On Fri, 20 Nov 2020 at 11:37, Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> wrote: > > On Fri, Nov 20, 2020 at 11:34:14AM +0100, Ard Biesheuvel wrote: > > On Fri, 20 Nov 2020 at 11:09, Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> wrote: > > > > > > On Fri, Nov 20, 2020 at 10:24:44AM +0100, Ard Biesheuvel wrote: > > > > > > > > OK, I'll apply this on top > > > > > > > > diff --git a/crypto/Kconfig b/crypto/Kconfig > > > > index 9ff2d687e334..959ee48f66a8 100644 > > > > --- a/crypto/Kconfig > > > > +++ b/crypto/Kconfig > > > > @@ -202,7 +202,7 @@ config CRYPTO_AUTHENC > > > > config CRYPTO_TEST > > > > tristate "Testing module" > > > > depends on m || CRYPTO_MANAGER_EXTRA_TESTS > > > > - select CRYPTO_MANAGER > > > > + depends on CRYPTO_MANAGER > > > > > > How about just removing the depends line altogether? > > > > That may break the build, and therefore randconfig build testing: > > > > crypto/tcrypt.o: In function `do_mult_aead_op': > > tcrypt.c:(.text+0x180): undefined reference to `crypto_aead_encrypt' > > tcrypt.c:(.text+0x194): undefined reference to `crypto_aead_decrypt' > > Did you keep the select CRYPTO_MANAGER? That should bring in > everything that's needed. > Ah right, you mean the /other/ depends line. Yeah, that would work, but enabling it as a builtin produces a lot of bogus output if you fail to set the tcrypt.mode=xxx kernel command line option, so it is really only intended for deliberate use.
