Merge pull request pq-code-package#623 from pq-code-package/nonpublic…

…_symbols

Add marker for non-public API

.clang-format

@@ -21,3 +21,5 @@ Macros:
  # Make this artifically long to avoid function bodies after short contracts
  - __contract__(x)={ void a; void b; void c; void d; void e; void f; } void abcdefghijklmnopqrstuvw()
  - __loop__(x)={}
+ # Make this artifically long to force line break
+ - MLKEM_NATIVE_INTERNAL_API=void abcdefghijklmnopqrstuvwabcdefghijklmnopqrstuvwabcdefg();

examples/monolithic_build/Makefile

@@ -52,6 +52,8 @@ CFLAGS := \
 	-std=c90 \
 	-pedantic \
 	-MMD
+# Set this flag to give all non-global functions internal linkage
+CFLAGS += -DMLKEM_NATIVE_MONOBUILD
 
 BINARY_NAME_FULL=$(BUILD_DIR)/$(BIN)
 

examples/monolithic_build/mlkem_native_monobuild.c

@@ -234,6 +234,16 @@
 #undef MLKEM_NATIVE_FIPS202_BACKEND_NAME
 #endif
 
+/* mlkem/common.h */
+#if defined(MLKEM_NATIVE_INTERNAL_API)
+#undef MLKEM_NATIVE_INTERNAL_API
+#endif
+
+/* mlkem/common.h */
+#if defined(MLKEM_NATIVE_INTERNAL_API)
+#undef MLKEM_NATIVE_INTERNAL_API
+#endif
+
 /* mlkem/common.h */
 #if defined(MLKEM_ASM_NAMESPACE)
 #undef MLKEM_ASM_NAMESPACE

mlkem/cbd.c

@@ -131,6 +131,7 @@ static void cbd3(poly *r, const uint8_t buf[3 * MLKEM_N / 4])
 }
 #endif /* MLKEM_ETA1 == 3 */
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_cbd_eta1(poly *r, const uint8_t buf[MLKEM_ETA1 * MLKEM_N / 4])
 {
 #if MLKEM_ETA1 == 2
@@ -142,6 +143,7 @@ void poly_cbd_eta1(poly *r, const uint8_t buf[MLKEM_ETA1 * MLKEM_N / 4])
 #endif
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_cbd_eta2(poly *r, const uint8_t buf[MLKEM_ETA2 * MLKEM_N / 4])
 {
 #if MLKEM_ETA2 == 2

mlkem/cbd.h

@@ -20,6 +20,7 @@
  * Arguments:   - poly *r: pointer to output polynomial
  *              - const uint8_t *buf: pointer to input byte array
  **************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void poly_cbd_eta1(poly *r, const uint8_t buf[MLKEM_ETA1 * MLKEM_N / 4])
 __contract__(
   requires(memory_no_alias(r, sizeof(poly)))
@@ -39,6 +40,7 @@ __contract__(
  * Arguments:   - poly *r: pointer to output polynomial
  *              - const uint8_t *buf: pointer to input byte array
  **************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void poly_cbd_eta2(poly *r, const uint8_t buf[MLKEM_ETA2 * MLKEM_N / 4])
 __contract__(
   requires(memory_no_alias(r, sizeof(poly)))

mlkem/common.h

@@ -32,6 +32,14 @@
 #define MLKEM_NATIVE_FIPS202_BACKEND_NAME C
 #endif
 
+/* For a monobuild (where all compilation units are merged into one), mark
+ * all non-public API as static since they don't need external linkage. */
+#if !defined(MLKEM_NATIVE_MONOBUILD)
+#define MLKEM_NATIVE_INTERNAL_API
+#else
+#define MLKEM_NATIVE_INTERNAL_API static
+#endif
+
 /* On Apple platforms, we need to emit leading underscore
  * in front of assembly symbols. We thus introducee a separate
  * namespace wrapper for ASM symbols. */

mlkem/indcpa.c

@@ -279,6 +279,7 @@ __contract__(
 #endif /* MLKEM_USE_NATIVE_NTT_CUSTOM_ORDER */
 
 /* Not static for benchmarking */
+MLKEM_NATIVE_INTERNAL_API
 void gen_matrix(polyvec *a, const uint8_t seed[MLKEM_SYMBYTES], int transposed)
 {
   int i;
@@ -409,6 +410,7 @@ __contract__(
 
 STATIC_ASSERT(NTT_BOUND + MLKEM_Q < INT16_MAX, indcpa_enc_bound_0)
 
+MLKEM_NATIVE_INTERNAL_API
 void indcpa_keypair_derand(uint8_t pk[MLKEM_INDCPA_PUBLICKEYBYTES],
                            uint8_t sk[MLKEM_INDCPA_SECRETKEYBYTES],
                            const uint8_t coins[MLKEM_SYMBYTES])
@@ -472,6 +474,7 @@ STATIC_ASSERT(INVNTT_BOUND + MLKEM_ETA1 < INT16_MAX, indcpa_enc_bound_0)
 STATIC_ASSERT(INVNTT_BOUND + MLKEM_ETA2 + MLKEM_Q < INT16_MAX,
               indcpa_enc_bound_1)
 
+MLKEM_NATIVE_INTERNAL_API
 void indcpa_enc(uint8_t c[MLKEM_INDCPA_BYTES],
                 const uint8_t m[MLKEM_INDCPA_MSGBYTES],
                 const uint8_t pk[MLKEM_INDCPA_PUBLICKEYBYTES],
@@ -531,6 +534,7 @@ void indcpa_enc(uint8_t c[MLKEM_INDCPA_BYTES],
 /* Check that the arithmetic in indcpa_dec() does not overflow */
 STATIC_ASSERT(INVNTT_BOUND + MLKEM_Q < INT16_MAX, indcpa_dec_bound_0)
 
+MLKEM_NATIVE_INTERNAL_API
 void indcpa_dec(uint8_t m[MLKEM_INDCPA_MSGBYTES],
                 const uint8_t c[MLKEM_INDCPA_BYTES],
                 const uint8_t sk[MLKEM_INDCPA_SECRETKEYBYTES])

mlkem/indcpa.h

@@ -23,6 +23,7 @@
  *              - const uint8_t *seed: pointer to input seed
  *              - int transposed: boolean deciding whether A or A^T is generated
  **************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void gen_matrix(polyvec *a, const uint8_t seed[MLKEM_SYMBYTES], int transposed)
 __contract__(
   requires(memory_no_alias(a, sizeof(polyvec) * MLKEM_K))
@@ -47,6 +48,7 @@ __contract__(
  *              - const uint8_t *coins: pointer to input randomness
  *                             (of length MLKEM_SYMBYTES bytes)
  **************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void indcpa_keypair_derand(uint8_t pk[MLKEM_INDCPA_PUBLICKEYBYTES],
                            uint8_t sk[MLKEM_INDCPA_SECRETKEYBYTES],
                            const uint8_t coins[MLKEM_SYMBYTES])
@@ -74,6 +76,7 @@ __contract__(
  *              - const uint8_t *coins: pointer to input random coins used as
  *seed (of length MLKEM_SYMBYTES) to deterministically generate all randomness
  **************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void indcpa_enc(uint8_t c[MLKEM_INDCPA_BYTES],
                 const uint8_t m[MLKEM_INDCPA_MSGBYTES],
                 const uint8_t pk[MLKEM_INDCPA_PUBLICKEYBYTES],
@@ -100,6 +103,7 @@ __contract__(
  *              - const uint8_t *sk: pointer to input secret key
  *                                   (of length MLKEM_INDCPA_SECRETKEYBYTES)
  **************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void indcpa_dec(uint8_t m[MLKEM_INDCPA_MSGBYTES],
                 const uint8_t c[MLKEM_INDCPA_BYTES],
                 const uint8_t sk[MLKEM_INDCPA_SECRETKEYBYTES])

mlkem/ntt.c

@@ -127,6 +127,7 @@ __contract__(
  * the proof may need strengthening.
  */
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_ntt(poly *p)
 {
   int len, layer;
@@ -150,6 +151,7 @@ void poly_ntt(poly *p)
 /* Check that bound for native NTT implies contractual bound */
 STATIC_ASSERT(NTT_BOUND_NATIVE <= NTT_BOUND, invntt_bound)
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_ntt(poly *p)
 {
   POLY_BOUND_MSG(p, MLKEM_Q, "native ntt input");
@@ -201,6 +203,7 @@ __contract__(
   }
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_invntt_tomont(poly *p)
 {
   /*
@@ -236,13 +239,15 @@ void poly_invntt_tomont(poly *p)
 /* Check that bound for native invNTT implies contractual bound */
 STATIC_ASSERT(INVNTT_BOUND_NATIVE <= INVNTT_BOUND, invntt_bound)
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_invntt_tomont(poly *p)
 {
   intt_native(p);
   POLY_BOUND_MSG(p, INVNTT_BOUND_NATIVE, "native intt output");
 }
 #endif /* MLKEM_USE_NATIVE_INTT */
 
+MLKEM_NATIVE_INTERNAL_API
 void basemul_cached(int16_t r[2], const int16_t a[2], const int16_t b[2],
                     int16_t b_cached)
 {

mlkem/ntt.h

@@ -32,6 +32,7 @@ extern const int16_t zetas[128];
  *
  * Arguments:   - poly *p: pointer to in/output polynomial
  **************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void poly_ntt(poly *r)
 __contract__(
   requires(memory_no_alias(r, sizeof(poly)))
@@ -57,6 +58,7 @@ __contract__(
  *
  * Arguments:   - uint16_t *a: pointer to in/output polynomial
  **************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void poly_invntt_tomont(poly *r)
 __contract__(
   requires(memory_no_alias(r, sizeof(poly)))
@@ -85,6 +87,7 @@ __contract__(
  *            - b_cached: Some precomputed value, typically derived from
  *                   b1 and a twiddle factor. Can be an arbitary int16_t.
  ************************************************************/
+MLKEM_NATIVE_INTERNAL_API
 void basemul_cached(int16_t r[2], const int16_t a[2], const int16_t b[2],
                     int16_t b_cached)
 __contract__(

mlkem/poly.c

@@ -16,6 +16,7 @@
 #include "symmetric.h"
 #include "verify.h"
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_compress_du(uint8_t r[MLKEM_POLYCOMPRESSEDBYTES_DU], const poly *a)
 {
   int j;
@@ -80,6 +81,7 @@ void poly_compress_du(uint8_t r[MLKEM_POLYCOMPRESSEDBYTES_DU], const poly *a)
 }
 
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_decompress_du(poly *r, const uint8_t a[MLKEM_POLYCOMPRESSEDBYTES_DU])
 {
   int j;
@@ -139,6 +141,7 @@ void poly_decompress_du(poly *r, const uint8_t a[MLKEM_POLYCOMPRESSEDBYTES_DU])
 #endif
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_compress_dv(uint8_t r[MLKEM_POLYCOMPRESSEDBYTES_DV], const poly *a)
 {
   int i;
@@ -193,6 +196,7 @@ void poly_compress_dv(uint8_t r[MLKEM_POLYCOMPRESSEDBYTES_DV], const poly *a)
 #endif
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_decompress_dv(poly *r, const uint8_t a[MLKEM_POLYCOMPRESSEDBYTES_DV])
 {
   int i;
@@ -250,6 +254,7 @@ void poly_decompress_dv(poly *r, const uint8_t a[MLKEM_POLYCOMPRESSEDBYTES_DV])
 }
 
 #if !defined(MLKEM_USE_NATIVE_POLY_TOBYTES)
+MLKEM_NATIVE_INTERNAL_API
 void poly_tobytes(uint8_t r[MLKEM_POLYBYTES], const poly *a)
 {
   unsigned int i;
@@ -282,6 +287,7 @@ void poly_tobytes(uint8_t r[MLKEM_POLYBYTES], const poly *a)
   }
 }
 #else  /* MLKEM_USE_NATIVE_POLY_TOBYTES */
+MLKEM_NATIVE_INTERNAL_API
 void poly_tobytes(uint8_t r[MLKEM_POLYBYTES], const poly *a)
 {
   POLY_UBOUND(a, MLKEM_Q);
@@ -290,6 +296,7 @@ void poly_tobytes(uint8_t r[MLKEM_POLYBYTES], const poly *a)
 #endif /* MLKEM_USE_NATIVE_POLY_TOBYTES */
 
 #if !defined(MLKEM_USE_NATIVE_POLY_FROMBYTES)
+MLKEM_NATIVE_INTERNAL_API
 void poly_frombytes(poly *r, const uint8_t a[MLKEM_POLYBYTES])
 {
   int i;
@@ -309,12 +316,14 @@ void poly_frombytes(poly *r, const uint8_t a[MLKEM_POLYBYTES])
   POLY_UBOUND(r, 4096);
 }
 #else  /* MLKEM_USE_NATIVE_POLY_FROMBYTES */
+MLKEM_NATIVE_INTERNAL_API
 void poly_frombytes(poly *r, const uint8_t a[MLKEM_POLYBYTES])
 {
   poly_frombytes_native(r, a);
 }
 #endif /* MLKEM_USE_NATIVE_POLY_FROMBYTES */
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_frommsg(poly *r, const uint8_t msg[MLKEM_INDCPA_MSGBYTES])
 {
   int i;
@@ -341,6 +350,7 @@ void poly_frommsg(poly *r, const uint8_t msg[MLKEM_INDCPA_MSGBYTES])
   POLY_BOUND_MSG(r, MLKEM_Q, "poly_frommsg output");
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_tomsg(uint8_t msg[MLKEM_INDCPA_MSGBYTES], const poly *a)
 {
   int i;
@@ -361,6 +371,7 @@ void poly_tomsg(uint8_t msg[MLKEM_INDCPA_MSGBYTES], const poly *a)
   }
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_getnoise_eta1_4x(poly *r0, poly *r1, poly *r2, poly *r3,
                            const uint8_t seed[MLKEM_SYMBYTES], uint8_t nonce0,
                            uint8_t nonce1, uint8_t nonce2, uint8_t nonce3)
@@ -388,6 +399,7 @@ void poly_getnoise_eta1_4x(poly *r0, poly *r1, poly *r2, poly *r3,
   POLY_BOUND_MSG(r3, MLKEM_ETA1 + 1, "poly_getnoise_eta1_4x output 3");
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_getnoise_eta2(poly *r, const uint8_t seed[MLKEM_SYMBYTES],
                         uint8_t nonce)
 {
@@ -403,6 +415,7 @@ void poly_getnoise_eta2(poly *r, const uint8_t seed[MLKEM_SYMBYTES],
   POLY_BOUND_MSG(r, MLKEM_ETA1 + 1, "poly_getnoise_eta2 output");
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_getnoise_eta1122_4x(poly *r0, poly *r1, poly *r2, poly *r3,
                               const uint8_t seed[MLKEM_SYMBYTES],
                               uint8_t nonce0, uint8_t nonce1, uint8_t nonce2,
@@ -441,6 +454,7 @@ void poly_getnoise_eta1122_4x(poly *r0, poly *r1, poly *r2, poly *r3,
   POLY_BOUND_MSG(r3, MLKEM_ETA2 + 1, "poly_getnoise_eta1122_4x output 3");
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_basemul_montgomery_cached(poly *r, const poly *a, const poly *b,
                                     const poly_mulcache *b_cache)
 {
@@ -461,6 +475,7 @@ void poly_basemul_montgomery_cached(poly *r, const poly *a, const poly *b,
 }
 
 #if !defined(MLKEM_USE_NATIVE_POLY_TOMONT)
+MLKEM_NATIVE_INTERNAL_API
 void poly_tomont(poly *r)
 {
   int i;
@@ -476,6 +491,7 @@ void poly_tomont(poly *r)
   POLY_BOUND(r, MLKEM_Q);
 }
 #else  /* MLKEM_USE_NATIVE_POLY_TOMONT */
+MLKEM_NATIVE_INTERNAL_API
 void poly_tomont(poly *r)
 {
   poly_tomont_native(r);
@@ -484,6 +500,7 @@ void poly_tomont(poly *r)
 #endif /* MLKEM_USE_NATIVE_POLY_TOMONT */
 
 #if !defined(MLKEM_USE_NATIVE_POLY_REDUCE)
+MLKEM_NATIVE_INTERNAL_API
 void poly_reduce(poly *r)
 {
   int i;
@@ -501,13 +518,15 @@ void poly_reduce(poly *r)
   POLY_UBOUND(r, MLKEM_Q);
 }
 #else  /* MLKEM_USE_NATIVE_POLY_REDUCE */
+MLKEM_NATIVE_INTERNAL_API
 void poly_reduce(poly *r)
 {
   poly_reduce_native(r);
   POLY_UBOUND(r, MLKEM_Q);
 }
 #endif /* MLKEM_USE_NATIVE_POLY_REDUCE */
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_add(poly *r, const poly *b)
 {
   int i;
@@ -521,6 +540,7 @@ void poly_add(poly *r, const poly *b)
   }
 }
 
+MLKEM_NATIVE_INTERNAL_API
 void poly_sub(poly *r, const poly *b)
 {
   int i;
@@ -535,6 +555,7 @@ void poly_sub(poly *r, const poly *b)
 }
 
 #if !defined(MLKEM_USE_NATIVE_POLY_MULCACHE_COMPUTE)
+MLKEM_NATIVE_INTERNAL_API
 void poly_mulcache_compute(poly_mulcache *x, const poly *a)
 {
   int i;
@@ -547,6 +568,7 @@ void poly_mulcache_compute(poly_mulcache *x, const poly *a)
   POLY_BOUND(x, MLKEM_Q);
 }
 #else  /* MLKEM_USE_NATIVE_POLY_MULCACHE_COMPUTE */
+MLKEM_NATIVE_INTERNAL_API
 void poly_mulcache_compute(poly_mulcache *x, const poly *a)
 {
   poly_mulcache_compute_native(x, a);