| Message ID | 1403534015-28496-1-git-send-email-mbizon@freebox.fr |
|---|---|
| State | Rejected, archived |
| Delegated to: | David Miller |
| Headers | show |
From: Maxime Bizon <mbizon@freebox.fr> Date: Mon, 23 Jun 2014 16:33:35 +0200 > The bridge code only set PACKET_HOST on packets sent to the bridge mac > address, packets sent to other local mac addresses are sent to upper > layers, but ignored because they are marked as PACKET_OTHERHOST. > > Signed-off-by: Maxime Bizon <mbizon@freebox.fr> This looks fine to me, Stephen and other bridge hackers please review. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, 23 Jun 2014 16:33:35 +0200 Maxime Bizon <mbizon@freebox.fr> wrote: > The bridge code only set PACKET_HOST on packets sent to the bridge mac > address, packets sent to other local mac addresses are sent to upper > layers, but ignored because they are marked as PACKET_OTHERHOST. > > Signed-off-by: Maxime Bizon <mbizon@freebox.fr> What are you trying to do? Bridge device itself only has one MAC address. Are you trying to run macvlan on top of bridge? We only respond to ARP with bridge's MAC address. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, 2014-06-25 at 16:13 -0700, Stephen Hemminger wrote: > > The bridge code only set PACKET_HOST on packets sent to the bridge mac > > address, packets sent to other local mac addresses are sent to upper > > layers, but ignored because they are marked as PACKET_OTHERHOST. > > > > Signed-off-by: Maxime Bizon <mbizon@freebox.fr> > > What are you trying to do? Bridge device itself only has one MAC address. > Are you trying to run macvlan on top of bridge? > We only respond to ARP with bridge's MAC address. host1 uses a bridge with two interfaces: ethernet and wireless (AP mode using hostapd) for some reasons, hostapd kept adding and removing the interface from the bridge, triggering a mac address of the bridge change each time (I did not notice that). Another host had a static arp entry for host1, using the ethernet mac address, so its packets to host1 were randomly trashed depending on the current mac of the bridge. I have since fixed this by setting a static mac address on the bridge to prevent dynamic change, but I thought this patch would prevent other people from having the same bug in the future. What do you think ?
From: Maxime Bizon <mbizon@freebox.fr> Date: Thu, 26 Jun 2014 13:15:52 +0200 > for some reasons, hostapd kept adding and removing the interface from > the bridge, triggering a mac address of the bridge change each time (I > did not notice that). I think this is what you need to investigate rather than making changes to the bridge layer. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thursday 26 Jun 2014 à 13:35:50 (-0700), David Miller wrote: > > for some reasons, hostapd kept adding and removing the interface from > > the bridge, triggering a mac address of the bridge change each time (I > > did not notice that). > > I think this is what you need to investigate rather than making > changes to the bridge layer. Of course I fixed the problem with hostapd, I was just explaining how I noticed that behaviour in the bridge code. If that behaviour is intended then just forget that patch.
diff --git a/net/bridge/br_input.c b/net/bridge/br_input.c index 366c436..b9cc8554 100644 --- a/net/bridge/br_input.c +++ b/net/bridge/br_input.c @@ -120,6 +120,7 @@ int br_handle_frame_finish(struct sk_buff *skb) dst->is_local) { skb2 = skb; /* Do not forward the packet since it's local. */ + skb2->pkt_type = PACKET_HOST; skb = NULL; }
The bridge code only set PACKET_HOST on packets sent to the bridge mac address, packets sent to other local mac addresses are sent to upper layers, but ignored because they are marked as PACKET_OTHERHOST. Signed-off-by: Maxime Bizon <mbizon@freebox.fr> --- net/bridge/br_input.c | 1 + 1 file changed, 1 insertion(+)