Date: Mon, 03 Apr 2006 16:14:46 -0600 (MDT) From: Warner Losh <imp@bsdimp.com> To: jhb@freebsd.org Cc: jmg@freebsd.org, perforce@freebsd.org Subject: Re: PERFORCE change 94510 for review Message-ID: <20060403.161446.78731312.imp@bsdimp.com> In-Reply-To: <200604031503.41288.jhb@freebsd.org> References: <200604030641.k336f6iG055021@repoman.freebsd.org> <200604031503.41288.jhb@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
From: John Baldwin <jhb@freebsd.org> Subject: Re: PERFORCE change 94510 for review Date: Mon, 3 Apr 2006 15:03:40 -0400 > On Monday 03 April 2006 02:41, John-Mark Gurney wrote: > > http://perforce.freebsd.org/chv.cgi?CH=94510 > > > > Change 94510 by jmg@jmg_arlene on 2006/04/03 06:40:19 > > > > don't force the unit number to match the bus number, this breaks > > machines with multiple pci domains... how the alpha worked with > > this is beyond me (besides using custom pci bridge drivers?) > > Alpha has issues with hoses. That said, this might break some things. :( > If nothing else, it makes it nearly impossible now for people to get the > names of hints to override things like PCI routing via tunables correct. > I would prefer a structured solution that preserved the status quo on > machines w/o multiple domains. Maybe have the unit number be something > like domain * X + bus number, where X is some arbitrary constant like > 100 or 1000 (those are better for human parsing). We override the pci unit number to act as a bus number in passing data into the kernel for pciconf. Warner
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060403.161446.78731312.imp>