[eepro100] mdio_read() timed out

[Kevin Hansard]

Hi,

I have modified the driver to output the IO address and this shows (I think) that it is not a hardware error. Interface details are as follows:

eth0: OEM Intel PCI EtherExpress Pro100 at 0xe12dc000, 00:02:B3:B4:63:9A, IRQ 24.
eth1: OEM Intel PCI EtherExpress Pro100 at 0xe12de000, 00:02:B3:B4:63:9B, IRQ 25.
eth2: OEM Intel PCI EtherExpress Pro100 at 0xe12e0000, 00:02:B3:B4:63:A0, IRQ 48.
eth3: OEM Intel PCI EtherExpress Pro100 at 0xe12e2000, 00:02:B3:B4:63:A1, IRQ 49.
eth4: OEM Intel PCI EtherExpress Pro100 at 0xe12e4000, 00:02:B3:B4:63:8E, IRQ 56.
eth5: OEM Intel PCI EtherExpress Pro100 at 0xe12e6000, 00:02:B3:B4:63:8F, IRQ 57.

My latest errors are:

kernel: e12e0000: mdio_read() timed out with val = 08210000.
kernel: e12e0000: mdio_read() timed out with val = 08210000.
kernel: e12e0000: mdio_read() timed out with val = 08210000.
kernel: e12e4000: mdio_read() timed out with val = 08210000.
kernel: e12de000: mdio_read() timed out with val = 08210000.
kernel: e12dc000: mdio_read() timed out with val = 08210000.
kernel: e12dc000: mdio_read() timed out with val = 08210000.
kernel: e12de000: mdio_read() timed out with val = 08210000.
kernel: e12dc000: mdio_read() timed out with val = 08210000.
kernel: e12dc000: mdio_read() timed out with val = 08210000.
kernel: e12dc000: mdio_read() timed out with val = 08210000.
kernel: e12de000: mdio_read() timed out with val = 08210000.

So I have the time outs from eth2, eth4, eth1, and eth0. eth3 is an very lightly loaded interface with virtually no traffic on it which I guess is why we haven't seen a timeout on it yet.

I was hoping that it would be a hardware fault because that would be an easy fix.

Incidentally I did mess around with the boguscnt being set to 64*10000 and I did get the odd return (probably about 1 in 10)

kernel:  mdio_read() took 31316 ticks.
kernel:  mdio_read() took 7788 ticks.
kernel:  mdio_read() took 63163 ticks.
kernel:  mdio_read() took 506921 ticks.

Have you any more thoughts on what could be causing it. Could other kernel modules cause this error? The machine is running RainConnect from Rainfinity. This software load balances IP connections over multiple ISP's. The software installs its own kernel module and interfaces with your driver in some way (the software requires netdrivers-3.1 or later). When I stop RainConnect the errors don't seem to happen, though I haven't been able to stop it for very long. In addition with RC stopped the load on the machine is significantly less.
Thanks for all your help.
Kevin