Re: Finding mysterious 2.0.33 crashes

Doug Ledford (dledford@dialnet.net)
Sun, 15 Feb 1998 15:28:01 -0600


G. Sumner Hayes wrote:
> I've had no problems with corrupt files, sig 11s, or anything else
> that's indicative of a hardware problem -- I just used rpm to verify the
> md5sums on my installed software, and they all look clean. The problem
> is unrelated to the machine's load (CPU, Network, disk, or other) --
> the only time that it's crashed while anyone was logged in was when I
> was doing nothing more than read e-mail. All my hardware timings are
> conservative, and the machine's not overclocked or anything obvious like
> that.
>
> The good news:
>
> It's been two and a half weeks since I had a hang (that was 2.0.33, I'm
> using pre-2.0.34-2 now). If it stays that way, I'm going to file this
> problem away in my archives and ignore it until other weirdnesses start
> happening. It would be nice to know whether it was a hardware problem
> or a kernel problem, but if I can't reproduce it then it's going to be
> nigh impossible to determine that.

Actually, this goes back to the example I gave with the AMD CPU in my last
email. The problem actually may not have been in 2.0.33, it could have
shown up earlier. Something like certain CPUs don't compile the kernel
perfectly (like my 5x86 CPU, the was either a 2.0.32 to 2.0.33 upgrade, or a
2.0.31 to 2.0.32 upgrade) and the problem could go away with a make clean &&
make zlilo under a running 2.0.33. A single corrupt file could carry
forward several kernel versions if you patch up and the file doesn't get
recompiled. Anyway, the basic point to that was to double check your
hardware and make a clean kernel just to make sure we aren't chasing
something that was a latent problem that has already been fixed.

-- 

Doug Ledford <dledford@dialnet.net> Opinions expressed are my own, but they should be everybody's.

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu