public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: JiaJia <physicalmtea@gmail.com>
To: Peter Hilber <peter.hilber@oss.qualcomm.com>
Cc: mst@redhat.com, jasowang@redhat.com, xuanzhuo@linux.alibaba.com,
	eperezma@redhat.com, virtualization@lists.linux.dev,
	linux-kernel@vger.kernel.org, physicalmtea@gmail.com
Subject: [PATCH v2] virtio: rtc: tear down old virtqueues before restore
Date: Fri, 24 Apr 2026 11:14:53 +0800	[thread overview]
Message-ID: <69eaeca0.050a0220.408e2.027b@mx.google.com> (raw)
In-Reply-To: <iwb4axxcxti6bpgpde5cv3n7y46xaopvkbdd2kug3qhn2ehfim@ny45mixj7ncn>

virtio_device_restore() resets the device and restores the negotiated
features before calling ->restore(). viortc_freeze() intentionally
leaves the existing virtqueues in place so the alarm queue can still
wake the system, but viortc_restore() immediately calls
viortc_init_vqs() without first deleting those old queues.

If virtqueue reinitialization fails on virtio-pci, the transport error
path can run vp_del_vqs() against a newly allocated vp_dev->vqs array
while vdev->vqs still contains the old virtqueues. vp_del_vqs() then
looks up queue state through the new array and can dereference a NULL
info pointer in vp_del_vq(), crashing the guest kernel during restore.

Delete the stale virtqueues before rebuilding them. If restore fails
before virtio_device_ready(), reuse the remove path to stop the device.
Once the device is ready, return errors directly instead of deleting the
virtqueues again.

Signed-off-by: JiaJia <physicalmtea@gmail.com>
---
Hi Peter,

thanks for taking a look and for pointing this out. I took another look
at the restore path and agree that deleting the virtqueues from that
error path can race with `viortc_cb_alarmq()` once the device has been
readied.

This v2 follows the direction you suggested. It factors out
`__viortc_remove()`, reuses that path for restore failures before
`virtio_device_ready()`, routes `viortc_init_vqs()` failures in probe
through `err_reset_vdev` for consistent teardown, drops the
`viortc_del_vqs()` helper and the `viortc_msg_xfer()` NULL-vq guard,
and avoids deleting the virtqueues again after `virtio_device_ready()`.

Thanks,
JiaJia

v2:
- reuse the remove path for restore failures before virtio_device_ready()
- route viortc_init_vqs() failures in probe through err_reset_vdev
- drop the viortc_del_vqs() helper and the viortc_msg_xfer() NULL-vq guard
- avoid deleting virtqueues again after virtio_device_ready()

 drivers/virtio/virtio_rtc_driver.c | 28 ++++++++++++++++++++--------
 1 file changed, 20 insertions(+), 8 deletions(-)

diff --git a/drivers/virtio/virtio_rtc_driver.c b/drivers/virtio/virtio_rtc_driver.c
index a57d5e06e..4419735b0 100644
--- a/drivers/virtio/virtio_rtc_driver.c
+++ b/drivers/virtio/virtio_rtc_driver.c
@@ -1257,6 +1257,15 @@ static int viortc_init_vqs(struct viortc_dev *viortc)
 	return 0;
 }
 
+static void __viortc_remove(struct viortc_dev *viortc)
+{
+	struct virtio_device *vdev = viortc->vdev;
+
+	viortc_clocks_deinit(viortc);
+	virtio_reset_device(vdev);
+	vdev->config->del_vqs(vdev);
+}
+
 /**
  * viortc_probe() - probe a virtio_rtc virtio device
  * @vdev: virtio device
@@ -1282,7 +1291,7 @@ static int viortc_probe(struct virtio_device *vdev)
 
 	ret = viortc_init_vqs(viortc);
 	if (ret)
-		return ret;
+		goto err_reset_vdev;
 
 	virtio_device_ready(vdev);
 
@@ -1329,10 +1338,7 @@ static void viortc_remove(struct virtio_device *vdev)
 {
 	struct viortc_dev *viortc = vdev->priv;
 
-	viortc_clocks_deinit(viortc);
-
-	virtio_reset_device(vdev);
-	vdev->config->del_vqs(vdev);
+	__viortc_remove(viortc);
 }
 
 static int viortc_freeze(struct virtio_device *dev)
@@ -1353,9 +1359,11 @@ static int viortc_restore(struct virtio_device *dev)
 	bool notify = false;
 	int ret;
 
+	dev->config->del_vqs(dev);
+
 	ret = viortc_init_vqs(viortc);
 	if (ret)
-		return ret;
+		goto err_remove;
 
 	alarm_viortc_vq = &viortc->vqs[VIORTC_ALARMQ];
 	alarm_vq = alarm_viortc_vq->vq;
@@ -1364,7 +1372,7 @@ static int viortc_restore(struct virtio_device *dev)
 		ret = viortc_populate_vq(viortc, alarm_viortc_vq,
 					 VIORTC_ALARMQ_BUF_CAP, false);
 		if (ret)
-			return ret;
+			goto err_remove;
 
 		notify = virtqueue_kick_prepare(alarm_vq);
 	}
@@ -1372,8 +1380,12 @@ static int viortc_restore(struct virtio_device *dev)
 	virtio_device_ready(dev);
 
 	if (notify && !virtqueue_notify(alarm_vq))
-		ret = -EIO;
+		return -EIO;
+
+	return 0;
 
+err_remove:
+	__viortc_remove(viortc);
 	return ret;
 }
 
-- 
2.34.1

  reply	other threads:[~2026-04-24  4:08 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-04-17  9:08 [PATCH] virtio: rtc: tear down old virtqueues before restore JiaJia
2026-04-23 18:02 ` Peter Hilber
2026-04-24  3:14   ` JiaJia [this message]
2026-04-24 13:54     ` [PATCH v2] " Peter Hilber
2026-04-25  7:08       ` [PATCH v3] " physicalmtea
2026-04-25 15:10         ` Michael S. Tsirkin
2026-04-25 16:09           ` Jia Jia
2026-04-25 17:03             ` Michael S. Tsirkin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=69eaeca0.050a0220.408e2.027b@mx.google.com \
    --to=physicalmtea@gmail.com \
    --cc=eperezma@redhat.com \
    --cc=jasowang@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mst@redhat.com \
    --cc=peter.hilber@oss.qualcomm.com \
    --cc=virtualization@lists.linux.dev \
    --cc=xuanzhuo@linux.alibaba.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox