X-Git-Url: https://www.flypig.org.uk/git/?a=blobdiff_plain;f=src%2Fhashcat.c;h=744d7d72d3e1ea1ae06a9120255dcb36c290b852;hb=c713196d8be7b45b8ba68feeae4fd2093696455d;hp=fd8b1e8c31e4ade244db273c54446c344d92f761;hpb=3a492277482086dad7dc5099f4afa6561b21758a;p=hashcat.git diff --git a/src/hashcat.c b/src/hashcat.c index fd8b1e8..744d7d7 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 70 +#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 |", @@ -407,13 +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", - #ifdef HAVE_ADL - " --powertune-enable | | Enable automatic power tuning (AMD OverDrive 6 only) |", - #endif + " --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", @@ -572,6 +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", @@ -1533,6 +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); @@ -1549,6 +1566,7 @@ void status_display () const int num_corespeed = hm_get_corespeed_with_device_id (device_id); const int num_memoryspeed = hm_get_memoryspeed_with_device_id (device_id); const int num_buslanes = hm_get_buslanes_with_device_id (device_id); + const int num_throttle = hm_get_throttle_with_device_id (device_id); char output_buf[256] = { 0 }; @@ -1556,35 +1574,35 @@ void status_display () if (num_temperature >= 0) { - snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Temp:%uc", num_temperature); + snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Temp:%3uc", num_temperature); output_len = strlen (output_buf); } if (num_fanspeed >= 0) { - snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Fan:%u%%", num_fanspeed); + snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Fan:%3u%%", num_fanspeed); output_len = strlen (output_buf); } if (num_utilization >= 0) { - snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Util:%u%%", num_utilization); + snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Util:%3u%%", num_utilization); output_len = strlen (output_buf); } if (num_corespeed >= 0) { - snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Core:%uMhz", num_corespeed); + snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Core:%4uMhz", num_corespeed); output_len = strlen (output_buf); } if (num_memoryspeed >= 0) { - snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Mem:%uMhz", num_memoryspeed); + snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Mem:%4uMhz", num_memoryspeed); output_len = strlen (output_buf); } @@ -1596,6 +1614,13 @@ void status_display () output_len = strlen (output_buf); } + if (num_throttle == 1) + { + snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " *Throttled*"); + + output_len = strlen (output_buf); + } + if (output_len == 0) { snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " N/A"); @@ -1603,11 +1628,12 @@ void status_display () output_len = strlen (output_buf); } - log_info ("HWMon.GPU.#%d...:%s", device_id + 1, output_buf); + log_info ("HWMon.Dev.#%d...:%s", device_id + 1, output_buf); } hc_thread_mutex_unlock (mux_adl); } + #endif // HAVE_HWMON } @@ -1727,7 +1753,14 @@ static void status_benchmark () format_speed_display (hashes_dev_ms[device_id] * 1000, display_dev_cur, sizeof (display_dev_cur)); - log_info ("Speed.Dev.#%d.: %9sH/s (%0.2fms)", device_id + 1, display_dev_cur, exec_all_ms[device_id]); + if (data.devices_active >= 10) + { + log_info ("Speed.Dev.#%d: %9sH/s (%0.2fms)", device_id + 1, display_dev_cur, exec_all_ms[device_id]); + } + else + { + log_info ("Speed.Dev.#%d.: %9sH/s (%0.2fms)", device_id + 1, display_dev_cur, exec_all_ms[device_id]); + } } char display_all_cur[16] = { 0 }; @@ -1809,6 +1842,26 @@ static void generate_cached_kernel_amp_filename (const uint attack_kern, char *p snprintf (cached_file, 255, "%s/kernels/amp_a%d.%s.kernel", profile_dir, attack_kern, device_name_chksum); } +static char *filename_from_filepath (char *filepath) +{ + char *ptr = NULL; + + if ((ptr = strrchr (filepath, '/')) != NULL) + { + ptr++; + } + else if ((ptr = strrchr (filepath, '\\')) != NULL) + { + ptr++; + } + else + { + ptr = filepath; + } + + return ptr; +} + static uint convert_from_hex (char *line_buf, const uint line_len) { if (line_len & 1) return (line_len); // not in hex @@ -2223,14 +2276,14 @@ static void check_cracked (hc_device_param_t *device_param, const uint salt_pos) uint cpt_cracked = 0; + hc_thread_mutex_lock (mux_display); + for (uint i = 0; i < num_cracked; i++) { const uint hash_pos = cracked[i].hash_pos; if (data.digests_shown[hash_pos] == 1) continue; - hc_thread_mutex_lock (mux_display); - if ((data.opts_type & OPTS_TYPE_PT_NEVERCRACK) == 0) { data.digests_shown[hash_pos] = 1; @@ -2251,11 +2304,11 @@ static void check_cracked (hc_device_param_t *device_param, const uint salt_pos) if (data.salts_done == data.salts_cnt) data.devices_status = STATUS_CRACKED; - hc_thread_mutex_unlock (mux_display); - check_hash (device_param, &cracked[i]); } + hc_thread_mutex_unlock (mux_display); + myfree (cracked); if (cpt_cracked > 0) @@ -2291,6 +2344,214 @@ static void check_cracked (hc_device_param_t *device_param, const uint salt_pos) } } +// stolen from princeprocessor ;) + +typedef struct +{ + FILE *fp; + + char buf[BUFSIZ]; + int len; + +} out_t; + +static void out_flush (out_t *out) +{ + fwrite (out->buf, 1, out->len, out->fp); + + out->len = 0; +} + +static void out_push (out_t *out, const u8 *pw_buf, const int pw_len) +{ + char *ptr = out->buf + out->len; + + memcpy (ptr, pw_buf, pw_len); + + ptr[pw_len] = '\n'; + + out->len += pw_len + 1; + + if (out->len >= BUFSIZ - 100) + { + out_flush (out); + } +} + +static void process_stdout (hc_device_param_t *device_param, const uint pws_cnt) +{ + out_t out; + + out.fp = stdout; + out.len = 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; + + out_push (&out, plain_ptr, plain_len); + } + } + } + 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; + } + + out_push (&out, plain_ptr, plain_len); + } + } + } + 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; + + out_push (&out, plain_ptr, plain_len); + } + } + } + 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; + + out_push (&out, plain_ptr, plain_len); + } + } + } + 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; + + out_push (&out, plain_ptr, plain_len); + } + } + } + + out_flush (&out); +} + static void save_hash () { char *hashfile = data.hashfile; @@ -2328,8 +2589,6 @@ 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; @@ -2341,11 +2600,15 @@ 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 { @@ -2383,43 +2646,7 @@ 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; @@ -2482,21 +2709,50 @@ 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++; @@ -2709,8 +2965,15 @@ 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) @@ -2729,33 +2992,39 @@ 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; @@ -2764,7 +3033,7 @@ 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; @@ -2796,7 +3065,7 @@ 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); @@ -2805,7 +3074,7 @@ 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); } } @@ -2899,7 +3168,7 @@ 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 @@ -2907,11 +3176,11 @@ 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); @@ -2938,10 +3207,13 @@ 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; @@ -2958,28 +3230,33 @@ static void autotune (hc_device_param_t *device_param) const u32 kernel_power_max = device_param->device_processors * device_param->kernel_threads * kernel_accel_max; - run_kernel_memset (device_param, device_param->d_pws_buf, 7, kernel_power_max * sizeof (pw_t)); - - if (data.attack_exec == ATTACK_EXEC_OUTSIDE_KERNEL) + if (data.attack_kern == ATTACK_KERN_BF) { - run_kernel_memset (device_param, device_param->d_pws_amp_buf, 7, kernel_power_max * sizeof (pw_t)); + run_kernel_memset (device_param, device_param->d_pws_buf, 7, kernel_power_max * sizeof (pw_t)); } - - /* - for (u32 i = 0; i < kernel_power_max; i++) + else { - device_param->pws_buf[i].i[0] = i; - device_param->pws_buf[i].i[1] = 0x01234567; - device_param->pws_buf[i].pw_len = 7; - } + for (u32 i = 0; i < kernel_power_max; i++) + { + device_param->pws_buf[i].i[0] = i; + device_param->pws_buf[i].i[1] = 0x01234567; + 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); + 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) + 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, MIN (kernel_loops_max, KERNEL_RULES) * sizeof (kernel_rule_t), 0, NULL, NULL); + } + } + else { run_kernel_amp (device_param, kernel_power_max); } - */ #define VERIFIER_CNT 1 @@ -3122,6 +3399,10 @@ 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; @@ -3137,8 +3418,8 @@ static void autotune (hc_device_param_t *device_param) { clear_prompt (); - log_info ("Device #%u: autotuned kernel-accel to %u\n" - "Device #%u: autotuned kernel-loops to %u\n", + log_info ("- Device #%u: autotuned kernel-accel to %u\n" + "- Device #%u: autotuned kernel-loops to %u\n", device_param->device_id + 1, kernel_accel, device_param->device_id + 1, kernel_loops); @@ -3239,9 +3520,16 @@ 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; @@ -3399,7 +3687,7 @@ 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 (); @@ -3411,7 +3699,10 @@ 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 @@ -3805,11 +4096,11 @@ 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; - // these variables are mainly used for fan control (AMD only) + // these variables are mainly used for fan control int *fan_speed_chgd = (int *) mycalloc (data.devices_cnt, sizeof (int)); @@ -3818,12 +4109,10 @@ static void *thread_monitor (void *p) int *temp_diff_old = (int *) mycalloc (data.devices_cnt, sizeof (int)); int *temp_diff_sum = (int *) mycalloc (data.devices_cnt, sizeof (int)); - #ifdef HAVE_ADL int temp_threshold = 1; // degrees celcius int fan_speed_min = 15; // in percentage int fan_speed_max = 100; - #endif // HAVE_ADL time_t last_temp_check_time; #endif // HAVE_HWMON @@ -3875,7 +4164,7 @@ static void *thread_monitor (void *p) #ifdef HAVE_HWMON - if (1) + if (hwmon_check == 1) { hc_thread_mutex_lock (mux_adl); @@ -3885,35 +4174,47 @@ 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 (); - if (data.quiet == 0) fprintf (stdout, "%s", PROMPT); - if (data.quiet == 0) fflush (stdout); + log_info ("WARNING: Drivers temperature threshold hit on GPU #%d, expect performance to drop...", device_id + 1); - slowdown_warnings++; + if (slowdown_warnings == 2) + { + log_info (""); + } + + if (data.quiet == 0) fprintf (stdout, "%s", PROMPT); + if (data.quiet == 0) fflush (stdout); + + slowdown_warnings++; + } + } + else + { + slowdown_warnings = 0; + } } } - else - { - slowdown_warnings = 0; - } } hc_thread_mutex_unlock (mux_adl); @@ -3950,12 +4251,11 @@ static void *thread_monitor (void *p) break; } - #ifdef HAVE_ADL const int gpu_temp_retain = data.gpu_temp_retain; - if (gpu_temp_retain) // VENDOR_ID_AMD implied + if (gpu_temp_retain) { - if (data.hm_device[device_id].fan_supported == 1) + if (data.hm_device[device_id].fan_set_supported == 1) { int temp_cur = temperature; @@ -3995,7 +4295,20 @@ static void *thread_monitor (void *p) if ((freely_change_fan_speed == 1) || (fan_speed_must_change == 1)) { - hm_set_fanspeed_with_device_id_amd (device_id, fan_speed_new); + if (device_param->device_vendor_id == VENDOR_ID_AMD) + { + 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; } @@ -4005,7 +4318,6 @@ static void *thread_monitor (void *p) } } } - #endif // HAVE_ADL } hc_thread_mutex_unlock (mux_adl); @@ -4395,8 +4707,8 @@ 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; @@ -4408,16 +4720,55 @@ 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); @@ -4426,33 +4777,19 @@ 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); @@ -4465,7 +4802,7 @@ 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; @@ -4473,12 +4810,19 @@ static void *thread_calc_stdin (void *p) autotune (device_param); + return NULL; +} + +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 attack_kern = data.attack_kern; - const uint kernel_power = device_param->kernel_power; - while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT)) { hc_thread_mutex_lock (mux_dispatcher); @@ -4492,7 +4836,7 @@ 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); @@ -4526,6 +4870,8 @@ 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)) @@ -4542,25 +4888,6 @@ 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); @@ -4619,8 +4946,6 @@ 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; @@ -4628,7 +4953,7 @@ 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; @@ -4743,18 +5068,16 @@ 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; @@ -4763,8 +5086,6 @@ 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); @@ -4945,11 +5266,11 @@ 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; @@ -4964,10 +5285,10 @@ 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); } /** @@ -5350,8 +5671,8 @@ static uint hlfmt_detect (FILE *fp, uint max_check) // wrapper around mymalloc for ADL -#if defined(HAVE_HWMON) && defined(HAVE_ADL) -void *__stdcall ADL_Main_Memory_Alloc (const int iSize) +#if defined(HAVE_HWMON) +void *HC_API_CALL ADL_Main_Memory_Alloc (const int iSize) { return mymalloc (iSize); } @@ -5373,8 +5694,6 @@ 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; @@ -5409,7 +5728,7 @@ 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); @@ -5426,8 +5745,8 @@ 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; @@ -5437,7 +5756,7 @@ void SetConsoleWindowSize (const int x) int main (int argc, char **argv) { - #ifdef _WIN + #ifdef WIN SetConsoleWindowSize (132); #endif @@ -5506,6 +5825,7 @@ 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; @@ -5565,14 +5885,13 @@ 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; uint gpu_temp_retain = GPU_TEMP_RETAIN; - #ifdef HAVE_ADL uint powertune_enable = POWERTUNE_ENABLE; #endif - #endif uint logfile_disable = LOGFILE_DISABLE; uint segment_size = SEGMENT_SIZE; uint scrypt_tmto = SCRYPT_TMTO; @@ -5605,6 +5924,7 @@ 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' @@ -5644,6 +5964,7 @@ 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 @@ -5685,6 +6006,7 @@ 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}, @@ -5722,13 +6044,12 @@ 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}, {"gpu-temp-retain", required_argument, 0, IDX_GPU_TEMP_RETAIN}, - #ifdef HAVE_ADL {"powertune-enable", no_argument, 0, IDX_POWERTUNE_ENABLE}, - #endif #endif // HAVE_HWMON {"logfile-disable", no_argument, 0, IDX_LOGFILE_DISABLE}, {"truecrypt-keyfiles", required_argument, 0, IDX_TRUECRYPT_KEYFILES}, @@ -5943,6 +6264,7 @@ 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; @@ -5952,11 +6274,6 @@ int main (int argc, char **argv) uint workload_profile_chgd = 0; uint opencl_vector_width_chgd = 0; - #if defined(HAVE_HWMON) && defined(HAVE_ADL) - uint gpu_temp_retain_chgd = 0; - uint gpu_temp_abort_chgd = 0; - #endif - optind = 1; optopt = 0; option_index = 0; @@ -5988,6 +6305,7 @@ 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; @@ -6035,21 +6353,13 @@ 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); - #ifdef HAVE_ADL - gpu_temp_abort_chgd = 1; - #endif - break; - case IDX_GPU_TEMP_RETAIN: gpu_temp_retain = atoi (optarg); - #ifdef HAVE_ADL - gpu_temp_retain_chgd = 1; - #endif - break; - #ifdef HAVE_ADL + case IDX_GPU_TEMP_ABORT: gpu_temp_abort = atoi (optarg); break; + case IDX_GPU_TEMP_RETAIN: gpu_temp_retain = atoi (optarg); break; case IDX_POWERTUNE_ENABLE: powertune_enable = 1; break; - #endif #endif // HAVE_HWMON case IDX_LOGFILE_DISABLE: logfile_disable = 1; break; case IDX_TRUECRYPT_KEYFILES: truecrypt_keyfiles = optarg; break; @@ -6108,6 +6418,10 @@ 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); @@ -6133,7 +6447,7 @@ 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"); @@ -6183,7 +6497,7 @@ int main (int argc, char **argv) { if (outfile_format > 1) { - log_error ("ERROR: Mixing outfile-format > 1 is not allowed together with left parameter"); + log_error ("ERROR: Mixing outfile-format > 1 with left parameter is not allowed"); return (-1); } @@ -6200,7 +6514,7 @@ int main (int argc, char **argv) { if ((outfile_format > 7) && (outfile_format < 16)) { - log_error ("ERROR: Mixing outfile-format > 7 is not allowed together with show parameter"); + log_error ("ERROR: Mixing outfile-format > 7 with show parameter is not allowed"); return (-1); } @@ -6230,21 +6544,21 @@ int main (int argc, char **argv) if ((increment == 1) && (attack_mode == ATTACK_MODE_STRAIGHT)) { - log_error ("ERROR: increment is not allowed in attack-mode 0"); + log_error ("ERROR: Increment is not allowed in attack-mode 0"); return (-1); } if ((increment == 0) && (increment_min_chgd == 1)) { - log_error ("ERROR: increment-min is only supported together with increment switch"); + log_error ("ERROR: Increment-min is only supported combined with increment switch"); return (-1); } if ((increment == 0) && (increment_max_chgd == 1)) { - log_error ("ERROR: increment-max is only supported together with increment switch"); + log_error ("ERROR: Increment-max is only supported combined with increment switch"); return (-1); } @@ -6277,8 +6591,8 @@ int main (int argc, char **argv) { if (force == 0) { - log_info ("The manual use of the option -n (or --kernel-accel) is outdated"); - log_info ("Please consider using the option -w instead"); + log_info ("The manual use of the -n option (or --kernel-accel) is outdated"); + log_info ("Please consider using the -w option instead"); log_info ("You can use --force to override this but do not post error reports if you do so"); log_info (""); @@ -6304,8 +6618,8 @@ int main (int argc, char **argv) { if (force == 0) { - log_info ("The manual use of the option -u (or --kernel-loops) is outdated"); - log_info ("Please consider using the option -w instead"); + log_info ("The manual use of the -u option (or --kernel-loops) is outdated"); + log_info ("Please consider using the -w option instead"); log_info ("You can use --force to override this but do not post error reports if you do so"); log_info (""); @@ -6371,8 +6685,32 @@ 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; @@ -6430,25 +6768,6 @@ 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) { @@ -6459,13 +6778,13 @@ int main (int argc, char **argv) { if (show == 1) { - log_error ("ERROR: Mixing show parameter not supported with keyspace parameter"); + log_error ("ERROR: Combining show parameter with keyspace parameter is not allowed"); return (-1); } else if (left == 1) { - log_error ("ERROR: Mixing left parameter not supported wiht keyspace parameter"); + log_error ("ERROR: Combining left parameter with keyspace parameter is not allowed"); return (-1); } @@ -6481,6 +6800,26 @@ 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; + session = "stdout"; + opencl_vector_width = 1; + } + if (remove_timer_chgd == 1) { if (remove == 0) @@ -6573,6 +6912,14 @@ 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 */ @@ -6754,10 +7101,8 @@ int main (int argc, char **argv) data.skip = skip; data.limit = limit; #ifdef HAVE_HWMON - #ifdef HAVE_ADL data.powertune_enable = powertune_enable; #endif - #endif data.logfile_disable = logfile_disable; data.truecrypt_keyfiles = truecrypt_keyfiles; data.veracrypt_keyfiles = veracrypt_keyfiles; @@ -6829,12 +7174,14 @@ 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); @@ -6859,7 +7206,7 @@ int main (int argc, char **argv) logfile_top_uint (outfile_format); logfile_top_uint (potfile_disable); logfile_top_string (potfile_path); - #if defined(HAVE_HWMON) && defined(HAVE_ADL) + #if defined(HAVE_HWMON) logfile_top_uint (powertune_enable); #endif logfile_top_uint (scrypt_tmto); @@ -6954,17 +7301,21 @@ 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 - #ifdef HAVE_ADL - powertune_enable = 1; - #endif + 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 @@ -8396,6 +8747,21 @@ 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 = KERN_TYPE_STDOUT; + 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; @@ -10810,6 +11176,25 @@ 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); } @@ -10936,6 +11321,7 @@ 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; @@ -11269,7 +11655,7 @@ 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; @@ -11338,7 +11724,7 @@ int main (int argc, char **argv) uint hashes_avail = 0; - if (benchmark == 0) + if ((benchmark == 0) && (stdout_flag == 0)) { struct stat f; @@ -11520,6 +11906,10 @@ 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) { } @@ -11543,7 +11933,7 @@ int main (int argc, char **argv) if (hash_fmt_error) { - log_info ("WARNING: failed to parse hashes using the '%s' format", strhlfmt (hashlist_format)); + log_info ("WARNING: Failed to parse hashes using the '%s' format", strhlfmt (hashlist_format)); } else { @@ -11840,7 +12230,7 @@ int main (int argc, char **argv) if (parser_status < PARSER_GLOBAL_ZERO) { - log_info ("WARNING: Hashfile '%s' in line %u (%s): %s", data.hashfile, line_num, line_buf, strparser (parser_status)); + log_info ("WARNING: Hashfile '%s' on line %u (%s): %s", data.hashfile, line_num, line_buf, strparser (parser_status)); continue; } @@ -11853,7 +12243,7 @@ int main (int argc, char **argv) if (parser_status < PARSER_GLOBAL_ZERO) { - log_info ("WARNING: Hashfile '%s' in line %u (%s): %s", data.hashfile, line_num, line_buf, strparser (parser_status)); + log_info ("WARNING: Hashfile '%s' on line %u (%s): %s", data.hashfile, line_num, line_buf, strparser (parser_status)); continue; } @@ -11875,7 +12265,7 @@ int main (int argc, char **argv) if (parser_status < PARSER_GLOBAL_ZERO) { - log_info ("WARNING: Hashfile '%s' in line %u (%s): %s", data.hashfile, line_num, line_buf, strparser (parser_status)); + log_info ("WARNING: Hashfile '%s' on line %u (%s): %s", data.hashfile, line_num, line_buf, strparser (parser_status)); continue; } @@ -11894,7 +12284,7 @@ int main (int argc, char **argv) if (parser_status < PARSER_GLOBAL_ZERO) { - log_info ("WARNING: Hashfile '%s' in line %u (%s): %s", data.hashfile, line_num, line_buf, strparser (parser_status)); + log_info ("WARNING: Hashfile '%s' on line %u (%s): %s", data.hashfile, line_num, line_buf, strparser (parser_status)); continue; } @@ -12329,7 +12719,7 @@ int main (int argc, char **argv) return (0); } - if (keyspace == 0) + if ((keyspace == 0) && (stdout_flag == 0)) { if (hashes_cnt == 0) { @@ -13118,14 +13508,14 @@ int main (int argc, char **argv) if (result == -1) { - log_info ("WARNING: Skipping invalid or unsupported rule in file %s in line %u: %s", rp_file, rule_line, rule_buf); + log_info ("WARNING: Skipping invalid or unsupported rule in file %s on line %u: %s", rp_file, rule_line, rule_buf); continue; } if (cpu_rule_to_kernel_rule (rule_buf, rule_len, &kernel_rules_buf[kernel_rules_cnt]) == -1) { - log_info ("WARNING: Cannot convert rule for use on device in file %s in line %u: %s", rp_file, rule_line, rule_buf); + log_info ("WARNING: Cannot convert rule for use on device in file %s on line %u: %s", rp_file, rule_line, rule_buf); memset (&kernel_rules_buf[kernel_rules_cnt], 0, sizeof (kernel_rule_t)); // needs to be cleared otherwise we could have some remaining data @@ -13292,40 +13682,13 @@ int main (int argc, char **argv) } } - /** - * OpenCL platforms: For each platform check if we need to unset features that we can not use, eg: temp_retain - */ - - for (uint platform_id = 0; platform_id < platforms_cnt; platform_id++) - { - cl_platform_id platform = platforms[platform_id]; - - char platform_vendor[INFOSZ] = { 0 }; - - hc_clGetPlatformInfo (data.ocl, platform, CL_PLATFORM_VENDOR, sizeof (platform_vendor), platform_vendor, NULL); - - #ifdef HAVE_HWMON - #if defined(HAVE_NVML) || defined(HAVE_NVAPI) - if (strcmp (platform_vendor, CL_VENDOR_NV) == 0) - { - // make sure that we do not directly control the fan for NVidia - - gpu_temp_retain = 0; - - data.gpu_temp_retain = gpu_temp_retain; - } - #endif // HAVE_NVML || HAVE_NVAPI - #endif - } - - /** - * OpenCL device types: - * In case the user did not specify --opencl-device-types and the user runs hashcat in a system with only a CPU only he probably want to use that CPU. - * In such a case, automatically enable CPU device type support, since it's disabled by default. - */ - if (opencl_device_types == NULL) { + /** + * OpenCL device types: + * In case the user did not specify --opencl-device-types and the user runs hashcat in a system with only a CPU only he probably want to use that CPU. + */ + cl_device_type device_types_all = 0; for (uint platform_id = 0; platform_id < platforms_cnt; platform_id++) @@ -13348,16 +13711,34 @@ int main (int argc, char **argv) } } + // In such a case, automatically enable CPU device type support, since it's disabled by default. + if ((device_types_all & (CL_DEVICE_TYPE_GPU | CL_DEVICE_TYPE_ACCELERATOR)) == 0) { device_types_filter |= CL_DEVICE_TYPE_CPU; } + + // In another case, when the user uses --stdout, using CPU devices is much faster to setup + // If we have a CPU device, force it to be used + + if (stdout_flag == 1) + { + if (device_types_all & CL_DEVICE_TYPE_CPU) + { + device_types_filter = CL_DEVICE_TYPE_CPU; + } + } } /** * 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; @@ -13368,8 +13749,6 @@ int main (int argc, char **argv) for (uint platform_id = 0; platform_id < platforms_cnt; platform_id++) { - if ((opencl_platforms_filter & (1 << platform_id)) == 0) continue; - cl_platform_id platform = platforms[platform_id]; hc_clGetDeviceIDs (data.ocl, platform, CL_DEVICE_TYPE_ALL, DEVICES_MAX, platform_devices, &platform_devices_cnt); @@ -13421,6 +13800,32 @@ int main (int argc, char **argv) platform_vendor_id = VENDOR_ID_GENERIC; } + const uint platform_skipped = ((opencl_platforms_filter & (1 << platform_id)) == 0); + + if ((benchmark == 1 || quiet == 0) && (algorithm_pos == 0)) + { + if (machine_readable == 0) + { + if (platform_skipped == 0) + { + const int len = log_info ("OpenCL Platform #%u: %s", platform_id + 1, platform_vendor); + + char line[256] = { 0 }; + + for (int i = 0; i < len; i++) line[i] = '='; + + log_info (line); + } + else + { + log_info ("OpenCL Platform #%u: %s, skipped", platform_id + 1, platform_vendor); + log_info (""); + } + } + } + + if (platform_skipped == 1) continue; + for (uint platform_devices_id = 0; platform_devices_id < platform_devices_cnt; platform_devices_id++) { size_t param_value_size = 0; @@ -13614,7 +14019,7 @@ 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 a little endian device", device_id + 1); device_param->skipped = 1; } @@ -13627,7 +14032,7 @@ 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; } @@ -13640,7 +14045,7 @@ 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: No compiler available for device", device_id + 1); device_param->skipped = 1; } @@ -13653,7 +14058,7 @@ 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; } @@ -13670,14 +14075,14 @@ 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; } @@ -13692,7 +14097,7 @@ 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; } @@ -13707,9 +14112,16 @@ 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 a 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; + } } } @@ -13748,18 +14160,41 @@ int main (int argc, char **argv) // device_processor_cores - if (device_type & CL_DEVICE_TYPE_CPU) + if (device_param->device_type & CL_DEVICE_TYPE_GPU) { - cl_uint device_processor_cores = 1; - - device_param->device_processor_cores = device_processor_cores; - } + if ((device_param->platform_vendor_id == VENDOR_ID_AMD) && (device_param->device_vendor_id == VENDOR_ID_AMD)) + { + need_adl = 1; + } - if (device_type & CL_DEVICE_TYPE_GPU) - { - if (device_vendor_id == VENDOR_ID_AMD) + if ((device_param->platform_vendor_id == VENDOR_ID_NV) && (device_param->device_vendor_id == VENDOR_ID_NV)) { - cl_uint device_processor_cores = 0; + 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; + + device_param->device_processor_cores = device_processor_cores; + } + + if (device_type & CL_DEVICE_TYPE_GPU) + { + if (device_vendor_id == VENDOR_ID_AMD) + { + cl_uint device_processor_cores = 0; #define CL_DEVICE_WAVEFRONT_WIDTH_AMD 0x4043 @@ -13796,6 +14231,29 @@ 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 { @@ -13813,7 +14271,7 @@ int main (int argc, char **argv) { if (device_param->skipped == 0) { - log_info ("Device #%u: %s, %lu/%lu MB allocatable, %uMCU", + log_info ("- Device #%u: %s, %lu/%lu MB allocatable, %uMCU", device_id + 1, device_name, (unsigned int) (device_maxmem_alloc / 1024 / 1024), @@ -13822,7 +14280,7 @@ int main (int argc, char **argv) } else { - log_info ("Device #%u: %s, skipped", + log_info ("- Device #%u: %s, skipped", device_id + 1, device_name); } @@ -13859,8 +14317,8 @@ int main (int argc, char **argv) if (catalyst_broken == 1) { log_info (""); - log_info ("ATTENTION! The installed catalyst driver in your system is known to be broken!"); - log_info ("It will pass over cracked hashes and does not report them as cracked"); + log_info ("ATTENTION! The Catalyst driver installed on your system is known to be broken!"); + log_info ("It passes over cracked hashes and will not report them as cracked"); 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 (""); @@ -13871,8 +14329,8 @@ int main (int argc, char **argv) if (catalyst_warn == 1) { log_info (""); - log_info ("ATTENTION! Unsupported or incorrect installed catalyst driver detected!"); - log_info ("You are STRONGLY encouraged to use the official supported catalyst driver for good reasons"); + log_info ("ATTENTION! Unsupported or incorrectly installed Catalyst driver detected!"); + log_info ("You are STRONGLY encouraged to use the official supported catalyst driver"); log_info ("See hashcat's homepage for official supported catalyst drivers"); #ifdef _WIN log_info ("Also see: http://hashcat.net/wiki/doku.php?id=upgrading_amd_drivers_how_to"); @@ -13887,8 +14345,8 @@ int main (int argc, char **argv) { if (device_param->kernel_exec_timeout != 0) { - if (data.quiet == 0) log_info ("Device #%u: WARNING! Kernel exec timeout is not disabled, it might cause you errors of code 702", device_id + 1); - if (data.quiet == 0) log_info (" See the wiki on how to disable it: https://hashcat.net/wiki/doku.php?id=timeout_patch"); + if (data.quiet == 0) log_info ("- Device #%u: WARNING! Kernel exec timeout is not disabled, it might cause you errors of code 702", device_id + 1); + if (data.quiet == 0) log_info (" See the wiki on how to disable it: https://hashcat.net/wiki/doku.php?id=timeout_patch"); } } } @@ -13977,6 +14435,14 @@ int main (int argc, char **argv) devices_cnt++; } + + if ((benchmark == 1 || quiet == 0) && (algorithm_pos == 0)) + { + if (machine_readable == 0) + { + log_info (""); + } + } } if (keyspace == 0 && devices_active == 0) @@ -14004,114 +14470,125 @@ int main (int argc, char **argv) data.devices_active = devices_active; - if ((benchmark == 1 || quiet == 0) && (algorithm_pos == 0)) - { - if (machine_readable == 0) - { - log_info (""); - } - } - /** * HM devices: init */ #ifdef HAVE_HWMON - #if defined(HAVE_NVML) || defined(HAVE_NVAPI) - hm_attrs_t hm_adapters_nv[DEVICES_MAX] = { { { 0 }, 0, 0, 0, 0 } }; - #endif - - #ifdef HAVE_ADL - hm_attrs_t hm_adapters_amd[DEVICES_MAX] = { { { 0 }, 0, 0, 0, 0 } }; - #endif + 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) { - #if defined(WIN) && defined(HAVE_NVAPI) - NVAPI_PTR *nvapi = (NVAPI_PTR *) mymalloc (sizeof (NVAPI_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)); - if (nvapi_init (nvapi) == 0) - data.hm_nv = nvapi; + data.hm_adl = NULL; + data.hm_nvapi = NULL; + data.hm_nvml = NULL; + data.hm_xnvctrl = NULL; - if (data.hm_nv) + if ((need_nvml == 1) && (nvml_init (nvml) == 0)) { - if (hm_NvAPI_Initialize (data.hm_nv) == NVAPI_OK) + data.hm_nvml = nvml; + } + + if (data.hm_nvml) + { + 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++) { - NV_GPU_COOLER_SETTINGS pCoolerSettings; + unsigned int speed; + + if (hm_NVML_nvmlDeviceGetFanSpeed (data.hm_nvml, 0, hm_adapters_nvml[i].nvml, &speed) == NVML_SUCCESS) hm_adapters_nvml[i].fan_get_supported = 1; - pCoolerSettings.Version = GPU_COOLER_SETTINGS_VER | sizeof (NV_GPU_COOLER_SETTINGS); + hm_NVML_nvmlDeviceSetComputeMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_COMPUTEMODE_EXCLUSIVE_PROCESS); - if (hm_NvAPI_GPU_GetCoolerSettings (data.hm_nv, hm_adapters_nv[i].adapter_index.nv, 0, &pCoolerSettings) != NVAPI_NOT_SUPPORTED) hm_adapters_nv[i].fan_supported = 1; + hm_NVML_nvmlDeviceSetGpuOperationMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_GOM_ALL_ON); } } } - #endif // WIN && HAVE_NVAPI - #if defined(LINUX) && defined(HAVE_NVML) - NVML_PTR *nvml = (NVML_PTR *) mymalloc (sizeof (NVML_PTR)); - - if (nvml_init (nvml) == 0) - data.hm_nv = nvml; + if ((need_nvapi == 1) && (nvapi_init (nvapi) == 0)) + { + data.hm_nvapi = nvapi; + } - if (data.hm_nv) + if (data.hm_nvapi) { - if (hm_NVML_nvmlInit (data.hm_nv) == NVML_SUCCESS) + if (hm_NvAPI_Initialize (data.hm_nvapi) == NVAPI_OK) { - HM_ADAPTER_NV nvGPUHandle[DEVICES_MAX] = { 0 }; + HM_ADAPTER_NVAPI nvGPUHandle[DEVICES_MAX] = { 0 }; - int tmp_in = hm_get_adapter_index_nv (nvGPUHandle); + int tmp_in = hm_get_adapter_index_nvapi (nvGPUHandle); int tmp_out = 0; for (int i = 0; i < tmp_in; i++) { - hm_adapters_nv[tmp_out++].adapter_index.nv = nvGPUHandle[i]; + hm_adapters_nvapi[tmp_out++].nvapi = nvGPUHandle[i]; } + } + } - for (int i = 0; i < tmp_out; i++) + 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++) { - unsigned int speed; + hc_device_param_t *device_param = &data.devices_param[device_id]; - 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_supported = 1; + if ((device_param->device_type & CL_DEVICE_TYPE_GPU) == 0) continue; + + 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; } } } - #endif // LINUX && HAVE_NVML - - data.hm_amd = NULL; - #ifdef HAVE_ADL - ADL_PTR *adl = (ADL_PTR *) mymalloc (sizeof (ADL_PTR)); - - if (adl_init (adl) == 0) - data.hm_amd = adl; + 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); @@ -14125,12 +14602,12 @@ 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); } @@ -14139,9 +14616,8 @@ int main (int argc, char **argv) myfree (lpAdapterInfo); } } - #endif // HAVE_ADL - 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; } @@ -14151,34 +14627,19 @@ int main (int argc, char **argv) * OpenCL devices: allocate buffer for device specific information */ - #ifdef HAVE_HWMON - int *temp_retain_fanspeed_value = (int *) mycalloc (data.devices_cnt, sizeof (int)); + int *temp_retain_fanspeed_value = (int *) mycalloc (data.devices_cnt, sizeof (int)); + int *temp_retain_fanpolicy_value = (int *) mycalloc (data.devices_cnt, sizeof (int)); - #ifdef HAVE_ADL ADLOD6MemClockState *od_clock_mem_status = (ADLOD6MemClockState *) mycalloc (data.devices_cnt, sizeof (ADLOD6MemClockState)); int *od_power_control_status = (int *) mycalloc (data.devices_cnt, sizeof (int)); - #endif // ADL - #endif - /** - * enable custom signal handler(s) - */ - - if (benchmark == 0) - { - hc_signal (sigHandler_default); - } - else - { - hc_signal (sigHandler_benchmark); - } + unsigned int *nvml_power_limit = (unsigned int *) mycalloc (data.devices_cnt, sizeof (unsigned int)); /** * User-defined GPU temp handling */ - #ifdef HAVE_HWMON if (gpu_temp_disable == 1) { gpu_temp_abort = 0; @@ -14189,7 +14650,7 @@ int main (int argc, char **argv) { if (gpu_temp_abort < gpu_temp_retain) { - log_error ("ERROR: invalid values for gpu-temp-abort. Parameter gpu-temp-abort is less than gpu-temp-retain."); + log_error ("ERROR: Invalid values for gpu-temp-abort. Parameter gpu-temp-abort is less than gpu-temp-retain."); return (-1); } @@ -14200,6 +14661,19 @@ int main (int argc, char **argv) data.gpu_temp_retain = gpu_temp_retain; #endif + /** + * enable custom signal handler(s) + */ + + if (benchmark == 0) + { + hc_signal (sigHandler_default); + } + else + { + hc_signal (sigHandler_benchmark); + } + /** * inform the user */ @@ -14232,7 +14706,7 @@ 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"); } @@ -14259,6 +14733,8 @@ int main (int argc, char **argv) #endif } + #ifdef HAVE_HWMON + /** * HM devices: copy */ @@ -14275,31 +14751,34 @@ int main (int argc, char **argv) const uint platform_devices_id = device_param->platform_devices_id; - #if defined(HAVE_NVML) || defined(HAVE_NVAPI) - 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; } - #endif - #ifdef HAVE_ADL - 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; } - #endif } } /** - * 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 */ - #ifdef HAVE_ADL if (powertune_enable == 1) { hc_thread_mutex_lock (mux_adl); @@ -14310,38 +14789,177 @@ 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}; + + 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]); + } + + 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_adl, data.hm_device[device_id].adl, 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_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"); + + 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_adl, data.hm_device[device_id].adl, &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_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; + } + } } } } @@ -14349,7 +14967,7 @@ int main (int argc, char **argv) hc_thread_mutex_unlock (mux_adl); } - #endif // HAVE_ADK + #endif // HAVE_HWMON #ifdef DEBUG @@ -14358,8 +14976,6 @@ int main (int argc, char **argv) if (data.quiet == 0) log_info_nn ("Initializing device kernels and memory..."); - uint kernel_power_all = 0; - for (uint device_id = 0; device_id < data.devices_cnt; device_id++) { /** @@ -14420,6 +15036,10 @@ 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 */ @@ -14506,7 +15126,7 @@ int main (int argc, char **argv) if (size_scryptV > device_param->device_maxmem_alloc) { - if (quiet == 0) log_info ("WARNING: not enough device memory allocatable to use --scrypt-tmto %d, increasing...", tmto); + if (quiet == 0) log_info ("WARNING: Not enough device memory allocatable to use --scrypt-tmto %d, increasing...", tmto); continue; } @@ -14522,7 +15142,7 @@ int main (int argc, char **argv) if (data.salts_buf[0].scrypt_phy == 0) { - log_error ("ERROR: can't allocate enough device memory"); + log_error ("ERROR: Can't allocate enough device memory"); return -1; } @@ -14534,7 +15154,7 @@ 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; @@ -14542,7 +15162,7 @@ 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; @@ -14709,7 +15329,9 @@ 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 @@ -14763,7 +15385,7 @@ int main (int argc, char **argv) /* if (kernel_accel_max == 0) { - log_error ("Device #%u: Device does not provide enough allocatable device-memory to handle hash-type %u", device_id + 1, data.hash_mode); + log_error ("- Device #%u: Device does not provide enough allocatable device-memory to handle hash-type %u", device_id + 1, data.hash_mode); return -1; } @@ -14775,7 +15397,7 @@ int main (int argc, char **argv) /* if (kernel_accel_max < kernel_accel) { - if (quiet == 0) log_info ("Device #%u: Reduced maximum kernel-accel to %u", device_id + 1, kernel_accel_max); + if (quiet == 0) log_info ("- Device #%u: Reduced maximum kernel-accel to %u", device_id + 1, kernel_accel_max); device_param->kernel_accel = kernel_accel_max; } @@ -14789,15 +15411,6 @@ 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 */ @@ -14830,7 +15443,7 @@ int main (int argc, char **argv) */ #ifdef DEBUG - log_info ("Device #%u: build_opts '%s'\n", device_id + 1, build_opts); + log_info ("- Device #%u: build_opts '%s'\n", device_id + 1, build_opts); #endif /** @@ -14884,7 +15497,7 @@ int main (int argc, char **argv) { if (cached == 0) { - if (quiet == 0) log_info ("Device #%u: Kernel %s not found in cache! Building may take a while...", device_id + 1, cached_file); + if (quiet == 0) log_info ("- Device #%u: Kernel %s not found in cache! Building may take a while...", device_id + 1, filename_from_filepath (cached_file)); load_kernel (source_file, 1, kernel_lengths, kernel_sources); @@ -14915,7 +15528,7 @@ int main (int argc, char **argv) { device_param->skipped = true; - log_info ("Device #%u: Kernel %s build failure. Proceed without this device.", device_id + 1, source_file); + log_info ("- Device #%u: Kernel %s build failure. Proceeding without this device.", device_id + 1, source_file); continue; } @@ -14935,7 +15548,7 @@ int main (int argc, char **argv) else { #ifdef DEBUG - log_info ("Device #%u: Kernel %s (%ld bytes)", device_id + 1, cached_file, cst.st_size); + log_info ("- Device #%u: Kernel %s (%ld bytes)", device_id + 1, cached_file, cst.st_size); #endif load_kernel (cached_file, 1, kernel_lengths, kernel_sources); @@ -14948,7 +15561,7 @@ int main (int argc, char **argv) else { #ifdef DEBUG - log_info ("Device #%u: Kernel %s (%ld bytes)", device_id + 1, source_file, sst.st_size); + log_info ("- Device #%u: Kernel %s (%ld bytes)", device_id + 1, source_file, sst.st_size); #endif load_kernel (source_file, 1, kernel_lengths, kernel_sources); @@ -14995,7 +15608,7 @@ int main (int argc, char **argv) { device_param->skipped = true; - log_info ("Device #%u: Kernel %s build failure. Proceed without this device.", device_id + 1, source_file); + log_info ("- Device #%u: Kernel %s build failure. Proceeding without this device.", device_id + 1, source_file); } } @@ -15054,7 +15667,7 @@ int main (int argc, char **argv) if (cached == 0) { - if (quiet == 0) log_info ("Device #%u: Kernel %s not found in cache! Building may take a while...", device_id + 1, cached_file); + if (quiet == 0) log_info ("- Device #%u: Kernel %s not found in cache! Building may take a while...", device_id + 1, filename_from_filepath (cached_file)); if (quiet == 0) log_info (""); load_kernel (source_file, 1, kernel_lengths, kernel_sources); @@ -15067,7 +15680,7 @@ int main (int argc, char **argv) { device_param->skipped = true; - log_info ("Device #%u: Kernel %s build failure. Proceed without this device.", device_id + 1, source_file); + log_info ("- Device #%u: Kernel %s build failure. Proceeding without this device.", device_id + 1, source_file); continue; } @@ -15087,7 +15700,7 @@ int main (int argc, char **argv) else { #ifdef DEBUG - log_info ("Device #%u: Kernel %s (%ld bytes)", device_id + 1, cached_file, cst.st_size); + log_info ("- Device #%u: Kernel %s (%ld bytes)", device_id + 1, cached_file, cst.st_size); #endif load_kernel (cached_file, 1, kernel_lengths, kernel_sources); @@ -15156,7 +15769,7 @@ int main (int argc, char **argv) if (cached == 0) { - if (quiet == 0) log_info ("Device #%u: Kernel %s not found in cache! Building may take a while...", device_id + 1, cached_file); + if (quiet == 0) log_info ("- Device #%u: Kernel %s not found in cache! Building may take a while...", device_id + 1, filename_from_filepath (cached_file)); if (quiet == 0) log_info (""); load_kernel (source_file, 1, kernel_lengths, kernel_sources); @@ -15169,7 +15782,7 @@ int main (int argc, char **argv) { device_param->skipped = true; - log_info ("Device #%u: Kernel %s build failure. Proceed without this device.", device_id + 1, source_file); + log_info ("- Device #%u: Kernel %s build failure. Proceed without this device.", device_id + 1, source_file); continue; } @@ -15189,7 +15802,7 @@ int main (int argc, char **argv) else { #ifdef DEBUG - if (quiet == 0) log_info ("Device #%u: Kernel %s (%ld bytes)", device_id + 1, cached_file, cst.st_size); + if (quiet == 0) log_info ("- Device #%u: Kernel %s (%ld bytes)", device_id + 1, cached_file, cst.st_size); #endif load_kernel (cached_file, 1, kernel_lengths, kernel_sources); @@ -15637,72 +16250,64 @@ int main (int argc, char **argv) run_kernel_bzero (device_param, device_param->d_markov_css_buf, size_markov_css); } - /** - * Store thermal target temperature so we can send a notice to user - */ - #if defined(HAVE_HWMON) - if (gpu_temp_disable == 0) - { - const int gpu_temp_threshold_slowdown = hm_get_threshold_slowdown_with_device_id (device_id); - - data.hm_device[device_id].gpu_temp_threshold_slowdown = (gpu_temp_threshold_slowdown == -1) ? 100000 : gpu_temp_threshold_slowdown; - } - #endif /** * Store initial fanspeed if gpu_temp_retain is enabled */ - #if defined(HAVE_HWMON) && defined(HAVE_ADL) - int gpu_temp_retain_set = 0; - if (gpu_temp_disable == 0) { - if (gpu_temp_retain != 0) // VENDOR_ID_AMD implied + if (gpu_temp_retain != 0) { hc_thread_mutex_lock (mux_adl); - if (data.hm_device[device_id].fan_supported == 1) + if (data.hm_device[device_id].fan_get_supported == 1) { - if (gpu_temp_retain_chgd == 0) - { - uint cur_temp = 0; - uint default_temp = 0; + const int fanspeed = hm_get_fanspeed_with_device_id (device_id); + const int fanpolicy = hm_get_fanpolicy_with_device_id (device_id); - int ADL_rc = hm_ADL_Overdrive6_TargetTemperatureData_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, (int *) &cur_temp, (int *) &default_temp); + temp_retain_fanspeed_value[device_id] = fanspeed; + temp_retain_fanpolicy_value[device_id] = fanpolicy; - if (ADL_rc == ADL_OK) - { - #define GPU_TEMP_RETAIN_ABORT_DIFF 15 + // we also set it to tell the OS we take control over the fan and it's automatic controller + // if it was set to automatic. we do not control user-defined fanspeeds. - const uint gpu_temp_retain_target = default_temp - GPU_TEMP_RETAIN_ABORT_DIFF; + if (fanpolicy == 1) + { + data.hm_device[device_id].fan_set_supported = 1; - // special case with multi gpu setups: always use minimum retain + int rc = -1; - if (gpu_temp_retain_set == 0) - { - gpu_temp_retain = gpu_temp_retain_target; - gpu_temp_retain_set = 1; - } - else - { - gpu_temp_retain = MIN (gpu_temp_retain, gpu_temp_retain_target); - } - - if (gpu_temp_abort_chgd == 0) gpu_temp_abort = gpu_temp_retain + GPU_TEMP_RETAIN_ABORT_DIFF; + if (device_param->device_vendor_id == VENDOR_ID_AMD) + { + rc = hm_set_fanspeed_with_device_id_adl (device_id, fanspeed, 1); } - } + else if (device_param->device_vendor_id == VENDOR_ID_NV) + { + #ifdef LINUX + rc = set_fan_control (data.hm_xnvctrl, data.hm_device[device_id].xnvctrl, NV_CTRL_GPU_COOLER_MANUAL_CONTROL_TRUE); + #endif - const int fan_speed = hm_get_fanspeed_with_device_id (device_id); + #ifdef WIN + rc = hm_set_fanspeed_with_device_id_nvapi (device_id, fanspeed, 1); + #endif + } - temp_retain_fanspeed_value[device_id] = fan_speed; + if (rc == 0) + { + data.hm_device[device_id].fan_set_supported = 1; + } + else + { + log_info ("WARNING: Failed to set initial fan speed for device #%u", device_id + 1); - if (fan_speed == -1) + data.hm_device[device_id].fan_set_supported = 0; + } + } + else { - log_info ("WARNING: Failed to get current fan speed settings for gpu number: %i:", device_id + 1); - - temp_retain_fanspeed_value[device_id] = 0; + data.hm_device[device_id].fan_set_supported = 0; } } @@ -15710,126 +16315,9 @@ 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 && HAVE_ADL + #endif // HAVE_HWMON } - data.kernel_power_all = kernel_power_all; - if (data.quiet == 0) log_info_nn (""); /** @@ -15950,7 +16438,7 @@ int main (int argc, char **argv) if (keyspace == 1) { - log_error ("ERROR: keyspace parameter is not allowed together with a directory"); + log_error ("ERROR: Keyspace parameter is not allowed together with a directory"); return (-1); } @@ -16413,7 +16901,7 @@ int main (int argc, char **argv) if (keyspace == 1) { - log_error ("ERROR: keyspace parameter is not allowed together with a directory"); + log_error ("ERROR: Keyspace parameter is not allowed together with a directory"); return (-1); } @@ -16592,7 +17080,7 @@ int main (int argc, char **argv) if (keyspace == 1) { - log_error ("ERROR: keyspace parameter is not allowed together with a directory"); + log_error ("ERROR: Keyspace parameter is not allowed together with a directory"); return (-1); } @@ -16700,11 +17188,18 @@ 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; @@ -16713,9 +17208,12 @@ 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 @@ -16756,8 +17254,8 @@ int main (int argc, char **argv) { if (potfile_remove_cracks > 0) { - if (potfile_remove_cracks == 1) log_info ("INFO: removed 1 hash found in pot file\n"); - else log_info ("INFO: removed %u hashes found in pot file\n", potfile_remove_cracks); + if (potfile_remove_cracks == 1) log_info ("INFO: Removed 1 hash found in pot file\n"); + else log_info ("INFO: Removed %u hashes found in pot file\n", potfile_remove_cracks); } } @@ -17054,6 +17552,8 @@ 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++) @@ -17071,8 +17571,6 @@ 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; @@ -17088,8 +17586,6 @@ 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) @@ -17281,12 +17777,12 @@ int main (int argc, char **argv) { if (css_cnt < mask_min) { - log_info ("WARNING: skipping mask '%s' because it is smaller than the minimum password length", mask); + log_info ("WARNING: Skipping mask '%s' because it is smaller than the minimum password length", mask); } if (css_cnt > mask_max) { - log_info ("WARNING: skipping mask '%s' because it is larger than the maximum password length", mask); + log_info ("WARNING: Skipping mask '%s' because it is larger than the maximum password length", mask); } // skip to next mask @@ -17501,7 +17997,7 @@ int main (int argc, char **argv) if (data.words_cur > data.words_base) { - log_error ("ERROR: restore value greater keyspace"); + log_error ("ERROR: Restore value greater keyspace"); return (-1); } @@ -17531,26 +18027,6 @@ 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 */ @@ -17586,6 +18062,64 @@ 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) @@ -17617,12 +18151,6 @@ 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]; @@ -17637,8 +18165,6 @@ int main (int argc, char **argv) } } - // wait for crack threads to exit - hc_thread_wait (data.devices_cnt, c_threads); local_free (c_threads); @@ -17808,11 +18334,13 @@ 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) { @@ -17938,7 +18466,6 @@ int main (int argc, char **argv) #ifdef HAVE_HWMON if (gpu_temp_disable == 0) { - #ifdef HAVE_ADL if (gpu_temp_retain != 0) // VENDOR_ID_AMD is implied here { hc_thread_mutex_lock (mux_adl); @@ -17949,24 +18476,39 @@ int main (int argc, char **argv) if (device_param->skipped) continue; - if (data.hm_device[device_id].fan_supported == 1) + if (data.hm_device[device_id].fan_set_supported == 1) { - int fanspeed = temp_retain_fanspeed_value[device_id]; + int fanspeed = temp_retain_fanspeed_value[device_id]; + int fanpolicy = temp_retain_fanpolicy_value[device_id]; - if (fanspeed == -1) continue; + if (fanpolicy == 1) + { + int rc = -1; + + if (device_param->device_vendor_id == VENDOR_ID_AMD) + { + rc = hm_set_fanspeed_with_device_id_adl (device_id, fanspeed, 0); + } + else if (device_param->device_vendor_id == VENDOR_ID_NV) + { + #ifdef LINUX + rc = set_fan_control (data.hm_xnvctrl, data.hm_device[device_id].xnvctrl, NV_CTRL_GPU_COOLER_MANUAL_CONTROL_FALSE); + #endif - int rc = hm_set_fanspeed_with_device_id_amd (device_id, fanspeed); + #ifdef WIN + rc = hm_set_fanspeed_with_device_id_nvapi (device_id, fanspeed, fanpolicy); + #endif + } - if (rc == -1) log_info ("WARNING: Failed to restore default fan speed for gpu number: %i:", device_id); + if (rc == -1) log_info ("WARNING: Failed to restore default fan speed and policy for device #%", device_id + 1); + } } } hc_thread_mutex_unlock (mux_adl); } - #endif // HAVE_ADL } - #ifdef HAVE_ADL // reset power tuning if (powertune_enable == 1) // VENDOR_ID_AMD is implied here @@ -17979,89 +18521,106 @@ 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"); + // check powertune capabilities first, if not available then skip device - return (-1); - } - - 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 - ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel)); + 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"); - performance_state->iNumberOfPerformanceLevels = 2; + return (-1); + } - 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; + // clocks - 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"); + ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel)); - return (-1); + performance_state->iNumberOfPerformanceLevels = 2; + + 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; + + 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); } + } + } + + if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_NV) + { + unsigned int limit = nvml_power_limit[device_id]; - local_free (performance_state); + if (limit > 0) + { + hm_NVML_nvmlDeviceSetPowerManagementLimit (data.hm_nvml, 0, data.hm_device[device_id].nvml, limit); } } } hc_thread_mutex_unlock (mux_adl); } - #endif // HAVE_ADL if (gpu_temp_disable == 0) { - #if defined(HAVE_NVML) || defined(HAVE_NVAPI) - if (data.hm_nv) + if (data.hm_nvml) { - #if defined(LINUX) && defined(HAVE_NVML) + hm_NVML_nvmlShutdown (data.hm_nvml); - hm_NVML_nvmlShutdown (data.hm_nv); + nvml_close (data.hm_nvml); - nvml_close (data.hm_nv); + data.hm_nvml = NULL; + } - #elif defined(WIN) && (HAVE_NVAPI) + if (data.hm_nvapi) + { + hm_NvAPI_Unload (data.hm_nvapi); - hm_NvAPI_Unload (data.hm_nv); + nvapi_close (data.hm_nvapi); - nvapi_close (data.hm_nv); + data.hm_nvapi = NULL; + } - #endif + if (data.hm_xnvctrl) + { + hm_XNVCTRL_XCloseDisplay (data.hm_xnvctrl); + + xnvctrl_close (data.hm_xnvctrl); - data.hm_nv = NULL; + data.hm_xnvctrl = NULL; } - #endif - #ifdef HAVE_ADL - 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_adl); - adl_close (data.hm_amd); - data.hm_amd = NULL; + data.hm_adl = NULL; } - #endif } #endif // HAVE_HWMON @@ -18104,10 +18663,9 @@ int main (int argc, char **argv) #ifdef HAVE_HWMON local_free (temp_retain_fanspeed_value); - #ifdef HAVE_ADL local_free (od_clock_mem_status); local_free (od_power_control_status); - #endif // ADL + local_free (nvml_power_limit); #endif global_free (devices_param);