[VLAN]: translate IF_OPER_DORMANT to netif_dormant_on()
Stefan Rompf [Tue, 21 Mar 2006 01:11:41 +0000 (17:11 -0800)]
this patch adds support to the VLAN driver to translate IF_OPER_DORMANT of the
underlying device to netif_dormant_on(). Beside clean state forwarding, this
allows running independant userspace supplicants on both the real device and
the stacked VLAN. It depends on my RFC2863 patch.

Signed-off-by: Stefan Rompf <stefan@loplof.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/8021q/vlan.c

index fa76220..3948949 100644 (file)
@@ -69,7 +69,7 @@ static struct packet_type vlan_packet_type = {
 
 /* Bits of netdev state that are propagated from real device to virtual */
 #define VLAN_LINK_STATE_MASK \
-       ((1<<__LINK_STATE_PRESENT)|(1<<__LINK_STATE_NOCARRIER))
+       ((1<<__LINK_STATE_PRESENT)|(1<<__LINK_STATE_NOCARRIER)|(1<<__LINK_STATE_DORMANT))
 
 /* End of global variables definitions. */
 
@@ -344,6 +344,26 @@ static void vlan_setup(struct net_device *new_dev)
        new_dev->do_ioctl = vlan_dev_ioctl;
 }
 
+static void vlan_transfer_operstate(const struct net_device *dev, struct net_device *vlandev)
+{
+       /* Have to respect userspace enforced dormant state
+        * of real device, also must allow supplicant running
+        * on VLAN device
+        */
+       if (dev->operstate == IF_OPER_DORMANT)
+               netif_dormant_on(vlandev);
+       else
+               netif_dormant_off(vlandev);
+
+       if (netif_carrier_ok(dev)) {
+               if (!netif_carrier_ok(vlandev))
+                       netif_carrier_on(vlandev);
+       } else {
+               if (netif_carrier_ok(vlandev))
+                       netif_carrier_off(vlandev);
+       }
+}
+
 /*  Attach a VLAN device to a mac address (ie Ethernet Card).
  *  Returns the device that was created, or NULL if there was
  *  an error of some kind.
@@ -450,7 +470,7 @@ static struct net_device *register_vlan_device(const char *eth_IF_name,
        new_dev->flags = real_dev->flags;
        new_dev->flags &= ~IFF_UP;
 
-       new_dev->state = real_dev->state & VLAN_LINK_STATE_MASK;
+       new_dev->state = real_dev->state & ~(1<<__LINK_STATE_START);
 
        /* need 4 bytes for extra VLAN header info,
         * hope the underlying device can handle it.
@@ -498,6 +518,10 @@ static struct net_device *register_vlan_device(const char *eth_IF_name,
        if (register_netdevice(new_dev))
                goto out_free_newdev;
 
+       new_dev->iflink = real_dev->ifindex;
+       vlan_transfer_operstate(real_dev, new_dev);
+       linkwatch_fire_event(new_dev); /* _MUST_ call rfc2863_policy() */
+
        /* So, got the sucker initialized, now lets place
         * it into our local structure.
         */
@@ -573,25 +597,12 @@ static int vlan_device_event(struct notifier_block *unused, unsigned long event,
        switch (event) {
        case NETDEV_CHANGE:
                /* Propagate real device state to vlan devices */
-               flgs = dev->state & VLAN_LINK_STATE_MASK;
                for (i = 0; i < VLAN_GROUP_ARRAY_LEN; i++) {
                        vlandev = grp->vlan_devices[i];
                        if (!vlandev)
                                continue;
 
-                       if (netif_carrier_ok(dev)) {
-                               if (!netif_carrier_ok(vlandev))
-                                       netif_carrier_on(vlandev);
-                       } else {
-                               if (netif_carrier_ok(vlandev))
-                                       netif_carrier_off(vlandev);
-                       }
-
-                       if ((vlandev->state & VLAN_LINK_STATE_MASK) != flgs) {
-                               vlandev->state = (vlandev->state &~ VLAN_LINK_STATE_MASK) 
-                                       | flgs;
-                               netdev_state_change(vlandev);
-                       }
+                       vlan_transfer_operstate(dev, vlandev);
                }
                break;