Re: [PATCH] Revert "x86: fix 27-rc crash on vsmp due to paravirt during module load"

[Paul Gortmaker <paul.gortmaker@windriver.com>]

On 14-01-16 03:44 AM, Ingo Molnar wrote:
> 
> * Paul Gortmaker <paul.gortmaker@windriver.com> wrote:
> 
>> This reverts commit 05e12e1c4c09cd35ac9f4e6af1e42b0036375d72.
>>
>> The old change from __init --> __init_or_module doesn't make
>> sense anymore, given that the arch/x86/kernel/vsmp_64.c code
>> is only for non-modular builds.
>>
>> As of commit 70511134f61bd6e5eed19f767381f9fb3e762d49
>> ("Revert "x86: don't compile vsmp_64 for 32bit") this file
>> became obj-y and hence is only for built-in.  That makes
>> any "or_module" support redundant.
>>
>> We need to distinguish between the two in order to do some
>> header cleanups.  This revert starts that process.
>>
>> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
> 
> I suspect you want to carry this as part of your header cleanups, to 
> not have to wait for upstream propagation?

Either way is fine; I've already fed out some of the other
independent chunks to maintainers just to keep the size of
what goes in from me at rc1 a bit smaller.

> 
> Please fix the title: the original commit was right, but was mooted by 
> other changes. That's not a revert but a change that happens to be the 
> inverse of an earlier change.

Fair point; will fix.

Thanks,
Paul.
--

> 
> Thanks,
> 
> 	Ingo
>