Re: Using PXE+GRUB to boot Linux on a Thinkpad T20

[Donald Becker]

On 29 Jul 2002, Patrick J. LoPresti wrote:

> Donald Becker <address@hidden> writes:
> > On 29 Jul 2002, Patrick J. LoPresti wrote:
> > >     00:03.0 Class 0200: 10b7:6056 (rev 20)
> > 
> > What is the subsystem ID?
> > (The "-5" version has a subsystem ID of 0x655610b7.)
> 
> Hm, how would I determine that?  Or is the subsystem ID just the
> concatenation of the device and vendor IDs?  That is, did you mean
> "0x605610b7"?  Because that is definitely the device; see the probe
> messages from the updated 3c59x driver below.

No, the subsystem ID is different than the primary PCI ID.  For 3Com
cards the vendor portion is the same, 0x10b7, but the device section
is used to distinguish the implementation.

There was a specific reason I asked about the subsystem: the "-5"
version requires setting an undocumented register.  That code was added
in 0.99Xc and is not in the standard 99X release. 

> I have tried the new scyld.com driver.  It's better!  Or at least, it
> fails differently :-).  It prints the banner with no warnings:
> 
>   eth0: 3c59x.c:v0.99X 6/21/2002 Donald Becker, address@hidden
>   eth0: 3Com 3c1556B mini-PCI at 0x1400, e0:00:e0:00:e0:00, IRQ 11

That indicates that the 2.4 kernel was not properly activating the
interface.  The pci-scan support is needed to activate the card.

> And it no longer complains about not finding any MII transceivers.
> But that MAC address still looks odd...

The bogus MAC address is because the EEPROM isn't being read correctly.
This is very likely a "-5" interface.

Try the 0.99Xd test driver at
      ftp://ftp.scyld.com/pub/network/test/3c59x.c

-- 
Donald Becker                           address@hidden
Scyld Computing Corporation             http://www.scyld.com
410 Severn Ave. Suite 210               Second Generation Beowulf Clusters
Annapolis MD 21403                      410-990-9993