Updating PCI vendors database

[Philip Paeps]

On 2011-04-04 07:31:53 (-0700), Garrett Cooper <gcooper at FreeBSD.org> wrote:
> On Mon, Apr 4, 2011 at 7:19 AM, Philip Paeps <philip at freebsd.org> wrote:
> > On 2011-04-04 16:10:16 (+0200), Philip Paeps <philip at freebsd.org> wrote:
> >> Would anyone object if I updated this list to source the same database used
> >> by Linux distributions at http://pciids.sourceforge.net/v2.2/pci.ids?
> >
> > It occurs to me that people would want to verify that this list does actually
> > work and that we gain (rather than lose) coverage from it.
> >
> > A sanity test I've run on a couple of machines:
> >
> >  % fetch http://pciids.sourceforge.net/v2.2/pci.ids
> >  % pciconf -lv > /tmp/pciconflv.old
> >  % PCICONF_VENDOR_DATABASE=pci.ids pciconf -lv >/tmp/pciconflv.new
> >  % diff -u /tmp/pciconflv.old /tmp/pciconflv.new
> >
> > In all cases I've seen so far, the new list yields better (more correct and up
> > to date) results than the exising list.  In no cases has pciconf complained
> > about the new list.
> 
> I've copy-pasted the discussion I brought this up to Warner/Brooks
> several months ago for review.

I think at that point, the lists we're currently sourcing still existed.  As
of this morning, I don't seem to be able to find Craig Hart's list of PCI IDs
anywhere on the net.  It's certainly no long available at the address listed
in the source tree.

> The big problem is that the descriptions with the previous source and the
> new source clash, so this would cause a huge amount of diff churn;

This would be a problem I agree, but not a huge one.  If the churn would be
too large, I would be hesitant to push this to any stable branches.  But I
don't think the churn should be a problem for HEAD.  It doesn't look to me
like the churn is too large on the machines I've used it on.

Generally, the changes I've seen is devices which lacked a device description
now have one, and devices which had a wrong or incomplete description now have
a complete and (as far as I can tell) correct one.  This feels like an
improvement to me. :)

> plus I think there are a few entries missing from each area (at least there
> were the last time I looked -- maybe our pci_vendors is more spartan than
> the new source is today).

The new list is much more complete than the list we have currently.  For one
thing, it also includes subvendors and subdevices, which the current list
lacks.  Our pciconf doesn't care about those currently, but could be made to
care.

I also don't think we should underestimate the value of sharing a list with
Linux (especially if the licence on the list is friendly to sharing it) and
potentially other operating system vendors.

 - Philip

-- 
Philip Paeps                                    Please don't Cc me, I am
philip at freebsd.org                               subscribed to the list.