1 /* SPDX-License-Identifier: LGPL-2.1+ */
7 #include "alloc-util.h"
9 #include "gcrypt-util.h"
10 #include "hexdecoct.h"
11 #include "import-util.h"
13 #include "machine-pool.h"
14 #include "parse-util.h"
15 #include "pull-common.h"
17 #include "string-util.h"
19 #include "xattr-util.h"
21 PullJob
* pull_job_unref(PullJob
*j
) {
25 curl_glue_remove_and_free(j
->glue
, j
->curl
);
26 curl_slist_free_all(j
->request_header
);
28 safe_close(j
->disk_fd
);
30 import_compress_free(&j
->compress
);
32 if (j
->checksum_context
)
33 gcry_md_close(j
->checksum_context
);
37 strv_free(j
->old_etags
);
44 static void pull_job_finish(PullJob
*j
, int ret
) {
47 if (IN_SET(j
->state
, PULL_JOB_DONE
, PULL_JOB_FAILED
))
51 j
->state
= PULL_JOB_DONE
;
52 j
->progress_percent
= 100;
53 log_info("Download of %s complete.", j
->url
);
55 j
->state
= PULL_JOB_FAILED
;
63 static int pull_job_restart(PullJob
*j
) {
65 char *chksum_url
= NULL
;
67 r
= import_url_change_last_component(j
->url
, "SHA256SUMS", &chksum_url
);
73 j
->state
= PULL_JOB_INIT
;
74 j
->payload
= mfree(j
->payload
);
76 j
->payload_allocated
= 0;
77 j
->written_compressed
= 0;
78 j
->written_uncompressed
= 0;
80 r
= pull_job_begin(j
);
87 void pull_job_curl_on_finished(CurlGlue
*g
, CURL
*curl
, CURLcode result
) {
93 if (curl_easy_getinfo(curl
, CURLINFO_PRIVATE
, (char **)&j
) != CURLE_OK
)
96 if (!j
|| IN_SET(j
->state
, PULL_JOB_DONE
, PULL_JOB_FAILED
))
99 if (result
!= CURLE_OK
) {
100 log_error("Transfer failed: %s", curl_easy_strerror(result
));
105 code
= curl_easy_getinfo(curl
, CURLINFO_RESPONSE_CODE
, &status
);
106 if (code
!= CURLE_OK
) {
107 log_error("Failed to retrieve response code: %s", curl_easy_strerror(code
));
110 } else if (status
== 304) {
111 log_info("Image already downloaded. Skipping download.");
112 j
->etag_exists
= true;
115 } else if (status
>= 300) {
116 if (status
== 404 && j
->style
== VERIFICATION_PER_FILE
) {
118 /* retry pull job with SHA256SUMS file */
119 r
= pull_job_restart(j
);
123 code
= curl_easy_getinfo(j
->curl
, CURLINFO_RESPONSE_CODE
, &status
);
124 if (code
!= CURLE_OK
) {
125 log_error("Failed to retrieve response code: %s", curl_easy_strerror(code
));
131 j
->style
= VERIFICATION_PER_DIRECTORY
;
136 log_error("HTTP request to %s failed with code %li.", j
->url
, status
);
139 } else if (status
< 200) {
140 log_error("HTTP request to %s finished with unexpected code %li.", j
->url
, status
);
145 if (j
->state
!= PULL_JOB_RUNNING
) {
146 log_error("Premature connection termination.");
151 if (j
->content_length
!= (uint64_t) -1 &&
152 j
->content_length
!= j
->written_compressed
) {
153 log_error("Download truncated.");
158 if (j
->checksum_context
) {
161 k
= gcry_md_read(j
->checksum_context
, GCRY_MD_SHA256
);
163 log_error("Failed to get checksum.");
168 j
->checksum
= hexmem(k
, gcry_md_get_algo_dlen(GCRY_MD_SHA256
));
174 log_debug("SHA256 of %s is %s.", j
->url
, j
->checksum
);
177 if (j
->disk_fd
>= 0 && j
->allow_sparse
) {
178 /* Make sure the file size is right, in case the file was
179 * sparse and we just seeked for the last part */
181 if (ftruncate(j
->disk_fd
, j
->written_uncompressed
) < 0) {
182 r
= log_error_errno(errno
, "Failed to truncate file: %m");
187 (void) fsetxattr(j
->disk_fd
, "user.source_etag", j
->etag
, strlen(j
->etag
), 0);
189 (void) fsetxattr(j
->disk_fd
, "user.source_url", j
->url
, strlen(j
->url
), 0);
192 struct timespec ut
[2];
194 timespec_store(&ut
[0], j
->mtime
);
196 (void) futimens(j
->disk_fd
, ut
);
198 (void) fd_setcrtime(j
->disk_fd
, j
->mtime
);
205 pull_job_finish(j
, r
);
208 static int pull_job_write_uncompressed(const void *p
, size_t sz
, void *userdata
) {
209 PullJob
*j
= userdata
;
218 if (j
->written_uncompressed
+ sz
< j
->written_uncompressed
)
219 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW
),
220 "File too large, overflow");
222 if (j
->written_uncompressed
+ sz
> j
->uncompressed_max
)
223 return log_error_errno(SYNTHETIC_ERRNO(EFBIG
),
224 "File overly large, refusing");
226 if (j
->disk_fd
>= 0) {
229 n
= sparse_write(j
->disk_fd
, p
, sz
, 64);
231 n
= write(j
->disk_fd
, p
, sz
);
236 return log_error_errno((int) n
, "Failed to write file: %m");
238 return log_error_errno(SYNTHETIC_ERRNO(EIO
), "Short write");
241 if (!GREEDY_REALLOC(j
->payload
, j
->payload_allocated
, j
->payload_size
+ sz
))
244 memcpy(j
->payload
+ j
->payload_size
, p
, sz
);
245 j
->payload_size
+= sz
;
248 j
->written_uncompressed
+= sz
;
253 static int pull_job_write_compressed(PullJob
*j
, void *p
, size_t sz
) {
262 if (j
->written_compressed
+ sz
< j
->written_compressed
)
263 return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW
), "File too large, overflow");
265 if (j
->written_compressed
+ sz
> j
->compressed_max
)
266 return log_error_errno(SYNTHETIC_ERRNO(EFBIG
), "File overly large, refusing.");
268 if (j
->content_length
!= (uint64_t) -1 &&
269 j
->written_compressed
+ sz
> j
->content_length
)
270 return log_error_errno(SYNTHETIC_ERRNO(EFBIG
),
271 "Content length incorrect.");
273 if (j
->checksum_context
)
274 gcry_md_write(j
->checksum_context
, p
, sz
);
276 r
= import_uncompress(&j
->compress
, p
, sz
, pull_job_write_uncompressed
, j
);
280 j
->written_compressed
+= sz
;
285 static int pull_job_open_disk(PullJob
*j
) {
290 if (j
->on_open_disk
) {
291 r
= j
->on_open_disk(j
);
296 if (j
->disk_fd
>= 0) {
297 /* Check if we can do sparse files */
299 if (lseek(j
->disk_fd
, SEEK_SET
, 0) == 0)
300 j
->allow_sparse
= true;
303 return log_error_errno(errno
, "Failed to seek on file descriptor: %m");
305 j
->allow_sparse
= false;
309 if (j
->calc_checksum
) {
310 initialize_libgcrypt(false);
312 if (gcry_md_open(&j
->checksum_context
, GCRY_MD_SHA256
, 0) != 0)
313 return log_error_errno(SYNTHETIC_ERRNO(EIO
),
314 "Failed to initialize hash context.");
320 static int pull_job_detect_compression(PullJob
*j
) {
321 _cleanup_free_
uint8_t *stub
= NULL
;
328 r
= import_uncompress_detect(&j
->compress
, j
->payload
, j
->payload_size
);
330 return log_error_errno(r
, "Failed to initialize compressor: %m");
334 log_debug("Stream is compressed: %s", import_compress_type_to_string(j
->compress
.type
));
336 r
= pull_job_open_disk(j
);
340 /* Now, take the payload we read so far, and decompress it */
342 stub_size
= j
->payload_size
;
346 j
->payload_allocated
= 0;
348 j
->state
= PULL_JOB_RUNNING
;
350 r
= pull_job_write_compressed(j
, stub
, stub_size
);
357 static size_t pull_job_write_callback(void *contents
, size_t size
, size_t nmemb
, void *userdata
) {
358 PullJob
*j
= userdata
;
359 size_t sz
= size
* nmemb
;
367 case PULL_JOB_ANALYZING
:
368 /* Let's first check what it actually is */
370 if (!GREEDY_REALLOC(j
->payload
, j
->payload_allocated
, j
->payload_size
+ sz
)) {
375 memcpy(j
->payload
+ j
->payload_size
, contents
, sz
);
376 j
->payload_size
+= sz
;
378 r
= pull_job_detect_compression(j
);
384 case PULL_JOB_RUNNING
:
386 r
= pull_job_write_compressed(j
, contents
, sz
);
393 case PULL_JOB_FAILED
:
398 assert_not_reached("Impossible state.");
404 pull_job_finish(j
, r
);
408 static size_t pull_job_header_callback(void *contents
, size_t size
, size_t nmemb
, void *userdata
) {
409 PullJob
*j
= userdata
;
410 size_t sz
= size
* nmemb
;
411 _cleanup_free_
char *length
= NULL
, *last_modified
= NULL
;
418 if (IN_SET(j
->state
, PULL_JOB_DONE
, PULL_JOB_FAILED
)) {
423 assert(j
->state
== PULL_JOB_ANALYZING
);
425 r
= curl_header_strdup(contents
, sz
, "ETag:", &etag
);
434 if (strv_contains(j
->old_etags
, j
->etag
)) {
435 log_info("Image already downloaded. Skipping download.");
436 j
->etag_exists
= true;
437 pull_job_finish(j
, 0);
444 r
= curl_header_strdup(contents
, sz
, "Content-Length:", &length
);
450 (void) safe_atou64(length
, &j
->content_length
);
452 if (j
->content_length
!= (uint64_t) -1) {
453 char bytes
[FORMAT_BYTES_MAX
];
455 if (j
->content_length
> j
->compressed_max
) {
456 log_error("Content too large.");
461 log_info("Downloading %s for %s.", format_bytes(bytes
, sizeof(bytes
), j
->content_length
), j
->url
);
467 r
= curl_header_strdup(contents
, sz
, "Last-Modified:", &last_modified
);
473 (void) curl_parse_http_time(last_modified
, &j
->mtime
);
478 r
= j
->on_header(j
, contents
, sz
);
486 pull_job_finish(j
, r
);
490 static int pull_job_progress_callback(void *userdata
, curl_off_t dltotal
, curl_off_t dlnow
, curl_off_t ultotal
, curl_off_t ulnow
) {
491 PullJob
*j
= userdata
;
500 percent
= ((100 * dlnow
) / dltotal
);
501 n
= now(CLOCK_MONOTONIC
);
503 if (n
> j
->last_status_usec
+ USEC_PER_SEC
&&
504 percent
!= j
->progress_percent
&&
506 char buf
[FORMAT_TIMESPAN_MAX
];
508 if (n
- j
->start_usec
> USEC_PER_SEC
&& dlnow
> 0) {
509 char y
[FORMAT_BYTES_MAX
];
512 done
= n
- j
->start_usec
;
513 left
= (usec_t
) (((double) done
* (double) dltotal
) / dlnow
) - done
;
515 log_info("Got %u%% of %s. %s left at %s/s.",
518 format_timespan(buf
, sizeof(buf
), left
, USEC_PER_SEC
),
519 format_bytes(y
, sizeof(y
), (uint64_t) ((double) dlnow
/ ((double) done
/ (double) USEC_PER_SEC
))));
521 log_info("Got %u%% of %s.", percent
, j
->url
);
523 j
->progress_percent
= percent
;
524 j
->last_status_usec
= n
;
533 int pull_job_new(PullJob
**ret
, const char *url
, CurlGlue
*glue
, void *userdata
) {
534 _cleanup_(pull_job_unrefp
) PullJob
*j
= NULL
;
535 _cleanup_free_
char *u
= NULL
;
550 .state
= PULL_JOB_INIT
,
552 .userdata
= userdata
,
554 .content_length
= (uint64_t) -1,
555 .start_usec
= now(CLOCK_MONOTONIC
),
556 .compressed_max
= 64LLU * 1024LLU * 1024LLU * 1024LLU, /* 64GB safety limit */
557 .uncompressed_max
= 64LLU * 1024LLU * 1024LLU * 1024LLU, /* 64GB safety limit */
558 .style
= VERIFICATION_STYLE_UNSET
,
567 int pull_job_begin(PullJob
*j
) {
572 if (j
->state
!= PULL_JOB_INIT
)
575 r
= curl_glue_make(&j
->curl
, j
->url
, j
);
579 if (!strv_isempty(j
->old_etags
)) {
580 _cleanup_free_
char *cc
= NULL
, *hdr
= NULL
;
582 cc
= strv_join(j
->old_etags
, ", ");
586 hdr
= strappend("If-None-Match: ", cc
);
590 if (!j
->request_header
) {
591 j
->request_header
= curl_slist_new(hdr
, NULL
);
592 if (!j
->request_header
)
595 struct curl_slist
*l
;
597 l
= curl_slist_append(j
->request_header
, hdr
);
601 j
->request_header
= l
;
605 if (j
->request_header
) {
606 if (curl_easy_setopt(j
->curl
, CURLOPT_HTTPHEADER
, j
->request_header
) != CURLE_OK
)
610 if (curl_easy_setopt(j
->curl
, CURLOPT_WRITEFUNCTION
, pull_job_write_callback
) != CURLE_OK
)
613 if (curl_easy_setopt(j
->curl
, CURLOPT_WRITEDATA
, j
) != CURLE_OK
)
616 if (curl_easy_setopt(j
->curl
, CURLOPT_HEADERFUNCTION
, pull_job_header_callback
) != CURLE_OK
)
619 if (curl_easy_setopt(j
->curl
, CURLOPT_HEADERDATA
, j
) != CURLE_OK
)
622 if (curl_easy_setopt(j
->curl
, CURLOPT_XFERINFOFUNCTION
, pull_job_progress_callback
) != CURLE_OK
)
625 if (curl_easy_setopt(j
->curl
, CURLOPT_XFERINFODATA
, j
) != CURLE_OK
)
628 if (curl_easy_setopt(j
->curl
, CURLOPT_NOPROGRESS
, 0) != CURLE_OK
)
631 r
= curl_glue_add(j
->glue
, j
->curl
);
635 j
->state
= PULL_JOB_ANALYZING
;