Re: [PATCH] ARM: tegra: make tegra_cpu_reset_handler_enable() __init

[Peter De Schrijver <pdeschrijver@nvidia.com>]

On Mon, Jul 23, 2012 at 10:09:30PM +0200, Stephen Warren wrote:
> On 07/18/2012 09:01 AM, Peter De Schrijver wrote:
> > On Mon, Jun 18, 2012 at 11:01:50PM +0200, Stephen Warren wrote:
> >> From: Stephen Warren <swarren@nvidia.com>
> >>
> >> This solves a section mismatch warning. I hadn't noticed this before,
> >> because my compiler was inlining tegra_cpu_reset_handler_enable() inside
> >> tegra_cpu_reset_handler_init(), which is already __init, but I switched
> >> compilers and it stopped doing that.
> > 
> > Why does this generate a section mismatch warning? I see why calling a
> > a __init marked function from a non __init marked function is a problem, but
> > the opposite should be ok no?
> 
> The issue is that tegra_cpu_reset_handler_enable() itself (which was not
> __init but the patch made to be __init) was referencing variable
> __tegra_cpu_reset_handler_{start,end}. This wasn't noticed before
> because when tegra_cpu_reset_handler_enable() was inlined into
> tegra_cpu_reset_handler_init(), tegra_cpu_reset_handler_enable() was
> effectively __init. However, when built as a separate function, you
> ended up with a non-__init function referencing other things that were
> __init.
> 
> In other words, this is indeed nothing to do with __init function
> tegra_cpu_reset_handler_init() calling non-__init function
> tegra_cpu_reset_handler_enable().

Ah. That makes sense :) We need to be careful with this when introducing
system suspend support because some of this work needs to be redone on resume.

Cheers,

Peter.