Re: Bug#160507: hotplug: module removal makes shutdown take a long time and sometimes triggers Oops...
On Sun, 15 Sep 2002 23:26, Adam Heath wrote:
> On Sun, 15 Sep 2002, Russell Coker wrote:
> > Yes, and there are many such bugs. I think that they were considering
> > removing the module unload functionality from the kernel to resolve such
> > issues. I've been meaning to listen to the MP3 from the Linux Kernel
> > Summit on this topic...
>
> Er, just because you remove the ability to unload a module, doesn't mean
> you fix the bug. The bug is now just hidden.
>
> *All* code should properly clean up after itself, in all cases.
I recommend that you download and listen to the mp3 of Rusty's talk from the
2002 Linux Kernel Summit. It's a 57 minute talk about the problems with the
current module system and how it's virtually impossible to safely unload
modules. Rusty is proposing that modules NEVER be unloaded, and in the
discussion I get the impression that Alan Cox and Theodore Ts'o are taking
serious note of the issue.
I would not be surprised if kernel 2.6 does not support unloading modules at
all!
In the situation regarding my bug report there is no need to unload the
modules to load a new version, no need to unload them to save memory (the
machine is being rebooted and all memory will be free in a matter of
seconds), and no good reason for unloading them.
Russell Coker
Reply to: