asm-x86/byteorder.h: clean up for userspace

asm-x86/byteorder.h: clean up for userspace

[Mike Frysinger]

Since asm-x86/byteorder.h is exported to userspace, use __asm__ rather than
asm in its code.

Signed-Off-By: Mike Frysinger <vapier@gentoo.org>
---
diff --git a/include/asm-x86/byteorder.h b/include/asm-x86/byteorder.h
index 1f2d6d5..fe2f2e5 100644
--- a/include/asm-x86/byteorder.h
+++ b/include/asm-x86/byteorder.h
@@ -30,13 +30,13 @@ static __inline__ __attribute_const__ __u64 ___arch__swab64(__u64 val)
 	} v;
 	v.u = val;
 #ifdef CONFIG_X86_BSWAP
-	asm("bswapl %0 ; bswapl %1 ; xchgl %0,%1"
+	__asm__("bswapl %0 ; bswapl %1 ; xchgl %0,%1"
 	    : "=r" (v.s.a), "=r" (v.s.b)
 	    : "0" (v.s.a), "1" (v.s.b));
 #else
 	v.s.a = ___arch__swab32(v.s.a);
 	v.s.b = ___arch__swab32(v.s.b);
-	asm("xchgl %0,%1" : "=r" (v.s.a), "=r" (v.s.b) : "0" (v.s.a), "1" (v.s.b));
+	__asm__("xchgl %0,%1" : "=r" (v.s.a), "=r" (v.s.b) : "0" (v.s.a), "1" (v.s.b));
 #endif
 	return v.u;
 }

Re: asm-x86/byteorder.h: clean up for userspace

[Ingo Molnar]

* Mike Frysinger <vapier@gentoo.org> wrote:

> Since asm-x86/byteorder.h is exported to userspace, use __asm__ rather 
> than asm in its code.

thanks Mike, i have applied your patch to x86.git - as a fix to be 
pushed into 2.6.24.

	Ingo

Re: asm-x86/byteorder.h: clean up for userspace

[Christoph Hellwig]

On Mon, Dec 31, 2007 at 01:12:45PM -0500, Mike Frysinger wrote:
> Since asm-x86/byteorder.h is exported to userspace, use __asm__ rather than
> asm in its code.

The correct fix is to not export it to userspace.

Re: asm-x86/byteorder.h: clean up for userspace

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 546 bytes --]

On Tuesday 01 January 2008, Christoph Hellwig wrote:
> On Mon, Dec 31, 2007 at 01:12:45PM -0500, Mike Frysinger wrote:
> > Since asm-x86/byteorder.h is exported to userspace, use __asm__ rather
> > than asm in its code.
>
> The correct fix is to not export it to userspace.

which is what people complained last time and upon further investigation, it 
isnt trivial to remove.  so until then, it should be made clean.  especially 
considering the proposed changes are perfectly fine when used inside and 
outside of the kernel.
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 835 bytes --]

Re: asm-x86/byteorder.h: clean up for userspace

[H. Peter Anvin]

Christoph Hellwig wrote:
> On Mon, Dec 31, 2007 at 01:12:45PM -0500, Mike Frysinger wrote:
>> Since asm-x86/byteorder.h is exported to userspace, use __asm__ rather than
>> asm in its code.
> 
> The correct fix is to not export it to userspace.

Unfortunately, it's historically been provided, for over 15 years.  It's 
also trivial to export, without funnies, and it's *useful* to userspace, 
as it provides an interface sorely lacking from the stock libc interfaces.

So please do provide it, even though it's arguably the wrong thing.  The 
downside is worse.

	-hpa

Re: asm-x86/byteorder.h: clean up for userspace

[Adrian Bunk]

On Tue, Jan 01, 2008 at 01:19:42PM -0800, H. Peter Anvin wrote:
> Christoph Hellwig wrote:
>> On Mon, Dec 31, 2007 at 01:12:45PM -0500, Mike Frysinger wrote:
>>> Since asm-x86/byteorder.h is exported to userspace, use __asm__ rather than
>>> asm in its code.
>>
>> The correct fix is to not export it to userspace.
>
> Unfortunately, it's historically been provided, for over 15 years.  It's 
> also trivial to export, without funnies, and it's *useful* to userspace, as 
> it provides an interface sorely lacking from the stock libc interfaces.
>...

Userspace either has to #define CONFIG_X86_BSWAP or it'll get the slow 
versions of these functions...

Leaking CONFIG_ variables to userspace is not really funny - I remember 
e.g. what tricks MySQL does (did?) for (ab)using asm-i386/atomic.h in 
userspace.

> 	-hpa

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: asm-x86/byteorder.h: clean up for userspace

[H. Peter Anvin]

Adrian Bunk wrote:
> 
> Userspace either has to #define CONFIG_X86_BSWAP or it'll get the slow 
> versions of these functions...
> 
> Leaking CONFIG_ variables to userspace is not really funny - I remember 
> e.g. what tricks MySQL does (did?) for (ab)using asm-i386/atomic.h in 
> userspace.
> 

True.  CONFIG_X86_BSWAP isn't appropriate for userspace anyway, 
userspace needs to use the appropriate gcc intrinsics for 486+.

	-hpa