static void init(struct fmt_main *self) {
char *kpc;
global_work_size = MAX_KEYS_PER_CRYPT;
opencl_init("$JOHN/md4_kernel.cl", ocl_gpu_id, platform_id);
crypt_kernel = clCreateKernel(program[ocl_gpu_id], "md4", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
if( ((kpc = getenv("LWS")) == NULL) || (atoi(kpc) == 0)) {
create_clobj(MD4_NUM_KEYS);
opencl_find_best_workgroup(self);
release_clobj();
}else {
local_work_size = atoi(kpc);
}
if( (kpc = getenv("GWS")) == NULL){
max_keys_per_crypt = MD4_NUM_KEYS;
create_clobj(MD4_NUM_KEYS);
} else {
if (atoi(kpc) == 0){
//user chose to die of boredom
max_keys_per_crypt = MD4_NUM_KEYS;
create_clobj(MD4_NUM_KEYS);
find_best_kpc();
} else {
max_keys_per_crypt = atoi(kpc);
create_clobj(max_keys_per_crypt);
}
}
fprintf(stderr, "Local work size (LWS) %d, Global work size (GWS) %d\n",(int)local_work_size, max_keys_per_crypt);
self->params.max_keys_per_crypt = max_keys_per_crypt;
}
static void init(struct fmt_main *self)
{
host_pass = calloc(KEYS_PER_CRYPT, sizeof(pwsafe_pass));
host_hash = calloc(KEYS_PER_CRYPT, sizeof(pwsafe_hash));
host_salt = calloc(1, sizeof(pwsafe_salt));
opencl_init("$JOHN/pwsafe_kernel.cl", ocl_gpu_id, platform_id);
///Alocate memory on the GPU
mem_salt =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_READ_ONLY, saltsize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code, "Error while allocating memory for salt");
mem_in =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code, "Error while allocating memory for passwords");
mem_out =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code, "Error while allocating memory for hashes");
///Assign kernel parameters
crypt_kernel = clCreateKernel(program[ocl_gpu_id], KERNEL_NAME, &ret_code);
HANDLE_CLERROR(ret_code, "Error while creating kernel");
clSetKernelArg(crypt_kernel, 0, sizeof(mem_in), &mem_in);
clSetKernelArg(crypt_kernel, 1, sizeof(mem_out), &mem_out);
clSetKernelArg(crypt_kernel, 2, sizeof(mem_salt), &mem_salt);
opencl_find_best_workgroup(self);
//local_work_size=256;
atexit(release_all);
}
static void init(struct fmt_main *_self)
{
CRC32_t crc;
char build_opts[64];
cl_int cl_error;
self = _self;
CRC32_Init(&crc);
snprintf(build_opts, sizeof(build_opts),
"-DPLAINTEXT_LENGTH=%d -DHASH_LOOPS=%d",
PLAINTEXT_LENGTH, HASH_LOOPS);
opencl_init("$JOHN/kernels/7z_kernel.cl",
gpu_id, build_opts);
sevenzip_init = clCreateKernel(program[gpu_id], "sevenzip_init",
&cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
crypt_kernel = clCreateKernel(program[gpu_id], "sevenzip_crypt",
&cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
if (pers_opts.target_enc == UTF_8)
self->params.plaintext_length = MIN(125, 3 * PLAINTEXT_LENGTH);
}
static void init(struct fmt_main *_self)
{
char build_opts[64];
static char valgo[sizeof(ALGORITHM_NAME) + 8] = "";
self = _self;
opencl_preinit();
/* VLIW5 does better with just 2x vectors due to GPR pressure */
if (!options.v_width && amd_vliw5(device_info[gpu_id]))
v_width = 2;
else
v_width = opencl_get_vector_width(gpu_id, sizeof(cl_int));
if (v_width > 1) {
/* Run vectorized kernel */
snprintf(valgo, sizeof(valgo),
ALGORITHM_NAME " %ux", v_width);
self->params.algorithm_name = valgo;
}
snprintf(build_opts, sizeof(build_opts),
"-DHASH_LOOPS=%u -DOUTLEN=%u "
"-DPLAINTEXT_LENGTH=%u -DV_WIDTH=%u",
HASH_LOOPS, OUTLEN, PLAINTEXT_LENGTH, v_width);
opencl_init("$JOHN/kernels/pbkdf1_hmac_sha1_kernel.cl",
gpu_id, build_opts);
pbkdf1_init = clCreateKernel(program[gpu_id], "pbkdf1_init", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel");
crypt_kernel = pbkdf1_loop = clCreateKernel(program[gpu_id], "pbkdf1_loop", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel");
pbkdf1_final = clCreateKernel(program[gpu_id], "pbkdf1_final", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel");
}
static void fmt_ssha_init(struct fmt_main *pFmt)
{
char *kpc;
opencl_init("$JOHN/ssha_opencl_kernel.cl", gpu_id);
// create kernel to execute
crypt_kernel = clCreateKernel(program[gpu_id], "sha1_crypt_kernel", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
if( (kpc = getenv("LWS")) == NULL){
create_clobj(SSHA_NUM_KEYS);
find_best_workgroup();
release_clobj();
}else {
local_work_size = atoi(kpc);
}
if( (kpc = getenv("KPC")) == NULL){
max_keys_per_crypt = SSHA_NUM_KEYS;
create_clobj(SSHA_NUM_KEYS);
} else {
if (atoi(kpc) == 0){
//user chose to die of boredom
max_keys_per_crypt = SSHA_NUM_KEYS;
create_clobj(SSHA_NUM_KEYS);
find_best_kpc();
} else {
max_keys_per_crypt = atoi(kpc);
create_clobj(max_keys_per_crypt);
}
}
printf("Local work size (LWS) %d, Keys per crypt (KPC) %d\n",(int)local_work_size,max_keys_per_crypt);
pFmt->params.max_keys_per_crypt = max_keys_per_crypt;
}
static void init(struct fmt_main *self)
{
char build_opts[96];
size_t gws_limit = 4 << 20;
if (pers_opts.target_enc == UTF_8)
max_len = self->params.plaintext_length =
MIN(125, 3 * PLAINTEXT_LENGTH);
snprintf(build_opts, sizeof(build_opts),
"-D%s -DPLAINTEXT_LENGTH=%u",
cp_id2macro(pers_opts.target_enc), PLAINTEXT_LENGTH);
opencl_init("$JOHN/kernels/oldoffice_kernel.cl", gpu_id, build_opts);
/* create kernels to execute */
oldoffice_utf16 = clCreateKernel(program[gpu_id], "oldoffice_utf16", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
crypt_kernel = oldoffice_md5 =
clCreateKernel(program[gpu_id], "oldoffice_md5", &ret_code);
oldoffice_sha1 =
clCreateKernel(program[gpu_id], "oldoffice_sha1", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
// Initialize openCL tuning (library) for this format.
opencl_init_auto_setup(SEED, 0, NULL,
warn, 3, self, create_clobj, release_clobj,
sizeof(mid_t), gws_limit);
// Auto tune execution from shared/included code.
autotune_run(self, 1, gws_limit, 1000000000);
}
static void init(struct fmt_main *self)
{
char *temp;
if ((temp = getenv("LWS")))
local_work_size = atoi(temp);
else
local_work_size = cpu(device_info[ocl_gpu_id]) ? 1 : 64;
if ((temp = getenv("GWS")))
global_work_size = atoi(temp);
else
global_work_size = MAX_KEYS_PER_CRYPT;
opencl_init("$JOHN/kernels/sha512_kernel.cl", ocl_gpu_id);
gkey = mem_calloc(global_work_size * sizeof(sha512_key));
ghash = mem_calloc(global_work_size * sizeof(sha512_hash));
///Allocate memory on the GPU
mem_in =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for passwords");
mem_out =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for hashes");
mem_binary =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_READ_ONLY, sizeof(uint64_t), NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for binary");
mem_cmp =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_WRITE_ONLY, sizeof(uint32_t), NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for cmp_all result");
///Assign crypt kernel parameters
crypt_kernel = clCreateKernel(program[ocl_gpu_id], KERNEL_NAME, &ret_code);
HANDLE_CLERROR(ret_code,"Error while creating crypt_kernel");
clSetKernelArg(crypt_kernel, 0, sizeof(mem_in), &mem_in);
clSetKernelArg(crypt_kernel, 1, sizeof(mem_out), &mem_out);
///Assign cmp kernel parameters
cmp_kernel = clCreateKernel(program[ocl_gpu_id], CMP_KERNEL_NAME, &ret_code);
HANDLE_CLERROR(ret_code,"Error while creating cmp_kernel");
clSetKernelArg(cmp_kernel, 0, sizeof(mem_binary), &mem_binary);
clSetKernelArg(cmp_kernel, 1, sizeof(mem_out), &mem_out);
clSetKernelArg(cmp_kernel, 2, sizeof(mem_cmp), &mem_cmp);
self->params.max_keys_per_crypt = global_work_size;
if (!local_work_size)
opencl_find_best_workgroup(self);
self->params.min_keys_per_crypt = local_work_size;
if (options.verbosity > 2)
fprintf(stderr, "Local worksize (LWS) %d, Global worksize (GWS) %d\n",(int)local_work_size, (int)global_work_size);
}
static void init(struct fmt_main *_self)
{
self = _self;
opencl_init("$JOHN/kernels/phpass_kernel.cl", gpu_id, NULL);
crypt_kernel = clCreateKernel(program[gpu_id], "phpass", &cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
}
int main( int argc, char** argv )
{
std::cout << ">>Opencl_init" << std::endl;
opencl_init();
std::cout << ">>buildProgramFromFile" << std::endl;
buildProgramFromFile();
std::cout << ">>calc_flow" << std::endl;
calc_flow();
}
static void fmt_ssha_init(struct fmt_main *self)
{
char *temp;
cl_ulong maxsize;
global_work_size = 0;
opencl_init("$JOHN/ssha_kernel.cl", ocl_gpu_id, platform_id);
// create kernel to execute
crypt_kernel = clCreateKernel(program[ocl_gpu_id], "sha1_crypt_kernel", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
HANDLE_CLERROR(clGetKernelWorkGroupInfo(crypt_kernel, devices[ocl_gpu_id], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize), &maxsize, NULL), "Query max work group size");
if ((temp = cfg_get_param(SECTION_OPTIONS, SUBSECTION_OPENCL, LWS_CONFIG)))
local_work_size = atoi(temp);
if ((temp = cfg_get_param(SECTION_OPTIONS, SUBSECTION_OPENCL, GWS_CONFIG)))
global_work_size = atoi(temp);
if ((temp = getenv("LWS")))
local_work_size = atoi(temp);
if ((temp = getenv("GWS")))
global_work_size = atoi(temp);
if (!local_work_size) {
int temp = global_work_size;
local_work_size = maxsize;
global_work_size = global_work_size ? global_work_size : 4 * maxsize;
create_clobj(global_work_size, self);
opencl_find_best_workgroup_limit(self, maxsize);
release_clobj();
global_work_size = temp;
}
if (local_work_size > maxsize) {
fprintf(stderr, "LWS %d is too large for this GPU. Max allowed is %d, using that.\n", (int)local_work_size, (int)maxsize);
local_work_size = maxsize;
}
if (!global_work_size)
find_best_gws(getenv("GWS") == NULL ? 0 : 1, self);
if (global_work_size < local_work_size)
global_work_size = local_work_size;
fprintf(stderr, "Local worksize (LWS) %d, Global worksize (GWS) %d\n", (int)local_work_size, (int)global_work_size);
create_clobj(global_work_size, self);
atexit(release_clobj);
}
static void init(struct fmt_main *self)
{
cl_int cl_error;
global_work_size = MAX_KEYS_PER_CRYPT;
inbuffer =
(zip_password *) malloc(sizeof(zip_password) *
MAX_KEYS_PER_CRYPT);
outbuffer =
(zip_hash *) malloc(sizeof(zip_hash) * MAX_KEYS_PER_CRYPT);
/* Zeroize the lengths in case crypt_all() is called with some keys still
* not set. This may happen during self-tests. */
{
int i;
for (i = 0; i < MAX_KEYS_PER_CRYPT; i++)
inbuffer[i].length = 0;
}
cracked = mem_calloc_tiny(sizeof(*cracked) *
KEYS_PER_CRYPT, MEM_ALIGN_WORD);
//listOpenCLdevices();
opencl_init("$JOHN/zip_kernel.cl", ocl_gpu_id, platform_id);
/// Alocate memory
mem_in =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&cl_error);
HANDLE_CLERROR(cl_error, "Error allocating mem in");
mem_setting =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_READ_ONLY, settingsize,
NULL, &cl_error);
HANDLE_CLERROR(cl_error, "Error allocating mem setting");
mem_out =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&cl_error);
HANDLE_CLERROR(cl_error, "Error allocating mem out");
crypt_kernel = clCreateKernel(program[ocl_gpu_id], "zip", &cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
HANDLE_CLERROR(clSetKernelArg(crypt_kernel, 0, sizeof(mem_in),
&mem_in), "Error while setting mem_in kernel argument");
HANDLE_CLERROR(clSetKernelArg(crypt_kernel, 1, sizeof(mem_out),
&mem_out), "Error while setting mem_out kernel argument");
HANDLE_CLERROR(clSetKernelArg(crypt_kernel, 2, sizeof(mem_setting),
&mem_setting), "Error while setting mem_salt kernel argument");
opencl_find_best_workgroup(self);
atexit(release_all);
}
static void init(struct fmt_main *_self)
{
self = _self;
opencl_init("$JOHN/kernels/pwsafe_kernel.cl", gpu_id, NULL);
init_kernel = clCreateKernel(program[gpu_id], KERNEL_INIT_NAME, &ret_code);
HANDLE_CLERROR(ret_code, "Error while creating init kernel");
crypt_kernel = clCreateKernel(program[gpu_id], KERNEL_RUN_NAME, &ret_code);
HANDLE_CLERROR(ret_code, "Error while creating crypt kernel");
finish_kernel = clCreateKernel(program[gpu_id], KERNEL_FINISH_NAME, &ret_code);
HANDLE_CLERROR(ret_code, "Error while creating finish kernel");
}
static void init(struct fmt_main *_self)
{
char build_opts[64];
self = _self;
snprintf(build_opts, sizeof(build_opts),
"-DKEYLEN=%d -DSALTLEN=%d -DOUTLEN=%d",
(int)sizeof(inbuffer->v),
(int)sizeof(currentsalt.salt),
(int)sizeof(outbuffer->v));
opencl_init("$JOHN/kernels/pbkdf2_hmac_sha1_unsplit_kernel.cl",
gpu_id, build_opts);
crypt_kernel = clCreateKernel(program[gpu_id], "derive_key", &cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
}
static void init(struct fmt_main *self)
{
global_work_size = MAX_KEYS_PER_CRYPT;
opencl_init("$JOHN/sha512_kernel.cl", ocl_gpu_id, platform_id);
///Alocate memory on the GPU
mem_in =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for passwords");
mem_out =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for hashes");
mem_binary =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_READ_ONLY, sizeof(uint64_t), NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for binary");
mem_cmp =
clCreateBuffer(context[ocl_gpu_id], CL_MEM_WRITE_ONLY, sizeof(uint32_t), NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for cmp_all result");
///Assign crypt kernel parameters
crypt_kernel = clCreateKernel(program[ocl_gpu_id], KERNEL_NAME, &ret_code);
HANDLE_CLERROR(ret_code,"Error while creating crypt_kernel");
clSetKernelArg(crypt_kernel, 0, sizeof(mem_in), &mem_in);
clSetKernelArg(crypt_kernel, 1, sizeof(mem_out), &mem_out);
///Assign cmp kernel parameters
cmp_kernel = clCreateKernel(program[ocl_gpu_id], CMP_KERNEL_NAME, &ret_code);
HANDLE_CLERROR(ret_code,"Error while creating cmp_kernel");
clSetKernelArg(cmp_kernel, 0, sizeof(mem_binary), &mem_binary);
clSetKernelArg(cmp_kernel, 1, sizeof(mem_out), &mem_out);
clSetKernelArg(cmp_kernel, 2, sizeof(mem_cmp), &mem_cmp);
opencl_find_best_workgroup(self);
fprintf(stderr, "Global work size = %lld\n",(long long)global_work_size);
atexit(release_all);
}
static void fmt_ssha_init(struct fmt_main *pFmt)
{
char *temp;
opencl_init("$JOHN/ssha_kernel.cl", gpu_id, platform_id);
// create kernel to execute
crypt_kernel = clCreateKernel(program[gpu_id], "sha1_crypt_kernel", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
if ((temp = cfg_get_param(SECTION_OPTIONS, SUBSECTION_OPENCL,
LWS_CONFIG)))
local_work_size = atoi(temp);
if ((temp = getenv("LWS")))
local_work_size = atoi(temp);
if (!local_work_size) {
create_clobj(SSHA_NUM_KEYS);
find_best_workgroup();
release_clobj();
}
if ((temp = cfg_get_param(SECTION_OPTIONS, SUBSECTION_OPENCL,
KPC_CONFIG)))
max_keys_per_crypt = atoi(temp);
else
max_keys_per_crypt = SSHA_NUM_KEYS;
if ((temp = getenv("KPC")))
max_keys_per_crypt = atoi(temp);
if (max_keys_per_crypt) {
create_clobj(max_keys_per_crypt);
} else {
//user chose to die of boredom
max_keys_per_crypt = SSHA_NUM_KEYS;
create_clobj(SSHA_NUM_KEYS);
find_best_kpc();
}
printf("Local work size (LWS) %d, Keys per crypt (KPC) %d\n",(int)local_work_size,max_keys_per_crypt);
pFmt->params.max_keys_per_crypt = max_keys_per_crypt;
}
static void init(struct fmt_main *_self)
{
char build_opts[128];
self = _self;
snprintf(build_opts, sizeof(build_opts),
"-DHASH_LOOPS=%u -DPLAINTEXT_LENGTH=%d -DMAX_SALT_SIZE=%d",
HASH_LOOPS, PLAINTEXT_LENGTH, MAX_SALT_SIZE);
opencl_init("$JOHN/kernels/pbkdf2_hmac_sha512_kernel.cl",
gpu_id, build_opts);
crypt_kernel = clCreateKernel(program[gpu_id], KERNEL_NAME, &cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
split_kernel =
clCreateKernel(program[gpu_id], SPLIT_KERNEL_NAME, &cl_error);
HANDLE_CLERROR(cl_error, "Error creating split kernel");
}
static void init(struct fmt_main *self)
{
char build_opts[64];
if (pers_opts.target_enc == UTF_8) {
max_len = self->params.plaintext_length = 3 * PLAINTEXT_LENGTH;
tests[1].plaintext = "\xC3\xBC"; // German u-umlaut in UTF-8
tests[1].ciphertext = "$mskrb5$$$958db4ddb514a6cc8be1b1ccf82b0191$090408357a6f41852d17f3b4bb4634adfd388db1be64d3fe1a1d75ee4338d2a4aea387e5";
tests[2].plaintext = "\xC3\x9C\xC3\x9C"; // 2x uppercase of them
tests[2].ciphertext = "$mskrb5$$$057cd5cb706b3de18e059912b1f057e3$fe2e561bd4e42767e972835ea99f08582ba526e62a6a2b6f61364e30aca7c6631929d427";
} else {
if (CP_to_Unicode[0xfc] == 0x00fc) {
tests[1].plaintext = "\xFC"; // German u-umlaut in many ISO-8859-x
tests[1].ciphertext = "$mskrb5$$$958db4ddb514a6cc8be1b1ccf82b0191$090408357a6f41852d17f3b4bb4634adfd388db1be64d3fe1a1d75ee4338d2a4aea387e5";
}
if (CP_to_Unicode[0xdc] == 0x00dc) {
tests[2].plaintext = "\xDC\xDC"; // 2x uppercase of them
tests[2].ciphertext = "$mskrb5$$$057cd5cb706b3de18e059912b1f057e3$fe2e561bd4e42767e972835ea99f08582ba526e62a6a2b6f61364e30aca7c6631929d427";
}
}
snprintf(build_opts, sizeof(build_opts),
"-D%s -DPLAINTEXT_LENGTH=%u",
cp_id2macro(pers_opts.target_enc), PLAINTEXT_LENGTH);
opencl_init("$JOHN/kernels/krb5pa-md5_kernel.cl", gpu_id, build_opts);
/* create kernels to execute */
krb5pa_md5_nthash = clCreateKernel(program[gpu_id], "krb5pa_md5_nthash", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
crypt_kernel = clCreateKernel(program[gpu_id], "krb5pa_md5_final", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
//Initialize openCL tuning (library) for this format.
opencl_init_auto_setup(SEED, 0, NULL,
warn, 2, self, create_clobj, release_clobj,
PLAINTEXT_LENGTH, 0);
//Auto tune execution from shared/included code.
autotune_run(self, 1, 0, 200);
}
static void init(struct fmt_main *pFmt)
{
assert(sizeof(hccap_t) == HCCAP_SIZE);
inbuffer =
(wpapsk_password *) malloc(sizeof(wpapsk_password) *
MAX_KEYS_PER_CRYPT);
outbuffer =
(wpapsk_hash *) malloc(sizeof(wpapsk_hash) * MAX_KEYS_PER_CRYPT);
mic = (mic_t *) malloc(sizeof(mic_t) * MAX_KEYS_PER_CRYPT);
//listOpenCLdevices();
opencl_init("$JOHN/wpapsk_kernel.cl", gpu_id, platform_id);
/// Alocate memory
cl_int cl_error;
mem_in =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&cl_error);
HANDLE_CLERROR(cl_error, "Error alocating mem in");
mem_setting =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, settingsize,
NULL, &cl_error);
HANDLE_CLERROR(cl_error, "Error alocating mem setting");
mem_out =
clCreateBuffer(context[gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&cl_error);
HANDLE_CLERROR(cl_error, "Error alocating mem out");
crypt_kernel = clCreateKernel(program[gpu_id], "wpapsk", &cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
clSetKernelArg(crypt_kernel, 0, sizeof(mem_in), &mem_in);
clSetKernelArg(crypt_kernel, 1, sizeof(mem_out), &mem_out);
clSetKernelArg(crypt_kernel, 2, sizeof(mem_setting), &mem_setting);
find_best_workgroup();
atexit(release_all);
}
static void init(struct fmt_main *self)
{
cl_ulong maxsize;
opencl_init("$JOHN/kernels/pbkdf2_hmac_sha512_unsplit_kernel.cl", ocl_gpu_id);
set_lws_gws(DEFAULT_LWS,DEFAULT_GWS);
crypt_kernel = clCreateKernel(program[ocl_gpu_id], KERNEL_NAME, &cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
create_clobj(global_work_size, self);
/* Note: we ask for the kernels' max sizes, not the device's! */
HANDLE_CLERROR(clGetKernelWorkGroupInfo(crypt_kernel,
devices[ocl_gpu_id], CL_KERNEL_WORK_GROUP_SIZE,
sizeof(maxsize), &maxsize, NULL), "Query max workgroup size");
while (local_work_size > maxsize)
local_work_size >>= 1;
self->params.min_keys_per_crypt = local_work_size;
self->params.max_keys_per_crypt = global_work_size;
}
static void init(struct fmt_main *pFmt)
{
opencl_init("$JOHN/cryptmd5_opencl_kernel.cl", gpu_id);
///Alocate memory on the GPU
cl_int cl_error;
mem_in =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&cl_error);
mem_salt =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, saltsize, NULL,
&cl_error);
mem_out =
clCreateBuffer(context[gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&cl_error);
///Assign kernel parameters
crypt_kernel = clCreateKernel(program[gpu_id], KERNEL_NAME, &cl_error);
clSetKernelArg(crypt_kernel, 0, sizeof(mem_in), &mem_in);
clSetKernelArg(crypt_kernel, 1, sizeof(mem_out), &mem_out);
clSetKernelArg(crypt_kernel, 2, sizeof(mem_salt), &mem_salt);
find_best_workgroup();
//atexit(release_all);
}
static void init(struct fmt_main *self)
{
opencl_init("$JOHN/kernels/pwsafe_kernel.cl", gpu_id, NULL);
init_kernel = clCreateKernel(program[gpu_id], KERNEL_INIT_NAME, &ret_code);
HANDLE_CLERROR(ret_code, "Error while creating init kernel");
crypt_kernel = clCreateKernel(program[gpu_id], KERNEL_RUN_NAME, &ret_code);
HANDLE_CLERROR(ret_code, "Error while creating crypt kernel");
finish_kernel = clCreateKernel(program[gpu_id], KERNEL_FINISH_NAME, &ret_code);
HANDLE_CLERROR(ret_code, "Error while creating finish kernel");
//Initialize openCL tuning (library) for this format.
self->methods.crypt_all = crypt_all_benchmark;
opencl_init_auto_setup(SEED, ROUNDS_DEFAULT/8, split_events,
warn, 2, self, create_clobj,
release_clobj, sizeof(pwsafe_pass), 0);
//Auto tune execution from shared/included code.
autotune_run(self, ROUNDS_DEFAULT, 0,
(cpu(device_info[gpu_id]) ? 500000000ULL : 1000000000ULL));
self->methods.crypt_all = crypt_all;
}
static void init(struct fmt_main *self)
{
size_t maxsize;
/* Read LWS/GWS prefs from config or environment */
opencl_get_user_preferences(OCL_CONFIG);
if (!local_work_size)
local_work_size = cpu(device_info[gpu_id]) ? 1 : 64;
if (!global_work_size)
global_work_size = MAX_KEYS_PER_CRYPT;
opencl_init("$JOHN/kernels/sha512_kernel.cl", gpu_id, NULL);
gkey = mem_calloc(global_work_size * sizeof(sha512_key));
ghash = mem_calloc(global_work_size * sizeof(sha512_hash));
///Allocate memory on the GPU
mem_in =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for passwords");
mem_out =
clCreateBuffer(context[gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for hashes");
mem_binary =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, sizeof(uint64_t), NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for binary");
mem_cmp =
clCreateBuffer(context[gpu_id], CL_MEM_WRITE_ONLY, sizeof(uint32_t), NULL,
&ret_code);
HANDLE_CLERROR(ret_code,"Error while allocating memory for cmp_all result");
///Assign crypt kernel parameters
crypt_kernel = clCreateKernel(program[gpu_id], KERNEL_NAME, &ret_code);
HANDLE_CLERROR(ret_code,"Error while creating crypt_kernel");
clSetKernelArg(crypt_kernel, 0, sizeof(mem_in), &mem_in);
clSetKernelArg(crypt_kernel, 1, sizeof(mem_out), &mem_out);
///Assign cmp kernel parameters
cmp_kernel = clCreateKernel(program[gpu_id], CMP_KERNEL_NAME, &ret_code);
HANDLE_CLERROR(ret_code,"Error while creating cmp_kernel");
clSetKernelArg(cmp_kernel, 0, sizeof(mem_binary), &mem_binary);
clSetKernelArg(cmp_kernel, 1, sizeof(mem_out), &mem_out);
clSetKernelArg(cmp_kernel, 2, sizeof(mem_cmp), &mem_cmp);
/* Note: we ask for the kernel's max size, not the device's! */
maxsize = get_kernel_max_lws(gpu_id, crypt_kernel);
if (local_work_size > maxsize) {
local_work_size = maxsize;
global_work_size = (global_work_size + local_work_size - 1) / local_work_size * local_work_size;
}
self->params.max_keys_per_crypt = global_work_size;
if (!local_work_size)
opencl_find_best_workgroup(self);
self->params.min_keys_per_crypt = local_work_size;
if (options.verbosity > 2)
fprintf(stderr, "Local worksize (LWS) %d, Global worksize (GWS) %d\n",(int)local_work_size, (int)global_work_size);
}
size_t select_device(int jtrUniqDevNo, struct fmt_main *fmt) {
cl_int err;
const char *errMsg;
size_t memAllocSz;
active_dev_ctr++;
opencl_init("$JOHN/kernels/pbkdf2_kernel.cl", jtrUniqDevNo, NULL);
globalObj[jtrUniqDevNo].krnl[0] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_preprocess_short", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_preprocess_short FAILED\n");
return 0;
}
globalObj[jtrUniqDevNo].krnl[1] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_preprocess_long", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_preprocess_long FAILED\n");
return 0;
}
globalObj[jtrUniqDevNo].krnl[2] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_iter", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_iter FAILED\n");
return 0;
}
globalObj[jtrUniqDevNo].krnl[3] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_postprocess", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_postprocess FAILED\n");
return 0;
}
errMsg = "Create Buffer FAILED";
memAllocSz = 4 * MAX_KEYS_PER_CRYPT * sizeof(cl_uint);
memAllocSz = memAllocSz < get_max_mem_alloc_size(jtrUniqDevNo) ? memAllocSz : get_max_mem_alloc_size(jtrUniqDevNo) / 4 * 4;
globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_ONLY, memAllocSz, NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu == (cl_mem)0)
HANDLE_CLERROR(err,errMsg );
globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_ONLY, (MAX_SALT_LENGTH / 2 + 1) * sizeof(cl_uint), NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_WRITE_ONLY, memAllocSz, NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
memAllocSz = MAX_KEYS_PER_CRYPT * sizeof(temp_buf);
memAllocSz = memAllocSz < get_max_mem_alloc_size(jtrUniqDevNo) ? memAllocSz : get_max_mem_alloc_size(jtrUniqDevNo) / 4 * 4;
globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_WRITE, memAllocSz, NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
memAllocSz = 5 * MAX_KEYS_PER_CRYPT * sizeof(cl_uint);
memAllocSz = memAllocSz < get_max_mem_alloc_size(jtrUniqDevNo) ? memAllocSz : get_max_mem_alloc_size(jtrUniqDevNo) / 4 * 4;
globalObj[jtrUniqDevNo].gpu_buffer.hmac_sha1_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_WRITE, memAllocSz, NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu), "Set Kernel 0 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 1, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu), "Set Kernel 0 Arg 1 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 3, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 0 Arg 3 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[1], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu), "Set Kernel 1 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[1], 1, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 1 Arg 1 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[1], 2, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.hmac_sha1_gpu), "Set Kernel 1 Arg 2 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[2], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 2 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[3], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 3 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[3], 1, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu), "Set Kernel 3 Arg 1 :FAILED");
if (!local_work_size)
find_best_workgroup(jtrUniqDevNo, quick_bechmark(jtrUniqDevNo));
else {
size_t maxsize, maxsize2;
globalObj[jtrUniqDevNo].lws = local_work_size;
// Obey limits
HANDLE_CLERROR(clGetKernelWorkGroupInfo(globalObj[jtrUniqDevNo].krnl[0], devices[jtrUniqDevNo], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize), &maxsize, NULL), "Error querying max LWS");
HANDLE_CLERROR(clGetKernelWorkGroupInfo(globalObj[jtrUniqDevNo].krnl[1], devices[jtrUniqDevNo], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize2), &maxsize2, NULL), "Error querying max LWS");
if (maxsize2 > maxsize)
maxsize = maxsize2;
HANDLE_CLERROR(clGetKernelWorkGroupInfo(globalObj[jtrUniqDevNo].krnl[2], devices[jtrUniqDevNo], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize2), &maxsize2, NULL), "Error querying max LWS");
if (maxsize2 > maxsize)
maxsize = maxsize2;
HANDLE_CLERROR(clGetKernelWorkGroupInfo(globalObj[jtrUniqDevNo].krnl[3], devices[jtrUniqDevNo], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize2), &maxsize2, NULL), "Error querying max LWS");
if (maxsize2 > maxsize)
maxsize = maxsize2;
while (globalObj[jtrUniqDevNo].lws > maxsize)
globalObj[jtrUniqDevNo].lws /= 2;
if (options.verbosity > 3)
fprintf(stderr, "Local worksize (LWS) forced to "Zu"\n", globalObj[jtrUniqDevNo].lws);
globalObj[jtrUniqDevNo].exec_time_inv = 1;
}
if (!global_work_size)
find_best_gws(jtrUniqDevNo, fmt);
else {
if (options.verbosity > 3)
fprintf(stderr, "Global worksize (GWS) forced to "Zu"\n", global_work_size);
fmt -> params.max_keys_per_crypt = global_work_size;
fmt -> params.min_keys_per_crypt = max_lws();
}
return globalObj[jtrUniqDevNo].lws;
}
static void init(struct fmt_main *self)
{
#ifdef CL_VERSION_1_0
char *temp;
cl_ulong maxsize;
global_work_size = 0;
opencl_init("$JOHN/rar_kernel.cl", ocl_gpu_id, platform_id);
// create kernel to execute
crypt_kernel = clCreateKernel(program[ocl_gpu_id], "SetCryptKeys", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
/* We mimic the lengths of cRARk for comparisons */
if (get_device_type(ocl_gpu_id) == CL_DEVICE_TYPE_GPU) {
#ifndef DEBUG
self->params.benchmark_comment = " (6 characters)";
#endif
self->params.tests = gpu_tests;
#if defined(DEBUG) && !defined(ALWAYS_OPENCL)
fprintf(stderr, "Note: will use CPU for some self-tests, and Single mode.\n");
#endif
}
if ((temp = cfg_get_param(SECTION_OPTIONS, SUBSECTION_OPENCL, LWS_CONFIG)))
local_work_size = atoi(temp);
if ((temp = cfg_get_param(SECTION_OPTIONS, SUBSECTION_OPENCL, GWS_CONFIG)))
global_work_size = atoi(temp);
if ((temp = getenv("LWS")))
local_work_size = atoi(temp);
if ((temp = getenv("GWS")))
global_work_size = atoi(temp);
/* Note: we ask for this kernel's max size, not the device's! */
HANDLE_CLERROR(clGetKernelWorkGroupInfo(crypt_kernel, devices[ocl_gpu_id], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize), &maxsize, NULL), "Query max work group size");
#ifdef DEBUG
fprintf(stderr, "Max allowed local work size %d\n", (int)maxsize);
#endif
if (!local_work_size) {
if (get_device_type(ocl_gpu_id) == CL_DEVICE_TYPE_CPU) {
if (get_platform_vendor_id(platform_id) == INTEL)
local_work_size = 8;
else
local_work_size = 1;
} else {
local_work_size = 64;
}
}
if (local_work_size > maxsize) {
fprintf(stderr, "LWS %d is too large for this GPU. Max allowed is %d, using that.\n", (int)local_work_size, (int)maxsize);
local_work_size = maxsize;
}
if (!global_work_size)
find_best_gws(temp == NULL ? 0 : 1);
if (global_work_size < local_work_size)
global_work_size = local_work_size;
fprintf(stderr, "Local worksize (LWS) %d, Global worksize (GWS) %d\n", (int)local_work_size, (int)global_work_size);
create_clobj(global_work_size);
#ifdef DEBUG
{
cl_ulong loc_mem_size;
HANDLE_CLERROR(clGetKernelWorkGroupInfo(crypt_kernel, devices[ocl_gpu_id], CL_KERNEL_LOCAL_MEM_SIZE, sizeof(loc_mem_size), &loc_mem_size, NULL), "Query local memory usage");
fprintf(stderr, "Kernel using %lu bytes of local memory out of %lu available\n", loc_mem_size, get_local_memory_size(ocl_gpu_id));
}
#endif
atexit(release_clobj);
*mkpc = VF * global_work_size;
#endif /* OpenCL */
#if defined (_OPENMP)
omp_t = omp_get_max_threads();
self->params.min_keys_per_crypt *= omp_t;
#ifndef CL_VERSION_1_0 /* OpenCL gets to decide */
*mkpc = omp_t * OMP_SCALE * MAX_KEYS_PER_CRYPT;
#endif
init_locks();
#endif /* _OPENMP */
if (options.utf8)
self->params.plaintext_length = PLAINTEXT_LENGTH * 3;
unpack_data = mem_calloc_tiny(sizeof(unpack_data_t) * omp_t, MEM_ALIGN_WORD);
cracked = mem_calloc_tiny(sizeof(*cracked) * *mkpc, MEM_ALIGN_WORD);
#ifndef CL_VERSION_1_0
saved_key = mem_calloc_tiny(UNICODE_LENGTH * *mkpc, MEM_ALIGN_NONE);
saved_len = mem_calloc_tiny(sizeof(*saved_len) * *mkpc, MEM_ALIGN_WORD);
saved_salt = mem_calloc_tiny(8, MEM_ALIGN_NONE);
aes_key = mem_calloc_tiny(16 * *mkpc, MEM_ALIGN_NONE);
aes_iv = mem_calloc_tiny(16 * *mkpc, MEM_ALIGN_NONE);
#endif
/* OpenSSL init */
init_aesni();
SSL_load_error_strings();
SSL_library_init();
OpenSSL_add_all_algorithms();
#ifndef __APPLE__
atexit(openssl_cleanup);
#endif
/* CRC-32 table init, do it before we start multithreading */
{
CRC32_t crc;
CRC32_Init(&crc);
}
}
