Re: kernel module (un)loader

[William Knop]

>On Mon, 26 Jun 2000 23:50:06 EDT,
>"William Knop" <[e-mail]> wrote:
> >I'm not a kernel-elite, but I've been wracking my brain over a way to 
>find a
> >way to be able to isolate a module in the event it crashes. The module 
>would
> >then be unloaded and then a clean one loaded in its place. Perhaps this
> >could be integrated into the intelligent module (un)loader or there could 
>be
> >a deamon to oversee this. Perhaps the kernel could be locked when crash 
>is
> >detected, and then modules could be dealt with-- not too sure. I'd
> >definitely be willing to code something like this, but I need a little
> >direction. Thanks!
>
>Define "crash".  If the module corrupts critical kernel code or fails
>with interrupts disabled then the entire kernel is dead, reboot time.
>If the module fails and leaves a non-zero use count on something then
>the module cannot be unloaded, not even manually.  Only if a module
>runs clean and has no dangling resources can it be removed.  In which
>case, how do you tell it crashed?
>

How about inserting a #define that tracks module usage, so upon the event of 
corruption or a kernel panic, use counts and various kernel kernel resources 
may be cleared and the correct module may be reloaded? I know Linux in 
Monolithic, but there must be some method of keeping tabs on this stuff to 
attempt crash recovery.

Monolithic with all the pros of micro and none of the cons. That's what I 
aim for. I don't think implementation would slow linux at all.

-Will
________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com