vppinfra: native AES-CTR implementation

Type: feature
Change-Id: I7ef3277edaeb266fbd3c8c9355d4443002ed2311
Signed-off-by: Damjan Marion <damarion@cisco.com>

src/plugins/crypto_native/aes_cbc.c

@@ -26,21 +26,18 @@
 #endif
 
 #if defined(__VAES__) && defined(__AVX512F__)
-#define N		  16
 #define u8xN		  u8x64
 #define u32xN		  u32x16
 #define u32xN_min_scalar  u32x16_min_scalar
 #define u32xN_is_all_zero u32x16_is_all_zero
 #define u32xN_splat	  u32x16_splat
 #elif defined(__VAES__)
-#define N		  8
 #define u8xN		  u8x32
 #define u32xN		  u32x8
 #define u32xN_min_scalar  u32x8_min_scalar
 #define u32xN_is_all_zero u32x8_is_all_zero
 #define u32xN_splat	  u32x8_splat
 #else
-#define N		  4
 #define u8xN		  u8x16
 #define u32xN		  u32x4
 #define u32xN_min_scalar  u32x4_min_scalar
@@ -58,17 +55,17 @@ aes_ops_enc_aes_cbc (vlib_main_t * vm, vnet_crypto_op_t * ops[],
   u32 i, j, count, n_left = n_ops;
   u32xN placeholder_mask = { };
   u32xN len = { };
-  vnet_crypto_key_index_t key_index[N];
-  u8 *src[N] = { };
-  u8 *dst[N] = { };
+  vnet_crypto_key_index_t key_index[N_AES_BYTES];
+  u8 *src[N_AES_BYTES] = {};
+  u8 *dst[N_AES_BYTES] = {};
   u8xN r[4] = {};
   u8xN k[15][4] = {};
 
-  for (i = 0; i < N; i++)
+  for (i = 0; i < N_AES_BYTES; i++)
     key_index[i] = ~0;
 
 more:
-  for (i = 0; i < N; i++)
+  for (i = 0; i < N_AES_BYTES; i++)
     if (len[i] == 0)
       {
 	if (n_left == 0)
@@ -160,16 +157,16 @@ more:
 
       for (j = 1; j < rounds; j++)
 	{
-	  r[0] = aes_enc_round (r[0], k[j][0]);
-	  r[1] = aes_enc_round (r[1], k[j][1]);
-	  r[2] = aes_enc_round (r[2], k[j][2]);
-	  r[3] = aes_enc_round (r[3], k[j][3]);
+	  r[0] = aes_enc_round_x1 (r[0], k[j][0]);
+	  r[1] = aes_enc_round_x1 (r[1], k[j][1]);
+	  r[2] = aes_enc_round_x1 (r[2], k[j][2]);
+	  r[3] = aes_enc_round_x1 (r[3], k[j][3]);
 	}
 
-      r[0] = aes_enc_last_round (r[0], k[j][0]);
-      r[1] = aes_enc_last_round (r[1], k[j][1]);
-      r[2] = aes_enc_last_round (r[2], k[j][2]);
-      r[3] = aes_enc_last_round (r[3], k[j][3]);
+      r[0] = aes_enc_last_round_x1 (r[0], k[j][0]);
+      r[1] = aes_enc_last_round_x1 (r[1], k[j][1]);
+      r[2] = aes_enc_last_round_x1 (r[2], k[j][2]);
+      r[3] = aes_enc_last_round_x1 (r[3], k[j][3]);
 
       aes_block_store (dst[0] + i, r[0]);
       aes_block_store (dst[1] + i, r[1]);
@@ -201,7 +198,7 @@ more:
 
   len -= u32xN_splat (count);
 
-  for (i = 0; i < N; i++)
+  for (i = 0; i < N_AES_BYTES; i++)
     {
       src[i] += count;
       dst[i] += count;

src/vppinfra/CMakeLists.txt

@@ -132,6 +132,7 @@ set(VPPINFRA_HEADERS
   crypto/ghash.h
   crypto/aes.h
   crypto/aes_cbc.h
+  crypto/aes_ctr.h
   crypto/aes_gcm.h
   crypto/poly1305.h
   dlist.h
@@ -285,6 +286,7 @@ endif(VPP_BUILD_VPPINFRA_TESTS)
 
 set(test_files
   test/aes_cbc.c
+  test/aes_ctr.c
   test/aes_gcm.c
   test/poly1305.c
   test/array_mask.c

src/vppinfra/crypto/aes.h

@@ -15,8 +15,8 @@
  *------------------------------------------------------------------
  */
 
-#ifndef __aesni_h__
-#define __aesni_h__
+#ifndef __aes_h__
+#define __aes_h__
 
 typedef enum
 {
@@ -35,7 +35,7 @@ aes_block_load (u8 * p)
 }
 
 static_always_inline u8x16
-aes_enc_round (u8x16 a, u8x16 k)
+aes_enc_round_x1 (u8x16 a, u8x16 k)
 {
 #if defined (__AES__)
   return (u8x16) _mm_aesenc_si128 ((__m128i) a, (__m128i) k);
@@ -97,7 +97,7 @@ aes_dec_last_round_x2 (u8x32 a, u8x32 k)
 #endif
 
 static_always_inline u8x16
-aes_enc_last_round (u8x16 a, u8x16 k)
+aes_enc_last_round_x1 (u8x16 a, u8x16 k)
 {
 #if defined (__AES__)
   return (u8x16) _mm_aesenclast_si128 ((__m128i) a, (__m128i) k);
@@ -109,13 +109,13 @@ aes_enc_last_round (u8x16 a, u8x16 k)
 #ifdef __x86_64__
 
 static_always_inline u8x16
-aes_dec_round (u8x16 a, u8x16 k)
+aes_dec_round_x1 (u8x16 a, u8x16 k)
 {
   return (u8x16) _mm_aesdec_si128 ((__m128i) a, (__m128i) k);
 }
 
 static_always_inline u8x16
-aes_dec_last_round (u8x16 a, u8x16 k)
+aes_dec_last_round_x1 (u8x16 a, u8x16 k)
 {
   return (u8x16) _mm_aesdeclast_si128 ((__m128i) a, (__m128i) k);
 }
@@ -133,8 +133,8 @@ aes_encrypt_block (u8x16 block, const u8x16 * round_keys, aes_key_size_t ks)
   int rounds = AES_KEY_ROUNDS (ks);
   block ^= round_keys[0];
   for (int i = 1; i < rounds; i += 1)
-    block = aes_enc_round (block, round_keys[i]);
-  return aes_enc_last_round (block, round_keys[rounds]);
+    block = aes_enc_round_x1 (block, round_keys[i]);
+  return aes_enc_last_round_x1 (block, round_keys[rounds]);
 }
 
 static_always_inline u8x16
@@ -427,13 +427,67 @@ aes_key_enc_to_dec (u8x16 * ke, u8x16 * kd, aes_key_size_t ks)
 
   kd[rounds / 2] = aes_inv_mix_column (ke[rounds / 2]);
 }
+#if defined(__VAES__) && defined(__AVX512F__)
+#define N_AES_LANES		   4
+#define aes_load_partial(p, n)	   u8x64_load_partial ((u8 *) (p), n)
+#define aes_store_partial(v, p, n) u8x64_store_partial (v, (u8 *) (p), n)
+#define aes_reflect(r)		   u8x64_reflect_u8x16 (r)
+typedef u8x64 aes_data_t;
+typedef u8x64u aes_mem_t;
+typedef u32x16 aes_counter_t;
+#elif defined(__VAES__)
+#define N_AES_LANES		   2
+#define aes_load_partial(p, n)	   u8x32_load_partial ((u8 *) (p), n)
+#define aes_store_partial(v, p, n) u8x32_store_partial (v, (u8 *) (p), n)
+#define aes_reflect(r)		   u8x32_reflect_u8x16 (r)
+typedef u8x32 aes_data_t;
+typedef u8x32u aes_mem_t;
+typedef u32x8 aes_counter_t;
+#else
+#define N_AES_LANES		   1
+#define aes_load_partial(p, n)	   u8x16_load_partial ((u8 *) (p), n)
+#define aes_store_partial(v, p, n) u8x16_store_partial (v, (u8 *) (p), n)
+#define aes_reflect(r)		   u8x16_reflect (r)
+typedef u8x16 aes_data_t;
+typedef u8x16u aes_mem_t;
+typedef u32x4 aes_counter_t;
+#endif
 
-#endif /* __aesni_h__ */
+#define N_AES_BYTES (N_AES_LANES * 16)
 
-/*
- * fd.io coding-style-patch-verification: ON
- *
- * Local Variables:
- * eval: (c-set-style "gnu")
- * End:
- */
+typedef union
+{
+  u8x16 x1;
+  u8x32 x2;
+  u8x64 x4;
+  u8x16 lanes[4];
+} aes_expaned_key_t;
+
+static_always_inline void
+aes_enc_round (aes_data_t *r, const aes_expaned_key_t *ek, uword n_blocks)
+{
+  for (int i = 0; i < n_blocks; i++)
+#if N_AES_LANES == 4
+    r[i] = aes_enc_round_x4 (r[i], ek->x4);
+#elif N_AES_LANES == 2
+    r[i] = aes_enc_round_x2 (r[i], ek->x2);
+#else
+    r[i] = aes_enc_round_x1 (r[i], ek->x1);
+#endif
+}
+
+static_always_inline void
+aes_enc_last_round (aes_data_t *r, aes_data_t *d, const aes_expaned_key_t *ek,
+		    uword n_blocks)
+{
+  for (int i = 0; i < n_blocks; i++)
+#if N_AES_LANES == 4
+    d[i] ^= r[i] = aes_enc_last_round_x4 (r[i], ek->x4);
+#elif N_AES_LANES == 2
+    d[i] ^= r[i] = aes_enc_last_round_x2 (r[i], ek->x2);
+#else
+    d[i] ^= r[i] = aes_enc_last_round_x1 (r[i], ek->x1);
+#endif
+}
+
+#endif /* __aes_h__ */

src/vppinfra/crypto/aes_cbc.h

@@ -30,8 +30,8 @@ clib_aes_cbc_encrypt (const aes_cbc_key_data_t *kd, const u8 *src, uword len,
 #if __x86_64__
       r = u8x16_xor3 (r, *(u8x16u *) (src + i), k[0]);
       for (j = 1; j < rounds; j++)
-	r = aes_enc_round (r, k[j]);
-      r = aes_enc_last_round (r, k[rounds]);
+	r = aes_enc_round_x1 (r, k[j]);
+      r = aes_enc_last_round_x1 (r, k[rounds]);
 #else
       r ^= *(u8x16u *) (src + i);
       for (j = 1; j < rounds - 1; j++)
@@ -85,16 +85,16 @@ aes_cbc_dec (const u8x16 *k, u8x16u *src, u8x16u *dst, u8x16u *iv, int count,
 
       for (int i = 1; i < rounds; i++)
 	{
-	  r[0] = aes_dec_round (r[0], k[i]);
-	  r[1] = aes_dec_round (r[1], k[i]);
-	  r[2] = aes_dec_round (r[2], k[i]);
-	  r[3] = aes_dec_round (r[3], k[i]);
+	  r[0] = aes_dec_round_x1 (r[0], k[i]);
+	  r[1] = aes_dec_round_x1 (r[1], k[i]);
+	  r[2] = aes_dec_round_x1 (r[2], k[i]);
+	  r[3] = aes_dec_round_x1 (r[3], k[i]);
 	}
 
-      r[0] = aes_dec_last_round (r[0], k[rounds]);
-      r[1] = aes_dec_last_round (r[1], k[rounds]);
-      r[2] = aes_dec_last_round (r[2], k[rounds]);
-      r[3] = aes_dec_last_round (r[3], k[rounds]);
+      r[0] = aes_dec_last_round_x1 (r[0], k[rounds]);
+      r[1] = aes_dec_last_round_x1 (r[1], k[rounds]);
+      r[2] = aes_dec_last_round_x1 (r[2], k[rounds]);
+      r[3] = aes_dec_last_round_x1 (r[3], k[rounds]);
 #else
       for (int i = 0; i < rounds - 1; i++)
 	{
@@ -125,8 +125,8 @@ aes_cbc_dec (const u8x16 *k, u8x16u *src, u8x16u *dst, u8x16u *iv, int count,
 #if __x86_64__
       r[0] ^= k[0];
       for (int i = 1; i < rounds; i++)
-	r[0] = aes_dec_round (r[0], k[i]);
-      r[0] = aes_dec_last_round (r[0], k[rounds]);
+	r[0] = aes_dec_round_x1 (r[0], k[i]);
+      r[0] = aes_dec_last_round_x1 (r[0], k[rounds]);
 #else
       c[0] = r[0] = src[0];
       for (int i = 0; i < rounds - 1; i++)
@@ -469,8 +469,8 @@ aes2_cbc_dec (const u8x16 *k, u8x32u *src, u8x32u *dst, u8x16u *iv, int count,
     {
       u8x16 rl = *(u8x16u *) src ^ k[0];
       for (i = 1; i < rounds; i++)
-	rl = aes_dec_round (rl, k[i]);
-      rl = aes_dec_last_round (rl, k[i]);
+	rl = aes_dec_round_x1 (rl, k[i]);
+      rl = aes_dec_last_round_x1 (rl, k[i]);
       *(u8x16u *) dst = rl ^ u8x32_extract_hi (f);
     }
 }

src/vppinfra/crypto/aes_ctr.h

@@ -0,0 +1,190 @@
+/* SPDX-License-Identifier: Apache-2.0
+ * Copyright(c) 2024 Cisco Systems, Inc.
+ */
+
+#ifndef __crypto_aes_ctr_h__
+#define __crypto_aes_ctr_h__
+
+#include <vppinfra/clib.h>
+#include <vppinfra/vector.h>
+#include <vppinfra/cache.h>
+#include <vppinfra/string.h>
+#include <vppinfra/crypto/aes.h>
+
+typedef struct
+{
+  const aes_expaned_key_t exp_key[AES_KEY_ROUNDS (AES_KEY_256) + 1];
+} aes_ctr_key_data_t;
+
+typedef struct
+{
+  const aes_expaned_key_t exp_key[AES_KEY_ROUNDS (AES_KEY_256) + 1];
+  aes_counter_t ctr;		   /* counter (reflected) */
+  u8 keystream_bytes[N_AES_BYTES]; /* keystream leftovers */
+  u32 n_keystream_bytes;	   /* number of keystream leftovers */
+} aes_ctr_ctx_t;
+
+static_always_inline aes_counter_t
+aes_ctr_one_block (aes_ctr_ctx_t *ctx, aes_counter_t ctr, const u8 *src,
+		   u8 *dst, u32 n_parallel, u32 n_bytes, int rounds, int last)
+{
+  u32 __clib_aligned (N_AES_BYTES)
+  inc[] = { N_AES_LANES, 0, 0, 0, N_AES_LANES, 0, 0, 0,
+	    N_AES_LANES, 0, 0, 0, N_AES_LANES, 0, 0, 0 };
+  const aes_expaned_key_t *k = ctx->exp_key;
+  const aes_mem_t *sv = (aes_mem_t *) src;
+  aes_mem_t *dv = (aes_mem_t *) dst;
+  aes_data_t d[4], t[4];
+  u32 r;
+
+  n_bytes -= (n_parallel - 1) * N_AES_BYTES;
+
+  /* AES First Round */
+  for (int i = 0; i < n_parallel; i++)
+    {
+#if N_AES_LANES == 4
+      t[i] = k[0].x4 ^ (u8x64) aes_reflect ((u8x64) ctr);
+#elif N_AES_LANES == 2
+      t[i] = k[0].x2 ^ (u8x32) aes_reflect ((u8x32) ctr);
+#else
+      t[i] = k[0].x1 ^ (u8x16) aes_reflect ((u8x16) ctr);
+#endif
+      ctr += *(aes_counter_t *) inc;
+    }
+
+  /* Load Data */
+  for (int i = 0; i < n_parallel - last; i++)
+    d[i] = sv[i];
+
+  if (last)
+    d[n_parallel - 1] =
+      aes_load_partial ((u8 *) (sv + n_parallel - 1), n_bytes);
+
+  /* AES Intermediate Rounds */
+  for (r = 1; r < rounds; r++)
+    aes_enc_round (t, k + r, n_parallel);
+
+  /* AES Last Round */
+  aes_enc_last_round (t, d, k + r, n_parallel);
+
+  /* Store Data */
+  for (int i = 0; i < n_parallel - last; i++)
+    dv[i] = d[i];
+
+  if (last)
+    {
+      aes_store_partial (d[n_parallel - 1], dv + n_parallel - 1, n_bytes);
+      *(aes_data_t *) ctx->keystream_bytes = t[n_parallel - 1];
+      ctx->n_keystream_bytes = N_AES_BYTES - n_bytes;
+    }
+
+  return ctr;
+}
+
+static_always_inline void
+clib_aes_ctr_init (aes_ctr_ctx_t *ctx, const aes_ctr_key_data_t *kd,
+		   const u8 *iv, aes_key_size_t ks)
+{
+  u32x4 ctr = (u32x4) u8x16_reflect (*(u8x16u *) iv);
+#if N_AES_LANES == 4
+  ctx->ctr = (aes_counter_t) u32x16_splat_u32x4 (ctr) +
+	     (u32x16){ 0, 0, 0, 0, 1, 0, 0, 0, 2, 0, 0, 0, 3, 0, 0, 0 };
+#elif N_AES_LANES == 2
+  ctx->ctr = (aes_counter_t) u32x8_splat_u32x4 (ctr) +
+	     (u32x8){ 0, 0, 0, 0, 1, 0, 0, 0 };
+#else
+  ctx->ctr = ctr;
+#endif
+  for (int i = 0; i < AES_KEY_ROUNDS (ks) + 1; i++)
+    ((aes_expaned_key_t *) ctx->exp_key)[i] = kd->exp_key[i];
+  ctx->n_keystream_bytes = 0;
+}
+
+static_always_inline void
+clib_aes_ctr_transform (aes_ctr_ctx_t *ctx, const u8 *src, u8 *dst,
+			u32 n_bytes, aes_key_size_t ks)
+{
+  int r = AES_KEY_ROUNDS (ks);
+  aes_counter_t ctr = ctx->ctr;
+
+  if (ctx->n_keystream_bytes)
+    {
+      u8 *ks = ctx->keystream_bytes + N_AES_BYTES - ctx->n_keystream_bytes;
+
+      if (ctx->n_keystream_bytes >= n_bytes)
+	{
+	  for (int i = 0; i < n_bytes; i++)
+	    dst[i] = src[i] ^ ks[i];
+	  ctx->n_keystream_bytes -= n_bytes;
+	  return;
+	}
+
+      for (int i = 0; i < ctx->n_keystream_bytes; i++)
+	dst++[0] = src++[0] ^ ks[i];
+
+      n_bytes -= ctx->n_keystream_bytes;
+      ctx->n_keystream_bytes = 0;
+    }
+
+  /* main loop */
+  for (int n = 4 * N_AES_BYTES; n_bytes >= n; n_bytes -= n, dst += n, src += n)
+    ctr = aes_ctr_one_block (ctx, ctr, src, dst, 4, n, r, 0);
+
+  if (n_bytes)
+    {
+      if (n_bytes > 3 * N_AES_BYTES)
+	ctr = aes_ctr_one_block (ctx, ctr, src, dst, 4, n_bytes, r, 1);
+      else if (n_bytes > 2 * N_AES_BYTES)
+	ctr = aes_ctr_one_block (ctx, ctr, src, dst, 3, n_bytes, r, 1);
+      else if (n_bytes > N_AES_BYTES)
+	ctr = aes_ctr_one_block (ctx, ctr, src, dst, 2, n_bytes, r, 1);
+      else
+	ctr = aes_ctr_one_block (ctx, ctr, src, dst, 1, n_bytes, r, 1);
+    }
+  else
+    ctx->n_keystream_bytes = 0;
+
+  ctx->ctr = ctr;
+}
+
+static_always_inline void
+clib_aes_ctr_key_expand (aes_ctr_key_data_t *kd, const u8 *key,
+			 aes_key_size_t ks)
+{
+  u8x16 ek[AES_KEY_ROUNDS (AES_KEY_256) + 1];
+  aes_expaned_key_t *k = (aes_expaned_key_t *) kd->exp_key;
+
+  /* expand AES key */
+  aes_key_expand (ek, key, ks);
+  for (int i = 0; i < AES_KEY_ROUNDS (ks) + 1; i++)
+    k[i].lanes[0] = k[i].lanes[1] = k[i].lanes[2] = k[i].lanes[3] = ek[i];
+}
+
+static_always_inline void
+clib_aes128_ctr (const aes_ctr_key_data_t *kd, const u8 *src, u32 n_bytes,
+		 const u8 *iv, u8 *dst)
+{
+  aes_ctr_ctx_t ctx;
+  clib_aes_ctr_init (&ctx, kd, iv, AES_KEY_128);
+  clib_aes_ctr_transform (&ctx, src, dst, n_bytes, AES_KEY_128);
+}
+
+static_always_inline void
+clib_aes192_ctr (const aes_ctr_key_data_t *kd, const u8 *src, u32 n_bytes,
+		 const u8 *iv, u8 *dst)
+{
+  aes_ctr_ctx_t ctx;
+  clib_aes_ctr_init (&ctx, kd, iv, AES_KEY_192);
+  clib_aes_ctr_transform (&ctx, src, dst, n_bytes, AES_KEY_192);
+}
+
+static_always_inline void
+clib_aes256_ctr (const aes_ctr_key_data_t *kd, const u8 *src, u32 n_bytes,
+		 const u8 *iv, u8 *dst)
+{
+  aes_ctr_ctx_t ctx;
+  clib_aes_ctr_init (&ctx, kd, iv, AES_KEY_256);
+  clib_aes_ctr_transform (&ctx, src, dst, n_bytes, AES_KEY_256);
+}
+
+#endif /* __crypto_aes_ctr_h__ */