]> git.saurik.com Git - apt.git/blob - methods/http.cc
25e31de9a3126becc17db843f4e4e59a2a1837c5
[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 <apt-pkg/fileutl.h>
29 #include <apt-pkg/acquire-method.h>
30 #include <apt-pkg/error.h>
31 #include <apt-pkg/hashes.h>
32 #include <apt-pkg/netrc.h>
33
34 #include <sys/stat.h>
35 #include <sys/time.h>
36 #include <utime.h>
37 #include <unistd.h>
38 #include <signal.h>
39 #include <stdio.h>
40 #include <errno.h>
41 #include <string.h>
42 #include <iostream>
43 #include <map>
44 #include <apti18n.h>
45
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 using namespace std;
56
57 string HttpMethod::FailFile;
58 int HttpMethod::FailFd = -1;
59 time_t HttpMethod::FailTime = 0;
60 unsigned long PipelineDepth = 10;
61 unsigned long TimeOut = 120;
62 bool AllowRedirect = false;
63 bool Debug = false;
64 URI Proxy;
65
66 unsigned long CircleBuf::BwReadLimit=0;
67 unsigned long CircleBuf::BwTickReadData=0;
68 struct timeval CircleBuf::BwReadTick={0,0};
69 const unsigned int CircleBuf::BW_HZ=10;
70
71 // CircleBuf::CircleBuf - Circular input buffer /*{{{*/
72 // ---------------------------------------------------------------------
73 /* */
74 CircleBuf::CircleBuf(unsigned long Size) : Size(Size), Hash(0)
75 {
76 Buf = new unsigned char[Size];
77 Reset();
78
79 CircleBuf::BwReadLimit = _config->FindI("Acquire::http::Dl-Limit",0)*1024;
80 }
81 /*}}}*/
82 // CircleBuf::Reset - Reset to the default state /*{{{*/
83 // ---------------------------------------------------------------------
84 /* */
85 void CircleBuf::Reset()
86 {
87 InP = 0;
88 OutP = 0;
89 StrPos = 0;
90 MaxGet = (unsigned int)-1;
91 OutQueue = string();
92 if (Hash != 0)
93 {
94 delete Hash;
95 Hash = new Hashes;
96 }
97 };
98 /*}}}*/
99 // CircleBuf::Read - Read from a FD into the circular buffer /*{{{*/
100 // ---------------------------------------------------------------------
101 /* This fills up the buffer with as much data as is in the FD, assuming it
102 is non-blocking.. */
103 bool CircleBuf::Read(int Fd)
104 {
105 unsigned long BwReadMax;
106
107 while (1)
108 {
109 // Woops, buffer is full
110 if (InP - OutP == Size)
111 return true;
112
113 // what's left to read in this tick
114 BwReadMax = CircleBuf::BwReadLimit/BW_HZ;
115
116 if(CircleBuf::BwReadLimit) {
117 struct timeval now;
118 gettimeofday(&now,0);
119
120 unsigned long d = (now.tv_sec-CircleBuf::BwReadTick.tv_sec)*1000000 +
121 now.tv_usec-CircleBuf::BwReadTick.tv_usec;
122 if(d > 1000000/BW_HZ) {
123 CircleBuf::BwReadTick = now;
124 CircleBuf::BwTickReadData = 0;
125 }
126
127 if(CircleBuf::BwTickReadData >= BwReadMax) {
128 usleep(1000000/BW_HZ);
129 return true;
130 }
131 }
132
133 // Write the buffer segment
134 int Res;
135 if(CircleBuf::BwReadLimit) {
136 Res = read(Fd,Buf + (InP%Size),
137 BwReadMax > LeftRead() ? LeftRead() : BwReadMax);
138 } else
139 Res = read(Fd,Buf + (InP%Size),LeftRead());
140
141 if(Res > 0 && BwReadLimit > 0)
142 CircleBuf::BwTickReadData += Res;
143
144 if (Res == 0)
145 return false;
146 if (Res < 0)
147 {
148 if (errno == EAGAIN)
149 return true;
150 return false;
151 }
152
153 if (InP == 0)
154 gettimeofday(&Start,0);
155 InP += Res;
156 }
157 }
158 /*}}}*/
159 // CircleBuf::Read - Put the string into the buffer /*{{{*/
160 // ---------------------------------------------------------------------
161 /* This will hold the string in and fill the buffer with it as it empties */
162 bool CircleBuf::Read(string Data)
163 {
164 OutQueue += Data;
165 FillOut();
166 return true;
167 }
168 /*}}}*/
169 // CircleBuf::FillOut - Fill the buffer from the output queue /*{{{*/
170 // ---------------------------------------------------------------------
171 /* */
172 void CircleBuf::FillOut()
173 {
174 if (OutQueue.empty() == true)
175 return;
176 while (1)
177 {
178 // Woops, buffer is full
179 if (InP - OutP == Size)
180 return;
181
182 // Write the buffer segment
183 unsigned long Sz = LeftRead();
184 if (OutQueue.length() - StrPos < Sz)
185 Sz = OutQueue.length() - StrPos;
186 memcpy(Buf + (InP%Size),OutQueue.c_str() + StrPos,Sz);
187
188 // Advance
189 StrPos += Sz;
190 InP += Sz;
191 if (OutQueue.length() == StrPos)
192 {
193 StrPos = 0;
194 OutQueue = "";
195 return;
196 }
197 }
198 }
199 /*}}}*/
200 // CircleBuf::Write - Write from the buffer into a FD /*{{{*/
201 // ---------------------------------------------------------------------
202 /* This empties the buffer into the FD. */
203 bool CircleBuf::Write(int Fd)
204 {
205 while (1)
206 {
207 FillOut();
208
209 // Woops, buffer is empty
210 if (OutP == InP)
211 return true;
212
213 if (OutP == MaxGet)
214 return true;
215
216 // Write the buffer segment
217 int Res;
218 Res = write(Fd,Buf + (OutP%Size),LeftWrite());
219
220 if (Res == 0)
221 return false;
222 if (Res < 0)
223 {
224 if (errno == EAGAIN)
225 return true;
226
227 return false;
228 }
229
230 if (Hash != 0)
231 Hash->Add(Buf + (OutP%Size),Res);
232
233 OutP += Res;
234 }
235 }
236 /*}}}*/
237 // CircleBuf::WriteTillEl - Write from the buffer to a string /*{{{*/
238 // ---------------------------------------------------------------------
239 /* This copies till the first empty line */
240 bool CircleBuf::WriteTillEl(string &Data,bool Single)
241 {
242 // We cheat and assume it is unneeded to have more than one buffer load
243 for (unsigned long I = OutP; I < InP; I++)
244 {
245 if (Buf[I%Size] != '\n')
246 continue;
247 ++I;
248
249 if (Single == false)
250 {
251 if (I < InP && Buf[I%Size] == '\r')
252 ++I;
253 if (I >= InP || Buf[I%Size] != '\n')
254 continue;
255 ++I;
256 }
257
258 Data = "";
259 while (OutP < I)
260 {
261 unsigned long Sz = LeftWrite();
262 if (Sz == 0)
263 return false;
264 if (I - OutP < Sz)
265 Sz = I - OutP;
266 Data += string((char *)(Buf + (OutP%Size)),Sz);
267 OutP += Sz;
268 }
269 return true;
270 }
271 return false;
272 }
273 /*}}}*/
274 // CircleBuf::Stats - Print out stats information /*{{{*/
275 // ---------------------------------------------------------------------
276 /* */
277 void CircleBuf::Stats()
278 {
279 if (InP == 0)
280 return;
281
282 struct timeval Stop;
283 gettimeofday(&Stop,0);
284 /* float Diff = Stop.tv_sec - Start.tv_sec +
285 (float)(Stop.tv_usec - Start.tv_usec)/1000000;
286 clog << "Got " << InP << " in " << Diff << " at " << InP/Diff << endl;*/
287 }
288 /*}}}*/
289
290 // ServerState::ServerState - Constructor /*{{{*/
291 // ---------------------------------------------------------------------
292 /* */
293 ServerState::ServerState(URI Srv,HttpMethod *Owner) : Owner(Owner),
294 In(64*1024), Out(4*1024),
295 ServerName(Srv)
296 {
297 Reset();
298 }
299 /*}}}*/
300 // ServerState::Open - Open a connection to the server /*{{{*/
301 // ---------------------------------------------------------------------
302 /* This opens a connection to the server. */
303 bool ServerState::Open()
304 {
305 // Use the already open connection if possible.
306 if (ServerFd != -1)
307 return true;
308
309 Close();
310 In.Reset();
311 Out.Reset();
312 Persistent = true;
313
314 // Determine the proxy setting
315 string SpecificProxy = _config->Find("Acquire::http::Proxy::" + ServerName.Host);
316 if (!SpecificProxy.empty())
317 {
318 if (SpecificProxy == "DIRECT")
319 Proxy = "";
320 else
321 Proxy = SpecificProxy;
322 }
323 else
324 {
325 string DefProxy = _config->Find("Acquire::http::Proxy");
326 if (!DefProxy.empty())
327 {
328 Proxy = DefProxy;
329 }
330 else
331 {
332 char* result = getenv("http_proxy");
333 Proxy = result ? result : "";
334 }
335 }
336
337 // Parse no_proxy, a , separated list of domains
338 if (getenv("no_proxy") != 0)
339 {
340 if (CheckDomainList(ServerName.Host,getenv("no_proxy")) == true)
341 Proxy = "";
342 }
343
344 // Determine what host and port to use based on the proxy settings
345 int Port = 0;
346 string Host;
347 if (Proxy.empty() == true || Proxy.Host.empty() == true)
348 {
349 if (ServerName.Port != 0)
350 Port = ServerName.Port;
351 Host = ServerName.Host;
352 }
353 else
354 {
355 if (Proxy.Port != 0)
356 Port = Proxy.Port;
357 Host = Proxy.Host;
358 }
359
360 // Connect to the remote server
361 if (Connect(Host,Port,"http",80,ServerFd,TimeOut,Owner) == false)
362 return false;
363
364 return true;
365 }
366 /*}}}*/
367 // ServerState::Close - Close a connection to the server /*{{{*/
368 // ---------------------------------------------------------------------
369 /* */
370 bool ServerState::Close()
371 {
372 close(ServerFd);
373 ServerFd = -1;
374 return true;
375 }
376 /*}}}*/
377 // ServerState::RunHeaders - Get the headers before the data /*{{{*/
378 // ---------------------------------------------------------------------
379 /* Returns 0 if things are OK, 1 if an IO error occurred and 2 if a header
380 parse error occurred */
381 ServerState::RunHeadersResult ServerState::RunHeaders()
382 {
383 State = Header;
384
385 Owner->Status(_("Waiting for headers"));
386
387 Major = 0;
388 Minor = 0;
389 Result = 0;
390 Size = 0;
391 StartPos = 0;
392 Encoding = Closes;
393 HaveContent = false;
394 time(&Date);
395
396 do
397 {
398 string Data;
399 if (In.WriteTillEl(Data) == false)
400 continue;
401
402 if (Debug == true)
403 clog << Data;
404
405 for (string::const_iterator I = Data.begin(); I < Data.end(); I++)
406 {
407 string::const_iterator J = I;
408 for (; J != Data.end() && *J != '\n' && *J != '\r';J++);
409 if (HeaderLine(string(I,J)) == false)
410 return RUN_HEADERS_PARSE_ERROR;
411 I = J;
412 }
413
414 // 100 Continue is a Nop...
415 if (Result == 100)
416 continue;
417
418 // Tidy up the connection persistance state.
419 if (Encoding == Closes && HaveContent == true)
420 Persistent = false;
421
422 return RUN_HEADERS_OK;
423 }
424 while (Owner->Go(false,this) == true);
425
426 return RUN_HEADERS_IO_ERROR;
427 }
428 /*}}}*/
429 // ServerState::RunData - Transfer the data from the socket /*{{{*/
430 // ---------------------------------------------------------------------
431 /* */
432 bool ServerState::RunData()
433 {
434 State = Data;
435
436 // Chunked transfer encoding is fun..
437 if (Encoding == Chunked)
438 {
439 while (1)
440 {
441 // Grab the block size
442 bool Last = true;
443 string Data;
444 In.Limit(-1);
445 do
446 {
447 if (In.WriteTillEl(Data,true) == true)
448 break;
449 }
450 while ((Last = Owner->Go(false,this)) == true);
451
452 if (Last == false)
453 return false;
454
455 // See if we are done
456 unsigned long Len = strtol(Data.c_str(),0,16);
457 if (Len == 0)
458 {
459 In.Limit(-1);
460
461 // We have to remove the entity trailer
462 Last = true;
463 do
464 {
465 if (In.WriteTillEl(Data,true) == true && Data.length() <= 2)
466 break;
467 }
468 while ((Last = Owner->Go(false,this)) == true);
469 if (Last == false)
470 return false;
471 return !_error->PendingError();
472 }
473
474 // Transfer the block
475 In.Limit(Len);
476 while (Owner->Go(true,this) == true)
477 if (In.IsLimit() == true)
478 break;
479
480 // Error
481 if (In.IsLimit() == false)
482 return false;
483
484 // The server sends an extra new line before the next block specifier..
485 In.Limit(-1);
486 Last = true;
487 do
488 {
489 if (In.WriteTillEl(Data,true) == true)
490 break;
491 }
492 while ((Last = Owner->Go(false,this)) == true);
493 if (Last == false)
494 return false;
495 }
496 }
497 else
498 {
499 /* Closes encoding is used when the server did not specify a size, the
500 loss of the connection means we are done */
501 if (Encoding == Closes)
502 In.Limit(-1);
503 else
504 In.Limit(Size - StartPos);
505
506 // Just transfer the whole block.
507 do
508 {
509 if (In.IsLimit() == false)
510 continue;
511
512 In.Limit(-1);
513 return !_error->PendingError();
514 }
515 while (Owner->Go(true,this) == true);
516 }
517
518 return Owner->Flush(this) && !_error->PendingError();
519 }
520 /*}}}*/
521 // ServerState::HeaderLine - Process a header line /*{{{*/
522 // ---------------------------------------------------------------------
523 /* */
524 bool ServerState::HeaderLine(string Line)
525 {
526 if (Line.empty() == true)
527 return true;
528
529 // The http server might be trying to do something evil.
530 if (Line.length() >= MAXLEN)
531 return _error->Error(_("Got a single header line over %u chars"),MAXLEN);
532
533 string::size_type Pos = Line.find(' ');
534 if (Pos == string::npos || Pos+1 > Line.length())
535 {
536 // Blah, some servers use "connection:closes", evil.
537 Pos = Line.find(':');
538 if (Pos == string::npos || Pos + 2 > Line.length())
539 return _error->Error(_("Bad header line"));
540 Pos++;
541 }
542
543 // Parse off any trailing spaces between the : and the next word.
544 string::size_type Pos2 = Pos;
545 while (Pos2 < Line.length() && isspace(Line[Pos2]) != 0)
546 Pos2++;
547
548 string Tag = string(Line,0,Pos);
549 string Val = string(Line,Pos2);
550
551 if (stringcasecmp(Tag.c_str(),Tag.c_str()+4,"HTTP") == 0)
552 {
553 // Evil servers return no version
554 if (Line[4] == '/')
555 {
556 int const elements = sscanf(Line.c_str(),"HTTP/%u.%u %u%[^\n]",&Major,&Minor,&Result,Code);
557 if (elements == 3)
558 {
559 Code[0] = '\0';
560 if (Debug == true)
561 clog << "HTTP server doesn't give Reason-Phrase for " << Result << std::endl;
562 }
563 else if (elements != 4)
564 return _error->Error(_("The HTTP server sent an invalid reply header"));
565 }
566 else
567 {
568 Major = 0;
569 Minor = 9;
570 if (sscanf(Line.c_str(),"HTTP %u%[^\n]",&Result,Code) != 2)
571 return _error->Error(_("The HTTP server sent an invalid reply header"));
572 }
573
574 /* Check the HTTP response header to get the default persistance
575 state. */
576 if (Major < 1)
577 Persistent = false;
578 else
579 {
580 if (Major == 1 && Minor <= 0)
581 Persistent = false;
582 else
583 Persistent = true;
584 }
585
586 return true;
587 }
588
589 if (stringcasecmp(Tag,"Content-Length:") == 0)
590 {
591 if (Encoding == Closes)
592 Encoding = Stream;
593 HaveContent = true;
594
595 // The length is already set from the Content-Range header
596 if (StartPos != 0)
597 return true;
598
599 if (sscanf(Val.c_str(),"%lu",&Size) != 1)
600 return _error->Error(_("The HTTP server sent an invalid Content-Length header"));
601 return true;
602 }
603
604 if (stringcasecmp(Tag,"Content-Type:") == 0)
605 {
606 HaveContent = true;
607 return true;
608 }
609
610 if (stringcasecmp(Tag,"Content-Range:") == 0)
611 {
612 HaveContent = true;
613
614 if (sscanf(Val.c_str(),"bytes %lu-%*u/%lu",&StartPos,&Size) != 2)
615 return _error->Error(_("The HTTP server sent an invalid Content-Range header"));
616 if ((unsigned)StartPos > Size)
617 return _error->Error(_("This HTTP server has broken range support"));
618 return true;
619 }
620
621 if (stringcasecmp(Tag,"Transfer-Encoding:") == 0)
622 {
623 HaveContent = true;
624 if (stringcasecmp(Val,"chunked") == 0)
625 Encoding = Chunked;
626 return true;
627 }
628
629 if (stringcasecmp(Tag,"Connection:") == 0)
630 {
631 if (stringcasecmp(Val,"close") == 0)
632 Persistent = false;
633 if (stringcasecmp(Val,"keep-alive") == 0)
634 Persistent = true;
635 return true;
636 }
637
638 if (stringcasecmp(Tag,"Last-Modified:") == 0)
639 {
640 if (RFC1123StrToTime(Val.c_str(), Date) == false)
641 return _error->Error(_("Unknown date format"));
642 return true;
643 }
644
645 if (stringcasecmp(Tag,"Location:") == 0)
646 {
647 Location = Val;
648 return true;
649 }
650
651 return true;
652 }
653 /*}}}*/
654
655 // HttpMethod::SendReq - Send the HTTP request /*{{{*/
656 // ---------------------------------------------------------------------
657 /* This places the http request in the outbound buffer */
658 void HttpMethod::SendReq(FetchItem *Itm,CircleBuf &Out)
659 {
660 URI Uri = Itm->Uri;
661
662 // The HTTP server expects a hostname with a trailing :port
663 char Buf[1000];
664 string ProperHost = Uri.Host;
665 if (Uri.Port != 0)
666 {
667 sprintf(Buf,":%u",Uri.Port);
668 ProperHost += Buf;
669 }
670
671 // Just in case.
672 if (Itm->Uri.length() >= sizeof(Buf))
673 abort();
674
675 /* Build the request. We include a keep-alive header only for non-proxy
676 requests. This is to tweak old http/1.0 servers that do support keep-alive
677 but not HTTP/1.1 automatic keep-alive. Doing this with a proxy server
678 will glitch HTTP/1.0 proxies because they do not filter it out and
679 pass it on, HTTP/1.1 says the connection should default to keep alive
680 and we expect the proxy to do this */
681 if (Proxy.empty() == true || Proxy.Host.empty())
682 sprintf(Buf,"GET %s HTTP/1.1\r\nHost: %s\r\nConnection: keep-alive\r\n",
683 QuoteString(Uri.Path,"~").c_str(),ProperHost.c_str());
684 else
685 {
686 /* Generate a cache control header if necessary. We place a max
687 cache age on index files, optionally set a no-cache directive
688 and a no-store directive for archives. */
689 sprintf(Buf,"GET %s HTTP/1.1\r\nHost: %s\r\n",
690 Itm->Uri.c_str(),ProperHost.c_str());
691 }
692 // generate a cache control header (if needed)
693 if (_config->FindB("Acquire::http::No-Cache",false) == true)
694 {
695 strcat(Buf,"Cache-Control: no-cache\r\nPragma: no-cache\r\n");
696 }
697 else
698 {
699 if (Itm->IndexFile == true)
700 {
701 sprintf(Buf+strlen(Buf),"Cache-Control: max-age=%u\r\n",
702 _config->FindI("Acquire::http::Max-Age",0));
703 }
704 else
705 {
706 if (_config->FindB("Acquire::http::No-Store",false) == true)
707 strcat(Buf,"Cache-Control: no-store\r\n");
708 }
709 }
710
711
712 string Req = Buf;
713
714 // Check for a partial file
715 struct stat SBuf;
716 if (stat(Itm->DestFile.c_str(),&SBuf) >= 0 && SBuf.st_size > 0)
717 {
718 // In this case we send an if-range query with a range header
719 sprintf(Buf,"Range: bytes=%li-\r\nIf-Range: %s\r\n",(long)SBuf.st_size - 1,
720 TimeRFC1123(SBuf.st_mtime).c_str());
721 Req += Buf;
722 }
723 else
724 {
725 if (Itm->LastModified != 0)
726 {
727 sprintf(Buf,"If-Modified-Since: %s\r\n",TimeRFC1123(Itm->LastModified).c_str());
728 Req += Buf;
729 }
730 }
731
732 if (Proxy.User.empty() == false || Proxy.Password.empty() == false)
733 Req += string("Proxy-Authorization: Basic ") +
734 Base64Encode(Proxy.User + ":" + Proxy.Password) + "\r\n";
735
736 maybe_add_auth (Uri, _config->FindFile("Dir::Etc::netrc"));
737 if (Uri.User.empty() == false || Uri.Password.empty() == false)
738 {
739 Req += string("Authorization: Basic ") +
740 Base64Encode(Uri.User + ":" + Uri.Password) + "\r\n";
741 }
742 Req += "User-Agent: " + _config->Find("Acquire::http::User-Agent",
743 "Debian APT-HTTP/1.3 ("VERSION")") + "\r\n\r\n";
744
745 if (Debug == true)
746 cerr << Req << endl;
747
748 Out.Read(Req);
749 }
750 /*}}}*/
751 // HttpMethod::Go - Run a single loop /*{{{*/
752 // ---------------------------------------------------------------------
753 /* This runs the select loop over the server FDs, Output file FDs and
754 stdin. */
755 bool HttpMethod::Go(bool ToFile,ServerState *Srv)
756 {
757 // Server has closed the connection
758 if (Srv->ServerFd == -1 && (Srv->In.WriteSpace() == false ||
759 ToFile == false))
760 return false;
761
762 fd_set rfds,wfds;
763 FD_ZERO(&rfds);
764 FD_ZERO(&wfds);
765
766 /* Add the server. We only send more requests if the connection will
767 be persisting */
768 if (Srv->Out.WriteSpace() == true && Srv->ServerFd != -1
769 && Srv->Persistent == true)
770 FD_SET(Srv->ServerFd,&wfds);
771 if (Srv->In.ReadSpace() == true && Srv->ServerFd != -1)
772 FD_SET(Srv->ServerFd,&rfds);
773
774 // Add the file
775 int FileFD = -1;
776 if (File != 0)
777 FileFD = File->Fd();
778
779 if (Srv->In.WriteSpace() == true && ToFile == true && FileFD != -1)
780 FD_SET(FileFD,&wfds);
781
782 // Add stdin
783 FD_SET(STDIN_FILENO,&rfds);
784
785 // Figure out the max fd
786 int MaxFd = FileFD;
787 if (MaxFd < Srv->ServerFd)
788 MaxFd = Srv->ServerFd;
789
790 // Select
791 struct timeval tv;
792 tv.tv_sec = TimeOut;
793 tv.tv_usec = 0;
794 int Res = 0;
795 if ((Res = select(MaxFd+1,&rfds,&wfds,0,&tv)) < 0)
796 {
797 if (errno == EINTR)
798 return true;
799 return _error->Errno("select",_("Select failed"));
800 }
801
802 if (Res == 0)
803 {
804 _error->Error(_("Connection timed out"));
805 return ServerDie(Srv);
806 }
807
808 // Handle server IO
809 if (Srv->ServerFd != -1 && FD_ISSET(Srv->ServerFd,&rfds))
810 {
811 errno = 0;
812 if (Srv->In.Read(Srv->ServerFd) == false)
813 return ServerDie(Srv);
814 }
815
816 if (Srv->ServerFd != -1 && FD_ISSET(Srv->ServerFd,&wfds))
817 {
818 errno = 0;
819 if (Srv->Out.Write(Srv->ServerFd) == false)
820 return ServerDie(Srv);
821 }
822
823 // Send data to the file
824 if (FileFD != -1 && FD_ISSET(FileFD,&wfds))
825 {
826 if (Srv->In.Write(FileFD) == false)
827 return _error->Errno("write",_("Error writing to output file"));
828 }
829
830 // Handle commands from APT
831 if (FD_ISSET(STDIN_FILENO,&rfds))
832 {
833 if (Run(true) != -1)
834 exit(100);
835 }
836
837 return true;
838 }
839 /*}}}*/
840 // HttpMethod::Flush - Dump the buffer into the file /*{{{*/
841 // ---------------------------------------------------------------------
842 /* This takes the current input buffer from the Server FD and writes it
843 into the file */
844 bool HttpMethod::Flush(ServerState *Srv)
845 {
846 if (File != 0)
847 {
848 // on GNU/kFreeBSD, apt dies on /dev/null because non-blocking
849 // can't be set
850 if (File->Name() != "/dev/null")
851 SetNonBlock(File->Fd(),false);
852 if (Srv->In.WriteSpace() == false)
853 return true;
854
855 while (Srv->In.WriteSpace() == true)
856 {
857 if (Srv->In.Write(File->Fd()) == false)
858 return _error->Errno("write",_("Error writing to file"));
859 if (Srv->In.IsLimit() == true)
860 return true;
861 }
862
863 if (Srv->In.IsLimit() == true || Srv->Encoding == ServerState::Closes)
864 return true;
865 }
866 return false;
867 }
868 /*}}}*/
869 // HttpMethod::ServerDie - The server has closed the connection. /*{{{*/
870 // ---------------------------------------------------------------------
871 /* */
872 bool HttpMethod::ServerDie(ServerState *Srv)
873 {
874 unsigned int LErrno = errno;
875
876 // Dump the buffer to the file
877 if (Srv->State == ServerState::Data)
878 {
879 // on GNU/kFreeBSD, apt dies on /dev/null because non-blocking
880 // can't be set
881 if (File->Name() != "/dev/null")
882 SetNonBlock(File->Fd(),false);
883 while (Srv->In.WriteSpace() == true)
884 {
885 if (Srv->In.Write(File->Fd()) == false)
886 return _error->Errno("write",_("Error writing to the file"));
887
888 // Done
889 if (Srv->In.IsLimit() == true)
890 return true;
891 }
892 }
893
894 // See if this is because the server finished the data stream
895 if (Srv->In.IsLimit() == false && Srv->State != ServerState::Header &&
896 Srv->Encoding != ServerState::Closes)
897 {
898 Srv->Close();
899 if (LErrno == 0)
900 return _error->Error(_("Error reading from server. Remote end closed connection"));
901 errno = LErrno;
902 return _error->Errno("read",_("Error reading from server"));
903 }
904 else
905 {
906 Srv->In.Limit(-1);
907
908 // Nothing left in the buffer
909 if (Srv->In.WriteSpace() == false)
910 return false;
911
912 // We may have got multiple responses back in one packet..
913 Srv->Close();
914 return true;
915 }
916
917 return false;
918 }
919 /*}}}*/
920 // HttpMethod::DealWithHeaders - Handle the retrieved header data /*{{{*/
921 // ---------------------------------------------------------------------
922 /* We look at the header data we got back from the server and decide what
923 to do. Returns DealWithHeadersResult (see http.h for details).
924 */
925 HttpMethod::DealWithHeadersResult
926 HttpMethod::DealWithHeaders(FetchResult &Res,ServerState *Srv)
927 {
928 // Not Modified
929 if (Srv->Result == 304)
930 {
931 unlink(Queue->DestFile.c_str());
932 Res.IMSHit = true;
933 Res.LastModified = Queue->LastModified;
934 return IMS_HIT;
935 }
936
937 /* Redirect
938 *
939 * Note that it is only OK for us to treat all redirection the same
940 * because we *always* use GET, not other HTTP methods. There are
941 * three redirection codes for which it is not appropriate that we
942 * redirect. Pass on those codes so the error handling kicks in.
943 */
944 if (AllowRedirect
945 && (Srv->Result > 300 && Srv->Result < 400)
946 && (Srv->Result != 300 // Multiple Choices
947 && Srv->Result != 304 // Not Modified
948 && Srv->Result != 306)) // (Not part of HTTP/1.1, reserved)
949 {
950 if (!Srv->Location.empty())
951 {
952 NextURI = Srv->Location;
953 return TRY_AGAIN_OR_REDIRECT;
954 }
955 /* else pass through for error message */
956 }
957
958 /* We have a reply we dont handle. This should indicate a perm server
959 failure */
960 if (Srv->Result < 200 || Srv->Result >= 300)
961 {
962 char err[255];
963 snprintf(err,sizeof(err)-1,"HttpError%i",Srv->Result);
964 SetFailReason(err);
965 _error->Error("%u %s",Srv->Result,Srv->Code);
966 if (Srv->HaveContent == true)
967 return ERROR_WITH_CONTENT_PAGE;
968 return ERROR_UNRECOVERABLE;
969 }
970
971 // This is some sort of 2xx 'data follows' reply
972 Res.LastModified = Srv->Date;
973 Res.Size = Srv->Size;
974
975 // Open the file
976 delete File;
977 File = new FileFd(Queue->DestFile,FileFd::WriteAny);
978 if (_error->PendingError() == true)
979 return ERROR_NOT_FROM_SERVER;
980
981 FailFile = Queue->DestFile;
982 FailFile.c_str(); // Make sure we dont do a malloc in the signal handler
983 FailFd = File->Fd();
984 FailTime = Srv->Date;
985
986 // Set the expected size
987 if (Srv->StartPos >= 0)
988 {
989 Res.ResumePoint = Srv->StartPos;
990 if (ftruncate(File->Fd(),Srv->StartPos) < 0)
991 _error->Errno("ftruncate", _("Failed to truncate file"));
992 }
993
994 // Set the start point
995 lseek(File->Fd(),0,SEEK_END);
996
997 delete Srv->In.Hash;
998 Srv->In.Hash = new Hashes;
999
1000 // Fill the Hash if the file is non-empty (resume)
1001 if (Srv->StartPos > 0)
1002 {
1003 lseek(File->Fd(),0,SEEK_SET);
1004 if (Srv->In.Hash->AddFD(File->Fd(),Srv->StartPos) == false)
1005 {
1006 _error->Errno("read",_("Problem hashing file"));
1007 return ERROR_NOT_FROM_SERVER;
1008 }
1009 lseek(File->Fd(),0,SEEK_END);
1010 }
1011
1012 SetNonBlock(File->Fd(),true);
1013 return FILE_IS_OPEN;
1014 }
1015 /*}}}*/
1016 // HttpMethod::SigTerm - Handle a fatal signal /*{{{*/
1017 // ---------------------------------------------------------------------
1018 /* This closes and timestamps the open file. This is neccessary to get
1019 resume behavoir on user abort */
1020 void HttpMethod::SigTerm(int)
1021 {
1022 if (FailFd == -1)
1023 _exit(100);
1024 close(FailFd);
1025
1026 // Timestamp
1027 struct utimbuf UBuf;
1028 UBuf.actime = FailTime;
1029 UBuf.modtime = FailTime;
1030 utime(FailFile.c_str(),&UBuf);
1031
1032 _exit(100);
1033 }
1034 /*}}}*/
1035 // HttpMethod::Fetch - Fetch an item /*{{{*/
1036 // ---------------------------------------------------------------------
1037 /* This adds an item to the pipeline. We keep the pipeline at a fixed
1038 depth. */
1039 bool HttpMethod::Fetch(FetchItem *)
1040 {
1041 if (Server == 0)
1042 return true;
1043
1044 // Queue the requests
1045 int Depth = -1;
1046 for (FetchItem *I = Queue; I != 0 && Depth < (signed)PipelineDepth;
1047 I = I->Next, Depth++)
1048 {
1049 // If pipelining is disabled, we only queue 1 request
1050 if (Server->Pipeline == false && Depth >= 0)
1051 break;
1052
1053 // Make sure we stick with the same server
1054 if (Server->Comp(I->Uri) == false)
1055 break;
1056 if (QueueBack == I)
1057 {
1058 QueueBack = I->Next;
1059 SendReq(I,Server->Out);
1060 continue;
1061 }
1062 }
1063
1064 return true;
1065 };
1066 /*}}}*/
1067 // HttpMethod::Configuration - Handle a configuration message /*{{{*/
1068 // ---------------------------------------------------------------------
1069 /* We stash the desired pipeline depth */
1070 bool HttpMethod::Configuration(string Message)
1071 {
1072 if (pkgAcqMethod::Configuration(Message) == false)
1073 return false;
1074
1075 AllowRedirect = _config->FindB("Acquire::http::AllowRedirect",true);
1076 TimeOut = _config->FindI("Acquire::http::Timeout",TimeOut);
1077 PipelineDepth = _config->FindI("Acquire::http::Pipeline-Depth",
1078 PipelineDepth);
1079 Debug = _config->FindB("Debug::Acquire::http",false);
1080 AutoDetectProxyCmd = _config->Find("Acquire::http::ProxyAutoDetect");
1081
1082 // Get the proxy to use
1083 AutoDetectProxy();
1084
1085 return true;
1086 }
1087 /*}}}*/
1088 // HttpMethod::Loop - Main loop /*{{{*/
1089 // ---------------------------------------------------------------------
1090 /* */
1091 int HttpMethod::Loop()
1092 {
1093 typedef vector<string> StringVector;
1094 typedef vector<string>::iterator StringVectorIterator;
1095 map<string, StringVector> Redirected;
1096
1097 signal(SIGTERM,SigTerm);
1098 signal(SIGINT,SigTerm);
1099
1100 Server = 0;
1101
1102 int FailCounter = 0;
1103 while (1)
1104 {
1105 // We have no commands, wait for some to arrive
1106 if (Queue == 0)
1107 {
1108 if (WaitFd(STDIN_FILENO) == false)
1109 return 0;
1110 }
1111
1112 /* Run messages, we can accept 0 (no message) if we didn't
1113 do a WaitFd above.. Otherwise the FD is closed. */
1114 int Result = Run(true);
1115 if (Result != -1 && (Result != 0 || Queue == 0))
1116 return 100;
1117
1118 if (Queue == 0)
1119 continue;
1120
1121 // Connect to the server
1122 if (Server == 0 || Server->Comp(Queue->Uri) == false)
1123 {
1124 delete Server;
1125 Server = new ServerState(Queue->Uri,this);
1126 }
1127 /* If the server has explicitly said this is the last connection
1128 then we pre-emptively shut down the pipeline and tear down
1129 the connection. This will speed up HTTP/1.0 servers a tad
1130 since we don't have to wait for the close sequence to
1131 complete */
1132 if (Server->Persistent == false)
1133 Server->Close();
1134
1135 // Reset the pipeline
1136 if (Server->ServerFd == -1)
1137 QueueBack = Queue;
1138
1139 // Connnect to the host
1140 if (Server->Open() == false)
1141 {
1142 Fail(true);
1143 delete Server;
1144 Server = 0;
1145 continue;
1146 }
1147
1148 // Fill the pipeline.
1149 Fetch(0);
1150
1151 // Fetch the next URL header data from the server.
1152 switch (Server->RunHeaders())
1153 {
1154 case ServerState::RUN_HEADERS_OK:
1155 break;
1156
1157 // The header data is bad
1158 case ServerState::RUN_HEADERS_PARSE_ERROR:
1159 {
1160 _error->Error(_("Bad header data"));
1161 Fail(true);
1162 RotateDNS();
1163 continue;
1164 }
1165
1166 // The server closed a connection during the header get..
1167 default:
1168 case ServerState::RUN_HEADERS_IO_ERROR:
1169 {
1170 FailCounter++;
1171 _error->Discard();
1172 Server->Close();
1173 Server->Pipeline = false;
1174
1175 if (FailCounter >= 2)
1176 {
1177 Fail(_("Connection failed"),true);
1178 FailCounter = 0;
1179 }
1180
1181 RotateDNS();
1182 continue;
1183 }
1184 };
1185
1186 // Decide what to do.
1187 FetchResult Res;
1188 Res.Filename = Queue->DestFile;
1189 switch (DealWithHeaders(Res,Server))
1190 {
1191 // Ok, the file is Open
1192 case FILE_IS_OPEN:
1193 {
1194 URIStart(Res);
1195
1196 // Run the data
1197 bool Result = Server->RunData();
1198
1199 /* If the server is sending back sizeless responses then fill in
1200 the size now */
1201 if (Res.Size == 0)
1202 Res.Size = File->Size();
1203
1204 // Close the file, destroy the FD object and timestamp it
1205 FailFd = -1;
1206 delete File;
1207 File = 0;
1208
1209 // Timestamp
1210 struct utimbuf UBuf;
1211 time(&UBuf.actime);
1212 UBuf.actime = Server->Date;
1213 UBuf.modtime = Server->Date;
1214 utime(Queue->DestFile.c_str(),&UBuf);
1215
1216 // Send status to APT
1217 if (Result == true)
1218 {
1219 Res.TakeHashes(*Server->In.Hash);
1220 URIDone(Res);
1221 }
1222 else
1223 {
1224 if (Server->ServerFd == -1)
1225 {
1226 FailCounter++;
1227 _error->Discard();
1228 Server->Close();
1229
1230 if (FailCounter >= 2)
1231 {
1232 Fail(_("Connection failed"),true);
1233 FailCounter = 0;
1234 }
1235
1236 QueueBack = Queue;
1237 }
1238 else
1239 Fail(true);
1240 }
1241 break;
1242 }
1243
1244 // IMS hit
1245 case IMS_HIT:
1246 {
1247 URIDone(Res);
1248 break;
1249 }
1250
1251 // Hard server error, not found or something
1252 case ERROR_UNRECOVERABLE:
1253 {
1254 Fail();
1255 break;
1256 }
1257
1258 // Hard internal error, kill the connection and fail
1259 case ERROR_NOT_FROM_SERVER:
1260 {
1261 delete File;
1262 File = 0;
1263
1264 Fail();
1265 RotateDNS();
1266 Server->Close();
1267 break;
1268 }
1269
1270 // We need to flush the data, the header is like a 404 w/ error text
1271 case ERROR_WITH_CONTENT_PAGE:
1272 {
1273 Fail();
1274
1275 // Send to content to dev/null
1276 File = new FileFd("/dev/null",FileFd::WriteExists);
1277 Server->RunData();
1278 delete File;
1279 File = 0;
1280 break;
1281 }
1282
1283 // Try again with a new URL
1284 case TRY_AGAIN_OR_REDIRECT:
1285 {
1286 // Clear rest of response if there is content
1287 if (Server->HaveContent)
1288 {
1289 File = new FileFd("/dev/null",FileFd::WriteExists);
1290 Server->RunData();
1291 delete File;
1292 File = 0;
1293 }
1294
1295 /* Detect redirect loops. No more redirects are allowed
1296 after the same URI is seen twice in a queue item. */
1297 StringVector &R = Redirected[Queue->DestFile];
1298 bool StopRedirects = false;
1299 if (R.size() == 0)
1300 R.push_back(Queue->Uri);
1301 else if (R[0] == "STOP" || R.size() > 10)
1302 StopRedirects = true;
1303 else
1304 {
1305 for (StringVectorIterator I = R.begin(); I != R.end(); I++)
1306 if (Queue->Uri == *I)
1307 {
1308 R[0] = "STOP";
1309 break;
1310 }
1311
1312 R.push_back(Queue->Uri);
1313 }
1314
1315 if (StopRedirects == false)
1316 Redirect(NextURI);
1317 else
1318 Fail();
1319
1320 break;
1321 }
1322
1323 default:
1324 Fail(_("Internal error"));
1325 break;
1326 }
1327
1328 FailCounter = 0;
1329 }
1330
1331 return 0;
1332 }
1333 /*}}}*/
1334 // HttpMethod::AutoDetectProxy - auto detect proxy /*{{{*/
1335 // ---------------------------------------------------------------------
1336 /* */
1337 bool HttpMethod::AutoDetectProxy()
1338 {
1339 if (AutoDetectProxyCmd.empty())
1340 return true;
1341
1342 if (Debug)
1343 clog << "Using auto proxy detect command: " << AutoDetectProxyCmd << endl;
1344
1345 int Pipes[2] = {-1,-1};
1346 if (pipe(Pipes) != 0)
1347 return _error->Errno("pipe", "Failed to create Pipe");
1348
1349 pid_t Process = ExecFork();
1350 if (Process == 0)
1351 {
1352 dup2(Pipes[1],STDOUT_FILENO);
1353 SetCloseExec(STDOUT_FILENO,false);
1354
1355 const char *Args[2];
1356 Args[0] = AutoDetectProxyCmd.c_str();
1357 Args[1] = 0;
1358 execv(Args[0],(char **)Args);
1359 cerr << "Failed to exec method " << Args[0] << endl;
1360 _exit(100);
1361 }
1362 char buf[512];
1363 int InFd = Pipes[0];
1364 if (read(InFd, buf, sizeof(buf)) < 0)
1365 return _error->Errno("read", "Failed to read");
1366 ExecWait(Process, "ProxyAutoDetect");
1367
1368 if (Debug)
1369 clog << "auto detect command returned: '" << buf << "'" << endl;
1370
1371 if (strstr(buf, "http://") == buf)
1372 _config->Set("Acquire::http::proxy", _strstrip(buf));
1373
1374 return true;
1375 }
1376 /*}}}*/
1377
1378