @obj: The object to migrate. + * @ww: An optional struct i915_gem_ww_ctx. If NULL, the backend may + * not be successful in evicting other objects to make room for this object. + * @id: The region id to migrate to. + * + * Attempt to migrate the object to the desired memory region. The + * object backend must support migration and the object may not be + * pinned, (explicitly pinned pages or pinned vmas). The object must + * be locked. + * On successful completion, the object will have pages pointing to + * memory in the new region, but an async migration task may not have + * completed yet, and to accomplish that, i915_gem_object_wait_migration() + * must be called. + * + * This function is a bit more permissive than i915_gem_object_can_migrate() + * to allow for migrating objects where the caller knows exactly what is + * happening. For example within selftests. More specifically this + * function allows migrating I915_BO_ALLOC_USER objects to regions + * that are not in the list of allowable regions. + * + * Note: the @ww parameter is not used yet, but included to make sure + * callers put some effort into obtaining a valid ww ctx if one is + * available. + * + * Return: 0 on success. Negative error code on failure. In particular may + * return -ENXIO on lack of region space, -EDEADLK for deadlock avoidance + * if @ww is set, -EINTR or -ERESTARTSYS if signal pending, and + * -EBUSY if the object is pinned. + */ +int i915_gem_object_migrate(struct drm_i915_gem_object *obj, + struct i915_gem_ww_ctx *ww, + enum intel_region_id id) +{ + struct drm_i915_private *i915 = to_i915(obj->base.dev); + struct intel_memory_region *mr; + + GEM_BUG_ON(id >= INTEL_REGION_UNKNOWN); + GEM_BUG_ON(obj->mm.madv != I915_MADV_WILLNEED); + assert_object_held(obj); + + mr = i915->mm.regions[id]; + GEM_BUG_ON(!mr); + + if (obj->mm.region == mr) + return 0; + + if (!i915_gem_object_evictable(obj)) + return -EBUSY; + + if (!obj->ops->migrate) + return -EOPNOTSUPP; + + return obj->ops->migrate(obj, mr); +} + void i915_gem_init__objects(struct drm_i915_private *i915) { INIT_WORK(&i915->mm.free_work, __i915_gem_free_work); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h index ea3224a480c4..8cbd7a5334e2 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h @@ -17,6 +17,8 @@ #include "i915_gem_ww.h" #include "i915_vma_types.h" +enum intel_region_id; + /* * XXX: There is a prevalence of the assumption that we fit the * object's page count inside a 32bit _signed_ variable. Let's document @@ -597,6 +599,16 @@ bool i915_gem_object_migratable(struct drm_i915_gem_object *obj); bool i915_gem_object_validates_to_lmem(struct drm_i915_gem_object *obj); +int i915_gem_object_migrate(struct drm_i915_gem_object *obj, + struct i915_gem_ww_ctx *ww, + enum intel_region_id id); + +bool i915_gem_object_can_migrate(struct drm_i915_gem_object *obj, + enum intel_region_id id); + +int i915_gem_object_wait_migration(struct drm_i915_gem_object *obj, + unsigned int flags); + #ifdef CONFIG_MMU_NOTIFIER static inline bool i915_gem_object_is_userptr(struct drm_i915_gem_object *obj) 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 441f913c87e6..ef3de2ae9723 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h @@ -18,6 +18,7 @@ struct drm_i915_gem_object; struct intel_fronbuffer; +struct intel_memory_region; /* * struct i915_lut_handle tracks the fast lookups from handle to vma used @@ -77,6 +78,14 @@ struct drm_i915_gem_object_ops { * delayed_free - Override the default delayed free implementation */ void (*delayed_free)(struct drm_i915_gem_object *obj); + + /** + * migrate - Migrate object to a different region either for + * pinning or for as long as the object lock is held. + */ + int (*migrate)(struct drm_i915_gem_object *obj, + struct intel_memory_region *mr); + void (*release)(struct drm_i915_gem_object *obj); const struct vm_operations_struct *mmap_ops; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c index c39d982c4fa6..521ab740001a 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c @@ -617,7 +617,8 @@ struct ttm_device_funcs *i915_ttm_driver(void) return &i915_ttm_bo_driver; } -static int i915_ttm_get_pages(struct drm_i915_gem_object *obj) +static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj, + struct ttm_placement *placement) { struct ttm_buffer_object *bo = i915_gem_to_ttm(obj); struct ttm_operation_ctx ctx = { @@ -625,19 +626,12 @@ static int i915_ttm_get_pages(struct drm_i915_gem_object *obj) .no_wait_gpu = false, }; struct sg_table *st; - struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS]; - struct ttm_placement placement; int real_num_busy; int ret; - GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS); - - /* Move to the requested placement. */ - i915_ttm_placement_from_obj(obj, &requested, busy, &placement); - /* First try only the requested placement. No eviction. */ - real_num_busy = fetch_and_zero(&placement.num_busy_placement); - ret = ttm_bo_validate(bo, &placement, &ctx); + real_num_busy = fetch_and_zero(&placement->num_busy_placement); + ret = ttm_bo_validate(bo, placement, &ctx); if (ret) { ret = i915_ttm_err_to_gem(ret); /* @@ -652,8 +646,8 @@ static int i915_ttm_get_pages(struct drm_i915_gem_object *obj) * If the initial attempt fails, allow all accepted placements, * evicting if necessary. */ - placement.num_busy_placement = real_num_busy; - ret = ttm_bo_validate(bo, &placement, &ctx); + placement->num_busy_placement = real_num_busy; + ret = ttm_bo_validate(bo, placement, &ctx); if (ret) return i915_ttm_err_to_gem(ret); } @@ -668,6 +662,7 @@ static int i915_ttm_get_pages(struct drm_i915_gem_object *obj) i915_ttm_adjust_gem_after_move(obj); } + GEM_WARN_ON(obj->mm.pages); /* 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)) @@ -678,6 +673,63 @@ static int i915_ttm_get_pages(struct drm_i915_gem_object *obj) return ret; } +static int i915_ttm_get_pages(struct drm_i915_gem_object *obj) +{ + struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS]; + struct ttm_placement placement; + + GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS); + + /* Move to the requested placement. */ + i915_ttm_placement_from_obj(obj, &requested, busy, &placement); + + return __i915_ttm_get_pages(obj, &placement); +} + +/** + * DOC: Migration vs eviction + * + * GEM migration may not be the same as TTM migration / eviction. If + * the TTM core decides to evict an object it may be evicted to a + * TTM memory type that is not in the object's allowable GEM regions, or + * in fact theoretically to a TTM memory type that doesn't correspond to + * a GEM memory region. In that case the object's GEM region is not + * updated, and the data is migrated back to the GEM region at + * get_pages time. TTM may however set up CPU ptes to the object even + * when it is evicted. + * Gem forced migration using the i915_ttm_migrate() op, is allowed even + * to regions that are not in the object's list of allowable placements. + */ +static int i915_ttm_migrate(struct drm_i915_gem_object *obj, + struct intel_memory_region *mr) +{ + struct ttm_place requested; + struct ttm_placement placement; + int ret; + + i915_ttm_place_from_region(mr, &requested, obj->flags); + placement.num_placement = 1; + placement.num_busy_placement = 1; + placement.placement = &requested; + placement.busy_placement = &requested; + + ret = __i915_ttm_get_pages(obj, &placement); + if (ret) + return ret; + + /* + * Reinitialize the region bindings. This is primarily + * required for objects where the new region is not in + * its allowable placements. + */ + if (obj->mm.region != mr) { + i915_gem_object_release_memory_region(obj); + i915_gem_object_init_memory_region(obj, mr); + } + + return 0; +} + static void i915_ttm_put_pages(struct drm_i915_gem_object *obj, struct sg_table *st) { @@ -814,6 +866,7 @@ static const struct drm_i915_gem_object_ops i915_gem_ttm_obj_ops = { .truncate = i915_ttm_purge, .adjust_lru = i915_ttm_adjust_lru, .delayed_free = i915_ttm_delayed_free, + .migrate = i915_ttm_migrate, .mmap_offset = i915_ttm_mmap_offset, .mmap_ops = &vm_ops_ttm, }; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_wait.c b/drivers/gpu/drm/i915/gem/i915_gem_wait.c index 1070d3afdce7..190e221eaf81 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_wait.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_wait.c @@ -290,3 +290,22 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file) i915_gem_object_put(obj); return ret; } + +/** + * i915_gem_object_wait_migration - Sync an accelerated migration operation + * @obj: The migrating object. + * @flags: waiting flags. Currently supports only I915_WAIT_INTERRUPTIBLE. + * + * Wait for any pending async migration operation on the object, + * whether it's explicitly (i915_gem_object_migrate()) or implicitly + * (swapin, initial clearing) initiated. + * + * Return: 0 if successful, -ERESTARTSYS if a signal was hit during waiting. + */ +int i915_gem_object_wait_migration(struct drm_i915_gem_object *obj, + unsigned int flags) +{ + dma_fence_might_wait(); + /* NOP for now. */ + return 0; +}