poly1305: account for simd being toggled off midway

This is a very rare occurance, but we should account for it, so that the
calculations aren't wrong. Here we convert from base 2^26 back to base
2^64.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

3 files changed, 131 insertions, 26 deletions

diff --git a/src/crypto/zinc/poly1305/poly1305-arm-glue.h b/src/crypto/zinc/poly1305/poly1305-arm-glue.h
index dd3fa5a..8a3daf8 100644
--- a/src/crypto/zinc/poly1305/poly1305-arm-glue.h
+++ b/src/crypto/zinc/poly1305/poly1305-arm-glue.h
@@ -29,6 +29,58 @@ static void __init poly1305_fpu_init(void)
 #endif
 }
 
+#if defined(CONFIG_ARM64)
+struct poly1305_arch_internal {
+	union {
+		u32 h[5];
+		struct {
+			u64 h0, h1, h2;
+		};
+	};
+	u32 is_base2_26;
+	u64 r[2];
+};
+#elif defined(CONFIG_ARM)
+struct poly1305_arch_internal {
+	union {
+		u32 h[5];
+		struct {
+			u64 h0, h1;
+			u32 h2;
+		} __packed;
+	};
+	u32 r[4];
+	u32 is_base2_26;
+};
+#endif
+
+#if defined(ARM_USE_NEON)
+static void convert_to_base2_64(void *ctx)
+{
+	struct poly1305_arch_internal *state = ctx;
+	u32 cy;
+
+	if (!state->is_base2_26)
+		return;
+
+	cy = state->h[0] >> 26; state->h[0] &= 0x3ffffff; state->h[1] += cy;
+	cy = state->h[1] >> 26; state->h[1] &= 0x3ffffff; state->h[2] += cy;
+	cy = state->h[2] >> 26; state->h[2] &= 0x3ffffff; state->h[3] += cy;
+	cy = state->h[3] >> 26; state->h[3] &= 0x3ffffff; state->h[4] += cy;
+	state->h0 = ((u64)state->h[2] << 52) | ((u64)state->h[1] << 26) | state->h[0];
+	state->h1 = ((u64)state->h[4] << 40) | ((u64)state->h[3] << 14) | (state->h[2] >> 12);
+	state->h2 = state->h[4] >> 24;
+#define ULT(a, b) ((a ^ ((a ^ b) | ((a - b) ^ b))) >> (sizeof(a) * 8 - 1))
+	cy = (state->h2 >> 2) + (state->h2 & ~3ULL);
+	state->h2 &= 3;
+	state->h0 += cy;
+	state->h1 += (cy = ULT(state->h0, cy));
+	state->h2 += ULT(state->h1, cy);
+#undef ULT
+	state->is_base2_26 = 0;
+}
+#endif
+
 static inline bool poly1305_init_arch(void *ctx,
 				      const u8 key[POLY1305_KEY_SIZE])
 {
@@ -45,7 +97,9 @@ static inline bool poly1305_blocks_arch(void *ctx, const u8 *inp,
 		poly1305_blocks_neon(ctx, inp, len, padbit);
 		return true;
 	}
+	convert_to_base2_64(ctx);
 #endif
+
 	poly1305_blocks_arm(ctx, inp, len, padbit);
 	return true;
 }
@@ -59,7 +113,9 @@ static inline bool poly1305_emit_arch(void *ctx, u8 mac[POLY1305_MAC_SIZE],
 		poly1305_emit_neon(ctx, mac, nonce);
 		return true;
 	}
+	convert_to_base2_64(ctx);
 #endif
+
 	poly1305_emit_arm(ctx, mac, nonce);
 	return true;
 }
diff --git a/src/crypto/zinc/poly1305/poly1305-x86_64-glue.h b/src/crypto/zinc/poly1305/poly1305-x86_64-glue.h
index d6495cd..f821f34 100644
--- a/src/crypto/zinc/poly1305/poly1305-x86_64-glue.h
+++ b/src/crypto/zinc/poly1305/poly1305-x86_64-glue.h
@@ -61,13 +61,43 @@ static inline bool poly1305_init_arch(void *ctx,
 }
 
 struct poly1305_arch_internal {
-	u32 h[5];
-	u32 is_base2_26;
+	union {
+		struct {
+			u32 h[5];
+			u32 is_base2_26;
+		};
+		u64 hs[3];
+	};
 	u64 r[2];
 	u64 pad;
 	struct { u32 r2, r1, r4, r3; } rn[9];
 };
 
+static void convert_to_base2_64(void *ctx)
+{
+	struct poly1305_arch_internal *state = ctx;
+	u32 cy;
+
+	if (!state->is_base2_26)
+		return;
+
+	cy = state->h[0] >> 26; state->h[0] &= 0x3ffffff; state->h[1] += cy;
+	cy = state->h[1] >> 26; state->h[1] &= 0x3ffffff; state->h[2] += cy;
+	cy = state->h[2] >> 26; state->h[2] &= 0x3ffffff; state->h[3] += cy;
+	cy = state->h[3] >> 26; state->h[3] &= 0x3ffffff; state->h[4] += cy;
+	state->hs[0] = ((u64)state->h[2] << 52) | ((u64)state->h[1] << 26) | state->h[0];
+	state->hs[1] = ((u64)state->h[4] << 40) | ((u64)state->h[3] << 14) | (state->h[2] >> 12);
+	state->hs[2] = state->h[4] >> 24;
+#define ULT(a, b) ((a ^ ((a ^ b) | ((a - b) ^ b))) >> (sizeof(a) * 8 - 1))
+	cy = (state->hs[2] >> 2) + (state->hs[2] & ~3ULL);
+	state->hs[2] &= 3;
+	state->hs[0] += cy;
+	state->hs[1] += (cy = ULT(state->hs[0], cy));
+	state->hs[2] += ULT(state->hs[1], cy);
+#undef ULT
+	state->is_base2_26 = 0;
+}
+
 static inline bool poly1305_blocks_arch(void *ctx, const u8 *inp,
 					const size_t len, const u32 padbit,
 					simd_context_t *simd_context)
@@ -77,24 +107,32 @@ static inline bool poly1305_blocks_arch(void *ctx, const u8 *inp,
 	if (!poly1305_use_avx ||
 	    (len < (POLY1305_BLOCK_SIZE * 18) && !state->is_base2_26) ||
 	    !simd_use(simd_context))
-	    poly1305_blocks_x86_64(ctx, inp, len, padbit);
-	else
+		goto scalar;
+
 #ifdef CONFIG_AS_AVX512
-	if (poly1305_use_avx512)
+	if (poly1305_use_avx512) {
 		poly1305_blocks_avx512(ctx, inp, len, padbit);
-	else
+		return true;
+	}
 #endif
+
 #ifdef CONFIG_AS_AVX2
-	if (poly1305_use_avx2)
+	if (poly1305_use_avx2) {
 		poly1305_blocks_avx2(ctx, inp, len, padbit);
-	else
+		return true;
+	}
 #endif
+
 #ifdef CONFIG_AS_AVX
-	if (poly1305_use_avx)
+	if (poly1305_use_avx) {
 		poly1305_blocks_avx(ctx, inp, len, padbit);
-	else
+		return true;
+	}
 #endif
-		poly1305_blocks_x86_64(ctx, inp, len, padbit);
+
+scalar:
+	convert_to_base2_64(ctx);
+	poly1305_blocks_x86_64(ctx, inp, len, padbit);
 	return true;
 }
 
@@ -105,23 +143,17 @@ static inline bool poly1305_emit_arch(void *ctx, u8 mac[POLY1305_MAC_SIZE],
 	struct poly1305_arch_internal *state = ctx;
 
 	if (!poly1305_use_avx || !state->is_base2_26 ||!simd_use(simd_context))
-		poly1305_emit_x86_64(ctx, mac, nonce);
-	else
-#ifdef CONFIG_AS_AVX512
-	if (poly1305_use_avx512)
-		poly1305_emit_avx(ctx, mac, nonce);
-	else
-#endif
-#ifdef CONFIG_AS_AVX2
-	if (poly1305_use_avx2)
-		poly1305_emit_avx(ctx, mac, nonce);
-	else
-#endif
+		goto scalar;
+
 #ifdef CONFIG_AS_AVX
-	if (poly1305_use_avx)
+	if (poly1305_use_avx || poly1305_use_avx2 || poly1305_use_avx512) {
 		poly1305_emit_avx(ctx, mac, nonce);
-	else
+		return true;
+	}
 #endif
-		poly1305_emit_x86_64(ctx, mac, nonce);
+
+scalar:
+	convert_to_base2_64(ctx);
+	poly1305_emit_x86_64(ctx, mac, nonce);
 	return true;
 }
diff --git a/src/crypto/zinc/selftest/poly1305.h b/src/crypto/zinc/selftest/poly1305.h
index 1439c98..71bee1b 100644
--- a/src/crypto/zinc/selftest/poly1305.h
+++ b/src/crypto/zinc/selftest/poly1305.h
@@ -862,6 +862,23 @@ static bool __init poly1305_selftest(void)
 					i + 1, j);
 				success = false;
 			}
+
+			memset(out, 0, sizeof(out));
+			memset(&poly1305, 0, sizeof(poly1305));
+			poly1305_init(&poly1305, poly1305_testvecs[i].key);
+			poly1305_update(&poly1305, poly1305_testvecs[i].input,
+					j, &simd_context);
+			poly1305_update(&poly1305,
+					poly1305_testvecs[i].input + j,
+					poly1305_testvecs[i].ilen - j,
+					(simd_context_t []){ HAVE_NO_SIMD });
+			poly1305_final(&poly1305, out, &simd_context);
+			if (memcmp(out, poly1305_testvecs[i].output,
+				   POLY1305_MAC_SIZE)) {
+				pr_info("poly1305 self-test %zu (split %zu, mixed simd): FAIL\n",
+					i + 1, j);
+				success = false;
+			}
 			simd_relax(&simd_context);
 		}
 	}