headers
Bart De Schuymer | 1 May 2003 17:48
Picon

[PATCH] change pkt_type to PACKET_HOST earlier

Hello Stephen,

When a frame arrives on a bridge port and the destination MAC address equals 
the bridge's address, then I believe the packet type should be set to 
PACKET_HOST immediately (this is standard behaviour of a network driver). 
Currently, if the address of the physical device belonging to the bridge port 
isn't the address used for the logical bridge, then the packet type will be 
PACKET_OTHERHOST.
The current code does the change to PACKET_HOST for this case rather late: in 
br_pass_frame_up(). What I want is to be able to filter on PACKET_HOST in the 
ebtables PREROUTING chain. The patch below enables this.
Note that when the destination address is that of a bridge port belonging to 
another logical bridge on the box, the patch below won't change the packet 
type this early. Also note that brouted frames are not affected by this 
patch.

cheers,
Bart

--- linux-2.5.68/net/bridge/br_input.c.old	Thu May  1 17:09:06 2003
+++ linux-2.5.68/net/bridge/br_input.c	Thu May  1 17:22:24 2003
 <at>  <at>  -143,6 +143,9  <at>  <at>  int br_handle_frame(struct sk_buff *skb)
 			return -1;
 		}

+		if (!memcmp(p->br->dev.dev_addr, dest, ETH_ALEN))
+			skb->pkt_type = PACKET_HOST;
+
 		NF_HOOK(PF_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
 			br_handle_frame_finish);
(Continue reading)

Permalink | Reply | 0 comments |
headers
Stephen Hemminger | 1 May 2003 18:26
X-Face

Re: etherboot accross a bridge.

On Wed, 30 Apr 2003 06:19:54 -0400
david stone <polygons <at> ispwest.com> wrote:

> Alright here's my little setup:
> 
> One P90 with 3 100Mb/sec NICs and 3 10Mb/sec NIC. Most of my computers 
> on my little network are connected with this little switch that I made 
> with parts laying around. It appears to work just fine and I easily 
> upgraded my 10Mb/sec hub with it.
> 
> Here's my problem:
> 
> I'm trying to use Etherboot (http://www.etherboot.org) to setup a few 
> diskless clients. Everything works fine except that the thin clients 
> take close to an hour loading the kernel. This normaly takes a few 
> seconds. After the kernel loads, however, the network operates at normal 
> speeds. The first thing I tried was a new NIC; I replaced the RTL8029 
> with a 3com 509 NIC.  Amazingly this fixed the problem and makes another 
> one because all my systems but one use the RTL8029. I mailed the people 
> on the etherboot mailing list and they say the RTL8029 works fine with 
> etherboot and something's probably wrong with my network setup. So I 
> replaced the switch I built with a little hub to test see if it was the 
> problem. When I did this the RTL8029 worked like a charm. So what it 
> boils down to is that something in my switch is not right. The only 
> advice I can give as to why it behaves like it does is that the 
> etherboot driver is pool driven while the kernel driver is interrupt 
> driven. I mainly want to know if etherboot is known to work over a bridge.

Could you get a trace with tcpdump of good vs. bad etherboot?
There might be timing issues where initial packets are getting dropped.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Luke Gorrie | 2 May 2003 15:13
Picon

Re: Question about turning STP off.

"Kirkendall, Charles" <CharlesK <at> aiinet.com> writes:

> I need to completely disable spanning tree within the bridge driver,
> because we are using a Daemon for handling spanning tree.  I have
> come up with the following modifications and want to know if these
> will have any adverse affects.

I too want to entirely disable spanning tree. Does anyone have
feedback about the Goodness of Charles's patch?

Cheers,
Luke
Permalink | Reply | 0 comments |
headers
Anurag Aggarwal, Noida | 2 May 2003 15:23

RE: Question about turning STP off.

Correct me if I am wrong but doesn't the brctl utility (bridge-utils
package) allows one to turn STP on/off.
The syntax is brctl stp <bridge> <state>. 
Hope it helps.

-----Original Message-----
From: Luke Gorrie [mailto:luke <at> member.fsf.org]
Sent: 02 May, 2003 6:43 PM
To: Kirkendall, Charles
Cc: 'bridge <at> math.leidenuniv.nl'
Subject: Re: [Bridge] Question about turning STP off.

"Kirkendall, Charles" <CharlesK <at> aiinet.com> writes:

> I need to completely disable spanning tree within the bridge driver,
> because we are using a Daemon for handling spanning tree.  I have
> come up with the following modifications and want to know if these
> will have any adverse affects.

I too want to entirely disable spanning tree. Does anyone have
feedback about the Goodness of Charles's patch?

Cheers,
Luke

_______________________________________________
Bridge mailing list
Bridge <at> math.leidenuniv.nl
http://www.math.leidenuniv.nl/mailman/listinfo/bridge
(Continue reading)

Permalink | Reply | 0 comments |
headers
Lucas Aimaretto | 3 May 2003 05:45
Picon

bridge and router

Hi there,

this is the question: is it posible to implement a bridge on a linux
that switches ethernet frames but also, in the same box, a router,
lets say, statics routes with iptables, so it wil NAT all connectiones
to the internet?

If so, please I'll be glad to hear on how to do it. If it helps, I
have SuSE 7.2, kernel 2.4.20 and iptables 1.2.7a.

Are there any kernel pathces that need to be added to the kernel?

Thanks

regards

lucas
Permalink | Reply | 1 comment |
headers
Nick Fedchik | 4 May 2003 12:07

Re: bridge and router

Hello Lucas Aimaretto,
Saturday, May 3, 2003, 6:45:22, you wrote:
LA> Hi there,

LA> this is the question: is it posible to implement a bridge on a linux
LA> that switches ethernet frames but also, in the same box, a router,
LA> lets say, statics routes with iptables, so it wil NAT all connectiones
LA> to the internet?

LA> If so, please I'll be glad to hear on how to do it.
http://ebtables.sourceforge.net/br_fw_ia/br_fw_ia.html#section3

LA> If it helps, I
LA> have SuSE 7.2, kernel 2.4.20 and iptables 1.2.7a.

LA> Are there any kernel pathces that need to be added to the kernel?
http://ebtables.sourceforge.net/
http://ebtables.sourceforge.net/ebtables-faq.html

--

-- 
Best regards,
Nick Fedchik FNM3-RIPE(-UANIC)
R.E.D Team Leader/DIS/UkrSat ISP, Kiev, Ukraine
Permalink | Reply | 0 comments |
headers
Dave Lee | 5 May 2003 15:48
Picon
Favicon

Obtaining the bridge address

I have a question regarding remotely obtaining the
bridge address of a bridge by a directly attached
host.  If my understanding of the 802.1d specs is
correct, I should be able to send (fake) a BPDU
Config/topology change notification to try to tease
out a response from the bridge I'm attached to (or
when the master updates the spanning tree.  But from
what I understand/recall of STP, it's not clear to me
that I'll get a response unless there is a topology
change elsewhere.  So this may not work.

Is this true for both 802.3 and 802.11?  Is there
another way to always determine what the bridge
address is (should I be looking at LLC mechanisms)? 
Basically, I want to find out the MAC address (or
other unique identifier) for the device that I'm
plugged into on the network over both 802.3 and
802.11.

Thanks,
Dave

__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com
Permalink | Reply | 1 comment |
headers
haden | 6 May 2003 10:25
Picon

ip_conntrack and etables

Hi,

I've got major performance problems with etables on 2.4.20 kernel, it seems
(I've found one mail) that ip_conntrack couses problems.
I wonder how can this be fixed w/o removing ip_conntrack?

Thanks.
Permalink | Reply | 1 comment |
headers
Bart De Schuymer | 6 May 2003 18:08
Picon

Re: ip_conntrack and etables

On Tuesday 06 May 2003 10:25, haden <at> homelan.no-ip.org wrote:
> Hi,
>
> I've got major performance problems with etables on 2.4.20 kernel, it seems
> (I've found one mail) that ip_conntrack couses problems.
> I wonder how can this be fixed w/o removing ip_conntrack?

I would be surprized that ebtables gave you performance problems, more likely 
the bridge-nf code and the netfilter/iptables stuff. There are no performance 
test results under different loads known to me, so I can only guess.
Since the bridge ports are in promiscuous mode and the PREROUTING chains (and 
the connection tracking code) are traversed before the bridging decision, 
they see all packets. What you could try is DROP all traffic of which you 
know it won't be bridged, in the PREROUTING chain of ebtables.
What would be interesting is some results saying in which situations the 
performance is bad and in which it isn't.

--

-- 
cheers,
Bart
Permalink | Reply | 0 comments |
headers
liang jian | 7 May 2003 11:21
Picon
Favicon

ebtables question

 I use kernel 2.4.19 with ebtables and bridge-nf patch.

brctl addbr br0

brctl stp br0 off

brctl addif br0 eth0

ifconfig br0 0.0.0.0 up

ifconfig eth0 10.0.1.254 netmask 255.255.255.0 broadcast 10.0.1.255

ebtables -t broute -A BROUTING -p ipv4 -i eth0 --ip-dst 10.0.1.254 -j DROP

ebtables -t broute -A BROUTING -p arp -i eth0 -d MAC_OF_ETH0 -j DROP

10.0.1.23 ----->| 10.0.1.254  bridging Firewall  66.66.66.66 | -------->66.66.66.88

10.0.1.23's gateway with 10.0.1.254.

then from 10.0.1.23 ping 66.66.66.88 I have :

at br0 noting.

at eth0 have arp request and reply,but I don't ping 66.66.66.88.why??

 

if I br0 with eth0 and eth1 ports,and no ip address with them,10.0.1.23 ip address change
to 66.66.66.23 and gateway port to 66.66.66.88, 66.66.66.23 don't ping 66.66.66.88 too ?

In the case rules:

ebtables -t b route -A BROUTING -i eth0 -p ipv4 -j DROP

ebtables -t broute -A BROUTING -i eth1 -p ipv4 -j DROP

ebtables -t broute -A BROUTING -i eth0 -p arp -j DROP

ebtables -t broute -A BROUTING -i eth1 -p arp -j DROP



Do You Yahoo!?
"相见不如聊天!不出门一样面对面！网络摄像头对对派送中~赶快用你的雅虎电邮帐号参与吧…… 
_______________________________________________
Bridge mailing list
Bridge <at> math.leidenuniv.nl
http://www.math.leidenuniv.nl/mailman/listinfo/bridge
Permalink | Reply | 0 comments |

Gmane