Move all data tables from .text section to .rodata, and update the code to

fetch them correctly when building PIC. Also drop unused data, and remove
--no-execute-only from linker flags.

ok kettenis@

8 files changed, 109 insertions, 41 deletions

diff --git a/lib/libcrypto/aes/asm/aes-sparcv9.pl b/lib/libcrypto/aes/asm/aes-sparcv9.pl
index c57b3a2d6d3..08ac9830f14 100755
--- a/lib/libcrypto/aes/asm/aes-sparcv9.pl
+++ b/lib/libcrypto/aes/asm/aes-sparcv9.pl
@@ -79,7 +79,7 @@ $code.=<<___ if ($bits==64);
 .register	%g3,#scratch
 ___
 $code.=<<___;
-.section	".text",#alloc,#execinstr
+.section	".rodata",#alloc
 
 .align	256
 AES_Te:
@@ -185,6 +185,7 @@ $code.=<<___;
 .type	AES_Te,#object
 .size	AES_Te,(.-AES_Te)
 
+.section	".text",#alloc,#execinstr
 .align	64
 .skip	16
 _sparcv9_AES_encrypt:
@@ -512,18 +513,30 @@ _sparcv9_AES_encrypt:
 .align	32
 .globl	AES_encrypt
 AES_encrypt:
-	or	%o0,%o1,%g1
+	save	%sp,-$frame,%sp
+#ifdef __PIC__
+	sethi	%hi(_GLOBAL_OFFSET_TABLE_-4), %o5
+	rd	%pc, %o4
+	or	%o5, %lo(_GLOBAL_OFFSET_TABLE_+4), %o5
+	add	%o5, %o4, %o5
+#endif
+
+	or	%i0,%i1,%g1
 	andcc	%g1,3,%g0
 	bnz,pn	%xcc,.Lunaligned_enc
-	save	%sp,-$frame,%sp
+	 nop
 
 	ld	[%i0+0],%o0
 	ld	[%i0+4],%o1
 	ld	[%i0+8],%o2
 	ld	[%i0+12],%o3
 
-1:	call	.+8
-	add	%o7,AES_Te-1b,%o4
+#ifdef __PIC__
+	set	AES_Te, %o4
+	ldx	[%o4+%o5], %o4
+#else
+	set	AES_Te, %o4
+#endif
 	call	_sparcv9_AES_encrypt
 	mov	%i2,%o5
 
@@ -582,8 +595,12 @@ AES_encrypt:
 	or	%l7,%l6,%l6
 	or	%l4,%l6,%o3
 
-1:	call	.+8
-	add	%o7,AES_Te-1b,%o4
+#ifdef __PIC__
+	set	AES_Te, %o4
+	ldx	[%o4+%o5], %o4
+#else
+	set	AES_Te, %o4
+#endif
 	call	_sparcv9_AES_encrypt
 	mov	%i2,%o5
 
@@ -627,6 +644,7 @@ AES_encrypt:
 ___
 
 $code.=<<___;
+.section	".rodata",#alloc
 .align	256
 AES_Td:
 ___
@@ -731,6 +749,7 @@ $code.=<<___;
 .type	AES_Td,#object
 .size	AES_Td,(.-AES_Td)
 
+.section	".text",#alloc,#execinstr
 .align	64
 .skip	16
 _sparcv9_AES_decrypt:
@@ -1058,18 +1077,30 @@ _sparcv9_AES_decrypt:
 .align	32
 .globl	AES_decrypt
 AES_decrypt:
-	or	%o0,%o1,%g1
+	save	%sp,-$frame,%sp
+#ifdef __PIC__
+	sethi	%hi(_GLOBAL_OFFSET_TABLE_-4), %o5
+	rd	%pc, %o4
+	or	%o5, %lo(_GLOBAL_OFFSET_TABLE_+4), %o5
+	add	%o5, %o4, %o5
+#endif
+
+	or	%i0,%i1,%g1
 	andcc	%g1,3,%g0
 	bnz,pn	%xcc,.Lunaligned_dec
-	save	%sp,-$frame,%sp
+	 nop
 
 	ld	[%i0+0],%o0
 	ld	[%i0+4],%o1
 	ld	[%i0+8],%o2
 	ld	[%i0+12],%o3
 
-1:	call	.+8
-	add	%o7,AES_Td-1b,%o4
+#ifdef __PIC__
+	set	AES_Td, %o4
+	ldx	[%o4+%o5], %o4
+#else
+	set	AES_Td, %o4
+#endif
 	call	_sparcv9_AES_decrypt
 	mov	%i2,%o5
 
@@ -1128,8 +1159,12 @@ AES_decrypt:
 	or	%l7,%l6,%l6
 	or	%l4,%l6,%o3
 
-1:	call	.+8
-	add	%o7,AES_Td-1b,%o4
+#ifdef __PIC__
+	set	AES_Td, %o4
+	ldx	[%o4+%o5], %o4
+#else
+	set	AES_Td, %o4
+#endif
 	call	_sparcv9_AES_decrypt
 	mov	%i2,%o5
 
@@ -1176,6 +1211,7 @@ ___
 # As UltraSPARC T1, a.k.a. Niagara, has shared FPU, FP nops can have
 # undesired effect, so just omit them and sacrifice some portion of
 # percent in performance...
-$code =~ s/fmovs.*$//gem;
+$code =~ s/fmovs.*$//gm;
 
 print $code;
+close STDOUT;	# ensure flush
diff --git a/lib/libcrypto/arch/sparc64/Makefile.inc b/lib/libcrypto/arch/sparc64/Makefile.inc
index cb502b28a38..223c941f77e 100644
--- a/lib/libcrypto/arch/sparc64/Makefile.inc
+++ b/lib/libcrypto/arch/sparc64/Makefile.inc
@@ -1,9 +1,7 @@
-# $OpenBSD: Makefile.inc,v 1.5 2023/01/11 16:36:44 deraadt Exp $
+# $OpenBSD: Makefile.inc,v 1.6 2023/01/13 17:27:26 miod Exp $
 
 # sparc64-specific libcrypto build rules
 
-LDADD+= -Wl,--no-execute-only
-
 # aes
 SRCS+= aes_core.c aes_cbc.c 
 CFLAGS+= -DAES_ASM
diff --git a/lib/libcrypto/bn/asm/sparcv9-mont.pl b/lib/libcrypto/bn/asm/sparcv9-mont.pl
index fb44c014431..610ec1a968a 100644
--- a/lib/libcrypto/bn/asm/sparcv9-mont.pl
+++ b/lib/libcrypto/bn/asm/sparcv9-mont.pl
@@ -598,8 +598,6 @@ $code.=<<___;
 	add	$tp,8,$tp
 .type	$fname,#function
 .size	$fname,(.-$fname)
-.asciz	"Montgomery Multipltication for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
-.align	32
 ___
 $code =~ s/\`([^\`]*)\`/eval($1)/gem;
 print $code;
diff --git a/lib/libcrypto/bn/asm/sparcv9a-mont.pl b/lib/libcrypto/bn/asm/sparcv9a-mont.pl
index 4ebe15c5c01..7bb1725a0ed 100755
--- a/lib/libcrypto/bn/asm/sparcv9a-mont.pl
+++ b/lib/libcrypto/bn/asm/sparcv9a-mont.pl
@@ -860,8 +860,6 @@ $fname:
 	restore
 .type   $fname,#function
 .size	$fname,(.-$fname)
-.asciz	"Montgomery Multipltication for UltraSPARC, CRYPTOGAMS by <appro\@openssl.org>"
-.align	32
 ___
 
 $code =~ s/\`([^\`]*)\`/eval($1)/gem;
diff --git a/lib/libcrypto/modes/asm/ghash-sparcv9.pl b/lib/libcrypto/modes/asm/ghash-sparcv9.pl
index 70e7b044a3e..ce75045f096 100644
--- a/lib/libcrypto/modes/asm/ghash-sparcv9.pl
+++ b/lib/libcrypto/modes/asm/ghash-sparcv9.pl
@@ -67,7 +67,7 @@ $inp="%i2";
 $len="%i3";
 
 $code.=<<___;
-.section	".text",#alloc,#execinstr
+.section	".rodata",#alloc
 
 .align	64
 rem_4bit:
@@ -78,18 +78,30 @@ rem_4bit:
 .type	rem_4bit,#object
 .size	rem_4bit,(.-rem_4bit)
 
+.section	".text",#alloc,#execinstr
 .globl	gcm_ghash_4bit
 .align	32
 gcm_ghash_4bit:
 	save	%sp,-$frame,%sp
+#ifdef __PIC__
+	sethi	%hi(_GLOBAL_OFFSET_TABLE_-4), $tmp
+	rd	%pc, $rem
+	or	$tmp, %lo(_GLOBAL_OFFSET_TABLE_+4), $tmp
+	add	$tmp, $rem, $tmp
+#endif
+
 	ldub	[$inp+15],$nlo
 	ldub	[$Xi+15],$xi0
 	ldub	[$Xi+14],$xi1
 	add	$len,$inp,$len
 	add	$Htbl,8,$Htblo
 
-1:	call	.+8
-	add	%o7,rem_4bit-1b,$rem_4bit
+#ifdef __PIC__
+	set	rem_4bit, $rem_4bit
+	ldx	[$rem_4bit+$tmp], $rem_4bit
+#else
+	set	rem_4bit, $rem_4bit
+#endif
 
 .Louter:
 	xor	$xi0,$nlo,$nlo
@@ -223,11 +235,22 @@ $code.=<<___;
 .align	32
 gcm_gmult_4bit:
 	save	%sp,-$frame,%sp
+#ifdef __PIC__
+	sethi	%hi(_GLOBAL_OFFSET_TABLE_-4), $tmp
+	rd	%pc, $rem
+	or	$tmp, %lo(_GLOBAL_OFFSET_TABLE_+4), $tmp
+	add	$tmp, $rem, $tmp
+#endif
+
 	ldub	[$Xi+15],$nlo
 	add	$Htbl,8,$Htblo
 
-1:	call	.+8
-	add	%o7,rem_4bit-1b,$rem_4bit
+#ifdef __PIC__
+	set	rem_4bit, $rem_4bit
+	ldx	[$rem_4bit+$tmp], $rem_4bit
+#else
+	set	rem_4bit, $rem_4bit
+#endif
 
 	and	$nlo,0xf0,$nhi
 	and	$nlo,0x0f,$nlo
@@ -321,8 +344,6 @@ gcm_gmult_4bit:
 	restore
 .type	gcm_gmult_4bit,#function
 .size	gcm_gmult_4bit,(.-gcm_gmult_4bit)
-.asciz	"GHASH for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
-.align	4
 ___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
diff --git a/lib/libcrypto/sha/asm/sha1-sparcv9.pl b/lib/libcrypto/sha/asm/sha1-sparcv9.pl
index 8306fc88ccf..5235c59e634 100644
--- a/lib/libcrypto/sha/asm/sha1-sparcv9.pl
+++ b/lib/libcrypto/sha/asm/sha1-sparcv9.pl
@@ -275,7 +275,6 @@ $code.=<<___;
 	restore
 .type	sha1_block_data_order,#function
 .size	sha1_block_data_order,(.-sha1_block_data_order)
-.asciz	"SHA1 block transform for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
 ___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
diff --git a/lib/libcrypto/sha/asm/sha1-sparcv9a.pl b/lib/libcrypto/sha/asm/sha1-sparcv9a.pl
index e81a4dcb053..8e7674e9d23 100644
--- a/lib/libcrypto/sha/asm/sha1-sparcv9a.pl
+++ b/lib/libcrypto/sha/asm/sha1-sparcv9a.pl
@@ -396,7 +396,7 @@ ___
 }
 
 $code.=<<___;
-.section	".text",#alloc,#execinstr
+.section	".rodata",#alloc
 
 .align	64
 vis_const:
@@ -409,13 +409,22 @@ vis_const:
 .type	vis_const,#object
 .size	vis_const,(.-vis_const)
 
+.section	".text",#alloc,#execinstr
 .globl	sha1_block_data_order
 sha1_block_data_order:
 	save	%sp,-$frame,%sp
 	add	%fp,$bias-256,$base
 
-1:	call	.+8
-	add	%o7,vis_const-1b,$tmp0
+#ifdef __PIC__
+	sethi	%hi(_GLOBAL_OFFSET_TABLE_-4), %o5
+	rd	%pc, %o4
+	or	%o5, %lo(_GLOBAL_OFFSET_TABLE_+4), %o5
+	add	%o5, %o4, %o5
+	set	vis_const, %o4
+	ldx	[%o4+%o5], %o4
+#else
+	set	vis_const, %o4
+#endif
 
 	ldd	[$tmp0+0],$VK_00_19
 	ldd	[$tmp0+8],$VK_20_39
@@ -538,8 +547,6 @@ $code.=<<___;
 	restore
 .type	sha1_block_data_order,#function
 .size	sha1_block_data_order,(.-sha1_block_data_order)
-.asciz	"SHA1 block transform for SPARCv9a, CRYPTOGAMS by <appro\@openssl.org>"
-.align	4
 ___
 
 # Purpose of these subroutines is to explicitly encode VIS instructions,
diff --git a/lib/libcrypto/sha/asm/sha512-sparcv9.pl b/lib/libcrypto/sha/asm/sha512-sparcv9.pl
index 54241aab504..3c93799446a 100644
--- a/lib/libcrypto/sha/asm/sha512-sparcv9.pl
+++ b/lib/libcrypto/sha/asm/sha512-sparcv9.pl
@@ -305,9 +305,9 @@ $code.=<<___;
 	srlx	@X[(($i+9)/2)%8],32,$tmp1	! X[i+9]
 	xor	$tmp0,$tmp2,$tmp2		! sigma1(X[i+14])
 	srl	@X[($i/2)%8],0,$tmp0
+	add	$tmp2,$tmp1,$tmp1
 	add	$xi,$T1,$T1			! +=X[i]
 	xor	$tmp0,@X[($i/2)%8],@X[($i/2)%8]
-	add	$tmp2,$T1,$T1
 	add	$tmp1,$T1,$T1
 
 	srl	$T1,0,$T1
@@ -318,9 +318,9 @@ ___
 $code.=<<___;
 	srlx	@X[($i/2)%8],32,$tmp1		! X[i]
 	xor	$tmp0,$tmp2,$tmp2		! sigma1(X[i+14])
-	srl	@X[($i/2)%8],0,@X[($i/2)%8]
 	add	$xi,$T1,$T1			! +=X[i+9]
-	add	$tmp2,$T1,$T1
+	add	$tmp2,$tmp1,$tmp1
+	srl	@X[($i/2)%8],0,@X[($i/2)%8]
 	add	$tmp1,$T1,$T1
 
 	sllx	$T1,32,$tmp0
@@ -386,7 +386,7 @@ $code.=<<___ if ($bits==64);
 .register	%g3,#scratch
 ___
 $code.=<<___;
-.section	".text",#alloc,#execinstr
+.section	".rodata",#alloc
 
 .align	64
 K${label}:
@@ -457,9 +457,17 @@ ___
 }
 $code.=<<___;
 .size	K${label},.-K${label}
+
+.section	".text",#alloc,#execinstr
 .globl	sha${label}_block_data_order
 sha${label}_block_data_order:
 	save	%sp,`-$frame-$locals`,%sp
+#ifdef __PIC__
+	sethi	%hi(_GLOBAL_OFFSET_TABLE_-4), %o5
+	rd	%pc, %o4
+	or	%o5, %lo(_GLOBAL_OFFSET_TABLE_+4), %o5
+	add	%o5, %o4, %o5
+#endif
 	and	$inp,`$align-1`,$tmp31
 	sllx	$len,`log(16*$SZ)/log(2)`,$len
 	andn	$inp,`$align-1`,$inp
@@ -471,8 +479,12 @@ $code.=<<___ if ($SZ==8); # SHA512
 	sub	$tmp32,$tmp31,$tmp32
 ___
 $code.=<<___;
-.Lpic:	call	.+8
-	add	%o7,K${label}-.Lpic,$Ktbl
+#ifdef __PIC__
+	set	K${label}, $Ktbl
+	ldx	[$Ktbl+%o5], $Ktbl
+#else
+	set	K${label}, $Ktbl
+#endif
 
 	$LD	[$ctx+`0*$SZ`],$A
 	$LD	[$ctx+`1*$SZ`],$B
@@ -585,7 +597,6 @@ $code.=<<___;
 	restore
 .type	sha${label}_block_data_order,#function
 .size	sha${label}_block_data_order,(.-sha${label}_block_data_order)
-.asciz	"SHA${label} block transform for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
 ___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;