[src/trunk]: src/sys/crypto/aes/arch/arm Fix AES NEON code for big-endian sof...

[riastradh <riastradh%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/8ad1f7a9d94a
branches:  trunk
changeset: 974915:8ad1f7a9d94a
user:      riastradh <riastradh%NetBSD.org@localhost>
date:      Sun Aug 16 18:02:03 2020 +0000

description:
Fix AES NEON code for big-endian softfp ARM.

...which is how the kernel runs.  Switch to using __SOFTFP__ for
consistency with how it gets exposed to C, although I'm not sure how
to get it defined automagically in the toolchain for .S files so
that's set manually in files.aesneon for now.

diffstat:

 sys/crypto/aes/arch/arm/aes_neon_32.S |  32 +++++++++++++++++++++++++-------
 sys/crypto/aes/arch/arm/files.aesneon |   4 +++-
 2 files changed, 28 insertions(+), 8 deletions(-)

diffs (105 lines):

diff -r 30e1eb1c1879 -r 8ad1f7a9d94a sys/crypto/aes/arch/arm/aes_neon_32.S
--- a/sys/crypto/aes/arch/arm/aes_neon_32.S     Sun Aug 16 17:58:48 2020 +0000
+++ b/sys/crypto/aes/arch/arm/aes_neon_32.S     Sun Aug 16 18:02:03 2020 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: aes_neon_32.S,v 1.5 2020/08/08 14:47:01 riastradh Exp $        */
+/*     $NetBSD: aes_neon_32.S,v 1.6 2020/08/16 18:02:03 riastradh Exp $        */
 
 /*-
  * Copyright (c) 2020 The NetBSD Foundation, Inc.
@@ -28,7 +28,7 @@
 
 #include <arm/asm.h>
 
-RCSID("$NetBSD: aes_neon_32.S,v 1.5 2020/08/08 14:47:01 riastradh Exp $")
+RCSID("$NetBSD: aes_neon_32.S,v 1.6 2020/08/16 18:02:03 riastradh Exp $")
 
        .fpu    neon
 
@@ -228,15 +228,19 @@
  * aes_neon_enc1(const struct aesenc *enc@r0, uint8x16_t x@q0,
  *     unsigned nrounds@r1)
  *
- *     With -mfloat-abi=soft(fp) (here spelled `#ifdef _KERNEL'):
+ *     With -mfloat-abi=soft(fp) (i.e., __SOFTFP__):
  *
  * uint8x16_t@(r0,r1,r2,r3)
  * aes_neon_enc1(const struct aesenc *enc@r0,
  *     uint8x16_t x@(r2,r3,sp[0],sp[4]), nrounds@sp[8])
  */
 ENTRY(aes_neon_enc1)
-#ifdef _KERNEL
+#ifdef __SOFTFP__
+#ifdef __ARM_BIG_ENDIAN
+       vmov    d0, r3, r2              /* d0 := x lo */
+#else
        vmov    d0, r2, r3              /* d0 := x lo */
+#endif
        vldr    d1, [sp]                /* d1 := x hi */
        ldr     r1, [sp, #8]            /* r1 := nrounds */
 #endif
@@ -434,10 +438,15 @@
 
        vpop    {d8-d15}
        pop     {r4, r5, r6, r7, r8, r10, r11, lr}
-#ifdef _KERNEL
+#ifdef __SOFTFP__
+#ifdef __ARM_BIG_ENDIAN
+       vmov    r1, r0, d0
+       vmov    r3, r2, d1
+#else
        vmov    r0, r1, d0
        vmov    r2, r3, d1
 #endif
+#endif
        bx      lr
 END(aes_neon_enc1)
 
@@ -457,8 +466,12 @@
  *     uint8x16_t x@(r2,r3,sp[0],sp[4]), nrounds@sp[8])
  */
 ENTRY(aes_neon_dec1)
-#ifdef _KERNEL
+#ifdef __SOFTFP__
+#ifdef __ARM_BIG_ENDIAN
+       vmov    d0, r3, r2              /* d0 := x lo */
+#else
        vmov    d0, r2, r3              /* d0 := x lo */
+#endif
        vldr    d1, [sp]                /* d1 := x hi */
        ldr     r1, [sp, #8]            /* r1 := nrounds */
 #endif
@@ -669,9 +682,14 @@
 
        vpop    {d8-d15}
        pop     {r4, r5, r6, r7, r8, r10, r11, lr}
-#ifdef _KERNEL
+#ifdef __SOFTFP__
+#ifdef __ARM_BIG_ENDIAN
+       vmov    r1, r0, d0
+       vmov    r3, r2, d1
+#else
        vmov    r0, r1, d0
        vmov    r2, r3, d1
 #endif
+#endif
        bx      lr
 END(aes_neon_dec1)
diff -r 30e1eb1c1879 -r 8ad1f7a9d94a sys/crypto/aes/arch/arm/files.aesneon
--- a/sys/crypto/aes/arch/arm/files.aesneon     Sun Aug 16 17:58:48 2020 +0000
+++ b/sys/crypto/aes/arch/arm/files.aesneon     Sun Aug 16 18:02:03 2020 +0000
@@ -1,4 +1,4 @@
-#      $NetBSD: files.aesneon,v 1.3 2020/06/30 17:03:13 riastradh Exp $
+#      $NetBSD: files.aesneon,v 1.4 2020/08/16 18:02:03 riastradh Exp $
 
 ifdef aarch64
 makeoptions    aes     "COPTS.aes_neon.c"+="-march=armv8-a"
@@ -8,6 +8,8 @@
 makeoptions    aes     "COPTS.aes_neon_subr.c"+="-mfloat-abi=softfp -mfpu=neon"
 endif
 
+makeoptions    aes     "AOPTS.aes_neon_32.S"+="-D__SOFTFP__"
+
 file   crypto/aes/arch/arm/aes_neon.c          aes & (cpu_cortex | aarch64)
 file   crypto/aes/arch/arm/aes_neon_impl.c     aes & (cpu_cortex | aarch64)
 file   crypto/aes/arch/arm/aes_neon_subr.c     aes & (cpu_cortex | aarch64)