Re: Linux kdb (was Re: first thread)

[John Alvord <jalvo@cloud9.net>]

On Fri, 29 Jan 1999, Tigran Aivazian wrote:

> On Fri, 29 Jan 1999, Philipp Rumpf wrote:
> > > 1. crash(1) to analyse those (dumps not humans) in 0. as well as playing
> > > with live /dev/mem.
> > 
> > The name sounds a bit strange to me ...
> Why so? crash(1) (or crash(1M) if you are pedantically ATT-ish) is a
> standard UNIX command present in most UNIX variants. Therefore Linux
> should have it too.
> 
> The only problem with Linux is that it is so good and flexible that things
> normally done with crash in several hours, in Linux are done in seconds by
> just catting some files in /proc or reading the sources... That puts
> people off writing crash(1) for Linux, probably.

The main reason it hasn't been done is that Linus is absolutely opposed.
He believes that fixes developed from crash dumps typically cure symptoms
rather than correct the underlying problem. That seems a bit weird to me,
but has been expressed several times.

A secondary reason is that people are worried that the attempt to write
out the crash data will destroy user data, since the system is unstable.

John Alvord

Music, Management, Poetry and more...
           http://www.candlelist.org/kuilema
 
Cheap CDs @ http://www.cruzio.com/~billpeet/MusicByCandlelight
 



-
Linux-future: thinking about the future of the Linux kernel
http://humbolt.nl.linux.org/lists/