From owner-freebsd-current Wed Jul 30 11:34:34 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.5/8.8.5) id LAA22984 for current-outgoing; Wed, 30 Jul 1997 11:34:34 -0700 (PDT) Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.50]) by hub.freebsd.org (8.8.5/8.8.5) with SMTP id LAA22976 for ; Wed, 30 Jul 1997 11:34:31 -0700 (PDT) Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id LAA05510; Wed, 30 Jul 1997 11:31:43 -0700 From: Terry Lambert Message-Id: <199707301831.LAA05510@phaeton.artisoft.com> Subject: Re: pccard and -current; a long way to go. :-( To: jdp@polstra.com (John Polstra) Date: Wed, 30 Jul 1997 11:31:42 -0700 (MST) Cc: TLiddelow@cybec.com.au, msmith@atrad.adelaide.edu.au, current@FreeBSD.ORG In-Reply-To: <199707300429.VAA07677@austin.polstra.com> from "John Polstra" at Jul 29, 97 09:29:49 pm X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > These benefits are real, but I'm not sure they're enough. (If you > are looking for a facts-be-damned advocate of FreeBSD-ELF, I'm not > your guy.) There is plenty of room for intelligent disagreement > about it. I myself have about three opinions on the subject, all > in conflict with one another. :-) > > I am sure there would be other benefits, but I don't know enough > about them to lobby on their behalf. The people who do are the > ones who need to do the PR. Explain why FreeBSD would be a better > place if it used ELF. So far nobody's done a very good job of > that. "Pageable kernel data areas?" Yawn. What does it mean to > me as a user or as a developer? It means: o Demand loading for LKM's. o Demand loading of dependencies (LKM TCP requires LKM IP, so loading LKM TCP loads LKM IP). o The generic kernel can discard drivers it's not using. o You can page out memory and page it back in to let you rearrange the pages. This means you don't have to statically link ing your "QuickCam" or similar driver that needs a region of physically contiguous kernel memory (ie: it can only currently allocate this at boot time). o You can write glue to let BSD use NT drivers (NT drivers, by design, require section coloring) for video cards, SCSI controllers, ethernet cards, etc.. NT drivers come with most new hardware. o You can replace drivers in a kernel using a section archiver instead of rebuilding and relinking. > Would the system run faster? Yes. It would page user data less if it could reclaim the memory from unused drivers and/opr probe code for drivers that *are* used. > Use less memory? Yes. > How much faster? How much less? Depends. How many transient or unused drivers are in your current kernel? Probably a hell of a lot, if you run generic, like most people. > Who's going to implement it if we switch to ELF? I think the point is that unless we switch to ELF, no one will *ever* implement it. > Enquiring minds want to know. Gotta disagree with you there. ;-). Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.