0.99H vs. 0.99E on a Boomerang

[Andy Alsup]

I tried something over the weekend. While reading through a few lists,
I noticed a message from Sysuh in Korea. He said he had no problems
with kernel 2.0.31 and 3c59x.c v.031. But had problems with 2.0.36
3c59x.c v0.99H

I decided to try 2.0.31 with 3c59x.c ver 0.99H. It gave me some weird
messgaes about FIFOs the first boot, but after that it has been
running beatifully.

I used to be able to consistantly reproduce the problem in kernel
2.0.0pre4 3c59x.c v0.99H by sending a few (1500) ping packets with a
size of 38400. (Just something to make it think a little harder)

I have left this 2.0.31 v0.99H machine routing packets all weekend.
In one interface, out the other and back. Same thing: ping -s 38400 
--- it is still going --- 57,600 packets later!

I'm not a kernel head, so this doesn't mean much to me except that it
might just be something with the newer kernels rather than the
driver.

Hope that helps someone.

>>> "Andy Alsup" <AALSUP@bju.edu> 01/14 4:17 PM >>>
I also had similar problems with a Cyclone where it would work
beautifully for a few days (or hours under load), but would
suddenly,
without any error messages, just stop talking.

I have two Cyclones in one machine. Both have the same problem at
the
same time.

Traffic was as Rob described, sporadic, and comes in bursts. (telnet
is real fun when you can't see what you're typing)

After a soft reboot, it was business as usual.

I have tried both Full-Duplex and Half-duplex. Both with same
results. 

Incidently, I have tried all the recent releases of the driver E, F,
G, H to see if this went away and it seems to occur in all of
them---on both 2.1 and 2.0 versions of the kernel.

Also, when this happens, my 3c509 on a coax segment displays the
same
symptoms. Hmmm...

I have noticed seemingly consistent differences in the time it takes
to fail between revisions of the same model card. Firmware revision?

I have also set debug to a high value and got a lot of neat stuff,
but nothing that meant failure to me.

Anyone have anything to add?



>>> Rob Clark <clarkc2@rpi.edu> 01/10 5:00 PM >>>

Just to report my experience with this darn card so far:
0.99E: causes a skb_put kernel panic after about 3 weeks
0.99F & 0.99G: never got them to work right.
0.99H: seems okay for a while, then after 3 days suddenly loses its
mind
and acts like
       name resolution has failed.  Pings come through in bursts.

I'm sorry I can't be more precise.  I tried bumping up the debug
level, but
that would be
a lot of output over three days, just at debug level = 3.

Also (successfully) using a few 905B Vortex's with 0.99E and the
options set.

3Com SuperStack II hub
 
3c59x.c:v0.99H 11/17/98 Donald Becker
http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html 
eth0: 3Com 3c905 Boomerang 100baseTx at 0xfcc0,  00:60:08:31:09:aa,
IRQ 11
8K word-wide RAM 3:5 Rx:Tx split, autoselect/MII interface.
MII transceiver found at address 24, status 786f.
Enabling bus-master transmits and whole-frame receives.
eth0: MII #24 status 786f, link partner capability 40a1, setting
half-duplex.