[Pkg-pciutils-discuss] Bug#335665: kudzu patch to not use PCI_FILL_CLASS

[Martin Mares]

Hello!

> Aha.  That makes sense; I suspect most PCI devices don't allow the OS to
> write that field ;-)  I didn't like the 'read it byte-by-byte' part of
> the patch, so I changed it to pci_get_word().
> 
> Martin, now that we have a reason, what do you think to this patch?

Applied :)

(Available in the GIT repository and also as a -pre2.)

> --- pciutils-2.2.1/lib/generic.c.devicetype	2004-08-13 16:15:23.000000000 -0400
> +++ pciutils-2.2.1/lib/generic.c	2005-12-13 17:02:12.000000000 -0500
> @@ -46,7 +46,8 @@
>  	  d->func = t->func;
>  	  d->vendor_id = vd & 0xffff;
>  	  d->device_id = vd >> 16U;
> -	  d->known_fields = PCI_FILL_IDENT;
> +	  d->device_class = pci_read_word(t, PCI_CLASS_DEVICE);
> +	  d->known_fields = PCI_FILL_IDENT | PCI_FILL_CLASS;

Except for this bit -- why should we read it if nobody requested?

> --- pciutils-2.2.1/lib/example.c.devicetype	2000-03-09 03:38:33.000000000 -0500
> +++ pciutils-2.2.1/lib/example.c	2005-12-13 17:02:12.000000000 -0500
> @@ -21,7 +21,7 @@
>    pci_scan_bus(pacc);		/* We want to get the list of devices */
>    for(dev=pacc->devices; dev; dev=dev->next)	/* Iterate over all devices */
>      {
> -      pci_fill_info(dev, PCI_FILL_IDENT | PCI_FILL_BASES);	/* Fill in header info we need */
> +      pci_fill_info(dev, PCI_FILL_IDENT | PCI_FILL_BASES | PCI_FILL_CLASS);	/* Fill in header info we need */
>        c = pci_read_word(dev, PCI_CLASS_DEVICE);	/* Read config register directly */

This doesn't make much sense, I've tried to come with a better example.

				Have a nice fortnight
-- 
Martin `MJ' Mares   <mj at ucw.cz>   http://atrey.karlin.mff.cuni.cz/~mj/
Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth
IBM = Inside Black Magic