src/import/pull-job.c

   1 /* SPDX-License-Identifier: LGPL-2.1+ */
   2
   3 #include <sys/xattr.h>
   4
   5 #include "alloc-util.h"
   6 #include "fd-util.h"
   7 #include "gcrypt-util.h"
   8 #include "hexdecoct.h"
   9 #include "import-util.h"
  10 #include "io-util.h"
  11 #include "machine-pool.h"
  12 #include "parse-util.h"
  13 #include "pull-common.h"
  14 #include "pull-job.h"
  15 #include "string-util.h"
  16 #include "strv.h"
  17 #include "xattr-util.h"
  18
  19 PullJob* pull_job_unref(PullJob *j) {
  20         if (!j)
  21                 return NULL;
  22
  23         curl_glue_remove_and_free(j->glue, j->curl);
  24         curl_slist_free_all(j->request_header);
  25
  26         safe_close(j->disk_fd);
  27
  28         import_compress_free(&j->compress);
  29
  30         if (j->checksum_context)
  31                 gcry_md_close(j->checksum_context);
  32
  33         free(j->url);
  34         free(j->etag);
  35         strv_free(j->old_etags);
  36         free(j->payload);
  37         free(j->checksum);
  38
  39         return mfree(j);
  40 }
  41
  42 static void pull_job_finish(PullJob *j, int ret) {
  43         assert(j);
  44
  45         if (IN_SET(j->state, PULL_JOB_DONE, PULL_JOB_FAILED))
  46                 return;
  47
  48         if (ret == 0) {
  49                 j->state = PULL_JOB_DONE;
  50                 j->progress_percent = 100;
  51                 log_info("Download of %s complete.", j->url);
  52         } else {
  53                 j->state = PULL_JOB_FAILED;
  54                 j->error = ret;
  55         }
  56
  57         if (j->on_finished)
  58                 j->on_finished(j);
  59 }
  60
  61 static int pull_job_restart(PullJob *j) {
  62         int r;
  63         char *chksum_url = NULL;
  64
  65         r = import_url_change_last_component(j->url, "SHA256SUMS", &chksum_url);
  66         if (r < 0)
  67                 return r;
  68
  69         free(j->url);
  70         j->url = chksum_url;
  71         j->state = PULL_JOB_INIT;
  72         j->payload = mfree(j->payload);
  73         j->payload_size = 0;
  74         j->payload_allocated = 0;
  75         j->written_compressed = 0;
  76         j->written_uncompressed = 0;
  77
  78         r = pull_job_begin(j);
  79         if (r < 0)
  80                 return r;
  81
  82         return 0;
  83 }
  84
  85 void pull_job_curl_on_finished(CurlGlue *g, CURL *curl, CURLcode result) {
  86         PullJob *j = NULL;
  87         CURLcode code;
  88         long status;
  89         int r;
  90
  91         if (curl_easy_getinfo(curl, CURLINFO_PRIVATE, (char **)&j) != CURLE_OK)
  92                 return;
  93
  94         if (!j || IN_SET(j->state, PULL_JOB_DONE, PULL_JOB_FAILED))
  95                 return;
  96
  97         if (result != CURLE_OK) {
  98                 log_error("Transfer failed: %s", curl_easy_strerror(result));
  99                 r = -EIO;
 100                 goto finish;
 101         }
 102
 103         code = curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &status);
 104         if (code != CURLE_OK) {
 105                 log_error("Failed to retrieve response code: %s", curl_easy_strerror(code));
 106                 r = -EIO;
 107                 goto finish;
 108         } else if (status == 304) {
 109                 log_info("Image already downloaded. Skipping download.");
 110                 j->etag_exists = true;
 111                 r = 0;
 112                 goto finish;
 113         } else if (status >= 300) {
 114                 if (status == 404 && j->style == VERIFICATION_PER_FILE) {
 115
 116                         /* retry pull job with SHA256SUMS file */
 117                         r = pull_job_restart(j);
 118                         if (r < 0)
 119                                 goto finish;
 120
 121                         code = curl_easy_getinfo(j->curl, CURLINFO_RESPONSE_CODE, &status);
 122                         if (code != CURLE_OK) {
 123                                 log_error("Failed to retrieve response code: %s", curl_easy_strerror(code));
 124                                 r = -EIO;
 125                                 goto finish;
 126                         }
 127
 128                         if (status == 0) {
 129                                 j->style = VERIFICATION_PER_DIRECTORY;
 130                                 return;
 131                         }
 132                 }
 133
 134                 log_error("HTTP request to %s failed with code %li.", j->url, status);
 135                 r = -EIO;
 136                 goto finish;
 137         } else if (status < 200) {
 138                 log_error("HTTP request to %s finished with unexpected code %li.", j->url, status);
 139                 r = -EIO;
 140                 goto finish;
 141         }
 142
 143         if (j->state != PULL_JOB_RUNNING) {
 144                 log_error("Premature connection termination.");
 145                 r = -EIO;
 146                 goto finish;
 147         }
 148
 149         if (j->content_length != (uint64_t) -1 &&
 150             j->content_length != j->written_compressed) {
 151                 log_error("Download truncated.");
 152                 r = -EIO;
 153                 goto finish;
 154         }
 155
 156         if (j->checksum_context) {
 157                 uint8_t *k;
 158
 159                 k = gcry_md_read(j->checksum_context, GCRY_MD_SHA256);
 160                 if (!k) {
 161                         log_error("Failed to get checksum.");
 162                         r = -EIO;
 163                         goto finish;
 164                 }
 165
 166                 j->checksum = hexmem(k, gcry_md_get_algo_dlen(GCRY_MD_SHA256));
 167                 if (!j->checksum) {
 168                         r = log_oom();
 169                         goto finish;
 170                 }
 171
 172                 log_debug("SHA256 of %s is %s.", j->url, j->checksum);
 173         }
 174
 175         if (j->disk_fd >= 0 && j->allow_sparse) {
 176                 /* Make sure the file size is right, in case the file was
 177                  * sparse and we just seeked for the last part */
 178
 179                 if (ftruncate(j->disk_fd, j->written_uncompressed) < 0) {
 180                         r = log_error_errno(errno, "Failed to truncate file: %m");
 181                         goto finish;
 182                 }
 183
 184                 if (j->etag)
 185                         (void) fsetxattr(j->disk_fd, "user.source_etag", j->etag, strlen(j->etag), 0);
 186                 if (j->url)
 187                         (void) fsetxattr(j->disk_fd, "user.source_url", j->url, strlen(j->url), 0);
 188
 189                 if (j->mtime != 0) {
 190                         struct timespec ut[2];
 191
 192                         timespec_store(&ut[0], j->mtime);
 193                         ut[1] = ut[0];
 194                         (void) futimens(j->disk_fd, ut);
 195
 196                         (void) fd_setcrtime(j->disk_fd, j->mtime);
 197                 }
 198         }
 199
 200         r = 0;
 201
 202 finish:
 203         pull_job_finish(j, r);
 204 }
 205
 206 static int pull_job_write_uncompressed(const void *p, size_t sz, void *userdata) {
 207         PullJob *j = userdata;
 208         ssize_t n;
 209
 210         assert(j);
 211         assert(p);
 212
 213         if (sz <= 0)
 214                 return 0;
 215
 216         if (j->written_uncompressed + sz < j->written_uncompressed)
 217                 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW),
 218                                        "File too large, overflow");
 219
 220         if (j->written_uncompressed + sz > j->uncompressed_max)
 221                 return log_error_errno(SYNTHETIC_ERRNO(EFBIG),
 222                                        "File overly large, refusing");
 223
 224         if (j->disk_fd >= 0) {
 225
 226                 if (j->allow_sparse)
 227                         n = sparse_write(j->disk_fd, p, sz, 64);
 228                 else {
 229                         n = write(j->disk_fd, p, sz);
 230                         if (n < 0)
 231                                 n = -errno;
 232                 }
 233                 if (n < 0)
 234                         return log_error_errno((int) n, "Failed to write file: %m");
 235                 if ((size_t) n < sz)
 236                         return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short write");
 237         } else {
 238
 239                 if (!GREEDY_REALLOC(j->payload, j->payload_allocated, j->payload_size + sz))
 240                         return log_oom();
 241
 242                 memcpy(j->payload + j->payload_size, p, sz);
 243                 j->payload_size += sz;
 244         }
 245
 246         j->written_uncompressed += sz;
 247
 248         return 0;
 249 }
 250
 251 static int pull_job_write_compressed(PullJob *j, void *p, size_t sz) {
 252         int r;
 253
 254         assert(j);
 255         assert(p);
 256
 257         if (sz <= 0)
 258                 return 0;
 259
 260         if (j->written_compressed + sz < j->written_compressed)
 261                 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW), "File too large, overflow");
 262
 263         if (j->written_compressed + sz > j->compressed_max)
 264                 return log_error_errno(SYNTHETIC_ERRNO(EFBIG), "File overly large, refusing.");
 265
 266         if (j->content_length != (uint64_t) -1 &&
 267             j->written_compressed + sz > j->content_length)
 268                 return log_error_errno(SYNTHETIC_ERRNO(EFBIG),
 269                                        "Content length incorrect.");
 270
 271         if (j->checksum_context)
 272                 gcry_md_write(j->checksum_context, p, sz);
 273
 274         r = import_uncompress(&j->compress, p, sz, pull_job_write_uncompressed, j);
 275         if (r < 0)
 276                 return r;
 277
 278         j->written_compressed += sz;
 279
 280         return 0;
 281 }
 282
 283 static int pull_job_open_disk(PullJob *j) {
 284         int r;
 285
 286         assert(j);
 287
 288         if (j->on_open_disk) {
 289                 r = j->on_open_disk(j);
 290                 if (r < 0)
 291                         return r;
 292         }
 293
 294         if (j->disk_fd >= 0) {
 295                 /* Check if we can do sparse files */
 296
 297                 if (lseek(j->disk_fd, SEEK_SET, 0) == 0)
 298                         j->allow_sparse = true;
 299                 else {
 300                         if (errno != ESPIPE)
 301                                 return log_error_errno(errno, "Failed to seek on file descriptor: %m");
 302
 303                         j->allow_sparse = false;
 304                 }
 305         }
 306
 307         if (j->calc_checksum) {
 308                 initialize_libgcrypt(false);
 309
 310                 if (gcry_md_open(&j->checksum_context, GCRY_MD_SHA256, 0) != 0)
 311                         return log_error_errno(SYNTHETIC_ERRNO(EIO),
 312                                                "Failed to initialize hash context.");
 313         }
 314
 315         return 0;
 316 }
 317
 318 static int pull_job_detect_compression(PullJob *j) {
 319         _cleanup_free_ uint8_t *stub = NULL;
 320         size_t stub_size;
 321
 322         int r;
 323
 324         assert(j);
 325
 326         r = import_uncompress_detect(&j->compress, j->payload, j->payload_size);
 327         if (r < 0)
 328                 return log_error_errno(r, "Failed to initialize compressor: %m");
 329         if (r == 0)
 330                 return 0;
 331
 332         log_debug("Stream is compressed: %s", import_compress_type_to_string(j->compress.type));
 333
 334         r = pull_job_open_disk(j);
 335         if (r < 0)
 336                 return r;
 337
 338         /* Now, take the payload we read so far, and decompress it */
 339         stub = j->payload;
 340         stub_size = j->payload_size;
 341
 342         j->payload = NULL;
 343         j->payload_size = 0;
 344         j->payload_allocated = 0;
 345
 346         j->state = PULL_JOB_RUNNING;
 347
 348         r = pull_job_write_compressed(j, stub, stub_size);
 349         if (r < 0)
 350                 return r;
 351
 352         return 0;
 353 }
 354
 355 static size_t pull_job_write_callback(void *contents, size_t size, size_t nmemb, void *userdata) {
 356         PullJob *j = userdata;
 357         size_t sz = size * nmemb;
 358         int r;
 359
 360         assert(contents);
 361         assert(j);
 362
 363         switch (j->state) {
 364
 365         case PULL_JOB_ANALYZING:
 366                 /* Let's first check what it actually is */
 367
 368                 if (!GREEDY_REALLOC(j->payload, j->payload_allocated, j->payload_size + sz)) {
 369                         r = log_oom();
 370                         goto fail;
 371                 }
 372
 373                 memcpy(j->payload + j->payload_size, contents, sz);
 374                 j->payload_size += sz;
 375
 376                 r = pull_job_detect_compression(j);
 377                 if (r < 0)
 378                         goto fail;
 379
 380                 break;
 381
 382         case PULL_JOB_RUNNING:
 383
 384                 r = pull_job_write_compressed(j, contents, sz);
 385                 if (r < 0)
 386                         goto fail;
 387
 388                 break;
 389
 390         case PULL_JOB_DONE:
 391         case PULL_JOB_FAILED:
 392                 r = -ESTALE;
 393                 goto fail;
 394
 395         default:
 396                 assert_not_reached("Impossible state.");
 397         }
 398
 399         return sz;
 400
 401 fail:
 402         pull_job_finish(j, r);
 403         return 0;
 404 }
 405
 406 static size_t pull_job_header_callback(void *contents, size_t size, size_t nmemb, void *userdata) {
 407         PullJob *j = userdata;
 408         size_t sz = size * nmemb;
 409         _cleanup_free_ char *length = NULL, *last_modified = NULL;
 410         char *etag;
 411         int r;
 412
 413         assert(contents);
 414         assert(j);
 415
 416         if (IN_SET(j->state, PULL_JOB_DONE, PULL_JOB_FAILED)) {
 417                 r = -ESTALE;
 418                 goto fail;
 419         }
 420
 421         assert(j->state == PULL_JOB_ANALYZING);
 422
 423         r = curl_header_strdup(contents, sz, "ETag:", &etag);
 424         if (r < 0) {
 425                 log_oom();
 426                 goto fail;
 427         }
 428         if (r > 0) {
 429                 free(j->etag);
 430                 j->etag = etag;
 431
 432                 if (strv_contains(j->old_etags, j->etag)) {
 433                         log_info("Image already downloaded. Skipping download.");
 434                         j->etag_exists = true;
 435                         pull_job_finish(j, 0);
 436                         return sz;
 437                 }
 438
 439                 return sz;
 440         }
 441
 442         r = curl_header_strdup(contents, sz, "Content-Length:", &length);
 443         if (r < 0) {
 444                 log_oom();
 445                 goto fail;
 446         }
 447         if (r > 0) {
 448                 (void) safe_atou64(length, &j->content_length);
 449
 450                 if (j->content_length != (uint64_t) -1) {
 451                         char bytes[FORMAT_BYTES_MAX];
 452
 453                         if (j->content_length > j->compressed_max) {
 454                                 log_error("Content too large.");
 455                                 r = -EFBIG;
 456                                 goto fail;
 457                         }
 458
 459                         log_info("Downloading %s for %s.", format_bytes(bytes, sizeof(bytes), j->content_length), j->url);
 460                 }
 461
 462                 return sz;
 463         }
 464
 465         r = curl_header_strdup(contents, sz, "Last-Modified:", &last_modified);
 466         if (r < 0) {
 467                 log_oom();
 468                 goto fail;
 469         }
 470         if (r > 0) {
 471                 (void) curl_parse_http_time(last_modified, &j->mtime);
 472                 return sz;
 473         }
 474
 475         if (j->on_header) {
 476                 r = j->on_header(j, contents, sz);
 477                 if (r < 0)
 478                         goto fail;
 479         }
 480
 481         return sz;
 482
 483 fail:
 484         pull_job_finish(j, r);
 485         return 0;
 486 }
 487
 488 static int pull_job_progress_callback(void *userdata, curl_off_t dltotal, curl_off_t dlnow, curl_off_t ultotal, curl_off_t ulnow) {
 489         PullJob *j = userdata;
 490         unsigned percent;
 491         usec_t n;
 492
 493         assert(j);
 494
 495         if (dltotal <= 0)
 496                 return 0;
 497
 498         percent = ((100 * dlnow) / dltotal);
 499         n = now(CLOCK_MONOTONIC);
 500
 501         if (n > j->last_status_usec + USEC_PER_SEC &&
 502             percent != j->progress_percent &&
 503             dlnow < dltotal) {
 504                 char buf[FORMAT_TIMESPAN_MAX];
 505
 506                 if (n - j->start_usec > USEC_PER_SEC && dlnow > 0) {
 507                         char y[FORMAT_BYTES_MAX];
 508                         usec_t left, done;
 509
 510                         done = n - j->start_usec;
 511                         left = (usec_t) (((double) done * (double) dltotal) / dlnow) - done;
 512
 513                         log_info("Got %u%% of %s. %s left at %s/s.",
 514                                  percent,
 515                                  j->url,
 516                                  format_timespan(buf, sizeof(buf), left, USEC_PER_SEC),
 517                                  format_bytes(y, sizeof(y), (uint64_t) ((double) dlnow / ((double) done / (double) USEC_PER_SEC))));
 518                 } else
 519                         log_info("Got %u%% of %s.", percent, j->url);
 520
 521                 j->progress_percent = percent;
 522                 j->last_status_usec = n;
 523
 524                 if (j->on_progress)
 525                         j->on_progress(j);
 526         }
 527
 528         return 0;
 529 }
 530
 531 int pull_job_new(PullJob **ret, const char *url, CurlGlue *glue, void *userdata) {
 532         _cleanup_(pull_job_unrefp) PullJob *j = NULL;
 533         _cleanup_free_ char *u = NULL;
 534
 535         assert(url);
 536         assert(glue);
 537         assert(ret);
 538
 539         u = strdup(url);
 540         if (!u)
 541                 return -ENOMEM;
 542
 543         j = new(PullJob, 1);
 544         if (!j)
 545                 return -ENOMEM;
 546
 547         *j = (PullJob) {
 548                 .state = PULL_JOB_INIT,
 549                 .disk_fd = -1,
 550                 .userdata = userdata,
 551                 .glue = glue,
 552                 .content_length = (uint64_t) -1,
 553                 .start_usec = now(CLOCK_MONOTONIC),
 554                 .compressed_max = 64LLU * 1024LLU * 1024LLU * 1024LLU, /* 64GB safety limit */
 555                 .uncompressed_max = 64LLU * 1024LLU * 1024LLU * 1024LLU, /* 64GB safety limit */
 556                 .style = VERIFICATION_STYLE_UNSET,
 557                 .url = TAKE_PTR(u),
 558         };
 559
 560         *ret = TAKE_PTR(j);
 561
 562         return 0;
 563 }
 564
 565 int pull_job_begin(PullJob *j) {
 566         int r;
 567
 568         assert(j);
 569
 570         if (j->state != PULL_JOB_INIT)
 571                 return -EBUSY;
 572
 573         r = curl_glue_make(&j->curl, j->url, j);
 574         if (r < 0)
 575                 return r;
 576
 577         if (!strv_isempty(j->old_etags)) {
 578                 _cleanup_free_ char *cc = NULL, *hdr = NULL;
 579
 580                 cc = strv_join(j->old_etags, ", ");
 581                 if (!cc)
 582                         return -ENOMEM;
 583
 584                 hdr = strappend("If-None-Match: ", cc);
 585                 if (!hdr)
 586                         return -ENOMEM;
 587
 588                 if (!j->request_header) {
 589                         j->request_header = curl_slist_new(hdr, NULL);
 590                         if (!j->request_header)
 591                                 return -ENOMEM;
 592                 } else {
 593                         struct curl_slist *l;
 594
 595                         l = curl_slist_append(j->request_header, hdr);
 596                         if (!l)
 597                                 return -ENOMEM;
 598
 599                         j->request_header = l;
 600                 }
 601         }
 602
 603         if (j->request_header) {
 604                 if (curl_easy_setopt(j->curl, CURLOPT_HTTPHEADER, j->request_header) != CURLE_OK)
 605                         return -EIO;
 606         }
 607
 608         if (curl_easy_setopt(j->curl, CURLOPT_WRITEFUNCTION, pull_job_write_callback) != CURLE_OK)
 609                 return -EIO;
 610
 611         if (curl_easy_setopt(j->curl, CURLOPT_WRITEDATA, j) != CURLE_OK)
 612                 return -EIO;
 613
 614         if (curl_easy_setopt(j->curl, CURLOPT_HEADERFUNCTION, pull_job_header_callback) != CURLE_OK)
 615                 return -EIO;
 616
 617         if (curl_easy_setopt(j->curl, CURLOPT_HEADERDATA, j) != CURLE_OK)
 618                 return -EIO;
 619
 620         if (curl_easy_setopt(j->curl, CURLOPT_XFERINFOFUNCTION, pull_job_progress_callback) != CURLE_OK)
 621                 return -EIO;
 622
 623         if (curl_easy_setopt(j->curl, CURLOPT_XFERINFODATA, j) != CURLE_OK)
 624                 return -EIO;
 625
 626         if (curl_easy_setopt(j->curl, CURLOPT_NOPROGRESS, 0) != CURLE_OK)
 627                 return -EIO;
 628
 629         r = curl_glue_add(j->glue, j->curl);
 630         if (r < 0)
 631                 return r;
 632
 633         j->state = PULL_JOB_ANALYZING;
 634
 635         return 0;
 636 }