Date: Sun, 03 Jul 2011 14:15:46 +0200 From: "Ronald Klop" <ronald-freebsd8@klop.yi.org> To: "Alexey Dokuchaev" <danfe@freebsd.org>, "Warner Losh" <imp@bsdimp.com> Cc: svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org, Hans Petter Selasky <hselasky@freebsd.org> Subject: Re: svn commit: r223519 - head/etc/devd Message-ID: <op.vx1lolht8527sy@212-182-167-131.ip.telfort.nl> In-Reply-To: <9521FD1A-3D77-4F37-B37C-CC73B13BC6B5@bsdimp.com> References: <201106242132.p5OLW3DL071062@svn.freebsd.org> <op.vxofipu38527sy@pinky> <CAEA5F96-D0B6-4A82-A137-5886A7CE5718@bsdimp.com> <20110628084558.GA35392@FreeBSD.org> <9521FD1A-3D77-4F37-B37C-CC73B13BC6B5@bsdimp.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 28 Jun 2011 18:01:37 +0200, Warner Losh <imp@bsdimp.com> wrote: > > On Jun 28, 2011, at 2:45 AM, Alexey Dokuchaev wrote: > >> On Sun, Jun 26, 2011 at 10:14:13PM -0600, Warner Losh wrote: >>> Not in GENERIC. 9.0 is getting ready to feature freeze any minute, s= o >>> there's no time to do the testing that a radical change like that wou= ld >>> require. >> >> Fair enough. Right now, this work seems to cover only USB kernel =20 >> modules, >> is it so? Is expansion into generic PCI area also planned, provided =20 >> that >> you've already had some ideas and seem to be willing to pick up the wo= rk >> given enough support from the other developers (please excuse me if my >> understanding is wrong)? > > The reason we were able to relatively quickly generate this file from =20 > USB is because USB is very regular and uniform in its probe routines. = =20 > They all have (or could be converted to have) something approximating =20 > the same format. > > PCI isn't even remotely close to this, and the techniques that worked =20 > for USB (and some of the short cuts taken) can't work for PCI for us. = =20 > There's no uniformity in the PCI space, although many of the PCI driver= s =20 > use a similar design pattern. They would all have to be converted to =20 > using the same thing, which is kinda impractical at the moment (since =20 > many have additional, driver specific fields). While the tool is kinda= =20 > flexible enough to cope with this situation, more and more code winds u= p =20 > in the tool to cope than is appropriate for the tool. > > That's why this feature is experimental. For what it works for, it =20 > works great. For everything else, it is a lot of work to make it happy= =20 > on a wide-spread enough scale. It is very instructive for how to =20 > approach things, but embeds too much knowledge of the bus code into the= =20 > tool to generate these files than is appropriate (since it is duplicate= d =20 > information from the driver, not exported from the bus/driver). > > Warner Maybe something for a GSOC 2012 project?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?op.vx1lolht8527sy>