See if anyone can divine any greater meaning in these events:

  * My machine, medusa, is a StrongARMed RiscPC with 34MB of RAM,
    labelled `RiscPC' on the front (as opposed to -[67]00).

  * My Ethernet card is an old podule Ether3, which has always worked 
    fine.  After a reboot a few days ago, it stopped working, claiming
    `eth0: can't identify podule bus width' (or similar), and whatever
    I did with kernel options, it absolutely refused to budge again.

    The salient details are probably these: with `debug' as a kernel
    parameter, I get the following output:

        0:        [0053:00A4] ANT Ethernet (...)
        ether3_probe: write8 [01:00], read8 [05:05]
        ether3_probe: write16 [0101], read16 [0105]
        eth0: unable to identify podule bus width

    When I added a hack to force it to select BUS_16 if it couldn't 
    identify the card type, I then got a whole load of gibberish about
    failing the RAM test, the contents of which I've lost, unfortunately.

  * This morning, upon rebooting the machine, the Ethernet card magically
    came back to life!  With nothing more than a reboot, the Ethernet
    card started identifying and working again under Linux.
  
  * Suspicious of the card, I swiped a handy EtherB NIC slot card, and
    tried that out.  Booted up -- same complaint about not being able 
    to identify the podule bus width, but with my hack in place, it
    went on to boot up and apparently function absolutely normally.

    I rebooted a few times, and intermittently it would tell me that
    it hadn't identified the podule bus width, and was defaulting to 16,
    or just `work', without a problem.  This shouldn't be intermittent,
    surely?  Is the udelay() not big enough on the read and writes?

  * Here's the ultimately weird bit.  With the now-decided-to-work
    Ether3 in place, the machine boots up, lets me log in as root and
    type `halt' and shutdown in about 1:35 (repeated over three test runs).
   
    With the EtherB in place, exactly the same procedure takes 1:55 on
    average.  Where the hell are those extra 20 seconds going?

-- 
Chris <[EMAIL PROTECTED]>                         ( http://www.fluff.org/chris )

unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]

Reply via email to