Bart and Gavin, I am using Becker's drivers which are part of standard Linux distribution 2.4.20 downloaded from kernel.org. I am seeing the same symptoms with Linux 2.4.21 also.
The test program I am using generates sustained traffic of about 70 MB per TCP connection. As a result the generated traffic fully stresses the queues as opposed to burtsy traffic where queues have a chance of recovery. Additionally this is being done on a private network with few machines so the bridge internal IP table is not strained at all. I am planning to try out the standard Intel drivers this wekend and will let you know my observations. We can continue this discussion on private e-mail. Thanks Rajesh [EMAIL PROTECTED] On Saturday 19 July 2003 19:45, Rajesh Srivastava wrote: > I am using Intel EtherExpress and Intel EtherPro with standard 2.4.20 > drivers. Is that eepro100 (Becker's driver) or e100 (Intel's driver)? I'm willing to try determine the exact cause and fix it, if you have time for trying patches. If so, we can continue this discussion in private mail. > What is surprising here is that ifconfig shows that there are no dropped > packets, no collisions etc. on the hardware. > > Also total traffic on br0 is less than 1% then that of eth0, eth1,and eth3 > - is this normal. All bridged frames are accounted for ethx. Only traffic with the bridge itself is accounted for br0. Traffic to the bridge is counted twice, once for ethx and once for br0. cheers, Bart Rajesh _______________________________________________ Bridge mailing list [EMAIL PROTECTED] http://www.math.leidenuniv.nl/mailman/listinfo/bridge
