Present in the firmware package titles (0100000000000819, 010000000000081A, 010000000000081B and 010000000000081C) and installed into eMMC storage's boot partitions 0 and 1, "package1" contains the first Switch bootloader to run under the NVIDIA boot processor (an ARM7TDMI called "BPMP", "BPMP-Lite", "AVP" or "COP").

The boot ROM validates, copies to IRAM and executes this package by parsing it's information block from the BCT.

Bootloader

Split into two main stages, the bootloader is responsible for setting up hardware, generate keys and prepare the main CPU ("CCPLEX").

Stage 0

The code for this stage is stored in plaintext inside the package. By looking into the BCT's bootloader0_info (normal) or bootloader1_info (safe mode), the boot ROM starts executing this stage at address 0x40010020 in IRAM.

Initialization

The stack pointer is set.

// Set the stack pointer
*(u32 *)sp = 0x40008000;

// Jump to main
bootloader_main();

// Infinite loop
deadlock();

Main

The bootloader poisons the exception vectors, cleans up memory (.bss and init_array), sets up hardware devices (including the security engine and fuses), does all the necessary checks, generates keys and finally decrypts and executes the next stage.

// Poison all exception vectors
*(u32 *)0x6000F200 = panic();
*(u32 *)0x6000F204 = panic();
*(u32 *)0x6000F208 = panic();
*(u32 *)0x6000F20C = panic();
*(u32 *)0x6000F210 = panic();
*(u32 *)0x6000F214 = panic();
*(u32 *)0x6000F218 = panic();
*(u32 *)0x6000F21C = panic();

u32 bss_addr_end = bss_addr_start;
u32 bss_offset = 0;
u32 bss_size = bss_addr_end - bss_addr_start;

// Clear .bss region
// Never happens due to bss_size being set to 0
while (bss_offset < bss_size)
{
   *(u32 *)bss_addr_start + bss_offset = 0;
   bss_offset += 0x04;
}

u32 init_array_addr_end = init_array_addr_start;
u32 init_array_offset = init_array_addr_start;

// Call init methods
// Never happens due to init_array_addr_end being set to init_array_addr_start
while (init_array_offset < init_array_addr_end)
{
   u32 init_method_offset = *(u32 *)init_array_offset;

   call_init_method(init_method_offset + init_array_offset);
   init_array_offset += 0x04;
}

// Setup I2S1, I2S2, I2S3, I2S4, DISPLAY and VIC
enable_hw_devices();

// Program the SE clock and resets
// Uses RST_DEVICES_V, CLK_OUT_ENB_V, CLK_SOURCE_SE and CLK_V_SE
enable_se_clkrst();

// Set MISC_CLK_ENB
// This makes fuse registers visible
enable_misc_clk(0x01);

// Read FUSE_SKU_INFO and compare with 0x83
check_sku();

// Check configuration fuses
check_config_fuses();

u32 bct_iram_addr = 0x40000000;

// Check bootloader version from BCT
check_bootloader_ver(bct_iram_addr);

// Check anti-downgrade fuses
check_downgrade();

// Set FUSE_DIS_PGM
// Disables fuse programming until next reboot
disable_fuse_pgm();

// Setup memory controllers
enable_mem_ctl();

// Setup the security engine's address
set_se_addr(0x70012000);

// Check SE global config
check_se_status();

// Generate keys
keygen(bct_iram_addr);

u32 stage1_addr = 0x40013FE0;

// Decrypt next stage
stage1_addr = decrypt_stage1(stage1_addr);

u32 stage1_sp = 0x40007000;

// Set the stack pointer and jump to a stub responsible
// for cleaning up and branching into the actual stage 1
exec_stage1_stub(stage1_addr, stage1_stub_addr, stage1_sp);

return;

Panic

If a panic occurs, all sensitive memory contents are cleared, the security engine and fuse programming are disabled and the boot processor is left in a halted state.

// Clear all stack contents
clear_stack();

// Terminate the security engine
disable_se();

// Set FUSE_DIS_PGM
// Disables fuse programming until next reboot
disable_fuse_pgm();

// Clear temporary key storage memory
clear_mem();

// Clear the stage 1 binary
clear_stage1();

u32 FLOW_CTLR_HALT_COP_EVENTS = 0x60007004;
u32 FLOW_MODE_STOP = 0x40000000;
u32 HALT_COP_EVENT_JTAG = 0x10000000;

// Halt the boot processor
while (true)
   *(u32 *)FLOW_CTLR_HALT_COP_EVENTS = (FLOW_MODE_STOP | HALT_COP_EVENT_JTAG);

Anti-downgrade

See Anti-downgrade.

Memory controllers

After disabling fuse programming, the bootloader configures the EMC and MEM/MC. It additionally disables QSPI resets and programs a special aperture designed for AHB redirected access to IRAM.

u32 PERIPH_CLK_SOURCE_EMC = 0x6000619C;
u32 CLK_OUT_ENB_SET_H = 0x60006328;
u32 CLK_OUT_ENB_SET_X = 0x60006284;
u32 RST_DEVICES_SET_H = 0x60006308;
u32 RST_DEVICES_CLR_Y = 0x600062AC;
u32 MC_IRAM_REG_CTRL = 0x70019964;
u32 MC_IRAM_BOM = 0x7001965C;
u32 MC_IRAM_TOM = 0x70019660;

// Initialize EMC's clock source
u32 emc_clk_src_val = *(u32 *)PERIPH_CLK_SOURCE_EMC;
*(u32 *)PERIPH_CLK_SOURCE_EMC = (emc_clk_src_val | 0x40000000);

// Enable CLK_H_EMC
u32 clk_out_enb_h_val = *(u32 *)CLK_OUT_ENB_SET_H;
clk_out_enb_h_val &= ~(0x2000000);
clk_out_enb_h_val |= 0x2000000;
*(u32 *)CLK_OUT_ENB_SET_H = clk_out_enb_h_val;

// Enable CLK_H_MEM
clk_out_enb_h_val = *(u32 *)CLK_OUT_ENB_SET_H;
clk_out_enb_h_val &= ~(0x01);
clk_out_enb_h_val |= 0x01;
*(u32 *)CLK_OUT_ENB_SET_H = clk_out_enb_h_val;

// Enable CLK_X_EMC_DLL
u32 clk_out_enb_x_val = *(u32 *)CLK_OUT_ENB_SET_X;
clk_out_enb_x_val &= ~(0x4000);
clk_out_enb_x_val |= 0x4000;
*(u32 *)CLK_OUT_ENB_SET_X = clk_out_enb_x_val;

// Enable RST_H_EMC and RST_H_MEM
*(u32 *)RST_DEVICES_SET_H = 0x2000001;

// Wait a while
mdelay(0x05);

// Disable RST_Y_QSPI
u32 rst_clr_y_val = *(u32 *)RST_DEVICES_CLR_Y;
rst_clr_y_val &= ~(0x80000);
rst_clr_y_val |= 0x80000;
*(u32 *)RST_DEVICES_CLR_Y = rst_clr_y_val;

// Refresh MC_IRAM_REG_CTRL
// Should be set to 0 (MC_ENABLE_IRAM_CFG_WRITES)
u32 mc_iram_reg_ctrl_val = *(u32 *)MC_IRAM_REG_CTRL;
*(u32 *)MC_IRAM_REG_CTRL = mc_iram_reg_ctrl_val;

// Set base and top addresses for AHB redirected IRAM path
// This allows devices like the GPU to access this range
*(u32 *)MC_IRAM_BOM = 0x40000000;
*(u32 *)MC_IRAM_TOM = 0x4003F000;

// Read back MC_IRAM_REG_CTRL
mc_iram_reg_ctrl_val = *(u32 *)MC_IRAM_REG_CTRL;
	
return mc_iram_reg_ctrl_val;

Key generation

After the security engine is ready and before decrypting the next stage, the bootloader initializes and generates several keys into hardware keyslots. For more details on the Switch's cryptosystem, please see this page.

Selection

Depending on FUSE_RESERVED_ODM4 and FUSE_SPARE_BIT_5 different static seeds are selected for key generation.

// Initialize keyslots 0x0C and 0x0D as readable
init_keyslot(0x0C, 0x15);
init_keyslot(0x0D, 0x15);

// Find the BCT's data address from IRAM header
u32 bct_data_addr = *(u32 *)bct_imem_addr + 0x4C;
u32 bct_customer_data_addr = *(u32 *)bct_data_addr + 0x450;

// Wrapper to get unit type from FUSE_RESERVED_ODM4
// This tells if the device is retail or debug
bool is_retail = is_unit_retail();

u32 master_static_seed_addr = 0;
u32 master_static_seed_size = 0;

if (is_retail)
{
   // Read FUSE_SPARE_BIT_5
   // This tells which master key to use
   u32 master_key_ver = read_fuse_spare_bit_5();

   // Invalid for retail
   if (!master_key_ver)
      panic();
   else
   {
      master_static_seed_addr = static_seed1_addr;
      master_static_seed_size = 0x10;

      // Generate retail keys
      generate_retail_keys(bct_customer_data_addr, static_seed_addr, static_seed_size);
   }
}
else
{
   // Read FUSE_SPARE_BIT_5
   // This tells which master key to use
   u32 master_key_ver = read_fuse_spare_bit_5();

   // Use debug key set
   if (!master_key_ver)
   {
      // Read the first byte of the BCT RSA PSS signature
      u8 rsa_pss_1_byte = *(u8 *)bct_data_addr + 0x210;

      if (rsa_pss_1_byte == 0x11)
      {
         master_static_seed_addr = static_seed6_addr;
         master_static_seed_size = 0x10;
      }
      else	                          
      {
         master_static_seed_addr = static_seed7_addr;
         master_static_seed_size = 0x10;
      }

      // Generate debug keys
      generate_debug_keys(static_seed_addr, static_seed_size);
   }
   else
   {
      // Read the first byte of the BCT RSA PSS signature
      u8 rsa_pss_1_byte = *(u8 *)bct_data_addr + 0x210;

      if (rsa_pss_1_byte == 0x4F)		// Different key as in retail mode
      {
         master_static_seed_addr = static_seed0_addr;
         master_static_seed_size = 0x10;
      }
      else	                                // Same key as in retail mode
      {
         master_static_seed_addr = static_seed1_addr;
         master_static_seed_size = 0x10;
      }

      // Generate retail keys
      generate_retail_keys(bct_customer_data_addr, master_static_seed_addr, master_static_seed_size);
   }
}

// Initialize keyslots 0x0C and 0x0D as unreadable
init_keyslot(0x0C, 0xFF);
init_keyslot(0x0D, 0xFF);

return;
generate_retail_keys

In order to generate retail keys, the bootloader starts by initializing TSEC and grabbing it's device key. Using static seeds and the SBK, the keyblob injected into the BCT's customer_data is validated and decrypted. The resulting keys will then be used to generate the master static key and the master device key.

See the pseudocode bellow for the detailed process.

u32 in_addr = 0;
u32 in_size = 0;
u32 out_addr = 0;
u32 out_size = 0;
u32 keyslot = 0;
u32 keyslot_dst = 0;

// Get the TSEC device key
tsec_get_device_key(tsec_device_key_addr, 0x10);

// Install the TSEC device key into keyslot 0x0D
set_keyslot_data(0x0D, tsec_device_key_addr, 0x10);

in_addr = static_seed2_addr;
in_size = 0x10;
out_addr = keyblob_device_key_addr;
out_size = 0x10;
keyslot = 0x0D;

// Use the tsec_device_key (keyslot 0x0D) to decrypt the static_seed2
// This generates the keyblob_device_key
aes_ecb_decrypt(out_addr, out_size, keyslot, in_addr, in_size);

in_addr = keyblob_device_key_addr;
in_size = 0x10;
keyslot = 0x0E;
keyslot_dst = 0x0D;

// Use SBK (keyslot 0x0E) to further decrypt the
// keyblob_device_key and install it into keyslot 0x0D
// This will generate the keyblob_key
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

// Clear SBK and SSK keyslots
clear_keyslot(0x0E);
clear_keyslot(0x0F);

in_addr = static_seed4_addr;
in_size = 0x10;
keyslot = 0x0D;
keyslot_dst = 0x0B;

// Use keyblob_key (keyslot 0x0D) to decrypt the
// static_seed4_addr and install it to keyslot 0x0B
// This will generate the bct_mac_key
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

in_addr = bct_customer_data_addr + 0x10;
in_size = 0xA0;
out_addr = mac_addr;
out_size = 0x10;
keyslot = 0x0B;

// Use the bct_mac_key (keyslot 0x0B) to generate
// CMAC over bct_customer_data_addr + 0x10
aes_cmac(out_addr, out_size, keyslot, in_addr, in_size);

// Comapre the generated hash with the first
// 0x10 bytes of bct_customer_data
bool match = safe_memcmp(mac_addr, bct_customer_data_addr, 0x10);

// Hashes don't match
if (!match)
   panic();

in_addr = bct_customer_data_addr + 0x20;
in_size = 0x90;
ctr_addr = bct_customer_data_addr + 0x10;
ctr_size = 0x10;
out_addr = dec_payload_addr;
out_size = 0x90;
keyslot = 0x0D;

// AES-CTR decrypt
// Use the keyblob_key (keyslot 0x0D) to decrypt bct_customer_data_addr + 0x20
// using bct_customer_data_addr + 0x10 as CTR
aes_ctr_decrypt(out_addr, out_size, keyslot, in_addr, in_size, ctr_addr, ctr_size);

// Install the last decrypted keyblob key into keyslot 0x0B
// This is the pk11_key
set_keyslot_data(0x0B, dec_payload_addr + 0x80, 0x10);

// Install the first decrypted keyblob key into keyslot 0x0C
// This is the master_static_kek
set_keyslot_data(0x0C, dec_payload_addr, 0x10);

// Clear out the decrypted data
memclear(dec_payload_addr, 0x90);

in_addr = master_static_seed_addr;
in_size = master_static_seed_size;
keyslot = 0x0C;
keyslot_dst = 0x0C;

// Use the master_static_kek (keyslot 0x0C) to decrypt
// master_static_seed and install it into keyslot 0x0C
// This will generate the master_static_key
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

in_addr = static_seed3_addr;
in_size = 0x10;
keyslot = 0x0D;
keyslot_dst = 0x0D;

// Use keyblob_key (keyslot 0x0D) to decrypt
// static_seed3_addr and install it into keyslot 0x0D
// This will generate the master_device_key
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

return;
generate_debug_keys

In order to generate debug keys, the bootloader only uses static seeds, the SBK and the SSK.

See the pseudocode bellow for the detailed process.

u32 in_addr = 0;
u32 in_size = 0;
u32 keyslot = 0;
u32 keyslot_dst = 0;

in_addr = static_seed8_addr;
in_size = 0x10;
keyslot = 0x0E;
keyslot_dst = 0x0B;

// Use SBK (keyslot 0x0E) to decrypt the
// static_seed8 and install it to keyslot 0x0B
// This will generate debug_pk11_key
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

in_addr = static_seed5_addr;
in_size = 0x10;
keyslot = 0x0E;
keyslot_dst = 0x0C;

// Use SBK (keyslot 0x0E) to decrypt the
// static_seed5 and install it to keyslot 0x0C
// This will generate debug_master_static_kek
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

in_addr = static_seed9_addr;
in_size = 0x10;
keyslot = 0x0F;
keyslot_dst = 0x0D;

// Use SSK (keyslot 0x0F) to decrypt the
// static_seed9 and install it to keyslot 0x0D
// This will generate debug_keyblob_key
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

// Clear SBK and SSK keyslots
clear_keyslot(0x0E);
clear_keyslot(0x0F);

in_addr = master_static_seed_addr;
in_size = master_static_seed_size;
keyslot = 0x0C;
keyslot_dst = 0x0C;

// Use the debug_master_static_kek (keyslot 0x0C) to decrypt the
// master_static_seed and install it to keyslot 0x0C
// This will generate the master_static_key
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

in_addr = static_seed3_addr;
in_size = 0x10;
keyslot = 0x0D;
keyslot_dst = 0x0D;

// Use debug_keyblob_key (keyslot 0x0D) to decrypt the
// static_seed3 and install it to keyslot 0x0D
// This will generate the master_device_key
decrypt_keyslot(keyslot_dst, keyslot, in_addr, in_size);

return;

Stage 1 loading

After generating all the necessary keys, the bootloader proceeds to decrypt and launch the next stage.

Decryption

The encrypted stage 1 binary is stored inside the package prepended with it's CTR and total image size. After checking the image's size against an hardcoded value (can change on firmware updates), the image is AES-CTR decrypted and the keyslot used for decryption is immediately cleared.

// Maximum stage 1 size on firmware version 1.0.0
u32 max_stage1_size = 0x29000;

u32 stage1_image_size = *(u32 *)stage1_addr;
u32 stage1_image_ctr_addr = stage1_addr + 0x10; 
u32 stage1_image_addr = stage1_addr + 0x20;

// Validate stage 1's size
if (stage1_image_size > max_stage1_size)
   panic();

u32 in_addr = stage1_image_addr;
u32 in_size = stage1_image_size;
u32 ctr_addr = stage1_image_ctr_addr;
u32 ctr_size = 0x10;
u32 out_addr = stage1_image_addr;
u32 out_size = stage1_image_size;
u32 keyslot = 0x0B;

// AES-CTR decrypt
// Use the pk11_key (keyslot 0x0B) to decrypt stage 1 in place
aes_ctr_decrypt(out_addr, out_size, keyslot, in_addr, in_size, ctr_addr, ctr_size);

// Clear pk11_key keyslot
clear_keyslot(0x0B);

// Validate the decrypted stage 1 image
// Checks the "PK11" magic and some pk11 header fields
bool is_valid = check_stage1(stage1_image_addr, stage1_image_size);

// Invalid stage 1 image
if (!is_valid)
   panic();

u32 pk11_header_size = 0x20;
u32 pk11_sec0_offset = *(u32 *)stage1_image_addr + 0x14;
u32 pk11_sec1_size = *(u32 *)stage1_image_addr + 0x18;

// Calculate PK11 main entrypoint
u32 pk11_main_addr = (stage1_image_addr + pk11_header_size + pk11_sec0_offset + pk11_sec1_size);
		
return pk11_main_addr;

Stage 1

When decrypted, this stage is encapsulated in a header.

Header

Offset Size Description
0x0 4 Magic "PK11"
0x4 4 Warmboot blob's size
0x8 4 Warmboot blob's offset
0xC 4 Unknown
0x10 4 NX bootloader blob's size
0x14 4 NX bootloader blob's offset
0x18 4 Secure Monitor blob's size
0x1C 4 Secure Monitor blob's offset

Changelog

2.0.0

  • The encrypted binaries' order and calculation for next stage's entrypoint was changed.

Old layout (before 2.0.0):

1.- PK11 header
2.- Secure Monitor blob
3.- NX bootloader blob
4.- Warmboot blob

NX bootloader entrypoint is calculated as:
0x40013FE0 + 0x20 + 0x20 + NX bootloader blob's offset + Secure Monitor blob's size

New layout (2.0.0+):

1.- PK11 header
2.- Warmboot blob
3.- NX bootloader blob
4.- Secure Monitor blob

NX bootloader entrypoint is calculated as:
0x40013FE0 + 0x20 + 0x20 + NX bootloader blob's offset + Warmboot blob's size
  • Some AES-ECB decryption related code was refactored.

3.0.0

  • The functions set_se_addr() and check_se_status() are now called right after enabling the security engine clocks and resets.

See Switch_System_Flaws#Stage_1_Bootloader.

  • Keyslot 0x0A is now used instead of keyslot 0x0D for generating the master_device_key.