arm64: lib: Use .arch_extension
Subsystems (2)
PATCH 1/2 arm64: lib: Use .arch_extension
Currently when implementing optimised assembler routines using architecture extensions we override the base architecture along with enabling the new extensions, causing problems for in kernel BTI support which needs to raise the base architecture level for assembler files in order to generate BTI landing pads. We did this due to a lack of support for the .arch_extension gas feature in older versions of the clang built in assembler but since current versions of clang now have support for .arch_extension we can use that.
--- arch/arm64/lib/crc32.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/lib/crc32.S +1 -1
@@ -9,7 +9,7 @@ #include <asm/alternative.h> #include <asm/assembler.h> - .cpu generic+crc + .arch_extension crc .macro __crc32, c cmp x2, #16
Signature · 7 lines
2.20.1 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 2/2 arm64: crypto: Use .arch_extension
Currently when implementing optimised assembler routines using architecture extensions we override the base architecture along with enabling the new extensions, causing problems for in kernel BTI support which needs to raise the base architecture level for assembler files in order to generate BTI landing pads. We did this due to a lack of support for the .arch_extension gas feature in older versions of the clang built in assembler but since current versions of clang now have support for .arch_extension we can use that.
--- arch/arm64/crypto/aes-ce-ccm-core.S | 2 +- arch/arm64/crypto/aes-ce-core.S | 2 +- arch/arm64/crypto/aes-ce.S | 2 +- arch/arm64/crypto/crct10dif-ce-core.S | 2 +- arch/arm64/crypto/ghash-ce-core.S | 2 +- arch/arm64/crypto/sha1-ce-core.S | 2 +- arch/arm64/crypto/sha2-ce-core.S | 2 +- 7 files changed, 7 insertions(+), 7 deletions(-)
arch/arm64/crypto/aes-ce-ccm-core.S +1 -1
@@ -9,7 +9,7 @@ #include <asm/assembler.h> .text - .arch armv8-a+crypto + .arch_extension crypto /* * void ce_aes_ccm_auth_data(u8 mac[], u8 const in[], u32 abytes,
arch/arm64/crypto/aes-ce-core.S +1 -1
@@ -6,7 +6,7 @@ #include <linux/linkage.h> #include <asm/assembler.h> - .arch armv8-a+crypto + .arch_extension crypto SYM_FUNC_START(__aes_ce_encrypt) sub w3, w3, #2
arch/arm64/crypto/aes-ce.S +1 -1
@@ -12,7 +12,7 @@ #define AES_FUNC_START(func) SYM_FUNC_START(ce_ ## func) #define AES_FUNC_END(func) SYM_FUNC_END(ce_ ## func) - .arch armv8-a+crypto + .arch_extension crypto xtsmask .req v16 cbciv .req v16
arch/arm64/crypto/crct10dif-ce-core.S +1 -1
@@ -66,7 +66,7 @@ #include <asm/assembler.h> .text - .cpu generic+crypto + .arch_extension crypto init_crc .req w19 buf .req x20
arch/arm64/crypto/ghash-ce-core.S +1 -1
@@ -57,7 +57,7 @@ HH34 .req v19 .text - .arch armv8-a+crypto + .arch_extension crypto .macro __pmull_p64, rd, rn, rm pmull \rd\().1q, \rn\().1d, \rm\().1d
arch/arm64/crypto/sha1-ce-core.S +1 -1
@@ -9,7 +9,7 @@ #include <asm/assembler.h> .text - .arch armv8-a+crypto + .arch_extension crypto k0 .req v0 k1 .req v1
arch/arm64/crypto/sha2-ce-core.S +1 -1
@@ -9,7 +9,7 @@ #include <asm/assembler.h> .text - .arch armv8-a+crypto + .arch_extension crypto dga .req q20 dgav .req v20
Signature · 7 lines
2.20.1 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 2/2] arm64: crypto: Use .arch_extension
Quoted 10 lines · Mark Brown
Currently when implementing optimised assembler routines using architecture extensions we override the base architecture along with enabling the new extensions, causing problems for in kernel BTI support which needs to raise the base architecture level for assembler files in order to generate BTI landing pads. We did this due to a lack of support for the .arch_extension gas feature in older versions of the clang built in assembler but since current versions of clang now have support for .arch_extension we can use that.
Signed-off-by Mark Brown <[email protected]>
It make sense (I got burnt by this with the MTE patches).
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Catalin Marinas
maintainerQuoted 10 lines · Mark Brown
Currently when implementing optimised assembler routines using architecture extensions we override the base architecture along with enabling the new extensions, causing problems for in kernel BTI support which needs to raise the base architecture level for assembler files in order to generate BTI landing pads. We did this due to a lack of support for the .arch_extension gas feature in older versions of the clang built in assembler but since current versions of clang now have support for .arch_extension we can use that.
Signed-off-by Mark Brown <[email protected]>
I'll queue this through the arm64 tree. Thanks.
Signature · 6 lines
Catalin _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Quoted 9 lines · Mark Brown
Currently when implementing optimised assembler routines using architecture extensions we override the base architecture along with enabling the new extensions, causing problems for in kernel BTI support which needs to raise the base architecture level for assembler files in order to generate BTI landing pads. We did this due to a lack of support for the .arch_extension gas feature in older versions of the clang built in assembler but since current versions of clang now have support for .arch_extension we can use that.
This is not 100% accurate. Support for .arch_extension was added to GAS/aarch64 much later, so we should at least double check that it works on the oldest binutils that we do support.
Quoted 20 lines
Signed-off-by Mark Brown <[email protected]>--- arch/arm64/lib/crc32.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/arm64/lib/crc32.S +1
-2@@ -9,7 +9,7 @@ #include <asm/alternative.h> #include <asm/assembler.h> - .cpu generic+crc + .arch_extension crc .macro __crc32, c cmp x2, #16 --
Quoted 2 lines
2.20.1
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Quoted 10 lines · Ard Biesheuvel
Quoted 4 lines · Mark Brown
order to generate BTI landing pads. We did this due to a lack of support for the .arch_extension gas feature in older versions of the clang built in assembler but since current versions of clang now have support for .arch_extension we can use that.
This is not 100% accurate. Support for .arch_extension was added to GAS/aarch64 much later, so we should at least double check that it works on the oldest binutils that we do support.
Ah, OK - the information I figured out from history was misleading sorry.
We've already got quite a lot of usages of .arch_extension in the kernel for arm and a couple for arm64 (one in TF and another for LSE). It looks like the feature was added in binutils 2.26 which is newer than the current advertised minimum binutils version of 2.21 but dates from 2016 (the code was added in 2014 and looks like it might've appeared in downstream releases earlier) so is pretty old in arm64 terms. If that's not OK I've got an open coded version for BTI that does this with macros but it's obviously not as nice.
I do wish the binutils docs included information on when features were added, it'd make life easier :/
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Quoted 13 lines · Mark Brown
Quoted 10 lines · Ard Biesheuvel
Quoted 4 lines · Mark Brown
order to generate BTI landing pads. We did this due to a lack of support for the .arch_extension gas feature in older versions of the clang built in assembler but since current versions of clang now have support for .arch_extension we can use that.
This is not 100% accurate. Support for .arch_extension was added to GAS/aarch64 much later, so we should at least double check that it works on the oldest binutils that we do support.
Quoted 5 lines
Ah, OK - the information I figured out from history was misleading sorry.
We've already got quite a lot of usages of .arch_extension in the kernel for arm and a couple for arm64 (one in TF and another for LSE).
ARM had support for this way before it was added to AArch64 as well. In the LSE case, it doesn't matter since LSE itself was not implemented by the assembler before that. For other things, we should really avoid .arch_extension as long as we still support binutils < 2.26
Quoted 10 lines
It
looks like the feature was added in binutils 2.26 which is newer than the current advertised minimum binutils version of 2.21 but dates from 2016 (the code was added in 2014 and looks like it might've appeared in downstream releases earlier) so is pretty old in arm64 terms. If that's not OK I've got an open coded version for BTI that does this with macros but it's obviously not as nice.I do wish the binutils docs included information on when features were added, it'd make life easier :/
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Catalin Marinas
maintainerQuoted 17 lines · Ard Biesheuvel
Quoted 16 lines · Mark Brown
Quoted 10 lines · Ard Biesheuvel
Quoted 4 lines · Mark Brown
order to generate BTI landing pads. We did this due to a lack of support for the .arch_extension gas feature in older versions of the clang built in assembler but since current versions of clang now have support for .arch_extension we can use that.
This is not 100% accurate. Support for .arch_extension was added to GAS/aarch64 much later, so we should at least double check that it works on the oldest binutils that we do support.
Ah, OK - the information I figured out from history was misleading sorry.
Quoted 8 lines
Quoted 2 lines
We've already got quite a lot of usages of .arch_extension in the kernel for arm and a couple for arm64 (one in TF and another for LSE).
ARM had support for this way before it was added to AArch64 as well. In the LSE case, it doesn't matter since LSE itself was not implemented by the assembler before that. For other things, we should really avoid .arch_extension as long as we still support binutils < 2.26
Good point. So we can add .arch_extension for new features which are already dependent on newer gas options.
I'll drop this patch for now but we should try to find a solution because of the overriding effect of .arch. In self-contained .S files (like crypto) that's not an issue but once we start adding more architecture feature in a single file, things will get more complicated.
Signature · 6 lines
Catalin _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Quoted 4 lines · Catalin Marinas
I'll drop this patch for now but we should try to find a solution because of the overriding effect of .arch. In self-contained .S files (like crypto) that's not an issue but once we start adding more architecture feature in a single file, things will get more complicated.
It's fine I think, you can open code .arch_extension in the assembler.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel