Re: Linux 2.3.18ac8

Richard Dynes (rdynes@sun1.varcom.com)
Fri, 24 Sep 1999 10:22:22 -0400

Arnaldo Carvalho de Melo wrote:
>
> On 23 Sep 1999, Oystein Viggen wrote:
>
> > Alan Cox wrote:
> >
> > > Hopefully we have the network scan done this time.
> > >
> > >
> > > 2.3.18ac8
> > > o Hpefully finally kill the pci scan bugs (Petr Vandrovec)
> >
> > Pci scan worked for me (NIC was detected), but the network-driver
> > (3c59x on 3c905B) still won't work.
> >
> > This is written on 2.3.18ac8 w/3c59x.c from 2.3.18 plain. This
> > configuration seems to work well. (It booted fine, sound, networking
> > and X are working - What more could a man wish for? :)
>
> same problem here, same fix (3c59x.c from 2.3.18 plain). 8)
>

I may have the same / related problem here:

H/W Configuration:

Ziatech 5521 with two on-board Tulip 21143's, Tx interface. eth0-eth1
Znyx 414 or something: quad Tulip 21143, Tx (copper) interface.
eth2-eth5
Osicom dual Tulip (not sure what kind) 100Base Fx. eth6-eth7

Previously:

Linux 2.3.11, with tulip.c:v0.91 4/14/99 as a module

I was able to use the on-board tulip interfaces without any problems:
ifconfig up down, all around. No issues. The quad board I had to be
less free with.

I wasn't able to get the Fx board going easily, but I didn't spend
much time on it.

The tulip driver _would_ recognize all 8 tulip chips, and assign
logical names to them, eg eth0 - eth7.
Currently:

Linux 2.3.18ac8 ==> tulip.c:v0.91m 9/15/99, as a module

Currently:

Linux 2.3.18ac8 ==> tulip.c:v0.91m 9/15/99, as a module

Problem:

I cannot use eth0- I get the following repeating error:

> Sep 23 11:54:45 probe2 kernel: eth0: Tx hung, 7 vs. 0.

When I take the _same_ cable, and move it over to the (also on-board)
eth1 port, things work just fine, I can ifconfig the interface, telnet
to it, etc. Everything appears as normal.

I've never had the experience of the two port behaving differently:
they're the same chips, same interfaces, on the same processor.

So I got onto the linux-tulip list, and exchanged a bunch of mail with
Donald Becker, doing tests, etc.

Here a summary of the results:

I suspect (with Donald's help):

- that there's an irq mapping problem (/proc/interrupts isn't right)
- that some but not all of the interrupts from any ethernet device
are being accounted for under a single interrupt.
- that we receive packets on eth0, because they are dropped.
- that the packets on eth0 are not processed, because they are
dropped rather than received (eg promisc mode).
- that we get Hung tx errors on eth0.
- that two ports, eth6 and eth7, while visible via lspci and
tulip_diag, are not recognized (via tulip_probe?), and so the
logical
devices are not created.

Of these, all of them point at an irq mapping or pci scan problem,
except possibly the last, which may be another issue, since it's a
different manufacturer and media (FSC). I'm getting more data on this
board to work this one.

I'm open to suggestions.

-Richard

-- 
    Richard Dynes
    rdynes@varcom.com


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/