linux-kernel - Re: [PATCH -next] crypto: Fix build error when CRYPTO_BLAKE2S_X86=m && CRYPTO_ALGAPI2=m && CONFIG_CRYPTO_MANAGER_EXTRA

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 26 May 2022 00:25:06 -0700
From:   Eric Biggers <ebiggers@...nel.org>
To:     gaochao <gaochao49@...wei.com>
Cc:     herbert@...dor.apana.org.au, davem@...emloft.net,
        tglx@...utronix.de, mingo@...hat.com, bp@...en8.de,
        dave.hansen@...ux.intel.com, x86@...nel.org, hpa@...or.com,
        linux-crypto@...r.kernel.org, linux-kernel@...r.kernel.org,
        "Jason A. Donenfeld" <Jason@...c4.com>
Subject: Re: [PATCH -next] crypto: Fix build error when CRYPTO_BLAKE2S_X86=m
 && CRYPTO_ALGAPI2=m && CONFIG_CRYPTO_MANAGER_EXTRA_TESTS=y

+Cc Jason

On Tue, May 17, 2022 at 11:36:30AM +0800, gaochao wrote:
> If CRYPTO=m, CRYPTO_ALGAPI=m, CRYPTO_ALGAPI2=m, CRYPTO_BLAKE2S_X86=m,
> CONFIG_CRYPTO_MANAGER_EXTRA_TESTS=y
> bulding fails:
> 
> arch/x86/crypto/blake2s-glue.o: In function `blake2s_compress':
> (.text+0x5a): undefined reference to `crypto_simd_disabled_for_test'
> make: *** [vmlinux] Error 1
> 
> When CRYPTO_MANAGER_EXTRA_TESTS=y, blake2s_compress will call
> crypto_simd_disabled_for_test.
> When CRYPTO_ALGAPI2=m, crypto_algapi build as a module,
> but if CONFIG_CRYPTO_BLAKE2S_X86=m at the same time,
> libblake2s-x86_64.o build with obj-y, this will accuse the above error.
> 
> To fix this error:
> 1 Choose CRYPTO_ALGAPI2 for CRYPTO_BLAKE2S_X86
> when CRYPTO_MANAGER_EXTRA_TESTS=y.
> 2 build libblake2s-x86_64.o as a module when CONFIG_CRYPTO_BLAKE2S_X86=y
> 
> Fixes: 8fc5f2ad896b ("crypto: testmgr - Move crypto_simd_disabled_for_test out")
> Reported-by: Hulk Robot <hulkci@...wei.com>
> Signed-off-by: gaochao <gaochao49@...wei.com>
> ---
>  arch/x86/crypto/Makefile | 2 +-
>  crypto/Kconfig           | 1 +
>  2 files changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/crypto/Makefile b/arch/x86/crypto/Makefile
> index 2831685adf6f..54b2469fa49a 100644
> --- a/arch/x86/crypto/Makefile
> +++ b/arch/x86/crypto/Makefile
> @@ -63,7 +63,7 @@ sha512-ssse3-y := sha512-ssse3-asm.o sha512-avx-asm.o sha512-avx2-asm.o sha512_s
> 
>  obj-$(CONFIG_CRYPTO_BLAKE2S_X86) += blake2s-x86_64.o
>  blake2s-x86_64-y := blake2s-shash.o
> -obj-$(if $(CONFIG_CRYPTO_BLAKE2S_X86),y) += libblake2s-x86_64.o
> +obj-$(CONFIG_CRYPTO_BLAKE2S_X86) += libblake2s-x86_64.o
>  libblake2s-x86_64-y := blake2s-core.o blake2s-glue.o

I think the (somewhat unusual) use of "obj-$(if $(CONFIG_CRYPTO_BLAKE2S_X86),y)"
is working as intended.  The goal to build the x86 BLAKE2s library into the core
kernel if CONFIG_CRYPTO_BLAKE2S_X86 is set to either y or m.  This change would
break that.  And if we were to make this change, then there would no longer be
any need to have blake2s-x86_64 and libblake2s-x86_64 be separate modules.

> diff --git a/crypto/Kconfig b/crypto/Kconfig
> index 19197469cfab..e61598f8f8c5 100644
> --- a/crypto/Kconfig
> +++ b/crypto/Kconfig
> @@ -714,6 +714,7 @@ config CRYPTO_BLAKE2S_X86
>  	depends on X86 && 64BIT
>  	select CRYPTO_LIB_BLAKE2S_GENERIC
>  	select CRYPTO_ARCH_HAVE_LIB_BLAKE2S
> +	select CRYPTO_ALGAPI2 if CRYPTO_MANAGER_EXTRA_TESTS

I expect that we'll run into this problem in the future too.  So just fixing it
for CRYPTO_BLAKE2S_X86 specifically isn't ideal.

How about just making CRYPTO_MANAGER_EXTRA_TESTS depend on CRYPTO_ALGAPI=y?

Or alternatively, move the crypto_simd_disabled_for_test variable into a file in
lib/crypto/.  But the other suggestion would be a bit simpler for now.

- Eric