]> git.saurik.com Git - apt.git/blob - methods/http.cc
e505b816eb5d2afa58f75596d7099920d0e78a29
[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/error.h>
33 #include <apt-pkg/hashes.h>
34 #include <apt-pkg/netrc.h>
35
36 #include <sys/stat.h>
37 #include <sys/time.h>
38 #include <utime.h>
39 #include <unistd.h>
40 #include <signal.h>
41 #include <stdio.h>
42 #include <errno.h>
43 #include <string.h>
44 #include <iostream>
45 #include <map>
46
47 // Internet stuff
48 #include <netdb.h>
49
50 #include "config.h"
51 #include "connect.h"
52 #include "rfc2553emu.h"
53 #include "http.h"
54
55 #include <apti18n.h>
56 /*}}}*/
57 using namespace std;
58
59 string HttpMethod::FailFile;
60 int HttpMethod::FailFd = -1;
61 time_t HttpMethod::FailTime = 0;
62 unsigned long PipelineDepth = 10;
63 unsigned long TimeOut = 120;
64 bool AllowRedirect = false;
65 bool Debug = false;
66 URI Proxy;
67
68 unsigned long CircleBuf::BwReadLimit=0;
69 unsigned long CircleBuf::BwTickReadData=0;
70 struct timeval CircleBuf::BwReadTick={0,0};
71 const unsigned int CircleBuf::BW_HZ=10;
72
73 // CircleBuf::CircleBuf - Circular input buffer /*{{{*/
74 // ---------------------------------------------------------------------
75 /* */
76 CircleBuf::CircleBuf(unsigned long Size) : Size(Size), Hash(0)
77 {
78 Buf = new unsigned char[Size];
79 Reset();
80
81 CircleBuf::BwReadLimit = _config->FindI("Acquire::http::Dl-Limit",0)*1024;
82 }
83 /*}}}*/
84 // CircleBuf::Reset - Reset to the default state /*{{{*/
85 // ---------------------------------------------------------------------
86 /* */
87 void CircleBuf::Reset()
88 {
89 InP = 0;
90 OutP = 0;
91 StrPos = 0;
92 MaxGet = (unsigned int)-1;
93 OutQueue = string();
94 if (Hash != 0)
95 {
96 delete Hash;
97 Hash = new Hashes;
98 }
99 };
100 /*}}}*/
101 // CircleBuf::Read - Read from a FD into the circular buffer /*{{{*/
102 // ---------------------------------------------------------------------
103 /* This fills up the buffer with as much data as is in the FD, assuming it
104 is non-blocking.. */
105 bool CircleBuf::Read(int Fd)
106 {
107 unsigned long BwReadMax;
108
109 while (1)
110 {
111 // Woops, buffer is full
112 if (InP - OutP == Size)
113 return true;
114
115 // what's left to read in this tick
116 BwReadMax = CircleBuf::BwReadLimit/BW_HZ;
117
118 if(CircleBuf::BwReadLimit) {
119 struct timeval now;
120 gettimeofday(&now,0);
121
122 unsigned long d = (now.tv_sec-CircleBuf::BwReadTick.tv_sec)*1000000 +
123 now.tv_usec-CircleBuf::BwReadTick.tv_usec;
124 if(d > 1000000/BW_HZ) {
125 CircleBuf::BwReadTick = now;
126 CircleBuf::BwTickReadData = 0;
127 }
128
129 if(CircleBuf::BwTickReadData >= BwReadMax) {
130 usleep(1000000/BW_HZ);
131 return true;
132 }
133 }
134
135 // Write the buffer segment
136 int Res;
137 if(CircleBuf::BwReadLimit) {
138 Res = read(Fd,Buf + (InP%Size),
139 BwReadMax > LeftRead() ? LeftRead() : BwReadMax);
140 } else
141 Res = read(Fd,Buf + (InP%Size),LeftRead());
142
143 if(Res > 0 && BwReadLimit > 0)
144 CircleBuf::BwTickReadData += Res;
145
146 if (Res == 0)
147 return false;
148 if (Res < 0)
149 {
150 if (errno == EAGAIN)
151 return true;
152 return false;
153 }
154
155 if (InP == 0)
156 gettimeofday(&Start,0);
157 InP += Res;
158 }
159 }
160 /*}}}*/
161 // CircleBuf::Read - Put the string into the buffer /*{{{*/
162 // ---------------------------------------------------------------------
163 /* This will hold the string in and fill the buffer with it as it empties */
164 bool CircleBuf::Read(string Data)
165 {
166 OutQueue += Data;
167 FillOut();
168 return true;
169 }
170 /*}}}*/
171 // CircleBuf::FillOut - Fill the buffer from the output queue /*{{{*/
172 // ---------------------------------------------------------------------
173 /* */
174 void CircleBuf::FillOut()
175 {
176 if (OutQueue.empty() == true)
177 return;
178 while (1)
179 {
180 // Woops, buffer is full
181 if (InP - OutP == Size)
182 return;
183
184 // Write the buffer segment
185 unsigned long Sz = LeftRead();
186 if (OutQueue.length() - StrPos < Sz)
187 Sz = OutQueue.length() - StrPos;
188 memcpy(Buf + (InP%Size),OutQueue.c_str() + StrPos,Sz);
189
190 // Advance
191 StrPos += Sz;
192 InP += Sz;
193 if (OutQueue.length() == StrPos)
194 {
195 StrPos = 0;
196 OutQueue = "";
197 return;
198 }
199 }
200 }
201 /*}}}*/
202 // CircleBuf::Write - Write from the buffer into a FD /*{{{*/
203 // ---------------------------------------------------------------------
204 /* This empties the buffer into the FD. */
205 bool CircleBuf::Write(int Fd)
206 {
207 while (1)
208 {
209 FillOut();
210
211 // Woops, buffer is empty
212 if (OutP == InP)
213 return true;
214
215 if (OutP == MaxGet)
216 return true;
217
218 // Write the buffer segment
219 int Res;
220 Res = write(Fd,Buf + (OutP%Size),LeftWrite());
221
222 if (Res == 0)
223 return false;
224 if (Res < 0)
225 {
226 if (errno == EAGAIN)
227 return true;
228
229 return false;
230 }
231
232 if (Hash != 0)
233 Hash->Add(Buf + (OutP%Size),Res);
234
235 OutP += Res;
236 }
237 }
238 /*}}}*/
239 // CircleBuf::WriteTillEl - Write from the buffer to a string /*{{{*/
240 // ---------------------------------------------------------------------
241 /* This copies till the first empty line */
242 bool CircleBuf::WriteTillEl(string &Data,bool Single)
243 {
244 // We cheat and assume it is unneeded to have more than one buffer load
245 for (unsigned long I = OutP; I < InP; I++)
246 {
247 if (Buf[I%Size] != '\n')
248 continue;
249 ++I;
250
251 if (Single == false)
252 {
253 if (I < InP && Buf[I%Size] == '\r')
254 ++I;
255 if (I >= InP || Buf[I%Size] != '\n')
256 continue;
257 ++I;
258 }
259
260 Data = "";
261 while (OutP < I)
262 {
263 unsigned long Sz = LeftWrite();
264 if (Sz == 0)
265 return false;
266 if (I - OutP < Sz)
267 Sz = I - OutP;
268 Data += string((char *)(Buf + (OutP%Size)),Sz);
269 OutP += Sz;
270 }
271 return true;
272 }
273 return false;
274 }
275 /*}}}*/
276 // CircleBuf::Stats - Print out stats information /*{{{*/
277 // ---------------------------------------------------------------------
278 /* */
279 void CircleBuf::Stats()
280 {
281 if (InP == 0)
282 return;
283
284 struct timeval Stop;
285 gettimeofday(&Stop,0);
286 /* float Diff = Stop.tv_sec - Start.tv_sec +
287 (float)(Stop.tv_usec - Start.tv_usec)/1000000;
288 clog << "Got " << InP << " in " << Diff << " at " << InP/Diff << endl;*/
289 }
290 /*}}}*/
291
292 // ServerState::ServerState - Constructor /*{{{*/
293 // ---------------------------------------------------------------------
294 /* */
295 ServerState::ServerState(URI Srv,HttpMethod *Owner) : Owner(Owner),
296 In(64*1024), Out(4*1024),
297 ServerName(Srv)
298 {
299 Reset();
300 }
301 /*}}}*/
302 // ServerState::Open - Open a connection to the server /*{{{*/
303 // ---------------------------------------------------------------------
304 /* This opens a connection to the server. */
305 bool ServerState::Open()
306 {
307 // Use the already open connection if possible.
308 if (ServerFd != -1)
309 return true;
310
311 Close();
312 In.Reset();
313 Out.Reset();
314 Persistent = true;
315
316 // Determine the proxy setting
317 string SpecificProxy = _config->Find("Acquire::http::Proxy::" + ServerName.Host);
318 if (!SpecificProxy.empty())
319 {
320 if (SpecificProxy == "DIRECT")
321 Proxy = "";
322 else
323 Proxy = SpecificProxy;
324 }
325 else
326 {
327 string DefProxy = _config->Find("Acquire::http::Proxy");
328 if (!DefProxy.empty())
329 {
330 Proxy = DefProxy;
331 }
332 else
333 {
334 char* result = getenv("http_proxy");
335 Proxy = result ? result : "";
336 }
337 }
338
339 // Parse no_proxy, a , separated list of domains
340 if (getenv("no_proxy") != 0)
341 {
342 if (CheckDomainList(ServerName.Host,getenv("no_proxy")) == true)
343 Proxy = "";
344 }
345
346 // Determine what host and port to use based on the proxy settings
347 int Port = 0;
348 string Host;
349 if (Proxy.empty() == true || Proxy.Host.empty() == true)
350 {
351 if (ServerName.Port != 0)
352 Port = ServerName.Port;
353 Host = ServerName.Host;
354 }
355 else
356 {
357 if (Proxy.Port != 0)
358 Port = Proxy.Port;
359 Host = Proxy.Host;
360 }
361
362 // Connect to the remote server
363 if (Connect(Host,Port,"http",80,ServerFd,TimeOut,Owner) == false)
364 return false;
365
366 return true;
367 }
368 /*}}}*/
369 // ServerState::Close - Close a connection to the server /*{{{*/
370 // ---------------------------------------------------------------------
371 /* */
372 bool ServerState::Close()
373 {
374 close(ServerFd);
375 ServerFd = -1;
376 return true;
377 }
378 /*}}}*/
379 // ServerState::RunHeaders - Get the headers before the data /*{{{*/
380 // ---------------------------------------------------------------------
381 /* Returns 0 if things are OK, 1 if an IO error occurred and 2 if a header
382 parse error occurred */
383 ServerState::RunHeadersResult ServerState::RunHeaders()
384 {
385 State = Header;
386
387 Owner->Status(_("Waiting for headers"));
388
389 Major = 0;
390 Minor = 0;
391 Result = 0;
392 Size = 0;
393 StartPos = 0;
394 Encoding = Closes;
395 HaveContent = false;
396 time(&Date);
397
398 do
399 {
400 string Data;
401 if (In.WriteTillEl(Data) == false)
402 continue;
403
404 if (Debug == true)
405 clog << Data;
406
407 for (string::const_iterator I = Data.begin(); I < Data.end(); I++)
408 {
409 string::const_iterator J = I;
410 for (; J != Data.end() && *J != '\n' && *J != '\r';J++);
411 if (HeaderLine(string(I,J)) == false)
412 return RUN_HEADERS_PARSE_ERROR;
413 I = J;
414 }
415
416 // 100 Continue is a Nop...
417 if (Result == 100)
418 continue;
419
420 // Tidy up the connection persistance state.
421 if (Encoding == Closes && HaveContent == true)
422 Persistent = false;
423
424 return RUN_HEADERS_OK;
425 }
426 while (Owner->Go(false,this) == true);
427
428 return RUN_HEADERS_IO_ERROR;
429 }
430 /*}}}*/
431 // ServerState::RunData - Transfer the data from the socket /*{{{*/
432 // ---------------------------------------------------------------------
433 /* */
434 bool ServerState::RunData()
435 {
436 State = Data;
437
438 // Chunked transfer encoding is fun..
439 if (Encoding == Chunked)
440 {
441 while (1)
442 {
443 // Grab the block size
444 bool Last = true;
445 string Data;
446 In.Limit(-1);
447 do
448 {
449 if (In.WriteTillEl(Data,true) == true)
450 break;
451 }
452 while ((Last = Owner->Go(false,this)) == true);
453
454 if (Last == false)
455 return false;
456
457 // See if we are done
458 unsigned long Len = strtol(Data.c_str(),0,16);
459 if (Len == 0)
460 {
461 In.Limit(-1);
462
463 // We have to remove the entity trailer
464 Last = true;
465 do
466 {
467 if (In.WriteTillEl(Data,true) == true && Data.length() <= 2)
468 break;
469 }
470 while ((Last = Owner->Go(false,this)) == true);
471 if (Last == false)
472 return false;
473 return !_error->PendingError();
474 }
475
476 // Transfer the block
477 In.Limit(Len);
478 while (Owner->Go(true,this) == true)
479 if (In.IsLimit() == true)
480 break;
481
482 // Error
483 if (In.IsLimit() == false)
484 return false;
485
486 // The server sends an extra new line before the next block specifier..
487 In.Limit(-1);
488 Last = true;
489 do
490 {
491 if (In.WriteTillEl(Data,true) == true)
492 break;
493 }
494 while ((Last = Owner->Go(false,this)) == true);
495 if (Last == false)
496 return false;
497 }
498 }
499 else
500 {
501 /* Closes encoding is used when the server did not specify a size, the
502 loss of the connection means we are done */
503 if (Encoding == Closes)
504 In.Limit(-1);
505 else
506 In.Limit(Size - StartPos);
507
508 // Just transfer the whole block.
509 do
510 {
511 if (In.IsLimit() == false)
512 continue;
513
514 In.Limit(-1);
515 return !_error->PendingError();
516 }
517 while (Owner->Go(true,this) == true);
518 }
519
520 return Owner->Flush(this) && !_error->PendingError();
521 }
522 /*}}}*/
523 // ServerState::HeaderLine - Process a header line /*{{{*/
524 // ---------------------------------------------------------------------
525 /* */
526 bool ServerState::HeaderLine(string Line)
527 {
528 if (Line.empty() == true)
529 return true;
530
531 // The http server might be trying to do something evil.
532 if (Line.length() >= MAXLEN)
533 return _error->Error(_("Got a single header line over %u chars"),MAXLEN);
534
535 string::size_type Pos = Line.find(' ');
536 if (Pos == string::npos || Pos+1 > Line.length())
537 {
538 // Blah, some servers use "connection:closes", evil.
539 Pos = Line.find(':');
540 if (Pos == string::npos || Pos + 2 > Line.length())
541 return _error->Error(_("Bad header line"));
542 Pos++;
543 }
544
545 // Parse off any trailing spaces between the : and the next word.
546 string::size_type Pos2 = Pos;
547 while (Pos2 < Line.length() && isspace(Line[Pos2]) != 0)
548 Pos2++;
549
550 string Tag = string(Line,0,Pos);
551 string Val = string(Line,Pos2);
552
553 if (stringcasecmp(Tag.c_str(),Tag.c_str()+4,"HTTP") == 0)
554 {
555 // Evil servers return no version
556 if (Line[4] == '/')
557 {
558 int const elements = sscanf(Line.c_str(),"HTTP/%u.%u %u%[^\n]",&Major,&Minor,&Result,Code);
559 if (elements == 3)
560 {
561 Code[0] = '\0';
562 if (Debug == true)
563 clog << "HTTP server doesn't give Reason-Phrase for " << Result << std::endl;
564 }
565 else if (elements != 4)
566 return _error->Error(_("The HTTP server sent an invalid reply header"));
567 }
568 else
569 {
570 Major = 0;
571 Minor = 9;
572 if (sscanf(Line.c_str(),"HTTP %u%[^\n]",&Result,Code) != 2)
573 return _error->Error(_("The HTTP server sent an invalid reply header"));
574 }
575
576 /* Check the HTTP response header to get the default persistance
577 state. */
578 if (Major < 1)
579 Persistent = false;
580 else
581 {
582 if (Major == 1 && Minor <= 0)
583 Persistent = false;
584 else
585 Persistent = true;
586 }
587
588 return true;
589 }
590
591 if (stringcasecmp(Tag,"Content-Length:") == 0)
592 {
593 if (Encoding == Closes)
594 Encoding = Stream;
595 HaveContent = true;
596
597 // The length is already set from the Content-Range header
598 if (StartPos != 0)
599 return true;
600
601 if (sscanf(Val.c_str(),"%lu",&Size) != 1)
602 return _error->Error(_("The HTTP server sent an invalid Content-Length header"));
603 return true;
604 }
605
606 if (stringcasecmp(Tag,"Content-Type:") == 0)
607 {
608 HaveContent = true;
609 return true;
610 }
611
612 if (stringcasecmp(Tag,"Content-Range:") == 0)
613 {
614 HaveContent = true;
615
616 if (sscanf(Val.c_str(),"bytes %lu-%*u/%lu",&StartPos,&Size) != 2)
617 return _error->Error(_("The HTTP server sent an invalid Content-Range header"));
618 if ((unsigned)StartPos > Size)
619 return _error->Error(_("This HTTP server has broken range support"));
620 return true;
621 }
622
623 if (stringcasecmp(Tag,"Transfer-Encoding:") == 0)
624 {
625 HaveContent = true;
626 if (stringcasecmp(Val,"chunked") == 0)
627 Encoding = Chunked;
628 return true;
629 }
630
631 if (stringcasecmp(Tag,"Connection:") == 0)
632 {
633 if (stringcasecmp(Val,"close") == 0)
634 Persistent = false;
635 if (stringcasecmp(Val,"keep-alive") == 0)
636 Persistent = true;
637 return true;
638 }
639
640 if (stringcasecmp(Tag,"Last-Modified:") == 0)
641 {
642 if (RFC1123StrToTime(Val.c_str(), Date) == false)
643 return _error->Error(_("Unknown date format"));
644 return true;
645 }
646
647 if (stringcasecmp(Tag,"Location:") == 0)
648 {
649 Location = Val;
650 return true;
651 }
652
653 return true;
654 }
655 /*}}}*/
656
657 // HttpMethod::SendReq - Send the HTTP request /*{{{*/
658 // ---------------------------------------------------------------------
659 /* This places the http request in the outbound buffer */
660 void HttpMethod::SendReq(FetchItem *Itm,CircleBuf &Out)
661 {
662 URI Uri = Itm->Uri;
663
664 // The HTTP server expects a hostname with a trailing :port
665 char Buf[1000];
666 string ProperHost = Uri.Host;
667 if (Uri.Port != 0)
668 {
669 sprintf(Buf,":%u",Uri.Port);
670 ProperHost += Buf;
671 }
672
673 // Just in case.
674 if (Itm->Uri.length() >= sizeof(Buf))
675 abort();
676
677 /* Build the request. We include a keep-alive header only for non-proxy
678 requests. This is to tweak old http/1.0 servers that do support keep-alive
679 but not HTTP/1.1 automatic keep-alive. Doing this with a proxy server
680 will glitch HTTP/1.0 proxies because they do not filter it out and
681 pass it on, HTTP/1.1 says the connection should default to keep alive
682 and we expect the proxy to do this */
683 if (Proxy.empty() == true || Proxy.Host.empty())
684 sprintf(Buf,"GET %s HTTP/1.1\r\nHost: %s\r\nConnection: keep-alive\r\n",
685 QuoteString(Uri.Path,"~").c_str(),ProperHost.c_str());
686 else
687 {
688 /* Generate a cache control header if necessary. We place a max
689 cache age on index files, optionally set a no-cache directive
690 and a no-store directive for archives. */
691 sprintf(Buf,"GET %s HTTP/1.1\r\nHost: %s\r\n",
692 Itm->Uri.c_str(),ProperHost.c_str());
693 }
694 // generate a cache control header (if needed)
695 if (_config->FindB("Acquire::http::No-Cache",false) == true)
696 {
697 strcat(Buf,"Cache-Control: no-cache\r\nPragma: no-cache\r\n");
698 }
699 else
700 {
701 if (Itm->IndexFile == true)
702 {
703 sprintf(Buf+strlen(Buf),"Cache-Control: max-age=%u\r\n",
704 _config->FindI("Acquire::http::Max-Age",0));
705 }
706 else
707 {
708 if (_config->FindB("Acquire::http::No-Store",false) == true)
709 strcat(Buf,"Cache-Control: no-store\r\n");
710 }
711 }
712
713
714 string Req = Buf;
715
716 // Check for a partial file
717 struct stat SBuf;
718 if (stat(Itm->DestFile.c_str(),&SBuf) >= 0 && SBuf.st_size > 0)
719 {
720 // In this case we send an if-range query with a range header
721 sprintf(Buf,"Range: bytes=%li-\r\nIf-Range: %s\r\n",(long)SBuf.st_size - 1,
722 TimeRFC1123(SBuf.st_mtime).c_str());
723 Req += Buf;
724 }
725 else
726 {
727 if (Itm->LastModified != 0)
728 {
729 sprintf(Buf,"If-Modified-Since: %s\r\n",TimeRFC1123(Itm->LastModified).c_str());
730 Req += Buf;
731 }
732 }
733
734 if (Proxy.User.empty() == false || Proxy.Password.empty() == false)
735 Req += string("Proxy-Authorization: Basic ") +
736 Base64Encode(Proxy.User + ":" + Proxy.Password) + "\r\n";
737
738 maybe_add_auth (Uri, _config->FindFile("Dir::Etc::netrc"));
739 if (Uri.User.empty() == false || Uri.Password.empty() == false)
740 {
741 Req += string("Authorization: Basic ") +
742 Base64Encode(Uri.User + ":" + Uri.Password) + "\r\n";
743 }
744 Req += "User-Agent: " + _config->Find("Acquire::http::User-Agent",
745 "Debian APT-HTTP/1.3 ("VERSION")") + "\r\n\r\n";
746
747 if (Debug == true)
748 cerr << Req << endl;
749
750 Out.Read(Req);
751 }
752 /*}}}*/
753 // HttpMethod::Go - Run a single loop /*{{{*/
754 // ---------------------------------------------------------------------
755 /* This runs the select loop over the server FDs, Output file FDs and
756 stdin. */
757 bool HttpMethod::Go(bool ToFile,ServerState *Srv)
758 {
759 // Server has closed the connection
760 if (Srv->ServerFd == -1 && (Srv->In.WriteSpace() == false ||
761 ToFile == false))
762 return false;
763
764 fd_set rfds,wfds;
765 FD_ZERO(&rfds);
766 FD_ZERO(&wfds);
767
768 /* Add the server. We only send more requests if the connection will
769 be persisting */
770 if (Srv->Out.WriteSpace() == true && Srv->ServerFd != -1
771 && Srv->Persistent == true)
772 FD_SET(Srv->ServerFd,&wfds);
773 if (Srv->In.ReadSpace() == true && Srv->ServerFd != -1)
774 FD_SET(Srv->ServerFd,&rfds);
775
776 // Add the file
777 int FileFD = -1;
778 if (File != 0)
779 FileFD = File->Fd();
780
781 if (Srv->In.WriteSpace() == true && ToFile == true && FileFD != -1)
782 FD_SET(FileFD,&wfds);
783
784 // Add stdin
785 if (_config->FindB("Acquire::http::DependOnSTDIN", true) == true)
786 FD_SET(STDIN_FILENO,&rfds);
787
788 // Figure out the max fd
789 int MaxFd = FileFD;
790 if (MaxFd < Srv->ServerFd)
791 MaxFd = Srv->ServerFd;
792
793 // Select
794 struct timeval tv;
795 tv.tv_sec = TimeOut;
796 tv.tv_usec = 0;
797 int Res = 0;
798 if ((Res = select(MaxFd+1,&rfds,&wfds,0,&tv)) < 0)
799 {
800 if (errno == EINTR)
801 return true;
802 return _error->Errno("select",_("Select failed"));
803 }
804
805 if (Res == 0)
806 {
807 _error->Error(_("Connection timed out"));
808 return ServerDie(Srv);
809 }
810
811 // Handle server IO
812 if (Srv->ServerFd != -1 && FD_ISSET(Srv->ServerFd,&rfds))
813 {
814 errno = 0;
815 if (Srv->In.Read(Srv->ServerFd) == false)
816 return ServerDie(Srv);
817 }
818
819 if (Srv->ServerFd != -1 && FD_ISSET(Srv->ServerFd,&wfds))
820 {
821 errno = 0;
822 if (Srv->Out.Write(Srv->ServerFd) == false)
823 return ServerDie(Srv);
824 }
825
826 // Send data to the file
827 if (FileFD != -1 && FD_ISSET(FileFD,&wfds))
828 {
829 if (Srv->In.Write(FileFD) == false)
830 return _error->Errno("write",_("Error writing to output file"));
831 }
832
833 // Handle commands from APT
834 if (FD_ISSET(STDIN_FILENO,&rfds))
835 {
836 if (Run(true) != -1)
837 exit(100);
838 }
839
840 return true;
841 }
842 /*}}}*/
843 // HttpMethod::Flush - Dump the buffer into the file /*{{{*/
844 // ---------------------------------------------------------------------
845 /* This takes the current input buffer from the Server FD and writes it
846 into the file */
847 bool HttpMethod::Flush(ServerState *Srv)
848 {
849 if (File != 0)
850 {
851 // on GNU/kFreeBSD, apt dies on /dev/null because non-blocking
852 // can't be set
853 if (File->Name() != "/dev/null")
854 SetNonBlock(File->Fd(),false);
855 if (Srv->In.WriteSpace() == false)
856 return true;
857
858 while (Srv->In.WriteSpace() == true)
859 {
860 if (Srv->In.Write(File->Fd()) == false)
861 return _error->Errno("write",_("Error writing to file"));
862 if (Srv->In.IsLimit() == true)
863 return true;
864 }
865
866 if (Srv->In.IsLimit() == true || Srv->Encoding == ServerState::Closes)
867 return true;
868 }
869 return false;
870 }
871 /*}}}*/
872 // HttpMethod::ServerDie - The server has closed the connection. /*{{{*/
873 // ---------------------------------------------------------------------
874 /* */
875 bool HttpMethod::ServerDie(ServerState *Srv)
876 {
877 unsigned int LErrno = errno;
878
879 // Dump the buffer to the file
880 if (Srv->State == ServerState::Data)
881 {
882 // on GNU/kFreeBSD, apt dies on /dev/null because non-blocking
883 // can't be set
884 if (File->Name() != "/dev/null")
885 SetNonBlock(File->Fd(),false);
886 while (Srv->In.WriteSpace() == true)
887 {
888 if (Srv->In.Write(File->Fd()) == false)
889 return _error->Errno("write",_("Error writing to the file"));
890
891 // Done
892 if (Srv->In.IsLimit() == true)
893 return true;
894 }
895 }
896
897 // See if this is because the server finished the data stream
898 if (Srv->In.IsLimit() == false && Srv->State != ServerState::Header &&
899 Srv->Encoding != ServerState::Closes)
900 {
901 Srv->Close();
902 if (LErrno == 0)
903 return _error->Error(_("Error reading from server. Remote end closed connection"));
904 errno = LErrno;
905 return _error->Errno("read",_("Error reading from server"));
906 }
907 else
908 {
909 Srv->In.Limit(-1);
910
911 // Nothing left in the buffer
912 if (Srv->In.WriteSpace() == false)
913 return false;
914
915 // We may have got multiple responses back in one packet..
916 Srv->Close();
917 return true;
918 }
919
920 return false;
921 }
922 /*}}}*/
923 // HttpMethod::DealWithHeaders - Handle the retrieved header data /*{{{*/
924 // ---------------------------------------------------------------------
925 /* We look at the header data we got back from the server and decide what
926 to do. Returns DealWithHeadersResult (see http.h for details).
927 */
928 HttpMethod::DealWithHeadersResult
929 HttpMethod::DealWithHeaders(FetchResult &Res,ServerState *Srv)
930 {
931 // Not Modified
932 if (Srv->Result == 304)
933 {
934 unlink(Queue->DestFile.c_str());
935 Res.IMSHit = true;
936 Res.LastModified = Queue->LastModified;
937 return IMS_HIT;
938 }
939
940 /* Redirect
941 *
942 * Note that it is only OK for us to treat all redirection the same
943 * because we *always* use GET, not other HTTP methods. There are
944 * three redirection codes for which it is not appropriate that we
945 * redirect. Pass on those codes so the error handling kicks in.
946 */
947 if (AllowRedirect
948 && (Srv->Result > 300 && Srv->Result < 400)
949 && (Srv->Result != 300 // Multiple Choices
950 && Srv->Result != 304 // Not Modified
951 && Srv->Result != 306)) // (Not part of HTTP/1.1, reserved)
952 {
953 if (Srv->Location.empty() == true);
954 else if (Srv->Location[0] == '/' && Queue->Uri.empty() == false)
955 {
956 URI Uri = Queue->Uri;
957 if (Uri.Host.empty() == false)
958 {
959 if (Uri.Port != 0)
960 strprintf(NextURI, "http://%s:%u", Uri.Host.c_str(), Uri.Port);
961 else
962 NextURI = "http://" + Uri.Host;
963 }
964 else
965 NextURI.clear();
966 NextURI.append(DeQuoteString(Srv->Location));
967 return TRY_AGAIN_OR_REDIRECT;
968 }
969 else
970 {
971 NextURI = DeQuoteString(Srv->Location);
972 return TRY_AGAIN_OR_REDIRECT;
973 }
974 /* else pass through for error message */
975 }
976
977 /* We have a reply we dont handle. This should indicate a perm server
978 failure */
979 if (Srv->Result < 200 || Srv->Result >= 300)
980 {
981 char err[255];
982 snprintf(err,sizeof(err)-1,"HttpError%i",Srv->Result);
983 SetFailReason(err);
984 _error->Error("%u %s",Srv->Result,Srv->Code);
985 if (Srv->HaveContent == true)
986 return ERROR_WITH_CONTENT_PAGE;
987 return ERROR_UNRECOVERABLE;
988 }
989
990 // This is some sort of 2xx 'data follows' reply
991 Res.LastModified = Srv->Date;
992 Res.Size = Srv->Size;
993
994 // Open the file
995 delete File;
996 File = new FileFd(Queue->DestFile,FileFd::WriteAny);
997 if (_error->PendingError() == true)
998 return ERROR_NOT_FROM_SERVER;
999
1000 FailFile = Queue->DestFile;
1001 FailFile.c_str(); // Make sure we dont do a malloc in the signal handler
1002 FailFd = File->Fd();
1003 FailTime = Srv->Date;
1004
1005 // Set the expected size
1006 if (Srv->StartPos >= 0)
1007 {
1008 Res.ResumePoint = Srv->StartPos;
1009 if (ftruncate(File->Fd(),Srv->StartPos) < 0)
1010 _error->Errno("ftruncate", _("Failed to truncate file"));
1011 }
1012
1013 // Set the start point
1014 lseek(File->Fd(),0,SEEK_END);
1015
1016 delete Srv->In.Hash;
1017 Srv->In.Hash = new Hashes;
1018
1019 // Fill the Hash if the file is non-empty (resume)
1020 if (Srv->StartPos > 0)
1021 {
1022 lseek(File->Fd(),0,SEEK_SET);
1023 if (Srv->In.Hash->AddFD(File->Fd(),Srv->StartPos) == false)
1024 {
1025 _error->Errno("read",_("Problem hashing file"));
1026 return ERROR_NOT_FROM_SERVER;
1027 }
1028 lseek(File->Fd(),0,SEEK_END);
1029 }
1030
1031 SetNonBlock(File->Fd(),true);
1032 return FILE_IS_OPEN;
1033 }
1034 /*}}}*/
1035 // HttpMethod::SigTerm - Handle a fatal signal /*{{{*/
1036 // ---------------------------------------------------------------------
1037 /* This closes and timestamps the open file. This is neccessary to get
1038 resume behavoir on user abort */
1039 void HttpMethod::SigTerm(int)
1040 {
1041 if (FailFd == -1)
1042 _exit(100);
1043 close(FailFd);
1044
1045 // Timestamp
1046 struct utimbuf UBuf;
1047 UBuf.actime = FailTime;
1048 UBuf.modtime = FailTime;
1049 utime(FailFile.c_str(),&UBuf);
1050
1051 _exit(100);
1052 }
1053 /*}}}*/
1054 // HttpMethod::Fetch - Fetch an item /*{{{*/
1055 // ---------------------------------------------------------------------
1056 /* This adds an item to the pipeline. We keep the pipeline at a fixed
1057 depth. */
1058 bool HttpMethod::Fetch(FetchItem *)
1059 {
1060 if (Server == 0)
1061 return true;
1062
1063 // Queue the requests
1064 int Depth = -1;
1065 for (FetchItem *I = Queue; I != 0 && Depth < (signed)PipelineDepth;
1066 I = I->Next, Depth++)
1067 {
1068 // If pipelining is disabled, we only queue 1 request
1069 if (Server->Pipeline == false && Depth >= 0)
1070 break;
1071
1072 // Make sure we stick with the same server
1073 if (Server->Comp(I->Uri) == false)
1074 break;
1075 if (QueueBack == I)
1076 {
1077 QueueBack = I->Next;
1078 SendReq(I,Server->Out);
1079 continue;
1080 }
1081 }
1082
1083 return true;
1084 };
1085 /*}}}*/
1086 // HttpMethod::Configuration - Handle a configuration message /*{{{*/
1087 // ---------------------------------------------------------------------
1088 /* We stash the desired pipeline depth */
1089 bool HttpMethod::Configuration(string Message)
1090 {
1091 if (pkgAcqMethod::Configuration(Message) == false)
1092 return false;
1093
1094 AllowRedirect = _config->FindB("Acquire::http::AllowRedirect",true);
1095 TimeOut = _config->FindI("Acquire::http::Timeout",TimeOut);
1096 PipelineDepth = _config->FindI("Acquire::http::Pipeline-Depth",
1097 PipelineDepth);
1098 Debug = _config->FindB("Debug::Acquire::http",false);
1099 AutoDetectProxyCmd = _config->Find("Acquire::http::ProxyAutoDetect");
1100
1101 // Get the proxy to use
1102 AutoDetectProxy();
1103
1104 return true;
1105 }
1106 /*}}}*/
1107 // HttpMethod::Loop - Main loop /*{{{*/
1108 // ---------------------------------------------------------------------
1109 /* */
1110 int HttpMethod::Loop()
1111 {
1112 typedef vector<string> StringVector;
1113 typedef vector<string>::iterator StringVectorIterator;
1114 map<string, StringVector> Redirected;
1115
1116 signal(SIGTERM,SigTerm);
1117 signal(SIGINT,SigTerm);
1118
1119 Server = 0;
1120
1121 int FailCounter = 0;
1122 while (1)
1123 {
1124 // We have no commands, wait for some to arrive
1125 if (Queue == 0)
1126 {
1127 if (WaitFd(STDIN_FILENO) == false)
1128 return 0;
1129 }
1130
1131 /* Run messages, we can accept 0 (no message) if we didn't
1132 do a WaitFd above.. Otherwise the FD is closed. */
1133 int Result = Run(true);
1134 if (Result != -1 && (Result != 0 || Queue == 0))
1135 {
1136 if(FailReason.empty() == false ||
1137 _config->FindB("Acquire::http::DependOnSTDIN", true) == true)
1138 return 100;
1139 else
1140 return 0;
1141 }
1142
1143 if (Queue == 0)
1144 continue;
1145
1146 // Connect to the server
1147 if (Server == 0 || Server->Comp(Queue->Uri) == false)
1148 {
1149 delete Server;
1150 Server = new ServerState(Queue->Uri,this);
1151 }
1152 /* If the server has explicitly said this is the last connection
1153 then we pre-emptively shut down the pipeline and tear down
1154 the connection. This will speed up HTTP/1.0 servers a tad
1155 since we don't have to wait for the close sequence to
1156 complete */
1157 if (Server->Persistent == false)
1158 Server->Close();
1159
1160 // Reset the pipeline
1161 if (Server->ServerFd == -1)
1162 QueueBack = Queue;
1163
1164 // Connnect to the host
1165 if (Server->Open() == false)
1166 {
1167 Fail(true);
1168 delete Server;
1169 Server = 0;
1170 continue;
1171 }
1172
1173 // Fill the pipeline.
1174 Fetch(0);
1175
1176 // Fetch the next URL header data from the server.
1177 switch (Server->RunHeaders())
1178 {
1179 case ServerState::RUN_HEADERS_OK:
1180 break;
1181
1182 // The header data is bad
1183 case ServerState::RUN_HEADERS_PARSE_ERROR:
1184 {
1185 _error->Error(_("Bad header data"));
1186 Fail(true);
1187 RotateDNS();
1188 continue;
1189 }
1190
1191 // The server closed a connection during the header get..
1192 default:
1193 case ServerState::RUN_HEADERS_IO_ERROR:
1194 {
1195 FailCounter++;
1196 _error->Discard();
1197 Server->Close();
1198 Server->Pipeline = false;
1199
1200 if (FailCounter >= 2)
1201 {
1202 Fail(_("Connection failed"),true);
1203 FailCounter = 0;
1204 }
1205
1206 RotateDNS();
1207 continue;
1208 }
1209 };
1210
1211 // Decide what to do.
1212 FetchResult Res;
1213 Res.Filename = Queue->DestFile;
1214 switch (DealWithHeaders(Res,Server))
1215 {
1216 // Ok, the file is Open
1217 case FILE_IS_OPEN:
1218 {
1219 URIStart(Res);
1220
1221 // Run the data
1222 bool Result = Server->RunData();
1223
1224 /* If the server is sending back sizeless responses then fill in
1225 the size now */
1226 if (Res.Size == 0)
1227 Res.Size = File->Size();
1228
1229 // Close the file, destroy the FD object and timestamp it
1230 FailFd = -1;
1231 delete File;
1232 File = 0;
1233
1234 // Timestamp
1235 struct utimbuf UBuf;
1236 time(&UBuf.actime);
1237 UBuf.actime = Server->Date;
1238 UBuf.modtime = Server->Date;
1239 utime(Queue->DestFile.c_str(),&UBuf);
1240
1241 // Send status to APT
1242 if (Result == true)
1243 {
1244 Res.TakeHashes(*Server->In.Hash);
1245 URIDone(Res);
1246 }
1247 else
1248 {
1249 if (Server->ServerFd == -1)
1250 {
1251 FailCounter++;
1252 _error->Discard();
1253 Server->Close();
1254
1255 if (FailCounter >= 2)
1256 {
1257 Fail(_("Connection failed"),true);
1258 FailCounter = 0;
1259 }
1260
1261 QueueBack = Queue;
1262 }
1263 else
1264 Fail(true);
1265 }
1266 break;
1267 }
1268
1269 // IMS hit
1270 case IMS_HIT:
1271 {
1272 URIDone(Res);
1273 break;
1274 }
1275
1276 // Hard server error, not found or something
1277 case ERROR_UNRECOVERABLE:
1278 {
1279 Fail();
1280 break;
1281 }
1282
1283 // Hard internal error, kill the connection and fail
1284 case ERROR_NOT_FROM_SERVER:
1285 {
1286 delete File;
1287 File = 0;
1288
1289 Fail();
1290 RotateDNS();
1291 Server->Close();
1292 break;
1293 }
1294
1295 // We need to flush the data, the header is like a 404 w/ error text
1296 case ERROR_WITH_CONTENT_PAGE:
1297 {
1298 Fail();
1299
1300 // Send to content to dev/null
1301 File = new FileFd("/dev/null",FileFd::WriteExists);
1302 Server->RunData();
1303 delete File;
1304 File = 0;
1305 break;
1306 }
1307
1308 // Try again with a new URL
1309 case TRY_AGAIN_OR_REDIRECT:
1310 {
1311 // Clear rest of response if there is content
1312 if (Server->HaveContent)
1313 {
1314 File = new FileFd("/dev/null",FileFd::WriteExists);
1315 Server->RunData();
1316 delete File;
1317 File = 0;
1318 }
1319
1320 /* Detect redirect loops. No more redirects are allowed
1321 after the same URI is seen twice in a queue item. */
1322 StringVector &R = Redirected[Queue->DestFile];
1323 bool StopRedirects = false;
1324 if (R.size() == 0)
1325 R.push_back(Queue->Uri);
1326 else if (R[0] == "STOP" || R.size() > 10)
1327 StopRedirects = true;
1328 else
1329 {
1330 for (StringVectorIterator I = R.begin(); I != R.end(); I++)
1331 if (Queue->Uri == *I)
1332 {
1333 R[0] = "STOP";
1334 break;
1335 }
1336
1337 R.push_back(Queue->Uri);
1338 }
1339
1340 if (StopRedirects == false)
1341 Redirect(NextURI);
1342 else
1343 Fail();
1344
1345 break;
1346 }
1347
1348 default:
1349 Fail(_("Internal error"));
1350 break;
1351 }
1352
1353 FailCounter = 0;
1354 }
1355
1356 return 0;
1357 }
1358 /*}}}*/
1359 // HttpMethod::AutoDetectProxy - auto detect proxy /*{{{*/
1360 // ---------------------------------------------------------------------
1361 /* */
1362 bool HttpMethod::AutoDetectProxy()
1363 {
1364 if (AutoDetectProxyCmd.empty())
1365 return true;
1366
1367 if (Debug)
1368 clog << "Using auto proxy detect command: " << AutoDetectProxyCmd << endl;
1369
1370 int Pipes[2] = {-1,-1};
1371 if (pipe(Pipes) != 0)
1372 return _error->Errno("pipe", "Failed to create Pipe");
1373
1374 pid_t Process = ExecFork();
1375 if (Process == 0)
1376 {
1377 close(Pipes[0]);
1378 dup2(Pipes[1],STDOUT_FILENO);
1379 SetCloseExec(STDOUT_FILENO,false);
1380
1381 const char *Args[2];
1382 Args[0] = AutoDetectProxyCmd.c_str();
1383 Args[1] = 0;
1384 execv(Args[0],(char **)Args);
1385 cerr << "Failed to exec method " << Args[0] << endl;
1386 _exit(100);
1387 }
1388 char buf[512];
1389 int InFd = Pipes[0];
1390 close(Pipes[1]);
1391 int res = read(InFd, buf, sizeof(buf));
1392 ExecWait(Process, "ProxyAutoDetect", true);
1393
1394 if (res < 0)
1395 return _error->Errno("read", "Failed to read");
1396 if (res == 0)
1397 return _error->Warning("ProxyAutoDetect returned no data");
1398
1399 // add trailing \0
1400 buf[res] = 0;
1401
1402 if (Debug)
1403 clog << "auto detect command returned: '" << buf << "'" << endl;
1404
1405 if (strstr(buf, "http://") == buf)
1406 _config->Set("Acquire::http::proxy", _strstrip(buf));
1407
1408 return true;
1409 }
1410 /*}}}*/
1411
1412