Rolf Fokkens
2010-Dec-19 22:26 UTC
[Bridge] Why should static MAC address match one of the port MAC addresses - bug?
Hi All, Despite of what Stephen explains, I'm having the problem Michiel describes as well. I have a bridge br2 containing 1 (but potentially more) tap interface for kvm. Windows 7 VM doesn't like the bridge mac address to change, so I try to make it fixed. ifconfig br2 hw ether 02:00:00:00:01:02 Interface br2 becomes instantly unreachable from the VM (via the tap interface). No arp response or anything else. So, Michiel's claim seems to be true: the bridge mac address should apparently be one of the bridge interfaces. But I don't see the logic, this seems like a bug to me. kernel is Fedora kernel-2.6.35.9-64.fc14.x86_64. Cheers, Rolf> > ------------------------------------------------------------------------ > > * /To/: Michel Lammertink <mlammertink at xxxxxxxxx > <mailto:mlammertink at DOMAIN.HIDDEN>> > * /Subject/: Re: Why should static MAC address match one of the > port MAC addresses. > * /From/: Stephen Hemminger <shemminger at xxxxxxxxxxxxxxxxxxxx > <mailto:shemminger at DOMAIN.HIDDEN>> > * /Date/: Tue, 12 Oct 2010 09:44:08 -0700 > * /Cc/: bridge at xxxxxxxxxxxxxxxxxxxxxxxxxx > <mailto:bridge at DOMAIN.HIDDEN> > * /In-reply-to/: <4CB46549.2070901 at xxxxxxxxx > <mailto:4CB46549.2070901 at DOMAIN.HIDDEN>> > * /Organization/: Linux Foundation > * /References/: <4CB46549.2070901 at xxxxxxxxx > <mailto:4CB46549.2070901 at DOMAIN.HIDDEN>> > > ------------------------------------------------------------------------ > On Tue, 12 Oct 2010 15:40:25 +0200 Michel Lammertink<mlammertink at xxxxxxxxx> wrote: > > >/ Hi all,/ > >/ / > >/ I've read many messages on this mailing list with the question about/ > >/ setting a static MAC address on the bridge interface, but two questions/ > >/ remain:/ > >/ / > >/ 1. Why should the MAC address of the bridge match the address of one of/ > >/ its interfaces? Setting the MAC address to a different value does not/ > >/ generate an error, but the bridge is not working properly (Local data/ > >/ does not arrive)./ > > No longer a requirement. It was just a good way to ensure a good > address that was unique, and not being used by other systems. > > >/ 2. Why does the bridge by default choose the "lowest" MAC address,/ > >/ instead of the MAC address of the first port added, as is proposed by/ > >/ the standard [1]?/ > > It was a convenience in original design to choose a consistent value. > Since ports can be added in any order, using the lowest seemed like > a good idea for the original author. This has been maintained for > backwards compatibility. > > >/ [1] IEEE802.1D, Par. 7.12.5./ > >/ Quote: "The Bridge Address may be the individual MAC Address of a/ > >/ Bridge Port, in which case, use of the address of the lowest numbered/ > >/ Bridge Port (Port 1) is recommended."/ > >/ / > >/ Thanks in advance,/ > >/ / > >/ Michel./-------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.linux-foundation.org/pipermail/bridge/attachments/20101219/0911409a/attachment.htm
Stephen Hemminger
2010-Dec-20 18:37 UTC
[Bridge] Why should static MAC address match one of the port MAC addresses - bug?
On Sun, 19 Dec 2010 23:26:22 +0100 Rolf Fokkens <rolf.fokkens at wanadoo.nl> wrote:> Hi All, > > Despite of what Stephen explains, I'm having the problem Michiel > describes as well. I have a bridge br2 containing 1 (but potentially > more) tap interface for kvm. Windows 7 VM doesn't like the bridge mac > address to change, so I try to make it fixed. > > ifconfig br2 hw ether 02:00:00:00:01:02 > > Interface br2 becomes instantly unreachable from the VM (via the tap > interface). No arp response or anything else. > > So, Michiel's claim seems to be true: the bridge mac address should > apparently be one of the bridge interfaces. But I don't see the logic, > this seems like a bug to me. > > kernel is Fedora kernel-2.6.35.9-64.fc14.x86_64. > > Cheers, > > RolfThe address you change the address, it is like you changed the MAC address of a card on the network. Windows ARP cache is no longer valid and has to time out.