]> git.saurik.com Git - apt.git/blob - methods/http.cc
fddf8a78e096bc94a645dd938303d6bccdf09152
[apt.git] / methods / http.cc
1 // -*- mode: cpp; mode: fold -*-
2 // Description /*{{{*/
3 // $Id: http.cc,v 1.59 2004/05/08 19:42:35 mdz Exp $
4 /* ######################################################################
5
6 HTTP Acquire Method - This is the HTTP aquire method for APT.
7
8 It uses HTTP/1.1 and many of the fancy options there-in, such as
9 pipelining, range, if-range and so on.
10
11 It is based on a doubly buffered select loop. A groupe of requests are
12 fed into a single output buffer that is constantly fed out the
13 socket. This provides ideal pipelining as in many cases all of the
14 requests will fit into a single packet. The input socket is buffered
15 the same way and fed into the fd for the file (may be a pipe in future).
16
17 This double buffering provides fairly substantial transfer rates,
18 compared to wget the http method is about 4% faster. Most importantly,
19 when HTTP is compared with FTP as a protocol the speed difference is
20 huge. In tests over the internet from two sites to llug (via ATM) this
21 program got 230k/s sustained http transfer rates. FTP on the other
22 hand topped out at 170k/s. That combined with the time to setup the
23 FTP connection makes HTTP a vastly superior protocol.
24
25 ##################################################################### */
26 /*}}}*/
27 // Include Files /*{{{*/
28 #include <config.h>
29
30 #include <apt-pkg/fileutl.h>
31 #include <apt-pkg/acquire-method.h>
32 #include <apt-pkg/configuration.h>
33 #include <apt-pkg/error.h>
34 #include <apt-pkg/hashes.h>
35 #include <apt-pkg/netrc.h>
36
37 #include <sys/stat.h>
38 #include <sys/time.h>
39 #include <utime.h>
40 #include <unistd.h>
41 #include <signal.h>
42 #include <stdio.h>
43 #include <errno.h>
44 #include <string.h>
45 #include <climits>
46 #include <iostream>
47 #include <map>
48
49 // Internet stuff
50 #include <netdb.h>
51
52 #include "config.h"
53 #include "connect.h"
54 #include "rfc2553emu.h"
55 #include "http.h"
56
57 #include <apti18n.h>
58 /*}}}*/
59 using namespace std;
60
61 string HttpMethod::FailFile;
62 int HttpMethod::FailFd = -1;
63 time_t HttpMethod::FailTime = 0;
64 unsigned long PipelineDepth = 0;
65 unsigned long TimeOut = 120;
66 bool AllowRedirect = false;
67 bool Debug = false;
68 URI Proxy;
69
70 unsigned long long CircleBuf::BwReadLimit=0;
71 unsigned long long CircleBuf::BwTickReadData=0;
72 struct timeval CircleBuf::BwReadTick={0,0};
73 const unsigned int CircleBuf::BW_HZ=10;
74
75 // CircleBuf::CircleBuf - Circular input buffer /*{{{*/
76 // ---------------------------------------------------------------------
77 /* */
78 CircleBuf::CircleBuf(unsigned long long Size) : Size(Size), Hash(0)
79 {
80 Buf = new unsigned char[Size];
81 Reset();
82
83 CircleBuf::BwReadLimit = _config->FindI("Acquire::http::Dl-Limit",0)*1024;
84 }
85 /*}}}*/
86 // CircleBuf::Reset - Reset to the default state /*{{{*/
87 // ---------------------------------------------------------------------
88 /* */
89 void CircleBuf::Reset()
90 {
91 InP = 0;
92 OutP = 0;
93 StrPos = 0;
94 MaxGet = (unsigned long long)-1;
95 OutQueue = string();
96 if (Hash != 0)
97 {
98 delete Hash;
99 Hash = new Hashes;
100 }
101 };
102 /*}}}*/
103 // CircleBuf::Read - Read from a FD into the circular buffer /*{{{*/
104 // ---------------------------------------------------------------------
105 /* This fills up the buffer with as much data as is in the FD, assuming it
106 is non-blocking.. */
107 bool CircleBuf::Read(int Fd)
108 {
109 unsigned long long BwReadMax;
110
111 while (1)
112 {
113 // Woops, buffer is full
114 if (InP - OutP == Size)
115 return true;
116
117 // what's left to read in this tick
118 BwReadMax = CircleBuf::BwReadLimit/BW_HZ;
119
120 if(CircleBuf::BwReadLimit) {
121 struct timeval now;
122 gettimeofday(&now,0);
123
124 unsigned long long d = (now.tv_sec-CircleBuf::BwReadTick.tv_sec)*1000000 +
125 now.tv_usec-CircleBuf::BwReadTick.tv_usec;
126 if(d > 1000000/BW_HZ) {
127 CircleBuf::BwReadTick = now;
128 CircleBuf::BwTickReadData = 0;
129 }
130
131 if(CircleBuf::BwTickReadData >= BwReadMax) {
132 usleep(1000000/BW_HZ);
133 return true;
134 }
135 }
136
137 // Write the buffer segment
138 ssize_t Res;
139 if(CircleBuf::BwReadLimit) {
140 Res = read(Fd,Buf + (InP%Size),
141 BwReadMax > LeftRead() ? LeftRead() : BwReadMax);
142 } else
143 Res = read(Fd,Buf + (InP%Size),LeftRead());
144
145 if(Res > 0 && BwReadLimit > 0)
146 CircleBuf::BwTickReadData += Res;
147
148 if (Res == 0)
149 return false;
150 if (Res < 0)
151 {
152 if (errno == EAGAIN)
153 return true;
154 return false;
155 }
156
157 if (InP == 0)
158 gettimeofday(&Start,0);
159 InP += Res;
160 }
161 }
162 /*}}}*/
163 // CircleBuf::Read - Put the string into the buffer /*{{{*/
164 // ---------------------------------------------------------------------
165 /* This will hold the string in and fill the buffer with it as it empties */
166 bool CircleBuf::Read(string Data)
167 {
168 OutQueue += Data;
169 FillOut();
170 return true;
171 }
172 /*}}}*/
173 // CircleBuf::FillOut - Fill the buffer from the output queue /*{{{*/
174 // ---------------------------------------------------------------------
175 /* */
176 void CircleBuf::FillOut()
177 {
178 if (OutQueue.empty() == true)
179 return;
180 while (1)
181 {
182 // Woops, buffer is full
183 if (InP - OutP == Size)
184 return;
185
186 // Write the buffer segment
187 unsigned long long Sz = LeftRead();
188 if (OutQueue.length() - StrPos < Sz)
189 Sz = OutQueue.length() - StrPos;
190 memcpy(Buf + (InP%Size),OutQueue.c_str() + StrPos,Sz);
191
192 // Advance
193 StrPos += Sz;
194 InP += Sz;
195 if (OutQueue.length() == StrPos)
196 {
197 StrPos = 0;
198 OutQueue = "";
199 return;
200 }
201 }
202 }
203 /*}}}*/
204 // CircleBuf::Write - Write from the buffer into a FD /*{{{*/
205 // ---------------------------------------------------------------------
206 /* This empties the buffer into the FD. */
207 bool CircleBuf::Write(int Fd)
208 {
209 while (1)
210 {
211 FillOut();
212
213 // Woops, buffer is empty
214 if (OutP == InP)
215 return true;
216
217 if (OutP == MaxGet)
218 return true;
219
220 // Write the buffer segment
221 ssize_t Res;
222 Res = write(Fd,Buf + (OutP%Size),LeftWrite());
223
224 if (Res == 0)
225 return false;
226 if (Res < 0)
227 {
228 if (errno == EAGAIN)
229 return true;
230
231 return false;
232 }
233
234 if (Hash != 0)
235 Hash->Add(Buf + (OutP%Size),Res);
236
237 OutP += Res;
238 }
239 }
240 /*}}}*/
241 // CircleBuf::WriteTillEl - Write from the buffer to a string /*{{{*/
242 // ---------------------------------------------------------------------
243 /* This copies till the first empty line */
244 bool CircleBuf::WriteTillEl(string &Data,bool Single)
245 {
246 // We cheat and assume it is unneeded to have more than one buffer load
247 for (unsigned long long I = OutP; I < InP; I++)
248 {
249 if (Buf[I%Size] != '\n')
250 continue;
251 ++I;
252
253 if (Single == false)
254 {
255 if (I < InP && Buf[I%Size] == '\r')
256 ++I;
257 if (I >= InP || Buf[I%Size] != '\n')
258 continue;
259 ++I;
260 }
261
262 Data = "";
263 while (OutP < I)
264 {
265 unsigned long long Sz = LeftWrite();
266 if (Sz == 0)
267 return false;
268 if (I - OutP < Sz)
269 Sz = I - OutP;
270 Data += string((char *)(Buf + (OutP%Size)),Sz);
271 OutP += Sz;
272 }
273 return true;
274 }
275 return false;
276 }
277 /*}}}*/
278 // CircleBuf::Stats - Print out stats information /*{{{*/
279 // ---------------------------------------------------------------------
280 /* */
281 void CircleBuf::Stats()
282 {
283 if (InP == 0)
284 return;
285
286 struct timeval Stop;
287 gettimeofday(&Stop,0);
288 /* float Diff = Stop.tv_sec - Start.tv_sec +
289 (float)(Stop.tv_usec - Start.tv_usec)/1000000;
290 clog << "Got " << InP << " in " << Diff << " at " << InP/Diff << endl;*/
291 }
292 /*}}}*/
293 CircleBuf::~CircleBuf()
294 {
295 delete [] Buf;
296 delete Hash;
297 }
298
299 // ServerState::ServerState - Constructor /*{{{*/
300 // ---------------------------------------------------------------------
301 /* */
302 ServerState::ServerState(URI Srv,HttpMethod *Owner) : Owner(Owner),
303 In(64*1024), Out(4*1024),
304 ServerName(Srv)
305 {
306 Reset();
307 }
308 /*}}}*/
309 // ServerState::Open - Open a connection to the server /*{{{*/
310 // ---------------------------------------------------------------------
311 /* This opens a connection to the server. */
312 bool ServerState::Open()
313 {
314 // Use the already open connection if possible.
315 if (ServerFd != -1)
316 return true;
317
318 Close();
319 In.Reset();
320 Out.Reset();
321 Persistent = true;
322
323 // Determine the proxy setting
324 string SpecificProxy = _config->Find("Acquire::http::Proxy::" + ServerName.Host);
325 if (!SpecificProxy.empty())
326 {
327 if (SpecificProxy == "DIRECT")
328 Proxy = "";
329 else
330 Proxy = SpecificProxy;
331 }
332 else
333 {
334 string DefProxy = _config->Find("Acquire::http::Proxy");
335 if (!DefProxy.empty())
336 {
337 Proxy = DefProxy;
338 }
339 else
340 {
341 char* result = getenv("http_proxy");
342 Proxy = result ? result : "";
343 }
344 }
345
346 // Parse no_proxy, a , separated list of domains
347 if (getenv("no_proxy") != 0)
348 {
349 if (CheckDomainList(ServerName.Host,getenv("no_proxy")) == true)
350 Proxy = "";
351 }
352
353 // Determine what host and port to use based on the proxy settings
354 int Port = 0;
355 string Host;
356 if (Proxy.empty() == true || Proxy.Host.empty() == true)
357 {
358 if (ServerName.Port != 0)
359 Port = ServerName.Port;
360 Host = ServerName.Host;
361 }
362 else
363 {
364 if (Proxy.Port != 0)
365 Port = Proxy.Port;
366 Host = Proxy.Host;
367 }
368
369 // Connect to the remote server
370 if (Connect(Host,Port,"http",80,ServerFd,TimeOut,Owner) == false)
371 return false;
372
373 return true;
374 }
375 /*}}}*/
376 // ServerState::Close - Close a connection to the server /*{{{*/
377 // ---------------------------------------------------------------------
378 /* */
379 bool ServerState::Close()
380 {
381 close(ServerFd);
382 ServerFd = -1;
383 return true;
384 }
385 /*}}}*/
386 // ServerState::RunHeaders - Get the headers before the data /*{{{*/
387 // ---------------------------------------------------------------------
388 /* Returns 0 if things are OK, 1 if an IO error occurred and 2 if a header
389 parse error occurred */
390 ServerState::RunHeadersResult ServerState::RunHeaders()
391 {
392 State = Header;
393
394 Owner->Status(_("Waiting for headers"));
395
396 Major = 0;
397 Minor = 0;
398 Result = 0;
399 Size = 0;
400 StartPos = 0;
401 Encoding = Closes;
402 HaveContent = false;
403 time(&Date);
404
405 do
406 {
407 string Data;
408 if (In.WriteTillEl(Data) == false)
409 continue;
410
411 if (Debug == true)
412 clog << Data;
413
414 for (string::const_iterator I = Data.begin(); I < Data.end(); ++I)
415 {
416 string::const_iterator J = I;
417 for (; J != Data.end() && *J != '\n' && *J != '\r'; ++J);
418 if (HeaderLine(string(I,J)) == false)
419 return RUN_HEADERS_PARSE_ERROR;
420 I = J;
421 }
422
423 // 100 Continue is a Nop...
424 if (Result == 100)
425 continue;
426
427 // Tidy up the connection persistance state.
428 if (Encoding == Closes && HaveContent == true)
429 Persistent = false;
430
431 return RUN_HEADERS_OK;
432 }
433 while (Owner->Go(false,this) == true);
434
435 return RUN_HEADERS_IO_ERROR;
436 }
437 /*}}}*/
438 // ServerState::RunData - Transfer the data from the socket /*{{{*/
439 // ---------------------------------------------------------------------
440 /* */
441 bool ServerState::RunData()
442 {
443 State = Data;
444
445 // Chunked transfer encoding is fun..
446 if (Encoding == Chunked)
447 {
448 while (1)
449 {
450 // Grab the block size
451 bool Last = true;
452 string Data;
453 In.Limit(-1);
454 do
455 {
456 if (In.WriteTillEl(Data,true) == true)
457 break;
458 }
459 while ((Last = Owner->Go(false,this)) == true);
460
461 if (Last == false)
462 return false;
463
464 // See if we are done
465 unsigned long long Len = strtoull(Data.c_str(),0,16);
466 if (Len == 0)
467 {
468 In.Limit(-1);
469
470 // We have to remove the entity trailer
471 Last = true;
472 do
473 {
474 if (In.WriteTillEl(Data,true) == true && Data.length() <= 2)
475 break;
476 }
477 while ((Last = Owner->Go(false,this)) == true);
478 if (Last == false)
479 return false;
480 return !_error->PendingError();
481 }
482
483 // Transfer the block
484 In.Limit(Len);
485 while (Owner->Go(true,this) == true)
486 if (In.IsLimit() == true)
487 break;
488
489 // Error
490 if (In.IsLimit() == false)
491 return false;
492
493 // The server sends an extra new line before the next block specifier..
494 In.Limit(-1);
495 Last = true;
496 do
497 {
498 if (In.WriteTillEl(Data,true) == true)
499 break;
500 }
501 while ((Last = Owner->Go(false,this)) == true);
502 if (Last == false)
503 return false;
504 }
505 }
506 else
507 {
508 /* Closes encoding is used when the server did not specify a size, the
509 loss of the connection means we are done */
510 if (Encoding == Closes)
511 In.Limit(-1);
512 else
513 In.Limit(Size - StartPos);
514
515 // Just transfer the whole block.
516 do
517 {
518 if (In.IsLimit() == false)
519 continue;
520
521 In.Limit(-1);
522 return !_error->PendingError();
523 }
524 while (Owner->Go(true,this) == true);
525 }
526
527 return Owner->Flush(this) && !_error->PendingError();
528 }
529 /*}}}*/
530 // ServerState::HeaderLine - Process a header line /*{{{*/
531 // ---------------------------------------------------------------------
532 /* */
533 bool ServerState::HeaderLine(string Line)
534 {
535 if (Line.empty() == true)
536 return true;
537
538 string::size_type Pos = Line.find(' ');
539 if (Pos == string::npos || Pos+1 > Line.length())
540 {
541 // Blah, some servers use "connection:closes", evil.
542 Pos = Line.find(':');
543 if (Pos == string::npos || Pos + 2 > Line.length())
544 return _error->Error(_("Bad header line"));
545 Pos++;
546 }
547
548 // Parse off any trailing spaces between the : and the next word.
549 string::size_type Pos2 = Pos;
550 while (Pos2 < Line.length() && isspace(Line[Pos2]) != 0)
551 Pos2++;
552
553 string Tag = string(Line,0,Pos);
554 string Val = string(Line,Pos2);
555
556 if (stringcasecmp(Tag.c_str(),Tag.c_str()+4,"HTTP") == 0)
557 {
558 // Evil servers return no version
559 if (Line[4] == '/')
560 {
561 int const elements = sscanf(Line.c_str(),"HTTP/%3u.%3u %3u%359[^\n]",&Major,&Minor,&Result,Code);
562 if (elements == 3)
563 {
564 Code[0] = '\0';
565 if (Debug == true)
566 clog << "HTTP server doesn't give Reason-Phrase for " << Result << std::endl;
567 }
568 else if (elements != 4)
569 return _error->Error(_("The HTTP server sent an invalid reply header"));
570 }
571 else
572 {
573 Major = 0;
574 Minor = 9;
575 if (sscanf(Line.c_str(),"HTTP %3u%359[^\n]",&Result,Code) != 2)
576 return _error->Error(_("The HTTP server sent an invalid reply header"));
577 }
578
579 /* Check the HTTP response header to get the default persistance
580 state. */
581 if (Major < 1)
582 Persistent = false;
583 else
584 {
585 if (Major == 1 && Minor == 0)
586 Persistent = false;
587 else
588 Persistent = true;
589 }
590
591 return true;
592 }
593
594 if (stringcasecmp(Tag,"Content-Length:") == 0)
595 {
596 if (Encoding == Closes)
597 Encoding = Stream;
598 HaveContent = true;
599
600 // The length is already set from the Content-Range header
601 if (StartPos != 0)
602 return true;
603
604 Size = strtoull(Val.c_str(), NULL, 10);
605 if (Size >= std::numeric_limits<unsigned long long>::max())
606 return _error->Errno("HeaderLine", _("The HTTP server sent an invalid Content-Length header"));
607 return true;
608 }
609
610 if (stringcasecmp(Tag,"Content-Type:") == 0)
611 {
612 HaveContent = true;
613 return true;
614 }
615
616 if (stringcasecmp(Tag,"Content-Range:") == 0)
617 {
618 HaveContent = true;
619
620 if (sscanf(Val.c_str(),"bytes %llu-%*u/%llu",&StartPos,&Size) != 2)
621 return _error->Error(_("The HTTP server sent an invalid Content-Range header"));
622 if ((unsigned long long)StartPos > Size)
623 return _error->Error(_("This HTTP server has broken range support"));
624 return true;
625 }
626
627 if (stringcasecmp(Tag,"Transfer-Encoding:") == 0)
628 {
629 HaveContent = true;
630 if (stringcasecmp(Val,"chunked") == 0)
631 Encoding = Chunked;
632 return true;
633 }
634
635 if (stringcasecmp(Tag,"Connection:") == 0)
636 {
637 if (stringcasecmp(Val,"close") == 0)
638 Persistent = false;
639 if (stringcasecmp(Val,"keep-alive") == 0)
640 Persistent = true;
641 return true;
642 }
643
644 if (stringcasecmp(Tag,"Last-Modified:") == 0)
645 {
646 if (RFC1123StrToTime(Val.c_str(), Date) == false)
647 return _error->Error(_("Unknown date format"));
648 return true;
649 }
650
651 if (stringcasecmp(Tag,"Location:") == 0)
652 {
653 Location = Val;
654 return true;
655 }
656
657 return true;
658 }
659 /*}}}*/
660
661 // HttpMethod::SendReq - Send the HTTP request /*{{{*/
662 // ---------------------------------------------------------------------
663 /* This places the http request in the outbound buffer */
664 void HttpMethod::SendReq(FetchItem *Itm,CircleBuf &Out)
665 {
666 URI Uri = Itm->Uri;
667
668 // The HTTP server expects a hostname with a trailing :port
669 char Buf[1000];
670 string ProperHost = Uri.Host;
671 if (Uri.Port != 0)
672 {
673 sprintf(Buf,":%u",Uri.Port);
674 ProperHost += Buf;
675 }
676
677 // Just in case.
678 if (Itm->Uri.length() >= sizeof(Buf))
679 abort();
680
681 /* Build the request. We include a keep-alive header only for non-proxy
682 requests. This is to tweak old http/1.0 servers that do support keep-alive
683 but not HTTP/1.1 automatic keep-alive. Doing this with a proxy server
684 will glitch HTTP/1.0 proxies because they do not filter it out and
685 pass it on, HTTP/1.1 says the connection should default to keep alive
686 and we expect the proxy to do this */
687 if (Proxy.empty() == true || Proxy.Host.empty())
688 {
689 // see LP bugs #1003633 and #1086997. The "+" is encoded as a workaround
690 // for a amazon S3 bug
691 sprintf(Buf,"GET %s HTTP/1.1\r\nHost: %s\r\nConnection: keep-alive\r\n",
692 QuoteString(Uri.Path,"+~ ").c_str(),ProperHost.c_str());
693 }
694 else
695 {
696 /* Generate a cache control header if necessary. We place a max
697 cache age on index files, optionally set a no-cache directive
698 and a no-store directive for archives. */
699 sprintf(Buf,"GET %s HTTP/1.1\r\nHost: %s\r\n",
700 Itm->Uri.c_str(),ProperHost.c_str());
701 }
702 // generate a cache control header (if needed)
703 if (_config->FindB("Acquire::http::No-Cache",false) == true)
704 {
705 strcat(Buf,"Cache-Control: no-cache\r\nPragma: no-cache\r\n");
706 }
707 else
708 {
709 if (Itm->IndexFile == true)
710 {
711 sprintf(Buf+strlen(Buf),"Cache-Control: max-age=%u\r\n",
712 _config->FindI("Acquire::http::Max-Age",0));
713 }
714 else
715 {
716 if (_config->FindB("Acquire::http::No-Store",false) == true)
717 strcat(Buf,"Cache-Control: no-store\r\n");
718 }
719 }
720
721 // If we ask for uncompressed files servers might respond with content-
722 // negotation which lets us end up with compressed files we do not support,
723 // see 657029, 657560 and co, so if we have no extension on the request
724 // ask for text only. As a sidenote: If there is nothing to negotate servers
725 // seem to be nice and ignore it.
726 if (_config->FindB("Acquire::http::SendAccept", true) == true)
727 {
728 size_t const filepos = Itm->Uri.find_last_of('/');
729 string const file = Itm->Uri.substr(filepos + 1);
730 if (flExtension(file) == file)
731 strcat(Buf,"Accept: text/*\r\n");
732 }
733
734 string Req = Buf;
735
736 // Check for a partial file
737 struct stat SBuf;
738 if (stat(Itm->DestFile.c_str(),&SBuf) >= 0 && SBuf.st_size > 0)
739 {
740 // In this case we send an if-range query with a range header
741 sprintf(Buf,"Range: bytes=%lli-\r\nIf-Range: %s\r\n",(long long)SBuf.st_size - 1,
742 TimeRFC1123(SBuf.st_mtime).c_str());
743 Req += Buf;
744 }
745 else
746 {
747 if (Itm->LastModified != 0)
748 {
749 sprintf(Buf,"If-Modified-Since: %s\r\n",TimeRFC1123(Itm->LastModified).c_str());
750 Req += Buf;
751 }
752 }
753
754 if (Proxy.User.empty() == false || Proxy.Password.empty() == false)
755 Req += string("Proxy-Authorization: Basic ") +
756 Base64Encode(Proxy.User + ":" + Proxy.Password) + "\r\n";
757
758 maybe_add_auth (Uri, _config->FindFile("Dir::Etc::netrc"));
759 if (Uri.User.empty() == false || Uri.Password.empty() == false)
760 {
761 Req += string("Authorization: Basic ") +
762 Base64Encode(Uri.User + ":" + Uri.Password) + "\r\n";
763 }
764 Req += "User-Agent: " + _config->Find("Acquire::http::User-Agent",
765 "Debian APT-HTTP/1.3 (" PACKAGE_VERSION ")") + "\r\n\r\n";
766
767 if (Debug == true)
768 cerr << Req << endl;
769
770 Out.Read(Req);
771 }
772 /*}}}*/
773 // HttpMethod::Go - Run a single loop /*{{{*/
774 // ---------------------------------------------------------------------
775 /* This runs the select loop over the server FDs, Output file FDs and
776 stdin. */
777 bool HttpMethod::Go(bool ToFile,ServerState *Srv)
778 {
779 // Server has closed the connection
780 if (Srv->ServerFd == -1 && (Srv->In.WriteSpace() == false ||
781 ToFile == false))
782 return false;
783
784 fd_set rfds,wfds;
785 FD_ZERO(&rfds);
786 FD_ZERO(&wfds);
787
788 /* Add the server. We only send more requests if the connection will
789 be persisting */
790 if (Srv->Out.WriteSpace() == true && Srv->ServerFd != -1
791 && Srv->Persistent == true)
792 FD_SET(Srv->ServerFd,&wfds);
793 if (Srv->In.ReadSpace() == true && Srv->ServerFd != -1)
794 FD_SET(Srv->ServerFd,&rfds);
795
796 // Add the file
797 int FileFD = -1;
798 if (File != 0)
799 FileFD = File->Fd();
800
801 if (Srv->In.WriteSpace() == true && ToFile == true && FileFD != -1)
802 FD_SET(FileFD,&wfds);
803
804 // Add stdin
805 if (_config->FindB("Acquire::http::DependOnSTDIN", true) == true)
806 FD_SET(STDIN_FILENO,&rfds);
807
808 // Figure out the max fd
809 int MaxFd = FileFD;
810 if (MaxFd < Srv->ServerFd)
811 MaxFd = Srv->ServerFd;
812
813 // Select
814 struct timeval tv;
815 tv.tv_sec = TimeOut;
816 tv.tv_usec = 0;
817 int Res = 0;
818 if ((Res = select(MaxFd+1,&rfds,&wfds,0,&tv)) < 0)
819 {
820 if (errno == EINTR)
821 return true;
822 return _error->Errno("select",_("Select failed"));
823 }
824
825 if (Res == 0)
826 {
827 _error->Error(_("Connection timed out"));
828 return ServerDie(Srv);
829 }
830
831 // Handle server IO
832 if (Srv->ServerFd != -1 && FD_ISSET(Srv->ServerFd,&rfds))
833 {
834 errno = 0;
835 if (Srv->In.Read(Srv->ServerFd) == false)
836 return ServerDie(Srv);
837 }
838
839 if (Srv->ServerFd != -1 && FD_ISSET(Srv->ServerFd,&wfds))
840 {
841 errno = 0;
842 if (Srv->Out.Write(Srv->ServerFd) == false)
843 return ServerDie(Srv);
844 }
845
846 // Send data to the file
847 if (FileFD != -1 && FD_ISSET(FileFD,&wfds))
848 {
849 if (Srv->In.Write(FileFD) == false)
850 return _error->Errno("write",_("Error writing to output file"));
851 }
852
853 // Handle commands from APT
854 if (FD_ISSET(STDIN_FILENO,&rfds))
855 {
856 if (Run(true) != -1)
857 exit(100);
858 }
859
860 return true;
861 }
862 /*}}}*/
863 // HttpMethod::Flush - Dump the buffer into the file /*{{{*/
864 // ---------------------------------------------------------------------
865 /* This takes the current input buffer from the Server FD and writes it
866 into the file */
867 bool HttpMethod::Flush(ServerState *Srv)
868 {
869 if (File != 0)
870 {
871 // on GNU/kFreeBSD, apt dies on /dev/null because non-blocking
872 // can't be set
873 if (File->Name() != "/dev/null")
874 SetNonBlock(File->Fd(),false);
875 if (Srv->In.WriteSpace() == false)
876 return true;
877
878 while (Srv->In.WriteSpace() == true)
879 {
880 if (Srv->In.Write(File->Fd()) == false)
881 return _error->Errno("write",_("Error writing to file"));
882 if (Srv->In.IsLimit() == true)
883 return true;
884 }
885
886 if (Srv->In.IsLimit() == true || Srv->Encoding == ServerState::Closes)
887 return true;
888 }
889 return false;
890 }
891 /*}}}*/
892 // HttpMethod::ServerDie - The server has closed the connection. /*{{{*/
893 // ---------------------------------------------------------------------
894 /* */
895 bool HttpMethod::ServerDie(ServerState *Srv)
896 {
897 unsigned int LErrno = errno;
898
899 // Dump the buffer to the file
900 if (Srv->State == ServerState::Data)
901 {
902 // on GNU/kFreeBSD, apt dies on /dev/null because non-blocking
903 // can't be set
904 if (File->Name() != "/dev/null")
905 SetNonBlock(File->Fd(),false);
906 while (Srv->In.WriteSpace() == true)
907 {
908 if (Srv->In.Write(File->Fd()) == false)
909 return _error->Errno("write",_("Error writing to the file"));
910
911 // Done
912 if (Srv->In.IsLimit() == true)
913 return true;
914 }
915 }
916
917 // See if this is because the server finished the data stream
918 if (Srv->In.IsLimit() == false && Srv->State != ServerState::Header &&
919 Srv->Encoding != ServerState::Closes)
920 {
921 Srv->Close();
922 if (LErrno == 0)
923 return _error->Error(_("Error reading from server. Remote end closed connection"));
924 errno = LErrno;
925 return _error->Errno("read",_("Error reading from server"));
926 }
927 else
928 {
929 Srv->In.Limit(-1);
930
931 // Nothing left in the buffer
932 if (Srv->In.WriteSpace() == false)
933 return false;
934
935 // We may have got multiple responses back in one packet..
936 Srv->Close();
937 return true;
938 }
939
940 return false;
941 }
942 /*}}}*/
943 // HttpMethod::DealWithHeaders - Handle the retrieved header data /*{{{*/
944 // ---------------------------------------------------------------------
945 /* We look at the header data we got back from the server and decide what
946 to do. Returns DealWithHeadersResult (see http.h for details).
947 */
948 HttpMethod::DealWithHeadersResult
949 HttpMethod::DealWithHeaders(FetchResult &Res,ServerState *Srv)
950 {
951 // Not Modified
952 if (Srv->Result == 304)
953 {
954 unlink(Queue->DestFile.c_str());
955 Res.IMSHit = true;
956 Res.LastModified = Queue->LastModified;
957 return IMS_HIT;
958 }
959
960 /* Redirect
961 *
962 * Note that it is only OK for us to treat all redirection the same
963 * because we *always* use GET, not other HTTP methods. There are
964 * three redirection codes for which it is not appropriate that we
965 * redirect. Pass on those codes so the error handling kicks in.
966 */
967 if (AllowRedirect
968 && (Srv->Result > 300 && Srv->Result < 400)
969 && (Srv->Result != 300 // Multiple Choices
970 && Srv->Result != 304 // Not Modified
971 && Srv->Result != 306)) // (Not part of HTTP/1.1, reserved)
972 {
973 if (Srv->Location.empty() == true);
974 else if (Srv->Location[0] == '/' && Queue->Uri.empty() == false)
975 {
976 URI Uri = Queue->Uri;
977 if (Uri.Host.empty() == false)
978 {
979 if (Uri.Port != 0)
980 strprintf(NextURI, "http://%s:%u", Uri.Host.c_str(), Uri.Port);
981 else
982 NextURI = "http://" + Uri.Host;
983 }
984 else
985 NextURI.clear();
986 NextURI.append(DeQuoteString(Srv->Location));
987 return TRY_AGAIN_OR_REDIRECT;
988 }
989 else
990 {
991 NextURI = DeQuoteString(Srv->Location);
992 URI tmpURI = NextURI;
993 // Do not allow a redirection to switch protocol
994 if (tmpURI.Access == "http")
995 return TRY_AGAIN_OR_REDIRECT;
996 }
997 /* else pass through for error message */
998 }
999
1000 /* We have a reply we dont handle. This should indicate a perm server
1001 failure */
1002 if (Srv->Result < 200 || Srv->Result >= 300)
1003 {
1004 char err[255];
1005 snprintf(err,sizeof(err)-1,"HttpError%i",Srv->Result);
1006 SetFailReason(err);
1007 _error->Error("%u %s",Srv->Result,Srv->Code);
1008 if (Srv->HaveContent == true)
1009 return ERROR_WITH_CONTENT_PAGE;
1010 return ERROR_UNRECOVERABLE;
1011 }
1012
1013 // This is some sort of 2xx 'data follows' reply
1014 Res.LastModified = Srv->Date;
1015 Res.Size = Srv->Size;
1016
1017 // Open the file
1018 delete File;
1019 File = new FileFd(Queue->DestFile,FileFd::WriteAny);
1020 if (_error->PendingError() == true)
1021 return ERROR_NOT_FROM_SERVER;
1022
1023 FailFile = Queue->DestFile;
1024 FailFile.c_str(); // Make sure we dont do a malloc in the signal handler
1025 FailFd = File->Fd();
1026 FailTime = Srv->Date;
1027
1028 delete Srv->In.Hash;
1029 Srv->In.Hash = new Hashes;
1030
1031 // Set the expected size and read file for the hashes
1032 if (Srv->StartPos >= 0)
1033 {
1034 Res.ResumePoint = Srv->StartPos;
1035 File->Truncate(Srv->StartPos);
1036
1037 if (Srv->In.Hash->AddFD(*File,Srv->StartPos) == false)
1038 {
1039 _error->Errno("read",_("Problem hashing file"));
1040 return ERROR_NOT_FROM_SERVER;
1041 }
1042 }
1043
1044 SetNonBlock(File->Fd(),true);
1045 return FILE_IS_OPEN;
1046 }
1047 /*}}}*/
1048 // HttpMethod::SigTerm - Handle a fatal signal /*{{{*/
1049 // ---------------------------------------------------------------------
1050 /* This closes and timestamps the open file. This is neccessary to get
1051 resume behavoir on user abort */
1052 void HttpMethod::SigTerm(int)
1053 {
1054 if (FailFd == -1)
1055 _exit(100);
1056 close(FailFd);
1057
1058 // Timestamp
1059 struct utimbuf UBuf;
1060 UBuf.actime = FailTime;
1061 UBuf.modtime = FailTime;
1062 utime(FailFile.c_str(),&UBuf);
1063
1064 _exit(100);
1065 }
1066 /*}}}*/
1067 // HttpMethod::Fetch - Fetch an item /*{{{*/
1068 // ---------------------------------------------------------------------
1069 /* This adds an item to the pipeline. We keep the pipeline at a fixed
1070 depth. */
1071 bool HttpMethod::Fetch(FetchItem *)
1072 {
1073 if (Server == 0)
1074 return true;
1075
1076 // Queue the requests
1077 int Depth = -1;
1078 for (FetchItem *I = Queue; I != 0 && Depth < (signed)PipelineDepth;
1079 I = I->Next, Depth++)
1080 {
1081 // If pipelining is disabled, we only queue 1 request
1082 if (Server->Pipeline == false && Depth >= 0)
1083 break;
1084
1085 // Make sure we stick with the same server
1086 if (Server->Comp(I->Uri) == false)
1087 break;
1088 if (QueueBack == I)
1089 {
1090 QueueBack = I->Next;
1091 SendReq(I,Server->Out);
1092 continue;
1093 }
1094 }
1095
1096 return true;
1097 };
1098 /*}}}*/
1099 // HttpMethod::Configuration - Handle a configuration message /*{{{*/
1100 // ---------------------------------------------------------------------
1101 /* We stash the desired pipeline depth */
1102 bool HttpMethod::Configuration(string Message)
1103 {
1104 if (pkgAcqMethod::Configuration(Message) == false)
1105 return false;
1106
1107 AllowRedirect = _config->FindB("Acquire::http::AllowRedirect",true);
1108 TimeOut = _config->FindI("Acquire::http::Timeout",TimeOut);
1109 PipelineDepth = _config->FindI("Acquire::http::Pipeline-Depth",
1110 PipelineDepth);
1111 Debug = _config->FindB("Debug::Acquire::http",false);
1112 AutoDetectProxyCmd = _config->Find("Acquire::http::ProxyAutoDetect");
1113
1114 // Get the proxy to use
1115 AutoDetectProxy();
1116
1117 return true;
1118 }
1119 /*}}}*/
1120 // HttpMethod::Loop - Main loop /*{{{*/
1121 // ---------------------------------------------------------------------
1122 /* */
1123 int HttpMethod::Loop()
1124 {
1125 typedef vector<string> StringVector;
1126 typedef vector<string>::iterator StringVectorIterator;
1127 map<string, StringVector> Redirected;
1128
1129 signal(SIGTERM,SigTerm);
1130 signal(SIGINT,SigTerm);
1131
1132 Server = 0;
1133
1134 int FailCounter = 0;
1135 while (1)
1136 {
1137 // We have no commands, wait for some to arrive
1138 if (Queue == 0)
1139 {
1140 if (WaitFd(STDIN_FILENO) == false)
1141 return 0;
1142 }
1143
1144 /* Run messages, we can accept 0 (no message) if we didn't
1145 do a WaitFd above.. Otherwise the FD is closed. */
1146 int Result = Run(true);
1147 if (Result != -1 && (Result != 0 || Queue == 0))
1148 {
1149 if(FailReason.empty() == false ||
1150 _config->FindB("Acquire::http::DependOnSTDIN", true) == true)
1151 return 100;
1152 else
1153 return 0;
1154 }
1155
1156 if (Queue == 0)
1157 continue;
1158
1159 // Connect to the server
1160 if (Server == 0 || Server->Comp(Queue->Uri) == false)
1161 {
1162 delete Server;
1163 Server = new ServerState(Queue->Uri,this);
1164 }
1165 /* If the server has explicitly said this is the last connection
1166 then we pre-emptively shut down the pipeline and tear down
1167 the connection. This will speed up HTTP/1.0 servers a tad
1168 since we don't have to wait for the close sequence to
1169 complete */
1170 if (Server->Persistent == false)
1171 Server->Close();
1172
1173 // Reset the pipeline
1174 if (Server->ServerFd == -1)
1175 QueueBack = Queue;
1176
1177 // Connnect to the host
1178 if (Server->Open() == false)
1179 {
1180 Fail(true);
1181 delete Server;
1182 Server = 0;
1183 continue;
1184 }
1185
1186 // Fill the pipeline.
1187 Fetch(0);
1188
1189 // Fetch the next URL header data from the server.
1190 switch (Server->RunHeaders())
1191 {
1192 case ServerState::RUN_HEADERS_OK:
1193 break;
1194
1195 // The header data is bad
1196 case ServerState::RUN_HEADERS_PARSE_ERROR:
1197 {
1198 _error->Error(_("Bad header data"));
1199 Fail(true);
1200 RotateDNS();
1201 continue;
1202 }
1203
1204 // The server closed a connection during the header get..
1205 default:
1206 case ServerState::RUN_HEADERS_IO_ERROR:
1207 {
1208 FailCounter++;
1209 _error->Discard();
1210 Server->Close();
1211 Server->Pipeline = false;
1212
1213 if (FailCounter >= 2)
1214 {
1215 Fail(_("Connection failed"),true);
1216 FailCounter = 0;
1217 }
1218
1219 RotateDNS();
1220 continue;
1221 }
1222 };
1223
1224 // Decide what to do.
1225 FetchResult Res;
1226 Res.Filename = Queue->DestFile;
1227 switch (DealWithHeaders(Res,Server))
1228 {
1229 // Ok, the file is Open
1230 case FILE_IS_OPEN:
1231 {
1232 URIStart(Res);
1233
1234 // Run the data
1235 bool Result = Server->RunData();
1236
1237 /* If the server is sending back sizeless responses then fill in
1238 the size now */
1239 if (Res.Size == 0)
1240 Res.Size = File->Size();
1241
1242 // Close the file, destroy the FD object and timestamp it
1243 FailFd = -1;
1244 delete File;
1245 File = 0;
1246
1247 // Timestamp
1248 struct utimbuf UBuf;
1249 time(&UBuf.actime);
1250 UBuf.actime = Server->Date;
1251 UBuf.modtime = Server->Date;
1252 utime(Queue->DestFile.c_str(),&UBuf);
1253
1254 // Send status to APT
1255 if (Result == true)
1256 {
1257 Res.TakeHashes(*Server->In.Hash);
1258 URIDone(Res);
1259 }
1260 else
1261 {
1262 if (Server->ServerFd == -1)
1263 {
1264 FailCounter++;
1265 _error->Discard();
1266 Server->Close();
1267
1268 if (FailCounter >= 2)
1269 {
1270 Fail(_("Connection failed"),true);
1271 FailCounter = 0;
1272 }
1273
1274 QueueBack = Queue;
1275 }
1276 else
1277 Fail(true);
1278 }
1279 break;
1280 }
1281
1282 // IMS hit
1283 case IMS_HIT:
1284 {
1285 URIDone(Res);
1286 break;
1287 }
1288
1289 // Hard server error, not found or something
1290 case ERROR_UNRECOVERABLE:
1291 {
1292 Fail();
1293 break;
1294 }
1295
1296 // Hard internal error, kill the connection and fail
1297 case ERROR_NOT_FROM_SERVER:
1298 {
1299 delete File;
1300 File = 0;
1301
1302 Fail();
1303 RotateDNS();
1304 Server->Close();
1305 break;
1306 }
1307
1308 // We need to flush the data, the header is like a 404 w/ error text
1309 case ERROR_WITH_CONTENT_PAGE:
1310 {
1311 Fail();
1312
1313 // Send to content to dev/null
1314 File = new FileFd("/dev/null",FileFd::WriteExists);
1315 Server->RunData();
1316 delete File;
1317 File = 0;
1318 break;
1319 }
1320
1321 // Try again with a new URL
1322 case TRY_AGAIN_OR_REDIRECT:
1323 {
1324 // Clear rest of response if there is content
1325 if (Server->HaveContent)
1326 {
1327 File = new FileFd("/dev/null",FileFd::WriteExists);
1328 Server->RunData();
1329 delete File;
1330 File = 0;
1331 }
1332
1333 /* Detect redirect loops. No more redirects are allowed
1334 after the same URI is seen twice in a queue item. */
1335 StringVector &R = Redirected[Queue->DestFile];
1336 bool StopRedirects = false;
1337 if (R.empty() == true)
1338 R.push_back(Queue->Uri);
1339 else if (R[0] == "STOP" || R.size() > 10)
1340 StopRedirects = true;
1341 else
1342 {
1343 for (StringVectorIterator I = R.begin(); I != R.end(); ++I)
1344 if (Queue->Uri == *I)
1345 {
1346 R[0] = "STOP";
1347 break;
1348 }
1349
1350 R.push_back(Queue->Uri);
1351 }
1352
1353 if (StopRedirects == false)
1354 Redirect(NextURI);
1355 else
1356 Fail();
1357
1358 break;
1359 }
1360
1361 default:
1362 Fail(_("Internal error"));
1363 break;
1364 }
1365
1366 FailCounter = 0;
1367 }
1368
1369 return 0;
1370 }
1371 /*}}}*/
1372 // HttpMethod::AutoDetectProxy - auto detect proxy /*{{{*/
1373 // ---------------------------------------------------------------------
1374 /* */
1375 bool HttpMethod::AutoDetectProxy()
1376 {
1377 if (AutoDetectProxyCmd.empty())
1378 return true;
1379
1380 if (Debug)
1381 clog << "Using auto proxy detect command: " << AutoDetectProxyCmd << endl;
1382
1383 int Pipes[2] = {-1,-1};
1384 if (pipe(Pipes) != 0)
1385 return _error->Errno("pipe", "Failed to create Pipe");
1386
1387 pid_t Process = ExecFork();
1388 if (Process == 0)
1389 {
1390 close(Pipes[0]);
1391 dup2(Pipes[1],STDOUT_FILENO);
1392 SetCloseExec(STDOUT_FILENO,false);
1393
1394 const char *Args[2];
1395 Args[0] = AutoDetectProxyCmd.c_str();
1396 Args[1] = 0;
1397 execv(Args[0],(char **)Args);
1398 cerr << "Failed to exec method " << Args[0] << endl;
1399 _exit(100);
1400 }
1401 char buf[512];
1402 int InFd = Pipes[0];
1403 close(Pipes[1]);
1404 int res = read(InFd, buf, sizeof(buf));
1405 ExecWait(Process, "ProxyAutoDetect", true);
1406
1407 if (res < 0)
1408 return _error->Errno("read", "Failed to read");
1409 if (res == 0)
1410 return _error->Warning("ProxyAutoDetect returned no data");
1411
1412 // add trailing \0
1413 buf[res] = 0;
1414
1415 if (Debug)
1416 clog << "auto detect command returned: '" << buf << "'" << endl;
1417
1418 if (strstr(buf, "http://") == buf)
1419 _config->Set("Acquire::http::proxy", _strstrip(buf));
1420
1421 return true;
1422 }
1423 /*}}}*/
1424
1425