[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20251024161105.GA1625@sol>
Date: Fri, 24 Oct 2025 09:11:05 -0700
From: Eric Biggers <ebiggers@...nel.org>
To: Harald Freudenberger <freude@...ux.ibm.com>
Cc: Holger Dengler <dengler@...ux.ibm.com>,
David Howells <dhowells@...hat.com>,
Ard Biesheuvel <ardb@...nel.org>,
"Jason A . Donenfeld" <Jason@...c4.com>,
linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-s390@...r.kernel.org, linux-crypto@...r.kernel.org
Subject: Re: [PATCH 15/17] lib/crypto: s390/sha3: Migrate optimized code into
library
On Fri, Oct 24, 2025 at 04:24:00PM +0200, Harald Freudenberger wrote:
> On 2025-10-21 17:49, Eric Biggers wrote:
> > On Tue, Oct 21, 2025 at 10:43:00AM +0200, Holger Dengler wrote:
> > > Hi Eric,
> > >
> > > On 21/10/2025 09:24, Holger Dengler wrote:
> > > > On 20/10/2025 19:57, Eric Biggers wrote:
> > > [...]>> - Risk of bugs. QEMU doesn't support the s390 SHA-3
> > > instructions, so no
> > > >> one except the s390 folks can test the code. I can try to write code
> > > >> for you, but I can't test it. And the s390 SHA-3 code has had bugs;
> > > >> see commits 992b7066800f, 68279380266a5, 73c2437109c3.
> > > >>
> > > >> The first priority should be correctness.
> > > >
> > > > Let me figure out, if me and my colleagues can do the testing for you.
> > > > Unfortunately, I'll be unavailable for the next two weeks. But I'll come back
> > > > with a solution for the testing.
> > >
> > > I talked to Harald: we can do the testing for you on our development
> > > machines.
> > > Please send new series to us or provide them in your git repo.
> >
> > Thanks! I'll Cc both of you on v2 when I send it later. For now, this
> > series (v1) can be found in lore at
> > https://lore.kernel.org/linux-crypto/20251020005038.661542-1-ebiggers@kernel.org/T/#u
> > And as mentioned in the cover letter it's also retrievable from git:
> >
> > git fetch
> > https://git.kernel.org/pub/scm/linux/kernel/git/ebiggers/linux.git
> > sha3-lib-v1
> >
> > v1 already has the s390 optimized implementations of
> > sha3_absorb_blocks() and sha3_keccakf(). If you could enable the
> > following:
> >
> > CONFIG_CRYPTO_LIB_SHA3_KUNIT_TEST=y
> > CONFIG_CRYPTO_LIB_BENCHMARK=y
> >
> > ... and then show the results for sha3_kunit before and after the commit
> > "lib/crypto: s390/sha3: Migrate optimized code into library", that would
> > be helpful.
> >
> > In v2, I'll look into providing overrides for the one-shot functions
> > sha3_{224,256,384,512}() too. If it works out, I'll ask you to re-test
> > with that additional change as well.
> >
> > - Eric
>
> I pulled your repository and checked out the branch sha3-lib-v1 and
> while the kernel build runs I get link errors:
>
> ld: crypto/sha3.o: in function `crypto_sha3_512_digest':
> /root/ebiggers-linux/crypto/sha3.c:80:(.text+0xaa): undefined reference to
> `sha3_512'
> ld: crypto/sha3.o: in function `crypto_sha3_384_digest':
> /root/ebiggers-linux/crypto/sha3.c:73:(.text+0xea): undefined reference to
> `sha3_384'
> ld: crypto/sha3.o: in function `crypto_sha3_256_digest':
> /root/ebiggers-linux/crypto/sha3.c:66:(.text+0x12a): undefined reference to
> `sha3_256'
> ld: crypto/sha3.o: in function `crypto_sha3_224_digest':
> /root/ebiggers-linux/crypto/sha3.c:59:(.text+0x1aa): undefined reference to
> `sha3_224'
> ld: crypto/sha3.o: in function `sha3_final':
> /root/ebiggers-linux/./include/crypto/sha3.h:188:(.text+0x1f0): undefined
> reference to `__sha3_squeeze'
> ld: crypto/sha3.o: in function `sha3_update':
> /root/ebiggers-linux/./include/crypto/sha3.h:172:(.text+0x232): undefined
> reference to `__sha3_update'
>
> with a s390 defconfig kernel configuration an a s390 debug_defconfig kernel
> configuration.
Yes, as mentioned elsewhere in the thread the following fixup is needed:
diff --git a/crypto/Kconfig b/crypto/Kconfig
index a04595f9d0ca4..0ff68212cb20a 100644
--- a/crypto/Kconfig
+++ b/crypto/Kconfig
@@ -1005,6 +1005,7 @@ config CRYPTO_SHA512
config CRYPTO_SHA3
tristate "SHA-3"
select CRYPTO_HASH
+ select CRYPTO_LIB_SHA3
help
SHA-3 secure hash algorithms (FIPS 202, ISO/IEC 10118-3)
I'll fix that in v2.
But also note that if you enable the KUnit test as I suggested then this
error gets avoided too, since it selects the library. (That's why I
didn't notice it before sending -- I always had the KUnit test enabled.)
- Eric
Powered by blists - more mailing lists