" --bitmap-min=NUM Minimum number of bits allowed for bitmaps",
" --bitmap-max=NUM Maximum number of bits allowed for bitmaps",
" --cpu-affinity=STR Locks to CPU devices, separate with comma",
+ " --opencl-platforms=STR OpenCL platforms to use, separate with comma",
" -d, --opencl-devices=STR OpenCL devices to use, separate with comma",
" --opencl-device-types=STR OpenCL device-types to use, separate with comma, see references below",
- " --opencl-platform=NUM OpenCL platform to use, in case multiple platforms are present",
" -w, --workload-profile=NUM Enable a specific workload profile, see references below",
" -n, --kernel-accel=NUM Workload tuning: 1, 8, 40, 80, 160",
" -u, --kernel-loops=NUM Workload fine-tuning: 8 - 1024",
{
hc_thread_mutex_lock (mux_adl);
- for (uint i = 0; i < data.devices_cnt; i++)
+ for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
{
+ hc_device_param_t *device_param = &data.devices_param[device_id];
+
#define HM_STR_BUF_SIZE 255
- if (data.hm_device[i].fan_supported == 1)
+ if (data.hm_device[device_id].fan_supported == 1)
{
char utilization[HM_STR_BUF_SIZE];
char temperature[HM_STR_BUF_SIZE];
char fanspeed[HM_STR_BUF_SIZE];
- hm_device_val_to_str ((char *) utilization, HM_STR_BUF_SIZE, "%", hm_get_utilization_with_device_id (i));
- hm_device_val_to_str ((char *) temperature, HM_STR_BUF_SIZE, "c", hm_get_temperature_with_device_id (i));
+ hm_device_val_to_str ((char *) utilization, HM_STR_BUF_SIZE, "%", hm_get_utilization_with_device_id (device_id));
+ hm_device_val_to_str ((char *) temperature, HM_STR_BUF_SIZE, "c", hm_get_temperature_with_device_id (device_id));
- if (data.vendor_id == VENDOR_ID_AMD)
+ if (device_param->vendor_id == VENDOR_ID_AMD)
{
- hm_device_val_to_str ((char *) fanspeed, HM_STR_BUF_SIZE, "%", hm_get_fanspeed_with_device_id (i));
+ hm_device_val_to_str ((char *) fanspeed, HM_STR_BUF_SIZE, "%", hm_get_fanspeed_with_device_id (device_id));
}
-
- if (data.vendor_id == VENDOR_ID_NV)
+ else if (device_param->vendor_id == VENDOR_ID_NV)
{
#ifdef LINUX
- hm_device_val_to_str ((char *) fanspeed, HM_STR_BUF_SIZE, "%", hm_get_fanspeed_with_device_id (i));
+ hm_device_val_to_str ((char *) fanspeed, HM_STR_BUF_SIZE, "%", hm_get_fanspeed_with_device_id (device_id));
#else
- hm_device_val_to_str ((char *) fanspeed, HM_STR_BUF_SIZE, "rpm", hm_get_fanspeed_with_device_id (i));
+ hm_device_val_to_str ((char *) fanspeed, HM_STR_BUF_SIZE, "rpm", hm_get_fanspeed_with_device_id (device_id));
#endif
}
- log_info ("HWMon.GPU.#%d...: %s Util, %s Temp, %s Fan", i + 1, utilization, temperature, fanspeed);
+ log_info ("HWMon.GPU.#%d...: %s Util, %s Temp, %s Fan", device_id + 1, utilization, temperature, fanspeed);
}
else
{
char utilization[HM_STR_BUF_SIZE];
char temperature[HM_STR_BUF_SIZE];
- hm_device_val_to_str ((char *) utilization, HM_STR_BUF_SIZE, "%", hm_get_utilization_with_device_id (i));
- hm_device_val_to_str ((char *) temperature, HM_STR_BUF_SIZE, "c", hm_get_temperature_with_device_id (i));
+ hm_device_val_to_str ((char *) utilization, HM_STR_BUF_SIZE, "%", hm_get_utilization_with_device_id (device_id));
+ hm_device_val_to_str ((char *) temperature, HM_STR_BUF_SIZE, "c", hm_get_temperature_with_device_id (device_id));
- log_info ("HWMon.GPU.#%d...: %s Util, %s Temp, N/A Fan", i + 1, utilization, temperature);
+ log_info ("HWMon.GPU.#%d...: %s Util, %s Temp, N/A Fan", device_id + 1, utilization, temperature);
}
}
snprintf (source_file, 255, "%s/OpenCL/m%05d.cl", shared_dir, (int) kern_type);
}
-static void generate_cached_kernel_filename (const uint attack_exec, const uint attack_kern, const uint kern_type, char *profile_dir, char *device_name_chksum, int vendor_id, char *cached_file)
+static void generate_cached_kernel_filename (const uint attack_exec, const uint attack_kern, const uint kern_type, char *profile_dir, char *device_name_chksum, char *cached_file)
{
if (attack_exec == ATTACK_EXEC_INSIDE_KERNEL)
{
if (attack_kern == ATTACK_KERN_STRAIGHT)
- snprintf (cached_file, 255, "%s/kernels/%d/m%05d_a0.%s.kernel", profile_dir, vendor_id, (int) kern_type, device_name_chksum);
+ snprintf (cached_file, 255, "%s/kernels/m%05d_a0.%s.kernel", profile_dir, (int) kern_type, device_name_chksum);
else if (attack_kern == ATTACK_KERN_COMBI)
- snprintf (cached_file, 255, "%s/kernels/%d/m%05d_a1.%s.kernel", profile_dir, vendor_id, (int) kern_type, device_name_chksum);
+ snprintf (cached_file, 255, "%s/kernels/m%05d_a1.%s.kernel", profile_dir, (int) kern_type, device_name_chksum);
else if (attack_kern == ATTACK_KERN_BF)
- snprintf (cached_file, 255, "%s/kernels/%d/m%05d_a3.%s.kernel", profile_dir, vendor_id, (int) kern_type, device_name_chksum);
+ snprintf (cached_file, 255, "%s/kernels/m%05d_a3.%s.kernel", profile_dir, (int) kern_type, device_name_chksum);
}
else
{
- snprintf (cached_file, 255, "%s/kernels/%d/m%05d.%s.kernel", profile_dir, vendor_id, (int) kern_type, device_name_chksum);
+ snprintf (cached_file, 255, "%s/kernels/m%05d.%s.kernel", profile_dir, (int) kern_type, device_name_chksum);
}
}
}
}
-static void generate_cached_kernel_mp_filename (const uint opti_type, const uint opts_type, char *profile_dir, char *device_name_chksum, int vendor_id, char *cached_file)
+static void generate_cached_kernel_mp_filename (const uint opti_type, const uint opts_type, char *profile_dir, char *device_name_chksum, char *cached_file)
{
if ((opti_type & OPTI_TYPE_BRUTE_FORCE) && (opts_type & OPTS_TYPE_PT_GENERATE_BE))
{
- snprintf (cached_file, 255, "%s/kernels/%d/markov_be.%s.kernel", profile_dir, vendor_id, device_name_chksum);
+ snprintf (cached_file, 255, "%s/kernels/markov_be.%s.kernel", profile_dir, device_name_chksum);
}
else
{
- snprintf (cached_file, 255, "%s/kernels/%d/markov_le.%s.kernel", profile_dir, vendor_id, device_name_chksum);
+ snprintf (cached_file, 255, "%s/kernels/markov_le.%s.kernel", profile_dir, device_name_chksum);
}
}
snprintf (source_file, 255, "%s/OpenCL/amp_a%d.cl", shared_dir, attack_kern);
}
-static void generate_cached_kernel_amp_filename (const uint attack_kern, char *profile_dir, char *device_name_chksum, int vendor_id, char *cached_file)
+static void generate_cached_kernel_amp_filename (const uint attack_kern, char *profile_dir, char *device_name_chksum, char *cached_file)
{
- snprintf (cached_file, 255, "%s/kernels/%d/amp_a%d.%s.kernel", profile_dir, vendor_id, attack_kern, device_name_chksum);
+ snprintf (cached_file, 255, "%s/kernels/amp_a%d.%s.kernel", profile_dir, attack_kern, device_name_chksum);
}
static uint convert_from_hex (char *line_buf, const uint line_len)
static void run_kernel_bzero (hc_device_param_t *device_param, cl_mem buf, const uint size)
{
- if (data.vendor_id == VENDOR_ID_AMD)
+ if (device_param->vendor_id == VENDOR_ID_AMD)
{
+ // So far tested, amd is the only supporting this OpenCL 1.2 function without segfaulting
+
const cl_uchar zero = 0;
hc_clEnqueueFillBuffer (device_param->command_queue, buf, &zero, sizeof (cl_uchar), 0, size, 0, NULL, NULL);
}
-
- if (data.vendor_id == VENDOR_ID_NV)
+ else
{
// NOTE: clEnqueueFillBuffer () always fails with -59
- // IOW, it's not supported by Nvidia ForceWare <= 352.21,
+ // IOW, it's not supported by Nvidia ForceWare <= 352.21, also pocl segfaults
// How's that possible, OpenCL 1.2 support is advertised??
// We need to workaround...
myfree (tmp);
}
-
- if (data.vendor_id == VENDOR_ID_GENERIC)
- {
- const cl_uchar zero = 0;
-
- hc_clEnqueueFillBuffer (device_param->command_queue, buf, &zero, sizeof (cl_uchar), 0, size, 0, NULL, NULL);
- }
}
static int run_rule_engine (const int rule_len, const char *rule_buf)
uint increment_max = INCREMENT_MAX;
char *cpu_affinity = NULL;
char *opencl_devices = NULL;
- char *opencl_platform = NULL;
+ char *opencl_platforms = NULL;
char *opencl_device_types = NULL;
char *truecrypt_keyfiles = NULL;
uint workload_profile = WORKLOAD_PROFILE;
#define IDX_MARKOV_HCSTAT 0xff24
#define IDX_CPU_AFFINITY 0xff25
#define IDX_OPENCL_DEVICES 'd'
- #define IDX_OPENCL_PLATFORM 0xff72
+ #define IDX_OPENCL_PLATFORMS 0xff72
#define IDX_OPENCL_DEVICE_TYPES 0xff73
#define IDX_WORKLOAD_PROFILE 'w'
#define IDX_KERNEL_ACCEL 'n'
{"markov-hcstat", required_argument, 0, IDX_MARKOV_HCSTAT},
{"cpu-affinity", required_argument, 0, IDX_CPU_AFFINITY},
{"opencl-devices", required_argument, 0, IDX_OPENCL_DEVICES},
- {"opencl-platform", required_argument, 0, IDX_OPENCL_PLATFORM},
+ {"opencl-platforms", required_argument, 0, IDX_OPENCL_PLATFORMS},
{"opencl-device-types", required_argument, 0, IDX_OPENCL_DEVICE_TYPES},
{"workload-profile", required_argument, 0, IDX_WORKLOAD_PROFILE},
{"kernel-accel", required_argument, 0, IDX_KERNEL_ACCEL},
myfree (exec_path);
+ /**
+ * kernel cache, we need to make sure folder exist
+ */
+
+ int kernels_folder_size = strlen (profile_dir) + 1 + 7 + 1 + 1;
+
+ char *kernels_folder = (char *) mymalloc (kernels_folder_size);
+
+ snprintf (kernels_folder, kernels_folder_size - 1, "%s/kernels", profile_dir);
+
+ mkdir (kernels_folder, 0700);
+
+ myfree (kernels_folder);
+
/**
* session
*/
case IDX_HEX_WORDLIST: hex_wordlist = 1; break;
case IDX_CPU_AFFINITY: cpu_affinity = optarg; break;
case IDX_OPENCL_DEVICES: opencl_devices = optarg; break;
- case IDX_OPENCL_PLATFORM: opencl_platform = optarg; break;
+ case IDX_OPENCL_PLATFORMS: opencl_platforms = optarg; break;
case IDX_OPENCL_DEVICE_TYPES:
opencl_device_types = optarg; break;
case IDX_WORKLOAD_PROFILE: workload_profile = atoi (optarg); break;
logfile_top_string (custom_charset_4);
logfile_top_string (debug_file);
logfile_top_string (opencl_devices);
- logfile_top_string (opencl_platform);
+ logfile_top_string (opencl_platforms);
logfile_top_string (opencl_device_types);
logfile_top_string (induction_dir);
logfile_top_string (markov_hcstat);
logfile_top_string (truecrypt_keyfiles);
/**
- * device types filter
+ * OpenCL platform selection
*/
- cl_device_type device_types_filter = setup_device_types_filter (opencl_device_types);
+ uint opencl_platforms_filter = setup_opencl_platforms_filter (opencl_platforms);
/**
- * devices
+ * OpenCL device selection
*/
uint opencl_devicemask = devices_to_devicemask (opencl_devices);
+ /**
+ * OpenCL device type selection
+ */
+
+ cl_device_type device_types_filter = setup_device_types_filter (opencl_device_types);
+
/**
* benchmark
*/
data.kernel_rules_buf = kernel_rules_buf;
/**
- * platform
+ * OpenCL platforms: detect
*/
cl_platform_id CL_platforms[CL_PLATFORMS_MAX];
return (-1);
}
- int CL_platform_sel = 1;
-
- if (opencl_platform != NULL)
- {
- CL_platform_sel = atoi (opencl_platform);
- }
+ /**
+ * OpenCL platforms: For each platform check if broken or unset features that we can not use, eg: temp_retain
+ */
- if (CL_platforms_cnt > 1)
+ for (uint i = 0; i < CL_platforms_cnt; i++)
{
- if (opencl_platform == NULL)
- {
- log_error ("ERROR: Too many OpenCL compatible platforms found");
+ cl_platform_id CL_platform = CL_platforms[i];
- log_info ("Please select a single platform using the --opencl-platform option");
- log_info ("");
- log_info ("Available OpenCL platforms:");
- log_info ("");
+ char CL_platform_vendor[INFOSZ];
- for (uint i = 0; i < CL_platforms_cnt; i++)
- {
- char CL_platform_vendor[INFOSZ];
-
- memset (CL_platform_vendor, 0, sizeof (CL_platform_vendor));
+ memset (CL_platform_vendor, 0, sizeof (CL_platform_vendor));
- hc_clGetPlatformInfo (CL_platforms[i], CL_PLATFORM_VENDOR, sizeof (CL_platform_vendor), CL_platform_vendor, NULL);
+ hc_clGetPlatformInfo (CL_platform, CL_PLATFORM_VENDOR, sizeof (CL_platform_vendor), CL_platform_vendor, NULL);
- log_info ("* %d = %s", i + 1, CL_platform_vendor);
- }
+ if (strcmp (CL_platform_vendor, CL_VENDOR_NV) == 0)
+ {
+ // make sure that we do not directly control the fan for NVidia
- log_info ("");
+ gpu_temp_retain = 0;
- return (-1);
+ data.gpu_temp_retain = gpu_temp_retain;
}
- else
+ else if (strcmp (CL_platform_vendor, CL_VENDOR_POCL) == 0)
{
- if (CL_platform_sel < 1)
- {
- log_error ("ERROR: --opencl-platform < 1");
-
- return (-1);
- }
-
- if (CL_platform_sel > (int) CL_platforms_cnt)
+ if (force == 0)
{
- log_error ("ERROR: invalid OpenCL platforms selected");
+ log_info ("");
+ log_info ("ATTENTION! All pocl drivers are known to be broken due to broken LLVM <= 3.7");
+ log_info ("You are STRONGLY encouraged not to use it");
+ log_info ("You can use --force to override this but do not post error reports if you do so");
+ log_info ("");
return (-1);
}
}
}
- else
- {
- if (CL_platform_sel != 1)
- {
- log_error ("ERROR: OpenCL platform number %d is not available", CL_platform_sel);
-
- return (-1);
- }
- }
-
- // zero-indexed: not starting to count at 1, as user does
-
- CL_platform_sel -= 1;
-
- cl_platform_id CL_platform = CL_platforms[CL_platform_sel];
-
- char CL_platform_vendor[INFOSZ];
-
- memset (CL_platform_vendor, 0, sizeof (CL_platform_vendor));
-
- hc_clGetPlatformInfo (CL_platform, CL_PLATFORM_VENDOR, sizeof (CL_platform_vendor), CL_platform_vendor, NULL);
-
- uint vendor_id;
-
- if (strcmp (CL_platform_vendor, CL_VENDOR_AMD) == 0)
- {
- vendor_id = VENDOR_ID_AMD;
- }
- else if (strcmp (CL_platform_vendor, CL_VENDOR_NV) == 0)
- {
- vendor_id = VENDOR_ID_NV;
-
- // make sure that we do not directly control the fan for NVidia
-
- gpu_temp_retain = 0;
-
- data.gpu_temp_retain = gpu_temp_retain;
- }
- else if (strcmp (CL_platform_vendor, CL_VENDOR_POCL) == 0)
- {
- if (force == 0)
- {
- log_error ("");
- log_error ("ATTENTION! All pocl drivers are known to be broken due to broken LLVM <= 3.7");
- log_error ("You are STRONGLY encouraged not to use it");
- log_error ("You can use --force to override this but do not post error reports if you do so");
-
- return (-1);
- }
-
- vendor_id = VENDOR_ID_GENERIC;
- }
- else
- {
- vendor_id = VENDOR_ID_GENERIC;
- }
-
- if (vendor_id == VENDOR_ID_GENERIC)
- {
- log_error ("Warning: unknown OpenCL vendor '%s' detected", CL_platform_vendor);
-
- gpu_temp_disable = 1;
- }
-
- data.vendor_id = vendor_id;
/**
- * cached kernel path depends on vendor_id which we don't know, so create it here
+ * OpenCL devices: push all devices from all platforms into the same device array
*/
- int vendor_id_folder_size = strlen (profile_dir) + 1 + 7 + 1 + 10 + 1;
-
- char *vendor_id_folder = (char *) mymalloc (vendor_id_folder_size);
-
- snprintf (vendor_id_folder, vendor_id_folder_size - 1, "%s/kernels", profile_dir);
-
- mkdir (vendor_id_folder, 0700);
-
- snprintf (vendor_id_folder, vendor_id_folder_size - 1, "%s/kernels/%d", profile_dir, vendor_id);
-
- mkdir (vendor_id_folder, 0700);
-
- myfree (vendor_id_folder);
-
- /**
- * devices
- */
+ uint devices_plf[DEVICES_MAX]; // device number on platform, required for hardware-management mapping
cl_device_id devices_all[DEVICES_MAX];
cl_device_id devices[DEVICES_MAX];
uint devices_all_cnt = 0;
+ uint devices_cnt = 0;
- hc_clGetDeviceIDs (CL_platform, device_types_filter, DEVICES_MAX, devices_all, (uint *) &devices_all_cnt);
-
- int hm_adapters_all = devices_all_cnt;
-
- hm_attrs_t hm_adapter_all[DEVICES_MAX];
-
- memset (hm_adapter_all, 0, sizeof (hm_adapter_all));
-
- if (gpu_temp_disable == 0)
+ for (uint i = 0; i < CL_platforms_cnt; i++)
{
- if (vendor_id == VENDOR_ID_NV)
- {
- #ifdef LINUX
- HM_LIB hm_dll = hm_init ();
-
- data.hm_dll = hm_dll;
-
- if (hc_NVML_nvmlInit (hm_dll) == NVML_SUCCESS)
- {
- HM_ADAPTER_NV nvGPUHandle[DEVICES_MAX];
-
- int tmp_in = hm_get_adapter_index_nv (nvGPUHandle);
-
- int tmp_out = 0;
-
- for (int i = 0; i < tmp_in; i++)
- {
- hm_adapter_all[tmp_out++].adapter_index.nv = nvGPUHandle[i];
- }
-
- hm_adapters_all = tmp_out;
-
- for (int i = 0; i < tmp_out; i++)
- {
- unsigned int speed;
-
- if (hc_NVML_nvmlDeviceGetFanSpeed (hm_dll, 1, hm_adapter_all[i].adapter_index.nv, &speed) != NVML_ERROR_NOT_SUPPORTED) hm_adapter_all[i].fan_supported = 1;
- }
- }
- #endif
+ if ((opencl_platforms_filter & (1 << i)) == 0) continue;
- #ifdef WIN
- if (NvAPI_Initialize () == NVAPI_OK)
- {
- HM_ADAPTER_NV nvGPUHandle[DEVICES_MAX];
-
- int tmp_in = hm_get_adapter_index_nv (nvGPUHandle);
-
- int tmp_out = 0;
-
- for (int i = 0; i < tmp_in; i++)
- {
- hm_adapter_all[tmp_out++].adapter_index.nv = nvGPUHandle[i];
- }
+ cl_platform_id CL_platform = CL_platforms[i];
- hm_adapters_all = tmp_out;
+ cl_device_id devices_platform[DEVICES_MAX];
- for (int i = 0; i < tmp_out; i++)
- {
- NvU32 speed;
+ cl_uint devices_platform_cnt = 0;
- if (NvAPI_GPU_GetTachReading (hm_adapter_all[i].adapter_index.nv, &speed) != NVAPI_NOT_SUPPORTED) hm_adapter_all[i].fan_supported = 1;
- }
- }
- #endif
- }
+ hc_clGetDeviceIDs (CL_platform, device_types_filter, DEVICES_MAX, devices_platform, &devices_platform_cnt);
- if (vendor_id == VENDOR_ID_AMD)
+ for (uint j = 0; j < devices_platform_cnt; j++)
{
- HM_LIB hm_dll = hm_init ();
-
- data.hm_dll = hm_dll;
-
- if (hc_ADL_Main_Control_Create (hm_dll, ADL_Main_Memory_Alloc, 0) == ADL_OK)
- {
- // total number of adapters
-
- int hm_adapters_num;
-
- if (get_adapters_num_amd (hm_dll, &hm_adapters_num) != 0) return (-1);
-
- // adapter info
-
- LPAdapterInfo lpAdapterInfo = hm_get_adapter_info_amd (hm_dll, hm_adapters_num);
-
- if (lpAdapterInfo == NULL) return (-1);
-
- // get a list (of ids of) valid/usable adapters
-
- int num_adl_adapters = 0;
-
- uint32_t *valid_adl_device_list = hm_get_list_valid_adl_adapters (hm_adapters_num, &num_adl_adapters, lpAdapterInfo);
-
- if (num_adl_adapters > 0)
- {
- hc_thread_mutex_lock (mux_adl);
-
- // hm_get_opencl_busid_devid (hm_adapter_all, devices_all_cnt, devices_all);
-
- hm_get_adapter_index_amd (hm_adapter_all, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
-
- hm_get_overdrive_version (hm_dll, hm_adapter_all, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
- hm_check_fanspeed_control (hm_dll, hm_adapter_all, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
-
- hc_thread_mutex_unlock (mux_adl);
- }
-
- hm_adapters_all = num_adl_adapters;
+ devices_all[devices_all_cnt] = devices_platform[j];
+ devices_plf[devices_all_cnt] = j;
- myfree (valid_adl_device_list);
- myfree (lpAdapterInfo);
- }
+ devices_all_cnt++;
}
}
- if (hm_adapters_all == 0)
- {
- gpu_temp_disable = 1;
- }
-
- if (gpu_temp_disable == 1)
- {
- gpu_temp_abort = 0;
- gpu_temp_retain = 0;
- }
-
/**
* enable custom signal handler(s)
*/
* devices mask and properties
*/
- uint hm_adapter_index = 0;
-
- uint devices_cnt = 0;
-
for (uint device_all_id = 0; device_all_id < devices_all_cnt; device_all_id++)
{
- const uint device_id = devices_cnt;
-
- devices[device_id] = devices_all[device_all_id];
-
- cl_device_type device_type;
-
- hc_clGetDeviceInfo (devices[device_id], CL_DEVICE_TYPE, sizeof (device_type), &device_type, NULL);
-
// skip the device, if the user did specify a list of GPUs to skip
if (opencl_devicemask)
if ((device_all_id_mask & opencl_devicemask) != device_all_id_mask)
{
- if (quiet == 0 && algorithm_pos == 0) log_info ("Device #%d: skipped by user", device_all_id_mask + 1);
-
- if (device_type & CL_DEVICE_TYPE_GPU) hm_adapter_index++;
+ if (quiet == 0 && algorithm_pos == 0) log_info ("Device #%d: skipped by user", device_all_id + 1);
continue;
}
}
+ const uint device_id = devices_cnt;
+
+ devices[device_id] = devices_all[device_all_id];
+
+ devices_plf[device_id] = devices_plf[device_all_id];
+
char device_name[INFOSZ];
memset (device_name, 0, sizeof (device_name));
(unsigned int) max_compute_units);
}
- // copy hm_adapter info to data.hm_device[]
-
- if (device_type & CL_DEVICE_TYPE_GPU)
- {
- memcpy (&data.hm_device[device_id], &hm_adapter_all[hm_adapter_index], sizeof (hm_attrs_t));
-
- hm_adapter_index++;
- }
- // this would be the CPU case (but it should be zeroed out already):
- // else memset (&data.hm_device[device_id], 0, sizeof (hm_attrs_t));
-
devices_cnt++;
}
}
/**
- * inform the user
+ * User-defined GPU temp handling
*/
- // gpu temp sanity check
+ if (gpu_temp_disable == 1)
+ {
+ gpu_temp_abort = 0;
+ gpu_temp_retain = 0;
+ }
if ((gpu_temp_abort != 0) && (gpu_temp_retain != 0))
{
data.gpu_temp_abort = gpu_temp_abort;
data.gpu_temp_retain = gpu_temp_retain;
+ /**
+ * inform the user
+ */
+
if (data.quiet == 0)
{
log_info ("Hashes: %u hashes; %u unique digests, %u unique salts", hashes_cnt_orig, digests_cnt, salts_cnt);
}
}
+ if (data.quiet == 0) log_info ("");
+
/**
* devices init
*/
device_param->device_maxmem_alloc = max_mem_alloc_size;
+ cl_uint vendor_id = 0;
+
+ hc_clGetDeviceInfo (device, CL_DEVICE_VENDOR_ID, sizeof (vendor_id), &vendor_id, NULL);
+
+ device_param->vendor_id = vendor_id;
+
char tmp[INFOSZ], t1[64];
memset (tmp, 0, sizeof (tmp));
hc_clGetDeviceInfo (device, CL_DEVICE_VERSION, sizeof (tmp), &tmp, NULL);
+ if (strstr (tmp, "pocl"))
+ {
+ // pocl returns the real AMD vendor_id id in CL_DEVICE_VENDOR_ID which causes many problems because of hms and missing amd_bfe () etc
+ // we need to overwrite vendor_id to avoid this. maybe open pocl issue?
+
+ cl_uint vendor_id = 0xffff;
+
+ device_param->vendor_id = vendor_id;
+ }
+
memset (t1, 0, sizeof (t1));
sscanf (tmp, "%*16s %*16s %*16s (%[^)]16s)", t1);
// create some filename that is easier to read on cached folder
- snprintf (tmp, sizeof (tmp) - 1, "%s-%s-%s-%d", device_param->device_name, device_param->device_version, device_param->driver_version, COMPTIME);
+ snprintf (tmp, sizeof (tmp) - 1, "%u-%s-%s-%s-%d", device_param->vendor_id, device_param->device_name, device_param->device_version, device_param->driver_version, COMPTIME);
uint device_name_digest[4];
device_param->device_processor_cores = device_processor_cores;
}
-
- if (vendor_id == VENDOR_ID_NV)
+ else if (vendor_id == VENDOR_ID_NV)
{
cl_uint kernel_exec_timeout = 0;
device_param->sm_minor = sm_minor;
device_param->sm_major = sm_major;
}
+ else
+ {
+ cl_uint device_processor_cores = 1;
+
+ device_param->device_processor_cores = device_processor_cores;
+ }
}
/**
if (data.quiet == 0) log_info (" See the wiki on how to disable it: https://hashcat.net/wiki/doku.php?id=timeout_patch");
}
}
-
- if (vendor_id == VENDOR_ID_AMD)
+ else if (vendor_id == VENDOR_ID_AMD)
{
int catalyst_check = (force == 1) ? 0 : 1;
}
}
+ /**
+ * HM devices: init
+ */
+
+ hm_attrs_t hm_adapters_nv[DEVICES_MAX];
+ hm_attrs_t hm_adapters_amd[DEVICES_MAX];
+
+ memset (hm_adapters_nv, 0, sizeof (hm_adapters_nv));
+ memset (hm_adapters_amd, 0, sizeof (hm_adapters_amd));
+
+ if (gpu_temp_disable == 0)
+ {
+ HM_LIB hm_dll_nv = hm_init (VENDOR_ID_NV);
+
+ data.hm_dll_nv = hm_dll_nv;
+
+ if (hm_dll_nv)
+ {
+ #ifdef LINUX
+ if (hc_NVML_nvmlInit (hm_dll_nv) == NVML_SUCCESS)
+ {
+ HM_ADAPTER_NV nvGPUHandle[DEVICES_MAX];
+
+ int tmp_in = hm_get_adapter_index_nv (nvGPUHandle);
+
+ int tmp_out = 0;
+
+ for (int i = 0; i < tmp_in; i++)
+ {
+ hm_adapters_nv[tmp_out++].adapter_index.nv = nvGPUHandle[i];
+ }
+
+ for (int i = 0; i < tmp_out; i++)
+ {
+ unsigned int speed;
+
+ if (hc_NVML_nvmlDeviceGetFanSpeed (hm_dll_nv, 1, hm_adapters_nv[i].adapter_index.nv, &speed) != NVML_ERROR_NOT_SUPPORTED) hm_adapters_nv[i].fan_supported = 1;
+ }
+ }
+ #endif
+
+ #ifdef WIN
+ if (NvAPI_Initialize () == NVAPI_OK)
+ {
+ HM_ADAPTER_NV nvGPUHandle[DEVICES_MAX];
+
+ int tmp_in = hm_get_adapter_index_nv (nvGPUHandle);
+
+ int tmp_out = 0;
+
+ for (int i = 0; i < tmp_in; i++)
+ {
+ hm_adapters_nv[tmp_out++].adapter_index.nv = nvGPUHandle[i];
+ }
+
+ for (int i = 0; i < tmp_out; i++)
+ {
+ NvU32 speed;
+
+ if (NvAPI_GPU_GetTachReading (hm_adapters_nv[i].adapter_index.nv, &speed) != NVAPI_NOT_SUPPORTED) hm_adapters_nv[i].fan_supported = 1;
+ }
+ }
+ #endif
+ }
+
+ HM_LIB hm_dll_amd = hm_init (VENDOR_ID_AMD);
+
+ data.hm_dll_amd = hm_dll_amd;
+
+ if (hm_dll_amd)
+ {
+ if (hc_ADL_Main_Control_Create (hm_dll_amd, ADL_Main_Memory_Alloc, 0) == ADL_OK)
+ {
+ // total number of adapters
+
+ int hm_adapters_num;
+
+ if (get_adapters_num_amd (hm_dll_amd, &hm_adapters_num) != 0) return (-1);
+
+ // adapter info
+
+ LPAdapterInfo lpAdapterInfo = hm_get_adapter_info_amd (hm_dll_amd, hm_adapters_num);
+
+ if (lpAdapterInfo == NULL) return (-1);
+
+ // get a list (of ids of) valid/usable adapters
+
+ int num_adl_adapters = 0;
+
+ uint32_t *valid_adl_device_list = hm_get_list_valid_adl_adapters (hm_adapters_num, &num_adl_adapters, lpAdapterInfo);
+
+ if (num_adl_adapters > 0)
+ {
+ hc_thread_mutex_lock (mux_adl);
+
+ // hm_get_opencl_busid_devid (hm_adapters_amd, devices_all_cnt, devices_all);
+
+ hm_get_adapter_index_amd (hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
+
+ hm_get_overdrive_version (hm_dll_amd, hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
+ hm_check_fanspeed_control (hm_dll_amd, hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
+
+ hc_thread_mutex_unlock (mux_adl);
+ }
+
+ myfree (valid_adl_device_list);
+ myfree (lpAdapterInfo);
+ }
+ }
+ }
+
+ /**
+ * HM devices: copy
+ */
+
+ if (gpu_temp_disable == 0)
+ {
+ for (uint device_id = 0; device_id < devices_cnt; device_id++)
+ {
+ hc_device_param_t *device_param = &data.devices_param[device_id];
+
+ if ((device_param->device_type & CL_DEVICE_TYPE_GPU) == 0) continue;
+
+ cl_uint device_id_on_platform = devices_plf[device_id];
+
+ if (device_param->vendor_id == VENDOR_ID_NV)
+ {
+ memcpy (&data.hm_device[device_id], &hm_adapters_nv[device_id_on_platform], sizeof (hm_attrs_t));
+ }
+
+ if (device_param->vendor_id == VENDOR_ID_AMD)
+ {
+ memcpy (&data.hm_device[device_id], &hm_adapters_amd[device_id_on_platform], sizeof (hm_attrs_t));
+ }
+ }
+ }
+
/*
* Temporary fix:
* with AMD r9 295x cards it seems that we need to set the powertune value just AFTER the ocl init stuff
* Driver / ADL bug?
*/
- if (vendor_id == VENDOR_ID_AMD)
+ if (powertune_enable == 1)
{
- if (powertune_enable == 1)
- {
- hc_thread_mutex_lock (mux_adl);
+ hc_thread_mutex_lock (mux_adl);
- for (uint i = 0; i < devices_cnt; i++)
+ for (uint device_id = 0; device_id < devices_cnt; device_id++)
+ {
+ if (data.hm_device[device_id].od_version == 6)
{
- if (data.hm_device[i].od_version == 6)
+ // set powertune value only
+
+ int powertune_supported = 0;
+
+ int ADL_rc = 0;
+
+ if ((ADL_rc = hc_ADL_Overdrive6_PowerControl_Caps (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK)
{
- // set powertune value only
+ log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
- int powertune_supported = 0;
+ return (-1);
+ }
- int ADL_rc = 0;
+ if (powertune_supported != 0)
+ {
+ // powertune set
+ ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0};
- if ((ADL_rc = hc_ADL_Overdrive6_PowerControl_Caps (data.hm_dll, data.hm_device[i].adapter_index.amd, &powertune_supported)) != ADL_OK)
+ if ((ADL_rc = hc_ADL_Overdrive_PowerControlInfo_Get (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, &powertune)) != ADL_OK)
{
- log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
+ log_error ("ERROR: Failed to get current ADL PowerControl settings");
return (-1);
}
- if (powertune_supported != 0)
+ if ((ADL_rc = hc_ADL_Overdrive_PowerControl_Set (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, powertune.iMaxValue)) != ADL_OK)
{
- // powertune set
- ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0};
-
- if ((ADL_rc = hc_ADL_Overdrive_PowerControlInfo_Get (data.hm_dll, data.hm_device[i].adapter_index.amd, &powertune)) != ADL_OK)
- {
- log_error ("ERROR: Failed to get current ADL PowerControl settings");
-
- return (-1);
- }
-
- if ((ADL_rc = hc_ADL_Overdrive_PowerControl_Set (data.hm_dll, data.hm_device[i].adapter_index.amd, powertune.iMaxValue)) != ADL_OK)
- {
- log_error ("ERROR: Failed to set new ADL PowerControl values");
+ log_error ("ERROR: Failed to set new ADL PowerControl values");
- return (-1);
- }
+ return (-1);
}
}
}
-
- hc_thread_mutex_unlock (mux_adl);
}
+
+ hc_thread_mutex_unlock (mux_adl);
}
uint kernel_blocks_all = 0;
if (hash_mode == 8900)
{
- if (vendor_id == VENDOR_ID_AMD)
+ if (device_param->vendor_id == VENDOR_ID_AMD)
{
tmto_start = 1;
}
- else if (vendor_id == VENDOR_ID_NV)
+ else if (device_param->vendor_id == VENDOR_ID_NV)
{
tmto_start = 3;
}
}
else if (hash_mode == 9300)
{
- if (vendor_id == VENDOR_ID_AMD)
+ if (device_param->vendor_id == VENDOR_ID_AMD)
{
tmto_start = 3;
}
- else if (vendor_id == VENDOR_ID_NV)
+ else if (device_param->vendor_id == VENDOR_ID_NV)
{
tmto_start = 5;
}
uint shader_per_mp = 1;
- if (vendor_id == VENDOR_ID_AMD)
+ if (device_param->vendor_id == VENDOR_ID_AMD)
{
shader_per_mp = 8;
}
-
- if (vendor_id == VENDOR_ID_NV)
+ else if (device_param->vendor_id == VENDOR_ID_NV)
{
shader_per_mp = 32;
}
char build_opts[1024];
- // we don't have sm_* on AMD but it doesn't matter
+ // we don't have sm_* on vendors not NV but it doesn't matter
- sprintf (build_opts, "-I%s/ -DVENDOR_ID=%d -DCUDA_ARCH=%d", shared_dir, vendor_id, (device_param->sm_major * 100) + device_param->sm_minor);
+ sprintf (build_opts, "-I%s/ -DVENDOR_ID=%d -DCUDA_ARCH=%d", shared_dir, device_param->vendor_id, (device_param->sm_major * 100) + device_param->sm_minor);
/**
* a0 kernel, required for some fast hashes to make weak_hash_check work
memset (cached_file, 0, sizeof (cached_file));
- generate_cached_kernel_filename (attack_exec, ATTACK_KERN_STRAIGHT, kern_type, profile_dir, device_name_chksum, vendor_id, cached_file);
+ generate_cached_kernel_filename (attack_exec, ATTACK_KERN_STRAIGHT, kern_type, profile_dir, device_name_chksum, cached_file);
int cached = 1;
memset (cached_file, 0, sizeof (cached_file));
- generate_cached_kernel_filename (attack_exec, attack_kern, kern_type, profile_dir, device_name_chksum, vendor_id, cached_file);
+ generate_cached_kernel_filename (attack_exec, attack_kern, kern_type, profile_dir, device_name_chksum, cached_file);
int cached = 1;
memset (cached_file, 0, sizeof (cached_file));
- generate_cached_kernel_mp_filename (opti_type, opts_type, profile_dir, device_name_chksum, vendor_id, cached_file);
+ generate_cached_kernel_mp_filename (opti_type, opts_type, profile_dir, device_name_chksum, cached_file);
int cached = 1;
memset (cached_file, 0, sizeof (cached_file));
- generate_cached_kernel_amp_filename (attack_kern, profile_dir, device_name_chksum, vendor_id, cached_file);
+ generate_cached_kernel_amp_filename (attack_kern, profile_dir, device_name_chksum, cached_file);
int cached = 1;
uint cur_temp = 0;
uint default_temp = 0;
- int ADL_rc = hc_ADL_Overdrive6_TargetTemperatureData_Get (data.hm_dll, data.hm_device[device_id].adapter_index.amd, (int *) &cur_temp, (int *) &default_temp);
+ int ADL_rc = hc_ADL_Overdrive6_TargetTemperatureData_Get (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, (int *) &cur_temp, (int *) &default_temp);
if (ADL_rc == ADL_OK)
{
int powertune_supported = 0;
- if ((ADL_rc = hc_ADL_Overdrive6_PowerControl_Caps (data.hm_dll, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK)
+ if ((ADL_rc = hc_ADL_Overdrive6_PowerControl_Caps (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK)
{
log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0};
- if ((ADL_rc = hc_ADL_Overdrive_PowerControlInfo_Get (data.hm_dll, data.hm_device[device_id].adapter_index.amd, &powertune)) == ADL_OK)
+ if ((ADL_rc = hc_ADL_Overdrive_PowerControlInfo_Get (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, &powertune)) == ADL_OK)
{
- ADL_rc = hc_ADL_Overdrive_PowerControl_Get (data.hm_dll, data.hm_device[device_id].adapter_index.amd, &od_power_control_status[device_id]);
+ ADL_rc = hc_ADL_Overdrive_PowerControl_Get (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, &od_power_control_status[device_id]);
}
if (ADL_rc != ADL_OK)
return (-1);
}
- if ((ADL_rc = hc_ADL_Overdrive_PowerControl_Set (data.hm_dll, data.hm_device[device_id].adapter_index.amd, powertune.iMaxValue)) != ADL_OK)
+ if ((ADL_rc = hc_ADL_Overdrive_PowerControl_Set (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, powertune.iMaxValue)) != ADL_OK)
{
log_error ("ERROR: Failed to set new ADL PowerControl values");
od_clock_mem_status[device_id].state.iNumberOfPerformanceLevels = 2;
- if ((ADL_rc = hc_ADL_Overdrive_StateInfo_Get (data.hm_dll, data.hm_device[device_id].adapter_index.amd, ADL_OD6_GETSTATEINFO_CUSTOM_PERFORMANCE, &od_clock_mem_status[device_id])) != ADL_OK)
+ if ((ADL_rc = hc_ADL_Overdrive_StateInfo_Get (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, ADL_OD6_GETSTATEINFO_CUSTOM_PERFORMANCE, &od_clock_mem_status[device_id])) != ADL_OK)
{
log_error ("ERROR: Failed to get ADL memory and engine clock frequency");
ADLOD6Capabilities caps = {0, 0, 0, {0, 0, 0}, {0, 0, 0}, 0, 0};
- if ((ADL_rc = hc_ADL_Overdrive_Capabilities_Get (data.hm_dll, data.hm_device[device_id].adapter_index.amd, &caps)) != ADL_OK)
+ if ((ADL_rc = hc_ADL_Overdrive_Capabilities_Get (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, &caps)) != ADL_OK)
{
log_error ("ERROR: Failed to get ADL device capabilities");
performance_state->aLevels[0].iMemoryClock = memory_clock_profile_max;
performance_state->aLevels[1].iMemoryClock = memory_clock_profile_max;
- if ((ADL_rc = hc_ADL_Overdrive_State_Set (data.hm_dll, data.hm_device[device_id].adapter_index.amd, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK)
+ if ((ADL_rc = hc_ADL_Overdrive_State_Set (data.hm_dll_amd, data.hm_device[device_id].adapter_index.amd, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK)
{
log_info ("ERROR: Failed to set ADL performance state");
int powertune_supported = 0;
- if ((hc_ADL_Overdrive6_PowerControl_Caps (data.hm_dll, data.hm_device[i].adapter_index.amd, &powertune_supported)) != ADL_OK)
+ if ((hc_ADL_Overdrive6_PowerControl_Caps (data.hm_dll_amd, data.hm_device[i].adapter_index.amd, &powertune_supported)) != ADL_OK)
{
log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
{
// powercontrol settings
- if ((hc_ADL_Overdrive_PowerControl_Set (data.hm_dll, data.hm_device[i].adapter_index.amd, od_power_control_status[i])) != ADL_OK)
+ if ((hc_ADL_Overdrive_PowerControl_Set (data.hm_dll_amd, data.hm_device[i].adapter_index.amd, od_power_control_status[i])) != ADL_OK)
{
log_info ("ERROR: Failed to restore the ADL PowerControl values");
performance_state->aLevels[0].iMemoryClock = od_clock_mem_status[i].state.aLevels[0].iMemoryClock;
performance_state->aLevels[1].iMemoryClock = od_clock_mem_status[i].state.aLevels[1].iMemoryClock;
- if ((hc_ADL_Overdrive_State_Set (data.hm_dll, data.hm_device[i].adapter_index.amd, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK)
+ if ((hc_ADL_Overdrive_State_Set (data.hm_dll_amd, data.hm_device[i].adapter_index.amd, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK)
{
log_info ("ERROR: Failed to restore ADL performance state");
if (gpu_temp_disable == 0)
{
- if (vendor_id == VENDOR_ID_NV)
+ #ifdef LINUX
+ if (data.hm_dll_nv)
{
- #ifdef LINUX
- hc_NVML_nvmlShutdown (data.hm_dll);
- #endif
+ hc_NVML_nvmlShutdown (data.hm_dll_nv);
- #ifdef WIN
- NvAPI_Unload ();
- #endif
+ hm_close (data.hm_dll_nv);
}
+ #endif
- if (vendor_id == VENDOR_ID_AMD)
- {
- hc_ADL_Main_Control_Destroy (data.hm_dll);
-
- hm_close (data.hm_dll);
- }
+ #ifdef WIN
+ NvAPI_Unload ();
+ #endif
- #ifdef LINUX
- if (vendor_id == VENDOR_ID_NV)
+ if (data.hm_dll_amd)
{
- hm_close (data.hm_dll);
+ hc_ADL_Main_Control_Destroy (data.hm_dll_amd);
+
+ hm_close (data.hm_dll_amd);
}
- #endif
}
// free memory
projects / hashcat.git / blobdiff