Linux cryptographic layer development
 help / color / mirror / Atom feed
From: Eric Biggers <ebiggers@kernel.org>
To: Tal Zussman <tz2294@columbia.edu>
Cc: "Jason A. Donenfeld" <Jason@zx2c4.com>,
	Ard Biesheuvel <ardb@kernel.org>,
	linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] lib/crypto: ensure generated *.S files are removed on make clean
Date: Wed, 13 Aug 2025 19:40:12 -0700	[thread overview]
Message-ID: <20250814024012.GA1296@sol> (raw)
In-Reply-To: <CAKha_srSRA9HftM+zLeRVrONKmPdtm-wTXq3n2NC60Gynuvwyw@mail.gmail.com>

On Thu, Aug 14, 2025 at 03:47:05AM +0300, Tal Zussman wrote:
> On Thu, Aug 14, 2025 at 3:39 AM Tal Zussman <tz2294@columbia.edu> wrote:
> >
> > make clean does not check the kernel config when removing files. As
> > such, additions to clean-files under CONFIG_ARM or CONFIG_ARM64 are not
> > evaluated. For example, when building on arm64, this means that
> > lib/crypto/arm64/sha{256,512}-core.S are left over after make clean.
> >
> > Set clean-files unconditionally to ensure that make clean removes these
> > files.
> >
> > Fixes: e96cb9507f2d ("lib/crypto: sha256: Consolidate into single module")
> > Fixes: 24c91b62ac50 ("lib/crypto: arm/sha512: Migrate optimized SHA-512 code to library")
> > Fixes: 60e3f1e9b7a5 ("lib/crypto: arm64/sha512: Migrate optimized SHA-512 code to library")
> > Signed-off-by: Tal Zussman <tz2294@columbia.edu>
> > ---
> > An alternative approach is to rename the generated files to *.s and
> > remove the clean-files lines, as make clean removes *.s files
> > automatically. However, this would require explicitly defining the
> > corresponding *.o rules.
> > ---
> >  lib/crypto/Makefile | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/lib/crypto/Makefile b/lib/crypto/Makefile
> > index e4151be2ebd4..44f6a1fdc808 100644
> > --- a/lib/crypto/Makefile
> > +++ b/lib/crypto/Makefile
> > @@ -100,7 +100,6 @@ ifeq ($(CONFIG_ARM),y)
> >  libsha256-y += arm/sha256-ce.o arm/sha256-core.o
> >  $(obj)/arm/sha256-core.S: $(src)/arm/sha256-armv4.pl
> >         $(call cmd,perlasm)
> > -clean-files += arm/sha256-core.S
> >  AFLAGS_arm/sha256-core.o += $(aflags-thumb2-y)
> >  endif
> >
> > @@ -108,7 +107,6 @@ ifeq ($(CONFIG_ARM64),y)
> >  libsha256-y += arm64/sha256-core.o
> >  $(obj)/arm64/sha256-core.S: $(src)/arm64/sha2-armv8.pl
> >         $(call cmd,perlasm_with_args)
> > -clean-files += arm64/sha256-core.S
> >  libsha256-$(CONFIG_KERNEL_MODE_NEON) += arm64/sha256-ce.o
> >  endif
> >
> > @@ -132,7 +130,6 @@ ifeq ($(CONFIG_ARM),y)
> >  libsha512-y += arm/sha512-core.o
> >  $(obj)/arm/sha512-core.S: $(src)/arm/sha512-armv4.pl
> >         $(call cmd,perlasm)
> > -clean-files += arm/sha512-core.S
> >  AFLAGS_arm/sha512-core.o += $(aflags-thumb2-y)
> >  endif
> >
> > @@ -140,7 +137,6 @@ ifeq ($(CONFIG_ARM64),y)
> >  libsha512-y += arm64/sha512-core.o
> >  $(obj)/arm64/sha512-core.S: $(src)/arm64/sha2-armv8.pl
> >         $(call cmd,perlasm_with_args)
> > -clean-files += arm64/sha512-core.S
> >  libsha512-$(CONFIG_KERNEL_MODE_NEON) += arm64/sha512-ce-core.o
> >  endif
> >
> > @@ -167,3 +163,7 @@ obj-$(CONFIG_PPC) += powerpc/
> >  obj-$(CONFIG_RISCV) += riscv/
> >  obj-$(CONFIG_S390) += s390/
> >  obj-$(CONFIG_X86) += x86/
> > +
> > +# clean-files must be defined unconditionally
> > +clean-files += arm/sha256-core.S arm/sha256-core.S
> > +clean-files += arm64/sha512-core.S arm64/sha512-core.S
> 
> Sorry this is broken, needs the following fix on top.
> I'll fix in v2.
> 
> diff --git a/lib/crypto/Makefile b/lib/crypto/Makefile
> index 44f6a1fdc808..539d5d59a50e 100644
> --- a/lib/crypto/Makefile
> +++ b/lib/crypto/Makefile
> @@ -165,5 +165,5 @@ obj-$(CONFIG_S390) += s390/
>  obj-$(CONFIG_X86) += x86/
> 
>  # clean-files must be defined unconditionally
> -clean-files += arm/sha256-core.S arm/sha256-core.S
> -clean-files += arm64/sha512-core.S arm64/sha512-core.S
> +clean-files += arm/sha256-core.S arm/sha512-core.S
> +clean-files += arm64/sha256-core.S arm64/sha512-core.S

Hmm, interesting.  So 'make clean' works in kind of a unique way.  It
looks at $(ARCH), but not .config.  And though no CONFIG_* options are
set for it, it still processes all the subdirectories like
'obj-$(CONFIG_FOO) += subdir/'.  But it doesn't process the contents of
'if' statements conditional on $(CONFIG_FOO).  I guess it couldn't
really work in any other way, but it's a bit unique.

Yes, just defining the clean-files unconditionally looks like the best
option here.  Please go ahead and send out v2 with the corrected
filenames.  Thanks!

- Eric

      reply	other threads:[~2025-08-14  2:41 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-08-14  0:38 [PATCH] lib/crypto: ensure generated *.S files are removed on make clean Tal Zussman
2025-08-14  0:47 ` Tal Zussman
2025-08-14  2:40   ` Eric Biggers [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20250814024012.GA1296@sol \
    --to=ebiggers@kernel.org \
    --cc=Jason@zx2c4.com \
    --cc=ardb@kernel.org \
    --cc=linux-crypto@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=tz2294@columbia.edu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox