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 553EAC77B7E for ; Mon, 29 May 2023 15:23:38 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 22A2B10E02A; Mon, 29 May 2023 15:23:38 +0000 (UTC) Received: from mga14.intel.com (mga14.intel.com [192.55.52.115]) by gabe.freedesktop.org (Postfix) with ESMTPS id 2B2BA10E02A for ; Mon, 29 May 2023 15:23:35 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1685373815; x=1716909815; h=message-id:date:mime-version:subject:to:references:from: in-reply-to:content-transfer-encoding; bh=X1BSaKui2ObwtR86V73lNZr3xg8FFe3P29v/RS6o794=; b=Qoy8W4Drb62DK6MnjaKjw1bnBGNhm5rWWt+zerW3ig9FYMru5NIzBc5w NCFWGulOtglLlbOfoLD8a2KgwTgKRAWdkPgj8svyzeILQDMunQOQu8NSo 4lTl1PNwIkTZ28bRbxGbfm7VG+fYaBlThN5VP7gk9Cj+YdxZwcexS3VxO 6piLapYsTHULhXFGVEmSecycCz9v6tzrKhHMZA5Xpy/OrarrgZ497FRiz PV/R2kvqQBB5dvoIgu57+DRSpmOjBBRzQihxWhz1HvGkGuf77i3r7w47B 4DoAOwuXN+XJcJvRRFYDcWjCM0QEVx5TIo6E1xV69gcPjOfezG0Rlgs+0 A==; X-IronPort-AV: E=McAfee;i="6600,9927,10725"; a="354742143" X-IronPort-AV: E=Sophos;i="6.00,201,1681196400"; d="scan'208";a="354742143" Received: from orsmga004.jf.intel.com ([10.7.209.38]) by fmsmga103.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 29 May 2023 08:23:34 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6600,9927,10725"; a="830426167" X-IronPort-AV: E=Sophos;i="6.00,201,1681196400"; d="scan'208";a="830426167" Received: from olhomerx-mobl.ger.corp.intel.com (HELO [10.252.46.198]) ([10.252.46.198]) by orsmga004-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 29 May 2023 08:23:31 -0700 Message-ID: Date: Mon, 29 May 2023 17:23:29 +0200 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101 Firefox/102.0 Thunderbird/102.11.0 Content-Language: en-US To: =?UTF-8?Q?Thomas_Hellstr=c3=b6m?= , intel-xe@lists.freedesktop.org References: <20230526121101.1619278-1-maarten.lankhorst@linux.intel.com> <20230526121101.1619278-5-maarten.lankhorst@linux.intel.com> <10e2a992-dcc4-8316-8e8b-f674641bed87@linux.intel.com> <5fece445-89eb-0e97-e934-1c896fb27bd7@linux.intel.com> <30734f17-58d0-bf50-ac83-dc4e7d65550d@linux.intel.com> <0d897527-d80b-5fd9-f338-2a5d83d6c682@linux.intel.com> From: Maarten Lankhorst In-Reply-To: <0d897527-d80b-5fd9-f338-2a5d83d6c682@linux.intel.com> Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Subject: Re: [Intel-xe] [PATCH 4/5] drm/xe: Prevent evicting for page tables 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 2023-05-29 17:13, Thomas Hellström wrote: > > On 5/29/23 17:11, Maarten Lankhorst wrote: >> Hey, >> >> On 2023-05-29 17:02, Thomas Hellström wrote: >>> On 5/29/23 15:44, Maarten Lankhorst wrote: >>>> On 2023-05-26 14:35, Thomas Hellström wrote: >>>>> On 5/26/23 14:11, Maarten Lankhorst wrote: >>>>>> When creating page tables from xe_exec_ioctl, we may end up freeing >>>>>> memory we just validated. To be certain this does not happen, do not >>>>>> allow the current reservation to be evicted from the ioctl. >>>>>> >>>>>> Callchain: >>>>>> [  109.008522]  xe_bo_move_notify+0x5c/0xf0 [xe] >>>>>> [  109.008548]  xe_bo_move+0x90/0x510 [xe] >>>>>> [  109.008573]  ttm_bo_handle_move_mem+0xb7/0x170 [ttm] >>>>>> [  109.008581]  ttm_bo_swapout+0x15e/0x360 [ttm] >>>>>> [  109.008586]  ttm_device_swapout+0xc2/0x110 [ttm] >>>>>> [  109.008592]  ttm_global_swapout+0x47/0xc0 [ttm] >>>>>> [  109.008598]  ttm_tt_populate+0x7a/0x130 [ttm] >>>>>> [  109.008603]  ttm_bo_handle_move_mem+0x160/0x170 [ttm] >>>>>> [  109.008609]  ttm_bo_validate+0xe5/0x1d0 [ttm] >>>>>> [  109.008614]  ttm_bo_init_reserved+0xac/0x190 [ttm] >>>>>> [  109.008620]  __xe_bo_create_locked+0x153/0x260 [xe] >>>>>> [  109.008645]  xe_bo_create_locked_range+0x77/0x360 [xe] >>>>>> [  109.008671]  xe_bo_create_pin_map_at+0x33/0x1f0 [xe] >>>>>> [  109.008695]  xe_bo_create_pin_map+0x11/0x20 [xe] >>>>>> [  109.008721]  xe_pt_create+0x69/0xf0 [xe] >>>>>> [  109.008749]  xe_pt_stage_bind_entry+0x208/0x430 [xe] >>>>>> [  109.008776]  xe_pt_walk_range+0xe9/0x2a0 [xe] >>>>>> [  109.008802]  xe_pt_walk_range+0x223/0x2a0 [xe] >>>>>> [  109.008828]  xe_pt_walk_range+0x223/0x2a0 [xe] >>>>>> [  109.008853]  __xe_pt_bind_vma+0x28d/0xbd0 [xe] >>>>>> [  109.008878]  xe_vm_bind_vma+0xc7/0x2f0 [xe] >>>>>> [  109.008904]  xe_vm_rebind+0x72/0x160 [xe] >>>>>> [  109.008930]  xe_exec_ioctl+0x22b/0xa70 [xe] >>>>>> [  109.008955]  drm_ioctl_kernel+0xb9/0x150 [drm] >>>>>> [  109.008972]  drm_ioctl+0x210/0x430 [drm] >>>>>> [  109.008988]  __x64_sys_ioctl+0x85/0xb0 >>>>>> [  109.008990]  do_syscall_64+0x38/0x90 >>>>>> [  109.008991]  entry_SYSCALL_64_after_hwframe+0x72/0xdc >>>>>> >>>>>> Original warning: >>>>>> [ 5613.149126] WARNING: CPU: 3 PID: 45883 at drivers/gpu/drm/xe/xe_vm.c:504 xe_vm_unlock_dma_resv+0x43/0x50 [xe] >>>>>> ... >>>>>> [ 5613.226398] RIP: 0010:xe_vm_unlock_dma_resv+0x43/0x50 [xe] >>>>>> [ 5613.316098] Call Trace: >>>>>> [ 5613.318595]  >>>>>> [ 5613.320743]  xe_exec_ioctl+0x383/0x8a0 [xe] >>>>>> [ 5613.325278]  ? __is_insn_slot_addr+0x8e/0x110 >>>>>> [ 5613.329719]  ? __is_insn_slot_addr+0x8e/0x110 >>>>>> [ 5613.334116]  ? kernel_text_address+0x75/0xf0 >>>>>> [ 5613.338429]  ? __pfx_stack_trace_consume_entry+0x10/0x10 >>>>>> [ 5613.343778]  ? __kernel_text_address+0x9/0x40 >>>>>> [ 5613.348181]  ? unwind_get_return_address+0x1a/0x30 >>>>>> [ 5613.353013]  ? __pfx_stack_trace_consume_entry+0x10/0x10 >>>>>> [ 5613.358362]  ? arch_stack_walk+0x99/0xf0 >>>>>> [ 5613.362329]  ? rcu_read_lock_sched_held+0xb/0x70 >>>>>> [ 5613.366996]  ? lock_acquire+0x287/0x2f0 >>>>>> [ 5613.370873]  ? rcu_read_lock_sched_held+0xb/0x70 >>>>>> [ 5613.375530]  ? rcu_read_lock_sched_held+0xb/0x70 >>>>>> [ 5613.380181]  ? lock_release+0x225/0x2e0 >>>>>> [ 5613.384059]  ? __pfx_xe_exec_ioctl+0x10/0x10 [xe] >>>>>> [ 5613.389092]  drm_ioctl_kernel+0xc0/0x170 >>>>>> [ 5613.393068]  drm_ioctl+0x1b7/0x490 >>>>>> [ 5613.396519]  ? __pfx_xe_exec_ioctl+0x10/0x10 [xe] >>>>>> [ 5613.401547]  ? lock_release+0x225/0x2e0 >>>>>> [ 5613.405432]  __x64_sys_ioctl+0x8a/0xb0 >>>>>> [ 5613.409232]  do_syscall_64+0x37/0x90 >>>>>> >>>>>> Signed-off-by: Maarten Lankhorst >>>>>> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/239 >>>>> Did you look at passing around the ttm_operation_ctx, or a "allow_res_evict" bool? >>>>> In any case would be good to have this fixed asap, so >>>>> >>>>> Reviewed-by: Thomas Hellström >>>>> >>>> I considered it, but the original callchain was too long. I don't think there is any usecase >>>> in which we want to evict from the current context to make room for new pagetables for VM_BIND. >>>> Anything locked is most likely used, making room by evicting from current VM (or its bound extobjs) >>>> will likely lead to ENOSPC anyway. >>> Well, I think the use-case where this will cause problems is if we're doing a single VM_BIND on a brand new VRAM BO, and need to evict other VRAM bos from the same VM to make room. >>> >>> This will then ofc ENOSPC on the next exec, but if we were to introduce a two-pass validation scheme, where we explicitly move suitable BOs with multiple placement options to TT on the first ENOSPC, we could avoid that... >> Allowing same-reservation eviction will allow you to evict the BO its VM_BIND page table, leaving no entries to write. :-) > > Aren't those pinned? > > /Thomas Not the bo itself. ~Maarten