Hi all,
http://ftp.kernel.org/pub/linux/kernel/people/rusty
Unified boot/module parameter and module loader rewrite
updated to 2.4.14. I'm off to Linux Kongress, so I'll be difficult to
contact for 10 days or so.
Main TODOS:
1) Should the PARAM() macros also declare the variables?
Lots of people seem to like writing INT_MODULE_PARM macros...
2) Need a less-sucky /proc|/proc/sys patch, to add access to
parameters through that.
Cheers!
Rusty.
--
Premature optmztion is rt of all evl. --DK
In article <E166p1R-0004ll-00@wagner> you wrote:
> Hi all,
>
> http://ftp.kernel.org/pub/linux/kernel/people/rusty
>
> Unified boot/module parameter and module loader rewrite
> updated to 2.4.14. I'm off to Linux Kongress, so I'll be difficult to
> contact for 10 days or so.
I absolutly oppose to the cosmetic naming changes.
Please let module be be initialized by module_init() and exited by
module_exit(). We had a hard enough time to get it everywhere, not
to mention the name makes a lot of sense.
Also MODULE_PARAM should just stay, combined with Keith's proposal
to use it at boottime aswell (as KBUILD_OBJECT.<paramname>).
Christoph
--
Of course it doesn't work. We've performed a software upgrade.
In message <[email protected]> you write:
> In article <E166p1R-0004ll-00@wagner> you wrote:
> > http://ftp.kernel.org/pub/linux/kernel/people/rusty
> >
> > Unified boot/module parameter and module loader rewrite
> > updated to 2.4.14. I'm off to Linux Kongress, so I'll be difficult to
> > contact for 10 days or so.
>
> I absolutly oppose to the cosmetic naming changes.
Hi Christoph!
Um, me too. I should have reposted my previous explanation, sorry!
> Please let module be be initialized by module_init() and exited by
> module_exit(). We had a hard enough time to get it everywhere, not
> to mention the name makes a lot of sense.
Unfortunately, removal needs to be done in two stages, to sanely make
things like IPv6 modular (a deactivate, and a kill stage). It turns
out that this applies to loading as well, in case the loading fails
part way through (there's also a number of modules at the moment which
initialize in the wrong order, which can lead to an oops).
> Also MODULE_PARAM should just stay, combined with Keith's proposal
> to use it at boottime aswell (as KBUILD_OBJECT.<paramname>).
Firstly, Keith and I agree on KBUILD_OBJECT[/.]paramname, BTW. I'm
not sure it was his proposal originally, though: it's a pretty simple
and old idea.
The previous module param stuff was prone to user bugs (no type
checking), was not extensible, and required duplicated code for boot
time. It also did not have the option of appearing in /proc.
It is possible to write macros mapping the NEW macros to the OLD, but
not vice versa, otherwise I wouldn't change at all.
Hope that clarifies,
Rusty.
--
Premature optmztion is rt of all evl. --DK
Rusty Russell wrote:
>
> Hi all,
>
> http://ftp.kernel.org/pub/linux/kernel/people/rusty
>
> Unified boot/module parameter and module loader rewrite
> updated to 2.4.14. I'm off to Linux Kongress, so I'll be difficult to
> contact for 10 days or so.
>
> Main TODOS:
> 1) Should the PARAM() macros also declare the variables?
> Lots of people seem to like writing INT_MODULE_PARM macros...
>
> 2) Need a less-sucky /proc|/proc/sys patch, to add access to
> parameters through that.
I did have a look at it and I would like to express my full
encouragements
to yours efforts. Finally somebody out there caring about modularization
of the kernel, who doesn't continuously introduce sick stuff like...
We remember:
1. Highjacking System V IPC for autoloader communication.
2. "Persistant module storage" patches, which destabilize the
kernel.
3. inter_module* idiocy.
4. Overloaded system calls.