[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Segmentation faults



Mathew Eis wrote:
> > >>  "Starting inetd superserver:
> > >>  INIT: Id "1" spawning too fast: disabled for 5
> > minutes"
> > >
> > >Unfortunately, you have a broken 68LC040. There's some
> > bug in this
> > >processor in the Floating Point Exception-handling and
> > Memory-management
> > >part of it.
> >
> >   IMHO has the complaint from init  about 'respawning too
> > fast'
> > nothing to do with the processor!
> > I had the same message with LinuxPPC on an Apple PowerMac
> > G4 which is
> > quit a different machine.
> > I succeeded to remove the symptom but uptill now I did
> > not care about
> > solving the problem.
> > By looking at the screen I found after which command the
> > message
> > appeared for the first
> > time during boot and which command followed thereafter.
> > By scanning
> > the init scripts in the
> > order they are executed I found the offending
> > script/program.
> > As far as I can remember the "1" stands for a script or
> > program. I
> > renamed the offending file and the message did not
> > reappear.
> >
> 
> Yes, however, the "offending script/program" is most likely
> crashing because of the buggy LC040.
> 
> Removing a part of the system to prevent the message will
> not fix his main problem - random segmentation faults; not
> only in whatever program is crashing at startup, but dpkg,
> perl and others. These faults are caused by the 68040 CPU,
> and there are only 3 solutions I am aware of.
> 
> 1) Get a non-LC040 CPU
> 2) Fix the floating point emulator
> 3) Recompile all the packages with sofware fp emulation
> 
> > In my
> > hurry to prevent people from buing new processors I did
> > not yet find the name of the script/program, but when I
> > find it I will post it.]
> 
> The script that is crashing is probably in /etc/inittab

Yes, the original poster was talking about the second phase of the
debian install. At this point /etc/inittab contains 1 process:
dpkg-<something> that will start the second installation phase.
dpkg-<something> calls perl which segfaults because it hit the bug in
the 68LC040 cpu.

I can add one solution to the three above:
4) replace /etc/inittab with /etc/inittab.real and reboot to restart
init, them manually (dpkg -i) install all needed packages. Only
advisable to (Debian) die-hards.

I'll comment on the other solutions:
1)	There are also very few non-buggy LC040's, but the full 040 also get
you more speed and can easily be found on online-sales etc.
2)	Has been discussed by assembler and kernel guru's several times, but
still seems to hard/impossible/not-worth-it to do.
3)	Someone (sorry forgot your name) on either debian-68k or in
comp.os.linux.m68k is doing this for some basic packages. Watch for
announcements.

HTH, Erik
-- 
---------------------------------------------------------------------------
Erik C.J. Laan				elaan at dds.nl
---------------------------------------------------------------------------



Reply to: