Re-import mlkem-native and mldsa-native; drop custom meta headers

This commit re-imports mlkem-native (v1.3.0, 398050c87) and mldsa-native
(c800f268c) and removes the three "shadow" backend meta headers owned by
AWS-LC (ml_kem/mlkem_aarch64_meta.h, ml_dsa/mldsa_aarch64_meta.h, and
ml_dsa/mldsa_x86_64_meta.h). The AArch64 shadows existed to add a runtime
NEON gate with a C fallback; upstream now gates NEON in the vendored
meta.h itself. The x86_64 mldsa shadow trimmed the backend to the
assembly-backed subset. Upstream has since converted the remaining AVX2
C-intrinsic operations to proven assembly, so the full backend is
imported and its meta.h is used as-is.

The commit also removes a previous label-renaming at import time that
was in place to avoid a name clash between mlkem and mldsa labels:
Upstream has since made the labels globally unique (mlkem-native PR#1813
and mldsa-native PR#1310).

Finally, the commit removes a previous file-renaming at import time that
was in place to avoid a name clash between the mlkem and mldsa assembly
files, which are compiled into the same FIPS module. Upstream has since
prefixed the backend assembly filenames per project (mlkem_*_asm.S and
mldsa_*_asm.S), so the basenames are globally unique and the individual
.S files can again be built directly.

The mlkem importer gains -UMLK_CONFIG_NO_{KEYPAIR,ENCAPS,DECAPS}_API in
the assembly unifdef invocation, mirroring what the mldsa importer
already does: v1.3.0 added these terms to the backend guards, and without
resolving them the guards would not collapse, leaving the imported
assembly gated on macros that are no longer in scope once common.h is
replaced by the s2n-bignum header.

This commit also reverts the custom configs to the vendored capability
enum (mldsa: `MLD_SYS_CAP_X86_64_AVX2`), adjusts ml_dsa.c for the
`sig_len`-less upstream sign/verify API, and makes `GITHUB_SERVER_URL`
overridable in both importers.

Signed-off-by: Hanno Becker <beckphan@amazon.co.uk>

Author: hanno-becker

crypto/fipsmodule/CMakeLists.txt

@@ -461,55 +461,31 @@ endif()
 # mlkem-native assembly files can be compiled on Unix platforms for x86_64 and arm64 only.
 if((ARCH STREQUAL "aarch64") AND UNIX)
 
-  # Set the source directory for s2n-bignum assembly files
+  # Set the source directory for mlkem-native assembly files.
   set(MLKEM_NATIVE_DIR "${AWSLC_SOURCE_DIR}/crypto/fipsmodule/ml_kem")
 
-  set(MLKEM_NATIVE_AARCH64_ASM_SOURCES
-
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/intt.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/ntt.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/poly_mulcache_compute_asm.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/poly_reduce_asm.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/poly_tobytes_asm.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/poly_tomont_asm.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/polyvec_basemul_acc_montgomery_cached_asm_k2.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/polyvec_basemul_acc_montgomery_cached_asm_k3.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/polyvec_basemul_acc_montgomery_cached_asm_k4.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/rej_uniform_asm.S
-  )
+  # Every .S file in this directory is imported by importer.sh and must be
+  # compiled; glob so that refreshes which add/remove/rename files don't need a
+  # matching edit here. CONFIGURE_DEPENDS makes CMake re-run when the set of
+  # matching files changes.
+  file(GLOB MLKEM_NATIVE_AARCH64_ASM_SOURCES CONFIGURE_DEPENDS
+    "${MLKEM_NATIVE_DIR}/mlkem/native/aarch64/src/*.S")
 
   list(APPEND BCM_ASM_SOURCES ${MLKEM_NATIVE_AARCH64_ASM_SOURCES})
 
 endif()
 
 if((ARCH STREQUAL "x86_64") AND UNIX AND NOT MY_ASSEMBLER_IS_TOO_OLD_FOR_AVX)
 
-  # Set the source directory for s2n-bignum assembly files
+  # Set the source directory for mlkem-native assembly files.
   set(MLKEM_NATIVE_DIR "${AWSLC_SOURCE_DIR}/crypto/fipsmodule/ml_kem")
 
-  set(MLKEM_NATIVE_X86_64_ASM_SOURCES
-
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/intt.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/ntt.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/mulcache_compute.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/nttfrombytes.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/ntttobytes.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/nttunpack.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/reduce.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/tomont.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/polyvec_basemul_acc_montgomery_cached_asm_k2.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/polyvec_basemul_acc_montgomery_cached_asm_k3.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/polyvec_basemul_acc_montgomery_cached_asm_k4.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/rej_uniform_asm.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/poly_compress_d10.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/poly_compress_d11.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/poly_compress_d4.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/poly_compress_d5.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/poly_decompress_d10.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/poly_decompress_d11.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/poly_decompress_d4.S
-    ${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/poly_decompress_d5.S
-  )
+  # Every .S file in this directory is imported by importer.sh and must be
+  # compiled; glob so that refreshes which add/remove/rename files don't need a
+  # matching edit here. CONFIGURE_DEPENDS makes CMake re-run when the set of
+  # matching files changes.
+  file(GLOB MLKEM_NATIVE_X86_64_ASM_SOURCES CONFIGURE_DEPENDS
+    "${MLKEM_NATIVE_DIR}/mlkem/native/x86_64/src/*.S")
 
   list(APPEND BCM_ASM_SOURCES ${MLKEM_NATIVE_X86_64_ASM_SOURCES})
 
@@ -524,7 +500,7 @@ if(UNIX AND ((ARCH STREQUAL "x86_64" AND NOT MY_ASSEMBLER_IS_TOO_OLD_FOR_AVX) OR
   set(MLDSA_NATIVE_DIR "${AWSLC_SOURCE_DIR}/crypto/fipsmodule/ml_dsa")
 
   # Every .S file in this directory is imported by importer.sh and must be
-  # compiled; glob so that refreshes which add/remove files don't need a
+  # compiled; glob so that refreshes which add/remove/rename files don't need a
   # matching edit here. CONFIGURE_DEPENDS makes CMake re-run when the set of
   # matching files changes.
   if(ARCH STREQUAL "x86_64")
@@ -541,7 +517,6 @@ if(UNIX AND ((ARCH STREQUAL "x86_64" AND NOT MY_ASSEMBLER_IS_TOO_OLD_FOR_AVX) OR
 
 endif()
 
-
 if(FIPS_DELOCATE)
   if(FIPS_SHARED)
     message(FATAL_ERROR "Can't set both delocate and shared mode for FIPS build")

crypto/fipsmodule/ml_dsa/META.yml

@@ -1,5 +1,5 @@
 name: mldsa-native
 source: pq-code-package/mldsa-native.git
-branch: 08d40f9403a9ca80f160118bc63e96bf36627866
-commit: 08d40f9403a9ca80f160118bc63e96bf36627866
-imported-at: 2026-06-12T17:04:44+0000
+branch: c800f268c55dc35e2b928b60630a2f66d97083f8
+commit: c800f268c55dc35e2b928b60630a2f66d97083f8
+imported-at: 2026-08-03T15:28:02+0100

crypto/fipsmodule/ml_dsa/importer.sh

@@ -31,7 +31,7 @@
 # Dependencies:
 # - unifdef
 
-GITHUB_SERVER_URL=https://github.com/
+GITHUB_SERVER_URL=${GITHUB_SERVER_URL:=https://github.com/}
 GITHUB_REPOSITORY=${GITHUB_REPOSITORY:=pq-code-package/mldsa-native.git}
 GITHUB_SHA=${GITHUB_SHA:=main}
 
@@ -78,27 +78,22 @@ mkdir $SRC
 find $TMP/mldsa/src -maxdepth 1 -type f -exec cp {} $SRC \;
 
 # Copy x86_64 backend
-# We import all assembly (.S) files and shared headers/constants from the
-# upstream x86_64 backend. The AVX2 C-intrinsic .c files (rej_uniform,
-# decompose, use_hint, chknorm, polyz_unpack) are excluded — their includes
-# are stripped from the BCM below.
-#
-# The upstream meta.h advertises both assembly and C-intrinsic operations.
-# Rather than modify it, we keep a hand-maintained replacement in
-# ../mldsa_x86_64_meta.h (referenced via MLD_CONFIG_ARITH_BACKEND_FILE) that
-# declares only the assembly-backed subset. Upstream meta.h is not copied.
+# The x86_64 backend is fully assembly-backed: every native operation is
+# implemented by a proven .S kernel (proofs live in the upstream mldsa-native
+# repo). We import the upstream meta.h verbatim along with all assembly (.S)
+# files and shared headers/constants, so no hand-maintained meta.h shadow is
+# needed.
 mkdir -p $SRC/native/x86_64/src
 cp $TMP/mldsa/src/native/api.h $SRC/native
-cp $TMP/mldsa/src/native/x86_64/src/arith_native_x86_64.h $SRC/native/x86_64/src
-cp $TMP/mldsa/src/native/x86_64/src/consts.h $SRC/native/x86_64/src
-cp $TMP/mldsa/src/native/x86_64/src/consts.c $SRC/native/x86_64/src
-# NOTE: all imported .S files must have verified proofs in s2n-bignum.
+cp $TMP/mldsa/src/native/x86_64/meta.h $SRC/native/x86_64
+cp $TMP/mldsa/src/native/x86_64/src/*.h $SRC/native/x86_64/src
+cp $TMP/mldsa/src/native/x86_64/src/*.c $SRC/native/x86_64/src
 cp $TMP/mldsa/src/native/x86_64/src/*.S $SRC/native/x86_64/src
 
 # Copy aarch64 backend
-# Unlike x86_64, the aarch64 backend is 100% assembly — no C-intrinsic .c
-# files. The upstream meta.h is suitable as-is, so we copy it verbatim.
-# All assembly (.S) files have verified proofs in s2n-bignum.
+# Like x86_64, the aarch64 backend is fully assembly-backed (proofs live in the
+# upstream mldsa-native repo). The upstream meta.h is suitable as-is, so we copy
+# it verbatim.
 mkdir -p $SRC/native/aarch64/src
 cp $TMP/mldsa/src/native/aarch64/*.h $SRC/native/aarch64
 cp $TMP/mldsa/src/native/aarch64/src/* $SRC/native/aarch64/src
@@ -141,12 +136,6 @@ cp $TMP/mldsa/mldsa_native.h $SRC
 echo "Fixup include paths"
 sed "${SED_I[@]}" 's/#include "src\/\([^"]*\)"/#include "\1"/' $SRC/mldsa_native_bcm.c
 
-# Drop #include directives for the C-intrinsic .c files we did not import.
-# Only consts.c (shared with the assembly backend) is kept.
-echo "Strip C-intrinsic includes from mldsa_native_bcm.c"
-BCM=$SRC/mldsa_native_bcm.c
-sed "${SED_I[@]}" '/^#include "native\/x86_64\/src\/[^"]*\.c"/{/consts\.c/!d;}' "$BCM"
-
 # ================================================================
 # Fixup assembly backends to use s2n-bignum macros
 # ================================================================
@@ -188,16 +177,6 @@ for file in $SRC/native/aarch64/src/*.S $SRC/native/x86_64/src/*.S; do
     sed "${SED_I[@]}" "s/MLD_ASM_FN_SYMBOL($func_name)/S2N_BN_SYMBOL(mldsa_$func_name):/" "$file"
     sed "${SED_I[@]}" "s/MLD_ASM_FN_SIZE($func_name)/S2N_BN_SIZE_DIRECTIVE(mldsa_$func_name)/" "$file"
   fi
-
-  # Prefix local labels with `Lmldsa_` to avoid collisions with other
-  # backends linked into the same FIPS BCM module (e.g. mlkem-native's
-  # ntt.S also defines `Lntt_layer123_start`). The delocator rejects
-  # duplicate symbol names across the unified BCM input.
-  # Build the rename list from the local label definitions (`^Lfoo:`) in
-  # this file, then s/Lfoo/Lmldsa_foo/g across all occurrences.
-  for label in $(grep -oE '^L[a-z][a-zA-Z0-9_]*:' "$file" | sed 's/:$//' | sort -u); do
-    sed "${SED_I[@]}" "s/\b${label}\b/Lmldsa_${label#L}/g" "$file"
-  done
 done
 
 echo "Remove temporary artifacts ..."