Re-jig to create more space for the AES code

Move vtor into main code body, leaving only binary info and embedded block at start

Reduce initial stack size, then increase after AES code to overwrite the workspace

Reduce vector table IRQs (requires pico-sdk pr 2200)

Author: will-v-pi

enc_bootloader/CMakeLists.txt

@@ -40,13 +40,18 @@ if (NOT USE_PRECOMPILED)
     # use stack guards, as AES variables are written near the stack
     target_compile_definitions(enc_bootloader PRIVATE
         PICO_USE_STACK_GUARDS=1
-        PICO_STACK_SIZE=0x200
+        PICO_STACK_SIZE=0x100
         PICO_NO_PROGRAM_INFO=1
         # No heap is used
         PICO_HEAP_SIZE=0
         # These inits are not required
         PICO_RUNTIME_SKIP_INIT_PER_CORE_IRQ_PRIORITIES=1
-        PICO_RUNTIME_SKIP_INIT_BOOTROM_LOCKING_ENABLE=1)
+        PICO_RUNTIME_SKIP_INIT_BOOTROM_LOCKING_ENABLE=1
+        # Don't need any vtor irqs
+        PICO_VECTOR_TABLE_NUM_IRQS=0)
+
+    # print memory usage
+    target_link_options(enc_bootloader PUBLIC -Wl,--print-memory-usage)
 
     pico_minimize_runtime(enc_bootloader)
 

enc_bootloader/enc_bootloader.c

@@ -4,17 +4,11 @@
  * SPDX-License-Identifier: BSD-3-Clause
  */
 
-#ifdef DEBUG_PRINT
-#include <stdio.h>
-#endif
 #include <string.h>
 #include "pico/stdlib.h"
 #include "boot/picobin.h"
 #include "pico/bootrom.h"
 #include "hardware/structs/otp.h"
-#if USE_USB_DPRAM
-#include "hardware/structs/usb_dpram.h"
-#endif
 
 #include "pico/binary_info.h"
 
@@ -65,20 +59,8 @@ static void init_aes() {
     init_lut_map();
 }
 
-#if USE_USB_DPRAM
-uint8_t* workarea = (uint8_t*)USBCTRL_DPRAM_BASE;
-#else
-uint8_t* workarea = (uint8_t*)0x20080200; // AES Code & workspace from 0x20080180 -> 0x20081600
-#endif
-
 int main() {
-#ifdef DEBUG_PRINT
-    stdio_init_all();
-
-    printf("Decrypting the image\n");
-    printf("OTP Valid Keys %x\n", otp_hw->key_valid);
-    printf("Unlocking\n");
-#endif
+    // Unlock OTP page with hardcoded key
     for (int i=0; i<4; i++) {
         uint32_t key_i = ((i*2+1) << 24) | ((i*2+1) << 16) |
                          (i*2 << 8) | i*2;
@@ -101,12 +83,6 @@ int main() {
     memcpy(iv + 8, (void*)&iv2, sizeof(iv2));
     memcpy(iv + 12, (void*)&iv3, sizeof(iv3));
 
-#ifdef DEBUG_PRINT
-    printf("Pre decryption image begins with\n");
-    for (int i=0; i < 4; i++)
-        printf("%08x\n", *(uint32_t*)(data_start_addr + i*4));
-#endif
-
     init_aes();
     // Read key directly from OTP - guarded reads will throw a bus fault if there are any errors
     uint16_t* otp_data = (uint16_t*)OTP_DATA_GUARDED_BASE;
@@ -115,25 +91,20 @@ int main() {
     otp_hw->sw_lock[otp_key_page] = 0xf;
     ctr_crypt_s(iv, (void*)data_start_addr, data_size/16);
 
-#ifdef DEBUG_PRINT
-    printf("Post decryption image begins with\n");
-    for (int i=0; i < 4; i++)
-        printf("%08x\n", *(uint32_t*)(data_start_addr + i*4));
-
-    printf("Chaining into %x, size %x\n", data_start_addr, data_size);
+    // Increase stack limit by 0x100
+    pico_default_asm_volatile(
+            "mrs r0, msplim\n"
+            "subs r0, 0x100\n"
+            "msr msplim, r0"
+            :::"r0");
 
-    stdio_deinit_all();
-#endif
-
-    int rc = rom_chain_image(
-        workarea,
+    // Chain into decrypted image
+    rom_chain_image(
+        (uint8_t*)0x20080200, // AES Code & workspace from 0x20080030 -> 0x20081500
         4 * 1024,
         data_start_addr,
         data_size
     );
 
-#ifdef DEBUG_PRINT
-    stdio_init_all();
-    printf("Shouldn't return from ROM call %d\n", rc);
-#endif
+    __breakpoint();
 }

enc_bootloader/memmap_enc_bootloader.ld

@@ -23,10 +23,10 @@
 
 MEMORY
 {
-    RAM_START(rwx) : ORIGIN = 0x20080000, LENGTH = 0x180
-    SCRATCH_X(rwx) : ORIGIN = 0x20080180, LENGTH = 0xE80
-    SCRATCH_Y(rwx) : ORIGIN = 0x20081000, LENGTH = 0x800
-    RAM(rwx)       : ORIGIN = 0x20081800, LENGTH = 0x800
+    RAM_START(rwx) : ORIGIN = 0x20080000, LENGTH = 0x30
+    SCRATCH_X(rwx) : ORIGIN = 0x20080030, LENGTH = 0xFD0
+    SCRATCH_Y(rwx) : ORIGIN = 0x20081000, LENGTH = 0x700
+    RAM(rwx)       : ORIGIN = 0x20081700, LENGTH = 0x900
 }
 
 ENTRY(_entry_point)
@@ -41,17 +41,17 @@ SECTIONS
 
     .start_text : {
         __logical_binary_start = .;
-        /* Vectors require 512-byte alignment on v8-M when >48 IRQs are used,
-           so we would waste RAM if the vector table were not at the
-           start. */
-        KEEP (*(.vectors))
         KEEP (*(.binary_info_header))
         __binary_info_header_end = .;
         KEEP (*(.embedded_block))
         __embedded_block_end = .;
     } > RAM_START
 
     .text : {
+        /* Vectors require 512-byte alignment on v8-M when >48 IRQs are used,
+           so we would waste RAM if the vector table were not at the
+           start. */
+        KEEP (*(.vectors))
         __reset_start = .;
         KEEP (*(.reset))
         __reset_end = .;