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 616B5C021BD for ; Mon, 24 Feb 2025 13:43:56 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id CAE2810E3EE; Mon, 24 Feb 2025 13:42:15 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="ZemFphjw"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [198.175.65.17]) by gabe.freedesktop.org (Postfix) with ESMTPS id 1279410E129 for ; Mon, 24 Feb 2025 03:23:37 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1740367417; x=1771903417; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=CbZTKoFPJ+hHuaptQZd+fb5Ic6P/8TjkA4NtW0jdxa0=; b=ZemFphjwRT8TwpsmpnECNfCBgs8li+1Fw46rFWNaZ+Kf+XDQ0jM0Gj0o abmsw6HOJmKl2isgu5tHQkNG/Gj4GjLnaHSjI0+N76csJ6kGFsIzk/iYo xz49maOp6LOitRYilU+Dwqu8owr4TJWTuKNpt7UoiEzary2V12eF0h/vW 7BAAGJ3p0aye/2PEEPlZMBsIT/mp0k6A315CPSfeNSNbfS0Ryq4dx4d9L zfxy8X2mYYpfcwl09212ViefX9gLQ7a2iIcwo4NHvFxTIG8tc+tZbxQkh 7LeF3Gnv47IAdABFqRI6j44eH/A3jhIJsjIcGw+S0i1BzeoT9Fm8zntWB A==; X-CSE-ConnectionGUID: /w9g7aqiRdmw/ynca56Cqw== X-CSE-MsgGUID: pGSsT0V2T7mcgrtw/jUoPQ== X-IronPort-AV: E=McAfee;i="6700,10204,11354"; a="41126883" X-IronPort-AV: E=Sophos;i="6.13,309,1732608000"; d="scan'208";a="41126883" Received: from orviesa006.jf.intel.com ([10.64.159.146]) by orvoesa109.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 23 Feb 2025 19:23:36 -0800 X-CSE-ConnectionGUID: xDXALJckRGWmoAgbhgeCMg== X-CSE-MsgGUID: ek4oC3jbRniq81NKg8DUeA== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.13,309,1732608000"; d="scan'208";a="115907817" Received: from lstrano-desk.jf.intel.com ([10.54.39.91]) by orviesa006-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 23 Feb 2025 19:23:37 -0800 From: Matthew Brost To: intel-xe@lists.freedesktop.org Cc: thomas.hellstrom@linux.intel.com, matthew.auld@intel.com Subject: [PATCH v2 1/3] drm/xe/userptr: fix EFAULT handling Date: Sun, 23 Feb 2025 19:24:36 -0800 Message-Id: <20250224032438.3025464-2-matthew.brost@intel.com> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20250224032438.3025464-1-matthew.brost@intel.com> References: <20250224032438.3025464-1-matthew.brost@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 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" From: Matthew Auld Currently we treat EFAULT from hmm_range_fault() as a non-fatal error when called from xe_vm_userptr_pin() with the idea that we want to avoid killing the entire vm and chucking an error, under the assumption that the user just did an unmap or something, and has no intention of actually touching that memory from the GPU. At this point we have already zapped the PTEs so any access should generate a page fault, and if the pin fails there also it will then become fatal. However it looks like it's possible for the userptr vma to still be on the rebind list in preempt_rebind_work_func(), if we had to retry the pin again due to something happening in the caller before we did the rebind step, but in the meantime needing to re-validate the userptr and this time hitting the EFAULT. This might explain an internal user report of hitting: [ 191.738349] WARNING: CPU: 1 PID: 157 at drivers/gpu/drm/xe/xe_res_cursor.h:158 xe_pt_stage_bind.constprop.0+0x60a/0x6b0 [xe] [ 191.738551] Workqueue: xe-ordered-wq preempt_rebind_work_func [xe] [ 191.738616] RIP: 0010:xe_pt_stage_bind.constprop.0+0x60a/0x6b0 [xe] [ 191.738690] Call Trace: [ 191.738692] [ 191.738694] ? show_regs+0x69/0x80 [ 191.738698] ? __warn+0x93/0x1a0 [ 191.738703] ? xe_pt_stage_bind.constprop.0+0x60a/0x6b0 [xe] [ 191.738759] ? report_bug+0x18f/0x1a0 [ 191.738764] ? handle_bug+0x63/0xa0 [ 191.738767] ? exc_invalid_op+0x19/0x70 [ 191.738770] ? asm_exc_invalid_op+0x1b/0x20 [ 191.738777] ? xe_pt_stage_bind.constprop.0+0x60a/0x6b0 [xe] [ 191.738834] ? ret_from_fork_asm+0x1a/0x30 [ 191.738849] bind_op_prepare+0x105/0x7b0 [xe] [ 191.738906] ? dma_resv_reserve_fences+0x301/0x380 [ 191.738912] xe_pt_update_ops_prepare+0x28c/0x4b0 [xe] [ 191.738966] ? kmemleak_alloc+0x4b/0x80 [ 191.738973] ops_execute+0x188/0x9d0 [xe] [ 191.739036] xe_vm_rebind+0x4ce/0x5a0 [xe] [ 191.739098] ? trace_hardirqs_on+0x4d/0x60 [ 191.739112] preempt_rebind_work_func+0x76f/0xd00 [xe] Followed by NPD, when running some workload, since the sg was never actually populated but the vma is still marked for rebind when it should be skipped for this special EFAULT case. And from the logs it does seem like we hit this special EFAULT case before the explosions. v2 (MattB): - Move earlier Fixes: 521db22a1d70 ("drm/xe: Invalidate userptr VMA on page pin fault") Signed-off-by: Matthew Auld Cc: Matthew Brost Cc: Thomas Hellström Cc: # v6.10+ Reviewed-by: Matthew Brost Reviewed-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_vm.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index d664f2e418b2..ea2e287e6526 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -681,6 +681,18 @@ int xe_vm_userptr_pin(struct xe_vm *vm) err = xe_vma_userptr_pin_pages(uvma); if (err == -EFAULT) { list_del_init(&uvma->userptr.repin_link); + /* + * We might have already done the pin once already, but + * then had to retry before the re-bind happened, due + * some other condition in the caller, but in the + * meantime the userptr got dinged by the notifier such + * that we need to revalidate here, but this time we hit + * the EFAULT. In such a case make sure we remove + * ourselves from the rebind list to avoid going down in + * flames. + */ + if (!list_empty(&uvma->vma.combined_links.rebind)) + list_del_init(&uvma->vma.combined_links.rebind); /* Wait for pending binds */ xe_vm_lock(vm, false); -- 2.34.1