Re: [PATCH 1/2] drivers/clocksource/timer-of: Remove __init markings

From: Saravana Kannan
Date: Thu Apr 30 2020 - 12:53:29 EST


On Thu, Apr 30, 2020 at 5:57 AM Ulf Hansson <ulf.hansson@xxxxxxxxxx> wrote:
>
> On Tue, 28 Apr 2020 at 20:55, Daniel Lezcano <daniel.lezcano@xxxxxxxxxx> wrote:
> >
> > On 28/04/2020 20:23, Saravana Kannan wrote:
> > > On Tue, Apr 28, 2020 at 12:02 AM Daniel Lezcano
> > > <daniel.lezcano@xxxxxxxxxx> wrote:
> > >>
> > >>
> > >> Hi Saravana,
> > >
> > > You were replying to Sandeep :)
> >
> > Oh, right :)
> >
> > Sorry Sandeep. Thanks for taking the time to elaborate a clear statement
> > of the GKI.
> >
> > [ ... ]
> >
> > >> That was my understanding of the GKI, thanks for confirming.
> > >>
> > >> Putting apart the non-technical aspect of these changes, the benefit I
> > >> see is the memory usage optimization regarding the single kernel image.
> > >>
> > >> With the ARM64 defconfig, multiple platforms and their corresponding
> > >> drivers are compiled-in. It results in a big kernel image which fails to
> > >> load because of overlapping on DT load address (or something else). When
> > >> that is detected, it is fine to adjust the load addresses, otherwise it
> > >> is painful to narrow down the root cause.
> > >>
> > >> In order to prevent this, we have to customize the defconfig each
> > >> version release.
> > >
> > > Sorry, I'm not sure I understand where you are going with this. Are
> > > you agreeing to pick up this change?
> >
> > Right. I agree with the change but I would like to have Thomas opinion
> > on this before picking the patch.
> >
> > Thomas ?
>
> I am not Thomas :-) But just wanted to provide some feedback from my side.
>
> In general we are careful when deciding to export symbols. And at
> least, I think at least we should require one user of it before
> allowing it to be exported (I assume that is what is happening in
> patch2/2 - I couldn't find it)

I believe it is done in Patch 2/2, but I think you don't see them in
the diff because they are already called from the driver, but the
driver won't compile as a module due to the removal of __init markings
in the driver.
https://lore.kernel.org/lkml/182aae1ed5e5d2b124f1a32686e5566c9a27c980.1585021186.git.baolin.wang7@xxxxxxxxx/

-Saravana