Merge branch '2.4.x' into 2.5.x
[catacomb] / symm / chacha.c
index 5683c8e..90a4c67 100644 (file)
@@ -41,6 +41,7 @@
 #include "grand.h"
 #include "keysz.h"
 #include "paranoia.h"
+#include "rsvr.h"
 
 /*----- Global variables --------------------------------------------------*/
 
@@ -62,26 +63,43 @@ const octet chacha_keysz[] = { KSZ_SET, 32, 16, 10, 0 };
  *             the feedforward step.
  */
 
-CPU_DISPATCH(static, (void),
-            void, core, (unsigned r, const chacha_matrix src,
-                         chacha_matrix dest),
-            (r, src, dest),
-            pick_core, simple_core);
+CPU_DISPATCH(static, (void), void, core,
+            (unsigned r, const chacha_matrix src, chacha_matrix dest),
+            (r, src, dest), pick_core, simple_core);
 
 static void simple_core(unsigned r, const chacha_matrix src,
                        chacha_matrix dest)
   { CHACHA_nR(dest, src, r); CHACHA_FFWD(dest, src); }
 
-#ifdef CPUFAM_X86
-extern core__functype chacha_core_x86_sse2;
+#if CPUFAM_X86 || CPUFAM_AMD64
+extern core__functype chacha_core_x86ish_sse2;
+extern core__functype chacha_core_x86ish_avx;
+#endif
+
+#if CPUFAM_ARMEL
+extern core__functype chacha_core_arm_neon;
+#endif
+
+#if CPUFAM_ARM64
+extern core__functype chacha_core_arm64;
 #endif
 
 static core__functype *pick_core(void)
 {
-#ifdef CPUFAM_X86
-  DISPATCH_PICK_COND(chacha_core, chacha_core_x86_sse2,
+#if CPUFAM_X86 || CPUFAM_AMD64
+  DISPATCH_PICK_COND(chacha_core, chacha_core_x86ish_avx,
+                    cpu_feature_p(CPUFEAT_X86_AVX));
+  DISPATCH_PICK_COND(chacha_core, chacha_core_x86ish_sse2,
                     cpu_feature_p(CPUFEAT_X86_SSE2));
 #endif
+#if CPUFAM_ARMEL
+  DISPATCH_PICK_COND(chacha_core, chacha_core_arm_neon,
+                    cpu_feature_p(CPUFEAT_ARM_NEON));
+#endif
+#if CPUFAM_ARM64
+  DISPATCH_PICK_COND(chacha_core, chacha_core_arm64,
+                    cpu_feature_p(CPUFEAT_ARM_NEON));
+#endif
   DISPATCH_PICK_FALLBACK(chacha_core, simple_core);
 }
 
@@ -136,6 +154,8 @@ static void populate(chacha_matrix a, const void *key, size_t ksz)
 
 /*----- ChaCha implementation ---------------------------------------------*/
 
+static const octet zerononce[XCHACHA_NONCESZ];
+
 /* --- @chacha_init@ --- *
  *
  * Arguments:  @chacha_ctx *ctx@ = context to fill in
@@ -151,16 +171,15 @@ static void populate(chacha_matrix a, const void *key, size_t ksz)
 void chacha_init(chacha_ctx *ctx, const void *key, size_t ksz,
                  const void *nonce)
 {
-  static const octet zerononce[CHACHA_NONCESZ];
-
   populate(ctx->a, key, ksz);
   chacha_setnonce(ctx, nonce ? nonce : zerononce);
 }
 
-/* --- @chacha_setnonce@ --- *
+/* --- @chacha_setnonce{,_ietf}@ --- *
  *
  * Arguments:  @chacha_ctx *ctx@ = pointer to context
- *             @const void *nonce@ = the nonce (@CHACHA_NONCESZ@ bytes)
+ *             @const void *nonce@ = the nonce (@CHACHA_NONCESZ@ or
+ *                     @CHACHA_IETF_NONCESZ@ bytes)
  *
  * Returns:    ---
  *
@@ -178,10 +197,20 @@ void chacha_setnonce(chacha_ctx *ctx, const void *nonce)
   chacha_seek(ctx, 0);
 }
 
-/* --- @chacha_seek@, @chacha_seeku64@ --- *
+void chacha_setnonce_ietf(chacha_ctx *ctx, const void *nonce)
+{
+  const octet *n = nonce;
+
+  ctx->a[13] = LOAD32_L(n + 0);
+  ctx->a[14] = LOAD32_L(n + 4);
+  ctx->a[15] = LOAD32_L(n + 8);
+  chacha_seek_ietf(ctx, 0);
+}
+
+/* --- @chacha_seek{,u64,_ietf}@ --- *
  *
  * Arguments:  @chacha_ctx *ctx@ = pointer to context
- *             @unsigned long i@, @kludge64 i@ = new position to set
+ *             @unsigned long i@, @kludge64 i@, @uint32 i@ = new position
  *
  * Returns:    ---
  *
@@ -197,10 +226,13 @@ void chacha_seek(chacha_ctx *ctx, unsigned long i)
 void chacha_seeku64(chacha_ctx *ctx, kludge64 i)
 {
   ctx->a[12] = LO64(i); ctx->a[13] = HI64(i);
-  ctx->bufi = CHACHA_OUTSZ;
+  ctx->off = CHACHA_OUTSZ;
 }
 
-/* --- @chacha_tell@, @chacha_tellu64@ --- *
+void chacha_seek_ietf(chacha_ctx *ctx, uint32 i)
+  { ctx->a[12] = i; }
+
+/* --- @chacha_tell{,u64,_ietf}@ --- *
  *
  * Arguments:  @chacha_ctx *ctx@ = pointer to context
  *
@@ -212,9 +244,12 @@ unsigned long chacha_tell(chacha_ctx *ctx)
   { kludge64 i = chacha_tellu64(ctx); return (GET64(unsigned long, i)); }
 
 kludge64 chacha_tellu64(chacha_ctx *ctx)
-  { kludge64 i; SET64(i, ctx->a[9], ctx->a[8]); return (i); }
+  { kludge64 i; SET64(i, ctx->a[13], ctx->a[12]); return (i); }
+
+uint32 chacha_tell_ietf(chacha_ctx *ctx)
+  { return (ctx->a[12]); }
 
-/* --- @chacha{,12,8}_encrypt@ --- *
+/* --- @chacha{20,12,8}_encrypt@ --- *
  *
  * Arguments:  @chacha_ctx *ctx@ = pointer to context
  *             @const void *src@ = source buffer (or null)
@@ -231,6 +266,8 @@ kludge64 chacha_tellu64(chacha_ctx *ctx)
  *             to @dest@.
  */
 
+static const rsvr_policy policy = { 0, CHACHA_OUTSZ, CHACHA_OUTSZ };
+
 #define CHACHA_ENCRYPT(r, ctx, src, dest, sz)                          \
   chacha##r##_encrypt(ctx, src, dest, sz)
 #define DEFENCRYPT(r)                                                  \
@@ -240,41 +277,40 @@ kludge64 chacha_tellu64(chacha_ctx *ctx)
     chacha_matrix b;                                                   \
     const octet *s = src;                                              \
     octet *d = dest;                                                   \
-    size_t n;                                                          \
+    rsvr_plan plan;                                                    \
     kludge64 pos, delta;                                               \
                                                                        \
-    SALSA20_OUTBUF(ctx, d, s, sz);                                     \
-    if (!sz) return;                                                   \
-                                                                       \
-    if (!dest) {                                                       \
-      n = sz/CHACHA_OUTSZ;                                             \
-      pos = chacha_tellu64(ctx);                                       \
-      ASSIGN64(delta, n);                                              \
-      ADD64(pos, pos, delta);                                          \
-      chacha_seeku64(ctx, pos);                                                \
-      sz = sz%CHACHA_OUTSZ;                                            \
-    } else if (!src) {                                                 \
-      while (sz >= CHACHA_OUTSZ) {                                     \
-       core(r, ctx->a, b);                                             \
-       CHACHA_STEP(ctx->a);                                            \
-       SALSA20_GENFULL(b, d);                                          \
-       sz -= CHACHA_OUTSZ;                                             \
+    rsvr_mkplan(&plan, &policy, ctx->off, sz);                         \
+                                                                       \
+    if (plan.head) {                                                   \
+      if (!ctx->off) {                                                 \
+       core(r, ctx->a, b); CHACHA_STEP(ctx->a);                        \
+       SALSA20_PREPBUF(ctx, b);                                        \
       }                                                                        \
-    } else {                                                           \
-      while (sz >= CHACHA_OUTSZ) {                                     \
-       core(r, ctx->a, b);                                             \
-       CHACHA_STEP(ctx->a);                                            \
-       SALSA20_MIXFULL(b, d, s);                                       \
-       sz -= CHACHA_OUTSZ;                                             \
+      SALSA20_OUTBUF(ctx, d, s, plan.head);                            \
+    }                                                                  \
+                                                                       \
+    ctx->off -= plan.from_rsvr;                                                \
+                                                                       \
+    if (!d) {                                                          \
+      if (plan.from_input) {                                           \
+       pos = chacha_tellu64(ctx);                                      \
+       ASSIGN64(delta, plan.from_input/SALSA20_OUTSZ);                 \
+       ADD64(pos, pos, delta);                                         \
+       chacha_seeku64(ctx, pos);                                       \
       }                                                                        \
+    } else if (!s) while (plan.from_input) {                           \
+      core(r, ctx->a, b); CHACHA_STEP(ctx->a);                         \
+      SALSA20_GENFULL(b, d); plan.from_input -= CHACHA_OUTSZ;          \
+    } else while (plan.from_input) {                                   \
+      core(r, ctx->a, b); CHACHA_STEP(ctx->a);                         \
+      SALSA20_MIXFULL(b, d, s); plan.from_input -= CHACHA_OUTSZ;       \
     }                                                                  \
                                                                        \
-    if (sz) {                                                          \
-      core(r, ctx->a, b);                                              \
-      CHACHA_STEP(ctx->a);                                             \
+    if (plan.tail) {                                                   \
+      core(r, ctx->a, b); CHACHA_STEP(ctx->a);                         \
       SALSA20_PREPBUF(ctx, b);                                         \
-      SALSA20_OUTBUF(ctx, d, s, sz);                                   \
-      assert(!sz);                                                     \
+      SALSA20_OUTBUF(ctx, d, s, plan.tail);                            \
     }                                                                  \
   }
 CHACHA_VARS(DEFENCRYPT)
@@ -370,8 +406,6 @@ CHACHA_VARS(DEFHCHACHA)
   void XCHACHA_INIT(r, XCHACHA_CTX(r) *ctx,                            \
                    const void *key, size_t ksz, const void *nonce)     \
   {                                                                    \
-    static const octet zerononce[XCHACHA_NONCESZ];                     \
-                                                                       \
     populate(ctx->k, key, ksz);                                                \
     ctx->s.a[ 0] = CHACHA_A256;                                                \
     ctx->s.a[ 1] = CHACHA_B256;                                                \
@@ -409,7 +443,7 @@ CHACHA_VARS(DEFXINIT)
   }
 CHACHA_VARS(DEFXNONCE)
 
-/* --- @xchacha{20,12,8}_seek@, @xchacha{20,12,8}_seeku64@ --- *
+/* --- @xchacha{20,12,8}_seek{,u64}@ --- *
  *
  * Arguments:  @xchachaR_ctx *ctx@ = pointer to context
  *             @unsigned long i@, @kludge64 i@ = new position to set
@@ -426,7 +460,7 @@ CHACHA_VARS(DEFXNONCE)
  *             different.
  */
 
-/* --- @xchacha{20,12,8}_tell@, @xchacha{20,12,8}_tellu64@ --- *
+/* --- @xchacha{20,12,8}_tell{,u64}@ --- *
  *
  * Arguments:  @chacha_ctx *ctx@ = pointer to context
  *
@@ -438,7 +472,7 @@ CHACHA_VARS(DEFXNONCE)
  *             different.
  */
 
-/* --- @xchacha{,12,8}_encrypt@ --- *
+/* --- @xchacha{20,12,8}_encrypt@ --- *
  *
  * Arguments:  @xchachaR_ctx *ctx@ = pointer to context
  *             @const void *src@ = source buffer (or null)
@@ -476,20 +510,29 @@ typedef struct gctx { gcipher c; chacha_ctx ctx; } gctx;
 static void gsetiv(gcipher *c, const void *iv)
   { gctx *g = (gctx *)c; chacha_setnonce(&g->ctx, iv); }
 
+static void gsetiv_ietf(gcipher *c, const void *iv)
+  { gctx *g = (gctx *)c; chacha_setnonce_ietf(&g->ctx, iv); }
+
 static void gdestroy(gcipher *c)
   { gctx *g = (gctx *)c; BURN(*g); S_DESTROY(g); }
 
+static gcipher *ginit(const void *k, size_t sz, const gcipher_ops *ops)
+{
+  gctx *g = S_CREATE(gctx);
+  g->c.ops = ops;
+  chacha_init(&g->ctx, k, sz, 0);
+  return (&g->c);
+}
+
 #define DEFGCIPHER(r)                                                  \
                                                                        \
-  static const gcipher_ops gops_##r;                                   \
+  static const gcipher_ops gops_##r, gops_##r##_ietf;                  \
                                                                        \
   static gcipher *ginit_##r(const void *k, size_t sz)                  \
-  {                                                                    \
-    gctx *g = S_CREATE(gctx);                                          \
-    g->c.ops = &gops_##r;                                              \
-    chacha_init(&g->ctx, k, sz, 0);                                    \
-    return (&g->c);                                                    \
-  }                                                                    \
+    { return (ginit(k, sz, &gops_##r)); }                              \
+                                                                       \
+  static gcipher *ginit_##r##_ietf(const void *k, size_t sz)           \
+    { return (ginit(k, sz, &gops_##r##_ietf)); }                       \
                                                                        \
   static void gencrypt_##r(gcipher *c, const void *s,                  \
                           void *t, size_t sz)                          \
@@ -500,9 +543,19 @@ static void gdestroy(gcipher *c)
     gencrypt_##r, gencrypt_##r, gdestroy, gsetiv, 0                    \
   };                                                                   \
                                                                        \
+  static const gcipher_ops gops_##r##_ietf = {                         \
+    &chacha##r##_ietf,                                                 \
+    gencrypt_##r, gencrypt_##r, gdestroy, gsetiv_ietf, 0               \
+  };                                                                   \
+                                                                       \
   const gccipher chacha##r = {                                         \
     "chacha" #r, chacha_keysz,                                         \
     CHACHA_NONCESZ, ginit_##r                                          \
+  };                                                                   \
+                                                                       \
+  const gccipher chacha##r##_ietf = {                                  \
+    "chacha" #r "-ietf", chacha_keysz,                                 \
+    CHACHA_IETF_NONCESZ, ginit_##r##_ietf                              \
   };
 
 CHACHA_VARS(DEFGCIPHER)
@@ -541,7 +594,7 @@ CHACHA_VARS(DEFGCIPHER)
                                                                        \
   const gccipher xchacha##r = {                                                \
     "xchacha" #r, chacha_keysz,                                                \
-    CHACHA_NONCESZ, gxinit_##r                                         \
+    XCHACHA_NONCESZ, gxinit_##r                                                \
   };
 
 CHACHA_VARS(DEFGXCIPHER)
@@ -674,15 +727,41 @@ typedef struct grctx {
 static void gr_seek(void *r, kludge64 pos)
   { grctx *g = r; chacha_seeku64(&g->ctx, pos); }
 
+static void gr_seek_ietf(void *r, kludge64 pos)
+  { grctx *g = r; chacha_seek_ietf(&g->ctx, LO64(pos)); }
+
 static kludge64 gr_tell(void *r)
   { grctx *g = r; return (chacha_tellu64(&g->ctx)); }
 
+static kludge64 gr_tell_ietf(void *r)
+{
+  grctx *g = r;
+  kludge64 pos;
+
+  SET64(pos, 0, chacha_tell_ietf(&g->ctx));
+  return (pos);
+}
+
 static void gr_setnonce(void *r, const void *n)
   { grctx *g = r; chacha_setnonce(&g->ctx, n); }
 
+static void gr_setnonce_ietf(void *r, const void *n)
+  { grctx *g = r; chacha_setnonce_ietf(&g->ctx, n); }
+
 static void grdestroy(grand *r)
   { grctx *g = (grctx *)r; BURN(*g); S_DESTROY(g); }
 
+static grand *grinit(const void *k, size_t ksz, const void *n,
+                    const grand_ops *ops, const grops *myops)
+{
+    grctx *g = S_CREATE(grctx);
+    g->r.r.ops = ops;
+    g->r.ops = myops;
+    chacha_init(&g->ctx, k, ksz, 0);
+    if (n) myops->setnonce(g, n);
+    return (&g->r.r);
+}
+
 #define DEFGRAND(rr)                                                   \
                                                                        \
   static void gr_generate_##rr(void *r, void *b, size_t sz)            \
@@ -692,20 +771,33 @@ static void grdestroy(grand *r)
     { CHACHA_NONCESZ, gr_seek, gr_tell,                                        \
       gr_setnonce, gr_generate_##rr };                                 \
                                                                        \
+  static const grops grops_##rr##_ietf =                               \
+    { CHACHA_IETF_NONCESZ, gr_seek_ietf, gr_tell_ietf,                 \
+      gr_setnonce_ietf, gr_generate_##rr };                            \
+                                                                       \
   static const grand_ops grops_rand_##rr = {                           \
     "chacha" #rr, GRAND_CRYPTO, 0,                                     \
     grmisc, grdestroy, grword,                                         \
-    grbyte, grword, grand_range, grfill                                        \
+    grbyte, grword, grand_defaultrange, grfill                         \
+  };                                                                   \
+                                                                       \
+  static const grand_ops grops_rand_##rr##_ietf = {                    \
+    "chacha" #rr "-ietf", GRAND_CRYPTO, 0,                             \
+    grmisc, grdestroy, grword,                                         \
+    grbyte, grword, grand_defaultrange, grfill                         \
   };                                                                   \
                                                                        \
   grand *chacha##rr##_rand(const void *k, size_t ksz, const void *n)   \
+    { return (grinit(k, ksz, n, &grops_rand_##rr, &grops_##rr)); }     \
+                                                                       \
+  grand *chacha##rr##_ietf_rand(const void *k, size_t ksz,             \
+                               const void *n)                          \
   {                                                                    \
-    grctx *g = S_CREATE(g);                                            \
-    g->r.r.ops = &grops_rand_##rr;                                     \
-    g->r.ops = &grops_##rr;                                            \
-    chacha_init(&g->ctx, k, ksz, n);                                   \
-    return (&g->r.r);                                                  \
+    return (grinit(k, ksz, n,                                          \
+                  &grops_rand_##rr##_ietf,                             \
+                  &grops_##rr##_ietf));                                \
   }
+
 CHACHA_VARS(DEFGRAND)
 
 #define DEFXGRAND(rr)                                                  \
@@ -737,12 +829,12 @@ CHACHA_VARS(DEFGRAND)
   static const grand_ops grxops_rand_##rr = {                          \
     "xchacha" #rr, GRAND_CRYPTO, 0,                                    \
     grmisc, grxdestroy_##rr, grword,                                   \
-    grbyte, grword, grand_range, grfill                                        \
+    grbyte, grword, grand_defaultrange, grfill                         \
   };                                                                   \
                                                                        \
   grand *xchacha##rr##_rand(const void *k, size_t ksz, const void *n)  \
   {                                                                    \
-    grxctx_##rr *g = S_CREATE(g);                                      \
+    grxctx_##rr *g = S_CREATE(grxctx_##rr);                            \
     g->r.r.ops = &grxops_rand_##rr;                                    \
     g->r.ops = &grxops_##rr;                                           \
     XCHACHA_INIT(rr, &g->ctx, k, ksz, n);                              \
@@ -797,16 +889,28 @@ CHACHA_VARS(DEFXGRAND)
 CHACHA_VARS(DEFVCORE)
 
 #define CHACHA_CTX(r) chacha_ctx
-#define CHACHA_INIT(r, ctx, k, ksz, n) chacha_init(ctx, k, ksz, n)
-#define CHACHA_SEEKU64(r, ctx, i) chacha_seeku64(ctx, i)
-#define XCHACHA_SEEKU64(r, ctx, i) xchacha##r##_seeku64(ctx, i)
+
+#define CHACHA_TESTSETUP(r, ctx, k, ksz, n, nsz, p, psz) do {          \
+  kludge64 pos64;                                                      \
+  chacha_init(ctx, k, ksz, 0);                                         \
+  if (nsz == 8) chacha_setnonce(ctx, n);                               \
+  else if (nsz == 12) chacha_setnonce_ietf(ctx, n);                    \
+  if (psz == 8) { LOAD64_(pos64, p); chacha_seeku64(ctx, pos64); }     \
+  else if (psz == 4) chacha_seek_ietf(ctx, LOAD32(p));                 \
+} while (0)
+
+#define XCHACHA_TESTSETUP(r, ctx, k, ksz, n, nsz, p, psz) do {         \
+  kludge64 pos64;                                                      \
+  XCHACHA_INIT(r, ctx, k, ksz, 0);                                     \
+  if (nsz == 24) XCHACHA_SETNONCE(r, ctx, n);                          \
+  if (psz == 8) { LOAD64_(pos64, p); xchacha##r##_seeku64(ctx, pos64); } \
+} while (0)
 
 #define DEFxVENC(base, BASE, r)                                                \
   static int v_encrypt_##base##_##r(dstr *v)                           \
   {                                                                    \
     BASE##_CTX(r) ctx;                                                 \
     dstr d = DSTR_INIT;                                                        \
-    kludge64 pos;                                                      \
     const octet *p, *p0;                                               \
     octet *q;                                                          \
     size_t sz, sz0, step;                                              \
@@ -822,11 +926,8 @@ CHACHA_VARS(DEFVCORE)
     while (step < sz0 + skip) {                                                \
       step = step ? 3*step + 4 : 1;                                    \
       if (step > sz0 + skip) step = sz0 + skip;                                \
-      BASE##_INIT(r, &ctx, v[0].buf, v[0].len, v[1].buf);              \
-      if (v[2].len) {                                                  \
-       LOAD64_(pos, v[2].buf);                                         \
-       BASE##_SEEKU64(r, &ctx, pos);                                   \
-      }                                                                        \
+      BASE##_TESTSETUP(r, &ctx, v[0].buf, v[0].len,                    \
+                      v[1].buf, v[1].len, v[2].buf, v[2].len);         \
                                                                        \
       for (sz = skip; sz >= step; sz -= step)                          \
        BASE##_ENCRYPT(r, &ctx, 0, 0, step);                            \