[Bridge] NIC unicast macs table manipulation by bridge

[Bridge] NIC unicast macs table manipulation by bridge

[Or Gerlitz]

Using a Linux bridge I see that none of the interfaces is set to promiscuous
mode but on the other hand I don't see any dev_unicast_xxx calls in the
bridge code... for the case of one of the interfaces being a physical NIC
(e.g eth1), what's the magic that makes the NIC RX filtering be willing
to accept frames whos dest mac isn't the NIC "primary" unicast mac?

Also is there a method to display dev->uc_list (e.g similar to /proc/net/dev_mcast)?

Or.

Re: [Bridge] NIC unicast macs table manipulation by bridge

[Stephen Hemminger]

On Tue, 21 Apr 2009 14:19:00 +0300 (IDT)
Or Gerlitz <ogerlitz@voltaire.com> wrote:

> Using a Linux bridge I see that none of the interfaces is set to promiscuous
> mode but on the other hand I don't see any dev_unicast_xxx calls in the
> bridge code... for the case of one of the interfaces being a physical NIC
> (e.g eth1), what's the magic that makes the NIC RX filtering be willing
> to accept frames whos dest mac isn't the NIC "primary" unicast mac?
> 
> Also is there a method to display dev->uc_list (e.g similar to /proc/net/dev_mcast)?
> 
> Or.
Look closer...

/* called with RTNL */
int br_add_if(struct net_bridge *br, struct net_device *dev)
{
	struct net_bridge_port *p;
	int err = 0;

	if (dev->flags & IFF_LOOPBACK || dev->type != ARPHRD_ETHER)
		return -EINVAL;

	if (dev->netdev_ops->ndo_start_xmit == br_dev_xmit)
		return -ELOOP;

	if (dev->br_port != NULL)
		return -EBUSY;

	p = new_nbp(br, dev);
	if (IS_ERR(p))
		return PTR_ERR(p);

	err = dev_set_promiscuity(dev, 1);

This turns on promiscuous mode (and nests).

Re: [Bridge] NIC unicast macs table manipulation by bridge

[Or Gerlitz]

On Tue, 21 Apr 2009, Or Gerlitz wrote:

> Using a Linux bridge I see that none of the interfaces is set to promiscuous mode

OKay, sorry for that. I was looking (2.6.29) on the output of /sbin/ifconfig
which is installed on my system through net-tools-1.60-73 - seems like it
ignores the IFF_PROMISC bit. Now, when looking on the device sysfs flags,
I see now the IFF_PROMISC (0x100) bit being set.

It turns out that this ifconfig is even more misleading, using
$ ifconfig eth1 promisc and later -promisc makes the output to show as now the
device is in the promiscuous state, and with -promisc as it doesn't - but in both
cases, the IFF_PROMISC bit setting done by the bridge remains steady
that is both interfaces connected to the bridge have the 0x1103 flags bit mask...

As for my other question - is there a method to display dev->uc_list
(e.g similar to /proc/net/dev_mcast)?

thanks,

Or.