Re: [PATCH] remove useless MOD_{INC,DEC}_USE_COUNT from sunrpc

[Christoph Hellwig <hch@lst.de>]

On Sun, May 04, 2003 at 01:15:58PM -0700, David S. Miller wrote:
>    From: Christoph Hellwig <hch@lst.de>
>    Date: Sun, 4 May 2003 23:00:11 +0200
>    
>    Oh well, what about something like the following?
>  ...   
>    +	 */
>    +	if (!try_module_get(THIS_MODULE))
>    +		return -EBUSY;
> 
> Ahem... Why don't we just do this right? :-)
> 
> By this I mean provide some real registry thing in the
> main kernel image that we can use to do try_module_get()
> outside of the sunrpc module?

Please read the comment above that piece of code.  We always
enter this through an exported function so we know we currently
aren't unloadable.

> The other option is the make more progress in the area of
> two-stage module unload, and allowing cleanup() to return
> whether the module is unloadable or not.  This is being
> discussed on netdev so that we have some way to make ipv6
> modules work sanely (instead of putting try_module_get() in
> every other line, that simply isn't acceptable).

That's fine with me, but I won't sign up to do that work :)