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 A5858CD5BB0 for ; Fri, 22 May 2026 08:51:43 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id E440B10E1C1; Fri, 22 May 2026 08:51:42 +0000 (UTC) Received: from mail-oa1-f71.google.com (mail-oa1-f71.google.com [209.85.160.71]) by gabe.freedesktop.org (Postfix) with ESMTPS id 4BB3D10E1C1 for ; Fri, 22 May 2026 08:51:40 +0000 (UTC) Received: by mail-oa1-f71.google.com with SMTP id 586e51a60fabf-43a538158ccso8199021fac.0 for ; Fri, 22 May 2026 01:51:40 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1779439899; x=1780044699; h=content-transfer-encoding:cc:to:from:subject:message-id:in-reply-to :date:mime-version:x-gm-message-state:from:to:cc:subject:date :message-id:reply-to; bh=MIi3MMN5y++s1UIskbwnxi/GVcsH2WsJKkCbw/xDCoI=; b=kX4dbH8KYE8ro0ElxPtioFlRmCptZVdxxYmU1y5s09CI7KN8rv5102u5bj3cYS2cZN 5z2hWE6i7/wNJ4n4309QKni7JG1XbtpzifwHA/0n5ugLviwRIPeFFOXfw1hR/tWC5Otc dA1HwW+Q7z9UGTnVNI5+9s/6k9d1+9U+QEg0YtChHZzQj8ESI0KgDZKFZ4Ib0kXAeyJ4 25m1IC5XnBctnAN3HXtD2vApSCTeQaNIvLkf/+6s/h0kjGYz9NuGlfwm0UQW7SSIW67k Xmg9YarWiB9dg/aS2HvugcQubBwqZWkHzhSCNgwCQ5fNGcasZn3+GS4K+RIEg7qt+SL1 F7Ug== X-Forwarded-Encrypted: i=1; AFNElJ/iE2LUuSO0mQOALDreavRzTarnmZ0RLvOBdcZCFOlk3+ck6ZVrF4/qgkVMVqCkvEM6G9FejMQq3z4=@lists.freedesktop.org X-Gm-Message-State: AOJu0YzJXFS/XXrAlDg/gW9RTkVTvGKW8PNMfbjrWbkZyzxxB/feudcO MpHgchr+qFS7Z86ASV7rLhsh1yuD4yGvOuQP4tReqNWwIHKQa6EFVkr/cH/VbtuiwDYB0ZCQwDF 21VOZgMZU8u/P5LqRcQF9v7Mt/LGVeDhZh00VkQAxONihIz24U8b48YgAKNI= MIME-Version: 1.0 X-Received: by 2002:a05:6820:4512:b0:69d:7d6e:e198 with SMTP id 006d021491bc7-69d7eb8d767mr1117466eaf.16.1779439899588; Fri, 22 May 2026 01:51:39 -0700 (PDT) Date: Fri, 22 May 2026 01:51:39 -0700 In-Reply-To: <18b1d7267ace8b76.329809b8dd1e13b3.4134c0f069791aa6@ryasuoka-thinkpadx1carbongen9.tokyo.csb> X-Google-Appengine-App-Id: s~syzkaller X-Google-Appengine-App-Id-Alias: syzkaller Message-ID: <6a10191b.170a0220.2e5ec6.0005.GAE@google.com> Subject: Re: [PATCH v2] drm/virtio: abort virtqueue wait on device removal to avoid hung task From: syzbot To: ryasuoka@redhat.com Cc: airlied@redhat.com, dmitry.osipenko@collabora.com, dri-devel@lists.freedesktop.org, gurchetansingh@chromium.org, kraxel@redhat.com, linux-kernel@vger.kernel.org, maarten.lankhorst@linux.intel.com, mripard@kernel.org, olvaffe@gmail.com, ryasuoka@redhat.com, simona@ffwll.ch, tzimmermann@suse.de, virtualization@lists.linux.dev Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable 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" > Hi Dmitry, > Thank you for your review and the comment. > > On 21/05/2026 22:01, Dmitry Osipenko wrote: >> 21.05.2026 05:19, Ryosuke Yasuoka =D0=BF=D0=B8=D1=88=D0=B5=D1=82: >>> virtio_gpu_queue_ctrl_sgs() and virtio_gpu_queue_cursor() use >>> wait_event() without any abort condition when waiting for virtqueue >>> space. If the host device stops processing commands, these waits block >>> indefinitely inside a drm_dev_enter/exit() critical section. Since >>> drm_dev_unplug(), which is called in device removal and system shutdown >>> call path, blocks on synchronize_srcu() until all critical sections >>> complete, device removal and system shutdown also hang. >>>=20 >>> Add a vqs_released flag to virtio_gpu_device and include it in the >>> wait_event() condition. Set the flag and wake up both queues in a new >>> virtio_gpu_release_vqs() helper, called before drm_dev_unplug() in both >>> virtio_gpu_remove() and virtio_gpu_shutdown(). When the flag is set, th= e >>> wait returns immediately and the command is aborted, following the same >>> cleanup path as drm_dev_enter() failure. >>>=20 >>> Reported-by: syzbot+d6dd6f86d3aaf7eebe7406e45c1c6e549453f224@syzkaller.= appspotmail.com >>> Closes: https://syzkaller.appspot.com/bug?id=3Dd6dd6f86d3aaf7eebe7406e4= 5c1c6e549453f224 >>> Reported-by: syzbot+908bd910da5dd79b88de4cf7baf376cc873a922e@syzkaller.= appspotmail.com >>> Closes: https://syzkaller.appspot.com/bug?id=3D908bd910da5dd79b88de4cf7= baf376cc873a922e >>> Signed-off-by: Ryosuke Yasuoka >>> --- >>> Changes in v2: >>> - Update the commit message. >>> - Replace wait_event_timeout() with wait_event() using a compound >>> condition that includes a new vqs_released flag. >>> - Add virtio_gpu_release_vqs() helper to set the flag and wake up >>> both queues, called before drm_dev_unplug() in remove and shutdown >>> paths. >>> - Remove the hardcoded 5-second timeout. Recovery is now driven by >>> the driver flag instead of an arbitrary timeout value. >>> --- >>> drivers/gpu/drm/virtio/virtgpu_drv.c | 15 +++++++++++++++ >>> drivers/gpu/drm/virtio/virtgpu_drv.h | 1 + >>> drivers/gpu/drm/virtio/virtgpu_vq.c | 23 +++++++++++++++++++++-- >>> 3 files changed, 37 insertions(+), 2 deletions(-) >>>=20 >>> diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.c b/drivers/gpu/drm/vir= tio/virtgpu_drv.c >>> index a5ce96fb8a1d..e4fe5e0780f9 100644 >>> --- a/drivers/gpu/drm/virtio/virtgpu_drv.c >>> +++ b/drivers/gpu/drm/virtio/virtgpu_drv.c >>> @@ -119,10 +119,24 @@ static int virtio_gpu_probe(struct virtio_device = *vdev) >>> return ret; >>> } >>> =20 >>> +/* >>> + * Release pending virtqueue waits so the drm_dev_enter/exit() critica= l >>> + * sections complete before drm_dev_unplug() blocks on synchronize_src= u(). >>> + */ >>> +static void virtio_gpu_release_vqs(struct drm_device *dev) >>> +{ >>> + struct virtio_gpu_device *vgdev =3D dev->dev_private; >>> + >>> + vgdev->vqs_released =3D true; >>> + wake_up_all(&vgdev->ctrlq.ack_queue); >>> + wake_up_all(&vgdev->cursorq.ack_queue); >>> +} >>> + >>> static void virtio_gpu_remove(struct virtio_device *vdev) >>> { >>> struct drm_device *dev =3D vdev->priv; >>> =20 >>> + virtio_gpu_release_vqs(dev); >>> drm_dev_unplug(dev); >>> drm_atomic_helper_shutdown(dev); >>> virtio_gpu_deinit(dev); >>> @@ -133,6 +147,7 @@ static void virtio_gpu_shutdown(struct virtio_devic= e *vdev) >>> { >>> struct drm_device *dev =3D vdev->priv; >>> =20 >>> + virtio_gpu_release_vqs(dev); >>> /* stop talking to the device */ >>> drm_dev_unplug(dev); >>> } >>> diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.h b/drivers/gpu/drm/vir= tio/virtgpu_drv.h >>> index f17660a71a3e..0bd69a40857e 100644 >>> --- a/drivers/gpu/drm/virtio/virtgpu_drv.h >>> +++ b/drivers/gpu/drm/virtio/virtgpu_drv.h >>> @@ -235,6 +235,7 @@ struct virtio_gpu_device { >>> =20 >>> struct virtio_gpu_queue ctrlq; >>> struct virtio_gpu_queue cursorq; >>> + bool vqs_released; >>> struct kmem_cache *vbufs; >>> =20 >>> atomic_t pending_commands; >>> diff --git a/drivers/gpu/drm/virtio/virtgpu_vq.c b/drivers/gpu/drm/virt= io/virtgpu_vq.c >>> index 67865810a2e7..8057a9b7356d 100644 >>> --- a/drivers/gpu/drm/virtio/virtgpu_vq.c >>> +++ b/drivers/gpu/drm/virtio/virtgpu_vq.c >>> @@ -396,7 +396,19 @@ static int virtio_gpu_queue_ctrl_sgs(struct virtio= _gpu_device *vgdev, >>> if (vq->num_free < elemcnt) { >>> spin_unlock(&vgdev->ctrlq.qlock); >>> virtio_gpu_notify(vgdev); >>> - wait_event(vgdev->ctrlq.ack_queue, vq->num_free >=3D elemcnt); >>> + wait_event(vgdev->ctrlq.ack_queue, >>> + vq->num_free >=3D elemcnt || vgdev->vqs_released); >>> + /* >>> + * Set by virtio_gpu_release_vqs() to unblock >>> + * synchronize_srcu() wait in drm_dev_unplug(). >>> + */ >>> + if (vgdev->vqs_released) { >>> + if (fence && vbuf->objs) >>> + virtio_gpu_array_unlock_resv(vbuf->objs); >>> + free_vbuf(vgdev, vbuf); >>> + drm_dev_exit(idx); >>> + return -ENODEV; >>> + } >>> goto again; >>> } >>> =20 >>> @@ -566,7 +578,14 @@ static void virtio_gpu_queue_cursor(struct virtio_= gpu_device *vgdev, >>> ret =3D virtqueue_add_sgs(vq, sgs, outcnt, 0, vbuf, GFP_ATOMIC); >>> if (ret =3D=3D -ENOSPC) { >>> spin_unlock(&vgdev->cursorq.qlock); >>> - wait_event(vgdev->cursorq.ack_queue, vq->num_free >=3D outcnt); >>> + wait_event(vgdev->cursorq.ack_queue, >>> + vq->num_free >=3D outcnt || vgdev->vqs_released); >>> + /* See comment in virtio_gpu_queue_ctrl_sgs(). */ >>> + if (vgdev->vqs_released) { >>> + free_vbuf(vgdev, vbuf); >>> + drm_dev_exit(idx); >>> + return; >>> + } >>> spin_lock(&vgdev->cursorq.qlock); >>> goto retry; >>> } else { >>=20 >> What about other wait_event in the driver? Why only these? > > There are other wait_event() calls on vgdev->resp_wq in virtgpu_prime.c > and virtgpu_vram.c. These can also cause a stuck process when the host > device stops or does not respond and should be fixed. However, they are > not related to the issue being fixed here. I think they should be > addressed in a separate commit. > > The wait_event(vgdev->resp_wq) calls wait for a host response, not for > virtqueue ring space. They are not inside a drm_dev_enter/exit() > critical section, so they don't block drm_dev_unplug() -> > synchronize_srcu() and are not part of the deadlock reported by syzbot. > > However, if a process is stuck in wait_event(vgdev->resp_wq), there is > no way to recover other than host device recovery itself. We can still > remove the device and destroy the virtqueues while the process is stuck, > so the host response can never arrive. IIUC, the stuck thread holds a > drm_device reference, preventing vgdev from being freed; this is a > resource leak. > > This can be fixed by adding a wake_up_all() and a vqs_released check > similar to what is done for ctrlq/cursorq. However, I think this should > be a separate commit since the root cause and problem are different. > > Alternatively, the resp_wq wait_event() calls could be converted to > wait_event_interruptible() to fix the issue. But as you mentioned > earlier in the v1 comment, we need the wait_event_interruptible() > rework. > > What do you think? Should I include the resp_wq fix as a separate commit > in this patch series, or leave it for the _interruptible() rework? > > Best regards, > Ryosuke > > I see the command but can't find the corresponding bug. The email is sent to syzbot+HASH@syzkaller.appspotmail.com address but the HASH does not correspond to any known bug. Please double check the address.