libzypp 17.31.7
MediaMultiCurl.cc
Go to the documentation of this file.
1/*---------------------------------------------------------------------\
2| ____ _ __ __ ___ |
3| |__ / \ / / . \ . \ |
4| / / \ V /| _/ _/ |
5| / /__ | | | | | | |
6| /_____||_| |_| |_| |
7| |
8\---------------------------------------------------------------------*/
13#include <ctype.h>
14#include <sys/types.h>
15#include <signal.h>
16#include <sys/wait.h>
17#include <netdb.h>
18#include <arpa/inet.h>
19
20#include <vector>
21#include <iostream>
22#include <algorithm>
23
24
25#include <zypp/ZConfig.h>
26#include <zypp/base/Logger.h>
28#include <zypp-curl/parser/MetaLinkParser>
29#include <zypp/ManagedFile.h>
31#include <zypp-curl/auth/CurlAuthData>
32
33using std::endl;
34using namespace zypp::base;
35
36#undef CURLVERSION_AT_LEAST
37#define CURLVERSION_AT_LEAST(M,N,O) LIBCURL_VERSION_NUM >= ((((M)<<8)+(N))<<8)+(O)
38
39namespace zypp {
40 namespace media {
41
42
44
45
46class multifetchrequest;
47
48// Hack: we derive from MediaCurl just to get the storage space for
49// settings, url, curlerrors and the like
50
52 friend class multifetchrequest;
53
54public:
55 multifetchworker(int no, multifetchrequest &request, const Url &url);
57 void nextjob();
58 void run();
59 bool checkChecksum();
60 bool recheckChecksum();
61 void disableCompetition();
62
63 void checkdns();
64 void adddnsfd(fd_set &rset, int &maxfd);
65 void dnsevent(fd_set &rset);
66
68
69 int _state;
71
72 size_t _blkno;
73 off_t _blkstart;
74 size_t _blksize;
76
79 off_t _received;
80
81 double _avgspeed;
82 double _maxspeed;
83
85
86private:
87 void stealjob();
88
89 size_t writefunction(void *ptr, size_t size);
90 static size_t _writefunction(void *ptr, size_t size, size_t nmemb, void *stream);
91
92 size_t headerfunction(char *ptr, size_t size);
93 static size_t _headerfunction(void *ptr, size_t size, size_t nmemb, void *stream);
94
96 int _pass;
97 std::string _urlbuf;
98 off_t _off;
99 size_t _size;
101
102 pid_t _pid;
104};
105
106#define WORKER_STARTING 0
107#define WORKER_LOOKUP 1
108#define WORKER_FETCH 2
109#define WORKER_DISCARD 3
110#define WORKER_DONE 4
111#define WORKER_SLEEP 5
112#define WORKER_BROKEN 6
113
114
115
117public:
118 multifetchrequest(const MediaMultiCurl *context, const Pathname &filename, const Url &baseurl, CURLM *multi, FILE *fp, callback::SendReport<DownloadProgressReport> *report, MediaBlockList *blklist, off_t filesize);
120
121 void run(std::vector<Url> &urllist);
122
123protected:
124
125 static size_t makeBlksize ( size_t filesize );
126
127 friend class multifetchworker;
128
132
133 FILE *_fp;
137
138 CURLM *_multi;
139
140 std::list<multifetchworker *> _workers;
143
144 size_t _blkno;
145 size_t _defaultBlksize = 0; //< The blocksize to use if the metalink file does not specify one
146 off_t _blkoff;
155
158
162
163public:
164 double _timeout;
166 double _maxspeed;
168};
169
170constexpr auto MIN_REQ_MIRRS = 4;
171constexpr auto MAXURLS = 10;
172
174
175static double
177{
178 struct timeval tv;
179 if (gettimeofday(&tv, NULL))
180 return 0;
181 return tv.tv_sec + tv.tv_usec / 1000000.;
182}
183
184size_t
185multifetchworker::writefunction(void *ptr, size_t size)
186{
187 size_t len, cnt;
188 if (_state == WORKER_BROKEN)
189 return size ? 0 : 1;
190
191 double now = currentTime();
192
193 len = size > _size ? _size : size;
194 if (!len)
195 {
196 // kill this job?
197 return size;
198 }
199
200 if (_blkstart && _off == _blkstart)
201 {
202 // make sure that the server replied with "partial content"
203 // for http requests
204 char *effurl;
205 (void)curl_easy_getinfo(_curl, CURLINFO_EFFECTIVE_URL, &effurl);
206 if (effurl && !strncasecmp(effurl, "http", 4))
207 {
208 long statuscode = 0;
209 (void)curl_easy_getinfo(_curl, CURLINFO_RESPONSE_CODE, &statuscode);
210 if (statuscode != 206)
211 return size ? 0 : 1;
212 }
213 }
214
215 _blkreceived += len;
216 _received += len;
217
218 _request->_lastprogress = now;
219
220 if (_state == WORKER_DISCARD || !_request->_fp)
221 {
222 // block is no longer needed
223 // still calculate the checksum so that we can throw out bad servers
224 if (_request->_blklist)
225 _dig.update((const char *)ptr, len);
226 _off += len;
227 _size -= len;
228 return size;
229 }
230 if (fseeko(_request->_fp, _off, SEEK_SET))
231 return size ? 0 : 1;
232 cnt = fwrite(ptr, 1, len, _request->_fp);
233 if (cnt > 0)
234 {
235 _request->_fetchedsize += cnt;
236 if (_request->_blklist)
237 _dig.update((const char *)ptr, cnt);
238 _off += cnt;
239 _size -= cnt;
240 if (cnt == len)
241 return size;
242 }
243 return cnt;
244}
245
246size_t
247multifetchworker::_writefunction(void *ptr, size_t size, size_t nmemb, void *stream)
248{
249 multifetchworker *me = reinterpret_cast<multifetchworker *>(stream);
250 return me->writefunction(ptr, size * nmemb);
251}
252
253size_t
255{
256 size_t l = size;
257 if (l > 9 && !strncasecmp(p, "Location:", 9))
258 {
259 std::string line(p + 9, l - 9);
260 if (line[l - 10] == '\r')
261 line.erase(l - 10, 1);
262 XXX << "#" << _workerno << ": redirecting to" << line << endl;
263 return size;
264 }
265 if (l <= 14 || l >= 128 || strncasecmp(p, "Content-Range:", 14) != 0)
266 return size;
267 p += 14;
268 l -= 14;
269 while (l && (*p == ' ' || *p == '\t'))
270 p++, l--;
271 if (l < 6 || strncasecmp(p, "bytes", 5))
272 return size;
273 p += 5;
274 l -= 5;
275 char buf[128];
276 memcpy(buf, p, l);
277 buf[l] = 0;
278 unsigned long long start, off, filesize;
279 if (sscanf(buf, "%llu-%llu/%llu", &start, &off, &filesize) != 3)
280 return size;
281 if (_request->_filesize == (off_t)-1)
282 {
283 WAR << "#" << _workerno << ": setting request filesize to " << filesize << endl;
284 _request->_filesize = filesize;
285 if (_request->_totalsize == 0 && !_request->_blklist)
286 _request->_totalsize = filesize;
287 }
288 if (_request->_filesize != (off_t)filesize)
289 {
290 XXX << "#" << _workerno << ": filesize mismatch" << endl;
292 strncpy(_curlError, "filesize mismatch", CURL_ERROR_SIZE);
293 }
294 return size;
295}
296
297size_t
298multifetchworker::_headerfunction(void *ptr, size_t size, size_t nmemb, void *stream)
299{
300 multifetchworker *me = reinterpret_cast<multifetchworker *>(stream);
301 return me->headerfunction((char *)ptr, size * nmemb);
302}
303
305: MediaCurl(url, Pathname())
306{
307 _workerno = no;
308 _request = &request;
310 _competing = false;
311 _off = _blkstart = 0;
312 _size = _blksize = 0;
313 _pass = 0;
314 _blkno = 0;
315 _pid = 0;
316 _dnspipe = -1;
317 _blkreceived = 0;
318 _received = 0;
319 _blkstarttime = 0;
320 _avgspeed = 0;
321 _sleepuntil = 0;
323 _noendrange = false;
324
325 Url curlUrl( clearQueryString(url) );
326 _urlbuf = curlUrl.asString();
328 if (_curl)
329 XXX << "reused worker from pool" << endl;
330 if (!_curl && !(_curl = curl_easy_init()))
331 {
333 strncpy(_curlError, "curl_easy_init failed", CURL_ERROR_SIZE);
334 return;
335 }
336 try
337 {
338 setupEasy();
339 }
340 catch (Exception &ex)
341 {
342 curl_easy_cleanup(_curl);
343 _curl = 0;
345 strncpy(_curlError, "curl_easy_setopt failed", CURL_ERROR_SIZE);
346 return;
347 }
348 curl_easy_setopt(_curl, CURLOPT_PRIVATE, this);
349 curl_easy_setopt(_curl, CURLOPT_URL, _urlbuf.c_str());
350 curl_easy_setopt(_curl, CURLOPT_WRITEFUNCTION, &_writefunction);
351 curl_easy_setopt(_curl, CURLOPT_WRITEDATA, this);
352 if (_request->_filesize == off_t(-1) || !_request->_blklist || !_request->_blklist->haveChecksum(0))
353 {
354 curl_easy_setopt(_curl, CURLOPT_HEADERFUNCTION, &_headerfunction);
355 curl_easy_setopt(_curl, CURLOPT_HEADERDATA, this);
356 }
357 // if this is the same host copy authorization
358 // (the host check is also what curl does when doing a redirect)
359 // (note also that unauthorized exceptions are thrown with the request host)
361 {
365 if ( _settings.userPassword().size() )
366 {
367 curl_easy_setopt(_curl, CURLOPT_USERPWD, _settings.userPassword().c_str());
368 std::string use_auth = _settings.authType();
369 if (use_auth.empty())
370 use_auth = "digest,basic"; // our default
371 long auth = CurlAuthData::auth_type_str2long(use_auth);
372 if( auth != CURLAUTH_NONE)
373 {
374 XXX << "#" << _workerno << ": Enabling HTTP authentication methods: " << use_auth
375 << " (CURLOPT_HTTPAUTH=" << auth << ")" << std::endl;
376 curl_easy_setopt(_curl, CURLOPT_HTTPAUTH, auth);
377 }
378 }
379 }
380 checkdns();
381}
382
384{
385 if (_curl)
386 {
388 curl_multi_remove_handle(_request->_multi, _curl);
390 {
391#if CURLVERSION_AT_LEAST(7,15,5)
392 curl_easy_setopt(_curl, CURLOPT_MAX_RECV_SPEED_LARGE, (curl_off_t)0);
393#endif
394 curl_easy_setopt(_curl, CURLOPT_PRIVATE, (void *)0);
395 curl_easy_setopt(_curl, CURLOPT_WRITEFUNCTION, (void *)0);
396 curl_easy_setopt(_curl, CURLOPT_WRITEDATA, (void *)0);
397 curl_easy_setopt(_curl, CURLOPT_HEADERFUNCTION, (void *)0);
398 curl_easy_setopt(_curl, CURLOPT_HEADERDATA, (void *)0);
400 }
401 else
402 curl_easy_cleanup(_curl);
403 _curl = 0;
404 }
405 if (_pid)
406 {
407 kill(_pid, SIGKILL);
408 int status;
409 while (waitpid(_pid, &status, 0) == -1)
410 if (errno != EINTR)
411 break;
412 _pid = 0;
413 }
414 if (_dnspipe != -1)
415 {
416 close(_dnspipe);
417 _dnspipe = -1;
418 }
419 // the destructor in MediaCurl doesn't call disconnect() if
420 // the media is not attached, so we do it here manually
422}
423
424static inline bool env_isset(std::string name)
425{
426 const char *s = getenv(name.c_str());
427 return s && *s ? true : false;
428}
429
430void
432{
433 std::string host = _url.getHost();
434
435 if (host.empty())
436 return;
437
438 if (_request->_context->isDNSok(host))
439 return;
440
441 // no need to do dns checking for numeric hosts
442 char addrbuf[128];
443 if (inet_pton(AF_INET, host.c_str(), addrbuf) == 1)
444 return;
445 if (inet_pton(AF_INET6, host.c_str(), addrbuf) == 1)
446 return;
447
448 // no need to do dns checking if we use a proxy
449 if (!_settings.proxy().empty())
450 return;
451 if (env_isset("all_proxy") || env_isset("ALL_PROXY"))
452 return;
453 std::string schemeproxy = _url.getScheme() + "_proxy";
454 if (env_isset(schemeproxy))
455 return;
456 if (schemeproxy != "http_proxy")
457 {
458 std::transform(schemeproxy.begin(), schemeproxy.end(), schemeproxy.begin(), ::toupper);
459 if (env_isset(schemeproxy))
460 return;
461 }
462
463 XXX << "checking DNS lookup of " << host << endl;
464 int pipefds[2];
465 if (pipe(pipefds))
466 {
468 strncpy(_curlError, "DNS pipe creation failed", CURL_ERROR_SIZE);
469 return;
470 }
471 _pid = fork();
472 if (_pid == pid_t(-1))
473 {
474 close(pipefds[0]);
475 close(pipefds[1]);
476 _pid = 0;
478 strncpy(_curlError, "DNS checker fork failed", CURL_ERROR_SIZE);
479 return;
480 }
481 else if (_pid == 0)
482 {
483 close(pipefds[0]);
484 // XXX: close all other file descriptors
485 struct addrinfo *ai, aihints;
486 memset(&aihints, 0, sizeof(aihints));
487 aihints.ai_family = PF_UNSPEC;
488 int tstsock = socket(PF_INET6, SOCK_DGRAM | SOCK_CLOEXEC, 0);
489 if (tstsock == -1)
490 aihints.ai_family = PF_INET;
491 else
492 close(tstsock);
493 aihints.ai_socktype = SOCK_STREAM;
494 aihints.ai_flags = AI_CANONNAME;
495 unsigned int connecttimeout = _request->_connect_timeout;
496 if (connecttimeout)
497 alarm(connecttimeout);
498 signal(SIGALRM, SIG_DFL);
499 if (getaddrinfo(host.c_str(), NULL, &aihints, &ai))
500 _exit(1);
501 _exit(0);
502 }
503 close(pipefds[1]);
504 _dnspipe = pipefds[0];
506}
507
508void
509multifetchworker::adddnsfd(fd_set &rset, int &maxfd)
510{
511 if (_state != WORKER_LOOKUP)
512 return;
513 FD_SET(_dnspipe, &rset);
514 if (maxfd < _dnspipe)
515 maxfd = _dnspipe;
516}
517
518void
520{
521
522 if (_state != WORKER_LOOKUP || !FD_ISSET(_dnspipe, &rset))
523 return;
524 int status;
525 while (waitpid(_pid, &status, 0) == -1)
526 {
527 if (errno != EINTR)
528 return;
529 }
530 _pid = 0;
531 if (_dnspipe != -1)
532 {
533 close(_dnspipe);
534 _dnspipe = -1;
535 }
536 if (!WIFEXITED(status))
537 {
539 strncpy(_curlError, "DNS lookup failed", CURL_ERROR_SIZE);
541 return;
542 }
543 int exitcode = WEXITSTATUS(status);
544 XXX << "#" << _workerno << ": DNS lookup returned " << exitcode << endl;
545 if (exitcode != 0)
546 {
548 strncpy(_curlError, "DNS lookup failed", CURL_ERROR_SIZE);
550 return;
551 }
553 nextjob();
554}
555
556bool
558{
559 // XXX << "checkChecksum block " << _blkno << endl;
560 if (!_blksize || !_request->_blklist)
561 return true;
563}
564
565bool
567{
568 // XXX << "recheckChecksum block " << _blkno << endl;
569 if (!_request->_fp || !_blksize || !_request->_blklist)
570 return true;
571 if (fseeko(_request->_fp, _blkstart, SEEK_SET))
572 return false;
573 char buf[4096];
574 size_t l = _blksize;
575 _request->_blklist->createDigest(_dig); // resets digest
576 while (l)
577 {
578 size_t cnt = l > sizeof(buf) ? sizeof(buf) : l;
579 if (fread(buf, cnt, 1, _request->_fp) != 1)
580 return false;
581 _dig.update(buf, cnt);
582 l -= cnt;
583 }
585}
586
587
588void
590{
591 if (!_request->_stealing)
592 {
593 XXX << "start stealing!" << endl;
594 _request->_stealing = true;
595 }
596 multifetchworker *best = 0;
597 std::list<multifetchworker *>::iterator workeriter = _request->_workers.begin();
598 double now = 0;
599 for (; workeriter != _request->_workers.end(); ++workeriter)
600 {
601 multifetchworker *worker = *workeriter;
602 if (worker == this)
603 continue;
604 if (worker->_pass == -1)
605 continue; // do not steal!
606 if (worker->_state == WORKER_DISCARD || worker->_state == WORKER_DONE || worker->_state == WORKER_SLEEP || !worker->_blksize)
607 continue; // do not steal finished jobs
608 if (!worker->_avgspeed && worker->_blkreceived)
609 {
610 if (!now)
611 now = currentTime();
612 if (now > worker->_blkstarttime)
613 worker->_avgspeed = worker->_blkreceived / (now - worker->_blkstarttime);
614 }
615 if (!best || best->_pass > worker->_pass)
616 {
617 best = worker;
618 continue;
619 }
620 if (best->_pass < worker->_pass)
621 continue;
622 // if it is the same block, we want to know the best worker, otherwise the worst
623 if (worker->_blkstart == best->_blkstart)
624 {
625 if ((worker->_blksize - worker->_blkreceived) * best->_avgspeed < (best->_blksize - best->_blkreceived) * worker->_avgspeed)
626 best = worker;
627 }
628 else
629 {
630 if ((worker->_blksize - worker->_blkreceived) * best->_avgspeed > (best->_blksize - best->_blkreceived) * worker->_avgspeed)
631 best = worker;
632 }
633 }
634 if (!best)
635 {
638 _request->_finished = true;
639 return;
640 }
641 // do not sleep twice
642 if (_state != WORKER_SLEEP)
643 {
644 if (!_avgspeed && _blkreceived)
645 {
646 if (!now)
647 now = currentTime();
648 if (now > _blkstarttime)
650 }
651
652 // lets see if we should sleep a bit
653 XXX << "me #" << _workerno << ": " << _avgspeed << ", size " << best->_blksize << endl;
654 XXX << "best #" << best->_workerno << ": " << best->_avgspeed << ", size " << (best->_blksize - best->_blkreceived) << endl;
655 if (_avgspeed && best->_avgspeed && best->_blksize - best->_blkreceived > 0 &&
656 (best->_blksize - best->_blkreceived) * _avgspeed < best->_blksize * best->_avgspeed)
657 {
658 if (!now)
659 now = currentTime();
660 double sl = (best->_blksize - best->_blkreceived) / best->_avgspeed * 2;
661 if (sl > 1)
662 sl = 1;
663 XXX << "#" << _workerno << ": going to sleep for " << sl * 1000 << " ms" << endl;
664 _sleepuntil = now + sl;
667 return;
668 }
669 }
670
671 _competing = true;
672 best->_competing = true;
673 _blkstart = best->_blkstart;
674 _blksize = best->_blksize;
675 best->_pass++;
676 _pass = best->_pass;
677 _blkno = best->_blkno;
678 run();
679}
680
681void
683{
684 std::list<multifetchworker *>::iterator workeriter = _request->_workers.begin();
685 for (; workeriter != _request->_workers.end(); ++workeriter)
686 {
687 multifetchworker *worker = *workeriter;
688 if (worker == this)
689 continue;
690 if (worker->_blkstart == _blkstart)
691 {
692 if (worker->_state == WORKER_FETCH)
693 worker->_state = WORKER_DISCARD;
694 worker->_pass = -1; /* do not steal this one, we already have it */
695 }
696 }
697}
698
699
700void
702{
703 _noendrange = false;
704 if (_request->_stealing)
705 {
706 stealjob();
707 return;
708 }
709
710 MediaBlockList *blklist = _request->_blklist;
711 if (!blklist)
712 {
714 if (_request->_filesize != off_t(-1))
715 {
717 {
718 stealjob();
719 return;
720 }
724 }
725 DBG << "No BLOCKLIST falling back to chunk size: " << _request->_defaultBlksize << std::endl;
726 }
727 else
728 {
729 MediaBlock blk = blklist->getBlock(_request->_blkno);
730 while (_request->_blkoff >= (off_t)(blk.off + blk.size))
731 {
732 if (++_request->_blkno == blklist->numBlocks())
733 {
734 stealjob();
735 return;
736 }
737 blk = blklist->getBlock(_request->_blkno);
738 _request->_blkoff = blk.off;
739 }
740 _blksize = blk.off + blk.size - _request->_blkoff;
742 DBG << "Block: "<< _request->_blkno << " has no checksum falling back to default blocksize: " << _request->_defaultBlksize << std::endl;
744 }
745 }
749 run();
750}
751
752void
754{
755 char rangebuf[128];
756
758 return; // just in case...
759 if (_noendrange)
760 sprintf(rangebuf, "%llu-", (unsigned long long)_blkstart);
761 else
762 sprintf(rangebuf, "%llu-%llu", (unsigned long long)_blkstart, (unsigned long long)_blkstart + _blksize - 1);
763 XXX << "#" << _workerno << ": BLK " << _blkno << ":" << rangebuf << " " << _url << endl;
764 if (curl_easy_setopt(_curl, CURLOPT_RANGE, !_noendrange || _blkstart != 0 ? rangebuf : (char *)0) != CURLE_OK)
765 {
768 strncpy(_curlError, "curl_easy_setopt range failed", CURL_ERROR_SIZE);
769 return;
770 }
771 if (curl_multi_add_handle(_request->_multi, _curl) != CURLM_OK)
772 {
775 strncpy(_curlError, "curl_multi_add_handle failed", CURL_ERROR_SIZE);
776 return;
777 }
778 _request->_havenewjob = true;
779 _off = _blkstart;
780 _size = _blksize;
781 if (_request->_blklist)
782 _request->_blklist->createDigest(_dig); // resets digest
784
785 double now = currentTime();
786 _blkstarttime = now;
787 _blkreceived = 0;
788}
789
790
792
793
794multifetchrequest::multifetchrequest(const MediaMultiCurl *context, const Pathname &filename, const Url &baseurl, CURLM *multi, FILE *fp, callback::SendReport<DownloadProgressReport> *report, MediaBlockList *blklist, off_t filesize) : _context(context), _filename(filename), _baseurl(baseurl)
795{
796 _fp = fp;
797 _report = report;
798 _blklist = blklist;
799 _filesize = filesize;
800 _defaultBlksize = makeBlksize( filesize );
801 _multi = multi;
802 _stealing = false;
803 _havenewjob = false;
804 _blkno = 0;
805 if (_blklist)
807 else
808 _blkoff = 0;
809 _activeworkers = 0;
810 _lookupworkers = 0;
811 _sleepworkers = 0;
812 _minsleepuntil = 0;
813 _finished = false;
814 _fetchedsize = 0;
816 _totalsize = 0;
819 _periodavg = 0;
820 _timeout = 0;
822 _maxspeed = 0;
823 _maxworkers = 0;
824 if (blklist)
825 {
826 for (size_t blkno = 0; blkno < blklist->numBlocks(); blkno++)
827 {
828 MediaBlock blk = blklist->getBlock(blkno);
829 _totalsize += blk.size;
830 }
831 }
832 else if (filesize != off_t(-1))
833 _totalsize = filesize;
834}
835
837{
838 for (std::list<multifetchworker *>::iterator workeriter = _workers.begin(); workeriter != _workers.end(); ++workeriter)
839 {
840 multifetchworker *worker = *workeriter;
841 *workeriter = NULL;
842 delete worker;
843 }
844 _workers.clear();
845}
846
847void
848multifetchrequest::run(std::vector<Url> &urllist)
849{
850 int workerno = 0;
851 std::vector<Url>::iterator urliter = urllist.begin();
852 for (;;)
853 {
854 fd_set rset, wset, xset;
855 int maxfd, nqueue;
856
857 if (_finished)
858 {
859 XXX << "finished!" << endl;
860 break;
861 }
862
863 if ((int)_activeworkers < _maxworkers && urliter != urllist.end() && _workers.size() < MAXURLS)
864 {
865 // spawn another worker!
866 multifetchworker *worker = new multifetchworker(workerno++, *this, *urliter);
867 _workers.push_back(worker);
868 if (worker->_state != WORKER_BROKEN)
869 {
871 if (worker->_state != WORKER_LOOKUP)
872 {
873 worker->nextjob();
874 }
875 else
877 }
878 ++urliter;
879 continue;
880 }
881 if (!_activeworkers)
882 {
883 WAR << "No more active workers!" << endl;
884 // show the first worker error we find
885 for (std::list<multifetchworker *>::iterator workeriter = _workers.begin(); workeriter != _workers.end(); ++workeriter)
886 {
887 if ((*workeriter)->_state != WORKER_BROKEN)
888 continue;
889 ZYPP_THROW(MediaCurlException(_baseurl, "Server error", (*workeriter)->_curlError));
890 }
891 break;
892 }
893
894 FD_ZERO(&rset);
895 FD_ZERO(&wset);
896 FD_ZERO(&xset);
897
898 curl_multi_fdset(_multi, &rset, &wset, &xset, &maxfd);
899
900 if (_lookupworkers)
901 for (std::list<multifetchworker *>::iterator workeriter = _workers.begin(); workeriter != _workers.end(); ++workeriter)
902 (*workeriter)->adddnsfd(rset, maxfd);
903
904 timeval tv;
905 // if we added a new job we have to call multi_perform once
906 // to make it show up in the fd set. do not sleep in this case.
907 tv.tv_sec = 0;
908 tv.tv_usec = _havenewjob ? 0 : 200000;
910 {
911 if (_minsleepuntil == 0)
912 {
913 for (std::list<multifetchworker *>::iterator workeriter = _workers.begin(); workeriter != _workers.end(); ++workeriter)
914 {
915 multifetchworker *worker = *workeriter;
916 if (worker->_state != WORKER_SLEEP)
917 continue;
918 if (!_minsleepuntil || _minsleepuntil > worker->_sleepuntil)
919 _minsleepuntil = worker->_sleepuntil;
920 }
921 }
922 double sl = _minsleepuntil - currentTime();
923 if (sl < 0)
924 {
925 sl = 0;
926 _minsleepuntil = 0;
927 }
928 if (sl < .2)
929 tv.tv_usec = sl * 1000000;
930 }
931 int r = select(maxfd + 1, &rset, &wset, &xset, &tv);
932 if (r == -1 && errno != EINTR)
933 ZYPP_THROW(MediaCurlException(_baseurl, "select() failed", "unknown error"));
934 if (r != 0 && _lookupworkers)
935 for (std::list<multifetchworker *>::iterator workeriter = _workers.begin(); workeriter != _workers.end(); ++workeriter)
936 {
937 multifetchworker *worker = *workeriter;
938 if (worker->_state != WORKER_LOOKUP)
939 continue;
940 (*workeriter)->dnsevent(rset);
941 if (worker->_state != WORKER_LOOKUP)
943 }
944 _havenewjob = false;
945
946 // run curl
947 for (;;)
948 {
949 CURLMcode mcode;
950 int tasks;
951 mcode = curl_multi_perform(_multi, &tasks);
952 if (mcode == CURLM_CALL_MULTI_PERFORM)
953 continue;
954 if (mcode != CURLM_OK)
955 ZYPP_THROW(MediaCurlException(_baseurl, "curl_multi_perform", "unknown error"));
956 break;
957 }
958
959 double now = currentTime();
960
961 // update periodavg
962 if (now > _lastperiodstart + .5)
963 {
964 if (!_periodavg)
966 else
969 _lastperiodstart = now;
970 }
971
972 // wake up sleepers
973 if (_sleepworkers)
974 {
975 for (std::list<multifetchworker *>::iterator workeriter = _workers.begin(); workeriter != _workers.end(); ++workeriter)
976 {
977 multifetchworker *worker = *workeriter;
978 if (worker->_state != WORKER_SLEEP)
979 continue;
980 if (worker->_sleepuntil > now)
981 continue;
982 if (_minsleepuntil == worker->_sleepuntil)
983 _minsleepuntil = 0;
984 XXX << "#" << worker->_workerno << ": sleep done, wake up" << endl;
986 // nextjob chnages the state
987 worker->nextjob();
988 }
989 }
990
991 // collect all curl results, reschedule new jobs
992 CURLMsg *msg;
993 while ((msg = curl_multi_info_read(_multi, &nqueue)) != 0)
994 {
995 if (msg->msg != CURLMSG_DONE)
996 continue;
997 CURL *easy = msg->easy_handle;
998 CURLcode cc = msg->data.result;
999 multifetchworker *worker;
1000 if (curl_easy_getinfo(easy, CURLINFO_PRIVATE, &worker) != CURLE_OK)
1001 ZYPP_THROW(MediaCurlException(_baseurl, "curl_easy_getinfo", "unknown error"));
1002 if (worker->_blkreceived && now > worker->_blkstarttime)
1003 {
1004 if (worker->_avgspeed)
1005 worker->_avgspeed = (worker->_avgspeed + worker->_blkreceived / (now - worker->_blkstarttime)) / 2;
1006 else
1007 worker->_avgspeed = worker->_blkreceived / (now - worker->_blkstarttime);
1008 }
1009 XXX << "#" << worker->_workerno << ": BLK " << worker->_blkno << " done code " << cc << " speed " << worker->_avgspeed << endl;
1010 curl_multi_remove_handle(_multi, easy);
1011 if (cc == CURLE_HTTP_RETURNED_ERROR)
1012 {
1013 long statuscode = 0;
1014 (void)curl_easy_getinfo(easy, CURLINFO_RESPONSE_CODE, &statuscode);
1015 XXX << "HTTP status " << statuscode << endl;
1016 if (statuscode == 416 && !_blklist) /* Range error */
1017 {
1018 if (_filesize == off_t(-1))
1019 {
1020 if (!worker->_noendrange)
1021 {
1022 XXX << "#" << worker->_workerno << ": retrying with no end range" << endl;
1023 worker->_noendrange = true;
1024 worker->run();
1025 continue;
1026 }
1027 worker->_noendrange = false;
1028 worker->stealjob();
1029 continue;
1030 }
1031 if (worker->_blkstart >= _filesize)
1032 {
1033 worker->nextjob();
1034 continue;
1035 }
1036 }
1037 }
1038 if (cc == 0)
1039 {
1040 if (!worker->checkChecksum())
1041 {
1042 WAR << "#" << worker->_workerno << ": checksum error, disable worker" << endl;
1043 worker->_state = WORKER_BROKEN;
1044 strncpy(worker->_curlError, "checksum error", CURL_ERROR_SIZE);
1046 continue;
1047 }
1048 if (worker->_state == WORKER_FETCH)
1049 {
1050 if (worker->_competing)
1051 {
1052 worker->disableCompetition();
1053 // multiple workers wrote into this block. We already know that our
1054 // data was correct, but maybe some other worker overwrote our data
1055 // with something broken. Thus we have to re-check the block.
1056 if (!worker->recheckChecksum())
1057 {
1058 XXX << "#" << worker->_workerno << ": recheck checksum error, refetch block" << endl;
1059 // re-fetch! No need to worry about the bad workers,
1060 // they will now be set to DISCARD. At the end of their block
1061 // they will notice that they wrote bad data and go into BROKEN.
1062 worker->run();
1063 continue;
1064 }
1065 }
1066 _fetchedgoodsize += worker->_blksize;
1067 }
1068
1069 // make bad workers sleep a little
1070 double maxavg = 0;
1071 int maxworkerno = 0;
1072 int numbetter = 0;
1073 for (std::list<multifetchworker *>::iterator workeriter = _workers.begin(); workeriter != _workers.end(); ++workeriter)
1074 {
1075 multifetchworker *oworker = *workeriter;
1076 if (oworker->_state == WORKER_BROKEN)
1077 continue;
1078 if (oworker->_avgspeed > maxavg)
1079 {
1080 maxavg = oworker->_avgspeed;
1081 maxworkerno = oworker->_workerno;
1082 }
1083 if (oworker->_avgspeed > worker->_avgspeed)
1084 numbetter++;
1085 }
1086 if (maxavg && !_stealing)
1087 {
1088 double ratio = worker->_avgspeed / maxavg;
1089 ratio = 1 - ratio;
1090 if (numbetter < 3) // don't sleep that much if we're in the top two
1091 ratio = ratio * ratio;
1092 if (ratio > .01)
1093 {
1094 XXX << "#" << worker->_workerno << ": too slow ("<< ratio << ", " << worker->_avgspeed << ", #" << maxworkerno << ": " << maxavg << "), going to sleep for " << ratio * 1000 << " ms" << endl;
1095 worker->_sleepuntil = now + ratio;
1096 worker->_state = WORKER_SLEEP;
1097 _sleepworkers++;
1098 continue;
1099 }
1100 }
1101
1102 // do rate control (if requested)
1103 // should use periodavg, but that's not what libcurl does
1104 if (_maxspeed && now > _starttime)
1105 {
1106 double avg = _fetchedsize / (now - _starttime);
1107 avg = worker->_maxspeed * _maxspeed / avg;
1108 if (avg < _maxspeed / _maxworkers)
1109 avg = _maxspeed / _maxworkers;
1110 if (avg > _maxspeed)
1111 avg = _maxspeed;
1112 if (avg < 1024)
1113 avg = 1024;
1114 worker->_maxspeed = avg;
1115#if CURLVERSION_AT_LEAST(7,15,5)
1116 curl_easy_setopt(worker->_curl, CURLOPT_MAX_RECV_SPEED_LARGE, (curl_off_t)(avg));
1117#endif
1118 }
1119
1120 worker->nextjob();
1121 }
1122 else
1123 {
1124 worker->_state = WORKER_BROKEN;
1126 if (!_activeworkers && !(urliter != urllist.end() && _workers.size() < MAXURLS))
1127 {
1128 // end of workers reached! goodbye!
1129 worker->evaluateCurlCode(Pathname(), cc, false);
1130 }
1131 }
1132
1133 if ( _filesize > 0 && _fetchedgoodsize > _filesize ) {
1135 }
1136 }
1137
1138 // send report
1139 if (_report)
1140 {
1141 int percent = _totalsize ? (100 * (_fetchedgoodsize + _fetchedsize)) / (_totalsize + _fetchedsize) : 0;
1142
1143 double avg = 0;
1144 if (now > _starttime)
1145 avg = _fetchedsize / (now - _starttime);
1146 if (!(*(_report))->progress(percent, _baseurl, avg, _lastperiodstart == _starttime ? avg : _periodavg))
1147 ZYPP_THROW(MediaCurlException(_baseurl, "User abort", "cancelled"));
1148 }
1149
1150 if (_timeout && now - _lastprogress > _timeout)
1151 break;
1152 }
1153
1154 if (!_finished)
1156
1157 // print some download stats
1158 WAR << "overall result" << endl;
1159 for (std::list<multifetchworker *>::iterator workeriter = _workers.begin(); workeriter != _workers.end(); ++workeriter)
1160 {
1161 multifetchworker *worker = *workeriter;
1162 WAR << "#" << worker->_workerno << ": state: " << worker->_state << " received: " << worker->_received << " url: " << worker->_url << endl;
1163 }
1164}
1165
1166inline size_t multifetchrequest::makeBlksize ( size_t filesize )
1167{
1168 // this case should never happen because we never start a multi download if we do not know the filesize beforehand
1169 if ( filesize == 0 ) return 2 * 1024 * 1024;
1170 else if ( filesize < 2*256*1024 ) return filesize;
1171 else if ( filesize < 8*1024*1024 ) return 256*1024;
1172 else if ( filesize < 256*1024*1024 ) return 1024*1024;
1173 return 4*1024*1024;
1174}
1175
1177
1178
1179MediaMultiCurl::MediaMultiCurl(const Url &url_r, const Pathname & attach_point_hint_r)
1180 : MediaCurl(url_r, attach_point_hint_r)
1181{
1182 MIL << "MediaMultiCurl::MediaMultiCurl(" << url_r << ", " << attach_point_hint_r << ")" << endl;
1183 _multi = 0;
1185}
1186
1188{
1190 {
1191 curl_slist_free_all(_customHeadersMetalink);
1193 }
1194 if (_multi)
1195 {
1196 curl_multi_cleanup(_multi);
1197 _multi = 0;
1198 }
1199 std::map<std::string, CURL *>::iterator it;
1200 for (it = _easypool.begin(); it != _easypool.end(); it++)
1201 {
1202 CURL *easy = it->second;
1203 if (easy)
1204 {
1205 curl_easy_cleanup(easy);
1206 it->second = NULL;
1207 }
1208 }
1209}
1210
1212{
1214
1216 {
1217 curl_slist_free_all(_customHeadersMetalink);
1219 }
1220 struct curl_slist *sl = _customHeaders;
1221 for (; sl; sl = sl->next)
1222 _customHeadersMetalink = curl_slist_append(_customHeadersMetalink, sl->data);
1223 _customHeadersMetalink = curl_slist_append(_customHeadersMetalink, "Accept: */*, application/metalink+xml, application/metalink4+xml");
1224}
1225
1226static bool looks_like_metalink_fd(int fd)
1227{
1228 char buf[256], *p;
1229 int l;
1230 while ((l = pread(fd, buf, sizeof(buf) - 1, (off_t)0)) == -1 && errno == EINTR)
1231 ;
1232 if (l == -1)
1233 return 0;
1234 buf[l] = 0;
1235 p = buf;
1236 while (*p == ' ' || *p == '\t' || *p == '\r' || *p == '\n')
1237 p++;
1238 if (!strncasecmp(p, "<?xml", 5))
1239 {
1240 while (*p && *p != '>')
1241 p++;
1242 if (*p == '>')
1243 p++;
1244 while (*p == ' ' || *p == '\t' || *p == '\r' || *p == '\n')
1245 p++;
1246 }
1247 bool ret = !strncasecmp(p, "<metalink", 9) ? true : false;
1248 return ret;
1249}
1250
1251static bool looks_like_metalink(const Pathname & file)
1252{
1253 int fd;
1254 if ((fd = open(file.asString().c_str(), O_RDONLY|O_CLOEXEC)) == -1)
1255 return false;
1256 bool ret = looks_like_metalink_fd(fd);
1257 close(fd);
1258 DBG << "looks_like_metalink(" << file << "): " << ret << endl;
1259 return ret;
1260}
1261
1262// here we try to suppress all progress coming from a metalink download
1263// bsc#1021291: Nevertheless send alive trigger (without stats), so UIs
1264// are able to abort a hanging metalink download via callback response.
1265int MediaMultiCurl::progressCallback( void *clientp, double dltotal, double dlnow, double ultotal, double ulnow)
1266{
1268 if (!_curl)
1269 return MediaCurl::aliveCallback(clientp, dltotal, dlnow, ultotal, ulnow);
1270
1271 // bsc#408814: Don't report any sizes before we don't have data on disk. Data reported
1272 // due to redirection etc. are not interesting, but may disturb filesize checks.
1273 FILE *fp = 0;
1274 if ( curl_easy_getinfo( _curl, CURLINFO_PRIVATE, &fp ) != CURLE_OK || !fp )
1275 return MediaCurl::aliveCallback( clientp, dltotal, dlnow, ultotal, ulnow );
1276 if ( ftell( fp ) == 0 )
1277 return MediaCurl::aliveCallback( clientp, dltotal, 0.0, ultotal, ulnow );
1278
1279 // (no longer needed due to the filesize check above?)
1280 // work around curl bug that gives us old data
1281 long httpReturnCode = 0;
1282 if (curl_easy_getinfo(_curl, CURLINFO_RESPONSE_CODE, &httpReturnCode ) != CURLE_OK || httpReturnCode == 0)
1283 return MediaCurl::aliveCallback(clientp, dltotal, dlnow, ultotal, ulnow);
1284
1285 char *ptr = NULL;
1286 bool ismetalink = false;
1287 if (curl_easy_getinfo(_curl, CURLINFO_CONTENT_TYPE, &ptr) == CURLE_OK && ptr)
1288 {
1289 std::string ct = std::string(ptr);
1290 if (ct.find("application/metalink+xml") == 0 || ct.find("application/metalink4+xml") == 0)
1291 ismetalink = true;
1292 }
1293 if (!ismetalink && dlnow < 256)
1294 {
1295 // can't tell yet, ...
1296 return MediaCurl::aliveCallback(clientp, dltotal, dlnow, ultotal, ulnow);
1297 }
1298 if (!ismetalink)
1299 {
1300 fflush(fp);
1301 ismetalink = looks_like_metalink_fd(fileno(fp));
1302 DBG << "looks_like_metalink_fd: " << ismetalink << endl;
1303 }
1304 if (ismetalink)
1305 {
1306 // this is a metalink file change the expected filesize
1308 // we're downloading the metalink file. Just trigger aliveCallbacks
1309 curl_easy_setopt(_curl, CURLOPT_PROGRESSFUNCTION, &MediaCurl::aliveCallback);
1310 return MediaCurl::aliveCallback(clientp, dltotal, dlnow, ultotal, ulnow);
1311 }
1312 curl_easy_setopt(_curl, CURLOPT_PROGRESSFUNCTION, &MediaCurl::progressCallback);
1313 return MediaCurl::progressCallback(clientp, dltotal, dlnow, ultotal, ulnow);
1314}
1315
1316void MediaMultiCurl::doGetFileCopy( const OnMediaLocation &srcFile , const Pathname & target, callback::SendReport<DownloadProgressReport> & report, RequestOptions options ) const
1317{
1318 Pathname dest = target.absolutename();
1319 if( assert_dir( dest.dirname() ) )
1320 {
1321 DBG << "assert_dir " << dest.dirname() << " failed" << endl;
1322 ZYPP_THROW( MediaSystemException(getFileUrl(srcFile.filename()), "System error on " + dest.dirname().asString()) );
1323 }
1324
1325 ManagedFile destNew { target.extend( ".new.zypp.XXXXXX" ) };
1326 AutoFILE file;
1327 {
1328 AutoFREE<char> buf { ::strdup( (*destNew).c_str() ) };
1329 if( ! buf )
1330 {
1331 ERR << "out of memory for temp file name" << endl;
1332 ZYPP_THROW(MediaSystemException(getFileUrl(srcFile.filename()), "out of memory for temp file name"));
1333 }
1334
1335 AutoFD tmp_fd { ::mkostemp( buf, O_CLOEXEC ) };
1336 if( tmp_fd == -1 )
1337 {
1338 ERR << "mkstemp failed for file '" << destNew << "'" << endl;
1340 }
1341 destNew = ManagedFile( (*buf), filesystem::unlink );
1342
1343 file = ::fdopen( tmp_fd, "we" );
1344 if ( ! file )
1345 {
1346 ERR << "fopen failed for file '" << destNew << "'" << endl;
1348 }
1349 tmp_fd.resetDispose(); // don't close it here! ::fdopen moved ownership to file
1350 }
1351
1352 DBG << "dest: " << dest << endl;
1353 DBG << "temp: " << destNew << endl;
1354
1355 // set IFMODSINCE time condition (no download if not modified)
1356 if( PathInfo(target).isExist() && !(options & OPTION_NO_IFMODSINCE) )
1357 {
1358 curl_easy_setopt(_curl, CURLOPT_TIMECONDITION, CURL_TIMECOND_IFMODSINCE);
1359 curl_easy_setopt(_curl, CURLOPT_TIMEVALUE, (long)PathInfo(target).mtime());
1360 }
1361 else
1362 {
1363 curl_easy_setopt(_curl, CURLOPT_TIMECONDITION, CURL_TIMECOND_NONE);
1364 curl_easy_setopt(_curl, CURLOPT_TIMEVALUE, 0L);
1365 }
1366 // change header to include Accept: metalink
1367 curl_easy_setopt(_curl, CURLOPT_HTTPHEADER, _customHeadersMetalink);
1368 // change to our own progress funcion
1369 curl_easy_setopt(_curl, CURLOPT_PROGRESSFUNCTION, &progressCallback);
1370 curl_easy_setopt(_curl, CURLOPT_PRIVATE, (*file) ); // important to pass the FILE* explicitly (passing through varargs)
1371 try
1372 {
1373 MediaCurl::doGetFileCopyFile( srcFile, dest, file, report, options );
1374 }
1375 catch (Exception &ex)
1376 {
1377 curl_easy_setopt(_curl, CURLOPT_TIMECONDITION, CURL_TIMECOND_NONE);
1378 curl_easy_setopt(_curl, CURLOPT_TIMEVALUE, 0L);
1379 curl_easy_setopt(_curl, CURLOPT_HTTPHEADER, _customHeaders);
1380 curl_easy_setopt(_curl, CURLOPT_PRIVATE, (void *)0);
1381 ZYPP_RETHROW(ex);
1382 }
1383 curl_easy_setopt(_curl, CURLOPT_TIMECONDITION, CURL_TIMECOND_NONE);
1384 curl_easy_setopt(_curl, CURLOPT_TIMEVALUE, 0L);
1385 curl_easy_setopt(_curl, CURLOPT_HTTPHEADER, _customHeaders);
1386 curl_easy_setopt(_curl, CURLOPT_PRIVATE, (void *)0);
1387 long httpReturnCode = 0;
1388 CURLcode infoRet = curl_easy_getinfo(_curl, CURLINFO_RESPONSE_CODE, &httpReturnCode);
1389 if (infoRet == CURLE_OK)
1390 {
1391 DBG << "HTTP response: " + str::numstring(httpReturnCode) << endl;
1392 if ( httpReturnCode == 304
1393 || ( httpReturnCode == 213 && _url.getScheme() == "ftp" ) ) // not modified
1394 {
1395 DBG << "not modified: " << PathInfo(dest) << endl;
1396 return;
1397 }
1398 }
1399 else
1400 {
1401 WAR << "Could not get the response code." << endl;
1402 }
1403
1404 bool ismetalink = false;
1405
1406 char *ptr = NULL;
1407 if (curl_easy_getinfo(_curl, CURLINFO_CONTENT_TYPE, &ptr) == CURLE_OK && ptr)
1408 {
1409 std::string ct = std::string(ptr);
1410 if (ct.find("application/metalink+xml") == 0 || ct.find("application/metalink4+xml") == 0)
1411 ismetalink = true;
1412 }
1413
1414 if (!ismetalink)
1415 {
1416 // some proxies do not store the content type, so also look at the file to find
1417 // out if we received a metalink (bnc#649925)
1418 fflush(file);
1419 if (looks_like_metalink(destNew))
1420 ismetalink = true;
1421 }
1422
1423 if (ismetalink)
1424 {
1425 bool userabort = false;
1426 Pathname failedFile = ZConfig::instance().repoCachePath() / "MultiCurl.failed";
1427 file = nullptr; // explicitly close destNew before the parser reads it.
1428 try
1429 {
1430 MetaLinkParser mlp;
1431 mlp.parse(destNew);
1432 MediaBlockList bl = mlp.getBlockList();
1433
1434 /*
1435 * gihub issue libzipp:#277 Multicurl backend breaks with MirrorCache and Metalink with unknown filesize.
1436 * Fall back to a normal download if we have no knowledge about the filesize we want to download.
1437 */
1438 if ( !bl.haveFilesize() && ! srcFile.downloadSize() ) {
1439 XXX << "No filesize in metalink file and no expected filesize, aborting multicurl." << std::endl;
1440 ZYPP_THROW( MediaException("Multicurl requires filesize but none was provided.") );
1441 }
1442
1443 std::vector<Url> urls = mlp.getUrls();
1444 /*
1445 * bsc#1191609 In certain locations we do not receive a suitable number of metalink mirrors, and might even
1446 * download chunks serially from one and the same server. In those cases we need to fall back to a normal download.
1447 */
1448 if ( urls.size() < MIN_REQ_MIRRS ) {
1449 ZYPP_THROW( MediaException("Multicurl enabled but not enough mirrors provided") );
1450 }
1451
1452 XXX << bl << endl;
1453 file = fopen((*destNew).c_str(), "w+e");
1454 if (!file)
1456 if (PathInfo(target).isExist())
1457 {
1458 XXX << "reusing blocks from file " << target << endl;
1459 bl.reuseBlocks(file, target.asString());
1460 XXX << bl << endl;
1461 }
1462 if (bl.haveChecksum(1) && PathInfo(failedFile).isExist())
1463 {
1464 XXX << "reusing blocks from file " << failedFile << endl;
1465 bl.reuseBlocks(file, failedFile.asString());
1466 XXX << bl << endl;
1467 filesystem::unlink(failedFile);
1468 }
1469 Pathname df = srcFile.deltafile();
1470 if (!df.empty())
1471 {
1472 XXX << "reusing blocks from file " << df << endl;
1473 bl.reuseBlocks(file, df.asString());
1474 XXX << bl << endl;
1475 }
1476 try
1477 {
1478 multifetch(srcFile.filename(), file, &urls, &report, &bl, srcFile.downloadSize());
1479 }
1480 catch (MediaCurlException &ex)
1481 {
1482 userabort = ex.errstr() == "User abort";
1483 ZYPP_RETHROW(ex);
1484 }
1485 }
1486 catch (MediaFileSizeExceededException &ex) {
1487 ZYPP_RETHROW(ex);
1488 }
1489 catch (Exception &ex)
1490 {
1491 // something went wrong. fall back to normal download
1492 file = nullptr; // explicitly close destNew before moving it
1493 if (PathInfo(destNew).size() >= 63336)
1494 {
1495 ::unlink(failedFile.asString().c_str());
1496 filesystem::hardlinkCopy(destNew, failedFile);
1497 }
1498 if (userabort)
1499 {
1500 ZYPP_RETHROW(ex);
1501 }
1502 file = fopen((*destNew).c_str(), "w+e");
1503 if (!file)
1505
1506 // use the default progressCallback
1507 curl_easy_setopt(_curl, CURLOPT_PROGRESSFUNCTION, &MediaCurl::progressCallback);
1508 MediaCurl::doGetFileCopyFile(srcFile, dest, file, report, options | OPTION_NO_REPORT_START);
1509 }
1510 }
1511
1512 if (::fchmod( ::fileno(file), filesystem::applyUmaskTo( 0644 )))
1513 {
1514 ERR << "Failed to chmod file " << destNew << endl;
1515 }
1516
1517 file.resetDispose(); // we're going to close it manually here
1518 if (::fclose(file))
1519 {
1520 filesystem::unlink(destNew);
1521 ERR << "Fclose failed for file '" << destNew << "'" << endl;
1523 }
1524
1525 if ( rename( destNew, dest ) != 0 )
1526 {
1527 ERR << "Rename failed" << endl;
1529 }
1530 destNew.resetDispose(); // no more need to unlink it
1531
1532 DBG << "done: " << PathInfo(dest) << endl;
1533}
1534
1535void MediaMultiCurl::multifetch(const Pathname & filename, FILE *fp, std::vector<Url> *urllist, callback::SendReport<DownloadProgressReport> *report, MediaBlockList *blklist, off_t filesize) const
1536{
1537 Url baseurl(getFileUrl(filename));
1538 if (blklist && filesize == off_t(-1) && blklist->haveFilesize())
1539 filesize = blklist->getFilesize();
1540 if (blklist && !blklist->haveBlocks() && filesize != 0)
1541 blklist = 0;
1542 if (blklist && (filesize == 0 || !blklist->numBlocks()))
1543 {
1544 checkFileDigest(baseurl, fp, blklist);
1545 return;
1546 }
1547 if (filesize == 0)
1548 return;
1549 if (!_multi)
1550 {
1551 _multi = curl_multi_init();
1552 if (!_multi)
1554 }
1555
1556 multifetchrequest req(this, filename, baseurl, _multi, fp, report, blklist, filesize);
1557 req._timeout = _settings.timeout();
1561 if (req._maxworkers > MAXURLS)
1562 req._maxworkers = MAXURLS;
1563 if (req._maxworkers <= 0)
1564 req._maxworkers = 1;
1565 std::vector<Url> myurllist;
1566 for (std::vector<Url>::iterator urliter = urllist->begin(); urliter != urllist->end(); ++urliter)
1567 {
1568 try
1569 {
1570 std::string scheme = urliter->getScheme();
1571 if (scheme == "http" || scheme == "https" || scheme == "ftp" || scheme == "tftp")
1572 {
1573 checkProtocol(*urliter);
1574 myurllist.push_back(internal::propagateQueryParams(*urliter, _url));
1575 }
1576 }
1577 catch (...)
1578 {
1579 }
1580 }
1581 if (!myurllist.size())
1582 myurllist.push_back(baseurl);
1583 req.run(myurllist);
1584 checkFileDigest(baseurl, fp, blklist);
1585}
1586
1587void MediaMultiCurl::checkFileDigest(Url &url, FILE *fp, MediaBlockList *blklist) const
1588{
1589 if (!blklist || !blklist->haveFileChecksum())
1590 return;
1591 if (fseeko(fp, off_t(0), SEEK_SET))
1592 ZYPP_THROW(MediaCurlException(url, "fseeko", "seek error"));
1593 Digest dig;
1594 blklist->createFileDigest(dig);
1595 char buf[4096];
1596 size_t l;
1597 while ((l = fread(buf, 1, sizeof(buf), fp)) > 0)
1598 dig.update(buf, l);
1599 if (!blklist->verifyFileDigest(dig))
1600 ZYPP_THROW(MediaCurlException(url, "file verification failed", "checksum error"));
1601}
1602
1603bool MediaMultiCurl::isDNSok(const std::string &host) const
1604{
1605 return _dnsok.find(host) == _dnsok.end() ? false : true;
1606}
1607
1608void MediaMultiCurl::setDNSok(const std::string &host) const
1609{
1610 _dnsok.insert(host);
1611}
1612
1613CURL *MediaMultiCurl::fromEasyPool(const std::string &host) const
1614{
1615 if (_easypool.find(host) == _easypool.end())
1616 return 0;
1617 CURL *ret = _easypool[host];
1618 _easypool.erase(host);
1619 return ret;
1620}
1621
1622void MediaMultiCurl::toEasyPool(const std::string &host, CURL *easy) const
1623{
1624 CURL *oldeasy = _easypool[host];
1625 _easypool[host] = easy;
1626 if (oldeasy)
1627 curl_easy_cleanup(oldeasy);
1628}
1629
1630 } // namespace media
1631} // namespace zypp
std::optional< KeyManagerCtx > _context
Definition: KeyRing.cc:157
#define WORKER_STARTING
#define WORKER_DISCARD
#define WORKER_DONE
#define WORKER_LOOKUP
#define WORKER_FETCH
#define WORKER_SLEEP
#define WORKER_BROKEN
void resetDispose()
Set no dispose function.
Definition: AutoDispose.h:180
Store and operate with byte count.
Definition: ByteCount.h:31
static const Unit MB
1000^2 Byte
Definition: ByteCount.h:60
std::string asString(unsigned field_width_r=0, unsigned unit_width_r=1) const
Auto selected Unit and precision.
Definition: ByteCount.h:133
Compute Message Digests (MD5, SHA1 etc)
Definition: Digest.h:37
bool update(const char *bytes, size_t len)
feed data into digest computation algorithm
Definition: Digest.cc:252
Base class for Exception.
Definition: Exception.h:146
Describes a resource file located on a medium.
const ByteCount & downloadSize() const
The size of the resource on the server.
const Pathname & filename() const
The path to the resource on the medium.
const Pathname & deltafile() const
The existing deltafile that can be used to reduce download size ( zchunk or metalink )
Url manipulation class.
Definition: Url.h:92
std::string getScheme() const
Returns the scheme name of the URL.
Definition: Url.cc:533
std::string asString() const
Returns a default string representation of the Url object.
Definition: Url.cc:497
std::string getHost(EEncoding eflag=zypp::url::E_DECODED) const
Returns the hostname or IP from the URL authority.
Definition: Url.cc:588
Pathname repoCachePath() const
Path where the caches are kept (/var/cache/zypp)
Definition: ZConfig.cc:949
static ZConfig & instance()
Singleton ctor.
Definition: ZConfig.cc:832
Wrapper class for ::stat/::lstat.
Definition: PathInfo.h:221
Pathname extend(const std::string &r) const
Append string r to the last component of the path.
Definition: Pathname.h:173
Pathname dirname() const
Return all but the last component od this path.
Definition: Pathname.h:124
const std::string & asString() const
String representation.
Definition: Pathname.h:91
Pathname absolutename() const
Return this path, adding a leading '/' if relative.
Definition: Pathname.h:139
static long auth_type_str2long(std::string &auth_type_str)
Converts a string of comma separated list of authetication type names into a long of ORed CURLAUTH_* ...
Definition: curlauthdata.cc:50
bool haveChecksum(size_t blkno) const
void reuseBlocks(FILE *wfp, std::string filename)
bool createDigest(Digest &digest) const
const MediaBlock & getBlock(size_t blkno) const
return the offset/size of a block with number blkno
bool verifyDigest(size_t blkno, Digest &digest) const
bool haveBlocks() const
do we have a blocklist describing the file? set to true when addBlock() is called
bool createFileDigest(Digest &digest) const
size_t numBlocks() const
return the number of blocks in the blocklist
bool verifyFileDigest(Digest &digest) const
Implementation class for FTP, HTTP and HTTPS MediaHandler.
Definition: MediaCurl.h:32
virtual void setupEasy()
initializes the curl easy handle with the data from the url
Definition: MediaCurl.cc:407
@ OPTION_NO_IFMODSINCE
to not add a IFMODSINCE header if target exists
Definition: MediaCurl.h:43
@ OPTION_NO_REPORT_START
do not send a start ProgressReport
Definition: MediaCurl.h:45
static void resetExpectedFileSize(void *clientp, const ByteCount &expectedFileSize)
MediaMultiCurl needs to reset the expected filesize in case a metalink file is downloaded otherwise t...
Definition: MediaCurl.cc:1461
static int aliveCallback(void *clientp, double dltotal, double dlnow, double ultotal, double ulnow)
Callback sending just an alive trigger to the UI, without stats (e.g.
Definition: MediaCurl.cc:1404
static int progressCallback(void *clientp, double dltotal, double dlnow, double ultotal, double ulnow)
Callback reporting download progress.
Definition: MediaCurl.cc:1418
Url clearQueryString(const Url &url) const
Definition: MediaCurl.cc:370
char _curlError[CURL_ERROR_SIZE]
Definition: MediaCurl.h:169
void doGetFileCopyFile(const OnMediaLocation &srcFile, const Pathname &dest, FILE *file, callback::SendReport< DownloadProgressReport > &report, RequestOptions options=OPTION_NONE) const
Definition: MediaCurl.cc:1248
void checkProtocol(const Url &url) const
check the url is supported by the curl library
Definition: MediaCurl.cc:382
void evaluateCurlCode(const zypp::Pathname &filename, CURLcode code, bool timeout) const
Evaluates a curl return code and throws the right MediaException filename Filename being downloaded c...
Definition: MediaCurl.cc:840
static CURL * progressCallback_getcurl(void *clientp)
Definition: MediaCurl.cc:1434
virtual void disconnectFrom() override
Definition: MediaCurl.cc:710
curl_slist * _customHeaders
Definition: MediaCurl.h:170
Just inherits Exception to separate media exceptions.
Url url() const
Url used.
Definition: MediaHandler.h:503
const Url _url
Url to handle.
Definition: MediaHandler.h:113
virtual void setupEasy() override
initializes the curl easy handle with the data from the url
std::map< std::string, CURL * > _easypool
curl_slist * _customHeadersMetalink
void multifetch(const Pathname &filename, FILE *fp, std::vector< Url > *urllist, callback::SendReport< DownloadProgressReport > *report=0, MediaBlockList *blklist=0, off_t filesize=off_t(-1)) const
static int progressCallback(void *clientp, double dltotal, double dlnow, double ultotal, double ulnow)
void setDNSok(const std::string &host) const
MediaMultiCurl(const Url &url_r, const Pathname &attach_point_hint_r)
std::set< std::string > _dnsok
bool isDNSok(const std::string &host) const
CURL * fromEasyPool(const std::string &host) const
virtual void doGetFileCopy(const OnMediaLocation &srcFile, const Pathname &targetFilename, callback::SendReport< DownloadProgressReport > &_report, RequestOptions options=OPTION_NONE) const override
void checkFileDigest(Url &url, FILE *fp, MediaBlockList *blklist) const
void toEasyPool(const std::string &host, CURL *easy) const
Url getFileUrl(const Pathname &filename) const
concatenate the attach url and the filename to a complete download url
void parse(const Pathname &filename)
parse a file consisting of metalink xml data
MediaBlockList getBlockList() const
return the block list from the parsed metalink data
std::vector< Url > getUrls() const
return the download urls from the parsed metalink data
const std::string & password() const
auth password
long maxDownloadSpeed() const
Maximum download speed (bytes per second)
long connectTimeout() const
connection timeout
const std::string & authType() const
get the allowed authentication types
long timeout() const
transfer timeout
void setUsername(const std::string &val_r)
sets the auth username
std::string userPassword() const
returns the user and password as a user:pass string
const std::string & proxy() const
proxy host
long maxConcurrentConnections() const
Maximum number of concurrent connections for a single transfer.
void setPassword(const std::string &val_r)
sets the auth password
const std::string & username() const
auth username
void setAuthType(const std::string &val_r)
set the allowed authentication types
callback::SendReport< DownloadProgressReport > * _report
void run(std::vector< Url > &urllist)
std::list< multifetchworker * > _workers
multifetchrequest(const MediaMultiCurl *context, const Pathname &filename, const Url &baseurl, CURLM *multi, FILE *fp, callback::SendReport< DownloadProgressReport > *report, MediaBlockList *blklist, off_t filesize)
static size_t makeBlksize(size_t filesize)
const MediaMultiCurl * _context
size_t headerfunction(char *ptr, size_t size)
multifetchworker(int no, multifetchrequest &request, const Url &url)
multifetchrequest * _request
size_t writefunction(void *ptr, size_t size)
static size_t _writefunction(void *ptr, size_t size, size_t nmemb, void *stream)
static size_t _headerfunction(void *ptr, size_t size, size_t nmemb, void *stream)
void adddnsfd(fd_set &rset, int &maxfd)
zypp::Url propagateQueryParams(zypp::Url url_r, const zypp::Url &template_r)
Definition: curlhelper.cc:358
mode_t applyUmaskTo(mode_t mode_r)
Modify mode_r according to the current umask ( mode_r & ~getUmask() ).
Definition: PathInfo.h:789
ByteCount df(const Pathname &path_r)
Report free disk space on a mounted file system.
Definition: PathInfo.cc:1155
int unlink(const Pathname &path)
Like 'unlink'.
Definition: PathInfo.cc:700
int rename(const Pathname &oldpath, const Pathname &newpath)
Like 'rename'.
Definition: PathInfo.cc:742
int assert_dir(const Pathname &path, unsigned mode)
Like 'mkdir -p'.
Definition: PathInfo.cc:319
int hardlinkCopy(const Pathname &oldpath, const Pathname &newpath)
Create newpath as hardlink or copy of oldpath.
Definition: PathInfo.cc:883
static bool looks_like_metalink(const Pathname &file)
static bool looks_like_metalink_fd(int fd)
constexpr auto MAXURLS
static bool env_isset(std::string name)
static double currentTime()
constexpr auto MIN_REQ_MIRRS
std::string numstring(char n, int w=0)
Definition: String.h:289
Easy-to use interface to the ZYPP dependency resolver.
Definition: CodePitfalls.doc:2
AutoDispose< const Pathname > ManagedFile
A Pathname plus associated cleanup code to be executed when path is no longer needed.
Definition: ManagedFile.h:27
AutoDispose<int> calling ::close
Definition: AutoDispose.h:301
AutoDispose<FILE*> calling ::fclose
Definition: AutoDispose.h:312
a single block from the blocklist, consisting of an offset and a size
#define ZYPP_RETHROW(EXCPT)
Drops a logline and rethrows, updating the CodeLocation.
Definition: Exception.h:440
#define ZYPP_THROW(EXCPT)
Drops a logline and throws the Exception.
Definition: Exception.h:428
#define DBG
Definition: Logger.h:95
#define MIL
Definition: Logger.h:96
#define ERR
Definition: Logger.h:98
#define WAR
Definition: Logger.h:97
#define XXX
Definition: Logger.h:94