Add support for PowerPC 32-bit SHA256 assembly speedups

arch.mk

@@ -561,6 +561,18 @@ ifeq ($(ARCH),PPC)
   LDFLAGS+=-Wl,--gc-sections
 
   OBJS+=src/boot_ppc_start.o src/boot_ppc.o
+
+  ifeq ($(SPMATH),1)
+    MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_c32.o
+  endif
+
+  ifneq ($(NO_ASM),1)
+    # Use the SHA256 and SP math all assembly accelerations
+    CFLAGS+=-DWOLFSSL_SP_PPC
+    CFLAGS+=-DWOLFSSL_PPC32_ASM -DWOLFSSL_PPC32_ASM_INLINE
+    #CFLAGS+=-DWOLFSSL_PPC32_ASM_SMALL
+    MATH_OBJS+=./lib/wolfssl/wolfcrypt/src/port/ppc32/ppc32-sha256-asm_c.o
+  endif
 endif
 
 ifeq ($(TARGET),kinetis)
@@ -788,12 +800,7 @@ ifeq ($(TARGET),nxp_t1024)
   OBJS+=src/pci.o
   CFLAGS+=-DWOLFBOOT_USE_PCI
   UPDATE_OBJS:=src/update_ram.o
-  ifeq ($(SPMATH),1)
-    MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_c32.o
-  else
-    # Use the SP math all assembly accelerations
-    CFLAGS+=-DWOLFSSL_SP_PPC
-  endif
+
   SPI_TARGET=nxp
   OPTIMIZATION_LEVEL=0 # using default -Os causes issues with alignment
 endif
@@ -811,12 +818,6 @@ ifeq ($(TARGET),nxp_t2080)
   LDFLAGS+=-Wl,--as-needed # remove weak functions not used
   UPDATE_OBJS:=src/update_ram.o
   OBJS+=src/fdt.o
-  ifeq ($(SPMATH),1)
-    MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_c32.o
-  else
-    # Use the SP math all assembly accelerations
-    CFLAGS+=-DWOLFSSL_SP_PPC
-  endif
 endif
 
 ifeq ($(TARGET),nxp_p1021)
@@ -835,12 +836,6 @@ ifeq ($(TARGET),nxp_p1021)
   # Use PPC stdlib for memcpy, etc.
   #CFLAGS+=-DWOLFBOOT_USE_STDLIBC
 
-  ifeq ($(SPMATH),1)
-    MATH_OBJS += ./lib/wolfssl/wolfcrypt/src/sp_c32.o
-  else
-    # Use the SP math all assembly accelerations
-    CFLAGS+=-DWOLFSSL_SP_PPC
-  endif
   SPI_TARGET=nxp
 endif
 

config/examples/nxp-p1021.config

@@ -4,6 +4,7 @@ SIGN?=ECC384
 HASH?=SHA384
 IMAGE_HEADER_SIZE?=512
 DEBUG?=0
+DEBUG_SYMBOLS?=1
 DEBUG_UART?=1
 VTOR?=1
 CORTEX_M0?=0
@@ -16,12 +17,13 @@ ALLOW_DOWNGRADE?=0
 NVM_FLASH_WRITEONCE?=0
 WOLFBOOT_VERSION?=0
 NO_MPU?=0
-SPMATH?=0
-SPMATHALL?=1
+SPMATH?=1
+SPMATHALL?=0
 RAM_CODE?=0
 DUALBANK_SWAP?=0
 WOLFTPM?=0
 ELF?=1
+DEBUG_ELF?=0
 
 # Flash Sector (Block) Size (16KB)
 WOLFBOOT_SECTOR_SIZE=0x4000

docs/Targets.md

@@ -1990,20 +1990,23 @@ make DEBUG=1 wolfboot.bin
 # OR
 make wolfboot.bin
 
-# Sign application
+# Build test app
+make test-app/image.bin
+
+# Sign the ELF32 application
 # 1=version (can be any 32-bit value)
-./tools/keytools/sign \
+IMAGE_HEADER_SIZE=512 ./tools/keytools/sign \
     --ecc384 \
     --sha384 \
-    test-app/image.bin \
+    test-app/image.elf \
     wolfboot_signing_private_key.der \
     1
 
 ./tools/bin-assemble/bin-assemble \
   factory.bin \
     0x0        hal/nxp_p1021_stage1.bin \
     0x8000     wolfboot.bin \
-    0x200000   test-app/image.bin \
+    0x200000   test-app/image_v1_signed.bin \
     0x01F00000 fsl_qe_ucode_1021_10_A.bin
 ```
 

hal/nxp_p1021.c

@@ -514,6 +514,48 @@ static void udelay(uint32_t delay_us)
     wait_ticks(delay_us * DELAY_US);
 }
 
+#if 0 /* useful timer code */
+
+uint64_t hal_timer_ms(void)
+{
+    uint64_t val;
+    /* time base is updated every 8 CCB clocks */
+    uint64_t cntfrq = hal_get_bus_clk() / 8;
+    uint64_t cntpct = get_ticks();
+    val = (cntpct * 1000ULL) / cntfrq;
+    return val;
+}
+
+/* example usage */
+//uint64_t start = hal_get_tick_count();
+// do some work
+//wolfBoot_printf("done (%lu ms)\n", (uint32_t)hal_elapsed_time_ms(start));
+
+/* Calculate elapsed time in milliseconds, handling timer overflow properly */
+uint64_t hal_elapsed_time_ms(uint64_t start_ticks)
+{
+    uint64_t current_ticks, elapsed_ticks;
+    uint64_t cntfrq = hal_get_bus_clk() / 8;
+
+    current_ticks = get_ticks();
+
+    /* Handle timer overflow using unsigned arithmetic
+     * This works correctly even if the timer has rolled over,
+     * as long as the elapsed time is less than the full timer range
+     */
+    elapsed_ticks = current_ticks - start_ticks;
+
+    /* Convert elapsed ticks to milliseconds */
+    return (elapsed_ticks * 1000ULL) / cntfrq;
+}
+
+/* Get current tick count for use with hal_elapsed_time_ms() */
+uint64_t hal_get_tick_count(void)
+{
+    return get_ticks();
+}
+#endif
+
 /* ---- eSPI Driver ---- */
 #ifdef ENABLE_ESPI
 void hal_espi_init(uint32_t cs, uint32_t clock_hz, uint32_t mode)

hal/nxp_ppc.h

@@ -498,6 +498,7 @@
 /* Hardware Implementation-Dependent Registers */
 #define SPRN_HID0   0x3F0
 #define HID0_TBEN   (1 << 14) /* Time base enable */
+#define HID0_TBCLK  (1 << 13) /* select clock: 0=every 8 ccb clocks, 1=rising edge of RTC */
 #define HID0_ENMAS7 (1 << 7)  /* Enable hot-wire update of MAS7 register */
 #define HID0_EMCP   (1 << 31) /* Enable machine check pin */
 

include/elf.h

@@ -166,7 +166,7 @@ typedef struct elf64_program_header {
 #define GET_E32(name) (is_elf32 ? GET32(e32->name) : GET32(e64->name))
 
 typedef int (*elf_mmu_map_cb)(uint64_t, uint64_t, uint32_t);
-int elf_load_image_mmu(uint8_t *image, uintptr_t *entry, elf_mmu_map_cb mmu_cb);
+int elf_load_image_mmu(uint8_t *image, uintptr_t *pentry, elf_mmu_map_cb mmu_cb);
 int elf_load_image(uint8_t *image, uintptr_t *entry, int is_ext);
 int64_t elf_hdr_pht_combined_size(const unsigned char* ehdr);
 int elf_open(const unsigned char *ehdr, int *is_elf32);

src/elf.c

@@ -54,7 +54,7 @@ static int check_scatter_format(const unsigned char* ehdr, int is_elf32);
 /* Loader for elf32 or elf64 format program headers
  * Returns the entry point function
  */
-int elf_load_image_mmu(uint8_t *image, uintptr_t *entry, elf_mmu_map_cb mmu_cb)
+int elf_load_image_mmu(uint8_t *image, uintptr_t *pentry, elf_mmu_map_cb mmu_cb)
 {
     elf32_header* h32 = (elf32_header*)image;
     elf64_header* h64 = (elf64_header*)image;
@@ -86,6 +86,9 @@ int elf_load_image_mmu(uint8_t *image, uintptr_t *entry, elf_mmu_map_cb mmu_cb)
         is_elf32 ? 32 : 64, is_le ? "little" : "big");
 #endif
 
+    /* set entry point */
+    *pentry = GET_H64(entry);
+
     /* programs */
     entry_off = image + GET_H32(ph_offset);
     entry_size = GET_H16(ph_entry_size);
@@ -133,20 +136,29 @@ int elf_load_image_mmu(uint8_t *image, uintptr_t *entry, elf_mmu_map_cb mmu_cb)
             }
         }
 
-        memcpy((void*)(uintptr_t)vaddr, image + offset, file_size);
-        if (mem_size > file_size) {
-            memset((void*)(uintptr_t)(vaddr + file_size), 0,
-                   mem_size - file_size);
+        /* confirm the entry won't clobber any of the headers */
+        if ((uint8_t*)vaddr + file_size < image ||
+            (uint8_t*)vaddr > (entry_off + entry_count * entry_size))
+        {
+            memcpy((void*)vaddr, image + offset, file_size);
+            if (mem_size > file_size) {
+                memset((void*)(uintptr_t)(vaddr + file_size), 0,
+                    mem_size - file_size);
+            }
+        #ifdef ARCH_PPC
+            flush_cache(paddr, mem_size);
+        #endif
+        }
+    #ifdef DEBUG_ELF
+        else {
+            wolfBoot_printf("Section would collide with headers! Skipping\n");
         }
-    #ifdef ARCH_PPC
-        flush_cache(paddr, mem_size);
     #endif
-#endif
+#endif /* !ELF_PARSER */
     }
 
-    *entry = GET_H64(entry);
 #ifdef DEBUG_ELF
-    wolfBoot_printf("Entry point %p\r\n", (void*)*entry);
+    wolfBoot_printf("Entry point %p\r\n", (void*)*pentry);
 #endif
 
     return 0;

tools/keytools/user_settings.h

@@ -149,7 +149,6 @@
 #define NO_DES3
 #define NO_PWDBASED
 #define NO_WRITEV
-#define NO_FILESYSTEM
 #define NO_OLD_RNGNAME
 #define NO_WOLFSSL_DIR
 #define WOLFSSL_NO_SOCK

tools/test.mk

@@ -961,40 +961,40 @@ test-all: clean
 
 
 test-size-all:
-	make test-size SIGN=NONE LIMIT=4860 NO_ARM_ASM=1
+	make test-size SIGN=NONE LIMIT=4824 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=ED25519 LIMIT=11448 NO_ARM_ASM=1
+	make test-size SIGN=ED25519 LIMIT=11428 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=ECC256  LIMIT=17988 NO_ARM_ASM=1
+	make test-size SIGN=ECC256  LIMIT=17968 NO_ARM_ASM=1
 	make clean
 	make test-size SIGN=ECC256 NO_ASM=1 LIMIT=13536 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=RSA2048 LIMIT=11272 NO_ARM_ASM=1
+	make test-size SIGN=RSA2048 LIMIT=11264 NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=RSA2048 NO_ASM=1 LIMIT=11840 NO_ARM_ASM=1
+	make test-size SIGN=RSA2048 NO_ASM=1 LIMIT=11844 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=RSA4096 LIMIT=11556 NO_ARM_ASM=1
+	make test-size SIGN=RSA4096 LIMIT=11552 NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=RSA4096 NO_ASM=1 LIMIT=12128 NO_ARM_ASM=1
+	make test-size SIGN=RSA4096 NO_ASM=1 LIMIT=12132 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=ECC384 LIMIT=17556 NO_ARM_ASM=1
+	make test-size SIGN=ECC384 LIMIT=17464 NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=ECC384 NO_ASM=1 LIMIT=14924 NO_ARM_ASM=1
+	make test-size SIGN=ECC384 NO_ASM=1 LIMIT=14912 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=ED448 LIMIT=13464 NO_ARM_ASM=1
+	make test-size SIGN=ED448 LIMIT=13416 NO_ARM_ASM=1
 	make keysclean
-	make test-size SIGN=RSA3072 LIMIT=11408 NO_ARM_ASM=1
+	make test-size SIGN=RSA3072 LIMIT=11404 NO_ARM_ASM=1
 	make clean
-	make test-size SIGN=RSA3072 NO_ASM=1 LIMIT=11944 NO_ARM_ASM=1
+	make test-size SIGN=RSA3072 NO_ASM=1 LIMIT=11948 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=7536 NO_ARM_ASM=1
+		IMAGE_HEADER_SIZE?=5288 LIMIT=7516 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=8292 NO_ARM_ASM=1
 	make keysclean
 	make clean
-	make test-size SIGN=ML_DSA ML_DSA_LEVEL=2 LIMIT=20168 \
+	make test-size SIGN=ML_DSA ML_DSA_LEVEL=2 LIMIT=18866 \
 		 IMAGE_SIGNATURE_SIZE=2420 IMAGE_HEADER_SIZE?=8192