[apt.git] / apt-inst / contrib / extracttar.cc

// -*- mode: cpp; mode: fold -*-
// Description								/*{{{*/
// $Id: extracttar.cc,v 1.3 2001/05/27 23:47:09 jgg Exp $
/* ######################################################################

   Extract a Tar - Tar Extractor

   Some performance measurements showed that zlib performed quite poorly
   in comparision to a forked gzip process. This tar extractor makes use
   of the fact that dup'd file descriptors have the same seek pointer
   and that gzip will not read past the end of a compressed stream, 
   even if there is more data. We use the dup property to track extraction
   progress and the gzip feature to just feed gzip a fd in the middle
   of an AR file. 
   
   ##################################################################### */
									/*}}}*/
// Include Files							/*{{{*/
#ifdef __GNUG__
#pragma implementation "apt-pkg/extracttar.h"
#endif
#include <apt-pkg/extracttar.h>

#include <apt-pkg/error.h>
#include <apt-pkg/strutl.h>
#include <apt-pkg/configuration.h>
#include <system.h>

#include <stdlib.h>
#include <unistd.h>
#include <signal.h>
#include <fcntl.h>
#include <iostream.h>
									/*}}}*/

// The on disk header for a tar file.
struct ExtractTar::TarHeader
{
   char Name[100];
   char Mode[8];
   char UserID[8];
   char GroupID[8];
   char Size[12];
   char MTime[12];
   char Checksum[8];
   char LinkFlag;
   char LinkName[100];
   char MagicNumber[8];
   char UserName[32];
   char GroupName[32];
   char Major[8];
   char Minor[8];      
};
   
// ExtractTar::ExtractTar - Constructor					/*{{{*/
// ---------------------------------------------------------------------
/* */
ExtractTar::ExtractTar(FileFd &Fd,unsigned long Max) : File(Fd), 
                         MaxInSize(Max)

{
   GZPid = -1;
   InFd = -1;
   Eof = false;
}
									/*}}}*/
// ExtractTar::ExtractTar - Destructor					/*{{{*/
// ---------------------------------------------------------------------
/* */
ExtractTar::~ExtractTar()
{
   Done(false);
}
									/*}}}*/
// ExtractTar::Done - Reap the gzip sub process				/*{{{*/
// ---------------------------------------------------------------------
/* If the force flag is given then error messages are suppressed - this
   means we hit the end of the tar file but there was still gzip data. */
bool ExtractTar::Done(bool Force)
{
   InFd.Close();
   if (GZPid <= 0)
      return true;

   /* If there is a pending error then we are cleaning up gzip and are
      not interested in it's failures */
   if (_error->PendingError() == true)
      Force = true;
   
   // Make sure we clean it up!
   kill(GZPid,SIGINT);
   if (ExecWait(GZPid,_config->Find("dir::bin::gzip","/bin/gzip").c_str(),
		Force) == false)
   {
      GZPid = -1;
      return Force;
   }
   
   GZPid = -1;
   return true;
}
									/*}}}*/
// ExtractTar::StartGzip - Startup gzip					/*{{{*/
// ---------------------------------------------------------------------
/* This creates a gzip sub process that has its input as the file itself.
   If this tar file is embedded into something like an ar file then 
   gzip will efficiently ignore the extra bits. */
bool ExtractTar::StartGzip()
{
   int Pipes[2];
   if (pipe(Pipes) != 0)
      return _error->Errno("pipe","Failed to create pipes");
   
   // Fork off the process
   GZPid = ExecFork();

   // Spawn the subprocess
   if (GZPid == 0)
   {
      // Setup the FDs
      dup2(Pipes[1],STDOUT_FILENO);
      dup2(File.Fd(),STDIN_FILENO);
      int Fd = open("/dev/null",O_RDWR);
      if (Fd == -1)
	 _exit(101);
      dup2(Fd,STDERR_FILENO);
      close(Fd);
      SetCloseExec(STDOUT_FILENO,false);
      SetCloseExec(STDIN_FILENO,false);      
      SetCloseExec(STDERR_FILENO,false);
      
      const char *Args[3];
      Args[0] = _config->Find("dir::bin::gzip","/bin/gzip").c_str();
      Args[1] = "-d";
      Args[2] = 0;
      execv(Args[0],(char **)Args);
      cerr << "Failed to exec gzip " << Args[0] << endl;
      _exit(100);
   }

   // Fix up our FDs
   InFd.Fd(Pipes[0]);
   close(Pipes[1]);
   return true;
}
									/*}}}*/
// ExtractTar::Go - Perform extraction					/*{{{*/
// ---------------------------------------------------------------------
/* This reads each 512 byte block from the archive and extracts the header
   information into the Item structure. Then it resolves the UID/GID and
   invokes the correct processing function. */
bool ExtractTar::Go(pkgDirStream &Stream)
{   
   if (StartGzip() == false)
      return false;
   
   // Loop over all blocks
   string LastLongLink;
   string LastLongName;
   while (1)
   {
      bool BadRecord = false;      
      unsigned char Block[512];      
      if (InFd.Read(Block,sizeof(Block),true) == false)
	 return false;
      
      if (InFd.Eof() == true)
	 break;

      // Get the checksum
      TarHeader *Tar = (TarHeader *)Block;
      unsigned long CheckSum;
      if (StrToNum(Tar->Checksum,CheckSum,sizeof(Tar->Checksum),8) == false)
	 return _error->Error("Corrupted archive");
      
      /* Compute the checksum field. The actual checksum is blanked out
         with spaces so it is not included in the computation */
      unsigned long NewSum = 0;
      memset(Tar->Checksum,' ',sizeof(Tar->Checksum));
      for (int I = 0; I != sizeof(Block); I++)
	 NewSum += Block[I];
      
      /* Check for a block of nulls - in this case we kill gzip, GNU tar
       	 does this.. */
      if (NewSum == ' '*sizeof(Tar->Checksum))
	 return Done(true);
      
      if (NewSum != CheckSum)
	 return _error->Error("Tar Checksum failed, archive corrupted");
   
      // Decode all of the fields
      pkgDirStream::Item Itm;
      unsigned long UID;
      unsigned long GID;
      if (StrToNum(Tar->Mode,Itm.Mode,sizeof(Tar->Mode),8) == false ||
	  StrToNum(Tar->UserID,UID,sizeof(Tar->UserID),8) == false ||
	  StrToNum(Tar->GroupID,GID,sizeof(Tar->GroupID),8) == false ||
	  StrToNum(Tar->Size,Itm.Size,sizeof(Tar->Size),8) == false ||
	  StrToNum(Tar->MTime,Itm.MTime,sizeof(Tar->MTime),8) == false ||
	  StrToNum(Tar->Major,Itm.Major,sizeof(Tar->Major),8) == false ||
	  StrToNum(Tar->Minor,Itm.Minor,sizeof(Tar->Minor),8) == false)
	 return _error->Error("Corrupted archive");
      
      // Grab the filename
      if (LastLongName.empty() == false)
	 Itm.Name = (char *)LastLongName.c_str();
      else
      {
	 Tar->Name[sizeof(Tar->Name)] = 0;
	 Itm.Name = Tar->Name;
      }      
      if (Itm.Name[0] == '.' && Itm.Name[1] == '/' && Itm.Name[2] != 0)
	 Itm.Name += 2;
      
      // Grab the link target
      Tar->Name[sizeof(Tar->LinkName)] = 0;
      Itm.LinkTarget = Tar->LinkName;

      if (LastLongLink.empty() == false)
	 Itm.LinkTarget = (char *)LastLongLink.c_str();
      
      // Convert the type over
      switch (Tar->LinkFlag)
      {
	 case NormalFile0:
	 case NormalFile:
	 Itm.Type = pkgDirStream::Item::File;
	 break;
	 
	 case HardLink:
	 Itm.Type = pkgDirStream::Item::HardLink;
	 break;
	 
	 case SymbolicLink:
	 Itm.Type = pkgDirStream::Item::SymbolicLink;
	 break;
	 
	 case CharacterDevice:
	 Itm.Type = pkgDirStream::Item::CharDevice;
	 break;
	    
	 case BlockDevice:
	 Itm.Type = pkgDirStream::Item::BlockDevice;
	 break;
	 
	 case Directory:
	 Itm.Type = pkgDirStream::Item::Directory;
	 break;
	 
	 case FIFO:
	 Itm.Type = pkgDirStream::Item::FIFO;
	 break;

	 case GNU_LongLink:
	 {
	    unsigned long Length = Itm.Size;
	    unsigned char Block[512];
	    while (Length > 0)
	    {
	       if (InFd.Read(Block,sizeof(Block),true) == false)
		  return false;
	       if (Length <= sizeof(Block))
	       {
		  LastLongLink.append(Block,Block+sizeof(Block));
		  break;
	       }	       
	       LastLongLink.append(Block,Block+sizeof(Block));
	       Length -= sizeof(Block);
	    }
	    continue;
	 }
	 
	 case GNU_LongName:
	 {
	    unsigned long Length = Itm.Size;
	    unsigned char Block[512];
	    while (Length > 0)
	    {
	       if (InFd.Read(Block,sizeof(Block),true) == false)
		  return false;
	       if (Length < sizeof(Block))
	       {
		  LastLongName.append(Block,Block+sizeof(Block));
		  break;
	       }	       
	       LastLongName.append(Block,Block+sizeof(Block));
	       Length -= sizeof(Block);
	    }
	    continue;
	 }
	 
	 default:
	 BadRecord = true;
	 _error->Warning("Unkown TAR header type %u, member %s",(unsigned)Tar->LinkFlag,Tar->Name);
	 break;
      }
      
      int Fd = -1;
      if (BadRecord == false)
	 if (Stream.DoItem(Itm,Fd) == false)
	    return false;
      
      // Copy the file over the FD
      unsigned long Size = Itm.Size;
      while (Size != 0)
      {
	 unsigned char Junk[32*1024];
	 unsigned long Read = MIN(Size,sizeof(Junk));
	 if (InFd.Read(Junk,((Read+511)/512)*512) == false)
	    return false;
	 
	 if (BadRecord == false)
	 {
	    if (Fd > 0)
	    {
	       if (write(Fd,Junk,Read) != (signed)Read)
		  return Stream.Fail(Itm,Fd);
	    }
	    else
	    {
	       /* An Fd of -2 means to send to a special processing
		  function */
	       if (Fd == -2)
		  if (Stream.Process(Itm,Junk,Read,Itm.Size - Size) == false)
		     return Stream.Fail(Itm,Fd);
	    }
	 }
	 
	 Size -= Read;
      }
      
      // And finish up
      if (Itm.Size != 0 && BadRecord == false)
	 if (Stream.FinishedFile(Itm,Fd) == false)
	    return false;
      
      LastLongName.erase();
      LastLongLink.erase();
   }
   
   return Done(false);
}
									/*}}}*/
Commit	Line	Data
b2e465d6 AL	1	// -- mode: cpp; mode: fold --
b2e465d6 AL	2	// Description /{{{/
18255546	3	// $Id: extracttar.cc,v 1.3 2001/05/27 23:47:09 jgg 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>
18255546	33	#include <iostream.h>
b2e465d6 AL	34	/}}}/
	35
	36	// The on disk header for a tar file.
	37	struct ExtractTar::TarHeader
	38	{
	39	char Name[100];
	40	char Mode[8];
	41	char UserID[8];
	42	char GroupID[8];
	43	char Size[12];
	44	char MTime[12];
	45	char Checksum[8];
	46	char LinkFlag;
	47	char LinkName[100];
	48	char MagicNumber[8];
	49	char UserName[32];
	50	char GroupName[32];
	51	char Major[8];
	52	char Minor[8];
	53	};
	54
	55	// ExtractTar::ExtractTar - Constructor /{{{/
	56	// ---------------------------------------------------------------------
	57	/* */
	58	ExtractTar::ExtractTar(FileFd &Fd,unsigned long Max) : File(Fd),
	59	MaxInSize(Max)
	60
	61	{
	62	GZPid = -1;
	63	InFd = -1;
	64	Eof = false;
	65	}
	66	/}}}/
	67	// ExtractTar::ExtractTar - Destructor /{{{/
	68	// ---------------------------------------------------------------------
	69	/* */
	70	ExtractTar::~ExtractTar()
	71	{
	72	Done(false);
	73	}
	74	/}}}/
	75	// ExtractTar::Done - Reap the gzip sub process /{{{/
	76	// ---------------------------------------------------------------------
	77	/* If the force flag is given then error messages are suppressed - this
	78	means we hit the end of the tar file but there was still gzip data. */
	79	bool ExtractTar::Done(bool Force)
	80	{
	81	InFd.Close();
	82	if (GZPid <= 0)
	83	return true;
	84
	85	/* If there is a pending error then we are cleaning up gzip and are
	86	not interested in it's failures */
	87	if (_error->PendingError() == true)
	88	Force = true;
	89
	90	// Make sure we clean it up!
	91	kill(GZPid,SIGINT);
	92	if (ExecWait(GZPid,_config->Find("dir::bin::gzip","/bin/gzip").c_str(),
	93	Force) == false)
	94	{
	95	GZPid = -1;
	96	return Force;
	97	}
98
99	GZPid = -1;
100	return true;
101	}
102	/}}}/
103	// ExtractTar::StartGzip - Startup gzip /{{{/
104	// ---------------------------------------------------------------------
105	/* This creates a gzip sub process that has its input as the file itself.
106	If this tar file is embedded into something like an ar file then
107	gzip will efficiently ignore the extra bits. */
108	bool ExtractTar::StartGzip()
109	{
110	int Pipes[2];
111	if (pipe(Pipes) != 0)
112	return _error->Errno("pipe","Failed to create pipes");
113
114	// Fork off the process
115	GZPid = ExecFork();
116
117	// Spawn the subprocess
118	if (GZPid == 0)
119	{
120	// Setup the FDs
121	dup2(Pipes[1],STDOUT_FILENO);
122	dup2(File.Fd(),STDIN_FILENO);
123	int Fd = open("/dev/null",O_RDWR);
124	if (Fd == -1)
125	_exit(101);
126	dup2(Fd,STDERR_FILENO);
127	close(Fd);
128	SetCloseExec(STDOUT_FILENO,false);
129	SetCloseExec(STDIN_FILENO,false);
130	SetCloseExec(STDERR_FILENO,false);
131
132	const char *Args[3];
133	Args[0] = _config->Find("dir::bin::gzip","/bin/gzip").c_str();
134	Args[1] = "-d";
135	Args[2] = 0;
136	execv(Args[0],(char **)Args);
137	cerr << "Failed to exec gzip " << Args[0] << endl;
138	_exit(100);
139	}
140
141	// Fix up our FDs
142	InFd.Fd(Pipes[0]);
143	close(Pipes[1]);
144	return true;
145	}
146	/}}}/
147	// ExtractTar::Go - Perform extraction /{{{/
148	// ---------------------------------------------------------------------
149	/* This reads each 512 byte block from the archive and extracts the header
150	information into the Item structure. Then it resolves the UID/GID and
151	invokes the correct processing function. */
152	bool ExtractTar::Go(pkgDirStream &Stream)
153	{
154	if (StartGzip() == false)
155	return false;
156
157	// Loop over all blocks
158	string LastLongLink;
159	string LastLongName;
160	while (1)
161	{
162	bool BadRecord = false;
163	unsigned char Block[512];
164	if (InFd.Read(Block,sizeof(Block),true) == false)
165	return false;
166
167	if (InFd.Eof() == true)
168	break;
169
170	// Get the checksum
171	TarHeader Tar = (TarHeader )Block;
172	unsigned long CheckSum;
173	if (StrToNum(Tar->Checksum,CheckSum,sizeof(Tar->Checksum),8) == false)
174	return _error->Error("Corrupted archive");
175
176	/* Compute the checksum field. The actual checksum is blanked out
177	with spaces so it is not included in the computation */
178	unsigned long NewSum = 0;
179	memset(Tar->Checksum,' ',sizeof(Tar->Checksum));
180	for (int I = 0; I != sizeof(Block); I++)
181	NewSum += Block[I];
182
183	/* Check for a block of nulls - in this case we kill gzip, GNU tar
184	does this.. */
185	if (NewSum == ' '*sizeof(Tar->Checksum))
186	return Done(true);
187
188	if (NewSum != CheckSum)
189	return _error->Error("Tar Checksum failed, archive corrupted");
190
191	// Decode all of the fields
192	pkgDirStream::Item Itm;
193	unsigned long UID;
194	unsigned long GID;
195	if (StrToNum(Tar->Mode,Itm.Mode,sizeof(Tar->Mode),8) == false \|\|
196	StrToNum(Tar->UserID,UID,sizeof(Tar->UserID),8) == false \|\|
197	StrToNum(Tar->GroupID,GID,sizeof(Tar->GroupID),8) == false \|\|
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	/}}}/