> On Sun, 26 Apr 1998, Steven N. Hirsch wrote:
>
> > Hmm. Not so fast... With 2.1.98 + daemonless kmod I'm seeing some
> > strangeness during init and shutdown. Between the time that the first
> > module is loaded by kmod and the appearance of the login prompt, all
> > syslog and screen output is missing in action. Ditto on shutdown; no
> > screen output or logging after 'nfsd stop'.
> Oops, sorry. Didn't mean to be attributing incorrect information. My
> apologies
> > I've isolated this to the daemonless-kmod patch, and it's 100% repeatable
> > on a 6x86 box here. To make matters worse, my 486/100 box (with the same
> > kernel and patch level) refuses to run rpc.mountd during init. If I
> > start it by hand afterwards, all is well. When I put 'echo ..' statements
> > into the init script, nothing appears!
>
>
> Couple of things: I noticed that Adam added the '-s' option to the
> modprobe command line, which logs data to syslog. Maybe something is
> going wrong with this.
I'll try eliminating it, but doubt it's the root of the problem.
> Other possibility - I remember when testing my
> kmod alterations that there were some odd occasional memory corruptions.
> This was manifested by GPM registering mouse movements that I didn't make,
> and failure of sh scripts that I know are correct. Try inserting some
> dummy code into kmod and see if that changes the misbehavior. If it does,
> we may have some (un?)-related memory trashing. You're using gcc 2.7.2.3,
> not egcs or gcc2.8, right?
No, I'm primarily using egcs-1.0.2 w/ binutils-2.9.0.3. However, as a
control I rebuilt the kernel on my 486 box with 2.7.2.1 and binutils
2.8.x.y. No difference. And, to clarify my problem report, the failure
to run rpc.mountd is happening on _both_ platforms.
> Other thing - did the exact same setup work for 2.1.97 with Adam's patch?
> Only because the patch is exactly the same.
Turns out that the lack of console output was also occurring w/ 2.1.97 +
Adam. I don't recall seeing problems with mountd, though.
When you consider this wierdness along with Richard Henderson's report
about kmod not wanting to work in user space with all signals blocked,
there's an indication of underlying misbehavior. Now to find it..
Steve
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu