Re: [TuxOnIce-users] Do other people build TuxOnIce as modul…

Top Page
Attachments:
Message as email
+ (text/plain)
+ signature.asc (application/pgp-signature)
Delete this message
Reply to this message
Author: Raúl Sánchez Siles
Date:  
To: TuxOnIce users' list
Subject: Re: [TuxOnIce-users] Do other people build TuxOnIce as modules?
Hello all:

El Jueves, 17 de Julio de 2008, Luigi Bai escribió:
> I too build TOI as modules, but would happily switch if support were
> dropped. I like the flexibility of modules, but can see arguments against
> them; fiddling with initrd can be non-trivial, functionality that's
> critical at boot time is probably better integrated in the kernel, and
> patch/dev issues may not make the complexity worth the trade-off.
>
> But, that being said, I'd appeciate a separate patch to enable modules, if
> that's easy to maintain, even if mainline will only build an iintegrated
> kernel.
>
> Luigi
>


Being a TOI modules user I thought I should come into this discussion. I
tend to agree with Luigi, these are my comments.

I think using TOI as modules is the right think to do from the pure
conceptual point of view. This is, you only need to use TOI when you need to
hibernate, so why wasting resources having it when you are not using it. This
would exclude bits needed to be built into kernel to provide the
functionality.

I also liked this approach because when there is a new TOI version you don't
need to reboot in order to use it, even more when it fixes some important
bugs or provide major improvements. Please, take this point into mind.

From the practical POV having modules is almost useless these days. No
important TOI changes are happening (this is good) and I guess most people
(including myself) is not unloading the modules after resume. Also avoiding
modules usage makes TOI easier to build and to configure.

I can't evaluate the complexity of keeping the module code, I guess many
approaches could be taken, e.g.: merging TOI core directly bundling it into
built-in code with or without providing patches to build modules....

You should also evaluate having the ability to hibernating with
fixed/improved code (provided as modules) but maybe that's something that
could also be carried out using vmsplice.

All in all I think priority would be merging code into mainline so all Linux
users could enjoy this great functionallity. But also keeping developers
brain healthy so the fun is even longer ;)

HTH,

> > I've been splitting the monolithic patch into separate parts this
> > afternoon for linux-next purposes, and have put the build-as-modules
> > stuff into its own patch as part of this (I'm not going to seek to get
> > all those EXPORT_SYMBOLs merged). I'll keep it around and continue to
> > ship it as part of the monolithic patch.
> >
> > Regards,
> >
> > Nigel
>




-- 
     Raúl Sánchez Siles
----->Proud Debian user<-----
Linux registered user #416098