Very Computer

Have the 3c507 driver compiled into the kernel.  On boot, card is
detected correctly at IRQ 10, Mem 0300, etc.  With the card's native
state alterred by its DOS configurator app so that ZeroWaitStates are
disabled and Turbo mode is set to Standard, dhcpcd now communicates with
our dhcp server [when it before did not] and, viola, we have a working
network.  Sorta.  What happens is that nominal net traffic produces no
problems.  If, however, I run a TCP/IP app (same results whether
Netscape inside X or ncftp at the command prompt, for example),
approximately 50-60 seconds after starting up and while the app is
sending/retrieving data, this starts:

eth0: Command unit stopped, status xxxx, restarting -OR-
eth0: Rx unit stopped, status xxxx, restarting.

xxxx is replaced by a variety of four-digit numbers: 0000, a000, a040,
1220, 4040,  5020, and 5220, with 4040 being by far the most numerous
entry.  These stopped/restarting messages are repeated hundreds of time
(yes, hundreds, as /var/log/messages confirms: the last command was
repeated 696 [or 739[ times). These error messages will actually break
into the console as well as get listed in /var/log/messages.

Anyone have a clue what's going on?

My thanks,

Guy Stalnaker
--
*-------------------------------------------------------------*
                        J. Guy Stalnaker                      

 1210 W Dayton St Rm 4212                     wk. 608.263.8035
 Madison WI 53706                             fax 608.263.3846
*-------------------------------------------------------------*

Have the 3c507 driver compiled into the kernel.  On boot, card is
detected correctly at IRQ 10, Mem 0300, etc.  With the card's native
state alterred by its DOS configurator app so that ZeroWaitStates are
disabled and Turbo mode is set to Standard, dhcpcd now communicates with
our dhcp server [when it before did not] and, viola, we have a working
network.  Sorta.  What happens is that nominal net traffic produces no
problems.  If, however, I run a TCP/IP app (same results whether
Netscape inside X or ncftp at the command prompt, for example),
approximately 50-60 seconds after starting up and while the app is
sending/retrieving data, this starts:

eth0: Command unit stopped, status xxxx, restarting -OR-
eth0: Rx unit stopped, status xxxx, restarting.

xxxx is replaced by a variety of four-digit numbers: 0000, a000, a040,
1220, 4040,  5020, and 5220, with 4040 being by far the most numerous
entry.  These stopped/restarting messages are repeated hundreds of times
(yes, hundreds, as /var/log/messages confirms: the last command was
repeated 696 [or 739] times). These error messages will actually break
into the console as well as get listed in /var/log/messages.

Anyone have a clue what's going on?

My thanks,

Guy Stalnaker
--
*-------------------------------------------------------------*
                        J. Guy Stalnaker                      

 1210 W Dayton St Rm 4212                     wk. 608.263.8035
 Madison WI 53706                             fax 608.263.3846
*-------------------------------------------------------------*

Quote:>Here's the skinny:

>Have the 3c507 driver compiled into the kernel.  On boot, card is
>detected correctly at IRQ 10, Mem 0300, etc.  With the card's native

Quote:>state alterred by its DOS configurator app so that ZeroWaitStates are
>disabled and Turbo mode is set to Standard, dhcpcd now communicates with
>our dhcp server [when it before did not] and, viola, we have a working
>network.  Sorta.  What happens is that nominal net traffic produces no
>problems.  If, however, I run a TCP/IP app (same results whether
>Netscape inside X or ncftp at the command prompt, for example),
>approximately 50-60 seconds after starting up and while the app is
>sending/retrieving data, this starts:

>eth0: Command unit stopped, status xxxx, restarting -OR-
>eth0: Rx unit stopped, status xxxx, restarting.
[...]
>Anyone have a clue what's going on?

          4. Performance Tips

             4.1 General Concepts
*            4.2 ISA Bus Speed
*            4.3 Setting the TCP Rx Window
             4.4 Increasing NFS performance

       4.3 Setting the TCP Rx Window
       [...]

        Recent versions of the route command have the ability to set
        the size of this window on the fly.  Usually it is only for
        the local net that this window must be reduced, as computers
        that are behind a couple of routers or gateways are `buffered'
        enough to not pose a problem.  An example usage would be:

        ______________________________________________________________
                route add <whatever> ... window <win_size>
        ______________________________________________________________

        where win_size is the size of the window you wish to use (in
        bytes).  An 8 bit 3c503 card on an ISA bus operating at a
        speed of 8MHz or less would work well with a window size of
        about 4kB. Too large a window will cause overruns and dropped
        packets, and a drastic reduction in ethernet throughput.  You
        can check the operating status by doing a cat /proc/net/dev
        which will display any dropped or overrun conditions that
        occurred.

Mvh Vidar Andresen