[eepro100] Another wait_for_cmd_timeout sufferer...

[Andrew Reid]

  Greetings!

  I discovered this list through Google-searching on some 
error messages I've been getting, and in spite of the copious 
amounts of helpful advice in the archives, I still seem to be stuck.
I hope you'll forgive more traffic on this topic.

(Addendum: I also hope this isn't double-posted, I got some
subscription errors from the server on the 28th...)

  I have a Sony Vaio PCG-R505TS laptop computer, with built-in
Intel Pro/100 V ethernet, dual-boot between WindowsME and 
Debian Linux, (mostly "testing", for those who worry about
such things) and at various points, the 2.2.19 and 2.4.12 kernels.
On my home, 10BaseT-hub network, in with either kernel, and the
eepro100 driver bundled with the kernel source, I get a
lot of these command-wait time-outs, followed by a loss of network
functionality, which restores itself automatically after several
minutes, or can be restored sooner by restarting networking
(e.g. by "/etc/init.d/networking restart").  

  On 100BaseT networks, either at my workplace, or at home when
I connect directly to my server via cross-over cable, I have no
problems.  Also, (and this is galling) under WindowsME, there are
no problems even on 10BaseT.  So, I believe all the hardware
involved is OK.

  I have been able to set up the Intel driver under the 2.2.19
kernel, but that costs me ReiserFS support (for fast restarts,
important because the Sony's apm support is poor and it won't
sleep), and also the jog-shuttle device support (gives you 
a middle mouse-button, also important in X-windows.)
It's liveable, but it's not a good long-term solution.


  I had the same problem with missing symbols for the 
pci-scan.o and eepro100.o, somebody on the list here said
they solved it by hitting the books, but I can't find anything
relevant -- I compiled the eepro100 with "-I/usr/src/linux/include"
and "-I.", which I believe is correct (and was suggested on the
Scyld page), but this did not solve the problem.  Where is 
the M I am meant to be RTF-ing for this?  This was the 1.17b
version of the Scyld driver I was trying to build.


  Also, I saw a reference to using special options 
"-G 0 -w -w" to disable the sleep mode that is reported
by eepro100-diag, but to what command are these options?
Is this for pci-scan, eepro100, or eepro100-diag? 

 The Intel driver has its own problems under 2.4.12 -- 
it initially fails with a "__bad_udelay()" missing symbol,
but if you hack it up so that it never calls udelay(n) with
an argument n greater than 20000, that goes away.  Then it
claims "device not found".  (This is OT for this list, I
guess, but I'm trying to be complete.)


  The eepro100-diag program reports:

> eepro100-diag.c:v2.05 6/13/2001 Donald Becker (becker@scyld.com)
>  http://www.scyld.com/diag/index.html
> Index #1: Found a Intel i82562 Pro/100 V adapter at 0x3000.
> EEPROM contents, size 64x16:
>     00: 0008 1646 dbcc 1a03 0000 0201 4701 0000
>  0x08: 0000 0000 49a2 3013 8086 007f 0000 0000
>       ...
>   0x30: 002c 0000 0000 0000 0000 0000 0000 0000
>   0x38: 0000 0000 0000 4030 0000 0000 0000 2a85
>  The EEPROM checksum is correct.
> Intel EtherExpress Pro 10/100 EEPROM contents:
>   Station address 08:00:46:16:CC:DB.
>   Board assembly 000000-000, Physical connectors present: RJ45
>   Primary interface chip i82555 PHY #1.
>    Sleep mode is enabled.  This is not recommended.
>    Under high load the card may not respond to
>     PCI requests, and thus cause a master abort.

 
  So the obvious question is, can anyone point me in 
the right direction to getting *any* non-bundled module loaded
under 2.4.12?  And, in the spirit of collaboration, is there
anything I can investigate on my box that will help in building
a better driver?
  
                                -- A. Reid
                                   reidac@bellatlantic.net