src/intel/tools/error2aub.c

   1 /*
   2  * Copyright © 2018 Intel Corporation
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21  * IN THE SOFTWARE.
  22  *
  23  */
  24
  25 #include <assert.h>
  26 #include <getopt.h>
  27 #include <inttypes.h>
  28 #include <signal.h>
  29 #include <stdio.h>
  30 #include <stdlib.h>
  31 #include <string.h>
  32 #include <stdarg.h>
  33 #include <zlib.h>
  34
  35 #include "util/list.h"
  36
  37 #include "aub_write.h"
  38 #include "drm-uapi/i915_drm.h"
  39 #include "intel_aub.h"
  40
  41 static void __attribute__ ((format(__printf__, 2, 3)))
  42 fail_if(int cond, const char *format, ...)
  43 {
  44    va_list args;
  45
  46    if (!cond)
  47       return;
  48
  49    va_start(args, format);
  50    vfprintf(stderr, format, args);
  51    va_end(args);
  52
  53    raise(SIGTRAP);
  54 }
  55
  56 #define fail(...) fail_if(true, __VA_ARGS__)
  57
  58 static int zlib_inflate(uint32_t **ptr, int len)
  59 {
  60    struct z_stream_s zstream;
  61    void *out;
  62    const uint32_t out_size = 128*4096;  /* approximate obj size */
  63
  64    memset(&zstream, 0, sizeof(zstream));
  65
  66    zstream.next_in = (unsigned char *)*ptr;
  67    zstream.avail_in = 4*len;
  68
  69    if (inflateInit(&zstream) != Z_OK)
  70       return 0;
  71
  72    out = malloc(out_size);
  73    zstream.next_out = out;
  74    zstream.avail_out = out_size;
  75
  76    do {
  77       switch (inflate(&zstream, Z_SYNC_FLUSH)) {
  78       case Z_STREAM_END:
  79          goto end;
  80       case Z_OK:
  81          break;
  82       default:
  83          inflateEnd(&zstream);
  84          return 0;
  85       }
  86
  87       if (zstream.avail_out)
  88          break;
  89
  90       out = realloc(out, 2*zstream.total_out);
  91       if (out == NULL) {
  92          inflateEnd(&zstream);
  93          return 0;
  94       }
  95
  96       zstream.next_out = (unsigned char *)out + zstream.total_out;
  97       zstream.avail_out = zstream.total_out;
  98    } while (1);
  99  end:
 100    inflateEnd(&zstream);
 101    free(*ptr);
 102    *ptr = out;
 103    return zstream.total_out / 4;
 104 }
 105
 106 static int ascii85_decode(const char *in, uint32_t **out, bool inflate)
 107 {
 108    int len = 0, size = 1024;
 109
 110    *out = realloc(*out, sizeof(uint32_t)*size);
 111    if (*out == NULL)
 112       return 0;
 113
 114    while (*in >= '!' && *in <= 'z') {
 115       uint32_t v = 0;
 116
 117       if (len == size) {
 118          size *= 2;
 119          *out = realloc(*out, sizeof(uint32_t)*size);
 120          if (*out == NULL)
 121             return 0;
 122       }
 123
 124       if (*in == 'z') {
 125          in++;
 126       } else {
 127          v += in[0] - 33; v *= 85;
 128          v += in[1] - 33; v *= 85;
 129          v += in[2] - 33; v *= 85;
 130          v += in[3] - 33; v *= 85;
 131          v += in[4] - 33;
 132          in += 5;
 133       }
 134       (*out)[len++] = v;
 135    }
 136
 137    if (!inflate)
 138       return len;
 139
 140    return zlib_inflate(out, len);
 141 }
 142
 143 static void
 144 print_help(const char *progname, FILE *file)
 145 {
 146    fprintf(file,
 147            "Usage: %s [OPTION]... [FILE]\n"
 148            "Convert an Intel GPU i915 error state to an aub file.\n"
 149            "  -h, --help          display this help and exit\n"
 150            "  -o, --output=FILE   the output aub file (default FILE.aub)\n",
 151            progname);
 152 }
 153
 154 struct bo {
 155    enum address_space {
 156       PPGTT,
 157       GGTT,
 158    } gtt;
 159    enum bo_type {
 160       BO_TYPE_UNKNOWN = 0,
 161       BO_TYPE_BATCH,
 162       BO_TYPE_USER,
 163       BO_TYPE_CONTEXT,
 164       BO_TYPE_RINGBUFFER,
 165       BO_TYPE_STATUS,
 166       BO_TYPE_CONTEXT_WA,
 167    } type;
 168    const char *name;
 169    uint64_t addr;
 170    uint8_t *data;
 171    uint64_t size;
 172
 173    enum drm_i915_gem_engine_class engine_class;
 174    int engine_instance;
 175
 176    struct list_head link;
 177 };
 178
 179 static struct bo *
 180 find_or_create(struct list_head *bo_list, uint64_t addr,
 181                enum address_space gtt,
 182                enum drm_i915_gem_engine_class engine_class,
 183                int engine_instance)
 184 {
 185    list_for_each_entry(struct bo, bo_entry, bo_list, link) {
 186       if (bo_entry->addr == addr &&
 187           bo_entry->gtt == gtt &&
 188           bo_entry->engine_class == engine_class &&
 189           bo_entry->engine_instance == engine_instance)
 190          return bo_entry;
 191    }
 192
 193    struct bo *new_bo = calloc(1, sizeof(*new_bo));
 194    new_bo->addr = addr;
 195    new_bo->gtt = gtt;
 196    new_bo->engine_class = engine_class;
 197    new_bo->engine_instance = engine_instance;
 198    list_addtail(&new_bo->link, bo_list);
 199
 200    return new_bo;
 201 }
 202
 203 static void
 204 engine_from_name(const char *engine_name,
 205                  enum drm_i915_gem_engine_class *engine_class,
 206                  int *engine_instance)
 207 {
 208    const struct {
 209       const char *match;
 210       enum drm_i915_gem_engine_class engine_class;
 211       bool parse_instance;
 212    } rings[] = {
 213       { "rcs", I915_ENGINE_CLASS_RENDER, true },
 214       { "vcs", I915_ENGINE_CLASS_VIDEO, true },
 215       { "vecs", I915_ENGINE_CLASS_VIDEO_ENHANCE, true },
 216       { "bcs", I915_ENGINE_CLASS_COPY, true },
 217       { "global", I915_ENGINE_CLASS_INVALID, false },
 218       { "render command stream", I915_ENGINE_CLASS_RENDER, false },
 219       { "blt command stream", I915_ENGINE_CLASS_COPY, false },
 220       { "bsd command stream", I915_ENGINE_CLASS_VIDEO, false },
 221       { "vebox command stream", I915_ENGINE_CLASS_VIDEO_ENHANCE, false },
 222       { NULL, I915_ENGINE_CLASS_INVALID },
 223    }, *r;
 224
 225    for (r = rings; r->match; r++) {
 226       if (strncasecmp(engine_name, r->match, strlen(r->match)) == 0) {
 227          *engine_class = r->engine_class;
 228          if (r->parse_instance)
 229             *engine_instance = strtol(engine_name + strlen(r->match), NULL, 10);
 230          else
 231             *engine_instance = 0;
 232          return;
 233       }
 234    }
 235
 236    fail("Unknown engine %s\n", engine_name);
 237 }
 238
 239 int
 240 main(int argc, char *argv[])
 241 {
 242    int i, c;
 243    bool help = false, verbose;
 244    char *out_filename = NULL, *in_filename = NULL;
 245    const struct option aubinator_opts[] = {
 246       { "help",       no_argument,       NULL,     'h' },
 247       { "output",     required_argument, NULL,     'o' },
 248       { "verbose",    no_argument,       NULL,     'v' },
 249       { NULL,         0,                 NULL,     0 }
 250    };
 251
 252    i = 0;
 253    while ((c = getopt_long(argc, argv, "ho:v", aubinator_opts, &i)) != -1) {
 254       switch (c) {
 255       case 'h':
 256          help = true;
 257          break;
 258       case 'o':
 259          out_filename = strdup(optarg);
 260          break;
 261       case 'v':
 262          verbose = true;
 263          break;
 264       default:
 265          break;
 266       }
 267    }
 268
 269    if (optind < argc)
 270       in_filename = argv[optind++];
 271
 272    if (help || argc == 1 || !in_filename) {
 273       print_help(argv[0], stderr);
 274       return in_filename ? EXIT_SUCCESS : EXIT_FAILURE;
 275    }
 276
 277    if (out_filename == NULL) {
 278       int out_filename_size = strlen(in_filename) + 5;
 279       out_filename = malloc(out_filename_size);
 280       snprintf(out_filename, out_filename_size, "%s.aub", in_filename);
 281    }
 282
 283    FILE *err_file = fopen(in_filename, "r");
 284    fail_if(!err_file, "Failed to open error file \"%s\": %m\n", in_filename);
 285
 286    FILE *aub_file = fopen(out_filename, "w");
 287    fail_if(!aub_file, "Failed to open aub file \"%s\": %m\n", in_filename);
 288
 289    struct aub_file aub = {};
 290
 291    enum drm_i915_gem_engine_class active_engine_class = I915_ENGINE_CLASS_INVALID;
 292    int active_engine_instance = -1;
 293
 294    enum address_space active_gtt = PPGTT;
 295
 296    struct {
 297       struct {
 298          uint32_t ring_buffer_head;
 299          uint32_t ring_buffer_tail;
 300       } instances[3];
 301    } engines[I915_ENGINE_CLASS_VIDEO_ENHANCE + 1];
 302    memset(engines, 0, sizeof(engines));
 303
 304    int num_ring_bos = 0;
 305
 306    struct list_head bo_list;
 307    list_inithead(&bo_list);
 308
 309    struct bo *last_bo = NULL;
 310
 311    char *line = NULL;
 312    size_t line_size;
 313    while (getline(&line, &line_size, err_file) > 0) {
 314       const char *pci_id_start = strstr(line, "PCI ID");
 315       if (pci_id_start) {
 316          int pci_id;
 317          int matched = sscanf(line, "PCI ID: 0x%04x\n", &pci_id);
 318          fail_if(!matched, "Invalid error state file!\n");
 319
 320          aub_file_init(&aub, aub_file,
 321                        NULL, pci_id, "error_state");
 322          if (verbose)
 323             aub.verbose_log_file = stdout;
 324          fail_if(!aub_use_execlists(&aub),
 325                  "%s currently only works on gen8+\n", argv[0]);
 326          continue;
 327       }
 328
 329       if (strstr(line, " command stream:")) {
 330          engine_from_name(line, &active_engine_class, &active_engine_instance);
 331          continue;
 332       }
 333
 334       if (sscanf(line, "  ring->head: 0x%x\n",
 335                  &engines[
 336                     active_engine_class].instances[
 337                        active_engine_instance].ring_buffer_head) == 1) {
 338          continue;
 339       }
 340
 341       if (sscanf(line, "  ring->tail: 0x%x\n",
 342                  &engines[
 343                     active_engine_class].instances[
 344                        active_engine_instance].ring_buffer_tail) == 1) {
 345          continue;
 346       }
 347
 348       const char *active_start = "Active (";
 349       if (strncmp(line, active_start, strlen(active_start)) == 0) {
 350          char *ring = line + strlen(active_start);
 351
 352          engine_from_name(ring, &active_engine_class, &active_engine_instance);
 353          active_gtt = PPGTT;
 354
 355          char *count = strchr(ring, '[');
 356          fail_if(!count || sscanf(count, "[%d]:", &num_ring_bos) < 1,
 357                  "Failed to parse BO table header\n");
 358          continue;
 359       }
 360
 361       const char *global_start = "Pinned (global) [";
 362       if (strncmp(line, global_start, strlen(global_start)) == 0) {
 363          active_engine_class = I915_ENGINE_CLASS_INVALID;
 364          active_engine_instance = -1;
 365          active_gtt = GGTT;
 366          continue;
 367       }
 368
 369       if (num_ring_bos > 0) {
 370          unsigned hi, lo, size;
 371          if (sscanf(line, " %x_%x %d", &hi, &lo, &size) == 3) {
 372             assert(aub_use_execlists(&aub));
 373             struct bo *bo_entry = find_or_create(&bo_list, ((uint64_t)hi) << 32 | lo,
 374                                                  active_gtt,
 375                                                  active_engine_class,
 376                                                  active_engine_instance);
 377             bo_entry->size = size;
 378             num_ring_bos--;
 379          } else {
 380             fail("Not enough BO entries in the active table\n");
 381          }
 382          continue;
 383       }
 384
 385       if (line[0] == ':' || line[0] == '~') {
 386          if (!last_bo || last_bo->type == BO_TYPE_UNKNOWN)
 387             continue;
 388
 389          int count = ascii85_decode(line+1, (uint32_t **) &last_bo->data, line[0] == ':');
 390          fail_if(count == 0, "ASCII85 decode failed.\n");
 391          last_bo->size = count * 4;
 392          continue;
 393       }
 394
 395       char *dashes = strstr(line, " --- ");
 396       if (dashes) {
 397          dashes += 5;
 398
 399          engine_from_name(line, &active_engine_class, &active_engine_instance);
 400
 401          uint32_t hi, lo;
 402          char *bo_address_str = strchr(dashes, '=');
 403          if (!bo_address_str || sscanf(bo_address_str, "= 0x%08x %08x\n", &hi, &lo) != 2)
 404             continue;
 405
 406          const struct {
 407             const char *match;
 408             enum bo_type type;
 409             enum address_space gtt;
 410          } bo_types[] = {
 411             { "gtt_offset", BO_TYPE_BATCH,      PPGTT },
 412             { "user",       BO_TYPE_USER,       PPGTT },
 413             { "HW context", BO_TYPE_CONTEXT,    GGTT },
 414             { "ringbuffer", BO_TYPE_RINGBUFFER, GGTT },
 415             { "HW Status",  BO_TYPE_STATUS,     GGTT },
 416             { "WA context", BO_TYPE_CONTEXT_WA, GGTT },
 417             { "unknown",    BO_TYPE_UNKNOWN,    GGTT },
 418          }, *b;
 419
 420          for (b = bo_types; b->type != BO_TYPE_UNKNOWN; b++) {
 421             if (strncasecmp(dashes, b->match, strlen(b->match)) == 0)
 422                break;
 423          }
 424
 425          last_bo = find_or_create(&bo_list, ((uint64_t) hi) << 32 | lo,
 426                                   b->gtt,
 427                                   active_engine_class, active_engine_instance);
 428
 429          /* The batch buffer will appear twice as gtt_offset and user. Only
 430           * keep the batch type.
 431           */
 432          if (last_bo->type == BO_TYPE_UNKNOWN) {
 433             last_bo->type = b->type;
 434             last_bo->name = b->match;
 435          }
 436
 437          continue;
 438       }
 439    }
 440
 441    if (verbose) {
 442       fprintf(stdout, "BOs found:\n");
 443       list_for_each_entry(struct bo, bo_entry, &bo_list, link) {
 444          fprintf(stdout, "\t type=%i addr=0x%016" PRIx64 " size=%" PRIu64 "\n",
 445                  bo_entry->type, bo_entry->addr, bo_entry->size);
 446       }
 447    }
 448
 449    /* Find the batch that trigger the hang */
 450    struct bo *batch_bo = NULL;
 451    list_for_each_entry(struct bo, bo_entry, &bo_list, link) {
 452       if (bo_entry->type == BO_TYPE_BATCH) {
 453          batch_bo = bo_entry;
 454          break;
 455       }
 456    }
 457    fail_if(!batch_bo, "Failed to find batch buffer.\n");
 458
 459    /* Add all the BOs to the aub file */
 460    struct bo *hwsp_bo = NULL;
 461    list_for_each_entry(struct bo, bo_entry, &bo_list, link) {
 462       switch (bo_entry->type) {
 463       case BO_TYPE_BATCH:
 464          aub_map_ppgtt(&aub, bo_entry->addr, bo_entry->size);
 465          aub_write_trace_block(&aub, AUB_TRACE_TYPE_BATCH,
 466                                bo_entry->data, bo_entry->size, bo_entry->addr);
 467          break;
 468       case BO_TYPE_USER:
 469          aub_map_ppgtt(&aub, bo_entry->addr, bo_entry->size);
 470          aub_write_trace_block(&aub, AUB_TRACE_TYPE_NOTYPE,
 471                                bo_entry->data, bo_entry->size, bo_entry->addr);
 472          break;
 473       case BO_TYPE_CONTEXT:
 474          if (bo_entry->engine_class == batch_bo->engine_class &&
 475              bo_entry->engine_instance == batch_bo->engine_instance) {
 476             hwsp_bo = bo_entry;
 477
 478             uint32_t *context = (uint32_t *) (bo_entry->data + 4096 /* GuC */ + 4096 /* HWSP */);
 479
 480             if (context[1] == 0) {
 481                fprintf(stderr,
 482                        "Invalid context image data.\n"
 483                        "This is likely a kernel issue : https://bugs.freedesktop.org/show_bug.cgi?id=107691\n");
 484             }
 485
 486             /* Update the ring buffer at the last known location. */
 487             context[5] = engines[bo_entry->engine_class].instances[bo_entry->engine_instance].ring_buffer_head;
 488             context[7] = engines[bo_entry->engine_class].instances[bo_entry->engine_instance].ring_buffer_tail;
 489             fprintf(stdout, "engine start=0x%x head/tail=0x%x/0x%x\n",
 490                     context[9], context[5], context[7]);
 491
 492             /* The error state doesn't provide a dump of the page tables, so
 493              * we have to provide our own, that's easy enough.
 494              */
 495             context[49] = aub.pml4.phys_addr >> 32;
 496             context[51] = aub.pml4.phys_addr & 0xffffffff;
 497
 498             fprintf(stdout, "context dump:\n");
 499             for (int i = 0; i < 60; i++) {
 500                if (i % 4 == 0)
 501                   fprintf(stdout, "\n 0x%08" PRIx64 ": ", bo_entry->addr + 8192 + i * 4);
 502                fprintf(stdout, "0x%08x ", context[i]);
 503             }
 504             fprintf(stdout, "\n");
 505
 506          }
 507          aub_write_ggtt(&aub, bo_entry->addr, bo_entry->size, bo_entry->data);
 508          break;
 509       case BO_TYPE_RINGBUFFER:
 510       case BO_TYPE_STATUS:
 511       case BO_TYPE_CONTEXT_WA:
 512          aub_write_ggtt(&aub, bo_entry->addr, bo_entry->size, bo_entry->data);
 513          break;
 514       case BO_TYPE_UNKNOWN:
 515          if (bo_entry->gtt == PPGTT) {
 516             aub_map_ppgtt(&aub, bo_entry->addr, bo_entry->size);
 517             if (bo_entry->data) {
 518                aub_write_trace_block(&aub, AUB_TRACE_TYPE_NOTYPE,
 519                                      bo_entry->data, bo_entry->size, bo_entry->addr);
 520             }
 521          } else {
 522             if (bo_entry->size > 0) {
 523                void *zero_data = calloc(1, bo_entry->size);
 524                aub_write_ggtt(&aub, bo_entry->addr, bo_entry->size, zero_data);
 525                free(zero_data);
 526             }
 527          }
 528          break;
 529       default:
 530          break;
 531       }
 532    }
 533
 534    fail_if(!hwsp_bo, "Failed to find Context buffer.\n");
 535    aub_write_context_execlists(&aub, hwsp_bo->addr + 4096 /* skip GuC page */, hwsp_bo->engine_class);
 536
 537    /* Cleanup */
 538    list_for_each_entry_safe(struct bo, bo_entry, &bo_list, link) {
 539       list_del(&bo_entry->link);
 540       free(bo_entry->data);
 541       free(bo_entry);
 542    }
 543
 544    free(out_filename);
 545    free(line);
 546    if(err_file) {
 547       fclose(err_file);
 548    }
 549    if(aub.file) {
 550       aub_file_finish(&aub);
 551    } else if(aub_file) {
 552       fclose(aub_file);
 553    }
 554    return EXIT_SUCCESS;
 555 }
 556
 557 /* vim: set ts=8 sw=8 tw=0 cino=:0,(0 noet :*/