Re: libcrypto: fix assembly for clang

[Brent Cook]

On Wed, Jan 4, 2017 at 5:12 AM, Mark Kettenis 
wrote:

> > Date: Wed, 4 Jan 2017 11:04:43 +0100
> > From: Patrick Wildt 
> >
> > On Tue, Jan 03, 2017 at 02:16:25PM -0500, Ted Unangst wrote:
> > > Patrick Wildt wrote:
> > > >
> > > > Maybe they want to support older binutils that do not support the
> unified syntax?
> > > > What's our policy there?
> > >
> > > OpenBSD is just about the oldest binutils around. I wouldn't worry
> about
> > > anything older than what we have in base.
> >
> > Considering that, I'm going to ask for OKs then.
> >
> > ok?
>
> ok kettenis@, but lets give the LibreSSL people a chance to raise any
> portability concerns as they might have a better idea about LibreSSL
> portable deployment on ARM.
>
> ​
ok bcook@ - we don't even enable assembly on ARM in portable. Happy if
someone wants to help change that.

Re: libcrypto: fix assembly for clang

[Mark Kettenis]

> Date: Wed, 4 Jan 2017 11:04:43 +0100
> From: Patrick Wildt 
> 
> On Tue, Jan 03, 2017 at 02:16:25PM -0500, Ted Unangst wrote:
> > Patrick Wildt wrote:
> > > 
> > > Maybe they want to support older binutils that do not support the unified 
> > > syntax?
> > > What's our policy there?
> > 
> > OpenBSD is just about the oldest binutils around. I wouldn't worry about
> > anything older than what we have in base.
> 
> Considering that, I'm going to ask for OKs then.
> 
> ok?

ok kettenis@, but lets give the LibreSSL people a chance to raise any
portability concerns as they might have a better idea about LibreSSL
portable deployment on ARM.

Re: libcrypto: fix assembly for clang

[Patrick Wildt]

On Tue, Jan 03, 2017 at 02:16:25PM -0500, Ted Unangst wrote:
> Patrick Wildt wrote:
> > 
> > Maybe they want to support older binutils that do not support the unified 
> > syntax?
> > What's our policy there?
> 
> OpenBSD is just about the oldest binutils around. I wouldn't worry about
> anything older than what we have in base.

Considering that, I'm going to ask for OKs then.

ok?

Re: libcrypto: fix assembly for clang

[Ted Unangst]

Patrick Wildt wrote:
> 
> Maybe they want to support older binutils that do not support the unified 
> syntax?
> What's our policy there?

OpenBSD is just about the oldest binutils around. I wouldn't worry about
anything older than what we have in base.

libcrypto: fix assembly for clang

[Patrick Wildt]

Hi,

clang complains about the ARM assembly, since it expects the conditional
branch instructions to state the condition (in this case pl/ne) after
the "b" for branch.  We also need to state that we're using the unified
syntax, so that binutils 2.17 likes it as well.

OpenSSL fixed that issue in another way:

.text
#if defined(__thumb2__) || defined(__clang__)
.syntax unified
#endif
#if defined(__thumb2__)
.thumb
#else
.code   32
#endif

#ifdef  __clang__
#define ldrplb  ldrbpl
#define ldrneb  ldrbne
#endif

Maybe they want to support older binutils that do not support the unified 
syntax?
What's our policy there?

Comments?

Patrick


diff --git a/lib/libcrypto/modes/asm/ghash-armv4.pl 
b/lib/libcrypto/modes/asm/ghash-armv4.pl
index d91586ee292..4f8372d8974 100644
--- a/lib/libcrypto/modes/asm/ghash-armv4.pl
+++ b/lib/libcrypto/modes/asm/ghash-armv4.pl
@@ -110,6 +110,7 @@ $code=<<___;
 #include "arm_arch.h"
 
 .text
+.syntaxunified
 .code  32
 
 .type  rem_4bit,%object
@@ -182,7 +183,7 @@ gcm_ghash_4bit:
eor $Zlh,$Zlh,$Zhl,lsl#28
ldrh$Tll,[sp,$nlo]  @ rem_4bit[rem]
eor $Zhl,$Thl,$Zhl,lsr#4
-   ldrplb  $nlo,[$inp,$cnt]
+   ldrbpl  $nlo,[$inp,$cnt]
eor $Zhl,$Zhl,$Zhh,lsl#28
eor $Zhh,$Thh,$Zhh,lsr#4
 
@@ -192,7 +193,7 @@ gcm_ghash_4bit:
add $nhi,$nhi,$nhi
ldmia   $Thh,{$Tll-$Thh}@ load Htbl[nhi]
eor $Zll,$Tll,$Zll,lsr#4
-   ldrplb  $Tll,[$Xi,$cnt]
+   ldrbpl  $Tll,[$Xi,$cnt]
eor $Zll,$Zll,$Zlh,lsl#28
eor $Zlh,$Tlh,$Zlh,lsr#4
ldrh$Tlh,[sp,$nhi]
@@ -210,7 +211,7 @@ gcm_ghash_4bit:
add $inp,$inp,#16
mov $nhi,$Zll
 ___
-   ("cmp\t$inp,$len","ldrneb\t$nlo,[$inp,#15]");
+   ("cmp\t$inp,$len","ldrbne\t$nlo,[$inp,#15]");
 $code.=<<___;
bne .Louter
 
@@ -268,7 +269,7 @@ gcm_gmult_4bit:
eor $Zlh,$Zlh,$Zhl,lsl#28
ldrh$Tll,[$rem_4bit,$nlo]   @ rem_4bit[rem]
eor $Zhl,$Thl,$Zhl,lsr#4
-   ldrplb  $nlo,[$Xi,$cnt]
+   ldrbpl  $nlo,[$Xi,$cnt]
eor $Zhl,$Zhl,$Zhh,lsl#28
eor $Zhh,$Thh,$Zhh,lsr#4