[libvirt] [PATCH] nodedev: udev: Fix handling of wireless NIC

[Jim Meyering]

Dave Allan wrote:
> On Wed, May 26, 2010 at 03:55:44PM -0400, Cole Robinson wrote:
>> Wireless NICs were being ignored because we weren't correctly handling
>> device type. Fix this, as well as wireless NIC net subtype.
>
> Thanks for finding and fixing this.  I made a v2 patch with the change
> that since DEVTYPE == "wlan" is a definitive indication that the
> device is a network interface, that case should be with the other
> device types for which DEVTYPE is definitive.  v2 patch attached.
>
> Dave
>
>
>>From ccf89c57a6335c1c7c7e0b29ae09f4916b33622d Mon Sep 17 00:00:00 2001
> From: David Allan <dallan at redhat.com>
> Date: Thu, 27 May 2010 10:44:02 -0400
> Subject: [PATCH 1/1] v2 of Cole's wlan support
>
> * Incorporated Jim's feedback
>
> * Moved case of DEVTYPE == "wlan" up as it's definitive that we have a network interface.
>
> * Made comment more detailed about the wired case to explain better
>   how it differentiates between wired network interfaces and USB
>   devices.
...

>      int ret = -1;

Thanks Dave.

> +    const char *devtype = udev_device_get_devtype(device);
>      union _virNodeDevCapData *data = &def->caps->data;
>
> +    if (devtype && STREQ(devtype, "wlan")) {
> +        data->net.subtype = VIR_NODE_DEV_CAP_NET_80211;
> +    } else {
> +        data->net.subtype = VIR_NODE_DEV_CAP_NET_80203;
> +    }
> +
>      if (udevGetStringProperty(device,
>                                "INTERFACE",
>                                &data->net.ifname) == PROPERTY_ERROR) {
> @@ -1074,6 +1081,8 @@ static int udevGetDeviceType(struct udev_device *device,
>      int ret = 0;
>
>      devtype = udev_device_get_devtype(device);
> +    VIR_DEBUG("Found device type '%s' for device '%s'",
> +              devtype, udev_device_get_sysname(device));

Sorry I missed it the first time, but since devtype can be
NULL here, we have to avoid dereferencing it:
(it wouldn't cause trouble with glibc, but would segfault on others)

       VIR_DEBUG("Found device type '%s' for device '%s'",
                 NULLSTR(devtype), udev_device_get_sysname(device));

>      if (devtype != NULL && STREQ(devtype, "usb_device")) {
>          *type = VIR_NODE_DEV_CAP_USB_DEV;
> @@ -1105,17 +1114,25 @@ static int udevGetDeviceType(struct udev_device *device,
>          goto out;
>      }
>
> +    if (devtype != NULL && STREQ(devtype, "wlan")) {
> +        *type = VIR_NODE_DEV_CAP_NET;
> +        goto out;
> +    }
> +
>      if (udevGetUintProperty(device, "PCI_CLASS", &tmp, 16) == PROPERTY_FOUND) {
>          *type = VIR_NODE_DEV_CAP_PCI_DEV;
>          goto out;
>      }
>
> -    /* It does not appear that network interfaces set the device type
> -     * property. */
> -    if (devtype == NULL &&
> +    /* It does not appear that wired network interfaces set the
> +     * DEVTYPE property.  USB devices also have an INTERFACE property,
> +     * but they do set DEVTYPE, so if devtype is NULL and the
> +     * INTERFACE property exists, we have a network device. */
> +    if ((devtype == NULL) &&

I wouldn't bother to add parens above.

>          udevGetStringProperty(device,
>                                "INTERFACE",
>                                &tmp_string) == PROPERTY_FOUND) {
> +

IMHO, that added blank line detracts from readability.

>          VIR_FREE(tmp_string);
>          *type = VIR_NODE_DEV_CAP_NET;
>          goto out;