From mboxrd@z Thu Jan 1 00:00:00 1970 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 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 smtp.lore.kernel.org (Postfix) with ESMTPS id A3F1AC36002 for ; Mon, 24 Mar 2025 14:41:06 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 6B5C910E309; Mon, 24 Mar 2025 14:41:06 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="PfDuSrJn"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [192.198.163.17]) by gabe.freedesktop.org (Postfix) with ESMTPS id 66AB910E309 for ; Mon, 24 Mar 2025 14:41:05 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1742827265; x=1774363265; h=message-id:date:mime-version:subject:to:cc:references: from:in-reply-to:content-transfer-encoding; bh=rzt3medV+bcewlMI5gdo1WaF7ZmV3kH1iCSnEeblOgM=; b=PfDuSrJn6INvkNrMoMjD+5ZvYkgUKZJbN+QondNT4QunMCtZtduug+X9 dhGNk5lErK1rlTkiocBb6FUYNTn0CEy7E7shB4E9b4zkoHCjWh+Q17OZ8 5blTVKDdYS8JII46o639UqhyshjctyLpON4sTG0v0Kr3hvOswSDhrrBGj OeEGHS0k0Pkbex78mt3Z/I7Y0n0CX7L8NsgJkVfUUCuLCXXiGjtd0qMab aMbwjeJPqycgJa6p1KLS/QKaXqkQgsW0fUCtsXV9oQEhiKccn3L309ngV Ixei83roT3hW4L98BZC12oaNLpNXpXLzu2xRg0wKPpplm+yFEiBsk9QTI w==; X-CSE-ConnectionGUID: dRM93ItwTN62UBv5FHri5w== X-CSE-MsgGUID: 9QpcJFBaRnuqNXHkLO6LTQ== X-IronPort-AV: E=McAfee;i="6700,10204,11383"; a="43910936" X-IronPort-AV: E=Sophos;i="6.14,272,1736841600"; d="scan'208";a="43910936" Received: from fmviesa006.fm.intel.com ([10.60.135.146]) by fmvoesa111.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 24 Mar 2025 07:41:05 -0700 X-CSE-ConnectionGUID: fqVy8mq0SkiAv5NZ3+d4RA== X-CSE-MsgGUID: DuPetowBQau+Qg9E3YtqOg== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.14,272,1736841600"; d="scan'208";a="123864237" Received: from dhhellew-desk2.ger.corp.intel.com (HELO [10.245.245.252]) ([10.245.245.252]) by fmviesa006-auth.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 24 Mar 2025 07:41:04 -0700 Message-ID: <9b68e406-21b1-4500-ae3e-662c3109fe89@intel.com> Date: Mon, 24 Mar 2025 14:41:01 +0000 MIME-Version: 1.0 User-Agent: Mozilla Thunderbird Subject: Re: [PATCH v2 3/5] drm/xe/bo: Add a bo remove callback To: =?UTF-8?Q?Thomas_Hellstr=C3=B6m?= , intel-xe@lists.freedesktop.org Cc: himal.prasad.ghimiray@intel.com, Matthew Brost References: <20250321163416.45217-1-thomas.hellstrom@linux.intel.com> <20250321163416.45217-4-thomas.hellstrom@linux.intel.com> Content-Language: en-GB From: Matthew Auld In-Reply-To: <20250321163416.45217-4-thomas.hellstrom@linux.intel.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit X-BeenThere: intel-xe@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Intel Xe graphics driver List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: intel-xe-bounces@lists.freedesktop.org Sender: "Intel-xe" On 21/03/2025 16:34, Thomas Hellström wrote: > On device unbind, migrate exported bos, including pagemap bos to > system. This allows importers to take proper action without > disruption. In particular, SVM clients on remote devices may > continue as if nothing happened, and can chose a different > placement. > > The evict_flags() placement is chosen in such a way that bos that > aren't exported are purged. > > For pinned bos, we unmap DMA, but their pages are not freed yet > since we can't be 100% sure they are not accessed. > > v2: > - Address review comments. (Matthew Auld). > > Signed-off-by: Thomas Hellström > --- > drivers/gpu/drm/xe/xe_bo.c | 54 ++++++++++++++++++++---- > drivers/gpu/drm/xe/xe_bo.h | 2 + > drivers/gpu/drm/xe/xe_bo_evict.c | 62 ++++++++++++++++++++++++++-- > drivers/gpu/drm/xe/xe_bo_evict.h | 1 + > drivers/gpu/drm/xe/xe_device.c | 6 ++- > drivers/gpu/drm/xe/xe_device_types.h | 6 ++- > 6 files changed, 117 insertions(+), 14 deletions(-) > > diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c > index 64f9c936eea0..9d043d2c30fd 100644 > --- a/drivers/gpu/drm/xe/xe_bo.c > +++ b/drivers/gpu/drm/xe/xe_bo.c > @@ -55,6 +55,8 @@ static struct ttm_placement sys_placement = { > .placement = &sys_placement_flags, > }; > > +static struct ttm_placement purge_placement; > + > static const struct ttm_place tt_placement_flags[] = { > { > .fpfn = 0, > @@ -281,6 +283,8 @@ int xe_bo_placement_for_flags(struct xe_device *xe, struct xe_bo *bo, > static void xe_evict_flags(struct ttm_buffer_object *tbo, > struct ttm_placement *placement) > { > + struct xe_device *xe = container_of(tbo->bdev, typeof(*xe), ttm); > + bool device_unplugged = drm_dev_is_unplugged(&xe->drm); > struct xe_bo *bo; > > if (!xe_bo_is_xe_bo(tbo)) { > @@ -290,7 +294,7 @@ static void xe_evict_flags(struct ttm_buffer_object *tbo, > return; > } > > - *placement = sys_placement; > + *placement = device_unplugged ? purge_placement : sys_placement; > return; > } > > @@ -300,6 +304,11 @@ static void xe_evict_flags(struct ttm_buffer_object *tbo, > return; > } > > + if (device_unplugged && !tbo->base.dma_buf) { > + *placement = purge_placement; > + return; > + } > + > /* > * For xe, sg bos that are evicted to system just triggers a > * rebind of the sg list upon subsequent validation to XE_PL_TT. > @@ -657,11 +666,20 @@ static int xe_bo_move_dmabuf(struct ttm_buffer_object *ttm_bo, > struct xe_ttm_tt *xe_tt = container_of(ttm_bo->ttm, struct xe_ttm_tt, > ttm); > struct xe_device *xe = ttm_to_xe_device(ttm_bo->bdev); > + bool device_unplugged = drm_dev_is_unplugged(&xe->drm); > struct sg_table *sg; > > xe_assert(xe, attach); > xe_assert(xe, ttm_bo->ttm); > > + if (device_unplugged && new_res->mem_type == XE_PL_SYSTEM && > + ttm_bo->sg) { > + dma_resv_wait_timeout(ttm_bo->base.resv, DMA_RESV_USAGE_BOOKKEEP, > + false, MAX_SCHEDULE_TIMEOUT); > + dma_buf_unmap_attachment(attach, ttm_bo->sg, DMA_BIDIRECTIONAL); > + ttm_bo->sg = NULL; > + } > + > if (new_res->mem_type == XE_PL_SYSTEM) > goto out; > > @@ -1224,6 +1242,31 @@ int xe_bo_restore_pinned(struct xe_bo *bo) > return ret; > } > > +int xe_bo_dma_unmap_pinned(struct xe_bo *bo) > +{ > + struct ttm_buffer_object *ttm_bo = &bo->ttm; > + struct ttm_tt *tt = ttm_bo->ttm; > + > + if (tt) { > + struct xe_ttm_tt *xe_tt = container_of(tt, typeof(*xe_tt), ttm); > + > + if (ttm_bo->type == ttm_bo_type_sg && ttm_bo->sg) { > + dma_buf_unmap_attachment(ttm_bo->base.import_attach, > + ttm_bo->sg, > + DMA_BIDIRECTIONAL); > + ttm_bo->sg = NULL; > + xe_tt->sg = NULL; > + } else if (xe_tt->sg) { > + dma_unmap_sgtable(xe_tt->xe->drm.dev, xe_tt->sg, > + DMA_BIDIRECTIONAL, 0); > + sg_free_table(xe_tt->sg); > + xe_tt->sg = NULL; > + } > + } > + > + return 0; > +} > + > static unsigned long xe_ttm_io_mem_pfn(struct ttm_buffer_object *ttm_bo, > unsigned long page_offset) > { > @@ -2102,12 +2145,9 @@ int xe_bo_pin_external(struct xe_bo *bo) > if (err) > return err; > > - if (xe_bo_is_vram(bo)) { > - spin_lock(&xe->pinned.lock); > - list_add_tail(&bo->pinned_link, > - &xe->pinned.external_vram); > - spin_unlock(&xe->pinned.lock); > - } > + spin_lock(&xe->pinned.lock); > + list_add_tail(&bo->pinned_link, &xe->pinned.external); > + spin_unlock(&xe->pinned.lock); > } > > ttm_bo_pin(&bo->ttm); > diff --git a/drivers/gpu/drm/xe/xe_bo.h b/drivers/gpu/drm/xe/xe_bo.h > index ec3e4446d027..05479240bf75 100644 > --- a/drivers/gpu/drm/xe/xe_bo.h > +++ b/drivers/gpu/drm/xe/xe_bo.h > @@ -276,6 +276,8 @@ int xe_bo_evict(struct xe_bo *bo, bool force_alloc); > int xe_bo_evict_pinned(struct xe_bo *bo); > int xe_bo_restore_pinned(struct xe_bo *bo); > > +int xe_bo_dma_unmap_pinned(struct xe_bo *bo); > + > extern const struct ttm_device_funcs xe_ttm_funcs; > extern const char *const xe_mem_type_to_name[]; > > diff --git a/drivers/gpu/drm/xe/xe_bo_evict.c b/drivers/gpu/drm/xe/xe_bo_evict.c > index 1eeb3910450b..ba0260900868 100644 > --- a/drivers/gpu/drm/xe/xe_bo_evict.c > +++ b/drivers/gpu/drm/xe/xe_bo_evict.c > @@ -93,8 +93,8 @@ int xe_bo_evict_all(struct xe_device *xe) > } > } > > - ret = xe_bo_apply_to_pinned(xe, &xe->pinned.external_vram, > - &xe->pinned.external_vram, > + ret = xe_bo_apply_to_pinned(xe, &xe->pinned.external, > + &xe->pinned.external_evicted, > xe_bo_evict_pinned); Just one question here. Do you know if this changes the existing behaviour during suspend-resume for such buffers? Say we have pinned tt that was exported to another driver (not dynamic), will this now move it from tt -> system? If so, is that correct? Does it make sense to skip this entirely for suspend-resume for the external pinned case? I'm assuming it can't be in VRAM for this case, so seems like we don't need to do anything here? Otherwise looks good. > > /* > @@ -181,8 +181,8 @@ int xe_bo_restore_user(struct xe_device *xe) > return 0; > > /* Pinned user memory in VRAM should be validated on resume */ > - ret = xe_bo_apply_to_pinned(xe, &xe->pinned.external_vram, > - &xe->pinned.external_vram, > + ret = xe_bo_apply_to_pinned(xe, &xe->pinned.external_evicted, > + &xe->pinned.external, > xe_bo_restore_pinned); > > /* Wait for restore to complete */ > @@ -191,3 +191,57 @@ int xe_bo_restore_user(struct xe_device *xe) > > return ret; > } > + > +static void xe_bo_pci_dev_remove_pinned(struct xe_device *xe) > +{ > + struct xe_tile *tile; > + unsigned int id; > + > + (void) xe_bo_apply_to_pinned(xe, &xe->pinned.external, > + &xe->pinned.external, > + xe_bo_dma_unmap_pinned); > + for_each_tile(tile, xe, id) > + xe_tile_migrate_wait(tile); > + > + (void) xe_bo_apply_to_pinned(xe, &xe->pinned.kernel_bo_present, > + &xe->pinned.kernel_bo_present, > + xe_bo_dma_unmap_pinned); > + > +} > + > + > + > +/** > + * xe_bo_pci_dev_remove_all() - Handle bos when the pci_device is about to be removed > + * @xe: The xe device. > + * > + * On pci_device removal we need to drop all dma mappings and move > + * the data of exported bos out to system. This includes SVM bos and > + * exported dma-buf bos. This is done by evicting all bos, but > + * the evict placement in xe_evict_flags() is chosen such that all > + * bos except those mentioned are purged, and thus their memory > + * is released. > + * > + * For pinned bos, we're unmapping dma. > + */ > +void xe_bo_pci_dev_remove_all(struct xe_device *xe) > +{ > + unsigned int mem_type; > + > + /* > + * Move pagemap bos and exported dma-buf to system, and > + * purge everything else. > + */ > + for (mem_type = XE_PL_VRAM1; mem_type >= XE_PL_TT; --mem_type) { > + struct ttm_resource_manager *man = > + ttm_manager_type(&xe->ttm, mem_type); > + > + if (man) { > + int ret = ttm_resource_manager_evict_all(&xe->ttm, man); > + > + drm_WARN_ON(&xe->drm, ret); > + } > + } > + > + xe_bo_pci_dev_remove_pinned(xe); > +} > diff --git a/drivers/gpu/drm/xe/xe_bo_evict.h b/drivers/gpu/drm/xe/xe_bo_evict.h > index 746894798852..0f06140163d6 100644 > --- a/drivers/gpu/drm/xe/xe_bo_evict.h > +++ b/drivers/gpu/drm/xe/xe_bo_evict.h > @@ -12,4 +12,5 @@ int xe_bo_evict_all(struct xe_device *xe); > int xe_bo_restore_kernel(struct xe_device *xe); > int xe_bo_restore_user(struct xe_device *xe); > > +void xe_bo_pci_dev_remove_all(struct xe_device *xe); > #endif > diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c > index b2f656b2a563..739861edcd30 100644 > --- a/drivers/gpu/drm/xe/xe_device.c > +++ b/drivers/gpu/drm/xe/xe_device.c > @@ -23,6 +23,7 @@ > #include "regs/xe_gt_regs.h" > #include "regs/xe_regs.h" > #include "xe_bo.h" > +#include "xe_bo_evict.h" > #include "xe_debugfs.h" > #include "xe_devcoredump.h" > #include "xe_dma_buf.h" > @@ -468,8 +469,9 @@ struct xe_device *xe_device_create(struct pci_dev *pdev, > > spin_lock_init(&xe->pinned.lock); > INIT_LIST_HEAD(&xe->pinned.kernel_bo_present); > - INIT_LIST_HEAD(&xe->pinned.external_vram); > + INIT_LIST_HEAD(&xe->pinned.external); > INIT_LIST_HEAD(&xe->pinned.evicted); > + INIT_LIST_HEAD(&xe->pinned.external_evicted); > > xe->preempt_fence_wq = alloc_ordered_workqueue("xe-preempt-fence-wq", > WQ_MEM_RECLAIM); > @@ -925,6 +927,8 @@ void xe_device_remove(struct xe_device *xe) > xe_display_unregister(xe); > > drm_dev_unplug(&xe->drm); > + > + xe_bo_pci_dev_remove_all(xe); > } > > void xe_device_shutdown(struct xe_device *xe) > diff --git a/drivers/gpu/drm/xe/xe_device_types.h b/drivers/gpu/drm/xe/xe_device_types.h > index 4472bf849f2b..aa83ba414301 100644 > --- a/drivers/gpu/drm/xe/xe_device_types.h > +++ b/drivers/gpu/drm/xe/xe_device_types.h > @@ -426,8 +426,10 @@ struct xe_device { > struct list_head kernel_bo_present; > /** @pinned.evicted: pinned BO that have been evicted */ > struct list_head evicted; > - /** @pinned.external_vram: pinned external BO in vram*/ > - struct list_head external_vram; > + /** @pinned.external: pinned external and dma-buf. */ > + struct list_head external; > + /** @pinned.external: evicted pinned external and dma-buf. */ > + struct list_head external_evicted; > } pinned; > > /** @ufence_wq: user fence wait queue */