Put arm/aarch64 assembly functions in their own section.
This change causes each global arm or aarch64 asm function to be put
into its own section by default. This matches the behaviour of the
-ffunction-sections option to GCC and allows the --gc-sections option to
the linker to discard unused asm functions on a function-by-function
basis.
Sometimes several asm functions will share the same data an, in that
situation, the data is put into the section of one of the functions and
the section of the other function is merged with the added
“.global_with_section” directive.
Change-Id: I12c9b844d48d104d28beb816764358551eac4456
Reviewed-on: https://boringssl-review.googlesource.com/6003
Reviewed-by: Adam Langley <agl@google.com>
diff --git a/crypto/aes/asm/aes-armv4.pl b/crypto/aes/asm/aes-armv4.pl
index 882017a..16ff990 100644
--- a/crypto/aes/asm/aes-armv4.pl
+++ b/crypto/aes/asm/aes-armv4.pl
@@ -82,6 +82,7 @@
# endif
#endif
+.pushsection .text.asm_AES_encrypt,"ax",%progbits
.type AES_Te,%object
.align 5
AES_Te:
@@ -187,6 +188,7 @@
.word 0x10000000, 0x20000000, 0x40000000, 0x80000000
.word 0x1B000000, 0x36000000, 0, 0, 0, 0, 0, 0
.size AES_Te,.-AES_Te
+.popsection
@ void asm_AES_encrypt(const unsigned char *in, unsigned char *out,
@ const AES_KEY *key) {
@@ -440,7 +442,7 @@
ldr pc,[sp],#4 @ pop and return
.size _armv4_AES_encrypt,.-_armv4_AES_encrypt
-.global asm_AES_set_encrypt_key
+.global_with_section asm_AES_set_encrypt_key, asm_AES_encrypt
.hidden asm_AES_set_encrypt_key
.type asm_AES_set_encrypt_key,%function
.align 5
@@ -866,6 +868,7 @@
#endif
.size AES_set_enc2dec_key,.-AES_set_enc2dec_key
+.pushsection .text.asm_AES_decrypt,"ax",%progbits
.type AES_Td,%object
.align 5
AES_Td:
@@ -967,6 +970,7 @@
.byte 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26
.byte 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d
.size AES_Td,.-AES_Td
+.popsection
@ void asm_AES_decrypt(const unsigned char *in, unsigned char *out,
@ const AES_KEY *key) {
diff --git a/crypto/aes/asm/aesv8-armx.pl b/crypto/aes/asm/aesv8-armx.pl
index 121154a..d770fa3 100644
--- a/crypto/aes/asm/aesv8-armx.pl
+++ b/crypto/aes/asm/aesv8-armx.pl
@@ -72,10 +72,12 @@
$code.=<<___;
.align 5
+.pushsection .text.${prefix}_set_encrypt_key,"ax",%progbits
.Lrcon:
.long 0x01,0x01,0x01,0x01
.long 0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d // rotate-n-splat
.long 0x1b,0x1b,0x1b,0x1b
+.popsection
.globl ${prefix}_set_encrypt_key
.type ${prefix}_set_encrypt_key,%function
diff --git a/crypto/aes/asm/bsaes-armv7.pl b/crypto/aes/asm/bsaes-armv7.pl
index 2617fef..5fa3005 100644
--- a/crypto/aes/asm/bsaes-armv7.pl
+++ b/crypto/aes/asm/bsaes-armv7.pl
@@ -2049,7 +2049,7 @@
.size bsaes_xts_encrypt,.-bsaes_xts_encrypt
-.globl bsaes_xts_decrypt
+.global_with_section bsaes_xts_decrypt, bsaes_xts_encrypt
.hidden bsaes_xts_decrypt
.type bsaes_xts_decrypt,%function
.align 4
diff --git a/crypto/bn/asm/armv4-mont.pl b/crypto/bn/asm/armv4-mont.pl
index 4206fd8..c464a5b 100644
--- a/crypto/bn/asm/armv4-mont.pl
+++ b/crypto/bn/asm/armv4-mont.pl
@@ -86,8 +86,10 @@
#if __ARM_MAX_ARCH__>=7
.align 5
+.pushsection .text.bn_mul_mont,"ax",%progbits
.LOPENSSL_armcap:
.word OPENSSL_armcap_P-.Lbn_mul_mont
+.popsection
#endif
.global bn_mul_mont
diff --git a/crypto/perlasm/arm-xlate.pl b/crypto/perlasm/arm-xlate.pl
index 706fa70..32eeb25 100755
--- a/crypto/perlasm/arm-xlate.pl
+++ b/crypto/perlasm/arm-xlate.pl
@@ -18,6 +18,14 @@
if ($flavour =~ /linux/) { ".arch\t".join(',',@_); }
else { ""; }
};
+my $pushsection = sub {
+ if ($flavour !~ /ios/) { ".pushsection\t".join(',',@_); }
+ else { ""; }
+};
+my $popsection = sub {
+ if ($flavour !~ /ios/) { ".popsection"; }
+ else { ""; }
+};
my $fpu = sub {
if ($flavour =~ /linux/) { ".fpu\t".join(',',@_); }
else { ""; }
@@ -44,7 +52,8 @@
$$global = $name;
$ret;
};
-my $globl = sub {
+my $in_section = 0;
+my $raw_globl = sub {
my $name = shift;
my $global = \$GLOBALS{$name};
my $ret;
@@ -59,7 +68,39 @@
$$global = $name;
$ret;
};
+my $globl = sub {
+ my $name = shift;
+
+ if ($flavour !~ /ios/) {
+ if ($in_section == 1) {
+ printf ".popsection\n";
+ }
+ $in_section = 1;
+
+ printf ".pushsection .text.$name,\"ax\",%%progbits\n";
+ }
+ return $raw_globl->($name);
+
+};
my $global = $globl;
+my $global_with_section = sub {
+ my $arg = shift;
+
+ if ($arg =~ m/^([^,]*), *(.*)/) {
+ if ($flavour !~ /ios/) {
+ if ($in_section == 1) {
+ printf ".popsection\n";
+ }
+ $in_section = 1;
+
+ printf ".pushsection .text.$2,\"ax\",%%progbits\n";
+ }
+ return $raw_globl->($1);
+ } else {
+ printf STDERR "Expected two arguments to global_with_section\n";
+ exit(1);
+ }
+};
my $extern = sub {
&$globl(@_);
return; # return nothing
diff --git a/crypto/sha/asm/sha256-armv4.pl b/crypto/sha/asm/sha256-armv4.pl
index 7e07147..b293f84 100644
--- a/crypto/sha/asm/sha256-armv4.pl
+++ b/crypto/sha/asm/sha256-armv4.pl
@@ -187,6 +187,7 @@
# endif
#endif
+.pushsection .text.sha256_block_data_order,"ax",%progbits
.type K256,%object
.align 5
K256:
@@ -212,6 +213,7 @@
.LOPENSSL_armcap:
.word OPENSSL_armcap_P-.Lsha256_block_data_order
#endif
+.popsection
.align 5
.global sha256_block_data_order
@@ -471,7 +473,7 @@
.arch armv7-a
.fpu neon
-.global sha256_block_data_order_neon
+.global_with_section sha256_block_data_order_neon, sha256_block_data_order
.type sha256_block_data_order_neon,%function
.align 4
sha256_block_data_order_neon:
diff --git a/crypto/sha/asm/sha512-armv4.pl b/crypto/sha/asm/sha512-armv4.pl
index cd3662a..4c1cbae 100644
--- a/crypto/sha/asm/sha512-armv4.pl
+++ b/crypto/sha/asm/sha512-armv4.pl
@@ -224,6 +224,7 @@
# endif
#endif
+.pushsection .text.sha512_block_data_order,"ax",%progbits
.type K512,%object
.align 5
K512:
@@ -275,6 +276,7 @@
#else
.skip 32
#endif
+.popsection
.global sha512_block_data_order
.type sha512_block_data_order,%function
@@ -602,7 +604,7 @@
.arch armv7-a
.fpu neon
-.global sha512_block_data_order_neon
+.global_with_section sha512_block_data_order_neon, sha512_block_data_order
.type sha512_block_data_order_neon,%function
.align 4
sha512_block_data_order_neon:
