uint plain_len = 0;
- const uint il_cnt = device_param->kernel_params_buf32[27]; // ugly, i know
+ const uint il_cnt = device_param->kernel_params_buf32[30]; // ugly, i know
if (data.attack_mode == ATTACK_MODE_STRAIGHT)
{
{
uint num_elements = num;
- device_param->kernel_params_buf32[30] = data.combs_mode;
- device_param->kernel_params_buf32[31] = num;
+ device_param->kernel_params_buf32[33] = data.combs_mode;
+ device_param->kernel_params_buf32[34] = num;
uint kernel_threads = device_param->kernel_threads;
case KERN_RUN_3: kernel = device_param->kernel3; break;
}
- hc_clSetKernelArg (data.ocl, kernel, 21, sizeof (cl_uint), device_param->kernel_params[21]);
- hc_clSetKernelArg (data.ocl, kernel, 22, sizeof (cl_uint), device_param->kernel_params[22]);
- hc_clSetKernelArg (data.ocl, kernel, 23, sizeof (cl_uint), device_param->kernel_params[23]);
hc_clSetKernelArg (data.ocl, kernel, 24, sizeof (cl_uint), device_param->kernel_params[24]);
hc_clSetKernelArg (data.ocl, kernel, 25, sizeof (cl_uint), device_param->kernel_params[25]);
hc_clSetKernelArg (data.ocl, kernel, 26, sizeof (cl_uint), device_param->kernel_params[26]);
hc_clSetKernelArg (data.ocl, kernel, 29, sizeof (cl_uint), device_param->kernel_params[29]);
hc_clSetKernelArg (data.ocl, kernel, 30, sizeof (cl_uint), device_param->kernel_params[30]);
hc_clSetKernelArg (data.ocl, kernel, 31, sizeof (cl_uint), device_param->kernel_params[31]);
+ hc_clSetKernelArg (data.ocl, kernel, 32, sizeof (cl_uint), device_param->kernel_params[32]);
+ hc_clSetKernelArg (data.ocl, kernel, 33, sizeof (cl_uint), device_param->kernel_params[33]);
+ hc_clSetKernelArg (data.ocl, kernel, 34, sizeof (cl_uint), device_param->kernel_params[34]);
cl_event event;
loop_left = MIN (loop_left, loop_step);
- device_param->kernel_params_buf32[25] = loop_pos;
- device_param->kernel_params_buf32[26] = loop_left;
+ device_param->kernel_params_buf32[28] = loop_pos;
+ device_param->kernel_params_buf32[29] = loop_left;
run_kernel (KERN_RUN_2, device_param, pws_cnt, true, slow_iteration);
{
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
- device_param->kernel_params_buf32[27] = kernel_loops; // because there's two variables for inner iters for slow and fast hashes
+ device_param->kernel_params_buf32[28] = 0;
+ device_param->kernel_params_buf32[29] = kernel_loops; // not a bug, both need to be set
+ device_param->kernel_params_buf32[30] = kernel_loops; // because there's two variables for inner iters for slow and fast hashes
if (data.attack_exec == ATTACK_EXEC_INSIDE_KERNEL)
{
salt_t *salt_buf = &data.salts_buf[salt_pos];
- device_param->kernel_params_buf32[24] = salt_pos;
- device_param->kernel_params_buf32[28] = salt_buf->digests_cnt;
- device_param->kernel_params_buf32[29] = salt_buf->digests_offset;
+ device_param->kernel_params_buf32[27] = salt_pos;
+ device_param->kernel_params_buf32[31] = salt_buf->digests_cnt;
+ device_param->kernel_params_buf32[32] = salt_buf->digests_offset;
FILE *combs_fp = device_param->combs_fp;
device_param->innerloop_pos = innerloop_pos;
device_param->innerloop_left = innerloop_left;
- device_param->kernel_params_buf32[27] = innerloop_left;
+ device_param->kernel_params_buf32[30] = innerloop_left;
// i think we can get rid of this
if (innerloop_left == 0)
salt_t *salt_buf = &data.salts_buf[salt_pos];
- device_param->kernel_params_buf32[24] = salt_pos;
- device_param->kernel_params_buf32[27] = 1;
- device_param->kernel_params_buf32[28] = salt_buf->digests_cnt;
- device_param->kernel_params_buf32[29] = salt_buf->digests_offset;
- device_param->kernel_params_buf32[30] = 0;
- device_param->kernel_params_buf32[31] = 1;
+ device_param->kernel_params_buf32[27] = salt_pos;
+ device_param->kernel_params_buf32[30] = 1;
+ device_param->kernel_params_buf32[31] = salt_buf->digests_cnt;
+ device_param->kernel_params_buf32[32] = salt_buf->digests_offset;
+ device_param->kernel_params_buf32[33] = 0;
+ device_param->kernel_params_buf32[34] = 1;
char *dictfile_old = data.dictfile;
loop_left = MIN (loop_left, loop_step);
- device_param->kernel_params_buf32[25] = loop_pos;
- device_param->kernel_params_buf32[26] = loop_left;
+ device_param->kernel_params_buf32[28] = loop_pos;
+ device_param->kernel_params_buf32[29] = loop_left;
run_kernel (KERN_RUN_2, device_param, 1, false, 0);
}
* cleanup
*/
- device_param->kernel_params_buf32[24] = 0;
- device_param->kernel_params_buf32[25] = 0;
- device_param->kernel_params_buf32[26] = 0;
device_param->kernel_params_buf32[27] = 0;
device_param->kernel_params_buf32[28] = 0;
device_param->kernel_params_buf32[29] = 0;
device_param->kernel_params_buf32[30] = 0;
device_param->kernel_params_buf32[31] = 0;
+ device_param->kernel_params_buf32[32] = 0;
+ device_param->kernel_params_buf32[33] = 0;
+ device_param->kernel_params_buf32[34] = 0;
data.dictfile = dictfile_old;
if (hash_mode == 3000) kernel_threads = 64; // DES
if (hash_mode == 3200) kernel_threads = 8; // Blowfish
if (hash_mode == 7500) kernel_threads = 64; // RC4
- if (hash_mode == 8900) kernel_threads = 32; // scrypt
+ if (hash_mode == 8900) kernel_threads = 64; // Scrypt
if (hash_mode == 9000) kernel_threads = 8; // Blowfish
- if (hash_mode == 9300) kernel_threads = 32; // scrypt
+ if (hash_mode == 9300) kernel_threads = 64; // Scrypt
if (hash_mode == 9700) kernel_threads = 64; // RC4
if (hash_mode == 9710) kernel_threads = 64; // RC4
if (hash_mode == 9800) kernel_threads = 64; // RC4
// scryptV stuff
- size_t size_scryptV = 1;
+ size_t size_scrypt = 4;
if ((hash_mode == 8900) || (hash_mode == 9300))
{
+ // we need to check that all hashes have the same scrypt settings
+
+ const u32 scrypt_N = data.salts_buf[0].scrypt_N;
+ const u32 scrypt_r = data.salts_buf[0].scrypt_r;
+ const u32 scrypt_p = data.salts_buf[0].scrypt_p;
+
+ for (uint i = 1; i < salts_cnt; i++)
+ {
+ if ((data.salts_buf[i].scrypt_N != scrypt_N)
+ || (data.salts_buf[i].scrypt_r != scrypt_r)
+ || (data.salts_buf[i].scrypt_p != scrypt_p))
+ {
+ log_error ("ERROR: Mixed scrypt settings not supported");
+
+ return -1;
+ }
+ }
+
uint tmto_start = 0;
uint tmto_stop = 10;
else
{
// in case the user did not specify the tmto manually
- // use some values known to run best (tested on 290x for AMD and 980ti for NV)
- // but set the lower end only in case the user has a device with too less memory
+ // use some values known to run best (tested on 290x for AMD and GTX1080 for NV)
if (hash_mode == 8900)
{
if (device_param->device_vendor_id == VENDOR_ID_AMD)
{
- tmto_start = 1;
+ tmto_start = 3;
}
else if (device_param->device_vendor_id == VENDOR_ID_NV)
{
}
else if (device_param->device_vendor_id == VENDOR_ID_NV)
{
- tmto_start = 2;
+ tmto_start = 4;
}
}
}
+ data.scrypt_tmp_size = (128 * scrypt_r);
+
device_param->kernel_accel_min = 1;
device_param->kernel_accel_max = 8;
- for (uint tmto = tmto_start; tmto < tmto_stop; tmto++)
+ uint tmto;
+
+ for (tmto = tmto_start; tmto < tmto_stop; tmto++)
{
- // TODO: in theory the following calculation needs to be done per salt, not global
- // we assume all hashes have the same scrypt settings
+ size_scrypt = (128 * scrypt_r) * scrypt_N;
+
+ size_scrypt /= 1 << tmto;
- size_scryptV = (128 * data.salts_buf[0].scrypt_r) * data.salts_buf[0].scrypt_N;
+ size_scrypt *= device_param->device_processors * device_param->kernel_threads * device_param->kernel_accel_max;
- size_scryptV /= 1 << tmto;
+ if ((size_scrypt / 4) > device_param->device_maxmem_alloc)
+ {
+ if (quiet == 0) log_info ("WARNING: Not enough single-block device memory allocatable to use --scrypt-tmto %d, increasing...", tmto);
- size_scryptV *= device_param->device_processors * device_param->kernel_threads * device_param->kernel_accel_max;
+ continue;
+ }
- if (size_scryptV > device_param->device_maxmem_alloc)
+ if (size_scrypt > device_param->device_global_mem)
{
- 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 total device memory allocatable to use --scrypt-tmto %d, increasing...", tmto);
continue;
}
for (uint salts_pos = 0; salts_pos < data.salts_cnt; salts_pos++)
{
- data.salts_buf[salts_pos].scrypt_tmto = tmto;
- data.salts_buf[salts_pos].scrypt_phy = device_param->device_processors * device_param->kernel_threads * device_param->kernel_accel_max;
+ data.scrypt_tmto_final = tmto;
}
break;
}
- if (data.salts_buf[0].scrypt_phy == 0)
+ if (tmto == tmto_stop)
{
log_error ("ERROR: Can't allocate enough device memory");
return -1;
}
- if (quiet == 0) log_info ("SCRYPT tmto optimizer value set to: %u, mem: %u\n", data.salts_buf[0].scrypt_tmto, size_scryptV);
+ if (quiet == 0) log_info ("SCRYPT tmto optimizer value set to: %u, mem: %u\n", data.scrypt_tmto_final, size_scrypt);
}
+ size_t size_scrypt4 = size_scrypt / 4;
+
/**
* some algorithms need a fixed kernel-loops count
*/
case 7900: size_tmps = kernel_power_max * sizeof (drupal7_tmp_t); break;
case 8200: size_tmps = kernel_power_max * sizeof (pbkdf2_sha512_tmp_t); break;
case 8800: size_tmps = kernel_power_max * sizeof (androidfde_tmp_t); break;
- case 8900: size_tmps = kernel_power_max * sizeof (scrypt_tmp_t); break;
+ case 8900: size_tmps = kernel_power_max * data.scrypt_tmp_size; break;
case 9000: size_tmps = kernel_power_max * sizeof (pwsafe2_tmp_t); break;
case 9100: size_tmps = kernel_power_max * sizeof (lotus8_tmp_t); break;
case 9200: size_tmps = kernel_power_max * sizeof (pbkdf2_sha256_tmp_t); break;
- case 9300: size_tmps = kernel_power_max * sizeof (scrypt_tmp_t); break;
+ case 9300: size_tmps = kernel_power_max * data.scrypt_tmp_size; break;
case 9400: size_tmps = kernel_power_max * sizeof (office2007_tmp_t); break;
case 9500: size_tmps = kernel_power_max * sizeof (office2010_tmp_t); break;
case 9600: size_tmps = kernel_power_max * sizeof (office2013_tmp_t); break;
+ size_rules
+ size_rules_c
+ size_salts
- + size_scryptV
+ + size_scrypt4
+ + size_scrypt4
+ + size_scrypt4
+ + size_scrypt4
+ size_shown
+ size_tm
+ size_tmps;
snprintf (build_opts, sizeof (build_opts) - 1, "-I \"%s\"", cpath_real);
- myfree (cpath_real);
-
#else
snprintf (cpath, sizeof (cpath) - 1, "%s/OpenCL/", shared_dir);
snprintf (build_opts, sizeof (build_opts) - 1, "-I %s", cpath_real);
- myfree (cpath_real);
-
#endif
+ // include check
+ // this test needs to be done manually because of osx opencl runtime
+ // if there's a problem with permission, its not reporting back and erroring out silently
+
+ #define files_cnt 15
+
+ const char *files_names[files_cnt] =
+ {
+ "inc_cipher_aes256.cl",
+ "inc_cipher_serpent256.cl",
+ "inc_cipher_twofish256.cl",
+ "inc_common.cl",
+ "inc_comp_multi_bs.cl",
+ "inc_comp_multi.cl",
+ "inc_comp_single_bs.cl",
+ "inc_comp_single.cl",
+ "inc_hash_constants.h",
+ "inc_hash_functions.cl",
+ "inc_rp.cl",
+ "inc_rp.h",
+ "inc_simd.cl",
+ "inc_types.cl",
+ "inc_vendor.cl",
+ };
+
+ for (int i = 0; i < files_cnt; i++)
+ {
+ char path[1024] = { 0 };
+
+ snprintf (path, sizeof (path) - 1, "%s/%s", cpath_real, files_names[i]);
+
+ FILE *fd = fopen (path, "r");
+
+ if (fd == NULL)
+ {
+ log_error ("ERROR: %s: fopen(): %s", path, strerror (errno));
+
+ return -1;
+ }
+
+ char buf[1];
+
+ size_t n = fread (buf, 1, 1, fd);
+
+ if (n != 1)
+ {
+ log_error ("ERROR: %s: fread(): %s", path, strerror (errno));
+
+ return -1;
+ }
+
+ fclose (fd);
+ }
+
+ myfree (cpath_real);
+
// we don't have sm_* on vendors not NV but it doesn't matter
char build_opts_new[1024] = { 0 };
if (force_jit_compilation == 1500)
{
- snprintf (build_opts_update, sizeof (build_opts_update) - 1, "%s -DDESCRYPT_SALT=%d", build_opts, data.salts_buf[0].salt_buf[0]);
+ snprintf (build_opts_update, sizeof (build_opts_update) - 1, "%s -DDESCRYPT_SALT=%u", build_opts, data.salts_buf[0].salt_buf[0]);
}
else if (force_jit_compilation == 8900)
{
- snprintf (build_opts_update, sizeof (build_opts_update) - 1, "%s -DSCRYPT_N=%d -DSCRYPT_R=%d -DSCRYPT_P=%d -DSCRYPT_TMTO=%d", build_opts, data.salts_buf[0].scrypt_N, data.salts_buf[0].scrypt_r, data.salts_buf[0].scrypt_p, 1 << data.salts_buf[0].scrypt_tmto);
+ snprintf (build_opts_update, sizeof (build_opts_update) - 1, "%s -DSCRYPT_N=%u -DSCRYPT_R=%u -DSCRYPT_P=%u -DSCRYPT_TMTO=%u -DSCRYPT_TMP_ELEM=%u", build_opts, data.salts_buf[0].scrypt_N, data.salts_buf[0].scrypt_r, data.salts_buf[0].scrypt_p, 1 << data.scrypt_tmto_final, data.scrypt_tmp_size / 16);
}
else
{
device_param->d_digests_shown = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_WRITE, size_shown, NULL);
device_param->d_salt_bufs = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_ONLY, size_salts, NULL);
device_param->d_result = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_WRITE, size_results, NULL);
- device_param->d_scryptV_buf = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_WRITE, size_scryptV, NULL);
+ device_param->d_scryptV0_buf = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_WRITE, size_scrypt4, NULL);
+ device_param->d_scryptV1_buf = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_WRITE, size_scrypt4, NULL);
+ device_param->d_scryptV2_buf = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_WRITE, size_scrypt4, NULL);
+ device_param->d_scryptV3_buf = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_WRITE, size_scrypt4, NULL);
hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_bitmap_s1_a, CL_TRUE, 0, bitmap_size, bitmap_s1_a, 0, NULL, NULL);
hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_bitmap_s1_b, CL_TRUE, 0, bitmap_size, bitmap_s1_b, 0, NULL, NULL);
* kernel args
*/
- device_param->kernel_params_buf32[21] = bitmap_mask;
- device_param->kernel_params_buf32[22] = bitmap_shift1;
- device_param->kernel_params_buf32[23] = bitmap_shift2;
- device_param->kernel_params_buf32[24] = 0; // salt_pos
- device_param->kernel_params_buf32[25] = 0; // loop_pos
- device_param->kernel_params_buf32[26] = 0; // loop_cnt
- device_param->kernel_params_buf32[27] = 0; // kernel_rules_cnt
- device_param->kernel_params_buf32[28] = 0; // digests_cnt
- device_param->kernel_params_buf32[29] = 0; // digests_offset
- device_param->kernel_params_buf32[30] = 0; // combs_mode
- device_param->kernel_params_buf32[31] = 0; // gid_max
+ device_param->kernel_params_buf32[24] = bitmap_mask;
+ device_param->kernel_params_buf32[25] = bitmap_shift1;
+ device_param->kernel_params_buf32[26] = bitmap_shift2;
+ device_param->kernel_params_buf32[27] = 0; // salt_pos
+ device_param->kernel_params_buf32[28] = 0; // loop_pos
+ device_param->kernel_params_buf32[29] = 0; // loop_cnt
+ device_param->kernel_params_buf32[30] = 0; // kernel_rules_cnt
+ device_param->kernel_params_buf32[31] = 0; // digests_cnt
+ device_param->kernel_params_buf32[32] = 0; // digests_offset
+ device_param->kernel_params_buf32[33] = 0; // combs_mode
+ device_param->kernel_params_buf32[34] = 0; // gid_max
device_param->kernel_params[ 0] = (attack_exec == ATTACK_EXEC_INSIDE_KERNEL)
? &device_param->d_pws_buf
device_param->kernel_params[17] = &device_param->d_salt_bufs;
device_param->kernel_params[18] = &device_param->d_esalt_bufs;
device_param->kernel_params[19] = &device_param->d_result;
- device_param->kernel_params[20] = &device_param->d_scryptV_buf;
- device_param->kernel_params[21] = &device_param->kernel_params_buf32[21];
- device_param->kernel_params[22] = &device_param->kernel_params_buf32[22];
- device_param->kernel_params[23] = &device_param->kernel_params_buf32[23];
+ device_param->kernel_params[20] = &device_param->d_scryptV0_buf;
+ device_param->kernel_params[21] = &device_param->d_scryptV1_buf;
+ device_param->kernel_params[22] = &device_param->d_scryptV2_buf;
+ device_param->kernel_params[23] = &device_param->d_scryptV3_buf;
device_param->kernel_params[24] = &device_param->kernel_params_buf32[24];
device_param->kernel_params[25] = &device_param->kernel_params_buf32[25];
device_param->kernel_params[26] = &device_param->kernel_params_buf32[26];
device_param->kernel_params[29] = &device_param->kernel_params_buf32[29];
device_param->kernel_params[30] = &device_param->kernel_params_buf32[30];
device_param->kernel_params[31] = &device_param->kernel_params_buf32[31];
+ device_param->kernel_params[32] = &device_param->kernel_params_buf32[32];
+ device_param->kernel_params[33] = &device_param->kernel_params_buf32[33];
+ device_param->kernel_params[34] = &device_param->kernel_params_buf32[34];
device_param->kernel_params_mp_buf64[3] = 0;
device_param->kernel_params_mp_buf32[4] = 0;
hc_clGetKernelWorkGroupInfo (data.ocl, device_param->kernel2, device_param->device, CL_KERNEL_WORK_GROUP_SIZE, sizeof (size_t), &kernel_wgs_tmp, NULL); kernel_threads = MIN (kernel_threads, kernel_wgs_tmp);
hc_clGetKernelWorkGroupInfo (data.ocl, device_param->kernel3, device_param->device, CL_KERNEL_WORK_GROUP_SIZE, sizeof (size_t), &kernel_wgs_tmp, NULL); kernel_threads = MIN (kernel_threads, kernel_wgs_tmp);
- for (uint i = 0; i <= 20; i++)
+ for (uint i = 0; i <= 23; i++)
{
hc_clSetKernelArg (data.ocl, device_param->kernel1, i, sizeof (cl_mem), device_param->kernel_params[i]);
hc_clSetKernelArg (data.ocl, device_param->kernel2, i, sizeof (cl_mem), device_param->kernel_params[i]);
if (opts_type & OPTS_TYPE_HOOK23) hc_clSetKernelArg (data.ocl, device_param->kernel23, i, sizeof (cl_mem), device_param->kernel_params[i]);
}
- for (uint i = 21; i <= 31; i++)
+ for (uint i = 24; i <= 34; i++)
{
hc_clSetKernelArg (data.ocl, device_param->kernel1, i, sizeof (cl_uint), device_param->kernel_params[i]);
hc_clSetKernelArg (data.ocl, device_param->kernel2, i, sizeof (cl_uint), device_param->kernel_params[i]);
if (device_param->d_tmps) hc_clReleaseMemObject (data.ocl, device_param->d_tmps);
if (device_param->d_hooks) hc_clReleaseMemObject (data.ocl, device_param->d_hooks);
if (device_param->d_result) hc_clReleaseMemObject (data.ocl, device_param->d_result);
- if (device_param->d_scryptV_buf) hc_clReleaseMemObject (data.ocl, device_param->d_scryptV_buf);
+ if (device_param->d_scryptV0_buf) hc_clReleaseMemObject (data.ocl, device_param->d_scryptV0_buf);
+ if (device_param->d_scryptV1_buf) hc_clReleaseMemObject (data.ocl, device_param->d_scryptV1_buf);
+ if (device_param->d_scryptV2_buf) hc_clReleaseMemObject (data.ocl, device_param->d_scryptV2_buf);
+ if (device_param->d_scryptV3_buf) hc_clReleaseMemObject (data.ocl, device_param->d_scryptV3_buf);
if (device_param->d_root_css_buf) hc_clReleaseMemObject (data.ocl, device_param->d_root_css_buf);
if (device_param->d_markov_css_buf) hc_clReleaseMemObject (data.ocl, device_param->d_markov_css_buf);
if (device_param->d_tm_c) hc_clReleaseMemObject (data.ocl, device_param->d_tm_c);
projects / hashcat.git / blobdiff