Hi Herbert,
On Fri, Jun 10, 2022 at 08:09:24PM -0700, Eric Biggers wrote:
> On Fri, Jun 10, 2022 at 06:42:55PM -0700, Jason Self wrote:
> > On Fri, 10 Jun 2022 18:25:23 -0700
> > Jason Self <jason@xxxxxxxxxxxx> wrote:
> >
> > > On Fri, 10 Jun 2022 07:15:37 +0200
> > > Greg KH <greg@xxxxxxxxx> wrote:
> > >
> > > > On Thu, Jun 09, 2022 at 04:29:43PM -0700, Jason Self wrote:
> > > > > In building 5.15.46 & 5.10.121 with CRYPTO_LIB_CURVE25519=m I
> > > > > get the following. My workaround is to leave it as
> > > > > CRYPTO_LIB_CURVE25519=n for now.
> > > > >
> > > > > CONFIG_OR1K_1200=y
> > > > > CONFIG_OPENRISC_BUILTIN_DTB="or1ksim"
> > > > >
> > > > > sed 's/\.ko$/\.o/' modules.order | scripts/mod/modpost -o
> > > > > modules-only.symvers -i vmlinux.symvers -T - ERROR: modpost:
> > > > > "__crypto_memneq" [lib/crypto/libcurve25519.ko] undefined!
> > > > > make[1]: *** [scripts/Makefile.modpost:134:
> > > > > modules-only.symvers] Error 1 make[1]: *** Deleting file
> > > > > 'modules-only.symvers' make: *** [Makefile:1783: modules] Error
> > > > > 2
> > > >
> > > >
> > > > Is this a new problem, or has it always been there for these
> > > > kernel trees?
> > >
> > > It's new; it began in 5.15.45 & 5.10.120, which is when make
> > > oldconfig first prompted about CONFIG_CRYPTO_LIB_CURVE25519.
> >
> > The result of my git bisect between 5.15.44 and 5.15.45 tell me the
> > following. It's the same "lib/crypto: add prompts back to crypto
> > libraries" commit when I bisect between 5.10.119 and 5.10.120.
> >
> >
> > e16cc79b0f916069de223bdb567fa0bc2ccd18a5 is the first bad commit
> > commit e16cc79b0f916069de223bdb567fa0bc2ccd18a5
> > Author: Justin M. Forbes <jforbes@xxxxxxxxxxxxxxxxx>
> > Date: Thu Jun 2 22:23:23 2022 +0200
> >
> > lib/crypto: add prompts back to crypto libraries
> >
> > commit e56e18985596617ae426ed5997fb2e737cffb58b upstream.
> >
> > Commit 6048fdcc5f269 ("lib/crypto: blake2s: include as built-in")
> > took away a number of prompt texts from other crypto libraries.
> > This makes values flip from built-in to module when oldconfig
> > runs, and causes problems when these crypto libs need to be built
> > in for thingslike BIG_KEYS.
> >
> > Fixes: 6048fdcc5f269 ("lib/crypto: blake2s: include as built-in")
> > Cc: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx>
> > Cc: linux-crypto@xxxxxxxxxxxxxxx
> > Signed-off-by: Justin M. Forbes <jforbes@xxxxxxxxxxxxxxxxx>
> > [Jason: - moved menu into submenu of lib/ instead of root menu
> > - fixed chacha sub-dependencies for CONFIG_CRYPTO]
> > Signed-off-by: Jason A. Donenfeld <Jason@xxxxxxxxx>
> > Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> >
> > crypto/Kconfig | 2 --
> > lib/Kconfig | 2 ++
> > lib/crypto/Kconfig | 17 ++++++++++++-----
> > 3 files changed, 14 insertions(+), 7 deletions(-)
> > bisect run success
> >
>
> It looks like "crypto: memneq - move into lib/" is going to fix this
> (https://lore.kernel.org/linux-crypto/20220528102429.189731-1-Jason@xxxxxxxxx).
> At the moment it's queued in cryptodev/master. Herbert, are you planning to
> send it upstream soon?
Both of these commits:
https://git.kernel.org/pub/scm/linux/kernel/git/herbert/cryptodev-2.6.git/commit/?id=920b0442b9f884f55f4745b53430c80e71e90275
https://git.kernel.org/pub/scm/linux/kernel/git/herbert/cryptodev-2.6.git/commit/?id=2d16803c562ecc644803d42ba98a8e0aef9c014e
were marked as "[PATCH crypto]", rather than cryptodev, have a CC to
stable@, and have a fixes tag. So I think these would be better slated
for the crypto tree rather than the cryptodev tree, so that they make it
to Linus soonish.
Jason
