]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/import/pull-job.c
bus: report builtins in GetManagedObjects() (#5799)
[thirdparty/systemd.git] / src / import / pull-job.c
1 /***
2 This file is part of systemd.
3
4 Copyright 2015 Lennart Poettering
5
6 systemd is free software; you can redistribute it and/or modify it
7 under the terms of the GNU Lesser General Public License as published by
8 the Free Software Foundation; either version 2.1 of the License, or
9 (at your option) any later version.
10
11 systemd is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public License
17 along with systemd; If not, see <http://www.gnu.org/licenses/>.
18 ***/
19
20 #include <sys/xattr.h>
21
22 #include "alloc-util.h"
23 #include "fd-util.h"
24 #include "hexdecoct.h"
25 #include "io-util.h"
26 #include "machine-pool.h"
27 #include "parse-util.h"
28 #include "pull-job.h"
29 #include "string-util.h"
30 #include "strv.h"
31 #include "xattr-util.h"
32 #include "pull-common.h"
33 #include "import-util.h"
34
35 PullJob* pull_job_unref(PullJob *j) {
36 if (!j)
37 return NULL;
38
39 curl_glue_remove_and_free(j->glue, j->curl);
40 curl_slist_free_all(j->request_header);
41
42 safe_close(j->disk_fd);
43
44 import_compress_free(&j->compress);
45
46 if (j->checksum_context)
47 gcry_md_close(j->checksum_context);
48
49 free(j->url);
50 free(j->etag);
51 strv_free(j->old_etags);
52 free(j->payload);
53 free(j->checksum);
54
55 return mfree(j);
56 }
57
58 static void pull_job_finish(PullJob *j, int ret) {
59 assert(j);
60
61 if (j->state == PULL_JOB_DONE ||
62 j->state == PULL_JOB_FAILED)
63 return;
64
65 if (ret == 0) {
66 j->state = PULL_JOB_DONE;
67 j->progress_percent = 100;
68 log_info("Download of %s complete.", j->url);
69 } else {
70 j->state = PULL_JOB_FAILED;
71 j->error = ret;
72 }
73
74 if (j->on_finished)
75 j->on_finished(j);
76 }
77
78 static int pull_job_restart(PullJob *j) {
79 int r;
80 char *chksum_url = NULL;
81
82 r = import_url_change_last_component(j->url, "SHA256SUMS", &chksum_url);
83 if (r < 0)
84 return r;
85
86 free(j->url);
87 free(j->payload);
88 j->url = chksum_url;
89 j->state = PULL_JOB_INIT;
90 j->payload = NULL;
91 j->payload_size = 0;
92 j->payload_allocated = 0;
93 j->written_compressed = 0;
94 j->written_uncompressed = 0;
95 j->written_since_last_grow = 0;
96
97 r = pull_job_begin(j);
98 if (r < 0)
99 return r;
100
101 return 0;
102 }
103
104
105 void pull_job_curl_on_finished(CurlGlue *g, CURL *curl, CURLcode result) {
106 PullJob *j = NULL;
107 CURLcode code;
108 long status;
109 int r;
110
111 if (curl_easy_getinfo(curl, CURLINFO_PRIVATE, (char **)&j) != CURLE_OK)
112 return;
113
114 if (!j || j->state == PULL_JOB_DONE || j->state == PULL_JOB_FAILED)
115 return;
116
117 if (result != CURLE_OK) {
118 log_error("Transfer failed: %s", curl_easy_strerror(result));
119 r = -EIO;
120 goto finish;
121 }
122
123 code = curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &status);
124 if (code != CURLE_OK) {
125 log_error("Failed to retrieve response code: %s", curl_easy_strerror(code));
126 r = -EIO;
127 goto finish;
128 } else if (status == 304) {
129 log_info("Image already downloaded. Skipping download.");
130 j->etag_exists = true;
131 r = 0;
132 goto finish;
133 } else if (status >= 300) {
134 if (status == 404 && j->style == VERIFICATION_PER_FILE) {
135
136 /* retry pull job with SHA256SUMS file */
137 r = pull_job_restart(j);
138 if (r < 0)
139 goto finish;
140
141 code = curl_easy_getinfo(j->curl, CURLINFO_RESPONSE_CODE, &status);
142 if (code != CURLE_OK) {
143 log_error("Failed to retrieve response code: %s", curl_easy_strerror(code));
144 r = -EIO;
145 goto finish;
146 }
147
148 if (status == 0) {
149 j->style = VERIFICATION_PER_DIRECTORY;
150 return;
151 }
152 }
153
154 log_error("HTTP request to %s failed with code %li.", j->url, status);
155 r = -EIO;
156 goto finish;
157 } else if (status < 200) {
158 log_error("HTTP request to %s finished with unexpected code %li.", j->url, status);
159 r = -EIO;
160 goto finish;
161 }
162
163 if (j->state != PULL_JOB_RUNNING) {
164 log_error("Premature connection termination.");
165 r = -EIO;
166 goto finish;
167 }
168
169 if (j->content_length != (uint64_t) -1 &&
170 j->content_length != j->written_compressed) {
171 log_error("Download truncated.");
172 r = -EIO;
173 goto finish;
174 }
175
176 if (j->checksum_context) {
177 uint8_t *k;
178
179 k = gcry_md_read(j->checksum_context, GCRY_MD_SHA256);
180 if (!k) {
181 log_error("Failed to get checksum.");
182 r = -EIO;
183 goto finish;
184 }
185
186 j->checksum = hexmem(k, gcry_md_get_algo_dlen(GCRY_MD_SHA256));
187 if (!j->checksum) {
188 r = log_oom();
189 goto finish;
190 }
191
192 log_debug("SHA256 of %s is %s.", j->url, j->checksum);
193 }
194
195 if (j->disk_fd >= 0 && j->allow_sparse) {
196 /* Make sure the file size is right, in case the file was
197 * sparse and we just seeked for the last part */
198
199 if (ftruncate(j->disk_fd, j->written_uncompressed) < 0) {
200 r = log_error_errno(errno, "Failed to truncate file: %m");
201 goto finish;
202 }
203
204 if (j->etag)
205 (void) fsetxattr(j->disk_fd, "user.source_etag", j->etag, strlen(j->etag), 0);
206 if (j->url)
207 (void) fsetxattr(j->disk_fd, "user.source_url", j->url, strlen(j->url), 0);
208
209 if (j->mtime != 0) {
210 struct timespec ut[2];
211
212 timespec_store(&ut[0], j->mtime);
213 ut[1] = ut[0];
214 (void) futimens(j->disk_fd, ut);
215
216 (void) fd_setcrtime(j->disk_fd, j->mtime);
217 }
218 }
219
220 r = 0;
221
222 finish:
223 pull_job_finish(j, r);
224 }
225
226 static int pull_job_write_uncompressed(const void *p, size_t sz, void *userdata) {
227 PullJob *j = userdata;
228 ssize_t n;
229
230 assert(j);
231 assert(p);
232
233 if (sz <= 0)
234 return 0;
235
236 if (j->written_uncompressed + sz < j->written_uncompressed) {
237 log_error("File too large, overflow");
238 return -EOVERFLOW;
239 }
240
241 if (j->written_uncompressed + sz > j->uncompressed_max) {
242 log_error("File overly large, refusing");
243 return -EFBIG;
244 }
245
246 if (j->disk_fd >= 0) {
247
248 if (j->grow_machine_directory && j->written_since_last_grow >= GROW_INTERVAL_BYTES) {
249 j->written_since_last_grow = 0;
250 grow_machine_directory();
251 }
252
253 if (j->allow_sparse)
254 n = sparse_write(j->disk_fd, p, sz, 64);
255 else
256 n = write(j->disk_fd, p, sz);
257 if (n < 0)
258 return log_error_errno(errno, "Failed to write file: %m");
259 if ((size_t) n < sz) {
260 log_error("Short write");
261 return -EIO;
262 }
263 } else {
264
265 if (!GREEDY_REALLOC(j->payload, j->payload_allocated, j->payload_size + sz))
266 return log_oom();
267
268 memcpy(j->payload + j->payload_size, p, sz);
269 j->payload_size += sz;
270 }
271
272 j->written_uncompressed += sz;
273 j->written_since_last_grow += sz;
274
275 return 0;
276 }
277
278 static int pull_job_write_compressed(PullJob *j, void *p, size_t sz) {
279 int r;
280
281 assert(j);
282 assert(p);
283
284 if (sz <= 0)
285 return 0;
286
287 if (j->written_compressed + sz < j->written_compressed) {
288 log_error("File too large, overflow");
289 return -EOVERFLOW;
290 }
291
292 if (j->written_compressed + sz > j->compressed_max) {
293 log_error("File overly large, refusing.");
294 return -EFBIG;
295 }
296
297 if (j->content_length != (uint64_t) -1 &&
298 j->written_compressed + sz > j->content_length) {
299 log_error("Content length incorrect.");
300 return -EFBIG;
301 }
302
303 if (j->checksum_context)
304 gcry_md_write(j->checksum_context, p, sz);
305
306 r = import_uncompress(&j->compress, p, sz, pull_job_write_uncompressed, j);
307 if (r < 0)
308 return r;
309
310 j->written_compressed += sz;
311
312 return 0;
313 }
314
315 static int pull_job_open_disk(PullJob *j) {
316 int r;
317
318 assert(j);
319
320 if (j->on_open_disk) {
321 r = j->on_open_disk(j);
322 if (r < 0)
323 return r;
324 }
325
326 if (j->disk_fd >= 0) {
327 /* Check if we can do sparse files */
328
329 if (lseek(j->disk_fd, SEEK_SET, 0) == 0)
330 j->allow_sparse = true;
331 else {
332 if (errno != ESPIPE)
333 return log_error_errno(errno, "Failed to seek on file descriptor: %m");
334
335 j->allow_sparse = false;
336 }
337 }
338
339 if (j->calc_checksum) {
340 if (gcry_md_open(&j->checksum_context, GCRY_MD_SHA256, 0) != 0) {
341 log_error("Failed to initialize hash context.");
342 return -EIO;
343 }
344 }
345
346 return 0;
347 }
348
349 static int pull_job_detect_compression(PullJob *j) {
350 _cleanup_free_ uint8_t *stub = NULL;
351 size_t stub_size;
352
353 int r;
354
355 assert(j);
356
357 r = import_uncompress_detect(&j->compress, j->payload, j->payload_size);
358 if (r < 0)
359 return log_error_errno(r, "Failed to initialize compressor: %m");
360 if (r == 0)
361 return 0;
362
363 log_debug("Stream is compressed: %s", import_compress_type_to_string(j->compress.type));
364
365 r = pull_job_open_disk(j);
366 if (r < 0)
367 return r;
368
369 /* Now, take the payload we read so far, and decompress it */
370 stub = j->payload;
371 stub_size = j->payload_size;
372
373 j->payload = NULL;
374 j->payload_size = 0;
375 j->payload_allocated = 0;
376
377 j->state = PULL_JOB_RUNNING;
378
379 r = pull_job_write_compressed(j, stub, stub_size);
380 if (r < 0)
381 return r;
382
383 return 0;
384 }
385
386 static size_t pull_job_write_callback(void *contents, size_t size, size_t nmemb, void *userdata) {
387 PullJob *j = userdata;
388 size_t sz = size * nmemb;
389 int r;
390
391 assert(contents);
392 assert(j);
393
394 switch (j->state) {
395
396 case PULL_JOB_ANALYZING:
397 /* Let's first check what it actually is */
398
399 if (!GREEDY_REALLOC(j->payload, j->payload_allocated, j->payload_size + sz)) {
400 r = log_oom();
401 goto fail;
402 }
403
404 memcpy(j->payload + j->payload_size, contents, sz);
405 j->payload_size += sz;
406
407 r = pull_job_detect_compression(j);
408 if (r < 0)
409 goto fail;
410
411 break;
412
413 case PULL_JOB_RUNNING:
414
415 r = pull_job_write_compressed(j, contents, sz);
416 if (r < 0)
417 goto fail;
418
419 break;
420
421 case PULL_JOB_DONE:
422 case PULL_JOB_FAILED:
423 r = -ESTALE;
424 goto fail;
425
426 default:
427 assert_not_reached("Impossible state.");
428 }
429
430 return sz;
431
432 fail:
433 pull_job_finish(j, r);
434 return 0;
435 }
436
437 static size_t pull_job_header_callback(void *contents, size_t size, size_t nmemb, void *userdata) {
438 PullJob *j = userdata;
439 size_t sz = size * nmemb;
440 _cleanup_free_ char *length = NULL, *last_modified = NULL;
441 char *etag;
442 int r;
443
444 assert(contents);
445 assert(j);
446
447 if (j->state == PULL_JOB_DONE || j->state == PULL_JOB_FAILED) {
448 r = -ESTALE;
449 goto fail;
450 }
451
452 assert(j->state == PULL_JOB_ANALYZING);
453
454 r = curl_header_strdup(contents, sz, "ETag:", &etag);
455 if (r < 0) {
456 log_oom();
457 goto fail;
458 }
459 if (r > 0) {
460 free(j->etag);
461 j->etag = etag;
462
463 if (strv_contains(j->old_etags, j->etag)) {
464 log_info("Image already downloaded. Skipping download.");
465 j->etag_exists = true;
466 pull_job_finish(j, 0);
467 return sz;
468 }
469
470 return sz;
471 }
472
473 r = curl_header_strdup(contents, sz, "Content-Length:", &length);
474 if (r < 0) {
475 log_oom();
476 goto fail;
477 }
478 if (r > 0) {
479 (void) safe_atou64(length, &j->content_length);
480
481 if (j->content_length != (uint64_t) -1) {
482 char bytes[FORMAT_BYTES_MAX];
483
484 if (j->content_length > j->compressed_max) {
485 log_error("Content too large.");
486 r = -EFBIG;
487 goto fail;
488 }
489
490 log_info("Downloading %s for %s.", format_bytes(bytes, sizeof(bytes), j->content_length), j->url);
491 }
492
493 return sz;
494 }
495
496 r = curl_header_strdup(contents, sz, "Last-Modified:", &last_modified);
497 if (r < 0) {
498 log_oom();
499 goto fail;
500 }
501 if (r > 0) {
502 (void) curl_parse_http_time(last_modified, &j->mtime);
503 return sz;
504 }
505
506 if (j->on_header) {
507 r = j->on_header(j, contents, sz);
508 if (r < 0)
509 goto fail;
510 }
511
512 return sz;
513
514 fail:
515 pull_job_finish(j, r);
516 return 0;
517 }
518
519 static int pull_job_progress_callback(void *userdata, curl_off_t dltotal, curl_off_t dlnow, curl_off_t ultotal, curl_off_t ulnow) {
520 PullJob *j = userdata;
521 unsigned percent;
522 usec_t n;
523
524 assert(j);
525
526 if (dltotal <= 0)
527 return 0;
528
529 percent = ((100 * dlnow) / dltotal);
530 n = now(CLOCK_MONOTONIC);
531
532 if (n > j->last_status_usec + USEC_PER_SEC &&
533 percent != j->progress_percent &&
534 dlnow < dltotal) {
535 char buf[FORMAT_TIMESPAN_MAX];
536
537 if (n - j->start_usec > USEC_PER_SEC && dlnow > 0) {
538 char y[FORMAT_BYTES_MAX];
539 usec_t left, done;
540
541 done = n - j->start_usec;
542 left = (usec_t) (((double) done * (double) dltotal) / dlnow) - done;
543
544 log_info("Got %u%% of %s. %s left at %s/s.",
545 percent,
546 j->url,
547 format_timespan(buf, sizeof(buf), left, USEC_PER_SEC),
548 format_bytes(y, sizeof(y), (uint64_t) ((double) dlnow / ((double) done / (double) USEC_PER_SEC))));
549 } else
550 log_info("Got %u%% of %s.", percent, j->url);
551
552 j->progress_percent = percent;
553 j->last_status_usec = n;
554
555 if (j->on_progress)
556 j->on_progress(j);
557 }
558
559 return 0;
560 }
561
562 int pull_job_new(PullJob **ret, const char *url, CurlGlue *glue, void *userdata) {
563 _cleanup_(pull_job_unrefp) PullJob *j = NULL;
564
565 assert(url);
566 assert(glue);
567 assert(ret);
568
569 j = new0(PullJob, 1);
570 if (!j)
571 return -ENOMEM;
572
573 j->state = PULL_JOB_INIT;
574 j->disk_fd = -1;
575 j->userdata = userdata;
576 j->glue = glue;
577 j->content_length = (uint64_t) -1;
578 j->start_usec = now(CLOCK_MONOTONIC);
579 j->compressed_max = j->uncompressed_max = 64LLU * 1024LLU * 1024LLU * 1024LLU; /* 64GB safety limit */
580 j->style = VERIFICATION_STYLE_UNSET;
581
582 j->url = strdup(url);
583 if (!j->url)
584 return -ENOMEM;
585
586 *ret = j;
587 j = NULL;
588
589 return 0;
590 }
591
592 int pull_job_begin(PullJob *j) {
593 int r;
594
595 assert(j);
596
597 if (j->state != PULL_JOB_INIT)
598 return -EBUSY;
599
600 if (j->grow_machine_directory)
601 grow_machine_directory();
602
603 r = curl_glue_make(&j->curl, j->url, j);
604 if (r < 0)
605 return r;
606
607 if (!strv_isempty(j->old_etags)) {
608 _cleanup_free_ char *cc = NULL, *hdr = NULL;
609
610 cc = strv_join(j->old_etags, ", ");
611 if (!cc)
612 return -ENOMEM;
613
614 hdr = strappend("If-None-Match: ", cc);
615 if (!hdr)
616 return -ENOMEM;
617
618 if (!j->request_header) {
619 j->request_header = curl_slist_new(hdr, NULL);
620 if (!j->request_header)
621 return -ENOMEM;
622 } else {
623 struct curl_slist *l;
624
625 l = curl_slist_append(j->request_header, hdr);
626 if (!l)
627 return -ENOMEM;
628
629 j->request_header = l;
630 }
631 }
632
633 if (j->request_header) {
634 if (curl_easy_setopt(j->curl, CURLOPT_HTTPHEADER, j->request_header) != CURLE_OK)
635 return -EIO;
636 }
637
638 if (curl_easy_setopt(j->curl, CURLOPT_WRITEFUNCTION, pull_job_write_callback) != CURLE_OK)
639 return -EIO;
640
641 if (curl_easy_setopt(j->curl, CURLOPT_WRITEDATA, j) != CURLE_OK)
642 return -EIO;
643
644 if (curl_easy_setopt(j->curl, CURLOPT_HEADERFUNCTION, pull_job_header_callback) != CURLE_OK)
645 return -EIO;
646
647 if (curl_easy_setopt(j->curl, CURLOPT_HEADERDATA, j) != CURLE_OK)
648 return -EIO;
649
650 if (curl_easy_setopt(j->curl, CURLOPT_XFERINFOFUNCTION, pull_job_progress_callback) != CURLE_OK)
651 return -EIO;
652
653 if (curl_easy_setopt(j->curl, CURLOPT_XFERINFODATA, j) != CURLE_OK)
654 return -EIO;
655
656 if (curl_easy_setopt(j->curl, CURLOPT_NOPROGRESS, 0) != CURLE_OK)
657 return -EIO;
658
659 r = curl_glue_add(j->glue, j->curl);
660 if (r < 0)
661 return r;
662
663 j->state = PULL_JOB_ANALYZING;
664
665 return 0;
666 }