[Bridge] STP with only 2 switches
iamatul at comcast.net
Thu Aug 11 19:08:47 PDT 2005
One thing which would be useful to see is the Address Resolution table
in the switch. Also, if you chipset supports TX/RX counters per port,
see if these are changing for port 1 & port 2.
Also, ping has an option to transmit X number of packets. (At least in
busybox version of ping ). So, if there is no OOPS message and kernel
and rest of the system is working, then its just that the ping is not timing
out. Maybe, try strace on ping to see which system call hangs.
Alternately, kernel lockmeter patch can tell you the time taken in
ticks for each system call.
This wont solve it but help identify the problem area. Check the
write system call in your driver. Alternately, try static ARP entries.
----- Original Message -----
From: "Stephen Hemminger" <shemminger at osdl.org>
To: <sameer.neugui at tatainfotech.com>
Cc: <bridge at lists.osdl.org>
Sent: Thursday, August 11, 2005 10:26 AM
Subject: Re: [Bridge] STP with only 2 switches
> On Thu, 11 Aug 2005 16:33:01 +0530
> "sameer" <sameer.neugui at tatainfotech.com> wrote:
>> Hi all,
>> We are facing a problem with STP implementation. We have only 2
>> switches. Thats the constraint, but we feel STP should work.
>> We have developed a layer 2 switch. This uses Cirrus logic's ep9301
>> processor, embedded Linux 2.4.21 kernel as the OS and zarlink's switch
>> The observation:
>> 1)we have 2 switches, switch A and switch B
>> 2)port 1 and port 2 of switch A are connected to port 1 and port 2 of
>> switch B
>> 3)1 PC is connected to port 0 of switch A and another to COM port of
>> switch A to observe STP
> PC-A ---0 switch A 1 ---- 1 switch B
> 2 ---- 2
>> 4)We restart the 2 switches and observe that one of the links is
>> blocked and the other in forwarding state. STP is always enabled on
> That is good.
>> 6)We start to ping switch A and the switches and the two PC hang.
> How are you setting up switch A? what are the equivalent bridge
> commands and ip configuration? Are you assigning IP to eth device
> or br device?
> Define hang? is network not responsive? system not responsive?
> Did one of the switches produce a OOPS on console?
>> 7)the LEDs on the switch device that signify the activity are seen
>> continuously glowing.
> You may have created a packet loop because of STP problem.
> Simple test is to just unplug the port 2 connection.
>> 8)switch A and B, PC 1 and 2 cannot be accessed at all.
>> 9) We remove one of the links and the STP reconfigures and the switch
>> A can be pinged.
>> 10)Most of the time we have observed, although not a confirmed
>> pattern, that if we remove the blocked link, the STP just disables
>> the port on switch A (we have not observed the counter part on the
>> switch B
>> 11)Neither the switches nor the PCs were on the rest of the company
>> Its not a hang exactly as upon some topology change the system is
>> back to working. I also checked on the internet. On cisco site I
>> found something called unicast flooding.
>> 1)Is this the reason for the system to blog down?
>> 2) How do I find which version of STP I am using? The person who has
>> originally implemented has quit.
>> what would be the fix or some pointer in this regard?
> You could start by using a newer kernel. Using 2.6 might be impossible
> but linux-2.4.21 is old. You might compare the current 2.4 net/bridge
> directory with your kernel source.
Bridge mailing list
Bridge at lists.osdl.org
More information about the Bridge