diff mbox series

crypto: aegis128 - avoid spurious references crypto_aegis128_update_simd

Message ID 20201130122620.16640-1-ardb@kernel.org (mailing list archive)
State Accepted
Delegated to: Herbert Xu
Headers show
Series crypto: aegis128 - avoid spurious references crypto_aegis128_update_simd | expand

Commit Message

Ard Biesheuvel Nov. 30, 2020, 12:26 p.m. UTC
Geert reports that builds where CONFIG_CRYPTO_AEGIS128_SIMD is not set
may still emit references to crypto_aegis128_update_simd(), which
cannot be satisfied and therefore break the build. These references
only exist in functions that can be optimized away, but apparently,
the compiler is not always able to prove this.

So add some explicit checks for CONFIG_CRYPTO_AEGIS128_SIMD to help the
compiler figure this out.

Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
---
 crypto/aegis128-core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Geert Uytterhoeven Nov. 30, 2020, 12:42 p.m. UTC | #1
Hi Ard,

On Mon, Nov 30, 2020 at 1:26 PM Ard Biesheuvel <ardb@kernel.org> wrote:
> Geert reports that builds where CONFIG_CRYPTO_AEGIS128_SIMD is not set
> may still emit references to crypto_aegis128_update_simd(), which
> cannot be satisfied and therefore break the build. These references
> only exist in functions that can be optimized away, but apparently,
> the compiler is not always able to prove this.

The code is not unreachable. Both crypto_aegis128_encrypt_simd() and
crypto_aegis128_decrypt_simd() call crypto_aegis128_process_ad(..., true);

Gr{oetje,eeting}s,

                        Geert
Ard Biesheuvel Nov. 30, 2020, 12:47 p.m. UTC | #2
On Mon, 30 Nov 2020 at 13:42, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
>
> Hi Ard,
>
> On Mon, Nov 30, 2020 at 1:26 PM Ard Biesheuvel <ardb@kernel.org> wrote:
> > Geert reports that builds where CONFIG_CRYPTO_AEGIS128_SIMD is not set
> > may still emit references to crypto_aegis128_update_simd(), which
> > cannot be satisfied and therefore break the build. These references
> > only exist in functions that can be optimized away, but apparently,
> > the compiler is not always able to prove this.
>
> The code is not unreachable. Both crypto_aegis128_encrypt_simd() and
> crypto_aegis128_decrypt_simd() call crypto_aegis128_process_ad(..., true);
>

Those functions themselves can be optimized away too, as well as
struct aead_alg crypto_aegis128_alg_simd, which is the only thing that
refers to those functions, and is itself only referenced inside a 'if
(IS_ENABLED(CONFIG_CRYPTO_AEGIS128_SIMD))' conditional block. This is
why it works fine most of the time.
Geert Uytterhoeven Nov. 30, 2020, 1:09 p.m. UTC | #3
Hi Ard,

On Mon, Nov 30, 2020 at 1:47 PM Ard Biesheuvel <ardb@kernel.org> wrote:
> On Mon, 30 Nov 2020 at 13:42, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> > On Mon, Nov 30, 2020 at 1:26 PM Ard Biesheuvel <ardb@kernel.org> wrote:
> > > Geert reports that builds where CONFIG_CRYPTO_AEGIS128_SIMD is not set
> > > may still emit references to crypto_aegis128_update_simd(), which
> > > cannot be satisfied and therefore break the build. These references
> > > only exist in functions that can be optimized away, but apparently,
> > > the compiler is not always able to prove this.
> >
> > The code is not unreachable. Both crypto_aegis128_encrypt_simd() and
> > crypto_aegis128_decrypt_simd() call crypto_aegis128_process_ad(..., true);
> >
>
> Those functions themselves can be optimized away too, as well as
> struct aead_alg crypto_aegis128_alg_simd, which is the only thing that
> refers to those functions, and is itself only referenced inside a 'if
> (IS_ENABLED(CONFIG_CRYPTO_AEGIS128_SIMD))' conditional block. This is
> why it works fine most of the time.

I stand corrected: I missed the conditional registration of
crypto_aegis128_alg_simd.

Gr{oetje,eeting}s,

                        Geert
Herbert Xu Dec. 4, 2020, 7:17 a.m. UTC | #4
On Mon, Nov 30, 2020 at 01:26:20PM +0100, Ard Biesheuvel wrote:
> Geert reports that builds where CONFIG_CRYPTO_AEGIS128_SIMD is not set
> may still emit references to crypto_aegis128_update_simd(), which
> cannot be satisfied and therefore break the build. These references
> only exist in functions that can be optimized away, but apparently,
> the compiler is not always able to prove this.
> 
> So add some explicit checks for CONFIG_CRYPTO_AEGIS128_SIMD to help the
> compiler figure this out.
> 
> Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
> Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
> ---
>  crypto/aegis128-core.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)

Patch applied.  Thanks.
diff mbox series

Patch

diff --git a/crypto/aegis128-core.c b/crypto/aegis128-core.c
index 2b05f79475d3..89dc1c559689 100644
--- a/crypto/aegis128-core.c
+++ b/crypto/aegis128-core.c
@@ -89,7 +89,7 @@  static void crypto_aegis128_update_a(struct aegis_state *state,
 				     const union aegis_block *msg,
 				     bool do_simd)
 {
-	if (do_simd) {
+	if (IS_ENABLED(CONFIG_CRYPTO_AEGIS128_SIMD) && do_simd) {
 		crypto_aegis128_update_simd(state, msg);
 		return;
 	}
@@ -101,7 +101,7 @@  static void crypto_aegis128_update_a(struct aegis_state *state,
 static void crypto_aegis128_update_u(struct aegis_state *state, const void *msg,
 				     bool do_simd)
 {
-	if (do_simd) {
+	if (IS_ENABLED(CONFIG_CRYPTO_AEGIS128_SIMD) && do_simd) {
 		crypto_aegis128_update_simd(state, msg);
 		return;
 	}