Re: [PATCH] x86: fix unconditional arch/x86/kernel/pcspeaker.c?compiling

[Dmitri Vorobiev <dmitri.vorobiev@gmail.com>]

Michael Opdenacker пишет:
> On Friday 18 January 2008, Matt Mackall wrote:
>> Probably makes sense to define it right next to INPUT_PCSPKR in
>> drivers/input/Kconfig.
>>
>> Then do the appropriate fix for all arches mentioned in INPUT_PCSPKR.
>>
>> For extra points, you can move the duplicate pcspeaker.c code out of all
>> those arches and stash it somewhere in drivers/input. Presumably it's
>> possible to get it to link into the kernel even when INPUT is modular.
> 
> Here's the patch, after spending some time to get familiar with git.
> 
> The patch is against git x86/mm, and it seems to work fine on x86. However,
> on x86, you no longer have /sys/devices/platform/pcspkr after
> running "make allnoconfig". To get it, you have to explicitely add support
> to CONFIG_INPUT_PCSPKR (m or y). I hope this is fine.
> 
> I'm copying the MIPS maintainer as this patch touches his tree too.

This patch does not apply cleanly to linux-mips Git tree:

----> snip
patching file arch/x86/kernel/Makefile
Hunk #1 FAILED at 69.
1 out of 1 hunk FAILED -- saving rejects to file arch/x86/kernel/Makefile.rej
----> snip

I believe this is because the linux-mips tree and the -mm tree are not in sync.

> 
> On MIPS, there should be no impact though, as CONFIG_PCSPEAKER is set in
> defconfig files.
> 
> In other architectures where CONFIG_INPUT_PCSPKR can exist,
> there is a change: when CONFIG_INPUT_PCSPKR is set, the platform device
> will be added, while it didn't exist before. I hope this is fine.
> 

It seems tempting to include at least Alpha in this patch, because they
have a device initcall identical to the one, which you're removing from
the MIPS and x86 arch code.

Thanks,

Dmitri