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

// -*- mode: cpp; mode: fold -*-
// Description								/*{{{*/
// $Id: extracttar.cc,v 1.8.2.1 2004/01/16 18:58:50 mdz 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							/*{{{*/
#include<config.h>

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

#include <stdlib.h>
#include <unistd.h>
#include <signal.h>
#include <fcntl.h>
#include <iostream>

#include <apti18n.h>
									/*}}}*/

using namespace std;
    
// 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,string DecompressionProgram) : File(Fd), 
                         MaxInSize(Max), DecompressProg(DecompressionProgram)

{
   GZPid = -1;
   InFd = -1;
   Eof = false;
}
									/*}}}*/
// ExtractTar::ExtractTar - Destructor					/*{{{*/
// ---------------------------------------------------------------------
/* */
ExtractTar::~ExtractTar()
{
   // Error close
   Done(true);
}
									/*}}}*/
// 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);
   string confvar = string("dir::bin::") + DecompressProg;
   if (ExecWait(GZPid,_config->Find(confvar.c_str(),DecompressProg.c_str()).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];
      string confvar = string("dir::bin::") + DecompressProg;
      string argv0 = _config->Find(confvar.c_str(),DecompressProg.c_str());
      Args[0] = argv0.c_str();
      Args[1] = "-d";
      Args[2] = 0;
      execvp(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;
      if (StrToNum(Tar->Mode,Itm.Mode,sizeof(Tar->Mode),8) == false ||
          (Base256ToNum(Tar->UserID,Itm.UID,8) == false &&
	     StrToNum(Tar->UserID,Itm.UID,sizeof(Tar->UserID),8) == false) ||
          (Base256ToNum(Tar->GroupID,Itm.GID,8) == false &&
	     StrToNum(Tar->GroupID,Itm.GID,sizeof(Tar->GroupID),8) == false) ||
          (Base256ToNum(Tar->Size,Itm.Size,12) == false &&
	     StrToNum(Tar->Size,Itm.Size,sizeof(Tar->Size),8) == false) ||
          (Base256ToNum(Tar->MTime,Itm.MTime,12) == 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)-1] = 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)-1] = 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(_("Unknown 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,(unsigned long)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 (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 /{{{/
7db98ffc	3	// $Id: extracttar.cc,v 1.8.2.1 2004/01/16 18:58:50 mdz 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 /{{{/
ea542140	19	#include<config.h>
b2e465d6	20
ea542140	21	#include <apt-pkg/extracttar.h>
b2e465d6 AL	22	#include <apt-pkg/error.h>
	23	#include <apt-pkg/strutl.h>
	24	#include <apt-pkg/configuration.h>
aea7f4c8	25	#include <apt-pkg/macros.h>
b2e465d6 AL	26
	27	#include <stdlib.h>
	28	#include <unistd.h>
	29	#include <signal.h>
	30	#include <fcntl.h>
90f057fd	31	#include <iostream>
ea542140	32
d77559ac	33	#include <apti18n.h>
b2e465d6	34	/}}}/
d77559ac	35
584e4558	36	using namespace std;
4520bfdf	37
b2e465d6 AL	38	// The on disk header for a tar file.
	39	struct ExtractTar::TarHeader
	40	{
	41	char Name[100];
	42	char Mode[8];
	43	char UserID[8];
	44	char GroupID[8];
	45	char Size[12];
	46	char MTime[12];
	47	char Checksum[8];
	48	char LinkFlag;
	49	char LinkName[100];
	50	char MagicNumber[8];
	51	char UserName[32];
	52	char GroupName[32];
	53	char Major[8];
	54	char Minor[8];
	55	};
	56
	57	// ExtractTar::ExtractTar - Constructor /{{{/
	58	// ---------------------------------------------------------------------
	59	/* */
b21c0438 MZ	60	ExtractTar::ExtractTar(FileFd &Fd,unsigned long Max,string DecompressionProgram) : File(Fd),
b21c0438 MZ	61	MaxInSize(Max), DecompressProg(DecompressionProgram)
b2e465d6 AL	62
	63	{
	64	GZPid = -1;
	65	InFd = -1;
	66	Eof = false;
	67	}
	68	/}}}/
	69	// ExtractTar::ExtractTar - Destructor /{{{/
	70	// ---------------------------------------------------------------------
	71	/* */
	72	ExtractTar::~ExtractTar()
	73	{
4520bfdf AL	74	// Error close
4520bfdf AL	75	Done(true);
b2e465d6 AL	76	}
	77	/}}}/
	78	// ExtractTar::Done - Reap the gzip sub process /{{{/
	79	// ---------------------------------------------------------------------
	80	/* If the force flag is given then error messages are suppressed - this
	81	means we hit the end of the tar file but there was still gzip data. */
	82	bool ExtractTar::Done(bool Force)
	83	{
	84	InFd.Close();
	85	if (GZPid <= 0)
	86	return true;
	87
	88	/* If there is a pending error then we are cleaning up gzip and are
	89	not interested in it's failures */
	90	if (_error->PendingError() == true)
	91	Force = true;
	92
	93	// Make sure we clean it up!
	94	kill(GZPid,SIGINT);
b21c0438 MZ	95	string confvar = string("dir::bin::") + DecompressProg;
b21c0438 MZ	96	if (ExecWait(GZPid,_config->Find(confvar.c_str(),DecompressProg.c_str()).c_str(),
b2e465d6 AL	97	Force) == false)
	98	{
	99	GZPid = -1;
	100	return Force;
	101	}
	102
	103	GZPid = -1;
	104	return true;
	105	}
	106	/}}}/
	107	// ExtractTar::StartGzip - Startup gzip /{{{/
	108	// ---------------------------------------------------------------------
	109	/* This creates a gzip sub process that has its input as the file itself.
	110	If this tar file is embedded into something like an ar file then
	111	gzip will efficiently ignore the extra bits. */
	112	bool ExtractTar::StartGzip()
	113	{
	114	int Pipes[2];
	115	if (pipe(Pipes) != 0)
05eb7df0	116	return _error->Errno("pipe",_("Failed to create pipes"));
b2e465d6 AL	117
	118	// Fork off the process
	119	GZPid = ExecFork();
	120
	121	// Spawn the subprocess
	122	if (GZPid == 0)
	123	{
	124	// Setup the FDs
	125	dup2(Pipes[1],STDOUT_FILENO);
	126	dup2(File.Fd(),STDIN_FILENO);
	127	int Fd = open("/dev/null",O_RDWR);
	128	if (Fd == -1)
	129	_exit(101);
	130	dup2(Fd,STDERR_FILENO);
	131	close(Fd);
	132	SetCloseExec(STDOUT_FILENO,false);
	133	SetCloseExec(STDIN_FILENO,false);
	134	SetCloseExec(STDERR_FILENO,false);
	135
	136	const char *Args[3];
b21c0438	137	string confvar = string("dir::bin::") + DecompressProg;
a9be43ff MV	138	string argv0 = _config->Find(confvar.c_str(),DecompressProg.c_str());
a9be43ff MV	139	Args[0] = argv0.c_str();
b2e465d6 AL	140	Args[1] = "-d";
b2e465d6 AL	141	Args[2] = 0;
b21c0438	142	execvp(Args[0],(char **)Args);
05eb7df0	143	cerr << _("Failed to exec gzip ") << Args[0] << endl;
b2e465d6 AL	144	_exit(100);
	145	}
	146
	147	// Fix up our FDs
	148	InFd.Fd(Pipes[0]);
	149	close(Pipes[1]);
	150	return true;
	151	}
	152	/}}}/
	153	// ExtractTar::Go - Perform extraction /{{{/
	154	// ---------------------------------------------------------------------
	155	/* This reads each 512 byte block from the archive and extracts the header
	156	information into the Item structure. Then it resolves the UID/GID and
	157	invokes the correct processing function. */
	158	bool ExtractTar::Go(pkgDirStream &Stream)
	159	{
	160	if (StartGzip() == false)
	161	return false;
	162
	163	// Loop over all blocks
	164	string LastLongLink;
	165	string LastLongName;
	166	while (1)
	167	{
	168	bool BadRecord = false;
	169	unsigned char Block[512];
	170	if (InFd.Read(Block,sizeof(Block),true) == false)
	171	return false;
	172
	173	if (InFd.Eof() == true)
	174	break;
	175
	176	// Get the checksum
	177	TarHeader Tar = (TarHeader )Block;
	178	unsigned long CheckSum;
	179	if (StrToNum(Tar->Checksum,CheckSum,sizeof(Tar->Checksum),8) == false)
05eb7df0	180	return _error->Error(_("Corrupted archive"));
b2e465d6 AL	181
	182	/* Compute the checksum field. The actual checksum is blanked out
	183	with spaces so it is not included in the computation */
	184	unsigned long NewSum = 0;
	185	memset(Tar->Checksum,' ',sizeof(Tar->Checksum));
	186	for (int I = 0; I != sizeof(Block); I++)
	187	NewSum += Block[I];
	188
	189	/* Check for a block of nulls - in this case we kill gzip, GNU tar
	190	does this.. */
	191	if (NewSum == ' '*sizeof(Tar->Checksum))
	192	return Done(true);
	193
	194	if (NewSum != CheckSum)
db0db9fe	195	return _error->Error(_("Tar checksum failed, archive corrupted"));
b2e465d6 AL	196
	197	// Decode all of the fields
	198	pkgDirStream::Item Itm;
b2e465d6	199	if (StrToNum(Tar->Mode,Itm.Mode,sizeof(Tar->Mode),8) == false \|\|
54f2f0a3 NH	200	(Base256ToNum(Tar->UserID,Itm.UID,8) == false &&
	201	StrToNum(Tar->UserID,Itm.UID,sizeof(Tar->UserID),8) == false) \|\|
	202	(Base256ToNum(Tar->GroupID,Itm.GID,8) == false &&
	203	StrToNum(Tar->GroupID,Itm.GID,sizeof(Tar->GroupID),8) == false) \|\|
	204	(Base256ToNum(Tar->Size,Itm.Size,12) == false &&
	205	StrToNum(Tar->Size,Itm.Size,sizeof(Tar->Size),8) == false) \|\|
	206	(Base256ToNum(Tar->MTime,Itm.MTime,12) == false &&
	207	StrToNum(Tar->MTime,Itm.MTime,sizeof(Tar->MTime),8) == false) \|\|
b2e465d6 AL	208	StrToNum(Tar->Major,Itm.Major,sizeof(Tar->Major),8) == false \|\|
b2e465d6 AL	209	StrToNum(Tar->Minor,Itm.Minor,sizeof(Tar->Minor),8) == false)
05eb7df0	210	return _error->Error(_("Corrupted archive"));
b2e465d6 AL	211
	212	// Grab the filename
	213	if (LastLongName.empty() == false)
	214	Itm.Name = (char *)LastLongName.c_str();
	215	else
	216	{
3c8cda8b	217	Tar->Name[sizeof(Tar->Name)-1] = 0;
b2e465d6 AL	218	Itm.Name = Tar->Name;
	219	}
	220	if (Itm.Name[0] == '.' && Itm.Name[1] == '/' && Itm.Name[2] != 0)
	221	Itm.Name += 2;
	222
	223	// Grab the link target
3c8cda8b	224	Tar->Name[sizeof(Tar->LinkName)-1] = 0;
b2e465d6 AL	225	Itm.LinkTarget = Tar->LinkName;
	226
	227	if (LastLongLink.empty() == false)
	228	Itm.LinkTarget = (char *)LastLongLink.c_str();
	229
	230	// Convert the type over
	231	switch (Tar->LinkFlag)
	232	{
	233	case NormalFile0:
	234	case NormalFile:
	235	Itm.Type = pkgDirStream::Item::File;
	236	break;
	237
	238	case HardLink:
	239	Itm.Type = pkgDirStream::Item::HardLink;
	240	break;
	241
	242	case SymbolicLink:
	243	Itm.Type = pkgDirStream::Item::SymbolicLink;
	244	break;
	245
	246	case CharacterDevice:
	247	Itm.Type = pkgDirStream::Item::CharDevice;
	248	break;
	249
	250	case BlockDevice:
	251	Itm.Type = pkgDirStream::Item::BlockDevice;
	252	break;
	253
	254	case Directory:
	255	Itm.Type = pkgDirStream::Item::Directory;
	256	break;
	257
	258	case FIFO:
	259	Itm.Type = pkgDirStream::Item::FIFO;
	260	break;
	261
	262	case GNU_LongLink:
	263	{
	264	unsigned long Length = Itm.Size;
	265	unsigned char Block[512];
	266	while (Length > 0)
	267	{
	268	if (InFd.Read(Block,sizeof(Block),true) == false)
	269	return false;
	270	if (Length <= sizeof(Block))
	271	{
	272	LastLongLink.append(Block,Block+sizeof(Block));
	273	break;
	274	}
	275	LastLongLink.append(Block,Block+sizeof(Block));
	276	Length -= sizeof(Block);
	277	}
	278	continue;
	279	}
	280
	281	case GNU_LongName:
	282	{
	283	unsigned long Length = Itm.Size;
	284	unsigned char Block[512];
	285	while (Length > 0)
	286	{
	287	if (InFd.Read(Block,sizeof(Block),true) == false)
	288	return false;
289	if (Length < sizeof(Block))
290	{
291	LastLongName.append(Block,Block+sizeof(Block));
292	break;
293	}
294	LastLongName.append(Block,Block+sizeof(Block));
295	Length -= sizeof(Block);
296	}
297	continue;
298	}
299
300	default:
301	BadRecord = true;
4c6a9fad	302	_error->Warning(_("Unknown TAR header type %u, member %s"),(unsigned)Tar->LinkFlag,Tar->Name);
b2e465d6 AL	303	break;
	304	}
	305
	306	int Fd = -1;
	307	if (BadRecord == false)
	308	if (Stream.DoItem(Itm,Fd) == false)
	309	return false;
	310
	311	// Copy the file over the FD
	312	unsigned long Size = Itm.Size;
	313	while (Size != 0)
	314	{
	315	unsigned char Junk[32*1024];
42ab8223	316	unsigned long Read = min(Size,(unsigned long)sizeof(Junk));
b2e465d6 AL	317	if (InFd.Read(Junk,((Read+511)/512)*512) == false)
	318	return false;
	319
	320	if (BadRecord == false)
	321	{
	322	if (Fd > 0)
	323	{
	324	if (write(Fd,Junk,Read) != (signed)Read)
	325	return Stream.Fail(Itm,Fd);
	326	}
	327	else
	328	{
	329	/* An Fd of -2 means to send to a special processing
	330	function */
	331	if (Fd == -2)
	332	if (Stream.Process(Itm,Junk,Read,Itm.Size - Size) == false)
	333	return Stream.Fail(Itm,Fd);
	334	}
	335	}
	336
	337	Size -= Read;
	338	}
	339
	340	// And finish up
59152cdb	341	if (BadRecord == false)
b2e465d6 AL	342	if (Stream.FinishedFile(Itm,Fd) == false)
	343	return false;
	344
	345	LastLongName.erase();
	346	LastLongLink.erase();
	347	}
	348
	349	return Done(false);
	350	}
	351	/}}}/