]>
Commit | Line | Data |
---|---|---|
b2e465d6 AL |
1 | // -*- mode: cpp; mode: fold -*- |
2 | // Description /*{{{*/ | |
584e4558 | 3 | // $Id: extracttar.cc,v 1.6 2002/11/11 06:55:50 doogie Exp $ |
b2e465d6 AL |
4 | /* ###################################################################### |
5 | ||
6 | Extract a Tar - Tar Extractor | |
7 | ||
8 | Some performance measurements showed that zlib performed quite poorly | |
9 | in comparision to a forked gzip process. This tar extractor makes use | |
10 | of the fact that dup'd file descriptors have the same seek pointer | |
11 | and that gzip will not read past the end of a compressed stream, | |
12 | even if there is more data. We use the dup property to track extraction | |
13 | progress and the gzip feature to just feed gzip a fd in the middle | |
14 | of an AR file. | |
15 | ||
16 | ##################################################################### */ | |
17 | /*}}}*/ | |
18 | // Include Files /*{{{*/ | |
19 | #ifdef __GNUG__ | |
20 | #pragma implementation "apt-pkg/extracttar.h" | |
21 | #endif | |
22 | #include <apt-pkg/extracttar.h> | |
23 | ||
24 | #include <apt-pkg/error.h> | |
25 | #include <apt-pkg/strutl.h> | |
26 | #include <apt-pkg/configuration.h> | |
27 | #include <system.h> | |
28 | ||
29 | #include <stdlib.h> | |
30 | #include <unistd.h> | |
31 | #include <signal.h> | |
32 | #include <fcntl.h> | |
90f057fd | 33 | #include <iostream> |
b2e465d6 | 34 | /*}}}*/ |
584e4558 | 35 | using namespace std; |
4520bfdf | 36 | |
b2e465d6 AL |
37 | // The on disk header for a tar file. |
38 | struct ExtractTar::TarHeader | |
39 | { | |
40 | char Name[100]; | |
41 | char Mode[8]; | |
42 | char UserID[8]; | |
43 | char GroupID[8]; | |
44 | char Size[12]; | |
45 | char MTime[12]; | |
46 | char Checksum[8]; | |
47 | char LinkFlag; | |
48 | char LinkName[100]; | |
49 | char MagicNumber[8]; | |
50 | char UserName[32]; | |
51 | char GroupName[32]; | |
52 | char Major[8]; | |
53 | char Minor[8]; | |
54 | }; | |
55 | ||
56 | // ExtractTar::ExtractTar - Constructor /*{{{*/ | |
57 | // --------------------------------------------------------------------- | |
58 | /* */ | |
59 | ExtractTar::ExtractTar(FileFd &Fd,unsigned long Max) : File(Fd), | |
60 | MaxInSize(Max) | |
61 | ||
62 | { | |
63 | GZPid = -1; | |
64 | InFd = -1; | |
65 | Eof = false; | |
66 | } | |
67 | /*}}}*/ | |
68 | // ExtractTar::ExtractTar - Destructor /*{{{*/ | |
69 | // --------------------------------------------------------------------- | |
70 | /* */ | |
71 | ExtractTar::~ExtractTar() | |
72 | { | |
4520bfdf AL |
73 | // Error close |
74 | Done(true); | |
b2e465d6 AL |
75 | } |
76 | /*}}}*/ | |
77 | // ExtractTar::Done - Reap the gzip sub process /*{{{*/ | |
78 | // --------------------------------------------------------------------- | |
79 | /* If the force flag is given then error messages are suppressed - this | |
80 | means we hit the end of the tar file but there was still gzip data. */ | |
81 | bool ExtractTar::Done(bool Force) | |
82 | { | |
83 | InFd.Close(); | |
84 | if (GZPid <= 0) | |
85 | return true; | |
86 | ||
87 | /* If there is a pending error then we are cleaning up gzip and are | |
88 | not interested in it's failures */ | |
89 | if (_error->PendingError() == true) | |
90 | Force = true; | |
91 | ||
92 | // Make sure we clean it up! | |
93 | kill(GZPid,SIGINT); | |
94 | if (ExecWait(GZPid,_config->Find("dir::bin::gzip","/bin/gzip").c_str(), | |
95 | Force) == false) | |
96 | { | |
97 | GZPid = -1; | |
98 | return Force; | |
99 | } | |
100 | ||
101 | GZPid = -1; | |
102 | return true; | |
103 | } | |
104 | /*}}}*/ | |
105 | // ExtractTar::StartGzip - Startup gzip /*{{{*/ | |
106 | // --------------------------------------------------------------------- | |
107 | /* This creates a gzip sub process that has its input as the file itself. | |
108 | If this tar file is embedded into something like an ar file then | |
109 | gzip will efficiently ignore the extra bits. */ | |
110 | bool ExtractTar::StartGzip() | |
111 | { | |
112 | int Pipes[2]; | |
113 | if (pipe(Pipes) != 0) | |
114 | return _error->Errno("pipe","Failed to create pipes"); | |
115 | ||
116 | // Fork off the process | |
117 | GZPid = ExecFork(); | |
118 | ||
119 | // Spawn the subprocess | |
120 | if (GZPid == 0) | |
121 | { | |
122 | // Setup the FDs | |
123 | dup2(Pipes[1],STDOUT_FILENO); | |
124 | dup2(File.Fd(),STDIN_FILENO); | |
125 | int Fd = open("/dev/null",O_RDWR); | |
126 | if (Fd == -1) | |
127 | _exit(101); | |
128 | dup2(Fd,STDERR_FILENO); | |
129 | close(Fd); | |
130 | SetCloseExec(STDOUT_FILENO,false); | |
131 | SetCloseExec(STDIN_FILENO,false); | |
132 | SetCloseExec(STDERR_FILENO,false); | |
133 | ||
134 | const char *Args[3]; | |
135 | Args[0] = _config->Find("dir::bin::gzip","/bin/gzip").c_str(); | |
136 | Args[1] = "-d"; | |
137 | Args[2] = 0; | |
138 | execv(Args[0],(char **)Args); | |
139 | cerr << "Failed to exec gzip " << Args[0] << endl; | |
140 | _exit(100); | |
141 | } | |
142 | ||
143 | // Fix up our FDs | |
144 | InFd.Fd(Pipes[0]); | |
145 | close(Pipes[1]); | |
146 | return true; | |
147 | } | |
148 | /*}}}*/ | |
149 | // ExtractTar::Go - Perform extraction /*{{{*/ | |
150 | // --------------------------------------------------------------------- | |
151 | /* This reads each 512 byte block from the archive and extracts the header | |
152 | information into the Item structure. Then it resolves the UID/GID and | |
153 | invokes the correct processing function. */ | |
154 | bool ExtractTar::Go(pkgDirStream &Stream) | |
155 | { | |
156 | if (StartGzip() == false) | |
157 | return false; | |
158 | ||
159 | // Loop over all blocks | |
160 | string LastLongLink; | |
161 | string LastLongName; | |
162 | while (1) | |
163 | { | |
164 | bool BadRecord = false; | |
165 | unsigned char Block[512]; | |
166 | if (InFd.Read(Block,sizeof(Block),true) == false) | |
167 | return false; | |
168 | ||
169 | if (InFd.Eof() == true) | |
170 | break; | |
171 | ||
172 | // Get the checksum | |
173 | TarHeader *Tar = (TarHeader *)Block; | |
174 | unsigned long CheckSum; | |
175 | if (StrToNum(Tar->Checksum,CheckSum,sizeof(Tar->Checksum),8) == false) | |
176 | return _error->Error("Corrupted archive"); | |
177 | ||
178 | /* Compute the checksum field. The actual checksum is blanked out | |
179 | with spaces so it is not included in the computation */ | |
180 | unsigned long NewSum = 0; | |
181 | memset(Tar->Checksum,' ',sizeof(Tar->Checksum)); | |
182 | for (int I = 0; I != sizeof(Block); I++) | |
183 | NewSum += Block[I]; | |
184 | ||
185 | /* Check for a block of nulls - in this case we kill gzip, GNU tar | |
186 | does this.. */ | |
187 | if (NewSum == ' '*sizeof(Tar->Checksum)) | |
188 | return Done(true); | |
189 | ||
190 | if (NewSum != CheckSum) | |
191 | return _error->Error("Tar Checksum failed, archive corrupted"); | |
192 | ||
193 | // Decode all of the fields | |
194 | pkgDirStream::Item Itm; | |
b2e465d6 | 195 | if (StrToNum(Tar->Mode,Itm.Mode,sizeof(Tar->Mode),8) == false || |
4520bfdf AL |
196 | StrToNum(Tar->UserID,Itm.UID,sizeof(Tar->UserID),8) == false || |
197 | StrToNum(Tar->GroupID,Itm.GID,sizeof(Tar->GroupID),8) == false || | |
b2e465d6 AL |
198 | StrToNum(Tar->Size,Itm.Size,sizeof(Tar->Size),8) == false || |
199 | StrToNum(Tar->MTime,Itm.MTime,sizeof(Tar->MTime),8) == false || | |
200 | StrToNum(Tar->Major,Itm.Major,sizeof(Tar->Major),8) == false || | |
201 | StrToNum(Tar->Minor,Itm.Minor,sizeof(Tar->Minor),8) == false) | |
202 | return _error->Error("Corrupted archive"); | |
203 | ||
204 | // Grab the filename | |
205 | if (LastLongName.empty() == false) | |
206 | Itm.Name = (char *)LastLongName.c_str(); | |
207 | else | |
208 | { | |
209 | Tar->Name[sizeof(Tar->Name)] = 0; | |
210 | Itm.Name = Tar->Name; | |
211 | } | |
212 | if (Itm.Name[0] == '.' && Itm.Name[1] == '/' && Itm.Name[2] != 0) | |
213 | Itm.Name += 2; | |
214 | ||
215 | // Grab the link target | |
216 | Tar->Name[sizeof(Tar->LinkName)] = 0; | |
217 | Itm.LinkTarget = Tar->LinkName; | |
218 | ||
219 | if (LastLongLink.empty() == false) | |
220 | Itm.LinkTarget = (char *)LastLongLink.c_str(); | |
221 | ||
222 | // Convert the type over | |
223 | switch (Tar->LinkFlag) | |
224 | { | |
225 | case NormalFile0: | |
226 | case NormalFile: | |
227 | Itm.Type = pkgDirStream::Item::File; | |
228 | break; | |
229 | ||
230 | case HardLink: | |
231 | Itm.Type = pkgDirStream::Item::HardLink; | |
232 | break; | |
233 | ||
234 | case SymbolicLink: | |
235 | Itm.Type = pkgDirStream::Item::SymbolicLink; | |
236 | break; | |
237 | ||
238 | case CharacterDevice: | |
239 | Itm.Type = pkgDirStream::Item::CharDevice; | |
240 | break; | |
241 | ||
242 | case BlockDevice: | |
243 | Itm.Type = pkgDirStream::Item::BlockDevice; | |
244 | break; | |
245 | ||
246 | case Directory: | |
247 | Itm.Type = pkgDirStream::Item::Directory; | |
248 | break; | |
249 | ||
250 | case FIFO: | |
251 | Itm.Type = pkgDirStream::Item::FIFO; | |
252 | break; | |
253 | ||
254 | case GNU_LongLink: | |
255 | { | |
256 | unsigned long Length = Itm.Size; | |
257 | unsigned char Block[512]; | |
258 | while (Length > 0) | |
259 | { | |
260 | if (InFd.Read(Block,sizeof(Block),true) == false) | |
261 | return false; | |
262 | if (Length <= sizeof(Block)) | |
263 | { | |
264 | LastLongLink.append(Block,Block+sizeof(Block)); | |
265 | break; | |
266 | } | |
267 | LastLongLink.append(Block,Block+sizeof(Block)); | |
268 | Length -= sizeof(Block); | |
269 | } | |
270 | continue; | |
271 | } | |
272 | ||
273 | case GNU_LongName: | |
274 | { | |
275 | unsigned long Length = Itm.Size; | |
276 | unsigned char Block[512]; | |
277 | while (Length > 0) | |
278 | { | |
279 | if (InFd.Read(Block,sizeof(Block),true) == false) | |
280 | return false; | |
281 | if (Length < sizeof(Block)) | |
282 | { | |
283 | LastLongName.append(Block,Block+sizeof(Block)); | |
284 | break; | |
285 | } | |
286 | LastLongName.append(Block,Block+sizeof(Block)); | |
287 | Length -= sizeof(Block); | |
288 | } | |
289 | continue; | |
290 | } | |
291 | ||
292 | default: | |
293 | BadRecord = true; | |
294 | _error->Warning("Unkown TAR header type %u, member %s",(unsigned)Tar->LinkFlag,Tar->Name); | |
295 | break; | |
296 | } | |
297 | ||
298 | int Fd = -1; | |
299 | if (BadRecord == false) | |
300 | if (Stream.DoItem(Itm,Fd) == false) | |
301 | return false; | |
302 | ||
303 | // Copy the file over the FD | |
304 | unsigned long Size = Itm.Size; | |
305 | while (Size != 0) | |
306 | { | |
307 | unsigned char Junk[32*1024]; | |
308 | unsigned long Read = MIN(Size,sizeof(Junk)); | |
309 | if (InFd.Read(Junk,((Read+511)/512)*512) == false) | |
310 | return false; | |
311 | ||
312 | if (BadRecord == false) | |
313 | { | |
314 | if (Fd > 0) | |
315 | { | |
316 | if (write(Fd,Junk,Read) != (signed)Read) | |
317 | return Stream.Fail(Itm,Fd); | |
318 | } | |
319 | else | |
320 | { | |
321 | /* An Fd of -2 means to send to a special processing | |
322 | function */ | |
323 | if (Fd == -2) | |
324 | if (Stream.Process(Itm,Junk,Read,Itm.Size - Size) == false) | |
325 | return Stream.Fail(Itm,Fd); | |
326 | } | |
327 | } | |
328 | ||
329 | Size -= Read; | |
330 | } | |
331 | ||
332 | // And finish up | |
333 | if (Itm.Size != 0 && BadRecord == false) | |
334 | if (Stream.FinishedFile(Itm,Fd) == false) | |
335 | return false; | |
336 | ||
337 | LastLongName.erase(); | |
338 | LastLongLink.erase(); | |
339 | } | |
340 | ||
341 | return Done(false); | |
342 | } | |
343 | /*}}}*/ |