From: deraadt <deraadt@openbsd.org>
Date: Mon, 11 Dec 2017 05:48:01 +0000 (+0000)
Subject: http://repzret.org/p/repzret/
X-Git-Url: http://artulab.com/gitweb/?a=commitdiff_plain;h=3f624d5f689b2964adb044dbb5971a0310c8869d;p=openbsd

http://repzret.org/p/repzret/
My read of this:  Long time ago (Think Conan, not dinasaurs) during the race
to make speedier processors, a cpu vendor built a pipeline with a bad stall,
and proposed a tremendously hasky workaround.  A wizard adopted this into his
perl scroll, and failed to reflect later when no compiler adopted the practice.
This relic remains at the tail end of some functions in OpenSSL as
".byte 0xf3,0xc3".  Banish it straight to hell.
ok mlarkin, others also stared blankly
---

diff --git a/lib/libcrypto/aes/asm/aes-x86_64.pl b/lib/libcrypto/aes/asm/aes-x86_64.pl
index c37fd55648b..9072f603a94 100755
--- a/lib/libcrypto/aes/asm/aes-x86_64.pl
+++ b/lib/libcrypto/aes/asm/aes-x86_64.pl
@@ -352,7 +352,7 @@ ___
 ___
 	}
 $code.=<<___;
-	.byte	0xf3,0xc3			# rep ret
+	retq
 .size	_x86_64_AES_encrypt,.-_x86_64_AES_encrypt
 ___
 
@@ -580,7 +580,7 @@ $code.=<<___;
 	xor	4($key),$s1
 	xor	8($key),$s2
 	xor	12($key),$s3
-	.byte	0xf3,0xc3			# rep ret
+	retq
 .size	_x86_64_AES_encrypt_compact,.-_x86_64_AES_encrypt_compact
 ___
 
@@ -925,7 +925,7 @@ ___
 ___
 	}
 $code.=<<___;
-	.byte	0xf3,0xc3			# rep ret
+	retq
 .size	_x86_64_AES_decrypt,.-_x86_64_AES_decrypt
 ___
 
@@ -1179,7 +1179,7 @@ $code.=<<___;
 	xor	4($key),$s1
 	xor	8($key),$s2
 	xor	12($key),$s3
-	.byte	0xf3,0xc3			# rep ret
+	retq
 .size	_x86_64_AES_decrypt_compact,.-_x86_64_AES_decrypt_compact
 ___
 
@@ -1496,7 +1496,7 @@ $code.=<<___;
 .Lbadpointer:
 	mov	\$-1,%rax
 .Lexit:
-	.byte	0xf3,0xc3			# rep ret
+	retq
 .size	_x86_64_AES_set_encrypt_key,.-_x86_64_AES_set_encrypt_key
 ___
 
diff --git a/lib/libcrypto/camellia/asm/cmll-x86_64.pl b/lib/libcrypto/camellia/asm/cmll-x86_64.pl
index a171c654b2d..df6bf11a288 100644
--- a/lib/libcrypto/camellia/asm/cmll-x86_64.pl
+++ b/lib/libcrypto/camellia/asm/cmll-x86_64.pl
@@ -218,7 +218,7 @@ $code.=<<___;
 	mov	$t2,@S[2]
 	mov	$t3,@S[3]
 
-	.byte	0xf3,0xc3		# rep ret
+	retq
 .size	_x86_64_Camellia_encrypt,.-_x86_64_Camellia_encrypt
 
 # V1.x API
@@ -329,7 +329,7 @@ $code.=<<___;
 	mov	$t0,@S[2]
 	mov	$t1,@S[3]
 
-	.byte	0xf3,0xc3		# rep ret
+	retq
 .size	_x86_64_Camellia_decrypt,.-_x86_64_Camellia_decrypt
 ___
 
diff --git a/lib/libcrypto/perlasm/x86_64-xlate.pl b/lib/libcrypto/perlasm/x86_64-xlate.pl
index a8393d27309..5f7f5855490 100755
--- a/lib/libcrypto/perlasm/x86_64-xlate.pl
+++ b/lib/libcrypto/perlasm/x86_64-xlate.pl
@@ -149,7 +149,7 @@ my %globals;
 		    $epilogue = "movq	8(%rsp),%rdi\n\t" .
 				"movq	16(%rsp),%rsi\n\t";
 		}
-	    	$epilogue . ".byte	0xf3,0xc3";
+	    	$epilogue . "retq";
 	    } elsif ($self->{op} eq "call" && !$elf && $current_segment eq ".init") {
 		".p2align\t3\n\t.quad";
 	    } else {