From: Alexey Moksyakov Date: Thu, 16 Jan 2025 11:21:30 +0000 (+0300) Subject: Move rodata to .rodata section for armv8 X-Git-Tag: openssl-3.5.0-alpha1~728 X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=5b36728d974578f2c74e9f3d2ee6472187084882;p=thirdparty%2Fopenssl.git Move rodata to .rodata section for armv8 Adds missing files where asm code is generated by perl scripts and read only constant is used PR #24137 closes #23312 Signed-off-by: Alexey Moksyakov Reviewed-by: Saša Nedvědický Reviewed-by: Tom Cosgrove Reviewed-by: Tomas Mraz (Merged from https://github.com/openssl/openssl/pull/26440) --- diff --git a/crypto/sm4/asm/sm4-armv8.pl b/crypto/sm4/asm/sm4-armv8.pl index 7358a6e6a2c..0bf27ddb04d 100755 --- a/crypto/sm4/asm/sm4-armv8.pl +++ b/crypto/sm4/asm/sm4-armv8.pl @@ -118,7 +118,10 @@ ___ {{{ $code.=<<___; +.rodata .align 6 +.type _${prefix}_consts,%object +_${prefix}_consts: .Lck: .long 0x00070E15, 0x1C232A31, 0x383F464D, 0x545B6269 .long 0x70777E85, 0x8C939AA1, 0xA8AFB6BD, 0xC4CBD2D9 @@ -130,6 +133,9 @@ $code.=<<___; .long 0x10171E25, 0x2C333A41, 0x484F565D, 0x646B7279 .Lfk: .long 0xa3b1bac6, 0x56aa3350, 0x677d9197, 0xb27022dc +.size _${prefix}_consts,.-_${prefix}_consts + +.previous ___ }}} @@ -146,9 +152,11 @@ $code.=<<___; ${prefix}_set_encrypt_key: AARCH64_VALID_CALL_TARGET ld1 {$key0.4s},[$key] - adr $tmp,.Lfk + adrp $tmp, _${prefix}_consts + add $tmp,$tmp,#:lo12:.Lfk ld1 {$fkconst.4s},[$tmp] - adr $tmp,.Lck + adrp $tmp, _${prefix}_consts + add $tmp,$tmp,#:lo12:.Lck ld1 {$const0.4s,$const1.4s,$const2.4s,$const3.4s},[$tmp],64 ___ &rev32($key0, $key0); @@ -183,9 +191,11 @@ $code.=<<___; ${prefix}_set_decrypt_key: AARCH64_VALID_CALL_TARGET ld1 {$key0.4s},[$key] - adr $tmp,.Lfk + adrp $tmp, _${prefix}_consts + add $tmp,$tmp,#:lo12:.Lfk ld1 {$fkconst.4s},[$tmp] - adr $tmp, .Lck + adrp $tmp, _${prefix}_consts + add $tmp,$tmp,#:lo12:.Lck ld1 {$const0.4s,$const1.4s,$const2.4s,$const3.4s},[$tmp],64 ___ &rev32($key0, $key0); diff --git a/crypto/sm4/asm/vpsm4-armv8.pl b/crypto/sm4/asm/vpsm4-armv8.pl index 78357676a56..115bbc76da4 100755 --- a/crypto/sm4/asm/vpsm4-armv8.pl +++ b/crypto/sm4/asm/vpsm4-armv8.pl @@ -474,7 +474,8 @@ sub load_sbox () { my $data = shift; $code.=<<___; - adr $ptr,.Lsbox + adrp $ptr, _${prefix}_consts + add $ptr,$ptr,#:lo12:.Lsbox ld1 {@sbox[0].16b,@sbox[1].16b,@sbox[2].16b,@sbox[3].16b},[$ptr],#64 ld1 {@sbox[4].16b,@sbox[5].16b,@sbox[6].16b,@sbox[7].16b},[$ptr],#64 ld1 {@sbox[8].16b,@sbox[9].16b,@sbox[10].16b,@sbox[11].16b},[$ptr],#64 @@ -524,7 +525,8 @@ sub compute_tweak_vec() { my $std = shift; &rbit(@vtmp[2],$src,$std); $code.=<<___; - ldr @qtmp[0], .Lxts_magic + adrp $xtmp2, _${prefix}_consts + ldr @qtmp[0], [$xtmp2,#:lo12:.Lxts_magic] shl $des.16b, @vtmp[2].16b, #1 ext @vtmp[1].16b, @vtmp[2].16b, @vtmp[2].16b,#15 ushr @vtmp[1].16b, @vtmp[1].16b, #7 @@ -539,9 +541,10 @@ $code=<<___; .arch armv8-a .text -.type _vpsm4_consts,%object +.rodata +.type _${prefix}_consts,%object .align 7 -_vpsm4_consts: +_${prefix}_consts: .Lsbox: .byte 0xD6,0x90,0xE9,0xFE,0xCC,0xE1,0x3D,0xB7,0x16,0xB6,0x14,0xC2,0x28,0xFB,0x2C,0x05 .byte 0x2B,0x67,0x9A,0x76,0x2A,0xBE,0x04,0xC3,0xAA,0x44,0x13,0x26,0x49,0x86,0x06,0x99 @@ -575,7 +578,8 @@ _vpsm4_consts: .Lxts_magic: .quad 0x0101010101010187,0x0101010101010101 -.size _vpsm4_consts,.-_vpsm4_consts +.size _${prefix}_consts,.-_${prefix}_consts +.previous ___ {{{ @@ -592,13 +596,16 @@ ___ &load_sbox(); &rev32($vkey,$vkey); $code.=<<___; - adr $pointer,.Lshuffles + adrp $pointer, _${prefix}_consts + add $pointer,$pointer,#:lo12:.Lshuffles ld1 {$vmap.2d},[$pointer] - adr $pointer,.Lfk + adrp $pointer, _${prefix}_consts + add $pointer,$pointer,#:lo12:.Lfk ld1 {$vfk.2d},[$pointer] eor $vkey.16b,$vkey.16b,$vfk.16b mov $schedules,#32 - adr $pointer,.Lck + adrp $pointer, _${prefix}_consts + add $pointer,$pointer,#:lo12:.Lck movi @vtmp[0].16b,#64 cbnz $enc,1f add $keys,$keys,124 diff --git a/crypto/sm4/asm/vpsm4_ex-armv8.pl b/crypto/sm4/asm/vpsm4_ex-armv8.pl index f4bcdbad1bc..2bbdb3e5b5b 100644 --- a/crypto/sm4/asm/vpsm4_ex-armv8.pl +++ b/crypto/sm4/asm/vpsm4_ex-armv8.pl @@ -475,12 +475,13 @@ sub load_sbox () { my $data = shift; $code.=<<___; - ldr $MaskQ, .Lsbox_magic - ldr $TAHMatQ, .Lsbox_magic+16 - ldr $TALMatQ, .Lsbox_magic+32 - ldr $ATAHMatQ, .Lsbox_magic+48 - ldr $ATALMatQ, .Lsbox_magic+64 - ldr $ANDMaskQ, .Lsbox_magic+80 + adrp $xtmp1, _${prefix}_consts + ldr $MaskQ, [$xtmp1, #:lo12:.Lsbox_magic] + ldr $TAHMatQ, [$xtmp1, #:lo12:.Lsbox_magic+16] + ldr $TALMatQ, [$xtmp1, #:lo12:.Lsbox_magic+32] + ldr $ATAHMatQ, [$xtmp1, #:lo12:.Lsbox_magic+48] + ldr $ATALMatQ, [$xtmp1, #:lo12:.Lsbox_magic+64] + ldr $ANDMaskQ, [$xtmp1, #:lo12:.Lsbox_magic+80] ___ } @@ -525,7 +526,8 @@ sub compute_tweak_vec() { my $std = shift; &rbit(@vtmp[2],$src,$std); $code.=<<___; - ldr @qtmp[0], .Lxts_magic + adrp $xtmp2, _${prefix}_consts + ldr @qtmp[0], [$xtmp2, #:lo12:.Lxts_magic] shl $des.16b, @vtmp[2].16b, #1 ext @vtmp[1].16b, @vtmp[2].16b, @vtmp[2].16b,#15 ushr @vtmp[1].16b, @vtmp[1].16b, #7 @@ -540,6 +542,7 @@ $code=<<___; .arch armv8-a+crypto .text +.rodata .type _${prefix}_consts,%object .align 7 _${prefix}_consts: @@ -567,6 +570,7 @@ _${prefix}_consts: .quad 0x0f0f0f0f0f0f0f0f,0x0f0f0f0f0f0f0f0f .size _${prefix}_consts,.-_${prefix}_consts +.previous ___ {{{ @@ -583,13 +587,16 @@ ___ &load_sbox(); &rev32($vkey,$vkey); $code.=<<___; - adr $pointer,.Lshuffles + adrp $pointer,_${prefix}_consts + add $pointer,$pointer,#:lo12:.Lshuffles ld1 {$vmap.2d},[$pointer] - adr $pointer,.Lfk + adrp $pointer,_${prefix}_consts + add $pointer,$pointer,#:lo12:.Lfk ld1 {$vfk.2d},[$pointer] eor $vkey.16b,$vkey.16b,$vfk.16b mov $schedules,#32 - adr $pointer,.Lck + adrp $pointer,_${prefix}_consts + add $pointer,$pointer,#:lo12:.Lck movi @vtmp[0].16b,#64 cbnz $enc,1f add $keys,$keys,124