- /**
+/**
* Authors.....: Jens Steube <jens.steube@gmail.com>
* Gabriele Gristina <matrix@hashcat.net>
* magnum <john.magnum@hushmail.com>
* License.....: MIT
*/
-#ifdef OSX
+#ifdef DARWIN
#include <stdio.h>
#endif
#define NVIDIA_SPIN_DAMP 100
#define GPU_TEMP_DISABLE 0
#define GPU_TEMP_ABORT 90
-#define GPU_TEMP_RETAIN 65
+#define GPU_TEMP_RETAIN 75
#define WORKLOAD_PROFILE 2
#define KERNEL_ACCEL 0
#define KERNEL_LOOPS 0
* flush
*/
- #ifdef _WIN
- fputc ('\r', out);
- fputc ('\n', out);
- #endif
-
- #ifdef _POSIX
- fputc ('\n', out);
- #endif
-
+ fputs (EOL, out);
fflush (out);
}
if (data.devices_status == STATUS_CRACKED) break;
if (data.devices_status == STATUS_ABORTED) break;
if (data.devices_status == STATUS_QUIT) break;
+ if (data.devices_status == STATUS_BYPASS) break;
/**
* speed
if (data.devices_status == STATUS_CRACKED) break;
if (data.devices_status == STATUS_ABORTED) break;
if (data.devices_status == STATUS_QUIT) break;
+ if (data.devices_status == STATUS_BYPASS) break;
/**
* result
return (p);
}
- while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ while (data.shutdown_inner == 0)
{
hc_sleep (sleep_time);
time (&runtime_cur);
- int runtime_left = data.runtime_start + data.runtime - runtime_cur;
+ int runtime_left = data.proc_start + data.runtime - runtime_cur;
if (runtime_left <= 0)
{
if (status_left == 0)
{
- //hc_thread_mutex_lock (mux_display);
+ hc_thread_mutex_lock (mux_display);
if (data.quiet == 0) clear_prompt ();
if (data.quiet == 0) log_info ("");
- //hc_thread_mutex_unlock (mux_display);
+ hc_thread_mutex_unlock (mux_display);
status_left = data.status_timer;
}
uint check_left = outfile_check_timer; // or 1 if we want to check it at startup
- while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ while (data.shutdown_inner == 0)
{
hc_sleep (1);
const uint attack_kern = data.attack_kern;
- while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ while ((data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
{
hc_thread_mutex_lock (mux_dispatcher);
if (attack_mode == ATTACK_MODE_BF)
{
- while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ while ((data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
{
const uint work = get_work (device_param, -1);
u64 words_cur = 0;
- while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ while ((data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
{
u64 words_off = 0;
u64 words_fin = 0;
data.status = status;
}
- uint i_threads_cnt = 0;
+ uint outer_threads_cnt = 0;
- hc_thread_t *i_threads = (hc_thread_t *) mycalloc (10, sizeof (hc_thread_t));
+ hc_thread_t *outer_threads = (hc_thread_t *) mycalloc (10, sizeof (hc_thread_t));
+
+ data.shutdown_outer = 0;
if (keyspace == 0 && benchmark == 0 && stdout_flag == 0)
{
if ((data.wordlist_mode == WL_MODE_FILE) || (data.wordlist_mode == WL_MODE_MASK))
{
- hc_thread_create (i_threads[i_threads_cnt], thread_keypress, NULL);
+ hc_thread_create (outer_threads[outer_threads_cnt], thread_keypress, NULL);
- i_threads_cnt++;
+ outer_threads_cnt++;
}
}
if (cpu_rule_to_kernel_rule (rule_buf, rule_len, &kernel_rules_buf[kernel_rules_cnt]) == -1)
{
- log_info ("WARNING: Cannot convert rule for use on device in file %s on line %u: %s", rp_file, rule_line, rule_buf);
+ log_info ("WARNING: Cannot convert rule for use on OpenCL device in file %s on line %u: %s", rp_file, rule_line, rule_buf);
memset (&kernel_rules_buf[kernel_rules_cnt], 0, sizeof (kernel_rule_t)); // needs to be cleared otherwise we could have some remaining data
/* its so slow
if (rulefind (&kernel_rules_buf[kernel_rules_cnt], kernel_rules_buf, kernel_rules_cnt, sizeof (kernel_rule_t), sort_by_kernel_rule))
{
- log_info ("Duplicate rule for use on device in file %s in line %u: %s", rp_file, rule_line, rule_buf);
+ log_info ("Duplicate rule for use on OpenCL device in file %s in line %u: %s", rp_file, rule_line, rule_buf);
continue;
}
* generate NOP rules
*/
- if (kernel_rules_cnt == 0)
+ if ((rp_files_cnt == 0) && (rp_gen == 0))
{
kernel_rules_buf = (kernel_rule_t *) mymalloc (sizeof (kernel_rule_t));
data.kernel_rules_cnt = kernel_rules_cnt;
data.kernel_rules_buf = kernel_rules_buf;
+ if (kernel_rules_cnt == 0)
+ {
+ log_error ("ERROR: No valid rules left");
+
+ return (-1);
+ }
+
/**
* OpenCL platforms: detect
*/
device_param->device_name_chksum = device_name_chksum;
- // device_processor_cores
+ // vendor specific
if (device_param->device_type & CL_DEVICE_TYPE_GPU)
{
}
}
- // device_processor_cores
-
- if (device_type & CL_DEVICE_TYPE_CPU)
- {
- cl_uint device_processor_cores = 1;
-
- device_param->device_processor_cores = device_processor_cores;
- }
-
if (device_type & CL_DEVICE_TYPE_GPU)
{
- if (device_vendor_id == VENDOR_ID_AMD)
- {
- cl_uint device_processor_cores = 0;
-
- #define CL_DEVICE_WAVEFRONT_WIDTH_AMD 0x4043
-
- hc_clGetDeviceInfo (data.ocl, device_param->device, CL_DEVICE_WAVEFRONT_WIDTH_AMD, sizeof (device_processor_cores), &device_processor_cores, NULL);
-
- device_param->device_processor_cores = device_processor_cores;
- }
- else if (device_vendor_id == VENDOR_ID_NV)
+ if (device_vendor_id == VENDOR_ID_NV)
{
cl_uint kernel_exec_timeout = 0;
device_param->kernel_exec_timeout = kernel_exec_timeout;
- cl_uint device_processor_cores = 0;
-
- #define CL_DEVICE_WARP_SIZE_NV 0x4003
-
- hc_clGetDeviceInfo (data.ocl, device_param->device, CL_DEVICE_WARP_SIZE_NV, sizeof (device_processor_cores), &device_processor_cores, NULL);
-
- device_param->device_processor_cores = device_processor_cores;
-
cl_uint sm_minor = 0;
cl_uint sm_major = 0;
device_param->nvidia_spin_damp /= 100;
}
- else
- {
- cl_uint device_processor_cores = 1;
-
- device_param->device_processor_cores = device_processor_cores;
- }
}
// display results
if (hm_NVML_nvmlDeviceGetFanSpeed (data.hm_nvml, 0, hm_adapters_nvml[i].nvml, &speed) == NVML_SUCCESS) hm_adapters_nvml[i].fan_get_supported = 1;
- hm_NVML_nvmlDeviceSetComputeMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_COMPUTEMODE_EXCLUSIVE_PROCESS);
-
- hm_NVML_nvmlDeviceSetGpuOperationMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_GOM_ALL_ON);
+ // doesn't seem to create any advantages
+ //hm_NVML_nvmlDeviceSetComputeMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_COMPUTEMODE_EXCLUSIVE_PROCESS);
+ //hm_NVML_nvmlDeviceSetGpuOperationMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_GOM_ALL_ON);
}
}
}
* OpenCL devices: allocate buffer for device specific information
*/
- int *temp_retain_fanspeed_value = (int *) mycalloc (data.devices_cnt, sizeof (int));
- int *temp_retain_fanpolicy_value = (int *) mycalloc (data.devices_cnt, sizeof (int));
-
ADLOD6MemClockState *od_clock_mem_status = (ADLOD6MemClockState *) mycalloc (data.devices_cnt, sizeof (ADLOD6MemClockState));
int *od_power_control_status = (int *) mycalloc (data.devices_cnt, sizeof (int));
const char *device_name_chksum = device_param->device_name_chksum;
const u32 device_processors = device_param->device_processors;
- const u32 device_processor_cores = device_param->device_processor_cores;
/**
* create context for each device
if (hash_mode == 3000) kernel_threads = 64; // DES
if (hash_mode == 3200) kernel_threads = 8; // Blowfish
if (hash_mode == 7500) kernel_threads = 64; // RC4
+ if (hash_mode == 8900) kernel_threads = 32; // scrypt
if (hash_mode == 9000) kernel_threads = 8; // Blowfish
+ if (hash_mode == 9300) kernel_threads = 32; // scrypt
if (hash_mode == 9700) kernel_threads = 64; // RC4
if (hash_mode == 9710) kernel_threads = 64; // RC4
if (hash_mode == 9800) kernel_threads = 64; // RC4
}
}
+ device_param->kernel_accel_min = 1;
+ device_param->kernel_accel_max = 8;
+
for (uint tmto = tmto_start; tmto < tmto_stop; tmto++)
{
// TODO: in theory the following calculation needs to be done per salt, not global
size_scryptV /= 1 << tmto;
- size_scryptV *= device_processors * device_processor_cores;
+ size_scryptV *= device_param->device_processors * device_param->kernel_threads * device_param->kernel_accel_max;
if (size_scryptV > device_param->device_maxmem_alloc)
{
for (uint salts_pos = 0; salts_pos < data.salts_cnt; salts_pos++)
{
data.salts_buf[salts_pos].scrypt_tmto = tmto;
- data.salts_buf[salts_pos].scrypt_phy = device_processors * device_processor_cores;
+ data.salts_buf[salts_pos].scrypt_phy = device_param->device_processors * device_param->kernel_threads * device_param->kernel_accel_max;
}
break;
snprintf (build_opts, sizeof (build_opts) - 1, "-I \"%s\"", cpath_real);
- myfree (cpath_real);
-
#else
snprintf (cpath, sizeof (cpath) - 1, "%s/OpenCL/", shared_dir);
snprintf (build_opts, sizeof (build_opts) - 1, "-I %s", cpath_real);
- myfree (cpath_real);
-
#endif
+ // include check
+ // this test needs to be done manually because of osx opencl runtime
+ // if there's a problem with permission, its not reporting back and erroring out silently
+
+ #define files_cnt 15
+
+ const char *files_names[files_cnt] =
+ {
+ "inc_cipher_aes256.cl",
+ "inc_cipher_serpent256.cl",
+ "inc_cipher_twofish256.cl",
+ "inc_common.cl",
+ "inc_comp_multi_bs.cl",
+ "inc_comp_multi.cl",
+ "inc_comp_single_bs.cl",
+ "inc_comp_single.cl",
+ "inc_hash_constants.h",
+ "inc_hash_functions.cl",
+ "inc_rp.cl",
+ "inc_rp.h",
+ "inc_simd.cl",
+ "inc_types.cl",
+ "inc_vendor.cl",
+ };
+
+ for (int i = 0; i < files_cnt; i++)
+ {
+ char path[1024] = { 0 };
+
+ snprintf (path, sizeof (path) - 1, "%s/%s", cpath_real, files_names[i]);
+
+ FILE *fd = fopen (path, "r");
+
+ if (fd == NULL)
+ {
+ log_error ("ERROR: %s: fopen(): %s", path, strerror (errno));
+
+ return -1;
+ }
+
+ char buf[1];
+
+ size_t n = fread (buf, 1, 1, fd);
+
+ if (n != 1)
+ {
+ log_error ("ERROR: %s: fread(): %s", path, strerror (errno));
+
+ return -1;
+ }
+
+ fclose (fd);
+ }
+
+ myfree (cpath_real);
+
// we don't have sm_* on vendors not NV but it doesn't matter
char build_opts_new[1024] = { 0 };
- snprintf (build_opts_new, sizeof (build_opts_new) - 1, "%s -D VENDOR_ID=%u -D CUDA_ARCH=%d -D VECT_SIZE=%u -D DEVICE_TYPE=%u -D KERN_TYPE=%u -D _unroll -cl-std=CL1.1", build_opts, device_param->device_vendor_id, (device_param->sm_major * 100) + device_param->sm_minor, device_param->vector_width, (u32) device_param->device_type, kern_type);
+ snprintf (build_opts_new, sizeof (build_opts_new) - 1, "%s -D VENDOR_ID=%u -D CUDA_ARCH=%d -D VECT_SIZE=%u -D DEVICE_TYPE=%u -D DGST_R0=%u -D DGST_R1=%u -D DGST_R2=%u -D DGST_R3=%u -D DGST_ELEM=%u -D KERN_TYPE=%u -D _unroll -cl-std=CL1.1", build_opts, device_param->device_vendor_id, (device_param->sm_major * 100) + device_param->sm_minor, device_param->vector_width, (u32) device_param->device_type, data.dgst_pos0, data.dgst_pos1, data.dgst_pos2, data.dgst_pos3, data.dgst_size / 4, kern_type);
strncpy (build_opts, build_opts_new, sizeof (build_opts));
const int fanspeed = hm_get_fanspeed_with_device_id (device_id);
const int fanpolicy = hm_get_fanpolicy_with_device_id (device_id);
- temp_retain_fanspeed_value[device_id] = fanspeed;
- temp_retain_fanpolicy_value[device_id] = fanpolicy;
-
// we also set it to tell the OS we take control over the fan and it's automatic controller
// if it was set to automatic. we do not control user-defined fanspeeds.
* status and monitor threads
*/
- if ((data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ if ((data.devices_status != STATUS_BYPASS) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
{
data.devices_status = STATUS_STARTING;
}
- uint ni_threads_cnt = 0;
-
- hc_thread_t *ni_threads = (hc_thread_t *) mycalloc (10, sizeof (hc_thread_t));
+ uint inner_threads_cnt = 0;
- if (keyspace == 0 && benchmark == 0 && stdout_flag == 0)
- {
- hc_thread_create (ni_threads[ni_threads_cnt], thread_monitor, NULL);
+ hc_thread_t *inner_threads = (hc_thread_t *) mycalloc (10, sizeof (hc_thread_t));
- ni_threads_cnt++;
- }
+ data.shutdown_inner = 0;
/**
* Outfile remove
if (keyspace == 0 && benchmark == 0 && stdout_flag == 0)
{
+ hc_thread_create (inner_threads[inner_threads_cnt], thread_monitor, NULL);
+
+ inner_threads_cnt++;
+
if (outfile_check_timer != 0)
{
if (data.outfile_check_directory != NULL)
!((hash_mode >= 13700) && (hash_mode <= 13799)) &&
(hash_mode != 9000))
{
- hc_thread_create (ni_threads[ni_threads_cnt], thread_outfile_remove, NULL);
+ hc_thread_create (inner_threads[inner_threads_cnt], thread_outfile_remove, NULL);
- ni_threads_cnt++;
+ inner_threads_cnt++;
}
else
{
logfile_sub_msg ("START");
- data.devices_status = STATUS_INIT;
+ if ((data.devices_status != STATUS_BYPASS) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ {
+ data.devices_status = STATUS_INIT;
+ }
memset (data.words_progress_done, 0, data.salts_cnt * sizeof (u64));
memset (data.words_progress_rejected, 0, data.salts_cnt * sizeof (u64));
hc_thread_t *c_threads = (hc_thread_t *) mycalloc (data.devices_cnt, sizeof (hc_thread_t));
- data.devices_status = STATUS_AUTOTUNE;
+ if ((data.devices_status != STATUS_BYPASS) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ {
+ data.devices_status = STATUS_AUTOTUNE;
+ }
for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
{
* create cracker threads
*/
- data.devices_status = STATUS_RUNNING;
+ if ((data.devices_status != STATUS_BYPASS) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
+ {
+ data.devices_status = STATUS_RUNNING;
+ }
if (initial_restore_done == 0)
{
local_free (c_threads);
- if ((data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT) && (data.devices_status != STATUS_BYPASS))
+ if ((data.devices_status != STATUS_BYPASS) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
{
data.devices_status = STATUS_EXHAUSTED;
}
global_free (subid);
- // from this point we handle bypass as exhausted
+ // from this point we handle bypass as running
if (data.devices_status == STATUS_BYPASS)
{
- data.devices_status = STATUS_EXHAUSTED;
+ data.devices_status = STATUS_RUNNING;
+ }
+
+ // and overwrite benchmark aborts as well
+
+ if (data.benchmark == 1)
+ {
+ if (data.devices_status == STATUS_ABORTED)
+ {
+ data.devices_status = STATUS_RUNNING;
+ }
}
// finalize task
}
}
- // wait for non-interactive threads
+ // wait for inner threads
- for (uint thread_idx = 0; thread_idx < ni_threads_cnt; thread_idx++)
+ data.shutdown_inner = 1;
+
+ for (uint thread_idx = 0; thread_idx < inner_threads_cnt; thread_idx++)
{
- hc_thread_wait (1, &ni_threads[thread_idx]);
+ hc_thread_wait (1, &inner_threads[thread_idx]);
}
- local_free (ni_threads);
+ local_free (inner_threads);
// we dont need restore file anymore
if (data.restore_disable == 0)
#ifdef HAVE_HWMON
if (gpu_temp_disable == 0)
{
- if (gpu_temp_retain != 0) // VENDOR_ID_AMD is implied here
+ if (gpu_temp_retain != 0)
{
hc_thread_mutex_lock (mux_adl);
if (data.hm_device[device_id].fan_set_supported == 1)
{
- int fanspeed = temp_retain_fanspeed_value[device_id];
- int fanpolicy = temp_retain_fanpolicy_value[device_id];
+ int rc = -1;
- if (fanpolicy == 1)
+ if (device_param->device_vendor_id == VENDOR_ID_AMD)
{
- int rc = -1;
-
- if (device_param->device_vendor_id == VENDOR_ID_AMD)
- {
- rc = hm_set_fanspeed_with_device_id_adl (device_id, fanspeed, 0);
- }
- else if (device_param->device_vendor_id == VENDOR_ID_NV)
- {
- #ifdef LINUX
- rc = set_fan_control (data.hm_xnvctrl, data.hm_device[device_id].xnvctrl, NV_CTRL_GPU_COOLER_MANUAL_CONTROL_FALSE);
- #endif
-
- #ifdef WIN
- rc = hm_set_fanspeed_with_device_id_nvapi (device_id, fanspeed, fanpolicy);
- #endif
- }
+ rc = hm_set_fanspeed_with_device_id_adl (device_id, 100, 0);
+ }
+ else if (device_param->device_vendor_id == VENDOR_ID_NV)
+ {
+ #ifdef LINUX
+ rc = set_fan_control (data.hm_xnvctrl, data.hm_device[device_id].xnvctrl, NV_CTRL_GPU_COOLER_MANUAL_CONTROL_FALSE);
+ #endif
- if (rc == -1) log_info ("WARNING: Failed to restore default fan speed and policy for device #%", device_id + 1);
+ #ifdef WIN
+ rc = hm_set_fanspeed_with_device_id_nvapi (device_id, 100, 0);
+ #endif
}
+
+ if (rc == -1) log_info ("WARNING: Failed to restore default fan speed and policy for device #%", device_id + 1);
}
}
// reset power tuning
- if (powertune_enable == 1) // VENDOR_ID_AMD is implied here
+ if (powertune_enable == 1)
{
hc_thread_mutex_lock (mux_adl);
local_free (bitmap_s2_d);
#ifdef HAVE_HWMON
- local_free (temp_retain_fanspeed_value);
local_free (od_clock_mem_status);
local_free (od_power_control_status);
local_free (nvml_power_limit);
if (data.devices_status == STATUS_QUIT) break;
}
- // wait for interactive threads
+ // wait for outer threads
+
+ data.shutdown_outer = 1;
- for (uint thread_idx = 0; thread_idx < i_threads_cnt; thread_idx++)
+ for (uint thread_idx = 0; thread_idx < outer_threads_cnt; thread_idx++)
{
- hc_thread_wait (1, &i_threads[thread_idx]);
+ hc_thread_wait (1, &outer_threads[thread_idx]);
}
- local_free (i_threads);
+ local_free (outer_threads);
// destroy others mutex