Re: [Qemu-devel] [PATCH v2 2/8] virtio: enhance virtio_error messages

[Ladi Prosek <lprosek@redhat.com>]

On Thu, Jul 13, 2017 at 3:40 PM, Stefan Hajnoczi <stefanha@redhat.com> wrote:
> On Thu, Jul 13, 2017 at 01:02:31PM +0200, Ladi Prosek wrote:
>> +static const char *virtio_get_device_id(VirtIODevice *vdev)
>> +{
>> +    DeviceState *qdev = DEVICE(vdev);
>> +    while (qdev) {
>> +        /* Find the proxy object corresponding to the vdev backend */
>> +        Object *prop = object_property_get_link(OBJECT(qdev),
>> +                                                VIRTIO_PROP_BACKEND, NULL);
>> +        if (prop == OBJECT(vdev)) {
>> +            return qdev->id;
>> +        }
>> +        qdev = qdev->parent_bus->parent;
>> +    }
>> +    return NULL;
>> +}
>> +
>>  void GCC_FMT_ATTR(2, 3) virtio_error(VirtIODevice *vdev, const char *fmt, ...)
>>  {
>>      va_list ap;
>>
>> +    error_report_nolf("%s (id=%s): ", vdev->name, virtio_get_device_id(vdev));
>
> virtio_get_device_id() can return NULL.  POSIX does not guarantee that
> the printf(3) family functions handle "%s", NULL safely.  glibc prints
> "(null)" but other libc implementations crash (e.g. Solaris).
>
> http://pubs.opengroup.org/onlinepubs/9699919799/functions/fprintf.html
>
> Should the return NULL above have g_assert_not_reached()?  That would
> communicate the assumption that we never reach return NULL and it might
> silence static checkers like Coverity but I'm not sure.

virtio_get_device_id is expected to return NULL if the device has no
id assigned and I kind of liked the "(null)" output. I just failed to
realize that not all printf's will handle it. I'll definitely fix
this, thanks!