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"
33 PullJob
* pull_job_unref(PullJob
*j
) {
37 curl_glue_remove_and_free(j
->glue
, j
->curl
);
38 curl_slist_free_all(j
->request_header
);
40 safe_close(j
->disk_fd
);
42 import_compress_free(&j
->compress
);
44 if (j
->checksum_context
)
45 gcry_md_close(j
->checksum_context
);
49 strv_free(j
->old_etags
);
56 static void pull_job_finish(PullJob
*j
, int ret
) {
59 if (j
->state
== PULL_JOB_DONE
||
60 j
->state
== PULL_JOB_FAILED
)
64 j
->state
= PULL_JOB_DONE
;
65 j
->progress_percent
= 100;
66 log_info("Download of %s complete.", j
->url
);
68 j
->state
= PULL_JOB_FAILED
;
76 void pull_job_curl_on_finished(CurlGlue
*g
, CURL
*curl
, CURLcode result
) {
82 if (curl_easy_getinfo(curl
, CURLINFO_PRIVATE
, (char **)&j
) != CURLE_OK
)
85 if (!j
|| j
->state
== PULL_JOB_DONE
|| j
->state
== PULL_JOB_FAILED
)
88 if (result
!= CURLE_OK
) {
89 log_error("Transfer failed: %s", curl_easy_strerror(result
));
94 code
= curl_easy_getinfo(curl
, CURLINFO_RESPONSE_CODE
, &status
);
95 if (code
!= CURLE_OK
) {
96 log_error("Failed to retrieve response code: %s", curl_easy_strerror(code
));
99 } else if (status
== 304) {
100 log_info("Image already downloaded. Skipping download.");
101 j
->etag_exists
= true;
104 } else if (status
>= 300) {
105 log_error("HTTP request to %s failed with code %li.", j
->url
, status
);
108 } else if (status
< 200) {
109 log_error("HTTP request to %s finished with unexpected code %li.", j
->url
, status
);
114 if (j
->state
!= PULL_JOB_RUNNING
) {
115 log_error("Premature connection termination.");
120 if (j
->content_length
!= (uint64_t) -1 &&
121 j
->content_length
!= j
->written_compressed
) {
122 log_error("Download truncated.");
127 if (j
->checksum_context
) {
130 k
= gcry_md_read(j
->checksum_context
, GCRY_MD_SHA256
);
132 log_error("Failed to get checksum.");
137 j
->checksum
= hexmem(k
, gcry_md_get_algo_dlen(GCRY_MD_SHA256
));
143 log_debug("SHA256 of %s is %s.", j
->url
, j
->checksum
);
146 if (j
->disk_fd
>= 0 && j
->allow_sparse
) {
147 /* Make sure the file size is right, in case the file was
148 * sparse and we just seeked for the last part */
150 if (ftruncate(j
->disk_fd
, j
->written_uncompressed
) < 0) {
151 r
= log_error_errno(errno
, "Failed to truncate file: %m");
156 (void) fsetxattr(j
->disk_fd
, "user.source_etag", j
->etag
, strlen(j
->etag
), 0);
158 (void) fsetxattr(j
->disk_fd
, "user.source_url", j
->url
, strlen(j
->url
), 0);
161 struct timespec ut
[2];
163 timespec_store(&ut
[0], j
->mtime
);
165 (void) futimens(j
->disk_fd
, ut
);
167 (void) fd_setcrtime(j
->disk_fd
, j
->mtime
);
174 pull_job_finish(j
, r
);
177 static int pull_job_write_uncompressed(const void *p
, size_t sz
, void *userdata
) {
178 PullJob
*j
= userdata
;
187 if (j
->written_uncompressed
+ sz
< j
->written_uncompressed
) {
188 log_error("File too large, overflow");
192 if (j
->written_uncompressed
+ sz
> j
->uncompressed_max
) {
193 log_error("File overly large, refusing");
197 if (j
->disk_fd
>= 0) {
199 if (j
->grow_machine_directory
&& j
->written_since_last_grow
>= GROW_INTERVAL_BYTES
) {
200 j
->written_since_last_grow
= 0;
201 grow_machine_directory();
205 n
= sparse_write(j
->disk_fd
, p
, sz
, 64);
207 n
= write(j
->disk_fd
, p
, sz
);
209 return log_error_errno(errno
, "Failed to write file: %m");
210 if ((size_t) n
< sz
) {
211 log_error("Short write");
216 if (!GREEDY_REALLOC(j
->payload
, j
->payload_allocated
, j
->payload_size
+ sz
))
219 memcpy(j
->payload
+ j
->payload_size
, p
, sz
);
220 j
->payload_size
+= sz
;
223 j
->written_uncompressed
+= sz
;
224 j
->written_since_last_grow
+= sz
;
229 static int pull_job_write_compressed(PullJob
*j
, void *p
, size_t sz
) {
238 if (j
->written_compressed
+ sz
< j
->written_compressed
) {
239 log_error("File too large, overflow");
243 if (j
->written_compressed
+ sz
> j
->compressed_max
) {
244 log_error("File overly large, refusing.");
248 if (j
->content_length
!= (uint64_t) -1 &&
249 j
->written_compressed
+ sz
> j
->content_length
) {
250 log_error("Content length incorrect.");
254 if (j
->checksum_context
)
255 gcry_md_write(j
->checksum_context
, p
, sz
);
257 r
= import_uncompress(&j
->compress
, p
, sz
, pull_job_write_uncompressed
, j
);
261 j
->written_compressed
+= sz
;
266 static int pull_job_open_disk(PullJob
*j
) {
271 if (j
->on_open_disk
) {
272 r
= j
->on_open_disk(j
);
277 if (j
->disk_fd
>= 0) {
278 /* Check if we can do sparse files */
280 if (lseek(j
->disk_fd
, SEEK_SET
, 0) == 0)
281 j
->allow_sparse
= true;
284 return log_error_errno(errno
, "Failed to seek on file descriptor: %m");
286 j
->allow_sparse
= false;
290 if (j
->calc_checksum
) {
291 if (gcry_md_open(&j
->checksum_context
, GCRY_MD_SHA256
, 0) != 0) {
292 log_error("Failed to initialize hash context.");
300 static int pull_job_detect_compression(PullJob
*j
) {
301 _cleanup_free_
uint8_t *stub
= NULL
;
308 r
= import_uncompress_detect(&j
->compress
, j
->payload
, j
->payload_size
);
310 return log_error_errno(r
, "Failed to initialize compressor: %m");
314 log_debug("Stream is compressed: %s", import_compress_type_to_string(j
->compress
.type
));
316 r
= pull_job_open_disk(j
);
320 /* Now, take the payload we read so far, and decompress it */
322 stub_size
= j
->payload_size
;
326 j
->payload_allocated
= 0;
328 j
->state
= PULL_JOB_RUNNING
;
330 r
= pull_job_write_compressed(j
, stub
, stub_size
);
337 static size_t pull_job_write_callback(void *contents
, size_t size
, size_t nmemb
, void *userdata
) {
338 PullJob
*j
= userdata
;
339 size_t sz
= size
* nmemb
;
347 case PULL_JOB_ANALYZING
:
348 /* Let's first check what it actually is */
350 if (!GREEDY_REALLOC(j
->payload
, j
->payload_allocated
, j
->payload_size
+ sz
)) {
355 memcpy(j
->payload
+ j
->payload_size
, contents
, sz
);
356 j
->payload_size
+= sz
;
358 r
= pull_job_detect_compression(j
);
364 case PULL_JOB_RUNNING
:
366 r
= pull_job_write_compressed(j
, contents
, sz
);
373 case PULL_JOB_FAILED
:
378 assert_not_reached("Impossible state.");
384 pull_job_finish(j
, r
);
388 static size_t pull_job_header_callback(void *contents
, size_t size
, size_t nmemb
, void *userdata
) {
389 PullJob
*j
= userdata
;
390 size_t sz
= size
* nmemb
;
391 _cleanup_free_
char *length
= NULL
, *last_modified
= NULL
;
398 if (j
->state
== PULL_JOB_DONE
|| j
->state
== PULL_JOB_FAILED
) {
403 assert(j
->state
== PULL_JOB_ANALYZING
);
405 r
= curl_header_strdup(contents
, sz
, "ETag:", &etag
);
414 if (strv_contains(j
->old_etags
, j
->etag
)) {
415 log_info("Image already downloaded. Skipping download.");
416 j
->etag_exists
= true;
417 pull_job_finish(j
, 0);
424 r
= curl_header_strdup(contents
, sz
, "Content-Length:", &length
);
430 (void) safe_atou64(length
, &j
->content_length
);
432 if (j
->content_length
!= (uint64_t) -1) {
433 char bytes
[FORMAT_BYTES_MAX
];
435 if (j
->content_length
> j
->compressed_max
) {
436 log_error("Content too large.");
441 log_info("Downloading %s for %s.", format_bytes(bytes
, sizeof(bytes
), j
->content_length
), j
->url
);
447 r
= curl_header_strdup(contents
, sz
, "Last-Modified:", &last_modified
);
453 (void) curl_parse_http_time(last_modified
, &j
->mtime
);
458 r
= j
->on_header(j
, contents
, sz
);
466 pull_job_finish(j
, r
);
470 static int pull_job_progress_callback(void *userdata
, curl_off_t dltotal
, curl_off_t dlnow
, curl_off_t ultotal
, curl_off_t ulnow
) {
471 PullJob
*j
= userdata
;
480 percent
= ((100 * dlnow
) / dltotal
);
481 n
= now(CLOCK_MONOTONIC
);
483 if (n
> j
->last_status_usec
+ USEC_PER_SEC
&&
484 percent
!= j
->progress_percent
&&
486 char buf
[FORMAT_TIMESPAN_MAX
];
488 if (n
- j
->start_usec
> USEC_PER_SEC
&& dlnow
> 0) {
489 char y
[FORMAT_BYTES_MAX
];
492 done
= n
- j
->start_usec
;
493 left
= (usec_t
) (((double) done
* (double) dltotal
) / dlnow
) - done
;
495 log_info("Got %u%% of %s. %s left at %s/s.",
498 format_timespan(buf
, sizeof(buf
), left
, USEC_PER_SEC
),
499 format_bytes(y
, sizeof(y
), (uint64_t) ((double) dlnow
/ ((double) done
/ (double) USEC_PER_SEC
))));
501 log_info("Got %u%% of %s.", percent
, j
->url
);
503 j
->progress_percent
= percent
;
504 j
->last_status_usec
= n
;
513 int pull_job_new(PullJob
**ret
, const char *url
, CurlGlue
*glue
, void *userdata
) {
514 _cleanup_(pull_job_unrefp
) PullJob
*j
= NULL
;
520 j
= new0(PullJob
, 1);
524 j
->state
= PULL_JOB_INIT
;
526 j
->userdata
= userdata
;
528 j
->content_length
= (uint64_t) -1;
529 j
->start_usec
= now(CLOCK_MONOTONIC
);
530 j
->compressed_max
= j
->uncompressed_max
= 64LLU * 1024LLU * 1024LLU * 1024LLU; /* 64GB safety limit */
532 j
->url
= strdup(url
);
542 int pull_job_begin(PullJob
*j
) {
547 if (j
->state
!= PULL_JOB_INIT
)
550 if (j
->grow_machine_directory
)
551 grow_machine_directory();
553 r
= curl_glue_make(&j
->curl
, j
->url
, j
);
557 if (!strv_isempty(j
->old_etags
)) {
558 _cleanup_free_
char *cc
= NULL
, *hdr
= NULL
;
560 cc
= strv_join(j
->old_etags
, ", ");
564 hdr
= strappend("If-None-Match: ", cc
);
568 if (!j
->request_header
) {
569 j
->request_header
= curl_slist_new(hdr
, NULL
);
570 if (!j
->request_header
)
573 struct curl_slist
*l
;
575 l
= curl_slist_append(j
->request_header
, hdr
);
579 j
->request_header
= l
;
583 if (j
->request_header
) {
584 if (curl_easy_setopt(j
->curl
, CURLOPT_HTTPHEADER
, j
->request_header
) != CURLE_OK
)
588 if (curl_easy_setopt(j
->curl
, CURLOPT_WRITEFUNCTION
, pull_job_write_callback
) != CURLE_OK
)
591 if (curl_easy_setopt(j
->curl
, CURLOPT_WRITEDATA
, j
) != CURLE_OK
)
594 if (curl_easy_setopt(j
->curl
, CURLOPT_HEADERFUNCTION
, pull_job_header_callback
) != CURLE_OK
)
597 if (curl_easy_setopt(j
->curl
, CURLOPT_HEADERDATA
, j
) != CURLE_OK
)
600 if (curl_easy_setopt(j
->curl
, CURLOPT_XFERINFOFUNCTION
, pull_job_progress_callback
) != CURLE_OK
)
603 if (curl_easy_setopt(j
->curl
, CURLOPT_XFERINFODATA
, j
) != CURLE_OK
)
606 if (curl_easy_setopt(j
->curl
, CURLOPT_NOPROGRESS
, 0) != CURLE_OK
)
609 r
= curl_glue_add(j
->glue
, j
->curl
);
613 j
->state
= PULL_JOB_ANALYZING
;