From owner-freebsd-hackers Wed Jul 10 10:11:43 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id KAA26924 for hackers-outgoing; Wed, 10 Jul 1996 10:11:43 -0700 (PDT) Received: from brasil.moneng.mei.com (brasil.moneng.mei.com [151.186.109.160]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id KAA26917 for ; Wed, 10 Jul 1996 10:11:40 -0700 (PDT) Received: (from jgreco@localhost) by brasil.moneng.mei.com (8.7.Beta.1/8.7.Beta.1) id MAA25844; Wed, 10 Jul 1996 12:10:35 -0500 From: Joe Greco Message-Id: <199607101710.MAA25844@brasil.moneng.mei.com> Subject: Re: Some recent changes to GENERIC To: jkh@time.cdrom.com (Jordan K. Hubbard) Date: Wed, 10 Jul 1996 12:10:35 -0500 (CDT) Cc: hackers@FreeBSD.ORG In-Reply-To: <4462.837016183@time.cdrom.com> from "Jordan K. Hubbard" at Jul 10, 96 09:29:43 am X-Mailer: ELM [version 2.4 PL24] Content-Type: text Sender: owner-hackers@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > > However, I will definitely scream if anyone removes sio2/sio3. Disabled by > > default, MAYBE. Removed, NO. I have seen far too many people who have > > three or four STANDARD SERIAL PORTS and sio2/3 directly map to COM3/4. > > There should be NO reason to screw around with this. These lines support > > standard PC hardware. If you remove them, remove sio1 too because you > > obviously only need one serial port to do an install. > > I'll compromise with you - we keep ed1, which I'm going to fold over > since it seems so many WD/NE have a conflicting view of what a > "standard setting" is. sio2/3 have already gone and I can hardly > accept the argument that 4 port PCs are any kind of norm. I've seen > literally hundreds of different PC configurations and in all but a few > very rare cases, it's always the same 2 ser/1 par/1 game combo. One mouse, one modem, ports full. Now try to add anything. Think of how many serial gizmo's are out there. Any home hacker has a mouse and modem, as an absolute freaking minimum. Anybody doing anything interesting has more. Like... An ISDN terminal adaptor (often used in ADDITION to a modem). A terminal (a friend has one in his living room). A touchscreen (yes I have one). A printer (serial lines can drive printers further away). Another computer (common around here). I can't think right now .. arrrgh. but there are plenty of other serial devices you might want to be able to use, and having it not work "out of the box" is a liability to FreeBSD. > Enabling sio3 can also give you a bad headache if you have an ATI > chipset, and it was already disabled in GENERIC as it was, so really > I've only removed one port. It was? Really. NOT. # pwd /ftp/.0/systems/unix/FreeBSD/2.0.5-RELEASE/src # grep sio3 /tmp/2.0.5-RELEASE/src/sys/i386/conf/GENERIC device sio3 at isa? port "IO_COM4" tty irq 9 vector siointr # cd ../../2.1.0-RELEASE/src # pwd /ftp/.0/systems/unix/FreeBSD/2.1.0-RELEASE/src # grep sio3 /tmp/2.1.0-RELEASE/src/sys/i386/conf/GENERIC device sio3 at isa? port "IO_COM4" tty irq 9 vector siointr # If it got turned off later, that's sort of too bad, as I wasn't aware of (or don't remember) the change and I would have raised a stink earlier. I generally stick to -RELEASE'd code out of necessity. I'm not arguing that nobody has a problem with sio3 being there. Apparently the Stealth cards have a problem. Maybe others do too. That's a problem that HAS to be addressed. I'm just saying we are going about this all wrong. It should probably be addressed through engineering (i.e. "solving the problem") rather than brute force (i.e. "removing two devices to fix one other"). See, it can be a real pain in the ass on a 386DX/40 with 8MB RAM and a 100MB hard disk to find space to rebuild a kernel just to add a few devices. The "-c" thing was really nice when it came around because it eliminated the need to rebuild a kernel just to change settings on current devices. In my opinion, we are better off having a more flexible kernel where certain things can be turned on and off (or configured) as necessary... A short term solution would be to offer a config option to disable a device by default, but allow it to be re-enabled using -c. I have no (!!) objection if you were to build a GENERIC kernel with four serial ports, one enabled, three disabled. It gives the user the flexibility, at a VERY small additional cost to the user (i.e. having to reboot and do -c) instead of the VERY large potential cost of trying to recompile a kernel - which can also be daunting to a new user. A wonderful long term solution would be to expand the -c functionality to allow it to create new instances of a device for which a driver was available. THIS would be EXTREMELY powerful and would largely eliminate the need for users to build new kernels at all. Want more serial ports but only have sio0 config'd? Fine, boot -c, and add sio1, add sio2, add sio3... even if I had to set every frigging parameter by hand, the power that this would offer would be worth it! In any case, the real problem is that this is being done three days before a release. I vote that you leave it as it was. What you are doing is NOT a solution to the problem, it is a hack around the problem, which will cause other people OTHER problems. That in itself should be a convincing argument. IMHO, a real solution should have been engineered into -stable and tested and reviewed and hashed around. This 11th hour stuff is crap. Please forgive me if I sound just a little exasperated. ... Joe ------------------------------------------------------------------------------- Joe Greco - Systems Administrator jgreco@ns.sol.net Solaria Public Access UNIX - Milwaukee, WI 414/546-7968