Re: SOLVED: Erk! Something is *really* wrong here!
On Tue, 4 Mar 1997, Karl Ferguson wrote:
> FINALLY, someone with the same problem as me - thankyou! :-) (I thought I
> was going crazy <smirk>)
>
> Ok - I have FIXED this error (though who knows, after this email it could
> stuff up, but it's been ok for an hour now...)
>
> This is the step-by-step solution I did:
>
> 1. go into rex/binary-i386/base (*not* rex-fixed - it still happens there!)
> 2. dpkg -i *.deb - say no to update any config file except /etc/init.d/boot
> 3. changed /etc/lilo.conf to boot the original debian shipped kernel.
> 4. made sure that /etc/modules had "auto" commented out.
> 5. rebooted.
> 6. checked logins and wtmp/utmp - they seemed to work!!!
> 7. put back my old custom 2.0.29 kernel in the /etc/lilo.conf
> 8. rebooted.
> 9. system seems to work great!
>
> I dare *NOT* to install anything from bo/binary-i386/base until we can find
> WHAT is causing the problem.
I'm getting the same problem on some of my systems. I've got a good
suspicion that it's mgetty 1.1...here's why:
- on machines which have no modems connected to them, there is no
corruption. some of these have hundreds of telnet/ftp logins per day, so
wtmp/utmp gets written to often enough that corruption would show up if
it was going to.
- on machines with a lot of modem logins, i get a lot of wtmp corruption.
- on machines with only a few modem logins per day, i only get a little
corruption.
all of these machines track unstable pretty closely (some more closely
than others). I've just downgraded to mgetty 1.0-1 from rex-fixed on one
of the dialin machines... i'll check in a few hours to see if that makes
any difference.
maybe mgetty 1.1 was compiled on a system with glibc installed???
craig
Reply to: