Date: Sat, 23 Sep 1995 13:40:11 +0200 (MET DST) From: J Wunsch <j@uriah.heep.sax.de> To: freebsd-hackers@freebsd.org (FreeBSD hackers) Subject: Re: cvs commit: src/sys/i386/conf GENERIC Message-ID: <199509231140.NAA21450@uriah.heep.sax.de> In-Reply-To: <Pine.BSF.3.91.950923162419.320E-100000@jhome.DIALix.COM> from "Peter Wemm" at Sep 23, 95 05:01:33 pm
next in thread | previous in thread | raw e-mail | index | archive | help
(Moved out to -hackers, that's why almost fully quoted.)
sysv{shm,sem,msg}
> > LKM's?
>
> That's what I thought too when I spoke to Jordan shortly before he did it.
>
> When we were talking, I griped about the fact that if the SYSV* stuff was
> not loaded, and somebody tried to use them, the system mercilessly killed
> the processes without explaination, causing much confusion by the user.
You don't think the "Illegal system call" ain't an explanation? :-)
> There are a couple of things that spring to mind....
>
> First, instead of making the system call table for the sysv routines
> pointing to nosys(), they could point to something that does a tprintf()
> (or is that uprintf()) giving them a message _why_ they got killed,
> before calling nosys(). This may actually be useful in general if there
> was a generic nosys()-type wrapper that printed the name/number of the
> attempted syscall before zapping the process.
Do it. :)
> Second, this sort of thing is crying out for Terry's demand-loading LKM's,
> but I dont really want to get into that religion.. :-) If we hunted
> around, I'm sure we could find enough stuff that could be demand-loaded to
> make the space saving bigger than the cost of the kernel-linker on
> average. (ppp, slip, tun, sysv*, some of the fs's, etc)..
I think ppp and slip can be demand-loaded already. For tun, it's a
bit different (as it's for slip, FWIW): they are required for the
installation process, and demand-loading wouldn't make sense on the
install floppy.
> Third, I'd be happier if they were LKM's by default rather than in
> GENERIC, and the "as-shipped" /etc/sysconfig loaded them by default, with
> the abilility to disable them. I think that'd get around Jordan's need to
> ship a fully-functional system that Xaccel, XFree etc wont blow up on if
> the advertised SHM extension was used. I don't have the time/inclination
> to do that however, so I wont complain. If Jordan can make it fit on the
> floppy, as far as I'm concerned (since there's no "better" solutions to
> the vendor's needs yet), he can go right ahead.
He cannot make it fit on the installation floppy. :-)
The GENERIC in 2.2-current required a major deletia when i've last
attempted to build a 2.2 "release", including but not limited to the
entire PCI stuff and almost all Ethernet interface (except of the ed
driver since i'm using it :). Only after this, it did tightly fit
into the 0x200000 bytes limit again.
RELENG_2_1_0 is not yet as bloated.
--
cheers, J"org
joerg_wunsch@uriah.heep.sax.de -- http://www.sax.de/~joerg/
Never trust an operating system you don't have sources for. ;-)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199509231140.NAA21450>