[Qemu-devel] [PATCH] virtio-scsi: Unset hotplug handler when unrealize

Fam Zheng posted 1 patch 6 years, 11 months ago
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20170518102808.30046-1-famz@redhat.com
Test checkpatch passed
Test docker passed
Test s390x passed
hw/scsi/virtio-scsi.c | 3 +++
1 file changed, 3 insertions(+)
[Qemu-devel] [PATCH] virtio-scsi: Unset hotplug handler when unrealize
Posted by Fam Zheng 6 years, 11 months ago
This matches the qbus_set_hotplug_handler in realize and is important to
release the final reference to the embedded VirtIODevice so that it is
properly finalized.

A use-after-free is fixed with this patch, indirectly:
virtio_device_instance_finalize wasn't called at hot-unplug, and the
vdev->listener would be a dangling pointer in the global and the per
address space listener list. See also RHBZ 1449031.

Thanks to Paolo for explaining the reference counting.

Cc: qemu-stable@nongnu.org
Signed-off-by: Fam Zheng <famz@redhat.com>

---

Like suggested by Paolo, it seems to make sense to do this clean up in
bus_unparent, but given how confused I got when working on this bug, I'm
not confident in my reasoning about the ref count balance in other
callers.
---
 hw/scsi/virtio-scsi.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/hw/scsi/virtio-scsi.c b/hw/scsi/virtio-scsi.c
index 46a3e3f..f46f06d 100644
--- a/hw/scsi/virtio-scsi.c
+++ b/hw/scsi/virtio-scsi.c
@@ -918,6 +918,9 @@ void virtio_scsi_common_unrealize(DeviceState *dev, Error **errp)
 
 static void virtio_scsi_device_unrealize(DeviceState *dev, Error **errp)
 {
+    VirtIOSCSI *s = VIRTIO_SCSI(dev);
+
+    qbus_set_hotplug_handler(BUS(&s->bus), NULL, &error_abort);
     virtio_scsi_common_unrealize(dev, errp);
 }
 
-- 
2.9.3


Re: [Qemu-devel] [PATCH] virtio-scsi: Unset hotplug handler when unrealize
Posted by Paolo Bonzini 6 years, 11 months ago

On 18/05/2017 12:28, Fam Zheng wrote:
> Like suggested by Paolo, it seems to make sense to do this clean up in
> bus_unparent, but given how confused I got when working on this bug, I'm
> not confident in my reasoning about the ref count balance in other
> callers.

Queued this for now.

Paolo