Re: critical syslog bugs in busybox prevent further dbootstrap work
On Thu Jan 20, 2000 at 11:34:40PM -0500, Ben Collins wrote:
> On Thu, Jan 20, 2000 at 12:54:56AM -0700, Erik Andersen wrote:
> > On Wed Jan 19, 2000 at 11:57:04PM -0500, Adam Di Carlo wrote:
> > >
> > > Init is definately broken -- can we move the tail out of inittab and
> > > back into rcS, please??!
> > >
> > > Also, syslog is utterly broken as well. 'logger foo' makes no entry
> > > in /var/log/messages .
> > >
> > > There are a million bugs in dbootstrap but I've lost all my debugging
> > > capabilities, so I can't even really work on things until this is
> > > fixed.
> > I've committed changes to inittab and rcS changes to revert syslogd to being
> > started from rcS. I can't seem to test it though, since my floppy drive seems
> > to have just broken. If you can test it, that would be nice. If not... I'll
> > try tomorrow at work.
> > Hope this fixes things,
> This wont fix it, and can't. I've found that the problem stems from how
> and when the root fs is remounted read-write. We are going to have to make
> init remount the root fs read-write, before it runs any of the scripts.
> This is also causing the problem with the console and other tty's. Trying
> to open up a device read/write, when the filesystem is read-only.
> I will look at this tomorrow along with the new change-root patch.
While I agree with you in principle, and I'd like to find a more general
solution, in practice everything is working just fine now, and it is written
such that it ought to work just fine over serial console as well (though I
havn't tried it with the boot floppies, I have the same version of busybox init
running over serial console on some VME powerpc boxes at work and it works just
I spent some time poking at it this evening and making a couple of cosmetic
fixes to text formatting. Boots up and runs just fine. Syslogd is syslogging,
dbootstrap is bootstrapping, and things just seem to be working. I didn't
finish the install (I was testing on my wife's box onto a ramdisk), but I
walked throught it till I got to installing base (not enough space on the
ramdisk for that...), but the parts I tested looked just fine. And with tausq'
small x86 stuff, the kernel even detected and ran the installed tulip card, so
presumably (if I had a bit more HD space then the ramdisk I was poking at), I
could have run the net-install and installed potato w only 2 floppies. :)
Erik B. Andersen Web: http://www.xmission.com/~andersen/
--This message was written using 73% post-consumer electrons--