Subject: Re: easy ways to crash your NetBSD system
To: None <current-users@NetBSD.ORG>
From: Brad Walker <firstname.lastname@example.org>
Date: 04/07/1996 03:26:42
> From: blymn@bunya3 (Brett Lymn)
> Subject: Re: easy ways to crash your NetBSD system
> To: email@example.com (Johan Danielsson)
> Date: Sun, 7 Apr 1996 13:06:22 +0930 (CST)
> Cc: current-users@NetBSD.ORG
> Actually, in a lot of the cases panics are about the only thing you
> can do. What sort of handling do you want to do when the file system
> finds the inode it is trying to free is already free? Just say "gee
> I musta lost the plot there someplace, I will just continue on..." to
> do more damage to the file system. How do you handle a hard error in
> your swap space? Just kill the process and continue on? What if the
> process is inetd? or even init?
There are reasons why fork(2) returns an error as reported in
errno. One of the reasons is that the system has run out of resources
to handle the request. This doesn't mean the system should crash. It
should report the error.
As for file system errors, there are places where one could logically
roll back from the error and not panic the system.
And as for a hard error in swap space doesn't mean the system panics,
it means you retry the operation and if it fails again then you kill
the process with a faulting code of BUS_OBJERR or something similiar..
There are times when panic is appropriate. There are times when the
problem needs a better approach to the problem..