2 This file is part of systemd.
4 Copyright 2015 Lennart Poettering
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.
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.
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/>.
20 #include <sys/xattr.h>
22 #include "alloc-util.h"
24 #include "hexdecoct.h"
26 #include "machine-pool.h"
27 #include "parse-util.h"
29 #include "string-util.h"
31 #include "xattr-util.h"
32 #include "pull-common.h"
33 #include "import-util.h"
35 PullJob
* pull_job_unref(PullJob
*j
) {
39 curl_glue_remove_and_free(j
->glue
, j
->curl
);
40 curl_slist_free_all(j
->request_header
);
42 safe_close(j
->disk_fd
);
44 import_compress_free(&j
->compress
);
46 if (j
->checksum_context
)
47 gcry_md_close(j
->checksum_context
);
51 strv_free(j
->old_etags
);
58 static void pull_job_finish(PullJob
*j
, int ret
) {
61 if (j
->state
== PULL_JOB_DONE
||
62 j
->state
== PULL_JOB_FAILED
)
66 j
->state
= PULL_JOB_DONE
;
67 j
->progress_percent
= 100;
68 log_info("Download of %s complete.", j
->url
);
70 j
->state
= PULL_JOB_FAILED
;
78 static int pull_job_restart(PullJob
*j
) {
80 char *chksum_url
= NULL
;
82 r
= import_url_change_last_component(j
->url
, "SHA256SUMS", &chksum_url
);
89 j
->state
= PULL_JOB_INIT
;
92 j
->payload_allocated
= 0;
93 j
->written_compressed
= 0;
94 j
->written_uncompressed
= 0;
95 j
->written_since_last_grow
= 0;
97 r
= pull_job_begin(j
);
105 void pull_job_curl_on_finished(CurlGlue
*g
, CURL
*curl
, CURLcode result
) {
111 if (curl_easy_getinfo(curl
, CURLINFO_PRIVATE
, (char **)&j
) != CURLE_OK
)
114 if (!j
|| j
->state
== PULL_JOB_DONE
|| j
->state
== PULL_JOB_FAILED
)
117 if (result
!= CURLE_OK
) {
118 log_error("Transfer failed: %s", curl_easy_strerror(result
));
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
));
128 } else if (status
== 304) {
129 log_info("Image already downloaded. Skipping download.");
130 j
->etag_exists
= true;
133 } else if (status
>= 300) {
134 if (status
== 404 && j
->style
== VERIFICATION_PER_FILE
) {
136 /* retry pull job with SHA256SUMS file */
137 r
= pull_job_restart(j
);
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
));
149 j
->style
= VERIFICATION_PER_DIRECTORY
;
154 log_error("HTTP request to %s failed with code %li.", j
->url
, status
);
157 } else if (status
< 200) {
158 log_error("HTTP request to %s finished with unexpected code %li.", j
->url
, status
);
163 if (j
->state
!= PULL_JOB_RUNNING
) {
164 log_error("Premature connection termination.");
169 if (j
->content_length
!= (uint64_t) -1 &&
170 j
->content_length
!= j
->written_compressed
) {
171 log_error("Download truncated.");
176 if (j
->checksum_context
) {
179 k
= gcry_md_read(j
->checksum_context
, GCRY_MD_SHA256
);
181 log_error("Failed to get checksum.");
186 j
->checksum
= hexmem(k
, gcry_md_get_algo_dlen(GCRY_MD_SHA256
));
192 log_debug("SHA256 of %s is %s.", j
->url
, j
->checksum
);
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 */
199 if (ftruncate(j
->disk_fd
, j
->written_uncompressed
) < 0) {
200 r
= log_error_errno(errno
, "Failed to truncate file: %m");
205 (void) fsetxattr(j
->disk_fd
, "user.source_etag", j
->etag
, strlen(j
->etag
), 0);
207 (void) fsetxattr(j
->disk_fd
, "user.source_url", j
->url
, strlen(j
->url
), 0);
210 struct timespec ut
[2];
212 timespec_store(&ut
[0], j
->mtime
);
214 (void) futimens(j
->disk_fd
, ut
);
216 (void) fd_setcrtime(j
->disk_fd
, j
->mtime
);
223 pull_job_finish(j
, r
);
226 static int pull_job_write_uncompressed(const void *p
, size_t sz
, void *userdata
) {
227 PullJob
*j
= userdata
;
236 if (j
->written_uncompressed
+ sz
< j
->written_uncompressed
) {
237 log_error("File too large, overflow");
241 if (j
->written_uncompressed
+ sz
> j
->uncompressed_max
) {
242 log_error("File overly large, refusing");
246 if (j
->disk_fd
>= 0) {
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();
254 n
= sparse_write(j
->disk_fd
, p
, sz
, 64);
256 n
= write(j
->disk_fd
, p
, sz
);
258 return log_error_errno(errno
, "Failed to write file: %m");
259 if ((size_t) n
< sz
) {
260 log_error("Short write");
265 if (!GREEDY_REALLOC(j
->payload
, j
->payload_allocated
, j
->payload_size
+ sz
))
268 memcpy(j
->payload
+ j
->payload_size
, p
, sz
);
269 j
->payload_size
+= sz
;
272 j
->written_uncompressed
+= sz
;
273 j
->written_since_last_grow
+= sz
;
278 static int pull_job_write_compressed(PullJob
*j
, void *p
, size_t sz
) {
287 if (j
->written_compressed
+ sz
< j
->written_compressed
) {
288 log_error("File too large, overflow");
292 if (j
->written_compressed
+ sz
> j
->compressed_max
) {
293 log_error("File overly large, refusing.");
297 if (j
->content_length
!= (uint64_t) -1 &&
298 j
->written_compressed
+ sz
> j
->content_length
) {
299 log_error("Content length incorrect.");
303 if (j
->checksum_context
)
304 gcry_md_write(j
->checksum_context
, p
, sz
);
306 r
= import_uncompress(&j
->compress
, p
, sz
, pull_job_write_uncompressed
, j
);
310 j
->written_compressed
+= sz
;
315 static int pull_job_open_disk(PullJob
*j
) {
320 if (j
->on_open_disk
) {
321 r
= j
->on_open_disk(j
);
326 if (j
->disk_fd
>= 0) {
327 /* Check if we can do sparse files */
329 if (lseek(j
->disk_fd
, SEEK_SET
, 0) == 0)
330 j
->allow_sparse
= true;
333 return log_error_errno(errno
, "Failed to seek on file descriptor: %m");
335 j
->allow_sparse
= false;
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.");
349 static int pull_job_detect_compression(PullJob
*j
) {
350 _cleanup_free_
uint8_t *stub
= NULL
;
357 r
= import_uncompress_detect(&j
->compress
, j
->payload
, j
->payload_size
);
359 return log_error_errno(r
, "Failed to initialize compressor: %m");
363 log_debug("Stream is compressed: %s", import_compress_type_to_string(j
->compress
.type
));
365 r
= pull_job_open_disk(j
);
369 /* Now, take the payload we read so far, and decompress it */
371 stub_size
= j
->payload_size
;
375 j
->payload_allocated
= 0;
377 j
->state
= PULL_JOB_RUNNING
;
379 r
= pull_job_write_compressed(j
, stub
, stub_size
);
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
;
396 case PULL_JOB_ANALYZING
:
397 /* Let's first check what it actually is */
399 if (!GREEDY_REALLOC(j
->payload
, j
->payload_allocated
, j
->payload_size
+ sz
)) {
404 memcpy(j
->payload
+ j
->payload_size
, contents
, sz
);
405 j
->payload_size
+= sz
;
407 r
= pull_job_detect_compression(j
);
413 case PULL_JOB_RUNNING
:
415 r
= pull_job_write_compressed(j
, contents
, sz
);
422 case PULL_JOB_FAILED
:
427 assert_not_reached("Impossible state.");
433 pull_job_finish(j
, r
);
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
;
447 if (j
->state
== PULL_JOB_DONE
|| j
->state
== PULL_JOB_FAILED
) {
452 assert(j
->state
== PULL_JOB_ANALYZING
);
454 r
= curl_header_strdup(contents
, sz
, "ETag:", &etag
);
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);
473 r
= curl_header_strdup(contents
, sz
, "Content-Length:", &length
);
479 (void) safe_atou64(length
, &j
->content_length
);
481 if (j
->content_length
!= (uint64_t) -1) {
482 char bytes
[FORMAT_BYTES_MAX
];
484 if (j
->content_length
> j
->compressed_max
) {
485 log_error("Content too large.");
490 log_info("Downloading %s for %s.", format_bytes(bytes
, sizeof(bytes
), j
->content_length
), j
->url
);
496 r
= curl_header_strdup(contents
, sz
, "Last-Modified:", &last_modified
);
502 (void) curl_parse_http_time(last_modified
, &j
->mtime
);
507 r
= j
->on_header(j
, contents
, sz
);
515 pull_job_finish(j
, r
);
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
;
529 percent
= ((100 * dlnow
) / dltotal
);
530 n
= now(CLOCK_MONOTONIC
);
532 if (n
> j
->last_status_usec
+ USEC_PER_SEC
&&
533 percent
!= j
->progress_percent
&&
535 char buf
[FORMAT_TIMESPAN_MAX
];
537 if (n
- j
->start_usec
> USEC_PER_SEC
&& dlnow
> 0) {
538 char y
[FORMAT_BYTES_MAX
];
541 done
= n
- j
->start_usec
;
542 left
= (usec_t
) (((double) done
* (double) dltotal
) / dlnow
) - done
;
544 log_info("Got %u%% of %s. %s left at %s/s.",
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
))));
550 log_info("Got %u%% of %s.", percent
, j
->url
);
552 j
->progress_percent
= percent
;
553 j
->last_status_usec
= n
;
562 int pull_job_new(PullJob
**ret
, const char *url
, CurlGlue
*glue
, void *userdata
) {
563 _cleanup_(pull_job_unrefp
) PullJob
*j
= NULL
;
569 j
= new0(PullJob
, 1);
573 j
->state
= PULL_JOB_INIT
;
575 j
->userdata
= userdata
;
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
;
582 j
->url
= strdup(url
);
592 int pull_job_begin(PullJob
*j
) {
597 if (j
->state
!= PULL_JOB_INIT
)
600 if (j
->grow_machine_directory
)
601 grow_machine_directory();
603 r
= curl_glue_make(&j
->curl
, j
->url
, j
);
607 if (!strv_isempty(j
->old_etags
)) {
608 _cleanup_free_
char *cc
= NULL
, *hdr
= NULL
;
610 cc
= strv_join(j
->old_etags
, ", ");
614 hdr
= strappend("If-None-Match: ", cc
);
618 if (!j
->request_header
) {
619 j
->request_header
= curl_slist_new(hdr
, NULL
);
620 if (!j
->request_header
)
623 struct curl_slist
*l
;
625 l
= curl_slist_append(j
->request_header
, hdr
);
629 j
->request_header
= l
;
633 if (j
->request_header
) {
634 if (curl_easy_setopt(j
->curl
, CURLOPT_HTTPHEADER
, j
->request_header
) != CURLE_OK
)
638 if (curl_easy_setopt(j
->curl
, CURLOPT_WRITEFUNCTION
, pull_job_write_callback
) != CURLE_OK
)
641 if (curl_easy_setopt(j
->curl
, CURLOPT_WRITEDATA
, j
) != CURLE_OK
)
644 if (curl_easy_setopt(j
->curl
, CURLOPT_HEADERFUNCTION
, pull_job_header_callback
) != CURLE_OK
)
647 if (curl_easy_setopt(j
->curl
, CURLOPT_HEADERDATA
, j
) != CURLE_OK
)
650 if (curl_easy_setopt(j
->curl
, CURLOPT_XFERINFOFUNCTION
, pull_job_progress_callback
) != CURLE_OK
)
653 if (curl_easy_setopt(j
->curl
, CURLOPT_XFERINFODATA
, j
) != CURLE_OK
)
656 if (curl_easy_setopt(j
->curl
, CURLOPT_NOPROGRESS
, 0) != CURLE_OK
)
659 r
= curl_glue_add(j
->glue
, j
->curl
);
663 j
->state
= PULL_JOB_ANALYZING
;