The MIPS Poly1305 implementation is generic MIPS code written such as to
support down to the original MIPS I and MIPS III ISA for the 32-bit and
64-bit variant respectively. Lift the current limitation then to enable
code for MIPSr1 ISA or newer processors only and have it available for
all MIPS processors.
Signed-off-by: Maciej W. Rozycki <[email protected]>
Fixes: a11d055e7a64 ("crypto: mips/poly1305 - incorporate OpenSSL/CRYPTOGAMS optimized implementation")
Cc: [email protected] # v5.5+
---
On Wed, 3 Mar 2021, Jason A. Donenfeld wrote:
> >> Would you mind sending this for 5.12 in an rc at some point, rather
> >> than waiting for 5.13? I'd like to see this backported to 5.10 and 5.4
> >> for OpenWRT.
> >
> > why is this so important for OpenWRT ? Just to select CRYPTO_POLY1305_MIPS
> > ?
>
> Yes. The performance boost on Octeon is significant for WireGuard users.
But that's the wrong fix for that purpose. I've skimmed over that module
and there's nothing MIPS64-specific there. In fact it's plain generic
MIPS assembly, with some R2 optimisations enabled where applicable but not
necessary (and then R6 tweaks, but that's irrelevant here).
As a matter of interest I have just built it successfully for a MIPS I
DECstation configuration:
$ file arch/mips/crypto/poly1305-mips.ko
arch/mips/crypto/poly1305-mips.ko: ELF 32-bit LSB relocatable, MIPS, MIPS-I version 1 (SYSV), BuildID[sha1]=d36384d94f60ba7deff638ca8a24500120b45b56, not stripped
$
Patch included, please apply.
So while your change is surely right, what you want is this really.
Maciej
---
arch/mips/crypto/Makefile | 4 ++--
crypto/Kconfig | 2 +-
drivers/net/Kconfig | 2 +-
3 files changed, 4 insertions(+), 4 deletions(-)
Index: linux/arch/mips/crypto/Makefile
===================================================================
--- linux.orig/arch/mips/crypto/Makefile
+++ linux/arch/mips/crypto/Makefile
@@ -12,8 +12,8 @@ AFLAGS_chacha-core.o += -O2 # needed to
obj-$(CONFIG_CRYPTO_POLY1305_MIPS) += poly1305-mips.o
poly1305-mips-y := poly1305-core.o poly1305-glue.o
-perlasm-flavour-$(CONFIG_CPU_MIPS32) := o32
-perlasm-flavour-$(CONFIG_CPU_MIPS64) := 64
+perlasm-flavour-$(CONFIG_32BIT) := o32
+perlasm-flavour-$(CONFIG_64BIT) := 64
quiet_cmd_perlasm = PERLASM $@
cmd_perlasm = $(PERL) $(<) $(perlasm-flavour-y) $(@)
Index: linux/crypto/Kconfig
===================================================================
--- linux.orig/crypto/Kconfig
+++ linux/crypto/Kconfig
@@ -772,7 +772,7 @@ config CRYPTO_POLY1305_X86_64
config CRYPTO_POLY1305_MIPS
tristate "Poly1305 authenticator algorithm (MIPS optimized)"
- depends on CPU_MIPS32 || (CPU_MIPS64 && 64BIT)
+ depends on MIPS
select CRYPTO_ARCH_HAVE_LIB_POLY1305
config CRYPTO_MD4
Index: linux/drivers/net/Kconfig
===================================================================
--- linux.orig/drivers/net/Kconfig
+++ linux/drivers/net/Kconfig
@@ -92,7 +92,7 @@ config WIREGUARD
select CRYPTO_POLY1305_ARM if ARM
select CRYPTO_CURVE25519_NEON if ARM && KERNEL_MODE_NEON
select CRYPTO_CHACHA_MIPS if CPU_MIPS32_R2
- select CRYPTO_POLY1305_MIPS if CPU_MIPS32 || (CPU_MIPS64 && 64BIT)
+ select CRYPTO_POLY1305_MIPS if MIPS
help
WireGuard is a secure, fast, and easy to use replacement for IPSec
that uses modern cryptography and clever networking tricks. It's
Hi Maciej,
On Wed, Mar 3, 2021 at 2:16 AM Maciej W. Rozycki <[email protected]> wrote:
>
> The MIPS Poly1305 implementation is generic MIPS code written such as to
> support down to the original MIPS I and MIPS III ISA for the 32-bit and
> 64-bit variant respectively. Lift the current limitation then to enable
> code for MIPSr1 ISA or newer processors only and have it available for
> all MIPS processors.
That sounds like a good solution to me. Thanks for doing the research
on it. Assuming your findings hold up:
Acked-by: Jason A. Donenfeld <[email protected]>
I'm also CC'ing Andy on this, who wrote the original assembly, in case
he has some last minute objection.
Jason
>
> Signed-off-by: Maciej W. Rozycki <[email protected]>
> Fixes: a11d055e7a64 ("crypto: mips/poly1305 - incorporate OpenSSL/CRYPTOGAMS optimized implementation")
> Cc: [email protected] # v5.5+
> ---
> On Wed, 3 Mar 2021, Jason A. Donenfeld wrote:
>
> > >> Would you mind sending this for 5.12 in an rc at some point, rather
> > >> than waiting for 5.13? I'd like to see this backported to 5.10 and 5.4
> > >> for OpenWRT.
> > >
> > > why is this so important for OpenWRT ? Just to select CRYPTO_POLY1305_MIPS
> > > ?
> >
> > Yes. The performance boost on Octeon is significant for WireGuard users.
>
> But that's the wrong fix for that purpose. I've skimmed over that module
> and there's nothing MIPS64-specific there. In fact it's plain generic
> MIPS assembly, with some R2 optimisations enabled where applicable but not
> necessary (and then R6 tweaks, but that's irrelevant here).
>
> As a matter of interest I have just built it successfully for a MIPS I
> DECstation configuration:
>
> $ file arch/mips/crypto/poly1305-mips.ko
> arch/mips/crypto/poly1305-mips.ko: ELF 32-bit LSB relocatable, MIPS, MIPS-I version 1 (SYSV), BuildID[sha1]=d36384d94f60ba7deff638ca8a24500120b45b56, not stripped
> $
>
> Patch included, please apply.
>
> So while your change is surely right, what you want is this really.
>
> Maciej
> ---
> arch/mips/crypto/Makefile | 4 ++--
> crypto/Kconfig | 2 +-
> drivers/net/Kconfig | 2 +-
> 3 files changed, 4 insertions(+), 4 deletions(-)
>
> Index: linux/arch/mips/crypto/Makefile
> ===================================================================
> --- linux.orig/arch/mips/crypto/Makefile
> +++ linux/arch/mips/crypto/Makefile
> @@ -12,8 +12,8 @@ AFLAGS_chacha-core.o += -O2 # needed to
> obj-$(CONFIG_CRYPTO_POLY1305_MIPS) += poly1305-mips.o
> poly1305-mips-y := poly1305-core.o poly1305-glue.o
>
> -perlasm-flavour-$(CONFIG_CPU_MIPS32) := o32
> -perlasm-flavour-$(CONFIG_CPU_MIPS64) := 64
> +perlasm-flavour-$(CONFIG_32BIT) := o32
> +perlasm-flavour-$(CONFIG_64BIT) := 64
>
> quiet_cmd_perlasm = PERLASM $@
> cmd_perlasm = $(PERL) $(<) $(perlasm-flavour-y) $(@)
> Index: linux/crypto/Kconfig
> ===================================================================
> --- linux.orig/crypto/Kconfig
> +++ linux/crypto/Kconfig
> @@ -772,7 +772,7 @@ config CRYPTO_POLY1305_X86_64
>
> config CRYPTO_POLY1305_MIPS
> tristate "Poly1305 authenticator algorithm (MIPS optimized)"
> - depends on CPU_MIPS32 || (CPU_MIPS64 && 64BIT)
> + depends on MIPS
> select CRYPTO_ARCH_HAVE_LIB_POLY1305
>
> config CRYPTO_MD4
> Index: linux/drivers/net/Kconfig
> ===================================================================
> --- linux.orig/drivers/net/Kconfig
> +++ linux/drivers/net/Kconfig
> @@ -92,7 +92,7 @@ config WIREGUARD
> select CRYPTO_POLY1305_ARM if ARM
> select CRYPTO_CURVE25519_NEON if ARM && KERNEL_MODE_NEON
> select CRYPTO_CHACHA_MIPS if CPU_MIPS32_R2
> - select CRYPTO_POLY1305_MIPS if CPU_MIPS32 || (CPU_MIPS64 && 64BIT)
> + select CRYPTO_POLY1305_MIPS if MIPS
> help
> WireGuard is a secure, fast, and easy to use replacement for IPSec
> that uses modern cryptography and clever networking tricks. It's
Hi,
> I'm also CC'ing Andy on this, who wrote the original assembly, in case
> he has some last minute objection.
Just "what took you so long":-) On potentially related note cryptogams
chacha-mips is as universal as poly1305-mips. "Universal" in sense that
it can be compiled for all MIPS stripes.
Cheers.
On Wed, Mar 3, 2021 at 1:31 PM Andy Polyakov <[email protected]> wrote:
>
> Hi,
>
> > I'm also CC'ing Andy on this, who wrote the original assembly, in case
> > he has some last minute objection.
>
> Just "what took you so long":-) On potentially related note cryptogams
> chacha-mips is as universal as poly1305-mips. "Universal" in sense that
> it can be compiled for all MIPS stripes.
Oh great. I didn't realize you had a ChaCha mips implementation. I'll
look into having that replace my junky mips32r2 one.
Jason
On Wed, Mar 03, 2021 at 02:16:04AM +0100, Maciej W. Rozycki wrote:
> The MIPS Poly1305 implementation is generic MIPS code written such as to
> support down to the original MIPS I and MIPS III ISA for the 32-bit and
> 64-bit variant respectively. Lift the current limitation then to enable
> code for MIPSr1 ISA or newer processors only and have it available for
> all MIPS processors.
>
> Signed-off-by: Maciej W. Rozycki <[email protected]>
> Fixes: a11d055e7a64 ("crypto: mips/poly1305 - incorporate OpenSSL/CRYPTOGAMS optimized implementation")
> Cc: [email protected] # v5.5+
> ---
> On Wed, 3 Mar 2021, Jason A. Donenfeld wrote:
>
> > >> Would you mind sending this for 5.12 in an rc at some point, rather
> > >> than waiting for 5.13? I'd like to see this backported to 5.10 and 5.4
> > >> for OpenWRT.
> > >
> > > why is this so important for OpenWRT ? Just to select CRYPTO_POLY1305_MIPS
> > > ?
> >
> > Yes. The performance boost on Octeon is significant for WireGuard users.
>
> But that's the wrong fix for that purpose. I've skimmed over that module
> and there's nothing MIPS64-specific there. In fact it's plain generic
> MIPS assembly, with some R2 optimisations enabled where applicable but not
> necessary (and then R6 tweaks, but that's irrelevant here).
>
> As a matter of interest I have just built it successfully for a MIPS I
> DECstation configuration:
>
> $ file arch/mips/crypto/poly1305-mips.ko
> arch/mips/crypto/poly1305-mips.ko: ELF 32-bit LSB relocatable, MIPS, MIPS-I version 1 (SYSV), BuildID[sha1]=d36384d94f60ba7deff638ca8a24500120b45b56, not stripped
> $
>
> Patch included, please apply.
>
> So while your change is surely right, what you want is this really.
>
> Maciej
> ---
> arch/mips/crypto/Makefile | 4 ++--
> crypto/Kconfig | 2 +-
> drivers/net/Kconfig | 2 +-
> 3 files changed, 4 insertions(+), 4 deletions(-)
applied to mips-fixes.
Thomas.
--
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea. [ RFC1925, 2.3 ]