Hi,
I'm trying to optimize NetBSD 7.0 on a Mac LC III. The LC III has 36 MB memory and no FPU. NetBSD 7.0 (everything except for X) was installed successfully on an external SCSI disk using the traditional method.
Good to know we still have some m68030s out there!
1) It takes about 2-3 days to compile a new kernel. Whether using the default or a customized kernel, the system takes about 20 minutes to run through all of its rc scripts during boot, even though almost everything is disabled in /etc/rc.conf. I see in /var/run/rc.log that many scripts are run; is there any automatic way to create a single script with just the things that I need based on the settings in /etc/rc.conf? I could create such a script manually, but I don't want to miss anything important.
It'd be interesting to log every thing that the startup scripts do, then consolidate it. It'd be even more interesting to see if there's some way to automate a run, consolidate, then run the consolidated script until a change is detected. I bet that'd be useful for other low resource / embedded systems.
2) When configuring a new kernel, is it sufficient to comment out entries, e.g.: ... #options M68040 options M68030 #options M68020 ... Or should I use "no options ..." instead?... no options M68040 options M68030 no options M68020 ...
I comment out if I'm editing the kernel configuration file itself, and use "no options" if I'm including another file.
3) Are the FPSP and FPU_EMULATE options needed for a 68030?
FPSP is for the m68040 because there are instructions it has to emulate even with an FPU. But you definitely do need FPU_EMULATE because you don't have an FPU unless you compile everything with -msoft-float.
4) Some commands (such as "ps aux") take a long time (minutes) to run. Is that normal? Here's the output from "ps aux" (I'm currently compiling a kernel and left those lines out):USER PID %CPU %MEM VSZ RSS TTY STAT STARTED TIME COMMAND root 0 0.0 78.1 0 28776 ? DKl Tue07AM 34:51.88 [system] root 1 0.0 2.3 3616 820 ? Is Tue07AM 0:02.24 init root 727 0.0 2.8 6060 1008 ? Ss Tue07AM 0:19.48 /usr/sbin/syslogd -s root 1835 0.0 2.2 3668 804 ? Is Tue08AM 0:01.00 /usr/sbin/inetd -l root 1966 0.0 2.9 3580 1052 ? Is Tue08AM 0:23.48 /usr/sbin/cron
Well, "normal" is a word open to interpretation. It illustrates that there are parts of NetBSD which may have gotten a little more resource hungry than usual. Older, more modest computers are often useful for finding performance bottlenecks.
Is it normal for the "[system]" RSS to be what appears to be the amount of available memory?
That's a lot for a system with 36 megs of memory... I'll have to check what mine is on my smallest memory system, which is a VAXstation VLC with 24 megs of memory...
5) Several years ago I used NetBSD 1.4 on an SE/30, and I don't remember it being as slow. Unfortunately, SE/30s seem to self-destruct over time, and I no longer have one available for testing. Would it be better to just use NetBSD 1.x on the LC III? My main reason for wanting to use 7.0 was to have a more modern gcc and related tools.
SE/30s not only need to have their motherboard recapped, but often work needs to be done on the power supply boards, too, to keep them happy.
6) Are there any plans to support HFS filesystems in addition to HFS+? To more easily move files to NetBSD, I had to install Mac OS 8.1 just so I could create an HFS+ volume. (I boot into NetBSD from Mac OS 7.6.1.) thanks for any information
I don't know if anyone's working on that, but for now you can use pkgsrc/sysutils/hfsutils. It's a little clunky, but you can use it to move files back and forth.
If you like, we can try out a size-optimized, m68030 optimized compile of NetBSD 7. Instead of compiling on an m68030 machine, though, I could compile it on a faster machine, if you'd like to try it.
John