X-Git-Url: https://www.flypig.org.uk/git/?a=blobdiff_plain;f=src%2Fhashcat.c;h=23b91c5eef2d420b11d5495befaa21ea7d3946b6;hb=e8aefb412cd45b7770bf7d3123b374aa822a0782;hp=76446bdb60f93d6e497bbb9d302528d3ecc51489;hpb=a23c0c471666c9a2bb8c95ac94f062ace0dd0135;p=hashcat.git diff --git a/src/hashcat.c b/src/hashcat.c index 76446bd..23b91c5 100644 --- a/src/hashcat.c +++ b/src/hashcat.c @@ -1,4 +1,4 @@ -/** + /** * Authors.....: Jens Steube * Gabriele Gristina * magnum @@ -33,6 +33,7 @@ double TARGET_MS_PROFILE[4] = { 2, 12, 96, 480 }; #define MARKOV_DISABLE 0 #define MARKOV_CLASSIC 0 #define BENCHMARK 0 +#define STDOUT_FLAG 0 #define RESTORE 0 #define RESTORE_TIMER 60 #define RESTORE_DISABLE 0 @@ -74,9 +75,10 @@ double TARGET_MS_PROFILE[4] = { 2, 12, 96, 480 }; #define SEPARATOR ':' #define BITMAP_MIN 16 #define BITMAP_MAX 24 +#define NVIDIA_SPIN_DAMP 100 #define GPU_TEMP_DISABLE 0 #define GPU_TEMP_ABORT 90 -#define GPU_TEMP_RETAIN 0 +#define GPU_TEMP_RETAIN 65 #define WORKLOAD_PROFILE 2 #define KERNEL_ACCEL 0 #define KERNEL_LOOPS 0 @@ -150,6 +152,8 @@ double TARGET_MS_PROFILE[4] = { 2, 12, 96, 480 }; #define NUM_DEFAULT_BENCHMARK_ALGORITHMS 143 +#define NVIDIA_100PERCENTCPU_WORKAROUND 100 + #define global_free(attr) \ { \ myfree ((void *) data.attr); \ @@ -164,6 +168,12 @@ double TARGET_MS_PROFILE[4] = { 2, 12, 96, 480 }; attr = NULL; \ } +#if defined(_WIN32) || defined(__WIN32__) || defined(__CYGWIN__) +#define HC_API_CALL __stdcall +#else +#define HC_API_CALL +#endif + static uint default_benchmark_algorithms[NUM_DEFAULT_BENCHMARK_ALGORITHMS] = { 900, @@ -380,6 +390,7 @@ const char *USAGE_BIG[] = " --outfile-autohex-disable | | Disable the use of $HEX[] in output plains |", " --outfile-check-timer | Num | Sets seconds between outfile checks to X | --outfile-check=30", " -p, --separator | Char | Separator char for hashlists and outfile | -p :", + " --stdout | | Do not crack a hash, instead print candidates only |", " --show | | Show cracked passwords only |", " --left | | Show un-cracked passwords only |", " --username | | Enable ignoring of usernames in hashfile |", @@ -407,11 +418,12 @@ const char *USAGE_BIG[] = " -w, --workload-profile | Num | Enable a specific workload profile, see pool below | -w 3", " -n, --kernel-accel | Num | Manual workload tuning, set outerloop step size to X | -n 64", " -u, --kernel-loops | Num | Manual workload tuning, set innerloop step size to X | -u 256", + " --nvidia-spin-damp | Num | Workaround NVidias CPU burning loop bug, in percent | --nvidia-spin-damp=50", " --gpu-temp-disable | | Disable temperature and fanspeed reads and triggers |", #ifdef HAVE_HWMON " --gpu-temp-abort | Num | Abort if GPU temperature reaches X degrees celsius | --gpu-temp-abort=100", " --gpu-temp-retain | Num | Try to retain GPU temperature at X degrees celsius | --gpu-temp-retain=95", - " --powertune-enable | | Enable automatic power tuning (AMD OverDrive 6 only) |", + " --powertune-enable | | Enable power tuning, restores settings when finished |", #endif " --scrypt-tmto | Num | Manually override TMTO value for scrypt to X | --scrypt-tmto=3", " -s, --skip | Num | Skip X words from the start | -s 1000000", @@ -570,6 +582,7 @@ const char *USAGE_BIG[] = " 22 | Juniper Netscreen/SSG (ScreenOS) | Operating-Systems", " 501 | Juniper IVE | Operating-Systems", " 5800 | Android PIN | Operating-Systems", + " 13800 | Windows 8+ phone PIN/Password | Operating-Systems", " 8100 | Citrix Netscaler | Operating-Systems", " 8500 | RACF | Operating-Systems", " 7200 | GRUB 2 | Operating-Systems", @@ -1531,6 +1544,12 @@ void status_display () } #ifdef HAVE_HWMON + + if (data.devices_status == STATUS_EXHAUSTED) return; + if (data.devices_status == STATUS_CRACKED) return; + if (data.devices_status == STATUS_ABORTED) return; + if (data.devices_status == STATUS_QUIT) return; + if (data.gpu_temp_disable == 0) { hc_thread_mutex_lock (mux_adl); @@ -1614,6 +1633,7 @@ void status_display () hc_thread_mutex_unlock (mux_adl); } + #endif // HAVE_HWMON } @@ -2297,6 +2317,175 @@ static void check_cracked (hc_device_param_t *device_param, const uint salt_pos) } } +static void process_stdout (hc_device_param_t *device_param, const uint pws_cnt) +{ + char out_buf[HCBUFSIZ] = { 0 }; + + uint plain_buf[16] = { 0 }; + + u8 *plain_ptr = (u8 *) plain_buf; + + uint plain_len = 0; + + const uint il_cnt = device_param->kernel_params_buf32[27]; // ugly, i know + + if (data.attack_mode == ATTACK_MODE_STRAIGHT) + { + pw_t pw; + + for (uint gidvid = 0; gidvid < pws_cnt; gidvid++) + { + gidd_to_pw_t (device_param, gidvid, &pw); + + const uint pos = device_param->innerloop_pos; + + for (uint il_pos = 0; il_pos < il_cnt; il_pos++) + { + for (int i = 0; i < 8; i++) + { + plain_buf[i] = pw.i[i]; + } + + plain_len = pw.pw_len; + + plain_len = apply_rules (data.kernel_rules_buf[pos + il_pos].cmds, &plain_buf[0], &plain_buf[4], plain_len); + + if (plain_len > data.pw_max) plain_len = data.pw_max; + + format_output (stdout, out_buf, plain_ptr, plain_len, 0, NULL, 0); + } + } + } + else if (data.attack_mode == ATTACK_MODE_COMBI) + { + pw_t pw; + + for (uint gidvid = 0; gidvid < pws_cnt; gidvid++) + { + gidd_to_pw_t (device_param, gidvid, &pw); + + for (uint il_pos = 0; il_pos < il_cnt; il_pos++) + { + for (int i = 0; i < 8; i++) + { + plain_buf[i] = pw.i[i]; + } + + plain_len = pw.pw_len; + + char *comb_buf = (char *) device_param->combs_buf[il_pos].i; + uint comb_len = device_param->combs_buf[il_pos].pw_len; + + if (data.combs_mode == COMBINATOR_MODE_BASE_LEFT) + { + memcpy (plain_ptr + plain_len, comb_buf, comb_len); + } + else + { + memmove (plain_ptr + comb_len, plain_ptr, plain_len); + + memcpy (plain_ptr, comb_buf, comb_len); + } + + plain_len += comb_len; + + if (data.pw_max != PW_DICTMAX1) + { + if (plain_len > data.pw_max) plain_len = data.pw_max; + } + + format_output (stdout, out_buf, plain_ptr, plain_len, 0, NULL, 0); + } + } + } + else if (data.attack_mode == ATTACK_MODE_BF) + { + for (uint gidvid = 0; gidvid < pws_cnt; gidvid++) + { + for (uint il_pos = 0; il_pos < il_cnt; il_pos++) + { + u64 l_off = device_param->kernel_params_mp_l_buf64[3] + gidvid; + u64 r_off = device_param->kernel_params_mp_r_buf64[3] + il_pos; + + uint l_start = device_param->kernel_params_mp_l_buf32[5]; + uint r_start = device_param->kernel_params_mp_r_buf32[5]; + + uint l_stop = device_param->kernel_params_mp_l_buf32[4]; + uint r_stop = device_param->kernel_params_mp_r_buf32[4]; + + sp_exec (l_off, (char *) plain_ptr + l_start, data.root_css_buf, data.markov_css_buf, l_start, l_start + l_stop); + sp_exec (r_off, (char *) plain_ptr + r_start, data.root_css_buf, data.markov_css_buf, r_start, r_start + r_stop); + + plain_len = data.css_cnt; + + format_output (stdout, out_buf, plain_ptr, plain_len, 0, NULL, 0); + } + } + } + else if (data.attack_mode == ATTACK_MODE_HYBRID1) + { + pw_t pw; + + for (uint gidvid = 0; gidvid < pws_cnt; gidvid++) + { + gidd_to_pw_t (device_param, gidvid, &pw); + + for (uint il_pos = 0; il_pos < il_cnt; il_pos++) + { + for (int i = 0; i < 8; i++) + { + plain_buf[i] = pw.i[i]; + } + + plain_len = pw.pw_len; + + u64 off = device_param->kernel_params_mp_buf64[3] + il_pos; + + uint start = 0; + uint stop = device_param->kernel_params_mp_buf32[4]; + + sp_exec (off, (char *) plain_ptr + plain_len, data.root_css_buf, data.markov_css_buf, start, start + stop); + + plain_len += start + stop; + + format_output (stdout, out_buf, plain_ptr, plain_len, 0, NULL, 0); + } + } + } + else if (data.attack_mode == ATTACK_MODE_HYBRID2) + { + pw_t pw; + + for (uint gidvid = 0; gidvid < pws_cnt; gidvid++) + { + gidd_to_pw_t (device_param, gidvid, &pw); + + for (uint il_pos = 0; il_pos < il_cnt; il_pos++) + { + for (int i = 0; i < 8; i++) + { + plain_buf[i] = pw.i[i]; + } + + plain_len = pw.pw_len; + + u64 off = device_param->kernel_params_mp_buf64[3] + il_pos; + + uint start = 0; + uint stop = device_param->kernel_params_mp_buf32[4]; + + memmove (plain_ptr + stop, plain_ptr, plain_len); + + sp_exec (off, (char *) plain_ptr, data.root_css_buf, data.markov_css_buf, start, start + stop); + + plain_len += start + stop; + + format_output (stdout, out_buf, plain_ptr, plain_len, 0, NULL, 0); + } + } + } +} + static void save_hash () { char *hashfile = data.hashfile; @@ -2334,8 +2523,6 @@ static void save_hash () if (data.hash_mode != 2500) { - char out_buf[HCBUFSIZ] = { 0 }; - if (data.username == 1) { user_t *user = data.hash_info[idx]->user; @@ -2347,11 +2534,15 @@ static void save_hash () fputc (separator, fp); } + char out_buf[HCBUFSIZ]; // scratch buffer + + out_buf[0] = 0; + ascii_digest (out_buf, salt_pos, digest_pos); fputs (out_buf, fp); - log_out (fp, ""); + fputc ('\n', fp); } else { @@ -2389,43 +2580,7 @@ static void save_hash () unlink (old_hashfile); } -static float find_kernel_power_div (const u64 total_left, const uint kernel_power_all) -{ - // function called only in case kernel_power_all > words_left - - float kernel_power_div = (float) (total_left) / kernel_power_all; - - kernel_power_div += kernel_power_div / 100; - - u32 kernel_power_new = (u32) (kernel_power_all * kernel_power_div); - - while (kernel_power_new < total_left) - { - kernel_power_div += kernel_power_div / 100; - - kernel_power_new = (u32) (kernel_power_all * kernel_power_div); - } - - if (data.quiet == 0) - { - clear_prompt (); - - //log_info (""); - - log_info ("INFO: approaching final keyspace, workload adjusted"); - log_info (""); - - fprintf (stdout, "%s", PROMPT); - - fflush (stdout); - } - - if ((kernel_power_all * kernel_power_div) < 8) return 1; - - return kernel_power_div; -} - -static void run_kernel (const uint kern_run, hc_device_param_t *device_param, const uint num, const uint event_update) +static void run_kernel (const uint kern_run, hc_device_param_t *device_param, const uint num, const uint event_update, const uint iteration) { uint num_elements = num; @@ -2488,21 +2643,50 @@ static void run_kernel (const uint kern_run, hc_device_param_t *device_param, co hc_clFlush (data.ocl, device_param->command_queue); + if (device_param->nvidia_spin_damp) + { + if (data.devices_status == STATUS_RUNNING) + { + if (iteration < EXPECTED_ITERATIONS) + { + switch (kern_run) + { + case KERN_RUN_1: if (device_param->exec_us_prev1[iteration]) usleep (device_param->exec_us_prev1[iteration] * device_param->nvidia_spin_damp); break; + case KERN_RUN_2: if (device_param->exec_us_prev2[iteration]) usleep (device_param->exec_us_prev2[iteration] * device_param->nvidia_spin_damp); break; + case KERN_RUN_3: if (device_param->exec_us_prev3[iteration]) usleep (device_param->exec_us_prev3[iteration] * device_param->nvidia_spin_damp); break; + } + } + } + } + hc_clWaitForEvents (data.ocl, 1, &event); - if (event_update) - { - cl_ulong time_start; - cl_ulong time_end; + cl_ulong time_start; + cl_ulong time_end; - hc_clGetEventProfilingInfo (data.ocl, event, CL_PROFILING_COMMAND_START, sizeof (time_start), &time_start, NULL); - hc_clGetEventProfilingInfo (data.ocl, event, CL_PROFILING_COMMAND_END, sizeof (time_end), &time_end, NULL); + hc_clGetEventProfilingInfo (data.ocl, event, CL_PROFILING_COMMAND_START, sizeof (time_start), &time_start, NULL); + hc_clGetEventProfilingInfo (data.ocl, event, CL_PROFILING_COMMAND_END, sizeof (time_end), &time_end, NULL); - const double exec_time = (double) (time_end - time_start) / 1000000.0; + const double exec_us = (double) (time_end - time_start) / 1000; + + if (data.devices_status == STATUS_RUNNING) + { + if (iteration < EXPECTED_ITERATIONS) + { + switch (kern_run) + { + case KERN_RUN_1: device_param->exec_us_prev1[iteration] = exec_us; break; + case KERN_RUN_2: device_param->exec_us_prev2[iteration] = exec_us; break; + case KERN_RUN_3: device_param->exec_us_prev3[iteration] = exec_us; break; + } + } + } + if (event_update) + { uint exec_pos = device_param->exec_pos; - device_param->exec_ms[exec_pos] = exec_time; + device_param->exec_ms[exec_pos] = exec_us / 1000; exec_pos++; @@ -2715,8 +2899,15 @@ static void run_kernel_bzero (hc_device_param_t *device_param, cl_mem buf, const */ } -static void choose_kernel (hc_device_param_t *device_param, const uint attack_exec, const uint attack_mode, const uint opts_type, const salt_t *salt_buf, const uint highest_pw_len, const uint pws_cnt) +static void choose_kernel (hc_device_param_t *device_param, const uint attack_exec, const uint attack_mode, const uint opts_type, const salt_t *salt_buf, const uint highest_pw_len, const uint pws_cnt, const uint fast_iteration) { + if (data.hash_mode == 2000) + { + process_stdout (device_param, pws_cnt); + + return; + } + if (attack_exec == ATTACK_EXEC_INSIDE_KERNEL) { if (attack_mode == ATTACK_MODE_BF) @@ -2735,33 +2926,39 @@ static void choose_kernel (hc_device_param_t *device_param, const uint attack_ex if (highest_pw_len < 16) { - run_kernel (KERN_RUN_1, device_param, pws_cnt, true); + run_kernel (KERN_RUN_1, device_param, pws_cnt, true, fast_iteration); } else if (highest_pw_len < 32) { - run_kernel (KERN_RUN_2, device_param, pws_cnt, true); + run_kernel (KERN_RUN_2, device_param, pws_cnt, true, fast_iteration); } else { - run_kernel (KERN_RUN_3, device_param, pws_cnt, true); + run_kernel (KERN_RUN_3, device_param, pws_cnt, true, fast_iteration); } } else { run_kernel_amp (device_param, pws_cnt); - run_kernel (KERN_RUN_1, device_param, pws_cnt, false); + run_kernel (KERN_RUN_1, device_param, pws_cnt, false, 0); if (opts_type & OPTS_TYPE_HOOK12) { - run_kernel (KERN_RUN_12, device_param, pws_cnt, false); + run_kernel (KERN_RUN_12, device_param, pws_cnt, false, 0); + + hc_clEnqueueReadBuffer (data.ocl, device_param->command_queue, device_param->d_hooks, CL_TRUE, 0, device_param->size_hooks, device_param->hooks_buf, 0, NULL, NULL); + + // do something with data + + hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_hooks, CL_TRUE, 0, device_param->size_hooks, device_param->hooks_buf, 0, NULL, NULL); } uint iter = salt_buf->salt_iter; uint loop_step = device_param->kernel_loops; - for (uint loop_pos = 0; loop_pos < iter; loop_pos += loop_step) + for (uint loop_pos = 0, slow_iteration = 0; loop_pos < iter; loop_pos += loop_step, slow_iteration++) { uint loop_left = iter - loop_pos; @@ -2770,7 +2967,7 @@ static void choose_kernel (hc_device_param_t *device_param, const uint attack_ex device_param->kernel_params_buf32[25] = loop_pos; device_param->kernel_params_buf32[26] = loop_left; - run_kernel (KERN_RUN_2, device_param, pws_cnt, true); + run_kernel (KERN_RUN_2, device_param, pws_cnt, true, slow_iteration); if (data.devices_status == STATUS_CRACKED) break; if (data.devices_status == STATUS_ABORTED) break; @@ -2802,7 +2999,7 @@ static void choose_kernel (hc_device_param_t *device_param, const uint attack_ex if (opts_type & OPTS_TYPE_HOOK23) { - run_kernel (KERN_RUN_23, device_param, pws_cnt, false); + run_kernel (KERN_RUN_23, device_param, pws_cnt, false, 0); hc_clEnqueueReadBuffer (data.ocl, device_param->command_queue, device_param->d_hooks, CL_TRUE, 0, device_param->size_hooks, device_param->hooks_buf, 0, NULL, NULL); @@ -2811,7 +3008,7 @@ static void choose_kernel (hc_device_param_t *device_param, const uint attack_ex hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_hooks, CL_TRUE, 0, device_param->size_hooks, device_param->hooks_buf, 0, NULL, NULL); } - run_kernel (KERN_RUN_3, device_param, pws_cnt, false); + run_kernel (KERN_RUN_3, device_param, pws_cnt, false, 0); } } @@ -2905,7 +3102,7 @@ static void run_copy (hc_device_param_t *device_param, const uint pws_cnt) static double try_run (hc_device_param_t *device_param, const u32 kernel_accel, const u32 kernel_loops) { - const u32 kernel_power = device_param->device_processors * device_param->kernel_threads * kernel_accel; + const u32 kernel_power_try = device_param->device_processors * device_param->kernel_threads * kernel_accel; device_param->kernel_params_buf32[25] = 0; device_param->kernel_params_buf32[26] = kernel_loops; // not a bug, both need to be set @@ -2913,11 +3110,11 @@ static double try_run (hc_device_param_t *device_param, const u32 kernel_accel, if (data.attack_exec == ATTACK_EXEC_INSIDE_KERNEL) { - run_kernel (KERN_RUN_1, device_param, kernel_power, true); + run_kernel (KERN_RUN_1, device_param, kernel_power_try, true, 0); } else { - run_kernel (KERN_RUN_2, device_param, kernel_power, true); + run_kernel (KERN_RUN_2, device_param, kernel_power_try, true, 0); } const double exec_ms_prev = get_avg_exec_time (device_param, 1); @@ -2944,10 +3141,13 @@ static void autotune (hc_device_param_t *device_param) if ((kernel_loops_min == kernel_loops_max) && (kernel_accel_min == kernel_accel_max)) { - try_run (device_param, kernel_accel, kernel_loops); - try_run (device_param, kernel_accel, kernel_loops); - try_run (device_param, kernel_accel, kernel_loops); - try_run (device_param, kernel_accel, kernel_loops); + if (data.hash_mode != 2000) + { + try_run (device_param, kernel_accel, kernel_loops); + try_run (device_param, kernel_accel, kernel_loops); + try_run (device_param, kernel_accel, kernel_loops); + try_run (device_param, kernel_accel, kernel_loops); + } device_param->kernel_accel = kernel_accel; device_param->kernel_loops = kernel_loops; @@ -2977,15 +3177,17 @@ static void autotune (hc_device_param_t *device_param) device_param->pws_buf[i].pw_len = 7 + (i & 7); } + hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_pws_buf, CL_TRUE, 0, kernel_power_max * sizeof (pw_t), device_param->pws_buf, 0, NULL, NULL); + } + + if (data.attack_exec == ATTACK_EXEC_INSIDE_KERNEL) + { if (data.kernel_rules_cnt > 1) { - hc_clEnqueueCopyBuffer (data.ocl, device_param->command_queue, device_param->d_rules, device_param->d_rules_c, 0, 0, kernel_loops_max * sizeof (kernel_rule_t), 0, NULL, NULL); + hc_clEnqueueCopyBuffer (data.ocl, device_param->command_queue, device_param->d_rules, device_param->d_rules_c, 0, 0, MIN (kernel_loops_max, KERNEL_RULES) * sizeof (kernel_rule_t), 0, NULL, NULL); } - - hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_pws_buf, CL_TRUE, 0, kernel_power_max * sizeof (pw_t), device_param->pws_buf, 0, NULL, NULL); } - - if (data.attack_exec == ATTACK_EXEC_OUTSIDE_KERNEL) + else { run_kernel_amp (device_param, kernel_power_max); } @@ -3131,6 +3333,10 @@ static void autotune (hc_device_param_t *device_param) memset (device_param->exec_ms, 0, EXEC_CACHE * sizeof (double)); + memset (device_param->exec_us_prev1, 0, EXPECTED_ITERATIONS * sizeof (double)); + memset (device_param->exec_us_prev2, 0, EXPECTED_ITERATIONS * sizeof (double)); + memset (device_param->exec_us_prev3, 0, EXPECTED_ITERATIONS * sizeof (double)); + // store device_param->kernel_accel = kernel_accel; @@ -3248,9 +3454,16 @@ static void run_cracker (hc_device_param_t *device_param, const uint pws_cnt) if (data.devices_status == STATUS_QUIT) break; if (data.devices_status == STATUS_BYPASS) break; + uint fast_iteration = 0; + uint innerloop_left = innerloop_cnt - innerloop_pos; - if (innerloop_left > innerloop_step) innerloop_left = innerloop_step; + if (innerloop_left > innerloop_step) + { + innerloop_left = innerloop_step; + + fast_iteration = 1; + } device_param->innerloop_pos = innerloop_pos; device_param->innerloop_left = innerloop_left; @@ -3408,7 +3621,7 @@ static void run_cracker (hc_device_param_t *device_param, const uint pws_cnt) hc_timer_set (&device_param->timer_speed); } - choose_kernel (device_param, data.attack_exec, data.attack_mode, data.opts_type, salt_buf, highest_pw_len, pws_cnt); + choose_kernel (device_param, data.attack_exec, data.attack_mode, data.opts_type, salt_buf, highest_pw_len, pws_cnt, fast_iteration); if (data.devices_status == STATUS_STOP_AT_CHECKPOINT) check_checkpoint (); @@ -3420,7 +3633,10 @@ static void run_cracker (hc_device_param_t *device_param, const uint pws_cnt) * result */ - check_cracked (device_param, salt_pos); + if (data.benchmark == 0) + { + check_cracked (device_param, salt_pos); + } /** * progress @@ -3814,7 +4030,7 @@ static void *thread_monitor (void *p) uint status_left = data.status_timer; #ifdef HAVE_HWMON - uint hwmon_check = 0; + uint hwmon_check = 0; int slowdown_warnings = 0; @@ -3882,7 +4098,7 @@ static void *thread_monitor (void *p) #ifdef HAVE_HWMON - if (1) + if (hwmon_check == 1) { hc_thread_mutex_lock (mux_adl); @@ -3892,35 +4108,47 @@ static void *thread_monitor (void *p) if (device_param->skipped) continue; - if ((data.devices_param[device_id].device_type & CL_DEVICE_TYPE_GPU) == 0) continue; + if (device_param->device_vendor_id == VENDOR_ID_NV) + { + if (data.hm_nvapi) + { + NV_GPU_PERF_POLICIES_INFO_PARAMS_V1 perfPolicies_info = { 0 }; + NV_GPU_PERF_POLICIES_STATUS_PARAMS_V1 perfPolicies_status = { 0 }; - const int temperature = hm_get_temperature_with_device_id (device_id); + perfPolicies_info.version = MAKE_NVAPI_VERSION (NV_GPU_PERF_POLICIES_INFO_PARAMS_V1, 1); + perfPolicies_status.version = MAKE_NVAPI_VERSION (NV_GPU_PERF_POLICIES_STATUS_PARAMS_V1, 1); - const int threshold = data.hm_device[device_id].gpu_temp_threshold_slowdown; + hm_NvAPI_GPU_GetPerfPoliciesInfo (data.hm_nvapi, data.hm_device[device_id].nvapi, &perfPolicies_info); - if (temperature >= threshold) - { - if (slowdown_warnings < 3) - { - if (data.quiet == 0) clear_prompt (); + perfPolicies_status.info_value = perfPolicies_info.info_value; - log_info ("WARNING: Drivers temperature threshold (%dc) hit on GPU #%d, expect performance to drop...", threshold, device_id + 1); + hm_NvAPI_GPU_GetPerfPoliciesStatus (data.hm_nvapi, data.hm_device[device_id].nvapi, &perfPolicies_status); - if (slowdown_warnings == 2) + if (perfPolicies_status.throttle & 2) { - log_info (""); - } + if (slowdown_warnings < 3) + { + if (data.quiet == 0) clear_prompt (); + + log_info ("WARNING: Drivers temperature threshold hit on GPU #%d, expect performance to drop...", device_id + 1); + + if (slowdown_warnings == 2) + { + log_info (""); + } - if (data.quiet == 0) fprintf (stdout, "%s", PROMPT); - if (data.quiet == 0) fflush (stdout); + if (data.quiet == 0) fprintf (stdout, "%s", PROMPT); + if (data.quiet == 0) fflush (stdout); - slowdown_warnings++; + slowdown_warnings++; + } + } + else + { + slowdown_warnings = 0; + } } } - else - { - slowdown_warnings = 0; - } } hc_thread_mutex_unlock (mux_adl); @@ -4003,11 +4231,17 @@ static void *thread_monitor (void *p) { if (device_param->device_vendor_id == VENDOR_ID_AMD) { - hm_set_fanspeed_with_device_id_amd (device_id, fan_speed_new, 1); + hm_set_fanspeed_with_device_id_adl (device_id, fan_speed_new, 1); } else if (device_param->device_vendor_id == VENDOR_ID_NV) { + #ifdef _WIN + hm_set_fanspeed_with_device_id_nvapi (device_id, fan_speed_new, 1); + #endif + #ifdef _LINUX + hm_set_fanspeed_with_device_id_xnvctrl (device_id, fan_speed_new); + #endif } fan_speed_chgd[device_id] = 1; @@ -4407,8 +4641,8 @@ static void *thread_outfile_remove (void *p) static void pw_add (hc_device_param_t *device_param, const u8 *pw_buf, const int pw_len) { - if (device_param->pws_cnt < device_param->kernel_power) - { + //if (device_param->pws_cnt < device_param->kernel_power) + //{ pw_t *pw = (pw_t *) device_param->pws_buf + device_param->pws_cnt; u8 *ptr = (u8 *) pw->i; @@ -4420,16 +4654,55 @@ static void pw_add (hc_device_param_t *device_param, const u8 *pw_buf, const int pw->pw_len = pw_len; device_param->pws_cnt++; + //} + //else + //{ + // fprintf (stderr, "BUG pw_add()!!\n"); + // + // return; + //} +} + +static void set_kernel_power_final (const u64 kernel_power_final) +{ + if (data.quiet == 0) + { + clear_prompt (); + + //log_info (""); + + log_info ("INFO: approaching final keyspace, workload adjusted"); + log_info (""); + + fprintf (stdout, "%s", PROMPT); + + fflush (stdout); } - else + + data.kernel_power_final = kernel_power_final; +} + +static u32 get_power (hc_device_param_t *device_param) +{ + const u64 kernel_power_final = data.kernel_power_final; + + if (kernel_power_final) { - fprintf (stderr, "BUG pw_add()!!\n"); + const double device_factor = (double) device_param->hardware_power / data.hardware_power_all; - return; + const u64 words_left_device = CEIL ((double) kernel_power_final * device_factor); + + // work should be at least the hardware power available without any accelerator + + const u64 work = MAX (words_left_device, device_param->hardware_power); + + return work; } + + return device_param->kernel_power; } -static uint get_work (hc_device_param_t *device_param, const u64 max, const bool allow_div) +static uint get_work (hc_device_param_t *device_param, const u64 max) { hc_thread_mutex_lock (mux_dispatcher); @@ -4438,33 +4711,19 @@ static uint get_work (hc_device_param_t *device_param, const u64 max, const bool device_param->words_off = words_cur; + const u64 kernel_power_all = data.kernel_power_all; + const u64 words_left = words_base - words_cur; - if (allow_div) + if (words_left < kernel_power_all) { - if (data.kernel_power_all > words_left) - { - if (data.kernel_power_div == 0) - { - data.kernel_power_div = find_kernel_power_div (words_left, data.kernel_power_all); - } - } - - if (data.kernel_power_div) + if (data.kernel_power_final == 0) { - if (device_param->kernel_power == device_param->kernel_power_user) - { - const u32 kernel_power_new = (float) device_param->kernel_power * data.kernel_power_div; - - if (kernel_power_new < device_param->kernel_power) - { - device_param->kernel_power = kernel_power_new; - } - } + set_kernel_power_final (words_left); } } - const uint kernel_power = device_param->kernel_power; + const u32 kernel_power = get_power (device_param); uint work = MIN (words_left, kernel_power); @@ -4477,7 +4736,7 @@ static uint get_work (hc_device_param_t *device_param, const u64 max, const bool return work; } -static void *thread_calc_stdin (void *p) +static void *thread_autotune (void *p) { hc_device_param_t *device_param = (hc_device_param_t *) p; @@ -4485,11 +4744,18 @@ static void *thread_calc_stdin (void *p) autotune (device_param); - char *buf = (char *) mymalloc (HCBUFSIZ); + return NULL; +} - const uint attack_kern = data.attack_kern; +static void *thread_calc_stdin (void *p) +{ + hc_device_param_t *device_param = (hc_device_param_t *) p; + + if (device_param->skipped) return NULL; + + char *buf = (char *) mymalloc (HCBUFSIZ); - const uint kernel_power = device_param->kernel_power; + 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)) { @@ -4504,7 +4770,7 @@ static void *thread_calc_stdin (void *p) uint words_cur = 0; - while (words_cur < kernel_power) + while (words_cur < device_param->kernel_power) { char *line_buf = fgets (buf, HCBUFSIZ - 1, stdin); @@ -4538,6 +4804,8 @@ static void *thread_calc_stdin (void *p) continue; } + // hmm that's always the case, or? + if (attack_kern == ATTACK_KERN_STRAIGHT) { if ((line_len < data.pw_min) || (line_len > data.pw_max)) @@ -4554,25 +4822,6 @@ static void *thread_calc_stdin (void *p) continue; } } - else if (attack_kern == ATTACK_KERN_COMBI) - { - // do not check if minimum restriction is satisfied (line_len >= data.pw_min) here - // since we still need to combine the plains - - if (line_len > data.pw_max) - { - hc_thread_mutex_lock (mux_counter); - - for (uint salt_pos = 0; salt_pos < data.salts_cnt; salt_pos++) - { - data.words_progress_rejected[salt_pos] += data.combs_cnt; - } - - hc_thread_mutex_unlock (mux_counter); - - continue; - } - } pw_add (device_param, (u8 *) line_buf, line_len); @@ -4631,8 +4880,6 @@ static void *thread_calc (void *p) if (device_param->skipped) return NULL; - autotune (device_param); - const uint attack_mode = data.attack_mode; const uint attack_kern = data.attack_kern; @@ -4640,7 +4887,7 @@ static void *thread_calc (void *p) { while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT)) { - const uint work = get_work (device_param, -1, true); + const uint work = get_work (device_param, -1); if (work == 0) break; @@ -4755,18 +5002,16 @@ static void *thread_calc (void *p) u64 words_off = 0; u64 words_fin = 0; - bool allow_div = true; - u64 max = -1; while (max) { - const uint work = get_work (device_param, max, allow_div); - - allow_div = false; + const uint work = get_work (device_param, max); if (work == 0) break; + max = 0; + words_off = device_param->words_off; words_fin = words_off + work; @@ -4775,8 +5020,6 @@ static void *thread_calc (void *p) for ( ; words_cur < words_off; words_cur++) get_next_word (wl_data, fd, &line_buf, &line_len); - max = 0; - for ( ; words_cur < words_fin; words_cur++) { get_next_word (wl_data, fd, &line_buf, &line_len); @@ -4957,11 +5200,11 @@ static void weak_hash_check (hc_device_param_t *device_param, const uint salt_po if (data.attack_exec == ATTACK_EXEC_INSIDE_KERNEL) { - run_kernel (KERN_RUN_1, device_param, 1, false); + run_kernel (KERN_RUN_1, device_param, 1, false, 0); } else { - run_kernel (KERN_RUN_1, device_param, 1, false); + run_kernel (KERN_RUN_1, device_param, 1, false, 0); uint loop_step = 16; @@ -4976,10 +5219,10 @@ static void weak_hash_check (hc_device_param_t *device_param, const uint salt_po device_param->kernel_params_buf32[25] = loop_pos; device_param->kernel_params_buf32[26] = loop_left; - run_kernel (KERN_RUN_2, device_param, 1, false); + run_kernel (KERN_RUN_2, device_param, 1, false, 0); } - run_kernel (KERN_RUN_3, device_param, 1, false); + run_kernel (KERN_RUN_3, device_param, 1, false, 0); } /** @@ -5363,7 +5606,7 @@ static uint hlfmt_detect (FILE *fp, uint max_check) // wrapper around mymalloc for ADL #if defined(HAVE_HWMON) -void *__stdcall ADL_Main_Memory_Alloc (const int iSize) +void *HC_API_CALL ADL_Main_Memory_Alloc (const int iSize) { return mymalloc (iSize); } @@ -5385,8 +5628,6 @@ static uint generate_bitmaps (const uint digests_cnt, const uint dgst_size, cons for (uint i = 0; i < digests_cnt; i++) { - if (data.digests_shown[i] == 1) continue; // can happen with potfile - uint *digest_ptr = (uint *) digests_buf_ptr; digests_buf_ptr += dgst_size; @@ -5421,7 +5662,7 @@ static uint generate_bitmaps (const uint digests_cnt, const uint dgst_size, cons * main */ -#ifdef _WIN +#ifdef WIN void SetConsoleWindowSize (const int x) { HANDLE h = GetStdHandle (STD_OUTPUT_HANDLE); @@ -5438,8 +5679,8 @@ void SetConsoleWindowSize (const int x) COORD co; - co.X = sr->Right + 1; - co.Y = sr->Bottom + 1; + co.X = sr->Right + 1; + co.Y = 9999; if (!SetConsoleScreenBufferSize (h, co)) return; @@ -5449,7 +5690,7 @@ void SetConsoleWindowSize (const int x) int main (int argc, char **argv) { - #ifdef _WIN + #ifdef WIN SetConsoleWindowSize (132); #endif @@ -5518,6 +5759,7 @@ int main (int argc, char **argv) uint version = VERSION; uint quiet = QUIET; uint benchmark = BENCHMARK; + uint stdout_flag = STDOUT_FLAG; uint show = SHOW; uint left = LEFT; uint username = USERNAME; @@ -5577,6 +5819,7 @@ int main (int argc, char **argv) uint workload_profile = WORKLOAD_PROFILE; uint kernel_accel = KERNEL_ACCEL; uint kernel_loops = KERNEL_LOOPS; + uint nvidia_spin_damp = NVIDIA_SPIN_DAMP; uint gpu_temp_disable = GPU_TEMP_DISABLE; #ifdef HAVE_HWMON uint gpu_temp_abort = GPU_TEMP_ABORT; @@ -5615,6 +5858,7 @@ int main (int argc, char **argv) #define IDX_FORCE 0xff08 #define IDX_RUNTIME 0xff09 #define IDX_BENCHMARK 'b' + #define IDX_STDOUT_FLAG 0xff77 #define IDX_HASH_MODE 'm' #define IDX_ATTACK_MODE 'a' #define IDX_RP_FILE 'r' @@ -5654,6 +5898,7 @@ int main (int argc, char **argv) #define IDX_WORKLOAD_PROFILE 'w' #define IDX_KERNEL_ACCEL 'n' #define IDX_KERNEL_LOOPS 'u' + #define IDX_NVIDIA_SPIN_DAMP 0xff79 #define IDX_GPU_TEMP_DISABLE 0xff29 #define IDX_GPU_TEMP_ABORT 0xff30 #define IDX_GPU_TEMP_RETAIN 0xff31 @@ -5695,6 +5940,7 @@ int main (int argc, char **argv) {"outfile-check-dir", required_argument, 0, IDX_OUTFILE_CHECK_DIR}, {"force", no_argument, 0, IDX_FORCE}, {"benchmark", no_argument, 0, IDX_BENCHMARK}, + {"stdout", no_argument, 0, IDX_STDOUT_FLAG}, {"restore", no_argument, 0, IDX_RESTORE}, {"restore-disable", no_argument, 0, IDX_RESTORE_DISABLE}, {"status", no_argument, 0, IDX_STATUS}, @@ -5732,6 +5978,7 @@ int main (int argc, char **argv) {"workload-profile", required_argument, 0, IDX_WORKLOAD_PROFILE}, {"kernel-accel", required_argument, 0, IDX_KERNEL_ACCEL}, {"kernel-loops", required_argument, 0, IDX_KERNEL_LOOPS}, + {"nvidia-spin-damp", required_argument, 0, IDX_NVIDIA_SPIN_DAMP}, {"gpu-temp-disable", no_argument, 0, IDX_GPU_TEMP_DISABLE}, #ifdef HAVE_HWMON {"gpu-temp-abort", required_argument, 0, IDX_GPU_TEMP_ABORT}, @@ -5951,6 +6198,7 @@ int main (int argc, char **argv) uint runtime_chgd = 0; uint kernel_loops_chgd = 0; uint kernel_accel_chgd = 0; + uint nvidia_spin_damp_chgd = 0; uint attack_mode_chgd = 0; uint outfile_format_chgd = 0; uint rp_gen_seed_chgd = 0; @@ -5991,6 +6239,7 @@ int main (int argc, char **argv) case IDX_LIMIT: limit = atoll (optarg); break; case IDX_KEYSPACE: keyspace = 1; break; case IDX_BENCHMARK: benchmark = 1; break; + case IDX_STDOUT_FLAG: stdout_flag = 1; break; case IDX_RESTORE: break; case IDX_RESTORE_DISABLE: restore_disable = 1; break; case IDX_STATUS: status = 1; break; @@ -6038,6 +6287,8 @@ int main (int argc, char **argv) kernel_accel_chgd = 1; break; case IDX_KERNEL_LOOPS: kernel_loops = atoi (optarg); kernel_loops_chgd = 1; break; + case IDX_NVIDIA_SPIN_DAMP: nvidia_spin_damp = atoi (optarg); + nvidia_spin_damp_chgd = 1; break; case IDX_GPU_TEMP_DISABLE: gpu_temp_disable = 1; break; #ifdef HAVE_HWMON case IDX_GPU_TEMP_ABORT: gpu_temp_abort = atoi (optarg); break; @@ -6090,6 +6341,9 @@ int main (int argc, char **argv) { log_info ("%s (%s) starting in benchmark-mode...", PROGNAME, VERSION_TAG); log_info (""); + log_info ("Note: Reported benchmark cracking speed = real cracking speed"); + log_info ("To verify, run hashcat like this: only_one_hash.txt -a 3 -w 3 ?b?b?b?b?b?b?b"); + log_info (""); } else { @@ -6101,6 +6355,10 @@ int main (int argc, char **argv) log_info ("%s (%s) starting in restore-mode...", PROGNAME, VERSION_TAG); log_info (""); } + else if (stdout_flag == 1) + { + // do nothing + } else { log_info ("%s (%s) starting...", PROGNAME, VERSION_TAG); @@ -6126,7 +6384,7 @@ int main (int argc, char **argv) return (-1); } - if (hash_mode_chgd && hash_mode > 13799) // just added to remove compiler warnings for hash_mode_chgd + if (hash_mode_chgd && hash_mode > 13800) // just added to remove compiler warnings for hash_mode_chgd { log_error ("ERROR: Invalid hash-type specified"); @@ -6364,8 +6622,32 @@ int main (int argc, char **argv) case ATTACK_MODE_HYBRID2: attack_kern = ATTACK_KERN_COMBI; break; } - if (benchmark == 0) + if (benchmark == 1) + { + if (myargv[optind] != 0) + { + log_error ("ERROR: Invalid argument for benchmark mode specified"); + + return (-1); + } + + if (attack_mode_chgd == 1) + { + if (attack_mode != ATTACK_MODE_BF) + { + log_error ("ERROR: Only attack-mode 3 allowed in benchmark mode"); + + return (-1); + } + } + } + else { + if (stdout_flag == 1) // no hash here + { + optind--; + } + if (keyspace == 1) { int num_additional_params = 1; @@ -6423,25 +6705,6 @@ int main (int argc, char **argv) return (-1); } } - else - { - if (myargv[optind] != 0) - { - log_error ("ERROR: Invalid argument for benchmark mode specified"); - - return (-1); - } - - if (attack_mode_chgd == 1) - { - if (attack_mode != ATTACK_MODE_BF) - { - log_error ("ERROR: Only attack-mode 3 allowed in benchmark mode"); - - return (-1); - } - } - } if (skip != 0 && limit != 0) { @@ -6474,6 +6737,24 @@ int main (int argc, char **argv) quiet = 1; } + if (stdout_flag == 1) + { + status_timer = 0; + restore_timer = 0; + restore_disable = 1; + restore = 0; + potfile_disable = 1; + weak_hash_threshold = 0; + gpu_temp_disable = 1; + hash_mode = 2000; + quiet = 1; + outfile_format = OUTFILE_FMT_PLAIN; + kernel_accel = 1024; + kernel_loops = 1024; + force = 1; + outfile_check_timer = 0; + } + if (remove_timer_chgd == 1) { if (remove == 0) @@ -6566,6 +6847,14 @@ int main (int argc, char **argv) weak_hash_threshold = 0; } + if (nvidia_spin_damp > 100) + { + log_error ("ERROR: setting --nvidia-spin-damp must be between 0 and 100 (inclusive)"); + + return (-1); + } + + /** * induction directory */ @@ -6820,12 +7109,14 @@ int main (int argc, char **argv) logfile_top_uint (attack_mode); logfile_top_uint (attack_kern); logfile_top_uint (benchmark); + logfile_top_uint (stdout_flag); logfile_top_uint (bitmap_min); logfile_top_uint (bitmap_max); logfile_top_uint (debug_mode); logfile_top_uint (force); logfile_top_uint (kernel_accel); logfile_top_uint (kernel_loops); + logfile_top_uint (nvidia_spin_damp); logfile_top_uint (gpu_temp_disable); #ifdef HAVE_HWMON logfile_top_uint (gpu_temp_abort); @@ -6945,15 +7236,21 @@ int main (int argc, char **argv) restore_disable = 1; potfile_disable = 1; weak_hash_threshold = 0; + nvidia_spin_damp = 0; gpu_temp_disable = 1; + outfile_check_timer = 0; #ifdef HAVE_HWMON - powertune_enable = 1; + if (powertune_enable == 1) + { + gpu_temp_disable = 0; + } #endif - data.status_timer = status_timer; - data.restore_timer = restore_timer; - data.restore_disable = restore_disable; + data.status_timer = status_timer; + data.restore_timer = restore_timer; + data.restore_disable = restore_disable; + data.outfile_check_timer = outfile_check_timer; /** * force attack mode to be bruteforce @@ -8385,6 +8682,21 @@ int main (int argc, char **argv) dgst_pos3 = 3; break; + case 2000: hash_type = HASH_TYPE_STDOUT; + salt_type = SALT_TYPE_NONE; + attack_exec = ATTACK_EXEC_INSIDE_KERNEL; + opts_type = OPTS_TYPE_PT_GENERATE_LE; + kern_type = 0; + dgst_size = DGST_SIZE_4_4; + parse_func = NULL; + sort_by_digest = NULL; + opti_type = 0; + dgst_pos0 = 0; + dgst_pos1 = 0; + dgst_pos2 = 0; + dgst_pos3 = 0; + break; + case 2100: hash_type = HASH_TYPE_DCC2; salt_type = SALT_TYPE_EMBEDDED; attack_exec = ATTACK_EXEC_OUTSIDE_KERNEL; @@ -10799,6 +11111,25 @@ int main (int argc, char **argv) dgst_pos3 = 3; break; + case 13800: hash_type = HASH_TYPE_SHA256; + salt_type = SALT_TYPE_EMBEDDED; + attack_exec = ATTACK_EXEC_INSIDE_KERNEL; + opts_type = OPTS_TYPE_PT_GENERATE_BE + | OPTS_TYPE_PT_UNICODE; + kern_type = KERN_TYPE_WIN8PHONE; + dgst_size = DGST_SIZE_4_8; + parse_func = win8phone_parse_hash; + sort_by_digest = sort_by_digest_4_8; + opti_type = OPTI_TYPE_ZERO_BYTE + | OPTI_TYPE_PRECOMPUTE_INIT + | OPTI_TYPE_EARLY_SKIP + | OPTI_TYPE_NOT_ITERATED + | OPTI_TYPE_RAW_HASH; + dgst_pos0 = 3; + dgst_pos1 = 7; + dgst_pos2 = 2; + dgst_pos3 = 6; + break; default: usage_mini_print (PROGNAME); return (-1); } @@ -10925,6 +11256,7 @@ int main (int argc, char **argv) case 13761: esalt_size = sizeof (tc_t); break; case 13762: esalt_size = sizeof (tc_t); break; case 13763: esalt_size = sizeof (tc_t); break; + case 13800: esalt_size = sizeof (win8phone_t); break; } data.esalt_size = esalt_size; @@ -11258,7 +11590,7 @@ int main (int argc, char **argv) break; case 7400: if (pw_max > 16) pw_max = 16; break; - case 7500: if (pw_max > 8) pw_max = 8; + case 7700: if (pw_max > 8) pw_max = 8; break; case 7900: if (pw_max > 48) pw_max = 48; break; @@ -11327,7 +11659,7 @@ int main (int argc, char **argv) uint hashes_avail = 0; - if (benchmark == 0) + if ((benchmark == 0) && (stdout_flag == 0)) { struct stat f; @@ -11509,6 +11841,10 @@ int main (int argc, char **argv) { // useless to read hash file for keyspace, cheat a little bit w/ optind } + else if (stdout_flag == 1) + { + // useless to read hash file for stdout, cheat a little bit w/ optind + } else if (hashes_avail == 0) { } @@ -12318,7 +12654,7 @@ int main (int argc, char **argv) return (0); } - if (keyspace == 0) + if ((keyspace == 0) && (stdout_flag == 0)) { if (hashes_cnt == 0) { @@ -13321,6 +13657,11 @@ int main (int argc, char **argv) * OpenCL devices: simply push all devices from all platforms into the same device array */ + int need_adl = 0; + int need_nvapi = 0; + int need_nvml = 0; + int need_xnvctrl = 0; + hc_device_param_t *devices_param = (hc_device_param_t *) mycalloc (DEVICES_MAX, sizeof (hc_device_param_t)); data.devices_param = devices_param; @@ -13577,7 +13918,7 @@ int main (int argc, char **argv) if (device_endian_little == CL_FALSE) { - if (data.quiet == 0) log_info ("Device #%u: WARNING: not little endian device", device_id + 1); + log_info ("Device #%u: WARNING: not little endian device", device_id + 1); device_param->skipped = 1; } @@ -13590,7 +13931,7 @@ int main (int argc, char **argv) if (device_available == CL_FALSE) { - if (data.quiet == 0) log_info ("Device #%u: WARNING: device not available", device_id + 1); + log_info ("Device #%u: WARNING: device not available", device_id + 1); device_param->skipped = 1; } @@ -13603,7 +13944,7 @@ int main (int argc, char **argv) if (device_compiler_available == CL_FALSE) { - if (data.quiet == 0) log_info ("Device #%u: WARNING: device no compiler available", device_id + 1); + log_info ("Device #%u: WARNING: device no compiler available", device_id + 1); device_param->skipped = 1; } @@ -13616,7 +13957,7 @@ int main (int argc, char **argv) if ((device_execution_capabilities & CL_EXEC_KERNEL) == 0) { - if (data.quiet == 0) log_info ("Device #%u: WARNING: device does not support executing kernels", device_id + 1); + log_info ("Device #%u: WARNING: device does not support executing kernels", device_id + 1); device_param->skipped = 1; } @@ -13633,14 +13974,14 @@ int main (int argc, char **argv) if (strstr (device_extensions, "base_atomics") == 0) { - if (data.quiet == 0) log_info ("Device #%u: WARNING: device does not support base atomics", device_id + 1); + log_info ("Device #%u: WARNING: device does not support base atomics", device_id + 1); device_param->skipped = 1; } if (strstr (device_extensions, "byte_addressable_store") == 0) { - if (data.quiet == 0) log_info ("Device #%u: WARNING: device does not support byte addressable store", device_id + 1); + log_info ("Device #%u: WARNING: device does not support byte addressable store", device_id + 1); device_param->skipped = 1; } @@ -13655,7 +13996,7 @@ int main (int argc, char **argv) if (device_local_mem_size < 32768) { - if (data.quiet == 0) log_info ("Device #%u: WARNING: device local mem size is too small", device_id + 1); + log_info ("Device #%u: WARNING: device local mem size is too small", device_id + 1); device_param->skipped = 1; } @@ -13670,9 +14011,16 @@ int main (int argc, char **argv) { if (device_param->device_vendor_id == VENDOR_ID_AMD_USE_INTEL) { - if (data.quiet == 0) log_info ("Device #%u: WARNING: not native intel opencl platform", device_id + 1); + if (data.force == 0) + { + if (algorithm_pos == 0) + { + log_info ("Device #%u: WARNING: not native intel opencl runtime, expect massive speed loss", device_id + 1); + log_info (" You can use --force to override this but do not post error reports if you do so"); + } - device_param->skipped = 1; + device_param->skipped = 1; + } } } @@ -13711,6 +14059,29 @@ int main (int argc, char **argv) // device_processor_cores + if (device_param->device_type & CL_DEVICE_TYPE_GPU) + { + if ((device_param->platform_vendor_id == VENDOR_ID_AMD) && (device_param->device_vendor_id == VENDOR_ID_AMD)) + { + need_adl = 1; + } + + if ((device_param->platform_vendor_id == VENDOR_ID_NV) && (device_param->device_vendor_id == VENDOR_ID_NV)) + { + need_nvml = 1; + + #ifdef _LINUX + need_xnvctrl = 1; + #endif + + #ifdef _WIN + need_nvapi = 1; + #endif + } + } + + // device_processor_cores + if (device_type & CL_DEVICE_TYPE_CPU) { cl_uint device_processor_cores = 1; @@ -13759,6 +14130,29 @@ int main (int argc, char **argv) device_param->sm_minor = sm_minor; device_param->sm_major = sm_major; + + // CPU burning loop damper + // Value is given as number between 0-100 + // By default 100% + + device_param->nvidia_spin_damp = (double) nvidia_spin_damp; + + if (nvidia_spin_damp_chgd == 0) + { + if (data.attack_mode == ATTACK_MODE_STRAIGHT) + { + /** + * the workaround is not a friend of rule based attacks + * the words from the wordlist combined with fast and slow rules cause + * fluctuations which cause inaccurate wait time estimations + * using a reduced damping percentage almost compensates this + */ + + device_param->nvidia_spin_damp = 64; + } + } + + device_param->nvidia_spin_damp /= 100; } else { @@ -13980,75 +14374,120 @@ int main (int argc, char **argv) */ #ifdef HAVE_HWMON - hm_attrs_t hm_adapters_nv[DEVICES_MAX] = { { { 0 }, 0, 0, 0, 0, 0 } }; - hm_attrs_t hm_adapters_amd[DEVICES_MAX] = { { { 0 }, 0, 0, 0, 0, 0 } }; + hm_attrs_t hm_adapters_adl[DEVICES_MAX] = { { 0 } }; + hm_attrs_t hm_adapters_nvapi[DEVICES_MAX] = { { 0 } }; + hm_attrs_t hm_adapters_nvml[DEVICES_MAX] = { { 0 } }; + hm_attrs_t hm_adapters_xnvctrl[DEVICES_MAX] = { { 0 } }; if (gpu_temp_disable == 0) { - NVML_PTR *nvml = (NVML_PTR *) mymalloc (sizeof (NVML_PTR)); + ADL_PTR *adl = (ADL_PTR *) mymalloc (sizeof (ADL_PTR)); + NVAPI_PTR *nvapi = (NVAPI_PTR *) mymalloc (sizeof (NVAPI_PTR)); + NVML_PTR *nvml = (NVML_PTR *) mymalloc (sizeof (NVML_PTR)); + XNVCTRL_PTR *xnvctrl = (XNVCTRL_PTR *) mymalloc (sizeof (XNVCTRL_PTR)); + + data.hm_adl = NULL; + data.hm_nvapi = NULL; + data.hm_nvml = NULL; + data.hm_xnvctrl = NULL; - if (nvml_init (nvml) == 0) - data.hm_nv = nvml; + if ((need_nvml == 1) && (nvml_init (nvml) == 0)) + { + data.hm_nvml = nvml; + } - if (data.hm_nv) + if (data.hm_nvml) { - if (hm_NVML_nvmlInit (data.hm_nv) == NVML_SUCCESS) + if (hm_NVML_nvmlInit (data.hm_nvml) == NVML_SUCCESS) { - HM_ADAPTER_NV nvGPUHandle[DEVICES_MAX] = { 0 }; + HM_ADAPTER_NVML nvmlGPUHandle[DEVICES_MAX] = { 0 }; - int tmp_in = hm_get_adapter_index_nv (nvGPUHandle); + int tmp_in = hm_get_adapter_index_nvml (nvmlGPUHandle); int tmp_out = 0; for (int i = 0; i < tmp_in; i++) { - hm_adapters_nv[tmp_out++].adapter_index.nv = nvGPUHandle[i]; + hm_adapters_nvml[tmp_out++].nvml = nvmlGPUHandle[i]; } for (int i = 0; i < tmp_out; i++) { unsigned int speed; - if (hm_NVML_nvmlDeviceGetFanSpeed (data.hm_nv, 1, hm_adapters_nv[i].adapter_index.nv, &speed) != NVML_ERROR_NOT_SUPPORTED) hm_adapters_nv[i].fan_get_supported = 1; + 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_nv, 1, hm_adapters_nv[i].adapter_index.nv, NVML_COMPUTEMODE_EXCLUSIVE_PROCESS); + hm_NVML_nvmlDeviceSetComputeMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_COMPUTEMODE_EXCLUSIVE_PROCESS); - hm_NVML_nvmlDeviceSetGpuOperationMode (data.hm_nv, 1, hm_adapters_nv[i].adapter_index.nv, NVML_GOM_ALL_ON); + hm_NVML_nvmlDeviceSetGpuOperationMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_GOM_ALL_ON); + } + } + } - unsigned int minLimit; - unsigned int maxLimit; + if ((need_nvapi == 1) && (nvapi_init (nvapi) == 0)) + { + data.hm_nvapi = nvapi; + } - if (hm_NVML_nvmlDeviceGetPowerManagementLimitConstraints (data.hm_nv, 1, hm_adapters_nv[i].adapter_index.nv, &minLimit, &maxLimit) == NVML_SUCCESS) - { - if (maxLimit > 0) - { - hm_NVML_nvmlDeviceSetPowerManagementLimit (data.hm_nv, 1, hm_adapters_nv[i].adapter_index.nv, maxLimit); - } - } + if (data.hm_nvapi) + { + if (hm_NvAPI_Initialize (data.hm_nvapi) == NVAPI_OK) + { + HM_ADAPTER_NVAPI nvGPUHandle[DEVICES_MAX] = { 0 }; + + int tmp_in = hm_get_adapter_index_nvapi (nvGPUHandle); + + int tmp_out = 0; + + for (int i = 0; i < tmp_in; i++) + { + hm_adapters_nvapi[tmp_out++].nvapi = nvGPUHandle[i]; } } } - data.hm_amd = NULL; + if ((need_xnvctrl == 1) && (xnvctrl_init (xnvctrl) == 0)) + { + data.hm_xnvctrl = xnvctrl; + } + + if (data.hm_xnvctrl) + { + if (hm_XNVCTRL_XOpenDisplay (data.hm_xnvctrl) == 0) + { + for (uint device_id = 0; device_id < data.devices_cnt; device_id++) + { + hc_device_param_t *device_param = &data.devices_param[device_id]; - ADL_PTR *adl = (ADL_PTR *) mymalloc (sizeof (ADL_PTR)); + if ((device_param->device_type & CL_DEVICE_TYPE_GPU) == 0) continue; - if (adl_init (adl) == 0) - data.hm_amd = adl; + hm_adapters_xnvctrl[device_id].xnvctrl = device_id; + + int speed = 0; + + if (get_fan_speed_current (data.hm_xnvctrl, device_id, &speed) == 0) hm_adapters_xnvctrl[device_id].fan_get_supported = 1; + } + } + } + + if ((need_adl == 1) && (adl_init (adl) == 0)) + { + data.hm_adl = adl; + } - if (data.hm_amd) + if (data.hm_adl) { - if (hm_ADL_Main_Control_Create (data.hm_amd, ADL_Main_Memory_Alloc, 0) == ADL_OK) + if (hm_ADL_Main_Control_Create (data.hm_adl, ADL_Main_Memory_Alloc, 0) == ADL_OK) { // total number of adapters int hm_adapters_num; - if (get_adapters_num_amd (data.hm_amd, &hm_adapters_num) != 0) return (-1); + if (get_adapters_num_adl (data.hm_adl, &hm_adapters_num) != 0) return (-1); // adapter info - LPAdapterInfo lpAdapterInfo = hm_get_adapter_info_amd (data.hm_amd, hm_adapters_num); + LPAdapterInfo lpAdapterInfo = hm_get_adapter_info_adl (data.hm_adl, hm_adapters_num); if (lpAdapterInfo == NULL) return (-1); @@ -14062,12 +14501,12 @@ int main (int argc, char **argv) { hc_thread_mutex_lock (mux_adl); - // hm_get_opencl_busid_devid (hm_adapters_amd, devices_all_cnt, devices_all); + // hm_get_opencl_busid_devid (hm_adapters_adl, devices_all_cnt, devices_all); - hm_get_adapter_index_amd (hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo); + hm_get_adapter_index_adl (hm_adapters_adl, valid_adl_device_list, num_adl_adapters, lpAdapterInfo); - hm_get_overdrive_version (data.hm_amd, hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo); - hm_check_fanspeed_control (data.hm_amd, hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo); + hm_get_overdrive_version (data.hm_adl, hm_adapters_adl, valid_adl_device_list, num_adl_adapters, lpAdapterInfo); + hm_check_fanspeed_control (data.hm_adl, hm_adapters_adl, valid_adl_device_list, num_adl_adapters, lpAdapterInfo); hc_thread_mutex_unlock (mux_adl); } @@ -14077,7 +14516,7 @@ int main (int argc, char **argv) } } - if (data.hm_amd == NULL && data.hm_nv == NULL) + if (data.hm_adl == NULL && data.hm_nvml == NULL && data.hm_xnvctrl == NULL) { gpu_temp_disable = 1; } @@ -14094,6 +14533,8 @@ int main (int argc, char **argv) int *od_power_control_status = (int *) mycalloc (data.devices_cnt, sizeof (int)); + unsigned int *nvml_power_limit = (unsigned int *) mycalloc (data.devices_cnt, sizeof (unsigned int)); + /** * User-defined GPU temp handling */ @@ -14164,7 +14605,7 @@ int main (int argc, char **argv) */ #ifdef HAVE_HWMON - if (gpu_temp_disable == 0 && data.hm_amd == NULL && data.hm_nv == NULL) + if (gpu_temp_disable == 0 && data.hm_adl == NULL && data.hm_nvml == NULL && data.hm_xnvctrl == NULL) { log_info ("Watchdog: Hardware Monitoring Interface not found on your system"); } @@ -14209,24 +14650,32 @@ int main (int argc, char **argv) const uint platform_devices_id = device_param->platform_devices_id; - if (device_param->device_vendor_id == VENDOR_ID_NV) + if (device_param->device_vendor_id == VENDOR_ID_AMD) { - memcpy (&data.hm_device[device_id], &hm_adapters_nv[platform_devices_id], sizeof (hm_attrs_t)); + data.hm_device[device_id].adl = hm_adapters_adl[platform_devices_id].adl; + data.hm_device[device_id].nvapi = 0; + data.hm_device[device_id].nvml = 0; + data.hm_device[device_id].xnvctrl = 0; + data.hm_device[device_id].od_version = hm_adapters_adl[platform_devices_id].od_version; + data.hm_device[device_id].fan_get_supported = hm_adapters_adl[platform_devices_id].fan_get_supported; + data.hm_device[device_id].fan_set_supported = 0; } - if (device_param->device_vendor_id == VENDOR_ID_AMD) + if (device_param->device_vendor_id == VENDOR_ID_NV) { - memcpy (&data.hm_device[device_id], &hm_adapters_amd[platform_devices_id], sizeof (hm_attrs_t)); + data.hm_device[device_id].adl = 0; + data.hm_device[device_id].nvapi = hm_adapters_nvapi[platform_devices_id].nvapi; + data.hm_device[device_id].nvml = hm_adapters_nvml[platform_devices_id].nvml; + data.hm_device[device_id].xnvctrl = hm_adapters_xnvctrl[platform_devices_id].xnvctrl; + data.hm_device[device_id].od_version = 0; + data.hm_device[device_id].fan_get_supported = hm_adapters_nvml[platform_devices_id].fan_get_supported; + data.hm_device[device_id].fan_set_supported = 0; } } } /** - * Temporary fix: - * with AMD r9 295x cards it seems that we need to set the powertune value just AFTER the ocl init stuff - * otherwise after hc_clCreateContext () etc, powertune value was set back to "normal" and cards unfortunately - * were not working @ full speed (setting hm_ADL_Overdrive_PowerControl_Set () here seems to fix the problem) - * Driver / ADL bug? + * powertune on user request */ if (powertune_enable == 1) @@ -14239,88 +14688,225 @@ int main (int argc, char **argv) if (device_param->skipped) continue; - if (data.hm_device[device_id].od_version == 6) + if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_AMD) { - // set powertune value only - - int powertune_supported = 0; - - int ADL_rc = 0; + /** + * Temporary fix: + * with AMD r9 295x cards it seems that we need to set the powertune value just AFTER the ocl init stuff + * otherwise after hc_clCreateContext () etc, powertune value was set back to "normal" and cards unfortunately + * were not working @ full speed (setting hm_ADL_Overdrive_PowerControl_Set () here seems to fix the problem) + * Driver / ADL bug? + */ - if ((ADL_rc = hm_ADL_Overdrive6_PowerControl_Caps (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK) + if (data.hm_device[device_id].od_version == 6) { - log_error ("ERROR: Failed to get ADL PowerControl Capabilities"); + int ADL_rc; - return (-1); - } + // check powertune capabilities first, if not available then skip device - if (powertune_supported != 0) - { - // powertune set - ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0}; + int powertune_supported = 0; - if ((ADL_rc = hm_ADL_Overdrive_PowerControlInfo_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune)) != ADL_OK) + if ((ADL_rc = hm_ADL_Overdrive6_PowerControl_Caps (data.hm_adl, data.hm_device[device_id].adl, &powertune_supported)) != ADL_OK) { - log_error ("ERROR: Failed to get current ADL PowerControl settings"); + log_error ("ERROR: Failed to get ADL PowerControl Capabilities"); return (-1); } - if ((ADL_rc = hm_ADL_Overdrive_PowerControl_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, powertune.iMaxValue)) != ADL_OK) + // first backup current value, we will restore it later + + if (powertune_supported != 0) { - log_error ("ERROR: Failed to set new ADL PowerControl values"); + // powercontrol settings - return (-1); - } - } - } - } + ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0}; - hc_thread_mutex_unlock (mux_adl); - } + if ((ADL_rc = hm_ADL_Overdrive_PowerControlInfo_Get (data.hm_adl, data.hm_device[device_id].adl, &powertune)) == ADL_OK) + { + ADL_rc = hm_ADL_Overdrive_PowerControl_Get (data.hm_adl, data.hm_device[device_id].adl, &od_power_control_status[device_id]); + } - #endif // HAVE_HWMON + if (ADL_rc != ADL_OK) + { + log_error ("ERROR: Failed to get current ADL PowerControl settings"); - #ifdef DEBUG - if (benchmark == 1) log_info ("Hashmode: %d", data.hash_mode); - #endif + return (-1); + } - if (data.quiet == 0) log_info_nn ("Initializing device kernels and memory..."); + if ((ADL_rc = hm_ADL_Overdrive_PowerControl_Set (data.hm_adl, data.hm_device[device_id].adl, powertune.iMaxValue)) != ADL_OK) + { + log_error ("ERROR: Failed to set new ADL PowerControl values"); - uint kernel_power_all = 0; + return (-1); + } - for (uint device_id = 0; device_id < data.devices_cnt; device_id++) - { - /** - * host buffer - */ + // clocks - hc_device_param_t *device_param = &data.devices_param[device_id]; + memset (&od_clock_mem_status[device_id], 0, sizeof (ADLOD6MemClockState)); - if (device_param->skipped) continue; + od_clock_mem_status[device_id].state.iNumberOfPerformanceLevels = 2; - /** - * device properties - */ + if ((ADL_rc = hm_ADL_Overdrive_StateInfo_Get (data.hm_adl, data.hm_device[device_id].adl, 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"); - 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; + return (-1); + } - /** - * create context for each device - */ + // Query capabilities only to see if profiles were not "damaged", if so output a warning but do accept the users profile settings - device_param->context = hc_clCreateContext (data.ocl, NULL, 1, &device_param->device, NULL, NULL); + ADLOD6Capabilities caps = {0, 0, 0, {0, 0, 0}, {0, 0, 0}, 0, 0}; - /** - * create command-queue - */ + if ((ADL_rc = hm_ADL_Overdrive_Capabilities_Get (data.hm_adl, data.hm_device[device_id].adl, &caps)) != ADL_OK) + { + log_error ("ERROR: Failed to get ADL device capabilities"); - // not supported with NV - // device_param->command_queue = hc_clCreateCommandQueueWithProperties (device_param->context, device_param->device, NULL); + return (-1); + } - device_param->command_queue = hc_clCreateCommandQueue (data.ocl, device_param->context, device_param->device, CL_QUEUE_PROFILING_ENABLE); + int engine_clock_max = caps.sEngineClockRange.iMax * 0.6666; + int memory_clock_max = caps.sMemoryClockRange.iMax * 0.6250; + + int warning_trigger_engine = (int) (0.25 * (float) engine_clock_max); + int warning_trigger_memory = (int) (0.25 * (float) memory_clock_max); + + int engine_clock_profile_max = od_clock_mem_status[device_id].state.aLevels[1].iEngineClock; + int memory_clock_profile_max = od_clock_mem_status[device_id].state.aLevels[1].iMemoryClock; + + // warning if profile has too low max values + + if ((engine_clock_max - engine_clock_profile_max) > warning_trigger_engine) + { + log_info ("WARN: the custom profile seems to have too low maximum engine clock values. You therefore may not reach full performance"); + } + + if ((memory_clock_max - memory_clock_profile_max) > warning_trigger_memory) + { + log_info ("WARN: the custom profile seems to have too low maximum memory clock values. You therefore may not reach full performance"); + } + + ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel)); + + performance_state->iNumberOfPerformanceLevels = 2; + + performance_state->aLevels[0].iEngineClock = engine_clock_profile_max; + performance_state->aLevels[1].iEngineClock = engine_clock_profile_max; + performance_state->aLevels[0].iMemoryClock = memory_clock_profile_max; + performance_state->aLevels[1].iMemoryClock = memory_clock_profile_max; + + if ((ADL_rc = hm_ADL_Overdrive_State_Set (data.hm_adl, data.hm_device[device_id].adl, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK) + { + log_info ("ERROR: Failed to set ADL performance state"); + + return (-1); + } + + local_free (performance_state); + } + + // set powertune value only + + if (powertune_supported != 0) + { + // powertune set + ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0}; + + if ((ADL_rc = hm_ADL_Overdrive_PowerControlInfo_Get (data.hm_adl, data.hm_device[device_id].adl, &powertune)) != ADL_OK) + { + log_error ("ERROR: Failed to get current ADL PowerControl settings"); + + return (-1); + } + + if ((ADL_rc = hm_ADL_Overdrive_PowerControl_Set (data.hm_adl, data.hm_device[device_id].adl, powertune.iMaxValue)) != ADL_OK) + { + log_error ("ERROR: Failed to set new ADL PowerControl values"); + + return (-1); + } + } + } + } + + if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_NV) + { + // first backup current value, we will restore it later + + unsigned int limit; + + int powertune_supported = 0; + + if (hm_NVML_nvmlDeviceGetPowerManagementLimit (data.hm_nvml, 0, data.hm_device[device_id].nvml, &limit) == NVML_SUCCESS) + { + powertune_supported = 1; + } + + // if backup worked, activate the maximum allowed + + if (powertune_supported != 0) + { + unsigned int minLimit; + unsigned int maxLimit; + + if (hm_NVML_nvmlDeviceGetPowerManagementLimitConstraints (data.hm_nvml, 0, data.hm_device[device_id].nvml, &minLimit, &maxLimit) == NVML_SUCCESS) + { + if (maxLimit > 0) + { + if (hm_NVML_nvmlDeviceSetPowerManagementLimit (data.hm_nvml, 0, data.hm_device[device_id].nvml, maxLimit) == NVML_SUCCESS) + { + // now we can be sure we need to reset later + + nvml_power_limit[device_id] = limit; + } + } + } + } + } + } + + hc_thread_mutex_unlock (mux_adl); + } + + #endif // HAVE_HWMON + + #ifdef DEBUG + if (benchmark == 1) log_info ("Hashmode: %d", data.hash_mode); + #endif + + if (data.quiet == 0) log_info_nn ("Initializing device kernels and memory..."); + + for (uint device_id = 0; device_id < data.devices_cnt; device_id++) + { + /** + * host buffer + */ + + hc_device_param_t *device_param = &data.devices_param[device_id]; + + if (device_param->skipped) continue; + + /** + * device properties + */ + + 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 + */ + + device_param->context = hc_clCreateContext (data.ocl, NULL, 1, &device_param->device, NULL, NULL); + + /** + * create command-queue + */ + + // not supported with NV + // device_param->command_queue = hc_clCreateCommandQueueWithProperties (device_param->context, device_param->device, NULL); + + device_param->command_queue = hc_clCreateCommandQueue (data.ocl, device_param->context, device_param->device, CL_QUEUE_PROFILING_ENABLE); /** * kernel threads: some algorithms need a fixed kernel-threads count @@ -14349,6 +14935,10 @@ int main (int argc, char **argv) if (hash_mode == 10500) kernel_threads = 64; // RC4 if (hash_mode == 13100) kernel_threads = 64; // RC4 + device_param->kernel_threads = kernel_threads; + + device_param->hardware_power = device_processors * kernel_threads; + /** * create input buffers on device : calculate size of fixed memory buffers */ @@ -14463,7 +15053,7 @@ int main (int argc, char **argv) * some algorithms need a fixed kernel-loops count */ - if (hash_mode == 1500) + if (hash_mode == 1500 && attack_mode == ATTACK_MODE_BF) { const u32 kernel_loops_fixed = 1024; @@ -14471,7 +15061,7 @@ int main (int argc, char **argv) device_param->kernel_loops_max = kernel_loops_fixed; } - if (hash_mode == 3000) + if (hash_mode == 3000 && attack_mode == ATTACK_MODE_BF) { const u32 kernel_loops_fixed = 1024; @@ -14638,7 +15228,9 @@ int main (int argc, char **argv) if ((opts_type & OPTS_TYPE_HOOK12) || (opts_type & OPTS_TYPE_HOOK23)) { - // none yet + switch (hash_mode) + { + } } // now check if all device-memory sizes which depend on the kernel_accel_max amplifier are within its boundaries @@ -14718,15 +15310,6 @@ int main (int argc, char **argv) device_param->size_tmps = size_tmps; device_param->size_hooks = size_hooks; - // do not confuse kernel_accel_max with kernel_accel here - - const u32 kernel_power = device_processors * kernel_threads * kernel_accel_max; - - device_param->kernel_threads = kernel_threads; - device_param->kernel_power_user = kernel_power; - - kernel_power_all += kernel_power; - /** * default building options */ @@ -15568,21 +16151,6 @@ int main (int argc, char **argv) #if defined(HAVE_HWMON) - /** - * Store thermal target temperature so we can send a notice to user - */ - - if (gpu_temp_disable == 0) - { - const int gpu_temp_threshold_slowdown = hm_get_threshold_slowdown_with_device_id (device_id); - const int gpu_temp_threshold_shutdown = hm_get_threshold_slowdown_with_device_id (device_id); - - data.hm_device[device_id].gpu_temp_threshold_slowdown = (gpu_temp_threshold_slowdown > 0) ? gpu_temp_threshold_slowdown : 10000; - data.hm_device[device_id].gpu_temp_threshold_shutdown = (gpu_temp_threshold_shutdown > 0) ? gpu_temp_threshold_shutdown : 10000; - - // we could use those numbers for gpu_temp_retain and gpu_temp_abort, too - } - /** * Store initial fanspeed if gpu_temp_retain is enabled */ @@ -15612,11 +16180,17 @@ int main (int argc, char **argv) if (device_param->device_vendor_id == VENDOR_ID_AMD) { - rc = hm_set_fanspeed_with_device_id_amd (device_id, fanspeed, 1); + rc = hm_set_fanspeed_with_device_id_adl (device_id, fanspeed, 1); } else if (device_param->device_vendor_id == VENDOR_ID_NV) { + #ifdef _WIN + rc = hm_set_fanspeed_with_device_id_nvapi (device_id, fanspeed, 1); + #endif + #ifdef _LINUX + rc = set_fan_control (data.hm_xnvctrl, data.hm_device[device_id].xnvctrl, NV_CTRL_GPU_COOLER_MANUAL_CONTROL_TRUE); + #endif } if (rc == 0) @@ -15625,7 +16199,7 @@ int main (int argc, char **argv) } else { - log_info ("WARNING: Failed to set initial fan speed for device #%u", device_id + 1); + //log_info ("WARNING: Failed to set initial fan speed for device #%u", device_id + 1); data.hm_device[device_id].fan_set_supported = 0; } @@ -15640,127 +16214,9 @@ int main (int argc, char **argv) } } - /** - * Store original powercontrol/clocks settings, set overdrive 6 performance tuning settings - */ - - if (powertune_enable == 1) // VENDOR_ID_AMD implied - { - hc_thread_mutex_lock (mux_adl); - - if (data.hm_device[device_id].od_version == 6) - { - int ADL_rc; - - // check powertune capabilities first, if not available then skip device - - int powertune_supported = 0; - - if ((ADL_rc = hm_ADL_Overdrive6_PowerControl_Caps (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK) - { - log_error ("ERROR: Failed to get ADL PowerControl Capabilities"); - - return (-1); - } - - if (powertune_supported != 0) - { - // powercontrol settings - - ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0}; - - if ((ADL_rc = hm_ADL_Overdrive_PowerControlInfo_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune)) == ADL_OK) - { - ADL_rc = hm_ADL_Overdrive_PowerControl_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &od_power_control_status[device_id]); - } - - if (ADL_rc != ADL_OK) - { - log_error ("ERROR: Failed to get current ADL PowerControl settings"); - - return (-1); - } - - if ((ADL_rc = hm_ADL_Overdrive_PowerControl_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, powertune.iMaxValue)) != ADL_OK) - { - log_error ("ERROR: Failed to set new ADL PowerControl values"); - - return (-1); - } - - // clocks - - memset (&od_clock_mem_status[device_id], 0, sizeof (ADLOD6MemClockState)); - - od_clock_mem_status[device_id].state.iNumberOfPerformanceLevels = 2; - - if ((ADL_rc = hm_ADL_Overdrive_StateInfo_Get (data.hm_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"); - - return (-1); - } - - // Query capabilities only to see if profiles were not "damaged", if so output a warning but do accept the users profile settings - - ADLOD6Capabilities caps = {0, 0, 0, {0, 0, 0}, {0, 0, 0}, 0, 0}; - - if ((ADL_rc = hm_ADL_Overdrive_Capabilities_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &caps)) != ADL_OK) - { - log_error ("ERROR: Failed to get ADL device capabilities"); - - return (-1); - } - - int engine_clock_max = caps.sEngineClockRange.iMax * 0.6666; - int memory_clock_max = caps.sMemoryClockRange.iMax * 0.6250; - - int warning_trigger_engine = (int) (0.25 * (float) engine_clock_max); - int warning_trigger_memory = (int) (0.25 * (float) memory_clock_max); - - int engine_clock_profile_max = od_clock_mem_status[device_id].state.aLevels[1].iEngineClock; - int memory_clock_profile_max = od_clock_mem_status[device_id].state.aLevels[1].iMemoryClock; - - // warning if profile has too low max values - - if ((engine_clock_max - engine_clock_profile_max) > warning_trigger_engine) - { - log_info ("WARN: the custom profile seems to have too low maximum engine clock values. You therefore may not reach full performance"); - } - - if ((memory_clock_max - memory_clock_profile_max) > warning_trigger_memory) - { - log_info ("WARN: the custom profile seems to have too low maximum memory clock values. You therefore may not reach full performance"); - } - - ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel)); - - performance_state->iNumberOfPerformanceLevels = 2; - - performance_state->aLevels[0].iEngineClock = engine_clock_profile_max; - performance_state->aLevels[1].iEngineClock = engine_clock_profile_max; - performance_state->aLevels[0].iMemoryClock = memory_clock_profile_max; - performance_state->aLevels[1].iMemoryClock = memory_clock_profile_max; - - if ((ADL_rc = hm_ADL_Overdrive_State_Set (data.hm_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"); - - return (-1); - } - - local_free (performance_state); - } - } - - hc_thread_mutex_unlock (mux_adl); - } - #endif // HAVE_HWMON } - data.kernel_power_all = kernel_power_all; - if (data.quiet == 0) log_info_nn (""); /** @@ -16631,11 +17087,18 @@ int main (int argc, char **argv) if (data.devices_status != STATUS_CRACKED) data.devices_status = STATUS_STARTING; - hc_thread_t i_thread = 0; + uint i_threads_cnt = 0; + + hc_thread_t *i_threads = (hc_thread_t *) mycalloc (10, sizeof (hc_thread_t)); if ((data.wordlist_mode == WL_MODE_FILE) || (data.wordlist_mode == WL_MODE_MASK)) { - hc_thread_create (i_thread, thread_keypress, &benchmark); + if (stdout_flag == 0) + { + hc_thread_create (i_threads[i_threads_cnt], thread_keypress, &benchmark); + + i_threads_cnt++; + } } if (wordlist_mode == WL_MODE_STDIN) data.status = 1; @@ -16644,9 +17107,12 @@ int main (int argc, char **argv) hc_thread_t *ni_threads = (hc_thread_t *) mycalloc (10, sizeof (hc_thread_t)); - hc_thread_create (ni_threads[ni_threads_cnt], thread_monitor, NULL); + if (stdout_flag == 0) + { + hc_thread_create (ni_threads[ni_threads_cnt], thread_monitor, NULL); - ni_threads_cnt++; + ni_threads_cnt++; + } /** * Outfile remove @@ -16985,6 +17451,8 @@ int main (int argc, char **argv) data.ms_paused = 0; + data.kernel_power_final = 0; + data.words_cur = rd->words_cur; for (uint device_id = 0; device_id < data.devices_cnt; device_id++) @@ -17002,8 +17470,6 @@ int main (int argc, char **argv) memset (device_param->exec_ms, 0, EXEC_CACHE * sizeof (double)); - device_param->kernel_power = device_param->kernel_power_user; - device_param->outerloop_pos = 0; device_param->outerloop_left = 0; device_param->innerloop_pos = 0; @@ -17019,8 +17485,6 @@ int main (int argc, char **argv) device_param->words_done = 0; } - data.kernel_power_div = 0; - // figure out some workload if (attack_mode == ATTACK_MODE_STRAIGHT) @@ -17462,26 +17926,6 @@ int main (int argc, char **argv) } } - /* - * Inform user about possible slow speeds - */ - - if ((wordlist_mode == WL_MODE_FILE) || (wordlist_mode == WL_MODE_MASK)) - { - if (data.words_base < kernel_power_all) - { - if (quiet == 0) - { - log_info ("ATTENTION!"); - log_info (" The wordlist or mask you are using is too small."); - log_info (" Therefore, hashcat is unable to utilize the full parallelization power of your device(s)."); - log_info (" The cracking speed will drop."); - log_info (" Workaround: https://hashcat.net/wiki/doku.php?id=frequently_asked_questions#how_to_create_more_work_for_full_speed"); - log_info (""); - } - } - } - /* * Update loopback file */ @@ -17517,6 +17961,64 @@ int main (int argc, char **argv) } } + /** + * create autotune threads + */ + + hc_thread_t *c_threads = (hc_thread_t *) mycalloc (data.devices_cnt, sizeof (hc_thread_t)); + + data.devices_status = STATUS_AUTOTUNE; + + for (uint device_id = 0; device_id < data.devices_cnt; device_id++) + { + hc_device_param_t *device_param = &devices_param[device_id]; + + hc_thread_create (c_threads[device_id], thread_autotune, device_param); + } + + hc_thread_wait (data.devices_cnt, c_threads); + + /* + * Inform user about possible slow speeds + */ + + uint hardware_power_all = 0; + + uint kernel_power_all = 0; + + for (uint device_id = 0; device_id < data.devices_cnt; device_id++) + { + hc_device_param_t *device_param = &devices_param[device_id]; + + hardware_power_all += device_param->hardware_power; + + kernel_power_all += device_param->kernel_power; + } + + data.hardware_power_all = hardware_power_all; // hardware_power_all is the same as kernel_power_all but without the influence of kernel_accel on the devices + + data.kernel_power_all = kernel_power_all; + + if ((wordlist_mode == WL_MODE_FILE) || (wordlist_mode == WL_MODE_MASK)) + { + if (data.words_base < kernel_power_all) + { + if (quiet == 0) + { + log_info ("ATTENTION!"); + log_info (" The wordlist or mask you are using is too small."); + log_info (" Therefore, hashcat is unable to utilize the full parallelization power of your device(s)."); + log_info (" The cracking speed will drop."); + log_info (" Workaround: https://hashcat.net/wiki/doku.php?id=frequently_asked_questions#how_to_create_more_work_for_full_speed"); + log_info (""); + } + } + } + + /** + * create cracker threads + */ + data.devices_status = STATUS_RUNNING; if (initial_restore_done == 0) @@ -17548,12 +18050,6 @@ int main (int argc, char **argv) data.runtime_start = runtime_start; - /** - * create cracker threads - */ - - hc_thread_t *c_threads = (hc_thread_t *) mycalloc (data.devices_cnt, sizeof (hc_thread_t)); - for (uint device_id = 0; device_id < data.devices_cnt; device_id++) { hc_device_param_t *device_param = &devices_param[device_id]; @@ -17568,8 +18064,6 @@ int main (int argc, char **argv) } } - // wait for crack threads to exit - hc_thread_wait (data.devices_cnt, c_threads); local_free (c_threads); @@ -17739,11 +18233,13 @@ int main (int argc, char **argv) // wait for interactive threads - if ((data.wordlist_mode == WL_MODE_FILE) || (data.wordlist_mode == WL_MODE_MASK)) + for (uint thread_idx = 0; thread_idx < i_threads_cnt; thread_idx++) { - hc_thread_wait (1, &i_thread); + hc_thread_wait (1, &i_threads[thread_idx]); } + local_free (i_threads); + // we dont need restore file anymore if (data.restore_disable == 0) { @@ -17890,11 +18386,17 @@ int main (int argc, char **argv) if (device_param->device_vendor_id == VENDOR_ID_AMD) { - rc = hm_set_fanspeed_with_device_id_amd (device_id, fanspeed, 0); + rc = hm_set_fanspeed_with_device_id_adl (device_id, fanspeed, 0); } else if (device_param->device_vendor_id == VENDOR_ID_NV) { + #ifdef _WIN + rc = hm_set_fanspeed_with_device_id_nvapi (device_id, fanspeed, fanpolicy); + #endif + #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); @@ -17918,49 +18420,62 @@ int main (int argc, char **argv) if (device_param->skipped) continue; - if (data.hm_device[device_id].od_version == 6) + if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_AMD) { - // check powertune capabilities first, if not available then skip device - - int powertune_supported = 0; - - if ((hm_ADL_Overdrive6_PowerControl_Caps (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK) + if (data.hm_device[device_id].od_version == 6) { - log_error ("ERROR: Failed to get ADL PowerControl Capabilities"); - - return (-1); - } + // check powertune capabilities first, if not available then skip device - if (powertune_supported != 0) - { - // powercontrol settings + int powertune_supported = 0; - if ((hm_ADL_Overdrive_PowerControl_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, od_power_control_status[device_id])) != ADL_OK) + if ((hm_ADL_Overdrive6_PowerControl_Caps (data.hm_adl, data.hm_device[device_id].adl, &powertune_supported)) != ADL_OK) { - log_info ("ERROR: Failed to restore the ADL PowerControl values"); + log_error ("ERROR: Failed to get ADL PowerControl Capabilities"); return (-1); } - // clocks + if (powertune_supported != 0) + { + // powercontrol settings + + if ((hm_ADL_Overdrive_PowerControl_Set (data.hm_adl, data.hm_device[device_id].adl, od_power_control_status[device_id])) != ADL_OK) + { + log_info ("ERROR: Failed to restore the ADL PowerControl values"); + + return (-1); + } - ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel)); + // clocks - performance_state->iNumberOfPerformanceLevels = 2; + ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel)); - performance_state->aLevels[0].iEngineClock = od_clock_mem_status[device_id].state.aLevels[0].iEngineClock; - performance_state->aLevels[1].iEngineClock = od_clock_mem_status[device_id].state.aLevels[1].iEngineClock; - performance_state->aLevels[0].iMemoryClock = od_clock_mem_status[device_id].state.aLevels[0].iMemoryClock; - performance_state->aLevels[1].iMemoryClock = od_clock_mem_status[device_id].state.aLevels[1].iMemoryClock; + performance_state->iNumberOfPerformanceLevels = 2; - if ((hm_ADL_Overdrive_State_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK) - { - log_info ("ERROR: Failed to restore ADL performance state"); + performance_state->aLevels[0].iEngineClock = od_clock_mem_status[device_id].state.aLevels[0].iEngineClock; + performance_state->aLevels[1].iEngineClock = od_clock_mem_status[device_id].state.aLevels[1].iEngineClock; + performance_state->aLevels[0].iMemoryClock = od_clock_mem_status[device_id].state.aLevels[0].iMemoryClock; + performance_state->aLevels[1].iMemoryClock = od_clock_mem_status[device_id].state.aLevels[1].iMemoryClock; - return (-1); + if ((hm_ADL_Overdrive_State_Set (data.hm_adl, data.hm_device[device_id].adl, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK) + { + log_info ("ERROR: Failed to restore ADL performance state"); + + return (-1); + } + + local_free (performance_state); } + } + } - local_free (performance_state); + if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_NV) + { + unsigned int limit = nvml_power_limit[device_id]; + + if (limit > 0) + { + hm_NVML_nvmlDeviceSetPowerManagementLimit (data.hm_nvml, 0, data.hm_device[device_id].nvml, limit); } } } @@ -17970,22 +18485,40 @@ int main (int argc, char **argv) if (gpu_temp_disable == 0) { - if (data.hm_nv) + if (data.hm_nvml) + { + hm_NVML_nvmlShutdown (data.hm_nvml); + + nvml_close (data.hm_nvml); + + data.hm_nvml = NULL; + } + + if (data.hm_nvapi) + { + hm_NvAPI_Unload (data.hm_nvapi); + + nvapi_close (data.hm_nvapi); + + data.hm_nvapi = NULL; + } + + if (data.hm_xnvctrl) { - hm_NVML_nvmlShutdown (data.hm_nv); + hm_XNVCTRL_XCloseDisplay (data.hm_xnvctrl); - nvml_close (data.hm_nv); + xnvctrl_close (data.hm_xnvctrl); - data.hm_nv = NULL; + data.hm_xnvctrl = NULL; } - if (data.hm_amd) + if (data.hm_adl) { - hm_ADL_Main_Control_Destroy (data.hm_amd); + hm_ADL_Main_Control_Destroy (data.hm_adl); - adl_close (data.hm_amd); + adl_close (data.hm_adl); - data.hm_amd = NULL; + data.hm_adl = NULL; } } #endif // HAVE_HWMON @@ -18031,6 +18564,7 @@ int main (int argc, char **argv) local_free (temp_retain_fanspeed_value); local_free (od_clock_mem_status); local_free (od_power_control_status); + local_free (nvml_power_limit); #endif global_free (devices_param);