transmit error causes network to hang

[richard@iguana.co.nz]

> Hi,
> I have an IBM 300GL with an Etherjet adapter (which is actually an intel
> etherexpress 100B).  I have had intermittent problems with the card.  I am
> running RedHat 6.1, right out of the box, no changes.  I downloaded,
> compiled, and installed the latest eepro100 module.  When the machine first
> comes up, the interface is ok.  I then try to use the network (mount a smb
> share or ftp a file) and the network will suddenly freeze.  I look at a
> netstat -in and see that the
> packets received (RX-OK) is no longer incrementing and that the transmit
> error (TX-ERR) column has incremented by one.   Bear with me, I dont know a
> lot about networking.  I just happened to notice that everytime the network
> hangs, this TX-ERR value has increased.  If I ifconfig the interface down
> and up, the adapter is reset and begins working again.  After a while,
> though, the same thing will happen again.  Is there an option I should
> configure for the kernel module?  I have since installed 3com 905B and that
> seems to be working ok.  I've tried monitoring the link with the mii-diag
> program but I dont seem to get any info that would give me a clue as to
> what is going on.
> 
> Any help would be greatly appreciated.

I suggest sticking with the 3com. There are numerous and ever-increasing
reports of problems with timeouts and similar on the etherexpress,
especially in late model cards, and the drivers under the current
development kernels. 

I have not seen a solid diagnosis or fix of the problem on this list, and
while I understand the point of Open Source is that one can fix problems
oneself, drivers for ethernet cards are deep voodoo and require a
significant investment in time in order to understand the operations, let
alone work out a fix for a problem that has obviously got those already
aquainted with the driver stumped. 

I recommend that you, and others in a similar position as users rather
than programmers, look at utilising other NICs. If you must use one of
these NICs, use the very latest driver for the card (To my knowledge, the
one with the least evidence of problems is version 1.09t of the eepro100.c
driver, available from
ftp://cesdis.gsfc.nasa.gov/pub/linux/drivers/test/eepro100.c

If your card is currently working correctly for you, don't upgrade :)
Given that nobody to my knowledge has solidly identified the problem,
upgrading could cause you problems.

As far as my experience has been thus far, 2.2.1* + the 1.09t driver above
has been the most stable.

Richard.

-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to linux-eepro100-request@beowulf.org