> > My only concern with the implementation is, in the kernel the combination
> > of tasklet_init/DECLARE_TAKSLET is seen in over ~400 plus files.
> > With the change(dropping unsigned long argument) there will be huge list
> > of patches migrating to the new api.
>
> Yeah, this is the main part of the work for making this change. When
> the timer API got changed, I had to do a two-stage change so we could
> convert the users incrementally, and then finalize the API change to
> drop the old style.
>
> Beyond that, yeah, everything you sent looks good. It's just the
> matter of building a series of patches to do it without breaking the
> world. (Though if it's small enough, maybe it could be a single patch?
> But I doubt that would be doable...)
Thank you.
I like the idea of two-stage approach. For first stage, I was thinking of all
the architectures with their default configs(Nothing gets broken when
defconf's are tried) and for second stage the rest of drivers etc..
Thoughts?
Either ways, I will convert and get them pushed to github, we could see how
best it could be taken to upstream once it's done.
- Allen
