devs/foxbox
1
0
Fork 0
forked from len0rd/rockbox
Code Pull requests Activity
foxbox/utils/atj2137/atjboottool/fwu.c
Nikita Burnashev e232f69214 atjboottool: gzipped fw files, option for big-endian fw, clarify ECIES in fwu 
Added fw modifications required to unpack real world player dumps.


Documented more fwu header fields, magic numbers and finite field arithmetics (extended Euclidean for inverse, long division for reducing modulo field_poly).

v3 encryption used is standard RC4 with the key additionally ciphered by the Elliptic Curve Integrated Encryption Scheme.

Either sect233k1 (NIST K-233) or sect163r2 (NIST B-163) curves can be used, with the former overwhelmingly prevailing, being hardwired in SDK's maker.exe. Using a private/public key scheme is superfluous because both are stored in the firmware, with the added level of complexity likely serving the purpose of obfuscation. The private key is generated at random with each invokation.

None of KDF or MAC from ECIES are used, RC4 key is directly xored with the shared secret. The random number r used to calculate rG isn't stored, but that's unimportant since only krG == rkG is actually used in the encryption.

Change-Id: Ieacf8cc744bc90c7c5582dd724b2c10a41bfc191
2023-05-17 13:19:31 -04:00

1237 lines
36 KiB
C
Raw Permalink Blame History

/***************************************************************************
* __________ __ ___.
* Open \______ \ ____ ____ | | _\_ |__ _______ ___
* Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
* Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
* Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
* \/ \/ \/ \/ \/
* $Id$
*
* Copyright (C) 2017 Amaury Pouly
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
* KIND, either express or implied.
*
****************************************************************************/
#include <string.h>
#include <stdlib.h>
#include "misc.h"
#include "fwu.h"
#include "afi.h"
#define check_field(v_exp, v_have, str_ok, str_bad) \
if((v_exp) != (v_have)) \
{ cprintf(RED, str_bad); return 1; } \
else { cprintf(RED, str_ok); }
#define check_field_soft(v_exp, v_have, str_ok, str_bad) \
if((v_exp) != (v_have)) \
{ cprintf(RED, str_bad); } \
else { cprintf(RED, str_ok); }
#define FWU_SIG_SIZE 16
#define FWU_BLOCK_SIZE 512
struct fwu_hdr_t
{
uint8_t sig[FWU_SIG_SIZE];
uint32_t fw_size;
uint32_t block_size;// always 512
uint8_t version;
uint8_t unk;
uint8_t sig2[FWU_SIG_SIZE];
} __attribute__((packed));
const uint8_t g_fwu_signature[FWU_SIG_SIZE] =
{
0x11, 0x22, 0x33, 0x44, 0x55, 0x66, 0x77, 0x88, 0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff, 0x75
};
struct fwu_crypto_hdr_t
{
uint8_t field0[16];
uint8_t unk;
uint8_t key[32];
} __attribute__((packed));
struct fwu_sector0_tail_t
{
uint8_t unk_2;
uint32_t unk_x808;
uint32_t unk_8;
uint8_t key_B[16];
uint8_t guid[16];
uint8_t unk_190;
uint8_t super_secret_xor[16];
uint8_t timestamp[8];
uint8_t unk_0;
uint8_t guid_filler[20];
uint8_t unk_1;
uint8_t check[20];
} __attribute__((packed));
struct fwu_block_A_hdr_t
{
uint16_t block_A_size;
uint8_t unk_0_a;
uint8_t unk_1_a;
uint8_t key_B[16];
uint8_t guid_filler[256];
uint8_t ec_sz;
uint8_t unk_0_b;
uint32_t unk_5;
uint32_t unk_x505;
uint16_t unk_1_b;
uint8_t timestamp[8];
} __attribute__((packed));
struct fwu_block_B_hdr_t
{
uint16_t block_B_size;
uint8_t unk_1_a;
uint16_t unk_1_b;
uint8_t timestamp[8];
uint16_t guid_filler_size;
} __attribute__((packed));
struct fwu_tail_t
{
uint8_t length; /* in blocks? it's always 1 */
uint8_t type; /* always 7 */
uint8_t reserved[14];
uint32_t fwu_checksum;
uint32_t flags; /* always 0x55aa55aa */
uint8_t desc[8]; /* always 'FwuTail' */
uint8_t fwu_crc_checksum[32]; /* always 0 */
uint8_t reserved2[444];
uint32_t fwutail_checksum;
} __attribute__((packed));
struct version_desc_t
{
uint8_t version;
uint8_t value;
uint8_t unk;
uint8_t sig2[FWU_SIG_SIZE];
};
struct version_desc_t g_version[] =
{
{ 1, 0xd, 0xd0, { 0x76, 0x5c, 0x50, 0x94, 0x69, 0xb0, 0xa7, 0x03, 0x10, 0xf1, 0x7e, 0xdb, 0x88, 0x90, 0x86, 0x9d } },
{ 1, 0xe, 0xd0, { 0x92, 0x22, 0x7a, 0x77, 0x08, 0x67, 0xae, 0x06, 0x16, 0x06, 0xb8, 0x65, 0xa6, 0x42, 0xf7, 0X52 } },
{ 3, 0x7e, 0xe1, { 0x3f, 0xad, 0xf8, 0xb0, 0x2e, 0xaf, 0x67, 0x49, 0xb9, 0x85, 0x5f, 0x63, 0x4e, 0x5e, 0x8e, 0x2e } },
};
#define NR_VERSIONS (int)(sizeof(g_version)/sizeof(g_version[0]))
typedef struct ec_point_t
{
uint32_t *x;
uint32_t *y;
}ec_point_t;
struct ec_info_t
{
int nr_bits;
int point_size;
uint32_t *ec_a; // size
uint32_t *ec_b; // size
uint32_t *field_poly; // size
uint32_t size;
ec_point_t pt_G;
ec_point_t pt_kG; // calculated ECIES public key
uint32_t field_bits;
int size_x2;
int nr_bytes;
int nr_dwords_m1;
int nr_dwords_x2;
int nr_dwords_x2_m1;
int nr_dwords;
};
struct ec_info_t g_ec_info;
struct fwu_block_A_hdr_t g_subblock_A;
uint8_t g_key_B[20];
uint8_t g_rc4_S[258];
uint8_t g_field_sz_byte;
ec_point_t g_public_key; // from block A
uint32_t *g_private_key; // from block B
#include "atj_tables.h"
#include <ctype.h>
void print_hex(const char *name, void *buf, size_t sz)
{
if(name)
cprintf(BLUE, "%s\n", name);
uint8_t *p = buf;
for(size_t i = 0; i < sz; i += 16)
{
if(name)
cprintf(OFF, " ");
for(size_t j = i; j < i + 16; j++)
if(j < sz)
cprintf(YELLOW, "%02x ", p[j]);
else
cprintf(OFF, " ");
cprintf(RED, " |");
for(size_t j = i; j < i + 16; j++)
cprintf(GREEN, "%c", (j < sz && isprint(p[j])) ? p[j] : '.');
cprintf(RED, "|\n");
}
}
void compute_checksum(uint8_t *buf, size_t size, uint8_t t[20])
{
memset(t, 0, 20);
for(size_t i = 0; i < size; i++)
t[i % 20] ^= buf[i];
for(int i = 0; i < 20; i++)
t[i] = ~t[i];
}
int check_block(uint8_t *buf, uint8_t ref[20], unsigned size)
{
uint8_t t[20];
compute_checksum(buf, size, t);
return memcmp(ref, t, 20);
}
int get_version(uint8_t *buf, unsigned long size)
{
(void) size;
struct fwu_hdr_t *hdr = (void *)buf;
for(int i = 0; i < NR_VERSIONS; i++)
if(hdr->version == g_version[i].value)
return i;
return -1;
}
static int decode_block_A(uint8_t block[1020])
{
uint8_t *p = &g_decode_A_table[32 * (block[998] & 31)];
uint8_t key[32];
for(int i = 0; i < 20; i++)
{
block[1000 + i] ^= p[i];
key[i] = block[1000 + i];
}
for(int i = 20; i < 32; i++)
key[i] = key[i - 20];
for(int i = 0; i < 31 * 32; i++)
block[i] ^= key[i % 32] ^ g_decode_A_table[i];
// FIXME dereferencing block - 1 is undefined behavior in standard C
return check_block(block - 1, block + 1000, 1001);
}
// https://en.wikipedia.org/wiki/RC4#Key-scheduling_algorithm_(KSA)
static void rc4_key_schedule(uint8_t *key, size_t keylength, uint8_t S[258])
{
for(int i = 0; i < 256; i++)
S[i] = i;
S[256] = S[257] = 0;
uint8_t j = 0;
for(int i = 0; i < 256; i++)
{
j = (j + S[i] + key[i % keylength]) % 256;
uint8_t tmp = S[i];
S[i] = S[j];
S[j] = tmp;
}
}
// https://en.wikipedia.org/wiki/RC4#Pseudo-random_generation_algorithm_(PRGA)
static void rc4_stream_cipher(uint8_t *buf, size_t size, uint8_t S[258])
{
uint8_t i = S[256];
uint8_t j = S[257];
for(size_t k = 0; k < size; k++)
{
i = (i + 1) % 256;
j = (j + S[i]) % 256;
uint8_t tmp = S[i];
S[i] = S[j];
S[j] = tmp;
buf[k] ^= S[(S[i] + S[j]) % 256];
}
}
static void rc4_cipher_block(uint8_t *keybuf, int keysize,
uint8_t *buf, int bufsize, uint8_t S[258])
{
rc4_key_schedule(keybuf, keysize, S);
rc4_stream_cipher(buf, bufsize, S);
}
static void rc4_key_swap(uint8_t *inbuf, uint8_t *outbuf, size_t size, int swap)
{
memcpy(outbuf, inbuf, size);
int a = swap & 0xf;
int b = (swap >> 4) + 16;
uint8_t v = outbuf[a];
outbuf[a] = outbuf[b];
outbuf[b] = v;
}
static void rc4_key_swap_and_decode(uint8_t keybuf[32], int swap,
uint8_t *buf, int bufsize, uint8_t S[258])
{
uint8_t keybuf_interm[32];
rc4_key_swap(keybuf, keybuf_interm, 32, swap);
rc4_cipher_block(keybuf_interm, 32, buf, bufsize, S);
}
static void gf_zero(void *buf, size_t size_dwords)
{
memset(buf, 0, 4 * size_dwords);
}
static void set_bit(int bit_pos, uint32_t *buf)
{
buf[bit_pos / 32] |= 1 << (bit_pos % 32);
}
static int fill_ec_info(uint8_t sz)
{
if(sz == 2) sz = 233;
else if(sz == 3) sz = 163;
else return 1;
g_ec_info.nr_bits = sz;
g_ec_info.nr_bytes = sz / 8 + (sz % 8 != 0);
g_ec_info.point_size = 2 * g_ec_info.nr_bytes;
g_ec_info.nr_dwords = sz / 32 + (sz % 32 != 0);
g_ec_info.size = 4 * g_ec_info.nr_dwords;
g_ec_info.size_x2 = 8 * g_ec_info.nr_dwords;
g_ec_info.nr_dwords_m1 = g_ec_info.nr_dwords - 1;
g_ec_info.nr_dwords_x2 = 2 * g_ec_info.nr_dwords;
g_ec_info.nr_dwords_x2_m1 = g_ec_info.nr_dwords_x2 - 1;
g_ec_info.pt_G.x = malloc(4 * g_ec_info.nr_dwords);
g_ec_info.pt_G.y = malloc(g_ec_info.size);
g_ec_info.pt_kG.x = malloc(g_ec_info.size);
g_ec_info.pt_kG.y = malloc(g_ec_info.size);
g_ec_info.field_poly = malloc(g_ec_info.size);
g_ec_info.ec_a = malloc(g_ec_info.size);
g_ec_info.ec_b = malloc(g_ec_info.size);
cprintf(BLUE, " Elliptic curve info:\n");
cprintf_field(" Field Bits: ", "%d\n", g_ec_info.nr_bits);
cprintf_field(" Field Bytes: ", "%d\n", g_ec_info.nr_bytes);
cprintf_field(" Point Size: ", "%d\n", g_ec_info.point_size);
cprintf_field(" Field DWords: ", "%d\n", g_ec_info.nr_dwords);
cprintf_field(" Size: ", "%d\n", g_ec_info.size);
return 0;
}
static int process_block_A(uint8_t block[1024])
{
cprintf(BLUE, "Block A\n");
int ret = decode_block_A(block + 4);
cprintf(GREEN, " Check: ");
check_field(ret, 0, "Pass\n", "Fail\n");
// print_hex("BlockA", block, 1024);
memcpy(&g_subblock_A, block, sizeof(g_subblock_A));
// assert(offsetof(struct fwu_block_A_hdr_t, ec_sz) == 276);
ret = fill_ec_info(g_subblock_A.ec_sz);
cprintf(GREEN, " Info: ");
check_field(ret, 0, "Pass\n", "Fail\n");
int tmp = 2 * g_ec_info.nr_bytes + 38;
int offset = 1004 - tmp + 5;
g_field_sz_byte = block[offset - 1];
g_public_key.x = malloc(g_ec_info.size);
g_public_key.y = malloc(g_ec_info.size);
memset(g_public_key.x, 0, g_ec_info.size);
memset(g_public_key.y, 0, g_ec_info.size);
memcpy(g_public_key.x, &block[offset], g_ec_info.nr_bytes);
int offset2 = g_ec_info.nr_bytes + offset;
memcpy(g_public_key.y, &block[offset2], g_ec_info.nr_bytes);
// assert(offsetof(struct fwu_block_A_hdr_t, unk_1_b) == 286);
cprintf_field(" Word: ", "%d ", g_subblock_A.unk_1_b);
check_field(g_subblock_A.unk_1_b, 1, "Ok\n", "Mismatch\n");
return 0;
}
static void decode_key_B(uint8_t buf[20], uint8_t buf2[16], uint8_t key[20])
{
for(int i = 0; i < 20; i++)
{
uint8_t v = buf[i] ^ g_decode_B_table[i];
key[i] = v;
buf[i] = v ^ buf2[i % 16];
}
}
static void decode_block_B(uint8_t *buf, uint8_t key[16], size_t size)
{
decode_key_B(&buf[size], key, g_key_B);
rc4_cipher_block(g_key_B, 20, buf, size, g_rc4_S);
}
static int find_last_bit_set(uint32_t *buf, bool a)
{
int i = a ? g_ec_info.nr_dwords_m1 : g_ec_info.nr_dwords_x2_m1;
while(i >= 0 && buf[i] == 0)
i--;
if(i < 0)
return -1;
for(int j = 31; j >= 0; j--)
if(buf[i] & (1 << j))
return 32 * i + j;
return -1; // unreachable
}
static void gf_copy(uint32_t *to, uint32_t *from)
{
for(int i = 0; i < g_ec_info.nr_dwords; i++)
to[i] = from[i];
}
static void gf_swap(uint32_t *a, uint32_t *b)
{
for(int i = 0; i < g_ec_info.nr_dwords; i++)
{
uint32_t c = a[i];
a[i] = b[i];
b[i] = c;
}
}
static void shift_left(uint32_t *buf, int nr_bits)
{
for(int i = g_ec_info.nr_dwords_m1; i >= 0; i--)
buf[i + (nr_bits / 32)] = buf[i];
memset(buf, 0, 4 * (nr_bits / 32));
size_t size = g_ec_info.nr_dwords + (nr_bits + 31) / 32;
nr_bits = nr_bits % 32;
uint32_t acc = 0;
for(size_t i = 0; i < size; i++)
{
uint32_t new_val = buf[i] << nr_bits | acc;
/* WARNING if nr_bits = 0 then the right shift by 32 is undefined and so
* the following code could break. The additional AND catches this case
* and make sure the result is 0 */
acc = ((1 << nr_bits) - 1) & (buf[i] >> (32 - nr_bits));
buf[i] = new_val;
}
}
static void gf_add_x2(uint32_t *res, uint32_t *a, uint32_t *b)
{
for(int i = 0; i < g_ec_info.nr_dwords_x2; i++)
res[i] = a[i] ^ b[i];
}
static void print_poly(const char *name, uint32_t *poly, int nr_dwords)
{
bool first = true;
cprintf(RED, "%s", name);
for(int dw = 0; dw < nr_dwords; dw++)
{
for(int i = 0; i < 32; i++)
{
if(!(poly[dw] & (1 << i)))
continue;
if(first)
first = false;
else
cprintf(OFF, "+");
cprintf(OFF, "x^%d", dw * 32 + i);
}
}
cprintf(OFF, "\n");
}
/* https://en.wikipedia.org/wiki/Extended_Euclidean_algorithm#Simple_algebraic_field_extensions
* invariant: p * s + a * t == r -> a * t == r (mod p)
* loop until only lowest bit set (r == 1) -> inverse in t */
static void gf_inverse(uint32_t *newt, uint32_t *val)
{
uint32_t *tmp = malloc(g_ec_info.size_x2);
uint32_t *r = malloc(g_ec_info.size_x2);
uint32_t *newr = malloc(g_ec_info.size_x2);
uint32_t *t = malloc(g_ec_info.size_x2);
gf_zero(tmp, g_ec_info.nr_dwords_x2);
/* newt := 1 */
gf_zero(newt, g_ec_info.nr_dwords);
*newt = 1;
/* t := 0 */
gf_zero(t, g_ec_info.nr_dwords);
/* newr := a */
gf_copy(newr, val);
/* r := p */
gf_copy(r, g_ec_info.field_poly);
for(int i = find_last_bit_set(newr, 1); i; i = find_last_bit_set(newr, 1))
{
/* pos := degree(newr) - degree(r) */
int pos = i - find_last_bit_set(r, 1);
if(pos < 0)
{
gf_swap(newr, r);
gf_swap(newt, t);
pos = -pos;
}
/* newr := newr - x^pos * r */
gf_copy(tmp, r);
shift_left(tmp, pos);
gf_add_x2(newr, newr, tmp);
/* newt := newt - x^pos * t */
gf_copy(tmp, t);
shift_left(tmp, pos);
gf_add_x2(newt, newt, tmp);
}
free(tmp);
free(r);
free(newr);
free(t);
}
static void shift_left_one(uint32_t *a)
{
int pos = find_last_bit_set(a, 0) / 32 + 1;
if(pos <= 0)
return;
uint32_t v = 0;
for(int i = 0; i < pos; i++)
{
uint32_t new_val = v | a[i] << 1;
v = a[i] >> 31;
a[i] = new_val;
}
if(v)
a[pos] = v;
}
#if 1
static void gf_mult(uint32_t *res, uint32_t *a2, uint32_t *a3)
{
uint32_t *tmp2 = malloc(g_ec_info.size_x2);
gf_zero(tmp2, g_ec_info.nr_dwords_x2);
gf_copy(tmp2, a3);
int pos = g_ec_info.nr_dwords;
uint32_t mask = 1;
for(int i = 0; i < 32; i++)
{
for(int j = 0; j < g_ec_info.nr_dwords; j++)
{
if(a2[j] & mask)
for(int k = 0; k < pos; k++)
res[j + k] ^= tmp2[k];
}
shift_left_one(tmp2);
mask <<= 1;
pos = find_last_bit_set(tmp2, 0) / 32 + 1;
}
free(tmp2);
}
#else
static void gf_mult(uint32_t *res, uint32_t *a2, uint32_t *a3)
{
for(int i = 0; i < 32 * g_ec_info.nr_dwords; i++)
for(int j = 0; j < 32 * g_ec_info.nr_dwords; j++)
{
int k = i + j;
uint32_t v1 = (a2[i / 32] >> (i % 32)) & 1;
uint32_t v2 = (a3[j / 32] >> (j % 32)) & 1;
res[k / 32] ^= (v1 * v2) << (k % 32);
}
}
#endif
// https://en.wikipedia.org/wiki/Polynomial_long_division#Pseudocode
static void gf_mod(uint32_t *r, uint32_t *field_poly)
{
uint32_t *tmp = malloc(g_ec_info.size_x2);
int deg_d = g_ec_info.field_bits;
int deg_r = find_last_bit_set(r, 0);
/* i := degree(lead(r) / lead(d)) */
for(int i = deg_r - deg_d; i >= 0; i = find_last_bit_set(r, 0) - deg_d)
{
/* r := r - x^i * d */
gf_zero(tmp, g_ec_info.nr_dwords_x2);
gf_copy(tmp, field_poly);
shift_left(tmp, i);
gf_add_x2(r, r, tmp);
}
free(tmp);
}
static void gf_add(uint32_t *res, uint32_t *a, uint32_t *b)
{
for(int i = 0; i < g_ec_info.nr_dwords; i++)
res[i] = a[i] ^ b[i];
}
static void print_point(const char *name, ec_point_t *ptr)
{
cprintf(BLUE, "%s\n", name);
print_poly(" x: ", ptr->x, g_ec_info.nr_dwords);
print_poly(" y: ", ptr->y, g_ec_info.nr_dwords);
}
static uint32_t g_gf_one[9] =
{
1, 0, 0, 0, 0, 0, 0, 0, 0
};
static void ec_double(ec_point_t *point, ec_point_t *res)
{
uint32_t *v2 = malloc(g_ec_info.size_x2);
uint32_t *v3 = malloc(g_ec_info.size_x2);
uint32_t *v4 = malloc(g_ec_info.size_x2);
uint32_t *v5 = malloc(g_ec_info.size_x2);
uint32_t *v6 = malloc(g_ec_info.size_x2);
gf_zero(res->x, g_ec_info.nr_dwords);
gf_zero(res->y, g_ec_info.nr_dwords);
gf_zero(v3, g_ec_info.nr_dwords_x2);
gf_zero(v6, g_ec_info.nr_dwords_x2);
gf_zero(v4, g_ec_info.nr_dwords_x2);
/* v4 := 1/x */
gf_inverse(v4, point->x);
gf_zero(v5, g_ec_info.nr_dwords_x2);
/* v5 := y/x */
gf_mult(v5, v4, point->y);
gf_mod(v5, g_ec_info.field_poly);
/* v2 := x + y/x (lambda) */
gf_add(v2, point->x, v5);
/* v4 := ec_a + lambda */
gf_add(v4, v2, g_ec_info.ec_a);
gf_zero(v3, g_ec_info.nr_dwords_x2);
/* v3 := lambda^2 */
gf_mult(v3, v2, v2);
gf_mod(v3, g_ec_info.field_poly);
/* x' := lambda + lambda^2 + ec_a */
gf_add(res->x, v4, v3);
gf_zero(v5, g_ec_info.nr_dwords_x2);
/* v4 := lambda + g_gf_one */
gf_add(v4, v2, g_gf_one);
/* v5 := (lambda + 1) * x' = lambda.x' + x' */
gf_mult(v5, v4, res->x);
gf_mod(v5, g_ec_info.field_poly);
gf_zero(v6, g_ec_info.nr_dwords_x2);
/* v6 := x1^2 */
gf_mult(v6, point->x, point->x);
gf_mod(v6, g_ec_info.field_poly);
/* y' = (lambda + g_gf_one) * x + x^2 = x^2 + lambda.x + x */
gf_add(res->y, v5, v6);
free(v2);
free(v3);
free(v4);
free(v5);
free(v6);
}
static void ec_add(ec_point_t *a1, ec_point_t *a2, ec_point_t *res)
{
uint32_t *v3 = malloc(g_ec_info.size_x2);
uint32_t *v4 = malloc(g_ec_info.size_x2);
uint32_t *v5 = malloc(g_ec_info.size_x2);
uint32_t *v6 = malloc(g_ec_info.size_x2);
uint32_t *v7 = malloc(g_ec_info.size_x2);
gf_zero(res->x, g_ec_info.nr_dwords);
gf_zero(res->y, g_ec_info.nr_dwords);
gf_zero(v4, g_ec_info.nr_dwords_x2);
gf_zero(v7, g_ec_info.nr_dwords_x2);
/* v5 = y1 + y2 */
gf_add(v5, a1->y, a2->y);
/* v6 = x1 + x2 */
gf_add(v6, a1->x, a2->x);
/* v7 = 1/(x1 + x2) */
gf_inverse(v7, v6);
gf_zero(v3, g_ec_info.nr_dwords_x2);
/* v3 = (y1 + y2) / (x1 + x2) (lambda) */
gf_mult(v3, v7, v5);
gf_mod(v3, g_ec_info.field_poly);
/* v5 = lambda + ec_a */
gf_add(v5, v3, g_ec_info.ec_a);
gf_zero(v4, g_ec_info.nr_dwords_x2);
/* v4 = lambda^2 */
gf_mult(v4, v3, v3);
gf_mod(v4, g_ec_info.field_poly);
/* v7 = lambda^2 + lambda + ec_a */
gf_add(v7, v5, v4);
/* x' = ec_a + x1 + x2 + lambda + lambda^2 */
gf_add(res->x, v7, v6);
/* v5 = x1 + x' */
gf_add(v5, a1->x, res->x);
/* v6 = x' + y1 */
gf_add(v6, res->x, a1->y);
gf_zero(v7, g_ec_info.nr_dwords_x2);
/* v7 = (x1 + x').lambda */
gf_mult(v7, v5, v3);
gf_mod(v7, g_ec_info.field_poly);
/* y' = (x1 + x').lambda + x' + y1 */
gf_add(res->y, v7, v6);
free(v3);
free(v4);
free(v5);
free(v6);
free(v7);
}
static int ec_mult(uint32_t *n, ec_point_t *point, ec_point_t *res)
{
ec_point_t res_others;
res_others.x = malloc(g_ec_info.size);
res_others.y = malloc(g_ec_info.size);
gf_zero(res->x, g_ec_info.nr_dwords);
gf_zero(res->y, g_ec_info.nr_dwords);
gf_zero(res_others.x, g_ec_info.nr_dwords);
gf_zero(res_others.y, g_ec_info.nr_dwords);
int pos = find_last_bit_set(n, 1);
/* res_other := point */
gf_copy(res_others.x, point->x);
gf_copy(res_others.y, point->y);
/* for all bit from SZ-1 downto 0 */
for(int bit = (pos % 32) - 1; bit >= 0; bit--)
{
/* res := 2 * res_other */
ec_double(&res_others, res);
/* res_other := res = 2 * res_other */
gf_copy(res_others.x, res->x);
gf_copy(res_others.y, res->y);
/* if bit of n is set */
if(n[pos / 32] & (1 << bit))
{
/* res := res_other + point */
ec_add(&res_others, point, res);
gf_copy(res_others.x, res->x);
gf_copy(res_others.y, res->y);
}
}
/* same but optimized */
for(int i = pos / 32 - 1; i >= 0; i--)
{
for(int bit = 31; bit >= 0; bit--)
{
ec_double(&res_others, res);
gf_copy(res_others.x, res->x);
gf_copy(res_others.y, res->y);
if(n[i] & (1 << bit))
{
ec_add(&res_others, point, res);
gf_copy(res_others.x, res->x);
gf_copy(res_others.y, res->y);
}
}
}
gf_copy(res->x, res_others.x);
gf_copy(res->y, res_others.y);
free(res_others.x);
free(res_others.y);
return 0;
}
static void xor_with_point(uint8_t *buf, ec_point_t *point)
{
/*
int sz = g_ec_info.nr_bytes - 1;
if(sz <= 32)
{
for(int i = 0; i < sz; i++)
buf[i] ^= point->x[i];
for(int i = sz; i < 32; i++)
buf[i] ^= point->y[i - sz];
}
else
for(int i = 0; i < 32; i++)
buf[i] ^= point->x[i];
*/
uint8_t *ptrA = (uint8_t *)point->x;
uint8_t *ptrB = (uint8_t *)point->y;
int sz = MIN(g_ec_info.nr_bytes - 1, 32);
for(int i = 0; i < sz; i++)
buf[i] ^= ptrA[i];
for(int i = sz; i < 32; i++)
buf[i] ^= ptrB[i - sz];
}
// https://en.wikipedia.org/wiki/Integrated_Encryption_Scheme#Formal_description_of_ECIES
static int xor_with_shared_secret(uint8_t *buf, ec_point_t *pt_rG, uint32_t *private_key)
{
ec_point_t shared_secret;
shared_secret.x = malloc(g_ec_info.size);
shared_secret.y = malloc(g_ec_info.size);
gf_zero(shared_secret.x, g_ec_info.nr_dwords);
gf_zero(shared_secret.y, g_ec_info.nr_dwords);
int ret = ec_mult(private_key, pt_rG, &shared_secret);
if(ret == 0)
xor_with_point(buf, &shared_secret);
free(shared_secret.x);
free(shared_secret.y);
return ret;
}
static int set_field_poly(uint32_t *field_poly, int field_sz)
{
gf_zero(field_poly, g_ec_info.nr_dwords);
g_ec_info.field_bits = 0;
if(field_sz == 4)
{
set_bit(0, field_poly);
set_bit(74, field_poly);
set_bit(233, field_poly);
g_ec_info.field_bits = 233;
return 0;
}
else if (field_sz == 5)
{
set_bit(0, field_poly);
set_bit(3, field_poly);
set_bit(6, field_poly);
set_bit(7, field_poly);
set_bit(163, field_poly);
g_ec_info.field_bits = 163;
return 0;
}
else
return 1;
}
static int ec_init(ec_point_t *ec_G, char field_sz)
{
int ret = set_field_poly(g_ec_info.field_poly, field_sz);
if(ret) return ret;
if(field_sz == 4)
{
gf_copy(ec_G->x, g_sect233k1_G_x);
gf_copy(ec_G->y, g_sect233k1_G_y);
gf_copy(g_ec_info.ec_a, g_sect233k1_a); // zero
gf_copy(g_ec_info.ec_b, g_sect233k1_b); // never used
return 0;
}
else if(field_sz == 6 ) // yet to find even a single specimen
{
gf_copy(ec_G->x, g_sect163r2_G_x);
gf_copy(ec_G->y, g_sect163r2_G_y);
gf_copy(g_ec_info.ec_a, g_sect163r2_a);
gf_copy(g_ec_info.ec_b, g_sect163r2_b);
return 0;
}
else
return 1;
}
static void create_guid(void *uid, int bit_size)
{
uint8_t *p = uid;
for(int i = 0; i < bit_size / 8; i++)
p[i] = rand() % 256;
}
static int process_block_B(uint8_t block[512])
{
struct fwu_block_B_hdr_t *p_hdr = (void *)block;
cprintf(BLUE, "Block B\n");
decode_block_B(block + 3, g_subblock_A.key_B, 492 - 3);
cprintf_field(" Word: ", "%d ", p_hdr->unk_1_b);
check_field(p_hdr->unk_1_b, 1, "Ok\n", "Mismatch\n");
int ret = check_block(block, block + 492, 492);
cprintf(GREEN, " Check: ");
check_field(ret, 0, "Pass\n", "Fail\n");
g_private_key = malloc(g_ec_info.size);
memset(g_private_key, 0, g_ec_info.size);
int offset = sizeof *p_hdr + p_hdr->guid_filler_size + 1;
memcpy(g_private_key, &block[offset], g_ec_info.nr_bytes);
return 0;
}
static int get_key_fwu_v3(size_t size, uint8_t *buf, uint8_t *blockA, uint8_t *blockB,
uint8_t *keybuf, uint8_t *blo)
{
(void) size;
uint8_t smallblock[512];
uint8_t bigblock[1024];
memset(smallblock, 0, sizeof(smallblock));
memset(bigblock, 0, sizeof(bigblock));
*blockA = buf[0x1ee] & 15;
*blockB = buf[0x1fe] & 15;
size_t offsetA = 512 * (1 + *blockA);
size_t offsetB = 512 * (1 + *blockB);
cprintf(BLUE, "Crypto\n");
cprintf_field(" Block A: ", "0x%zx\n", 512 + offsetA);
cprintf_field(" Block B: ", "0x%zx\n", 512 + offsetA + 1024 + offsetB);
memcpy(bigblock, &buf[512 + offsetA], sizeof(bigblock));
int ret = process_block_A(bigblock);
if(ret != 0)
return ret;
memcpy(smallblock, &buf[512 + offsetA + 1024 + offsetB], sizeof(smallblock));
ret = process_block_B(smallblock);
if(ret != 0)
return ret;
cprintf(BLUE, "Main\n");
struct fwu_crypto_hdr_t crypto_hdr;
memcpy(&crypto_hdr, buf + sizeof(struct fwu_hdr_t), sizeof(crypto_hdr));
cprintf_field(" Byte: ", "%d ", crypto_hdr.unk);
check_field(crypto_hdr.unk, 3, "Ok\n", "Mismatch\n");
size_t offset = sizeof(struct fwu_hdr_t) + sizeof(struct fwu_crypto_hdr_t);
ec_point_t pt_rG;
pt_rG.x = malloc(g_ec_info.size);
pt_rG.y = malloc(g_ec_info.size);
memset(pt_rG.x, 0, g_ec_info.size);
memset(pt_rG.y, 0, g_ec_info.size);
memcpy(pt_rG.x, buf + offset, g_ec_info.nr_bytes);
memcpy(pt_rG.y, buf + offset + g_ec_info.nr_bytes, g_ec_info.nr_bytes);
ret = ec_init(&g_ec_info.pt_G, g_field_sz_byte);
cprintf(GREEN, " Elliptic curve init: ");
check_field(ret, 0, "Pass\n", "Fail\n");
ec_mult(g_private_key, &g_ec_info.pt_G, &g_ec_info.pt_kG);
cprintf(GREEN, " Public key check: ");
if (memcmp(g_public_key.x, g_ec_info.pt_kG.x, g_ec_info.nr_bytes) ||
memcmp(g_public_key.y, g_ec_info.pt_kG.y, g_ec_info.nr_bytes))
{
cprintf(RED, "Fail\n");
return 1;
}
else
cprintf(RED, "Pass\n");
ret = xor_with_shared_secret(crypto_hdr.key, &pt_rG, g_private_key);
cprintf(GREEN, " ECIES decryption: ");
check_field(ret, 0, "Pass\n", "Fail\n");
memcpy(keybuf, crypto_hdr.key, 32);
offset += g_ec_info.point_size;
rc4_key_swap_and_decode(keybuf, 0, &buf[offset], 512 - offset, g_rc4_S);
int pos = *(uint16_t *)&buf[offset];
cprintf_field(" Filler size: ", "%d ", pos);
int tmp = offset + sizeof(struct fwu_sector0_tail_t);
check_field(pos, 510 - tmp, "Ok\n", "Mismatch\n");
struct fwu_sector0_tail_t tail;
memcpy(&tail, &buf[offset + 2 + pos], sizeof(tail));
cprintf_field(" Byte: ", "%d ", tail.unk_2);
check_field(tail.unk_2, 2, "Ok\n", "Invalid\n");
cprintf_field(" DWord: ", "0x%x ", tail.unk_x808);
check_field(tail.unk_x808, 0x808, "Ok\n", "Invalid\n");
cprintf_field(" DWord: ", "%d ", tail.unk_8);
check_field(tail.unk_8, 8, "Ok\n", "Invalid\n");
cprintf_field(" Byte: ", "%d ", tail.unk_190);
check_field(tail.unk_190, 190, "Ok\n", "Invalid\n");
/* encode super secret at random position in guid stream, never used */
memset(blo, 0, 512);
create_guid(smallblock, 476 * 8);
memcpy(smallblock + 476, tail.super_secret_xor, 16);
compute_checksum(smallblock, 492, blo + 492);
int bsz = blo[500];
memcpy(blo, smallblock, bsz);
memcpy(blo + bsz, tail.super_secret_xor, 16);
memcpy(blo + bsz + 16, smallblock + bsz, 476 - bsz);
rc4_cipher_block(blo + 492, 16, blo, 492, g_rc4_S);
ret = check_block(buf + sizeof(struct fwu_hdr_t), tail.check, 492 - sizeof(struct fwu_hdr_t));
cprintf(GREEN, " Check: ");
check_field(ret, 0, "Pass\n", "Fail\n");
ret = memcmp(g_subblock_A.key_B, tail.key_B, 16);
cprintf(GREEN, " Compare: ");
check_field(ret, 0, "Pass\n", "Fail\n");
/*
ret = memcmp(tail.guid, zero, sizeof(zero));
cprintf(GREEN, " Sanity: ");
check_field(ret, 0, "Pass\n", "Fail\n");
*/
return 0;
}
/* stolen from https://github.com/nfd/atj2127decrypt, I have no idea from where
* he got this sequence of code. This code is really weird, I copy verbatim
* his authors comment below. */
uint32_t atj2127_key[] =
{
0x42146ea2, 0x892c8e85, 0x9f9f6d27, 0x545fedc3,
0x09e5c0ca, 0x2dfa7e61, 0x4e5322e6, 0xb19185b9
};
/* decrypt a 512-byte sector */
static void atj2127_decrypt_sector(void *inbuf, size_t size,
uint32_t session_key[8], int rounds_to_perform)
{
uint32_t key[8];
for(int i = 0; i < 8; i++)
key[i] = atj2127_key[i] ^ session_key[i];
uint32_t *buf = inbuf;
if(size % 32)
cprintf(GREY, "Size is not a multiple of 32!!!\n");
while(rounds_to_perform > 0)
{
uint32_t rollover = buf[7] ^ session_key[7];
buf[0] ^= key[1];
buf[1] ^= key[2];
buf[2] ^= key[3];
buf[3] ^= key[4];
buf[4] ^= key[5];
buf[5] ^= key[6];
buf[6] ^= key[7];
buf[7] ^= key[1] ^ key[4];
key[1] = key[2];
key[2] = key[3];
key[3] = key[4];
key[4] = key[5];
key[5] = key[6];
key[6] = key[7];
key[7] = rollover;
buf += 8;
rounds_to_perform -= 1;
}
}
static void atj2127_decrypt(uint8_t *dst, const uint8_t *src, size_t size,
uint8_t keybuf[32], int rounds_to_perform)
{
cprintf(BLUE, "ATJ2127:\n");
cprintf_field(" Rounds: ", "%d\n", rounds_to_perform);
while(size > 0)
{
int sec_sz = MIN(size, 512);
memcpy(dst, src, sec_sz);
atj2127_decrypt_sector(dst, sec_sz, (uint32_t *)keybuf, rounds_to_perform);
src += sec_sz;
dst += sec_sz;
size -= sec_sz;
}
}
static int decrypt_fwu_v3(uint8_t *buf, size_t *size, uint8_t block[512], enum fwu_mode_t mode)
{
uint8_t blockA;
uint8_t blockB;
uint8_t keybuf[32];
struct fwu_hdr_t *hdr = (void *)buf;
memset(keybuf, 0, sizeof(keybuf));
int ret = get_key_fwu_v3(*size, buf, &blockA, &blockB, keybuf, block);
if(ret != 0)
return ret;
size_t file_size = *size;
/* the input buffer is reorganized based on two offsets (blockA and blockB),
* skip 2048 bytes of data used for crypto init */
*size = hdr->fw_size; /* use firmware size, not file size */
*size -= 512 + 1024 + 512; /* sector0 + blockA + blockB */
uint8_t *tmpbuf = malloc(*size);
memset(tmpbuf, 0, *size);
int offsetA = 512 * (1 + blockA);
int offsetB = 512 * (1 + blockB);
memcpy(tmpbuf, buf + 512, offsetA);
memcpy(tmpbuf + offsetA, buf + 512 + offsetA + 1024, offsetB);
memcpy(tmpbuf + offsetA + offsetB,
buf + 512 + offsetA + 1024 + offsetB + 512, *size - offsetA - offsetB);
/* stolen from https://github.com/nfd/atj2127decrypt, I have no idea from where
* he got this sequence of code. This code is really weird, I copy verbatim
* his authors comment below.
*
* This is really weird. This is passed to the decrypt-sector function and
* determines how much of each 512-byte sector to decrypt, where for every
* 32MB of size above the first 32MB, one 32 byte chunk of each sector
* (starting from the end) will remain unencrypted, up to a maximum of 480
* bytes of plaintext. Was this a speed-related thing? It just seems
* completely bizarre. */
/* NOTE: the original code uses the file length to determine how much
* to encrypt and not the size reported in the header. Since
* the file size can be different from the size reported in the header
* (the infamous 512 bytes described above), this might be wrong. */
int rounds_to_perform = 16 - (file_size >> 0x19);
if(rounds_to_perform <= 0)
rounds_to_perform = 1;
/* the ATJ213x and ATJ2127 do not use the same encryption at this point, and I
* don't see any obvious way to tell which encryption is used (since they
* use the same version above). */
bool is_atj2127 = false;
if(mode == FWU_AUTO)
{
uint8_t hdr_buf[512];
atj2127_decrypt(hdr_buf, tmpbuf, sizeof(hdr_buf), keybuf, rounds_to_perform);
is_atj2127 = afi_check(hdr_buf, sizeof(hdr_buf));
if(is_atj2127)
cprintf(BLUE, "File looks like an ATJ2127 firmware\n");
else
cprintf(BLUE, "File does not looks like an ATJ2127 firmware\n");
}
else if(mode == FWU_ATJ2127)
is_atj2127 = true;
if(is_atj2127)
atj2127_decrypt(buf, tmpbuf, *size, keybuf, rounds_to_perform);
else
{
rc4_key_schedule(keybuf, 32, g_rc4_S);
rc4_stream_cipher(tmpbuf, *size, g_rc4_S);
memcpy(buf, tmpbuf, *size);
}
return 0;
}
uint32_t fwu_checksum(void *buf, size_t size)
{
if(size % 4)
cprintf(GREY, "WARNING: checksum of buffer whose length is not a multiple of 4");
uint32_t *p = buf;
uint32_t sum = 0;
for(size_t i = 0; i < size / 4; i++)
sum += *p++;
return sum;
}
int fwu_decrypt(uint8_t *buf, size_t *size, enum fwu_mode_t mode)
{
struct fwu_hdr_t *hdr = (void *)buf;
if(*size < sizeof(struct fwu_hdr_t))
{
cprintf(GREY, "File too small\n");
return 1;
}
cprintf(BLUE, "Header\n");
cprintf(GREEN, " Signature:");
for(int i = 0; i < FWU_SIG_SIZE; i++)
cprintf(YELLOW, " %02x", hdr->sig[i]);
if(memcmp(hdr->sig, g_fwu_signature, FWU_SIG_SIZE) == 0)
cprintf(RED, " Ok\n");
else
{
cprintf(RED, " Mismatch\n");
return 1;
}
cprintf_field(" FW size: ", "%d ", hdr->fw_size);
if(hdr->fw_size == *size)
cprintf(RED, " Ok\n");
else if(hdr->fw_size < *size)
cprintf(RED, " Ok (file greater than firmware)\n");
else
{
cprintf(RED, " Error (file too small)\n");
return 1;
}
cprintf_field(" Block size: ", "%d ", hdr->block_size);
check_field(hdr->block_size, FWU_BLOCK_SIZE, "Ok\n", "Invalid\n");
cprintf_field(" Version: ", "0x%x ", hdr->version);
int ver = get_version(buf, *size);
if(ver < 0)
{
cprintf(RED, "(Unknown)\n");
return 1;
}
else
cprintf(RED, "(Ver. %d)\n", g_version[ver].version);
cprintf_field(" Unknown: ", "0x%x ", hdr->unk);
check_field(hdr->unk, g_version[ver].unk, "Ok\n", "Invalid\n");
cprintf(GREEN, " Signature:");
for(int i = 0; i < FWU_SIG_SIZE; i++)
cprintf(YELLOW, " %02x", hdr->sig2[i]);
if(memcmp(hdr->sig2, g_version[ver].sig2, FWU_SIG_SIZE) == 0)
cprintf(RED, " Ok\n");
else
{
cprintf(RED, " Mismatch\n");
return 2;
}
/* check whether the firmware has a FwuTail (as far as I know, there is no flag anywhere that
* indicates its presence or not) */
struct fwu_tail_t *tail = (void *)(buf + hdr->fw_size - sizeof(struct fwu_tail_t));
if(tail->flags == 0x55aa55aa && strcmp((char *)tail->desc, "FwuTail") == 0)
{
cprintf(BLUE, "Tail\n");
cprintf_field(" Length: ", "%d ", tail->length);
check_field_soft(tail->length, 1, "Ok\n", "Fail\n");
cprintf_field(" Type: ", "%d ", tail->type);
check_field_soft(tail->type, 7, "Ok\n", "Fail\n");
cprintf_field(" FW checksum: ", "%x ", tail->fwu_checksum);
check_field_soft(fwu_checksum(buf, hdr->fw_size - sizeof(struct fwu_tail_t)),
tail->fwu_checksum, "Ok\n", "Mismatch\n");
cprintf(GREEN, " FW CRC Checksum: ");
for(unsigned i = 0; i < sizeof(tail->fwu_crc_checksum); i++)
cprintf(YELLOW, "%02x", tail->fwu_crc_checksum[i]);
cprintf(RED, " Ignored (should be 0)\n");
cprintf_field(" Tail checksum: ", "%x ", tail->fwutail_checksum);
check_field_soft(fwu_checksum(tail, sizeof(struct fwu_tail_t) - 4),
tail->fwutail_checksum, "Ok\n", "Mismatch\n");
/* if it has a tail, the firmware size includes it, so we need to decrease it to avoid
* "decrypting" the tail and output garbage */
hdr->fw_size -= sizeof(struct fwu_tail_t);
}
else
cprintf(BLUE, "Firmware does not seem to have a tail\n");
if(g_version[ver].version == 3)
{
uint8_t block[512];
memset(block, 0, sizeof(block));
return decrypt_fwu_v3(buf, size, block, mode);
}
else
{
cprintf(GREY, "Unsupported version: %d\n", g_version[ver].version);
return 1;
}
}
bool fwu_check(uint8_t *buf, size_t size)
{
struct fwu_hdr_t *hdr = (void *)buf;
if(size < sizeof(struct fwu_hdr_t))
return false;
return memcmp(hdr->sig, g_fwu_signature, FWU_SIG_SIZE) == 0;
}
View git blame Copy permalink