Added Benchmark script. Added ARMASM support.

.github/workflows/test-configs.yml

@@ -70,28 +70,28 @@ jobs:
     with:
       arch: arm
       config-file: ./config/examples/imx-rt1040.config
-      make-args: PKA=1
+      make-args: PKA=1 NO_ARM_ASM=1
 
   imx_rt1050_test_pka:
     uses: ./.github/workflows/test-build-mcux-sdk.yml
     with:
       arch: arm
       config-file: ./config/examples/imx-rt1050.config
-      make-args: PKA=1
+      make-args: PKA=1 NO_ARM_ASM=1
 
   imx_rt1060_test_pka:
     uses: ./.github/workflows/test-build-mcux-sdk.yml
     with:
       arch: arm
       config-file: ./config/examples/imx-rt1060.config
-      make-args: PKA=1
+      make-args: PKA=1 NO_ARM_ASM=1
 
   imx_rt1064_test_pka:
     uses: ./.github/workflows/test-build-mcux-sdk.yml
     with:
       arch: arm
       config-file: ./config/examples/imx-rt1064.config
-      make-args: PKA=1
+      make-args: PKA=1 NO_ARM_ASM=1
 
   kinetis_k64f_test:
     uses: ./.github/workflows/test-build-mcux-sdk.yml

Makefile

@@ -318,6 +318,7 @@ clean:
 	$(Q)rm -f $(MACHINE_OBJ) $(MAIN_TARGET) $(LSCRIPT)
 	$(Q)rm -f $(OBJS)
 	$(Q)rm -f tools/keytools/otp/otp-keystore-gen
+	$(Q)rm -f .stack_usage
 	$(Q)$(MAKE) -C test-app -s clean
 	$(Q)$(MAKE) -C tools/check_config -s clean
 	$(Q)$(MAKE) -C stage1 -s clean
@@ -385,6 +386,13 @@ line-count-nrf52:
 line-count-x86:
 	cloc --force-lang-def cloc_lang_def.txt src/boot_x86_fsp.c src/boot_x86_fsp_payload.c src/boot_x86_fsp_start.S src/image.c src/keystore.c src/libwolfboot.c src/loader.c src/string.c src/update_disk.c src/x86/ahci.c src/x86/ata.c src/x86/common.c src/x86/gpt.c src/x86/hob.c src/pci.c src/x86/tgl_fsp.c hal/x86_fsp_tgl.c hal/x86_uart.c
 
+stack-usage: wolfboot.bin
+	$(Q)echo $(STACK_USAGE) > .stack_usage
+
+image-header-size: wolfboot.bin
+	$(Q)echo $(IMAGE_HEADER_SIZE) > .image_header_size
+
+
 cppcheck:
 	cppcheck -f --enable=warning --enable=portability \
 		--suppress="ctunullpointer" --suppress="nullPointer" \

arch.mk

@@ -188,7 +188,8 @@ ifeq ($(ARCH),ARM)
 ifeq ($(CORTEX_A5),1)
   FPU=-mfpu=vfp4-d16
   CFLAGS+=-mcpu=cortex-a5  -mtune=cortex-a5 -static -z noexecstack
-  LDLAGS+=-mcpu=cortex-a5 -mtune=cortex-a5  -mtune=cortex-a5 -static -z noexecstack  -Ttext 0x300000
+  LDLAGS+=-mcpu=cortex-a5 -mtune=cortex-a5  -mtune=cortex-a5 -static \
+		  -z noexecstack  -Ttext 0x300000
   # Cortex-A uses boot_arm32.o
   OBJS+=src/boot_arm32.o src/boot_arm32_start.o
   ifeq ($(NO_ASM),1)
@@ -198,11 +199,37 @@ ifeq ($(CORTEX_A5),1)
     OBJS+=./lib/wolfssl/wolfcrypt/src/port/arm/armv8-sha256.o
     OBJS+=./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-sha256-asm.o
     OBJS+=./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-sha256-asm_c.o
-    CFLAGS+=-DWOLFSSL_SP_ARM32_ASM -DWOLFSSL_ARMASM -DWOLFSSL_ARMASM_NO_HW_CRYPTO -DWOLFSSL_ARM_ARCH=7 -DWOLFSSL_ARMASM_INLINE -DWOLFSSL_ARMASM_NO_NEON
+    CFLAGS+=-DWOLFSSL_SP_ARM32_ASM -DWOLFSSL_ARMASM -DWOLFSSL_ARMASM_NO_HW_CRYPTO \
+			-DWOLFSSL_ARM_ARCH=7 -DWOLFSSL_ARMASM_INLINE -DWOLFSSL_ARMASM_NO_NEON
   endif
 else
   # All others use boot_arm.o
   OBJS+=src/boot_arm.o
+  ifneq ($(NO_ARM_ASM),1)
+    CORTEXM_ARM_EXTRA_OBJS= \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-aes.o \
+		  ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-chacha.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-sha256.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-sha512.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-aes-asm.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-aes-asm_c.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-sha256-asm.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-sha256-asm_c.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-sha512-asm.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-sha512-asm_c.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-sha3-asm.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-sha3-asm_c.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-chacha-asm.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/armv8-32-chacha-asm_c.o
+
+
+    CORTEXM_ARM_THUMB_EXTRA_OBJS= \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/thumb2-sha256-asm.o \
+          ./lib/wolfssl/wolfcrypt/src/port/arm/thumb2-sha256-asm_c.o
+
+    CORTEXM_ARM_EXTRA_CFLAGS+=-DWOLFSSL_ARMASM -DWOLFSSL_ARMASM_NO_HW_CRYPTO \
+							-DWOLFSSL_ARMASM_INLINE -DWOLFSSL_ARMASM_NO_NEON
+  endif
   ifeq ($(CORTEX_M33),1)
     CFLAGS+=-mcpu=cortex-m33 -DCORTEX_M33
     LDFLAGS+=-mcpu=cortex-m33
@@ -212,25 +239,22 @@ else
       endif
       CFLAGS+=-mcmse
       ifeq ($(WOLFCRYPT_TZ),1)
+        CORTEXM_ARM_EXTRA_OBJS=
+        CORTEXM_ARM_EXTRA_CFLAGS=
         SECURE_OBJS+=./src/wc_callable.o
         SECURE_OBJS+=./lib/wolfssl/wolfcrypt/src/random.o
         CFLAGS+=-DWOLFCRYPT_SECURE_MODE
         SECURE_LDFLAGS+=-Wl,--cmse-implib -Wl,--out-implib=./src/wc_secure_calls.o
       endif
     endif # TZEN=1
-    ifeq ($(NO_ASM),1)
       ifeq ($(SPMATH),1)
         ifeq ($(NO_ASM),1)
           MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_c32.o
         else
           CFLAGS+=-DWOLFSSL_SP_ASM -DWOLFSSL_SP_ARM_CORTEX_M_ASM
           MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_cortexm.o
-        endif
+          CFLAGS+=$(CORTEXM_ARM_EXTRA_CFLAGS) -DWOLFSSL_ARM_ARCH=8
-      endif
+          OBJS+=$(CORTEXM_ARM_EXTRA_OBJS)
-    else
-      ifeq ($(SPMATH),1)
-        CFLAGS+=-DWOLFSSL_SP_ASM -DWOLFSSL_SP_ARM_CORTEX_M_ASM
-        MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_cortexm.o
         endif
       endif
     else
@@ -243,6 +267,8 @@ else
         else
           CFLAGS+=-DWOLFSSL_SP_ASM -DWOLFSSL_SP_ARM_CORTEX_M_ASM
           MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_cortexm.o
+          CFLAGS+=$(CORTEXM_ARM_EXTRA_CFLAGS) -DWOLFSSL_ARM_ARCH=7
+          OBJS+=$(CORTEXM_ARM_EXTRA_OBJS)
         endif
        endif
     else
@@ -255,6 +281,9 @@ else
           else
             CFLAGS+=-DWOLFSSL_SP_ASM -DWOLFSSL_SP_ARM_THUMB_ASM
             MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_armthumb.o
+			# TODO: integrate thumb2-asm
+            #CFLAGS+=$(CORTEXM_ARM_EXTRA_CFLAGS) -DWOLFSSL_ARM_ARCH=6
+            #OBJS+=$(CORTEXM_ARM_THUMB_EXTRA_OBJS)
           endif
         endif
       else
@@ -269,6 +298,8 @@ else
             ifeq ($(SPMATH),1)
               CFLAGS+=-DWOLFSSL_SP_ASM -DWOLFSSL_SP_ARM_CORTEX_M_ASM -DWOLFSSL_SP_NO_UMAAL
               MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_cortexm.o
+              CFLAGS+=$(CORTEXM_ARM_EXTRA_CFLAGS) -DWOLFSSL_ARM_ARCH=7
+              OBJS+=$(CORTEXM_ARM_EXTRA_OBJS)
             endif
           endif
       else
@@ -284,6 +315,8 @@ else
           ifeq ($(SPMATH),1)
             CFLAGS+=-DWOLFSSL_SP_ASM -DWOLFSSL_SP_ARM_CORTEX_M_ASM
             MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_cortexm.o
+            CFLAGS+=$(CORTEXM_ARM_EXTRA_CFLAGS) -DWOLFSSL_ARM_ARCH=7
+            OBJS+=$(CORTEXM_ARM_EXTRA_OBJS)
           endif
         endif
       endif

docs/compile.md

@@ -42,6 +42,28 @@ By default, wolfBoot is compiled for ARM Cortex-M3/4/7. To compile for Cortex-M0
 
 `CORTEX_M0=1`
 
+### Speed vs. size
+
+On a number of targets, algorithm may be optimized automatically to use assembly
+optimizations. To disable assembly optimizations, use `NO_ASM=1`. This option will
+produce smaller code, but will also impact on the boot time.
+
+ARM-specific ARM optimizations affecting hash and symmetric key ciphers can be
+disabled with the option `NO_ARM_ASM=1`. This is useful for example when you want
+to use SP math optimizations for key verification, but exclude SHA2/AES optimizations
+to save some space.
+
+#### Example: ECC256 + SHA256 on STM32H7
+
+Benchmark footprint vs. boot time SHA of 100KB image + signature verification
+
+| Description | Selected options | wolfBoot size (B) | Boot time (s) |
+|-------------|------------------|-------------------|---------------|
+| Full ECC256 assembly optimizations. Fastest. | `SIGN=ECC256` | 21836 | .583 |
+| Optimize ECC only (SP math assembly only)  | `SIGN=ECC256 NO_ARM_ASM=1` | 18624 | .760 |
+| No assembly optimizations (smallest) | `SIGN=ECC256 NO_ASM=1` | 14416 | 3.356 |
+
+
 ### Flash partitions
 
 The file [include/target.h](../include/target.h) is generated according to the configured flash geometry,

include/user_settings.h

@@ -75,8 +75,8 @@ extern int tolower(int c);
 #   define ED25519_SMALL
 #   define NO_ED25519_SIGN
 #   define NO_ED25519_EXPORT
-#   define WOLFSSL_SHA512
 #   define USE_SLOW_SHA512
+#   define WOLFSSL_SHA512
 #endif
 
 /* ED448 and SHA3/SHAKE256 */
@@ -267,6 +267,9 @@ extern int tolower(int c);
     !defined(WOLFCRYPT_SECURE_MODE)
 #       define NO_SHA256
 #   endif
+#ifndef WOLFSSL_SHA512
+#define WOLFSSL_SHA512
+#endif
 #endif
 
 /* If SP math is enabled determine word size */
@@ -499,4 +502,8 @@ extern int tolower(int c);
 
 #endif /* WOLFBOOT_PKCS11_APP */
 
+#ifndef XTOLOWER
+#define XTOLOWER(x) (x)
+#endif
+
 #endif /* !_WOLFBOOT_USER_SETTINGS_H_ */

options.mk

@@ -448,7 +448,7 @@ ifeq ($(SIGN),XMSS)
   ifeq ($(WOLFBOOT_SMALL_STACK),1)
     $(error WOLFBOOT_SMALL_STACK with XMSS not supported)
   else
-    STACK_USAGE=2720
+    STACK_USAGE=9352
   endif
 endif
 

test-app/app_stm32h7.c

@@ -362,6 +362,9 @@ void uart_print(const char *s)
     }
 }
 
+#define FILLER_SIZE (100 * 1024)
+static volatile uint8_t filler_data[FILLER_SIZE] = { 0x01, 0x02, 0x03 };
+
 void main(void)
 {
     uint8_t firmware_version = 0;
@@ -373,6 +376,7 @@ void main(void)
     if (FIRMWARE_A)
         ld3_write(LED_INIT);
 
+    filler_data[FILLER_SIZE - 1] = 0xAA;
     /* LED Indicator of successful UART initialization. SUCCESS = ON, FAIL = OFF */
     if (uart_setup(115200) < 0)
         ld2_write(LED_OFF);

tools/config.mk

@@ -23,6 +23,7 @@ ifeq ($(ARCH),)
   CORTEX_M7?=0
   CORTEX_M3?=0
   NO_ASM?=0
+  NO_ARM_ASM?=0
   EXT_FLASH?=0
   SPI_FLASH?=0
   QSPI_FLASH?=0
@@ -104,5 +105,6 @@ CONFIG_VARS:= ARCH TARGET SIGN HASH MCUXSDK MCUXPRESSO MCUXPRESSO_CPU MCUXPRESSO
 	NXP_CUSTOM_DCD NXP_CUSTOM_DCD_OBJS \
 	FLASH_OTP_KEYSTORE \
 	KEYVAULT_OBJ_SIZE \
-	KEYVAULT_MAX_ITEMS
+	KEYVAULT_MAX_ITEMS \
+	NO_ARM_ASM
 

tools/scripts/benchmark.sh

@@ -0,0 +1,98 @@
+#!/bin/bash
+#
+function run_on_board() {
+    # GPIO2: RST
+    # GPIO3: BOOT (input)
+
+    if ! (st-flash reset &>/dev/null); then
+        echo -n "No data."
+    else
+    sleep 1
+        st-flash --connect-under-reset write factory.bin 0x8000000 &>/dev/null
+        sleep .2
+    echo "2" > /sys/class/gpio/export 2>/dev/null
+    echo "out" > /sys/class/gpio/gpio2/direction
+        echo "1" > /sys/class/gpio/gpio2/value # Release reset
+        echo "0" > /sys/class/gpio/gpio2/value # Keep reset low
+        sleep 1
+    echo -n " | "
+        echo "1" > /sys/class/gpio/gpio2/value # Release reset
+    START=`date +%s.%N`
+    while (test `cat /sys/class/gpio/gpio4/value` -eq 0); do
+       sleep .01
+    done
+    while (test `cat /sys/class/gpio/gpio4/value` -eq 0); do
+       sleep .01
+    done
+    END=`date +%s.%N`
+    echo "scale=3; $END/1 - $START/1 "| bc
+    echo "in" > /sys/class/gpio/gpio2/direction
+    echo "2" >/sys/class/gpio/unexport 2>/dev/null
+    fi
+}
+
+function set_benchmark {
+    NAME=$1
+    shift
+    CONFIG=$@
+    # Name
+    echo -n "| "
+    echo -n $NAME
+    echo -n " | "
+    # Configuration
+    echo -n $CONFIG | tr -d '\n'
+    echo -n " | "
+    make clean &>/dev/null
+    make keysclean &>/dev/null
+    make $@ factory.bin &>/dev/null || make $@ factory.bin
+    make $@ stack-usage &>/dev/null
+    make $@ image-header-size &>/dev/null
+    # Bootloader size
+    echo -n `ls -l wolfboot.bin | cut -d " " -f 5 | tr -d '\n'`
+    echo -n " | "
+    # Stack size
+    cat .stack_usage | tr -d '\n'
+    echo -n " | "
+    # Image header size
+    cat .image_header_size | tr -d '\n'
+    # Boot time
+    run_on_board 2>&1 | tr -d '\n'
+    echo " |"
+}
+
+echo "4" > /sys/class/gpio/export 2>/dev/null
+echo "2" > /sys/class/gpio/unexport 2>/dev/null
+make keytools &>/dev/null
+cp config/examples/stm32h7.config .config
+echo "in" > /sys/class/gpio/gpio4/direction
+# Output benchmark results in a Markdown table
+echo "| Name | Configuration | Bootloader size | Stack size | Image header size | Boot time |"
+echo "|------|---------------|-----------------|------------|-------------------|-----------|"
+
+
+set_benchmark "SHA2 only" SIGN=NONE
+set_benchmark "SHA384 only" SIGN=NONE HASH=SHA384
+set_benchmark "SHA3 only" SIGN=NONE HASH=SHA3
+set_benchmark "SHA2 only,small" SIGN=NONE NO_ASM=1
+set_benchmark "rsa2048" SIGN=RSA2048
+set_benchmark "rsa3072" SIGN=RSA3072
+set_benchmark "rsa4096" SIGN=RSA4096
+set_benchmark "rsa4096 with sha384" SIGN=RSA4096 HASH=SHA384
+set_benchmark "ecdsa256" SIGN=ECC256
+set_benchmark "ecdsa384" SIGN=ECC384
+set_benchmark "ecdsa521" SIGN=ECC521
+set_benchmark "ecdsa256 with small stack" SIGN=ECC384 WOLFBOOT_SMALL_STACK=1
+set_benchmark "ecdsa256 with fast math" SIGN=ECC384 SP_MATH=0
+set_benchmark "ecdsa256, no asm" SIGN=ECC256 NO_ASM=1
+set_benchmark "ecdsa384, no asm" SIGN=ECC384 NO_ASM=1
+set_benchmark "ecdsa521, no asm" SIGN=ECC521 NO_ASM=1
+set_benchmark "ecdsa384 with sha384" SIGN=ECC384 HASH=SHA384
+set_benchmark "ed25519 with sha384, small" SIGN=ED25519 HASH=SHA384 NO_ASM=1
+set_benchmark "ed25519 fast" SIGN=ED25519 NO_ASM=0
+set_benchmark "ed448" SIGN=ED448
+set_benchmark "ML_DSA-44" SIGN=ML_DSA ML_DSA_LEVEL=2 IMAGE_SIGNATURE_SIZE=2420 IMAGE_HEADER_SIZE=8192
+set_benchmark "ML_DSA-65" SIGN=ML_DSA ML_DSA_LEVEL=3 IMAGE_SIGNATURE_SIZE=3309 IMAGE_HEADER_SIZE=8192
+set_benchmark "ML_DSA-87" SIGN=ML_DSA ML_DSA_LEVEL=5 IMAGE_SIGNATURE_SIZE=4627 IMAGE_HEADER_SIZE=12288
+set_benchmark "LMS 1-10-8" SIGN=LMS LMS_LEVELS=1 LMS_HEIGHT=10 LMS_WINTERNITZ=8 IMAGE_HEADER_SIZE=4096 IMAGE_SIGNATURE_SIZE=1456
+set_benchmark "XMSS-SHA2_10_256'" XMSS_PARAMS='XMSS-SHA2_10_256' SIGN=XMSS IMAGE_SIGNATURE_SIZE=2500 IMAGE_HEADER_SIZE=8192
+

tools/test.mk

@@ -986,37 +986,37 @@ test-all: clean
 
 
 test-size-all:
-	make test-size SIGN=NONE LIMIT=4816
+	make test-size SIGN=NONE LIMIT=4816 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=ED25519 LIMIT=11396
+	make test-size SIGN=ED25519 LIMIT=11396 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=ECC256  LIMIT=17936
+	make test-size SIGN=ECC256  LIMIT=17936 NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=ECC256 NO_ASM=1 LIMIT=13480
+	make test-size SIGN=ECC256 NO_ASM=1 LIMIT=13480 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=RSA2048 LIMIT=11212
+	make test-size SIGN=RSA2048 LIMIT=11212  NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=RSA2048 NO_ASM=1 LIMIT=11788
+	make test-size SIGN=RSA2048 NO_ASM=1 LIMIT=11788 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=RSA4096 LIMIT=11500
+	make test-size SIGN=RSA4096 LIMIT=11500 NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=RSA4096 NO_ASM=1 LIMIT=12076
+	make test-size SIGN=RSA4096 NO_ASM=1 LIMIT=12076 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=ECC384 LIMIT=17504
+	make test-size SIGN=ECC384 LIMIT=17504 NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=ECC384 NO_ASM=1 LIMIT=14872
+	make test-size SIGN=ECC384 NO_ASM=1 LIMIT=14872 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=ED448 LIMIT=13408
+	make test-size SIGN=ED448 LIMIT=13408 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=RSA3072 LIMIT=11352
+	make test-size SIGN=RSA3072 LIMIT=11352 NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=RSA3072 NO_ASM=1 LIMIT=11892
+	make test-size SIGN=RSA3072 NO_ASM=1 LIMIT=11892 NO_ARM_ASM=1
 	make keysclean
 	make test-size SIGN=LMS LMS_LEVELS=2 LMS_HEIGHT=5 LMS_WINTERNITZ=8 \
 		WOLFBOOT_SMALL_STACK=0 IMAGE_SIGNATURE_SIZE=2644 \
-		IMAGE_HEADER_SIZE?=5288 LIMIT=7504
+		IMAGE_HEADER_SIZE?=5288 LIMIT=7504 NO_ARM_ASM=1
 	make keysclean
 	make test-size SIGN=XMSS XMSS_PARAMS='XMSS-SHA2_10_256' \
 		IMAGE_SIGNATURE_SIZE=2500 IMAGE_HEADER_SIZE?=4096 \
-		LIMIT=8220
+		LIMIT=8220 NO_ARM_ASM=1
 	make keysclean