Re: [Qemu-devel] [6490] Update #defines for PCI vendor and device IDs from OpenBIOS and Linux

[Blue Swirl <blauwirbel@gmail.com>]

On 2/1/09, Stuart Brady <sdbrady@ntlworld.com> wrote:
> On Sun, Feb 01, 2009 at 12:01:07PM +0000, Blue Swirl wrote:
>  > Update #defines for PCI vendor and device IDs from OpenBIOS and Linux
>
>
> Just a few questions...
>
>
>  > Modified: trunk/hw/grackle_pci.c
>  > ===================================================================
>  > --- trunk/hw/grackle_pci.c    2009-01-30 20:39:41 UTC (rev 6489)
>  > +++ trunk/hw/grackle_pci.c    2009-02-01 12:01:04 UTC (rev 6490)
>  > @@ -154,10 +154,8 @@
>  >
>  >  #if 0
>  >      /* PCI2PCI bridge same values as PearPC - check this */
>  > -    d->config[0x00] = 0x11; // vendor_id
>  > -    d->config[0x01] = 0x10;
>  > -    d->config[0x02] = 0x26; // device_id
>  > -    d->config[0x03] = 0x00;
>  > +    pci_config_set_vendor_id(d->config, PCI_VENDOR_ID_DEC);
>  > +    pci_config_set_device_id(d->config, PCI_DEVICE_ID_DEC_21154);
>
>
> Isn't the DEC 21154 is a Tulip-compatible NIC, and not a PCI bridge?
>  Yes, 0x1011 is the Vendor ID for DEC, and 0x0026 is the Device ID for
>  the 21154, but what was actually intended here?

I'd rather guess it's a PCI bridge.

>  > +#define PCI_DEVICE_ID_IBM_OPENPIC2       0xffff
>
> > +#define PCI_DEVICE_ID_APPLE_343S1201     0x0010
>  > +#define PCI_DEVICE_ID_APPLE_UNI_N_I_PCI  0x001e
>  > +#define PCI_DEVICE_ID_APPLE_UNI_N_PCI    0x001f
>
> > +#define PCI_DEVICE_ID_APPLE_UNI_N_KEYL   0x0022
>
> > +#define PCI_DEVICE_ID_REALTEK_RTL8029    0x8029
>  >  #define PCI_DEVICE_ID_REALTEK_8139       0x8139
>
>
> It probably wouldn't hurt to mark IDs without a corresponding definition
>  in Linux's pci_ids.h with a comment...  (Also, RTL8029 vs. 8139?...)

That's what ne2k uses (not 8139).

>  > +#define PCI_VENDOR_ID_QEMU               0x1234
>  > +#define PCI_DEVICE_ID_QEMU_VGA           0x1111
>
>
> I gather 0x1234/0x1111 was originally chosen by Bochs.  Unfortunately,
>  it is apparently allocated to 'Technical Corp.'  Would it be possible
>  to use something less confusing instead?

The IDs could be replaced with some real, original VGA compatible PCI
card, without any extra features. Bochs extensions should probably be
disabled then. Any suggestions?