Date: Tue, 01 Mar 2005 13:55:32 -0700 (MST) From: Warner Losh <imp@bsdimp.com> To: jhb@FreeBSD.org Cc: bmah@FreeBSD.org Subject: Re: cvs commit: src/release/doc/en_US.ISO8859-1/installation/common install.sgml Message-ID: <20050301.135532.39196899.imp@bsdimp.com> In-Reply-To: <200503011529.15251.jhb@FreeBSD.org> References: <86acpnmzih.fsf@xps.des.no> <20050301.094356.112814615.imp@bsdimp.com> <200503011529.15251.jhb@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
From: John Baldwin <jhb@FreeBSD.org> Subject: Re: cvs commit: src/release/doc/en_US.ISO8859-1/installation/common install.sgml Date: Tue, 1 Mar 2005 15:29:15 -0500 > On Tuesday 01 March 2005 11:43 am, M. Warner Losh wrote: > > In message: <86acpnmzih.fsf@xps.des.no> > > > > des@des.no (Dag-Erling Smørgrav) writes: > > : John Baldwin <jhb@FreeBSD.org> writes: > > : > How about a custom boot kernel? When I suggested 12MB rather than > > : > 7MB in the original PR, I was thinking of the case of building a > > : > stripped-down custom kernel. GENERIC certainly isn't going to fit > > : > well in 16MB. > > : > > : A custom kernel should work better. I figured out that the "missing" > > : memory is in fact the memory used by the kernel, so a system with a > > : trimmed kernel should have a lot more memory available. I'll run some > > : more tests... > > > > I've booted a stripped down (but not minimal) kernel with > > hw.physmem=10M to single user, but couldn't even get to multi-user > > with 16M: Too many processes and too much swapping (I didn't trim my > > enabled list on my laptop). If I booted a minimal kernel, and did > > some creative trimming, I think I can get down closer to 6M to 8M on a > > fairly small system (no acpi, etc), but it would be painful to run in > > that environemnt, unless you had special needs (eg, it was an embedded > > platform). > > Then I think we should just say 24MB is the minimum required for both install > and running given all the comments. It will also simplify the docs if we > just go with one number for minimum memory requirement. I agree. I neglected to explicitly state the point I was trying to make: While expert users might be able to make slower, smaller hardware configurations work, in general we should go with more conservative numbers. In the 'embedded' or 'embedding' documentation maybe we should mention the possibility of hand-crafted versions of FreeBSD running in much less memory that for normal users. Warner
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050301.135532.39196899.imp>