I'm getting some wierd errors on a Lance based eth card (they seem to
happen more under higher bandwidth usage, tho my average max usage is
probably 40-50k/sec, so its not what I would really consider 'loaded')
eth0 is connected to my internal LAN, and I am doing firewalling/masq
on the linux box. I am using modules for the eth cards (both) and a
source-build kernel customized to what I need. The problem doesnt seem to
be really that bothersome, but every once in a while I seem to lose
connectivity across the fiber, and if I check the logs i see one of these
errors right about the time when I lose connectivity, it could be
coincidental, or not. Messages in my syslog/dmesg are as follows:
eth1: Tx FIFO error! Status a2e2
eth1: Tx FIFO error! Status a2ea
I usually get a good pile of these (I think when im either ftping out or
someone is ftping from me, either way its Tx problem I think) in my logs,
and then:
eth1: transmit timed out, status 0062, resetting.
Ring data dump: dirty tx 105 cur_tx 121 (full) cur_rx 68.
and then a bunch of hexidecimal numbers in 12 columns. It is at this point
I believe I lose connectivity. A reboot seems to be the only thing that
fixes it.
The machine is a p5 100 with 64M ram, and isn't really very loaded at all
(averages 0.01) with an adaptec 2940UW SCSI controller and a seagate UW
hdd and 1 modem dialup (I have my home linux box dialed into it) and of
course the 2 eth cards (tulip is PCI, lance is ISA). This is running
RH6.0.
My setup is along the lines of:
Internal LAN
|
tulip(eth0)-[RH6box]-lance(eth1)
|
[Fiber Tranceiver]
|
1000 feet of multimode fiber
|
[Fiber Tranceiver]
|
ISP
Any ideas?
Thanks,
b. ellis
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
