retry without partial data after a 416 response
[ntk/apt.git] / methods / https.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 HTTPS Acquire Method - This is the HTTPS aquire method for APT.
7
8 It uses libcurl
9
10 ##################################################################### */
11 /*}}}*/
12 // Include Files /*{{{*/
13 #include <config.h>
14
15 #include <apt-pkg/fileutl.h>
16 #include <apt-pkg/acquire-method.h>
17 #include <apt-pkg/error.h>
18 #include <apt-pkg/hashes.h>
19 #include <apt-pkg/netrc.h>
20 #include <apt-pkg/configuration.h>
21
22 #include <sys/stat.h>
23 #include <sys/time.h>
24 #include <utime.h>
25 #include <unistd.h>
26 #include <signal.h>
27 #include <stdio.h>
28 #include <errno.h>
29 #include <string.h>
30 #include <iostream>
31 #include <sstream>
32
33 #include "config.h"
34 #include "https.h"
35 #include <apti18n.h>
36 /*}}}*/
37 using namespace std;
38
39 size_t
40 HttpsMethod::write_data(void *buffer, size_t size, size_t nmemb, void *userp)
41 {
42 HttpsMethod *me = (HttpsMethod *)userp;
43
44 if(me->File->Write(buffer, size*nmemb) != true)
45 return false;
46
47 return size*nmemb;
48 }
49
50 int
51 HttpsMethod::progress_callback(void *clientp, double dltotal, double dlnow,
52 double ultotal, double ulnow)
53 {
54 HttpsMethod *me = (HttpsMethod *)clientp;
55 if(dltotal > 0 && me->Res.Size == 0) {
56 me->Res.Size = (unsigned long long)dltotal;
57 me->URIStart(me->Res);
58 }
59 return 0;
60 }
61
62 void HttpsMethod::SetupProxy() /*{{{*/
63 {
64 URI ServerName = Queue->Uri;
65
66 // Curl should never read proxy settings from the environment, as
67 // we determine which proxy to use. Do this for consistency among
68 // methods and prevent an environment variable overriding a
69 // no-proxy ("DIRECT") setting in apt.conf.
70 curl_easy_setopt(curl, CURLOPT_PROXY, "");
71
72 // Determine the proxy setting - try https first, fallback to http and use env at last
73 string UseProxy = _config->Find("Acquire::https::Proxy::" + ServerName.Host,
74 _config->Find("Acquire::http::Proxy::" + ServerName.Host).c_str());
75
76 if (UseProxy.empty() == true)
77 UseProxy = _config->Find("Acquire::https::Proxy", _config->Find("Acquire::http::Proxy").c_str());
78
79 // User want to use NO proxy, so nothing to setup
80 if (UseProxy == "DIRECT")
81 return;
82
83 if (UseProxy.empty() == false)
84 {
85 // Parse no_proxy, a comma (,) separated list of domains we don't want to use
86 // a proxy for so we stop right here if it is in the list
87 if (getenv("no_proxy") != 0 && CheckDomainList(ServerName.Host,getenv("no_proxy")) == true)
88 return;
89 } else {
90 const char* result = getenv("https_proxy");
91 // FIXME: Fall back to http_proxy is to remain compatible with
92 // existing setups and behaviour of apt.conf. This should be
93 // deprecated in the future (including apt.conf). Most other
94 // programs do not fall back to http proxy settings and neither
95 // should Apt.
96 if (result == NULL)
97 result = getenv("http_proxy");
98 UseProxy = result == NULL ? "" : result;
99 }
100
101 // Determine what host and port to use based on the proxy settings
102 if (UseProxy.empty() == false)
103 {
104 Proxy = UseProxy;
105 if (Proxy.Port != 1)
106 curl_easy_setopt(curl, CURLOPT_PROXYPORT, Proxy.Port);
107 curl_easy_setopt(curl, CURLOPT_PROXY, Proxy.Host.c_str());
108 if (Proxy.User.empty() == false || Proxy.Password.empty() == false)
109 {
110 curl_easy_setopt(curl, CURLOPT_PROXYUSERNAME, Proxy.User.c_str());
111 curl_easy_setopt(curl, CURLOPT_PROXYPASSWORD, Proxy.Password.c_str());
112 }
113 }
114 } /*}}}*/
115 // HttpsMethod::Fetch - Fetch an item /*{{{*/
116 // ---------------------------------------------------------------------
117 /* This adds an item to the pipeline. We keep the pipeline at a fixed
118 depth. */
119 bool HttpsMethod::Fetch(FetchItem *Itm)
120 {
121 struct stat SBuf;
122 struct curl_slist *headers=NULL;
123 char curl_errorstr[CURL_ERROR_SIZE];
124 long curl_responsecode;
125 URI Uri = Itm->Uri;
126 string remotehost = Uri.Host;
127
128 // TODO:
129 // - http::Pipeline-Depth
130 // - error checking/reporting
131 // - more debug options? (CURLOPT_DEBUGFUNCTION?)
132
133 curl_easy_reset(curl);
134 SetupProxy();
135
136 maybe_add_auth (Uri, _config->FindFile("Dir::Etc::netrc"));
137
138 // callbacks
139 curl_easy_setopt(curl, CURLOPT_URL, static_cast<string>(Uri).c_str());
140 curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, write_data);
141 curl_easy_setopt(curl, CURLOPT_WRITEDATA, this);
142 curl_easy_setopt(curl, CURLOPT_PROGRESSFUNCTION, progress_callback);
143 curl_easy_setopt(curl, CURLOPT_PROGRESSDATA, this);
144 curl_easy_setopt(curl, CURLOPT_NOPROGRESS, false);
145 curl_easy_setopt(curl, CURLOPT_FILETIME, true);
146
147 // SSL parameters are set by default to the common (non mirror-specific) value
148 // if available (or a default one) and gets overload by mirror-specific ones.
149
150 // File containing the list of trusted CA.
151 string cainfo = _config->Find("Acquire::https::CaInfo","");
152 string knob = "Acquire::https::"+remotehost+"::CaInfo";
153 cainfo = _config->Find(knob.c_str(),cainfo.c_str());
154 if(cainfo.empty() == false)
155 curl_easy_setopt(curl, CURLOPT_CAINFO,cainfo.c_str());
156
157 // Check server certificate against previous CA list ...
158 bool peer_verify = _config->FindB("Acquire::https::Verify-Peer",true);
159 knob = "Acquire::https::" + remotehost + "::Verify-Peer";
160 peer_verify = _config->FindB(knob.c_str(), peer_verify);
161 curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, peer_verify);
162
163 // ... and hostname against cert CN or subjectAltName
164 bool verify = _config->FindB("Acquire::https::Verify-Host",true);
165 knob = "Acquire::https::"+remotehost+"::Verify-Host";
166 verify = _config->FindB(knob.c_str(),verify);
167 int const default_verify = (verify == true) ? 2 : 0;
168 curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, default_verify);
169
170 // Also enforce issuer of server certificate using its cert
171 string issuercert = _config->Find("Acquire::https::IssuerCert","");
172 knob = "Acquire::https::"+remotehost+"::IssuerCert";
173 issuercert = _config->Find(knob.c_str(),issuercert.c_str());
174 if(issuercert.empty() == false)
175 curl_easy_setopt(curl, CURLOPT_ISSUERCERT,issuercert.c_str());
176
177 // For client authentication, certificate file ...
178 string pem = _config->Find("Acquire::https::SslCert","");
179 knob = "Acquire::https::"+remotehost+"::SslCert";
180 pem = _config->Find(knob.c_str(),pem.c_str());
181 if(pem.empty() == false)
182 curl_easy_setopt(curl, CURLOPT_SSLCERT, pem.c_str());
183
184 // ... and associated key.
185 string key = _config->Find("Acquire::https::SslKey","");
186 knob = "Acquire::https::"+remotehost+"::SslKey";
187 key = _config->Find(knob.c_str(),key.c_str());
188 if(key.empty() == false)
189 curl_easy_setopt(curl, CURLOPT_SSLKEY, key.c_str());
190
191 // Allow forcing SSL version to SSLv3 or TLSv1 (SSLv2 is not
192 // supported by GnuTLS).
193 long final_version = CURL_SSLVERSION_DEFAULT;
194 string sslversion = _config->Find("Acquire::https::SslForceVersion","");
195 knob = "Acquire::https::"+remotehost+"::SslForceVersion";
196 sslversion = _config->Find(knob.c_str(),sslversion.c_str());
197 if(sslversion == "TLSv1")
198 final_version = CURL_SSLVERSION_TLSv1;
199 else if(sslversion == "SSLv3")
200 final_version = CURL_SSLVERSION_SSLv3;
201 curl_easy_setopt(curl, CURLOPT_SSLVERSION, final_version);
202
203 // CRL file
204 string crlfile = _config->Find("Acquire::https::CrlFile","");
205 knob = "Acquire::https::"+remotehost+"::CrlFile";
206 crlfile = _config->Find(knob.c_str(),crlfile.c_str());
207 if(crlfile.empty() == false)
208 curl_easy_setopt(curl, CURLOPT_CRLFILE, crlfile.c_str());
209
210 // cache-control
211 if(_config->FindB("Acquire::https::No-Cache",
212 _config->FindB("Acquire::http::No-Cache",false)) == false)
213 {
214 // cache enabled
215 if (_config->FindB("Acquire::https::No-Store",
216 _config->FindB("Acquire::http::No-Store",false)) == true)
217 headers = curl_slist_append(headers,"Cache-Control: no-store");
218 stringstream ss;
219 ioprintf(ss, "Cache-Control: max-age=%u", _config->FindI("Acquire::https::Max-Age",
220 _config->FindI("Acquire::http::Max-Age",0)));
221 headers = curl_slist_append(headers, ss.str().c_str());
222 } else {
223 // cache disabled by user
224 headers = curl_slist_append(headers, "Cache-Control: no-cache");
225 headers = curl_slist_append(headers, "Pragma: no-cache");
226 }
227 curl_easy_setopt(curl, CURLOPT_HTTPHEADER, headers);
228
229 // speed limit
230 int const dlLimit = _config->FindI("Acquire::https::Dl-Limit",
231 _config->FindI("Acquire::http::Dl-Limit",0))*1024;
232 if (dlLimit > 0)
233 curl_easy_setopt(curl, CURLOPT_MAX_RECV_SPEED_LARGE, dlLimit);
234
235 // set header
236 curl_easy_setopt(curl, CURLOPT_USERAGENT,
237 _config->Find("Acquire::https::User-Agent",
238 _config->Find("Acquire::http::User-Agent",
239 "Debian APT-CURL/1.0 (" PACKAGE_VERSION ")").c_str()).c_str());
240
241 // set timeout
242 int const timeout = _config->FindI("Acquire::https::Timeout",
243 _config->FindI("Acquire::http::Timeout",120));
244 curl_easy_setopt(curl, CURLOPT_CONNECTTIMEOUT, timeout);
245 //set really low lowspeed timeout (see #497983)
246 curl_easy_setopt(curl, CURLOPT_LOW_SPEED_LIMIT, DL_MIN_SPEED);
247 curl_easy_setopt(curl, CURLOPT_LOW_SPEED_TIME, timeout);
248
249 // set redirect options and default to 10 redirects
250 bool const AllowRedirect = _config->FindB("Acquire::https::AllowRedirect",
251 _config->FindB("Acquire::http::AllowRedirect",true));
252 curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, AllowRedirect);
253 curl_easy_setopt(curl, CURLOPT_MAXREDIRS, 10);
254
255 // debug
256 if(_config->FindB("Debug::Acquire::https", false))
257 curl_easy_setopt(curl, CURLOPT_VERBOSE, true);
258
259 // error handling
260 curl_errorstr[0] = '\0';
261 curl_easy_setopt(curl, CURLOPT_ERRORBUFFER, curl_errorstr);
262
263 // If we ask for uncompressed files servers might respond with content-
264 // negotation which lets us end up with compressed files we do not support,
265 // see 657029, 657560 and co, so if we have no extension on the request
266 // ask for text only. As a sidenote: If there is nothing to negotate servers
267 // seem to be nice and ignore it.
268 if (_config->FindB("Acquire::https::SendAccept", _config->FindB("Acquire::http::SendAccept", true)) == true)
269 {
270 size_t const filepos = Itm->Uri.find_last_of('/');
271 string const file = Itm->Uri.substr(filepos + 1);
272 if (flExtension(file) == file)
273 headers = curl_slist_append(headers, "Accept: text/*");
274 }
275
276 // if we have the file send an if-range query with a range header
277 if (stat(Itm->DestFile.c_str(),&SBuf) >= 0 && SBuf.st_size > 0)
278 {
279 char Buf[1000];
280 sprintf(Buf, "Range: bytes=%li-", (long) SBuf.st_size - 1);
281 headers = curl_slist_append(headers, Buf);
282 sprintf(Buf, "If-Range: %s", TimeRFC1123(SBuf.st_mtime).c_str());
283 headers = curl_slist_append(headers, Buf);
284 }
285 else if(Itm->LastModified > 0)
286 {
287 curl_easy_setopt(curl, CURLOPT_TIMECONDITION, CURL_TIMECOND_IFMODSINCE);
288 curl_easy_setopt(curl, CURLOPT_TIMEVALUE, Itm->LastModified);
289 }
290
291 // go for it - if the file exists, append on it
292 File = new FileFd(Itm->DestFile, FileFd::WriteAny);
293 if (File->Size() > 0)
294 File->Seek(File->Size() - 1);
295
296 // keep apt updated
297 Res.Filename = Itm->DestFile;
298
299 // get it!
300 CURLcode success = curl_easy_perform(curl);
301 curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &curl_responsecode);
302
303 long curl_servdate;
304 curl_easy_getinfo(curl, CURLINFO_FILETIME, &curl_servdate);
305
306 // If the server returns 200 OK but the If-Modified-Since condition is not
307 // met, CURLINFO_CONDITION_UNMET will be set to 1
308 long curl_condition_unmet = 0;
309 curl_easy_getinfo(curl, CURLINFO_CONDITION_UNMET, &curl_condition_unmet);
310
311 File->Close();
312
313 // cleanup
314 if(success != 0 || (curl_responsecode != 200 && curl_responsecode != 304))
315 {
316 _error->Error("%s", curl_errorstr);
317 // unlink, no need keep 401/404 page content in partial/
318 unlink(File->Name().c_str());
319 Fail();
320 return true;
321 }
322
323 // Timestamp
324 struct utimbuf UBuf;
325 if (curl_servdate != -1) {
326 UBuf.actime = curl_servdate;
327 UBuf.modtime = curl_servdate;
328 utime(File->Name().c_str(),&UBuf);
329 }
330
331 // check the downloaded result
332 struct stat Buf;
333 if (stat(File->Name().c_str(),&Buf) == 0)
334 {
335 Res.Filename = File->Name();
336 Res.LastModified = Buf.st_mtime;
337 Res.IMSHit = false;
338 if (curl_responsecode == 304 || curl_condition_unmet)
339 {
340 unlink(File->Name().c_str());
341 Res.IMSHit = true;
342 Res.LastModified = Itm->LastModified;
343 Res.Size = 0;
344 URIDone(Res);
345 return true;
346 }
347 Res.Size = Buf.st_size;
348 }
349
350 // take hashes
351 Hashes Hash;
352 FileFd Fd(Res.Filename, FileFd::ReadOnly);
353 Hash.AddFD(Fd);
354 Res.TakeHashes(Hash);
355
356 // keep apt updated
357 URIDone(Res);
358
359 // cleanup
360 Res.Size = 0;
361 delete File;
362 curl_slist_free_all(headers);
363
364 return true;
365 };
366
367 int main()
368 {
369 setlocale(LC_ALL, "");
370
371 HttpsMethod Mth;
372 curl_global_init(CURL_GLOBAL_SSL) ;
373
374 return Mth.Run();
375 }
376
377