Re: [tip:x86/vdso] x86/arch_prctl/vdso: Add ARCH_MAP_VDSO_*

[Dmitry Safonov <dsafonov@virtuozzo.com>]

On 10/26/2016 03:30 PM, Paul Bolle wrote:
> This is bit late, but we're still on v4.9-rc2, so anyhow.
>
> On Wed, 2016-09-14 at 12:34 -0700, tip-bot for Dmitry Safonov wrote:
>> Commit-ID:  2eefd8789698e89c4a5d610921dc3c1b66e3bd0d
>> Gitweb:     http://git.kernel.org/tip/2eefd8789698e89c4a5d610921dc3c1b66e3bd0d
>> Author:     Dmitry Safonov <dsafonov@virtuozzo.com>
>> AuthorDate: Mon, 5 Sep 2016 16:33:05 +0300
>> Committer:  Thomas Gleixner <tglx@linutronix.de>
>> CommitDate: Wed, 14 Sep 2016 21:28:09 +0200
>>
>> x86/arch_prctl/vdso: Add ARCH_MAP_VDSO_*
>>
>> Add API to change vdso blob type with arch_prctl.
>> As this is usefull only by needs of CRIU, expose
>> this interface under CONFIG_CHECKPOINT_RESTORE.
>>
>> Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
>> Acked-by: Andy Lutomirski <luto@kernel.org>
>> Cc: 0x7f454c46@gmail.com
>> Cc: oleg@redhat.com
>> Cc: linux-mm@kvack.org
>> Cc: gorcunov@openvz.org
>> Cc: xemul@virtuozzo.com
>> Link: http://lkml.kernel.org/r/20160905133308.28234-4-dsafonov@virtuozzo.com
>> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
>
>> --- a/arch/x86/include/uapi/asm/prctl.h
>> +++ b/arch/x86/include/uapi/asm/prctl.h
>
>>  #define ARCH_GET_FS 0x1003
>>  #define ARCH_GET_GS 0x1004
>>
>> +#ifdef CONFIG_CHECKPOINT_RESTORE
>> +# define ARCH_MAP_VDSO_X32	0x2001
>> +# define ARCH_MAP_VDSO_32	0x2002
>> +# define ARCH_MAP_VDSO_64	0x2003
>> +#endif
>> +
>>  #endif /* _ASM_X86_PRCTL_H */
>
> On my machine this header ends up in /usr/include/asm/prctl.h. But in
> userspace CONFIG_CHECKPOINT_RESTORE is meaningless. I think if you
> actually want to export these three macros to userspace the guard
> should read:
>     #if defined(CONFIG_CHECKPOINT_RESTORE) || !defined(__KERNEL__)
>
> And if you don't want to export these macros the guard should read:
>     #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(__KERNEL__)
>
> (In that case you're probably better of defining these macros outside
> of uapi.)
>
> I've only lightly tested those two alternatives, so please double
> check.
>

Hi Paul,

thanks for noting that!
Well, I think, I'll remove ifdeffery around those constants.
Let them be always defined, that would be the simplest solution.
Somehow I managed to forget that userspace knows nothing about enabled
kernel options.

Thanks again,
             Dmitry

>
> Paul Bolle
>