Re: eepro100 works again.

[Andrew Coulthurst <andyc@svarog.demon.co.uk>]

Hi Rogier,
Yes there are people with the eepro100 on this list !

I wrote some ( Ugly ) patches for the eepro100 to get it to work for me 
, and yes I had your problem about the code finding the wrong
information
about the io address. The patches on the Etherboot homepage contain an 
almost exact copy of the pci search routines implemented in the linux
kernel
version 2.0.33. These specifically look for an io address , rather than 
the first. 

Sincerely,

Andrew Coulthurst.



Rogier Wolff wrote:
> 
> Hi,
> 
> I've been bugged two or three times the last month about why
> Etherboot 4.0 doesn't work with the eepro100 anymore.
> 
> I just got out of my lazy bed, and fixed it (*). But there are some
> things that someone who wrote the code should comment on.....
> 
> First from config.c:
> 
> #ifdef  INCLUDE_PCI
>         pci_addrs[0] = 0;
>         eth_pci_init(pci_nic_list, pci_addrs);
>         pci_addrs[1] = 0;
> #endif
>         printf("Probing...");
>         for (t = NIC; t->nic_name != 0; ++t)
>         {
>                 printf("[%s]", t->nic_name);
>                 if ((*t->eth_probe)(&nic, t->probe_addrs))
>                         return (1);
>         }
> 
> puts the PCI IO address in the pci_addrs array. It then calls my probe
> function with the t->probe_addrs array, which doesn't tell me where my
> card is. I haven't the foggiest about how this works for the NE2000
> compatible PCI cards: as far as I can read the code they are probed at
> classical ISA (0x200, 0x300, 0x320 etc) adresses.
> 
> My current (very ugly) workaround is to export the pci_addrs array,
> and to reference that instead of probe_addrs.... I want something
> better.
> 
> Next, from pci.c:
> 
>                 for (i = 0; pcidev[i].vendor != 0; i++) {
>                         if (vendor == pcidev[i].vendor
>                                 && device == pcidev[i].dev_id) {
>                                 int i;
>                                 for (i=3;i>=0;i--) {
>                                         pcibios_read_config_dword(0, devfn,
>                                                 PCI_BASE_ADDRESS_0 + 4 *i, &pci_ioaddr);
>                                         if (pci_ioaddr & 0x01) break;
>                                 }
> 
>                                 /* Strip the I/O address out of the
>                                  * returned value */
>                                 pci_ioaddr &= PCI_BASE_ADDRESS_IO_MASK;
>                                 printf("Found %s at 0x%x\r\n",
>                                 pcidev[i].name, pci_ioaddr);
>                         }
>                 }
> 
> This is the NEW version. It looks for an -=IO=- address on the PCI
> device.  The old code found a noncacheable memory region as the
> address I can't really do anything with that.... My eepro100 boasts
> three adresses, the second one being the IO address that we need.....
> 
> Anybody with eepro100's lurking on this list?
> 
>                                         Roger.
> 
> (*) It loads a kernel, initialized the kernel, does the NFS-root
> mount, and then crashes. I probably didn't correctly install the NFS
> root for my client machine :-) Goes to show that I personally don't
> actually use it.... The machine boots perfectly from its harddisk...
> 
> --
> If it's there and you can see it, it's REAL      |___R.E.Wolff@BitWizard.nl  |
> If it's there and you can't see it, it's TRANSPARENT |  Tel: +31-15-2137555  |
> If it's not there and you can see it, it's VIRTUAL   |__FAX:_+31-15-2138217  |
> If it's not there and you can't see it, it's GONE! -- Roy Wilks, 1983  |_____|