Re: [RFC] new module format

From: Roman Zippel (zippel@linux-m68k.org)
Date: Thu Jul 18 2002 - 05:01:09 EST


Hi,

On Thu, 18 Jul 2002, Daniel Phillips wrote:

> But I don't see why this needs to be a two step process:
>
> module support calls mod->cleanup
> module code checks number of users
> if none, unregister all interfaces
> otherwise return -EBUSY
> if return wasn't -EBUSY, free all resources

So after you checked for users, someone starts using the module again,
before you were able to remove all interfaces -> OOPS.
This is exactly the reason why I prefer to make this explicit in the
module interface - the chances are higher to get this right.

> > It's possible to use the filesystem model, but it's unnecessary complex
> > and inflexible.
>
> What is the complex and inflexible part?

It has to work around the problem that cleanup_module() can't fail.

> > Another problem is that the more interfaces a module has (e.g. proc), the
> > harder it becomes to unload a module (or the easier it becomes to prevent
> > the unloading of a module).
>
> I don't see that this makes any difference at all.

It's currently impossible to force the unloading of a module, some user
can always keep a reference to the module. Even if you kill the process,
someone else can get a new reference, before you could unload the module.

> I'm proposing to add a return code to mod->cleanup (and pick a better
> name). Yes, every module will have to be fixed to use this interface, but
> why not?

I don't disagree, but if we break the module interface anyway, why don't
we clean it up properly?
(Patches that do that will follow shortly.)

> Anyway, you're proposing to do it backwards. We need to first ensure
> there are no users, then unregister the interfaces.

That's broken (see above).

bye, Roman

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Tue Jul 23 2002 - 22:00:26 EST