Support for tar -z option for uncompressing only
[oweals/busybox.git] / tar.c
1 /* vi: set sw=4 ts=4: */
2 /*
3  * Mini tar implementation for busybox 
4  *
5  * Note, that as of BusyBox-0.43, tar has been completely rewritten from the
6  * ground up.  It still has remnents of the old code lying about, but it is
7  * very different now (i.e. cleaner, less global variables, etc)
8  *
9  * Copyright (C) 2000 by Lineo, inc.
10  * Written by Erik Andersen <andersen@lineo.com>, <andersee@debian.org>
11  *
12  * Based in part in the tar implementation in sash
13  *  Copyright (c) 1999 by David I. Bell
14  *  Permission is granted to use, distribute, or modify this source,
15  *  provided that this copyright notice remains intact.
16  *  Permission to distribute sash derived code under the GPL has been granted.
17  *
18  * Based in part on the tar implementation from busybox-0.28
19  *  Copyright (C) 1995 Bruce Perens
20  *  This is free software under the GNU General Public License.
21  *
22  * This program is free software; you can redistribute it and/or modify
23  * it under the terms of the GNU General Public License as published by
24  * the Free Software Foundation; either version 2 of the License, or
25  * (at your option) any later version.
26  *
27  * This program is distributed in the hope that it will be useful,
28  * but WITHOUT ANY WARRANTY; without even the implied warranty of
29  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
30  * General Public License for more details.
31  *
32  * You should have received a copy of the GNU General Public License
33  * along with this program; if not, write to the Free Software
34  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
35  *
36  */
37
38
39 #include "busybox.h"
40 #define BB_DECLARE_EXTERN
41 #define bb_need_io_error
42 #define bb_need_name_longer_than_foo
43 #include "messages.c"
44 #include <stdio.h>
45 #include <dirent.h>
46 #include <errno.h>
47 #include <fcntl.h>
48 #include <signal.h>
49 #include <time.h>
50 #include <utime.h>
51 #include <sys/types.h>
52 #include <sys/sysmacros.h>
53 #include <getopt.h>
54
55 #ifdef BB_FEATURE_TAR_GZIP
56 extern int unzip(int in, int out);
57 extern int gunzip_init();
58 #endif
59
60 /* Tar file constants  */
61 #ifndef MAJOR
62 #define MAJOR(dev) (((dev)>>8)&0xff)
63 #define MINOR(dev) ((dev)&0xff)
64 #endif
65
66 #define NAME_SIZE       100
67
68 /* POSIX tar Header Block, from POSIX 1003.1-1990  */
69 struct TarHeader
70 {
71                                 /* byte offset */
72         char name[NAME_SIZE];         /*   0-99 */
73         char mode[8];                 /* 100-107 */
74         char uid[8];                  /* 108-115 */
75         char gid[8];                  /* 116-123 */
76         char size[12];                /* 124-135 */
77         char mtime[12];               /* 136-147 */
78         char chksum[8];               /* 148-155 */
79         char typeflag;                /* 156-156 */
80         char linkname[NAME_SIZE];     /* 157-256 */
81         char magic[6];                /* 257-262 */
82         char version[2];              /* 263-264 */
83         char uname[32];               /* 265-296 */
84         char gname[32];               /* 297-328 */
85         char devmajor[8];             /* 329-336 */
86         char devminor[8];             /* 337-344 */
87         char prefix[155];             /* 345-499 */
88         char padding[12];             /* 500-512 (pad to exactly the TAR_BLOCK_SIZE) */
89 };
90 typedef struct TarHeader TarHeader;
91
92
93 /* A few useful constants */
94 #define TAR_MAGIC          "ustar"        /* ustar and a null */
95 #define TAR_VERSION        "  "           /* Be compatable with GNU tar format */
96 #define TAR_MAGIC_LEN       6
97 #define TAR_VERSION_LEN     2
98 #define TAR_BLOCK_SIZE      512
99
100 /* A nice enum with all the possible tar file content types */
101 enum TarFileType 
102 {
103         REGTYPE  = '0',            /* regular file */
104         REGTYPE0 = '\0',           /* regular file (ancient bug compat)*/
105         LNKTYPE  = '1',            /* hard link */
106         SYMTYPE  = '2',            /* symbolic link */
107         CHRTYPE  = '3',            /* character special */
108         BLKTYPE  = '4',            /* block special */
109         DIRTYPE  = '5',            /* directory */
110         FIFOTYPE = '6',            /* FIFO special */
111         CONTTYPE = '7',            /* reserved */
112         GNULONGLINK = 'K',         /* GNU long (>100 chars) link name */
113         GNULONGNAME = 'L',         /* GNU long (>100 chars) file name */
114 };
115 typedef enum TarFileType TarFileType;
116
117 /* This struct ignores magic, non-numeric user name, 
118  * non-numeric group name, and the checksum, since
119  * these are all ignored by BusyBox tar. */ 
120 struct TarInfo
121 {
122         int              tarFd;          /* An open file descriptor for reading from the tarball */
123         char *           name;           /* File name */
124         mode_t           mode;           /* Unix mode, including device bits. */
125         uid_t            uid;            /* Numeric UID */
126         gid_t            gid;            /* Numeric GID */
127         size_t           size;           /* Size of file */
128         time_t           mtime;          /* Last-modified time */
129         enum TarFileType type;           /* Regular, directory, link, etc */
130         char *           linkname;       /* Name for symbolic and hard links */
131         long             devmajor;       /* Major number for special device */
132         long             devminor;       /* Minor number for special device */
133 };
134 typedef struct TarInfo TarInfo;
135
136 /* Local procedures to restore files from a tar file.  */
137 static int readTarFile(int tarFd, int extractFlag, int listFlag, 
138                 int tostdoutFlag, int verboseFlag, char** extractList,
139                 char** excludeList);
140
141 #ifdef BB_FEATURE_TAR_CREATE
142 /* Local procedures to save files into a tar file.  */
143 static int writeTarFile(const char* tarName, int verboseFlag, char **argv,
144                 char** excludeList);
145 #endif
146
147 #ifdef BB_FEATURE_TAR_GZIP
148 /* Signal handler for when child gzip process dies...  */
149 void child_died()
150 {
151         fflush(stdout);
152         fflush(stderr);
153         exit(EXIT_FAILURE);
154 }
155
156 static int tar_unzip_init(int tarFd)
157 {
158         int child_pid;
159         static int unzip_pipe[2];
160         /* Cope if child dies... Otherwise we block forever in read()... */
161         signal(SIGCHLD, child_died);
162
163         if (pipe(unzip_pipe)!=0)
164                 error_msg_and_die("pipe error\n");
165                         
166         if ( (child_pid = fork()) == -1)
167                 error_msg_and_die("fork failure\n");
168
169         if (child_pid==0) {
170                 /* child process */
171                 gunzip_init();
172                 unzip(tarFd, unzip_pipe[1]);
173                 exit(EXIT_SUCCESS);
174         }
175         else
176                 /* return fd of uncompressed data to parent process */
177                 return(unzip_pipe[0]);
178 }
179 #endif
180
181 extern int tar_main(int argc, char **argv)
182 {
183         char** excludeList=NULL;
184         char** extractList=NULL;
185         const char *tarName="-";
186 #if defined BB_FEATURE_TAR_EXCLUDE
187         int excludeListSize=0;
188         char *excludeFileName ="-";
189         FILE *fileList;
190         char file[256];
191 #endif
192 #if defined BB_FEATURE_TAR_GZIP
193         int unzipFlag    = FALSE;
194 #endif
195         int listFlag     = FALSE;
196         int extractFlag  = FALSE;
197         int createFlag   = FALSE;
198         int verboseFlag  = FALSE;
199         int tostdoutFlag = FALSE;
200         int status       = FALSE;
201         int firstOpt     = TRUE;
202         int stopIt;
203
204         if (argc <= 1)
205                 usage(tar_usage);
206
207         while (*(++argv) && (**argv == '-' || firstOpt == TRUE)) {
208                 firstOpt=FALSE;
209                 stopIt=FALSE;
210                 while (stopIt==FALSE && **argv) {
211                         switch (*((*argv)++)) {
212                                 case 'c':
213                                         if (extractFlag == TRUE || listFlag == TRUE)
214                                                 goto flagError;
215                                         createFlag = TRUE;
216                                         break;
217                                 case 'x':
218                                         if (listFlag == TRUE || createFlag == TRUE)
219                                                 goto flagError;
220                                         extractFlag = TRUE;
221                                         break;
222                                 case 't':
223                                         if (extractFlag == TRUE || createFlag == TRUE)
224                                                 goto flagError;
225                                         listFlag = TRUE;
226                                         break;
227 #ifdef BB_FEATURE_TAR_GZIP
228                                 case 'z':
229                                         unzipFlag = TRUE;
230                                         break;
231 #endif
232                                 case 'v':
233                                         verboseFlag = TRUE;
234                                         break;
235                                 case 'O':
236                                         tostdoutFlag = TRUE;
237                                         break;                                  
238                                 case 'f':
239                                         if (*tarName != '-')
240                                                 error_msg_and_die( "Only one 'f' option allowed\n");
241                                         tarName = *(++argv);
242                                         if (tarName == NULL)
243                                                 error_msg_and_die( "Option requires an argument: No file specified\n");
244                                         stopIt=TRUE;
245                                         break;
246 #if defined BB_FEATURE_TAR_EXCLUDE
247                                 case 'e':
248                                         if (strcmp(*argv, "xclude")==0) {
249                                                 excludeList=xrealloc( excludeList, sizeof(char**) * (excludeListSize+2));
250                                                 excludeList[excludeListSize] = *(++argv);
251                                                 if (excludeList[excludeListSize] == NULL)
252                                                         error_msg_and_die( "Option requires an argument: No file specified\n");
253                                                 /* Remove leading "/"s */
254                                                 if (*excludeList[excludeListSize] =='/')
255                                                         excludeList[excludeListSize] = (excludeList[excludeListSize])+1;
256                                                 /* Tack a NULL onto the end of the list */
257                                                 excludeList[++excludeListSize] = NULL;
258                                                 stopIt=TRUE;
259                                                 break;
260                                         }
261                                 case 'X':
262                                        if (*excludeFileName != '-')
263                                                error_msg_and_die("Only one 'X' option allowed\n");
264                                        excludeFileName = *(++argv);
265                                        if  (excludeFileName == NULL)
266                                                error_msg_and_die("Option requires an argument: No file specified\n");
267                                        fileList = fopen (excludeFileName, "rt");
268                                        if (! fileList)
269                                                error_msg_and_die("Exclude file: file not found\n");
270                                        while (!feof(fileList)) {
271                                                fscanf(fileList, "%s", file);
272                                                excludeList=xrealloc( excludeList, sizeof(char**) * (excludeListSize+2));
273                                                excludeList[excludeListSize] = malloc(sizeof(char) * (strlen(file)+1));
274                                                strcpy(excludeList[excludeListSize],file);
275                                                /* Remove leading "/"s */
276                                                if (*excludeList[excludeListSize] == '/')
277                                                        excludeList[excludeListSize] = (excludeList[excludeListSize])+1;
278                                                /* Tack a NULL onto the end of the list */
279                                                        excludeList[++excludeListSize] = NULL;
280                                        }
281  
282                                        fclose(fileList);
283                                        stopIt=TRUE;
284                                        break;
285 #endif
286                                 case '-':
287                                                 break;
288                                 default:
289                                         usage(tar_usage);
290                         }
291                 }
292         }
293
294         /* 
295          * Do the correct type of action supplying the rest of the
296          * command line arguments as the list of files to process.
297          */
298         if (createFlag == TRUE) {
299 #ifndef BB_FEATURE_TAR_CREATE
300                 error_msg_and_die( "This version of tar was not compiled with tar creation support.\n");
301 #else
302 #ifdef BB_FEATURE_TAR_GZIP
303                 if (unzipFlag==TRUE)
304                         error_msg_and_die("Creation of compressed not internally support by tar, pipe to busybox gunzip\n");
305 #endif
306                 status = writeTarFile(tarName, verboseFlag, argv, excludeList);
307 #endif
308         }
309         if (listFlag == TRUE || extractFlag == TRUE) {
310                 int tarFd;
311                 if (*argv)
312                         extractList = argv;
313                 /* Open the tar file for reading.  */
314                 if (!strcmp(tarName, "-"))
315                         tarFd = fileno(stdin);
316                 else
317                         tarFd = open(tarName, O_RDONLY);
318                 if (tarFd < 0)
319                         error_msg_and_die( "Error opening '%s': %s\n", tarName, strerror(errno));
320
321 #ifdef BB_FEATURE_TAR_GZIP      
322                 /* unzip tarFd in a seperate process */
323                 if (unzipFlag == TRUE)
324                         tarFd = tar_unzip_init(tarFd);
325 #endif                  
326                 status = readTarFile(tarFd, extractFlag, listFlag, tostdoutFlag,
327                                         verboseFlag, extractList, excludeList);
328         }
329
330         if (status == TRUE)
331                 return EXIT_SUCCESS;
332         else
333                 return EXIT_FAILURE;
334
335   flagError:
336         error_msg_and_die( "Exactly one of 'c', 'x' or 't' must be specified\n");
337 }
338                                         
339 static void
340 fixUpPermissions(TarInfo *header)
341 {
342         struct utimbuf t;
343         /* Now set permissions etc for the new file */
344         chown(header->name, header->uid, header->gid);
345         chmod(header->name, header->mode);
346         /* Reset the time */
347         t.actime = time(0);
348         t.modtime = header->mtime;
349         utime(header->name, &t);
350 }
351                                 
352 static int
353 tarExtractRegularFile(TarInfo *header, int extractFlag, int tostdoutFlag)
354 {
355         size_t  writeSize;
356         size_t  readSize;
357         size_t  actualWriteSz;
358         char    buffer[BUFSIZ];
359         size_t  size = header->size;
360         int outFd=fileno(stdout);
361
362         /* Open the file to be written, if a file is supposed to be written */
363         if (extractFlag==TRUE && tostdoutFlag==FALSE) {
364                 /* Create the path to the file, just in case it isn't there...
365                  * This should not screw up path permissions or anything. */
366                 create_path(header->name, 0777);
367                 if ((outFd=open(header->name, O_CREAT|O_TRUNC|O_WRONLY, 
368                                                 header->mode & ~S_IFMT)) < 0) {
369                         error_msg(io_error, header->name, strerror(errno)); 
370                         return( FALSE);
371                 }
372         }
373
374         /* Write out the file, if we are supposed to be doing that */
375         while ( size > 0 ) {
376                 actualWriteSz=0;
377                 if ( size > sizeof(buffer) )
378                         writeSize = readSize = sizeof(buffer);
379                 else {
380                         int mod = size % 512;
381                         if ( mod != 0 )
382                                 readSize = size + (512 - mod);
383                         else
384                                 readSize = size;
385                         writeSize = size;
386                 }
387                 if ( (readSize = full_read(header->tarFd, buffer, readSize)) <= 0 ) {
388                         /* Tarball seems to have a problem */
389                         error_msg("Unexpected EOF in archive\n"); 
390                         return( FALSE);
391                 }
392                 if ( readSize < writeSize )
393                         writeSize = readSize;
394
395                 /* Write out the file, if we are supposed to be doing that */
396                 if (extractFlag==TRUE) {
397
398                         if ((actualWriteSz=full_write(outFd, buffer, writeSize)) != writeSize ) {
399                                 /* Output file seems to have a problem */
400                                 error_msg(io_error, header->name, strerror(errno)); 
401                                 return( FALSE);
402                         }
403                 } else {
404                         actualWriteSz=writeSize;
405                 }
406
407                 size -= actualWriteSz;
408         }
409
410         /* Now we are done writing the file out, so try 
411          * and fix up the permissions and whatnot */
412         if (extractFlag==TRUE && tostdoutFlag==FALSE) {
413                 close(outFd);
414                 fixUpPermissions(header);
415         }
416         return( TRUE);
417 }
418
419 static int
420 tarExtractDirectory(TarInfo *header, int extractFlag, int tostdoutFlag)
421 {
422
423         if (extractFlag==FALSE || tostdoutFlag==TRUE)
424                 return( TRUE);
425
426         if (create_path(header->name, header->mode) != TRUE) {
427                 error_msg("%s: Cannot mkdir: %s\n", 
428                                 header->name, strerror(errno)); 
429                 return( FALSE);
430         }
431         /* make the final component, just in case it was
432          * omitted by create_path() (which will skip the
433          * directory if it doesn't have a terminating '/') */
434         if (mkdir(header->name, header->mode) == 0) {
435                 fixUpPermissions(header);
436         }
437         return( TRUE);
438 }
439
440 static int
441 tarExtractHardLink(TarInfo *header, int extractFlag, int tostdoutFlag)
442 {
443         if (extractFlag==FALSE || tostdoutFlag==TRUE)
444                 return( TRUE);
445
446         if (link(header->linkname, header->name) < 0) {
447                 error_msg("%s: Cannot create hard link to '%s': %s\n", 
448                                 header->name, header->linkname, strerror(errno)); 
449                 return( FALSE);
450         }
451
452         /* Now set permissions etc for the new directory */
453         fixUpPermissions(header);
454         return( TRUE);
455 }
456
457 static int
458 tarExtractSymLink(TarInfo *header, int extractFlag, int tostdoutFlag)
459 {
460         if (extractFlag==FALSE || tostdoutFlag==TRUE)
461                 return( TRUE);
462
463 #ifdef  S_ISLNK
464         if (symlink(header->linkname, header->name) < 0) {
465                 error_msg("%s: Cannot create symlink to '%s': %s\n", 
466                                 header->name, header->linkname, strerror(errno)); 
467                 return( FALSE);
468         }
469         /* Try to change ownership of the symlink.
470          * If libs doesn't support that, don't bother.
471          * Changing the pointed-to-file is the Wrong Thing(tm).
472          */
473 #if (__GLIBC__ >= 2) && (__GLIBC_MINOR__ >= 1)
474         lchown(header->name, header->uid, header->gid);
475 #endif
476
477         /* Do not change permissions or date on symlink,
478          * since it changes the pointed to file instead.  duh. */
479 #else
480         error_msg("%s: Cannot create symlink to '%s': %s\n", 
481                         header->name, header->linkname, 
482                         "symlinks not supported"); 
483 #endif
484         return( TRUE);
485 }
486
487 static int
488 tarExtractSpecial(TarInfo *header, int extractFlag, int tostdoutFlag)
489 {
490         if (extractFlag==FALSE || tostdoutFlag==TRUE)
491                 return( TRUE);
492
493         if (S_ISCHR(header->mode) || S_ISBLK(header->mode) || S_ISSOCK(header->mode)) {
494                 if (mknod(header->name, header->mode, makedev(header->devmajor, header->devminor)) < 0) {
495                         error_msg("%s: Cannot mknod: %s\n",
496                                 header->name, strerror(errno)); 
497                         return( FALSE);
498                 }
499         } else if (S_ISFIFO(header->mode)) {
500                 if (mkfifo(header->name, header->mode) < 0) {
501                         error_msg("%s: Cannot mkfifo: %s\n",
502                                 header->name, strerror(errno)); 
503                         return( FALSE);
504                 }
505         }
506
507         /* Now set permissions etc for the new directory */
508         fixUpPermissions(header);
509         return( TRUE);
510 }
511
512 /* Read an octal value in a field of the specified width, with optional
513  * spaces on both sides of the number and with an optional null character
514  * at the end.  Returns -1 on an illegal format.  */
515 static long getOctal(const char *cp, int size)
516 {
517         long val = 0;
518
519         for(;(size > 0) && (*cp == ' '); cp++, size--);
520         if ((size == 0) || !is_octal(*cp))
521                 return -1;
522         for(; (size > 0) && is_octal(*cp); size--) {
523                 val = val * 8 + *cp++ - '0';
524         }
525         for (;(size > 0) && (*cp == ' '); cp++, size--);
526         if ((size > 0) && *cp)
527                 return -1;
528         return val;
529 }
530
531
532 /* Parse the tar header and fill in the nice struct with the details */
533 static int
534 readTarHeader(struct TarHeader *rawHeader, struct TarInfo *header)
535 {
536         int i;
537         long chksum, sum=0;
538         unsigned char *s = (unsigned char *)rawHeader;
539
540         header->name  = rawHeader->name;
541         /* Check for and relativify any absolute paths */
542         if ( *(header->name) == '/' ) {
543                 static int alreadyWarned=FALSE;
544
545                 while (*(header->name) == '/')
546                         ++*(header->name);
547
548                 if (alreadyWarned == FALSE) {
549                         error_msg("Removing leading '/' from member names\n");
550                         alreadyWarned = TRUE;
551                 }
552         }
553
554         header->mode  = getOctal(rawHeader->mode, sizeof(rawHeader->mode));
555         header->uid   =  getOctal(rawHeader->uid, sizeof(rawHeader->uid));
556         header->gid   =  getOctal(rawHeader->gid, sizeof(rawHeader->gid));
557         header->size  = getOctal(rawHeader->size, sizeof(rawHeader->size));
558         header->mtime = getOctal(rawHeader->mtime, sizeof(rawHeader->mtime));
559         chksum = getOctal(rawHeader->chksum, sizeof(rawHeader->chksum));
560         header->type  = rawHeader->typeflag;
561         header->linkname  = rawHeader->linkname;
562         header->devmajor  = getOctal(rawHeader->devmajor, sizeof(rawHeader->devmajor));
563         header->devminor  = getOctal(rawHeader->devminor, sizeof(rawHeader->devminor));
564
565         /* Check the checksum */
566         for (i = sizeof(*rawHeader); i-- != 0;) {
567                 sum += *s++;
568         }
569         /* Remove the effects of the checksum field (replace 
570          * with blanks for the purposes of the checksum) */
571         s = rawHeader->chksum;
572         for (i = sizeof(rawHeader->chksum) ; i-- != 0;) {
573                 sum -= *s++;
574         }
575         sum += ' ' * sizeof(rawHeader->chksum);
576         if (sum == chksum )
577                 return ( TRUE);
578         return( FALSE);
579 }
580
581
582 /*
583  * Read a tar file and extract or list the specified files within it.
584  * If the list is empty than all files are extracted or listed.
585  */
586 extern int readTarFile(int tarFd, int extractFlag, int listFlag, 
587                 int tostdoutFlag, int verboseFlag, char** extractList,
588                 char** excludeList)
589 {
590         int status;
591         int errorFlag=FALSE;
592         int skipNextHeaderFlag=FALSE;
593         TarHeader rawHeader;
594         TarInfo header;
595         char** tmpList;
596
597         /* Set the umask for this process so it doesn't 
598          * screw up permission setting for us later. */
599         umask(0);
600
601         /* Read the tar file, and iterate over it one file at a time */
602         while ( (status = full_read(tarFd, (char*)&rawHeader, TAR_BLOCK_SIZE)) == TAR_BLOCK_SIZE ) {
603
604                 /* Try to read the header */
605                 if ( readTarHeader(&rawHeader, &header) == FALSE ) {
606                         if ( *(header.name) == '\0' ) {
607                                 goto endgame;
608                         } else {
609                                 errorFlag=TRUE;
610                                 error_msg("Bad tar header, skipping\n");
611                                 continue;
612                         }
613                 }
614                 if ( *(header.name) == '\0' )
615                                 goto endgame;
616                 header.tarFd = tarFd;
617
618                 /* Skip funky extra GNU headers that precede long files */
619                 if ( (header.type == GNULONGNAME) || (header.type == GNULONGLINK) ) {
620                         skipNextHeaderFlag=TRUE;
621                         if (tarExtractRegularFile(&header, FALSE, FALSE) == FALSE)
622                                 errorFlag = TRUE;
623                         continue;
624                 }
625                 if ( skipNextHeaderFlag == TRUE ) { 
626                         skipNextHeaderFlag=FALSE;
627                         error_msg(name_longer_than_foo, NAME_SIZE); 
628                         if (tarExtractRegularFile(&header, FALSE, FALSE) == FALSE)
629                                 errorFlag = TRUE;
630                         continue;
631                 }
632
633 #if defined BB_FEATURE_TAR_EXCLUDE
634                 {
635                         int skipFlag=FALSE;
636                         /* Check for excluded files....  */
637                         for (tmpList=excludeList; tmpList && *tmpList; tmpList++) {
638                                 /* Do some extra hoop jumping for when directory names
639                                  * end in '/' but the entry in tmpList doesn't */
640                                 if (strncmp( *tmpList, header.name, strlen(*tmpList))==0 || (
641                                                         header.name[strlen(header.name)-1]=='/'
642                                                         && strncmp( *tmpList, header.name, 
643                                                                 MIN(strlen(header.name)-1, strlen(*tmpList)))==0)) {
644                                         /* If it is a regular file, pretend to extract it with
645                                          * the extractFlag set to FALSE, so the junk in the tarball
646                                          * is properly skipped over */
647                                         if ( header.type==REGTYPE || header.type==REGTYPE0 ) {
648                                                 if (tarExtractRegularFile(&header, FALSE, FALSE) == FALSE)
649                                                         errorFlag = TRUE;
650                                         }
651                                         skipFlag=TRUE;
652                                         break;
653                                 }
654                         }
655                         /* There are not the droids you're looking for, move along */
656                         if (skipFlag==TRUE)
657                                 continue;
658                 }
659 #endif
660                 if (extractList != NULL) {
661                         int skipFlag = TRUE;
662                         for (tmpList = extractList; *tmpList != NULL; tmpList++) {
663                                 if (strncmp( *tmpList, header.name, strlen(*tmpList))==0 || (
664                                                         header.name[strlen(header.name)-1]=='/'
665                                                         && strncmp( *tmpList, header.name, 
666                                                                 MIN(strlen(header.name)-1, strlen(*tmpList)))==0)) {
667                                         /* If it is a regular file, pretend to extract it with
668                                          * the extractFlag set to FALSE, so the junk in the tarball
669                                          * is properly skipped over */
670                                         skipFlag = FALSE;
671                                         memmove(extractList+1, extractList,
672                                                                 sizeof(*extractList)*(tmpList-extractList));
673                                         extractList++;
674                                         break;
675                                 }
676                         }
677                         /* There are not the droids you're looking for, move along */
678                         if (skipFlag == TRUE) {
679                                 if ( header.type==REGTYPE || header.type==REGTYPE0 )
680                                         if (tarExtractRegularFile(&header, FALSE, FALSE) == FALSE)
681                                                 errorFlag = TRUE;
682                                 continue;
683                         }
684                 }
685
686                 if (listFlag == TRUE) {
687                         /* Special treatment if the list (-t) flag is on */
688                         if (verboseFlag == TRUE) {
689                                 int len, len1;
690                                 char buf[35];
691                                 struct tm *tm = localtime (&(header.mtime));
692
693                                 len=printf("%s ", mode_string(header.mode));
694                                 memset(buf, 0, 8*sizeof(char));
695                                 my_getpwuid(buf, header.uid);
696                                 if (! *buf)
697                                         len+=printf("%d", header.uid);
698                                 else
699                                         len+=printf("%s", buf);
700                                 memset(buf, 0, 8*sizeof(char));
701                                 my_getgrgid(buf, header.gid);
702                                 if (! *buf)
703                                         len+=printf("/%-d ", header.gid);
704                                 else
705                                         len+=printf("/%-s ", buf);
706
707                                 if (header.type==CHRTYPE || header.type==BLKTYPE) {
708                                         len1=snprintf(buf, sizeof(buf), "%ld,%-ld ", 
709                                                         header.devmajor, header.devminor);
710                                 } else {
711                                         len1=snprintf(buf, sizeof(buf), "%lu ", (long)header.size);
712                                 }
713                                 /* Jump through some hoops to make the columns match up */
714                                 for(;(len+len1)<31;len++)
715                                         printf(" ");
716                                 printf(buf);
717
718                                 /* Use ISO 8610 time format */
719                                 if (tm) { 
720                                         printf ("%04d-%02d-%02d %02d:%02d:%02d ", 
721                                                         tm->tm_year + 1900, tm->tm_mon + 1, tm->tm_mday, 
722                                                         tm->tm_hour, tm->tm_min, tm->tm_sec);
723                                 }
724                         }
725                         printf("%s", header.name);
726                         if (verboseFlag == TRUE) {
727                                 if (header.type==LNKTYPE)       /* If this is a link, say so */
728                                         printf(" link to %s", header.linkname);
729                                 else if (header.type==SYMTYPE)
730                                         printf(" -> %s", header.linkname);
731                         }
732                         printf("\n");
733                 }
734
735                 /* List contents if we are supposed to do that */
736                 if (verboseFlag == TRUE && extractFlag == TRUE) {
737                         /* Now the normal listing */
738                         FILE *vbFd = stdout;
739                         if (tostdoutFlag == TRUE)       // If the archive goes to stdout, verbose to stderr
740                                 vbFd = stderr;
741                         fprintf(vbFd, "%s\n", header.name);
742                 }
743                         
744                 /* Remove files if we would overwrite them */
745                 if (extractFlag == TRUE && tostdoutFlag == FALSE)
746                         unlink(header.name);
747
748                 /* If we got here, we can be certain we have a legitimate 
749                  * header to work with.  So work with it.  */
750                 switch ( header.type ) {
751                         case REGTYPE:
752                         case REGTYPE0:
753                                 /* If the name ends in a '/' then assume it is
754                                  * supposed to be a directory, and fall through */
755                                 if (header.name[strlen(header.name)-1] != '/') {
756                                         if (tarExtractRegularFile(&header, extractFlag, tostdoutFlag)==FALSE)
757                                                 errorFlag=TRUE;
758                                         break;
759                                 }
760                         case DIRTYPE:
761                                 if (tarExtractDirectory( &header, extractFlag, tostdoutFlag)==FALSE)
762                                         errorFlag=TRUE;
763                                 break;
764                         case LNKTYPE:
765                                 if (tarExtractHardLink( &header, extractFlag, tostdoutFlag)==FALSE)
766                                         errorFlag=TRUE;
767                                 break;
768                         case SYMTYPE:
769                                 if (tarExtractSymLink( &header, extractFlag, tostdoutFlag)==FALSE)
770                                         errorFlag=TRUE;
771                                 break;
772                         case CHRTYPE:
773                         case BLKTYPE:
774                         case FIFOTYPE:
775                                 if (tarExtractSpecial( &header, extractFlag, tostdoutFlag)==FALSE)
776                                         errorFlag=TRUE;
777                                 break;
778 #if 0
779                         /* Handled earlier */
780                         case GNULONGNAME:
781                         case GNULONGLINK:
782                                 skipNextHeaderFlag=TRUE;
783                                 break;
784 #endif
785                         default:
786                                 error_msg("Unknown file type '%c' in tar file\n", header.type);
787                                 close( tarFd);
788                                 return( FALSE);
789                 }
790         }
791         close(tarFd);
792         if (status > 0) {
793                 /* Bummer - we read a partial header */
794                 error_msg( "Error reading tar file: %s\n", strerror(errno));
795                 return ( FALSE);
796         }
797         else if (errorFlag==TRUE) {
798                 error_msg( "Error exit delayed from previous errors\n");
799                 return( FALSE);
800         } else 
801                 return( status);
802
803         /* Stuff to do when we are done */
804 endgame:
805         close( tarFd);
806         if (extractList != NULL) {
807                 for (; *extractList != NULL; extractList++) {
808                         error_msg("%s: Not found in archive\n", *extractList);
809                         errorFlag = TRUE;
810                 }
811         }
812         if ( *(header.name) == '\0' ) {
813                 if (errorFlag==TRUE)
814                         error_msg( "Error exit delayed from previous errors\n");
815                 else
816                         return( TRUE);
817         } 
818         return( FALSE);
819 }
820
821
822 #ifdef BB_FEATURE_TAR_CREATE
823
824 /*
825 ** writeTarFile(),  writeFileToTarball(), and writeTarHeader() are
826 ** the only functions that deal with the HardLinkInfo structure.
827 ** Even these functions use the xxxHardLinkInfo() functions.
828 */
829 typedef struct HardLinkInfo HardLinkInfo;
830 struct HardLinkInfo
831 {
832         HardLinkInfo *next;           /* Next entry in list */
833         dev_t dev;                    /* Device number */
834         ino_t ino;                    /* Inode number */
835         short linkCount;              /* (Hard) Link Count */
836         char name[1];                 /* Start of filename (must be last) */
837 };
838
839 /* Some info to be carried along when creating a new tarball */
840 struct TarBallInfo
841 {
842         char* fileName;               /* File name of the tarball */
843         int tarFd;                    /* Open-for-write file descriptor
844                                                                          for the tarball */
845         struct stat statBuf;          /* Stat info for the tarball, letting
846                                                                          us know the inode and device that the
847                                                                          tarball lives, so we can avoid trying 
848                                                                          to include the tarball into itself */
849         int verboseFlag;              /* Whether to print extra stuff or not */
850         char** excludeList;           /* List of files to not include */
851         HardLinkInfo *hlInfoHead;     /* Hard Link Tracking Information */
852         HardLinkInfo *hlInfo;         /* Hard Link Info for the current file */
853 };
854 typedef struct TarBallInfo TarBallInfo;
855
856
857 /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */
858 static void
859 addHardLinkInfo (HardLinkInfo **hlInfoHeadPtr, dev_t dev, ino_t ino,
860                 short linkCount, const char *name)
861 {
862         /* Note: hlInfoHeadPtr can never be NULL! */
863         HardLinkInfo *hlInfo;
864
865         hlInfo = (HardLinkInfo *)xmalloc(sizeof(HardLinkInfo)+strlen(name)+1);
866         if (hlInfo) {
867                 hlInfo->next = *hlInfoHeadPtr;
868                 *hlInfoHeadPtr = hlInfo;
869                 hlInfo->dev = dev;
870                 hlInfo->ino = ino;
871                 hlInfo->linkCount = linkCount;
872                 strcpy(hlInfo->name, name);
873         }
874         return;
875 }
876
877 static void
878 freeHardLinkInfo (HardLinkInfo **hlInfoHeadPtr)
879 {
880         HardLinkInfo *hlInfo = NULL;
881         HardLinkInfo *hlInfoNext = NULL;
882
883         if (hlInfoHeadPtr) {
884                 hlInfo = *hlInfoHeadPtr;
885                 while (hlInfo) {
886                         hlInfoNext = hlInfo->next;
887                         free(hlInfo);
888                         hlInfo = hlInfoNext;
889                 }
890                 *hlInfoHeadPtr = NULL;
891         }
892         return;
893 }
894
895 /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */
896 static HardLinkInfo *
897 findHardLinkInfo (HardLinkInfo *hlInfo, dev_t dev, ino_t ino)
898 {
899         while(hlInfo) {
900                 if ((ino == hlInfo->ino) && (dev == hlInfo->dev))
901                         break;
902                 hlInfo = hlInfo->next;
903         }
904         return(hlInfo);
905 }
906
907 /* Put an octal string into the specified buffer.
908  * The number is zero and space padded and possibly null padded.
909  * Returns TRUE if successful.  */ 
910 static int putOctal (char *cp, int len, long value)
911 {
912         int tempLength;
913         char tempBuffer[32];
914         char *tempString = tempBuffer;
915
916         /* Create a string of the specified length with an initial space,
917          * leading zeroes and the octal number, and a trailing null.  */
918         sprintf (tempString, "%0*lo", len - 1, value);
919
920         /* If the string is too large, suppress the leading space.  */
921         tempLength = strlen (tempString) + 1;
922         if (tempLength > len) {
923                 tempLength--;
924                 tempString++;
925         }
926
927         /* If the string is still too large, suppress the trailing null.  */
928         if (tempLength > len)
929                 tempLength--;
930
931         /* If the string is still too large, fail.  */
932         if (tempLength > len)
933                 return FALSE;
934
935         /* Copy the string to the field.  */
936         memcpy (cp, tempString, len);
937
938         return TRUE;
939 }
940
941 /* Write out a tar header for the specified file/directory/whatever */
942 static int
943 writeTarHeader(struct TarBallInfo *tbInfo, const char *fileName, struct stat *statbuf)
944 {
945         long chksum=0;
946         struct TarHeader header;
947 #if defined BB_FEATURE_TAR_EXCLUDE
948         char** tmpList;
949 #endif
950         const unsigned char *cp = (const unsigned char *) &header;
951         ssize_t size = sizeof(struct TarHeader);
952                 
953         memset( &header, 0, size);
954
955         if (*fileName=='/') {
956                 static int alreadyWarned=FALSE;
957                 if (alreadyWarned==FALSE) {
958                         error_msg("Removing leading '/' from member names\n");
959                         alreadyWarned=TRUE;
960                 }
961                 strncpy(header.name, fileName+1, sizeof(header.name)); 
962         }
963         else {
964                 strncpy(header.name, fileName, sizeof(header.name)); 
965         }
966
967 #if defined BB_FEATURE_TAR_EXCLUDE
968         /* Check for excluded files....  */
969         for (tmpList=tbInfo->excludeList; tmpList && *tmpList; tmpList++) {
970                 /* Do some extra hoop jumping for when directory names
971                  * end in '/' but the entry in tmpList doesn't */
972                 if (strncmp( *tmpList, header.name, strlen(*tmpList))==0 || (
973                                         header.name[strlen(header.name)-1]=='/'
974                                         && strncmp( *tmpList, header.name, 
975                                                 MIN(strlen(header.name)-1, strlen(*tmpList)))==0)) {
976                         /* Set the mode to something that is not a regular file, thereby
977                          * faking out writeTarFile into thinking that nothing further need
978                          * be done for this file.  Yes, I know this is ugly, but it works. */
979                         statbuf->st_mode = 0;
980                         return( TRUE);
981                 }
982         }
983 #endif
984
985         putOctal(header.mode, sizeof(header.mode), statbuf->st_mode);
986         putOctal(header.uid, sizeof(header.uid), statbuf->st_uid);
987         putOctal(header.gid, sizeof(header.gid), statbuf->st_gid);
988         putOctal(header.size, sizeof(header.size), 0); /* Regular file size is handled later */
989         putOctal(header.mtime, sizeof(header.mtime), statbuf->st_mtime);
990         strncpy(header.magic, TAR_MAGIC TAR_VERSION, 
991                         TAR_MAGIC_LEN + TAR_VERSION_LEN );
992
993         /* Enter the user and group names (default to root if it fails) */
994         my_getpwuid(header.uname, statbuf->st_uid);
995         if (! *header.uname)
996                 strcpy(header.uname, "root");
997         my_getgrgid(header.gname, statbuf->st_gid);
998         if (! *header.uname)
999                 strcpy(header.uname, "root");
1000
1001         if (tbInfo->hlInfo) {
1002                 /* This is a hard link */
1003                 header.typeflag = LNKTYPE;
1004                 strncpy(header.linkname, tbInfo->hlInfo->name, sizeof(header.linkname));
1005         } else if (S_ISLNK(statbuf->st_mode)) {
1006                 int link_size=0;
1007                 char buffer[BUFSIZ];
1008                 header.typeflag  = SYMTYPE;
1009                 link_size = readlink(fileName, buffer, sizeof(buffer) - 1);
1010                 if ( link_size < 0) {
1011                         error_msg("Error reading symlink '%s': %s\n", header.name, strerror(errno));
1012                         return ( FALSE);
1013                 }
1014                 buffer[link_size] = '\0';
1015                 strncpy(header.linkname, buffer, sizeof(header.linkname)); 
1016         } else if (S_ISDIR(statbuf->st_mode)) {
1017                 header.typeflag  = DIRTYPE;
1018                 strncat(header.name, "/", sizeof(header.name)); 
1019         } else if (S_ISCHR(statbuf->st_mode)) {
1020                 header.typeflag  = CHRTYPE;
1021                 putOctal(header.devmajor, sizeof(header.devmajor), MAJOR(statbuf->st_rdev));
1022                 putOctal(header.devminor, sizeof(header.devminor), MINOR(statbuf->st_rdev));
1023         } else if (S_ISBLK(statbuf->st_mode)) {
1024                 header.typeflag  = BLKTYPE;
1025                 putOctal(header.devmajor, sizeof(header.devmajor), MAJOR(statbuf->st_rdev));
1026                 putOctal(header.devminor, sizeof(header.devminor), MINOR(statbuf->st_rdev));
1027         } else if (S_ISFIFO(statbuf->st_mode)) {
1028                 header.typeflag  = FIFOTYPE;
1029         } else if (S_ISREG(statbuf->st_mode)) {
1030                 header.typeflag  = REGTYPE;
1031                 putOctal(header.size, sizeof(header.size), statbuf->st_size);
1032         } else {
1033                 error_msg("%s: Unknown file type\n", fileName);
1034                 return ( FALSE);
1035         }
1036
1037         /* Calculate and store the checksum (i.e. the sum of all of the bytes of
1038          * the header).  The checksum field must be filled with blanks for the
1039          * calculation.  The checksum field is formatted differently from the
1040          * other fields: it has [6] digits, a null, then a space -- rather than
1041          * digits, followed by a null like the other fields... */
1042         memset(header.chksum, ' ', sizeof(header.chksum));
1043         cp = (const unsigned char *) &header;
1044         while (size-- > 0)
1045                 chksum += *cp++;
1046         putOctal(header.chksum, 7, chksum);
1047         
1048         /* Now write the header out to disk */
1049         if ((size=full_write(tbInfo->tarFd, (char*)&header, sizeof(struct TarHeader))) < 0) {
1050                 error_msg(io_error, fileName, strerror(errno)); 
1051                 return ( FALSE);
1052         }
1053         /* Pad the header up to the tar block size */
1054         for (; size<TAR_BLOCK_SIZE; size++) {
1055                 write(tbInfo->tarFd, "\0", 1);
1056         }
1057         /* Now do the verbose thing (or not) */
1058         if (tbInfo->verboseFlag==TRUE) {
1059                 FILE *vbFd = stdout;
1060                 if (tbInfo->tarFd == fileno(stdout))    // If the archive goes to stdout, verbose to stderr
1061                         vbFd = stderr;
1062                 fprintf(vbFd, "%s\n", header.name);
1063         }
1064
1065         return ( TRUE);
1066 }
1067
1068
1069 static int writeFileToTarball(const char *fileName, struct stat *statbuf, void* userData)
1070 {
1071         struct TarBallInfo *tbInfo = (struct TarBallInfo *)userData;
1072
1073         /*
1074         ** Check to see if we are dealing with a hard link.
1075         ** If so -
1076         ** Treat the first occurance of a given dev/inode as a file while
1077         ** treating any additional occurances as hard links.  This is done
1078         ** by adding the file information to the HardLinkInfo linked list.
1079         */
1080         tbInfo->hlInfo = NULL;
1081         if (statbuf->st_nlink > 1) {
1082                 tbInfo->hlInfo = findHardLinkInfo(tbInfo->hlInfoHead, statbuf->st_dev, 
1083                                 statbuf->st_ino);
1084                 if (tbInfo->hlInfo == NULL)
1085                         addHardLinkInfo (&tbInfo->hlInfoHead, statbuf->st_dev,
1086                                         statbuf->st_ino, statbuf->st_nlink, fileName);
1087         }
1088
1089         /* It is against the rules to archive a socket */
1090         if (S_ISSOCK(statbuf->st_mode)) {
1091                 error_msg("%s: socket ignored\n", fileName);
1092                 return( TRUE);
1093         }
1094
1095         /* It is a bad idea to store the archive we are in the process of creating,
1096          * so check the device and inode to be sure that this particular file isn't
1097          * the new tarball */
1098         if (tbInfo->statBuf.st_dev == statbuf->st_dev &&
1099                         tbInfo->statBuf.st_ino == statbuf->st_ino) {
1100                 error_msg("%s: file is the archive; skipping\n", fileName);
1101                 return( TRUE);
1102         }
1103
1104         if (strlen(fileName) >= NAME_SIZE) {
1105                 error_msg(name_longer_than_foo, NAME_SIZE);
1106                 return ( TRUE);
1107         }
1108
1109         if (writeTarHeader(tbInfo, fileName, statbuf)==FALSE) {
1110                 return( FALSE);
1111         } 
1112
1113         /* Now, if the file is a regular file, copy it out to the tarball */
1114         if ((tbInfo->hlInfo == NULL)
1115         &&  (S_ISREG(statbuf->st_mode))) {
1116                 int  inputFileFd;
1117                 char buffer[BUFSIZ];
1118                 ssize_t size=0, readSize=0;
1119
1120                 /* open the file we want to archive, and make sure all is well */
1121                 if ((inputFileFd = open(fileName, O_RDONLY)) < 0) {
1122                         error_msg("%s: Cannot open: %s\n", fileName, strerror(errno));
1123                         return( FALSE);
1124                 }
1125                 
1126                 /* write the file to the archive */
1127                 while ( (size = full_read(inputFileFd, buffer, sizeof(buffer))) > 0 ) {
1128                         if (full_write(tbInfo->tarFd, buffer, size) != size ) {
1129                                 /* Output file seems to have a problem */
1130                                 error_msg(io_error, fileName, strerror(errno)); 
1131                                 return( FALSE);
1132                         }
1133                         readSize+=size;
1134                 }
1135                 if (size == -1) {
1136                         error_msg(io_error, fileName, strerror(errno)); 
1137                         return( FALSE);
1138                 }
1139                 /* Pad the file up to the tar block size */
1140                 for (; (readSize%TAR_BLOCK_SIZE) != 0; readSize++) {
1141                         write(tbInfo->tarFd, "\0", 1);
1142                 }
1143                 close( inputFileFd);
1144         }
1145
1146         return( TRUE);
1147 }
1148
1149 static int writeTarFile(const char* tarName, int verboseFlag, char **argv,
1150                 char** excludeList)
1151 {
1152         int tarFd=-1;
1153         int errorFlag=FALSE;
1154         ssize_t size;
1155         struct TarBallInfo tbInfo;
1156         tbInfo.verboseFlag = verboseFlag;
1157         tbInfo.hlInfoHead = NULL;
1158
1159         /* Make sure there is at least one file to tar up.  */
1160         if (*argv == NULL)
1161                 error_msg_and_die("Cowardly refusing to create an empty archive\n");
1162
1163         /* Open the tar file for writing.  */
1164         if (!strcmp(tarName, "-"))
1165                 tbInfo.tarFd = fileno(stdout);
1166         else
1167                 tbInfo.tarFd = open (tarName, O_WRONLY | O_CREAT | O_TRUNC, 0644);
1168         if (tbInfo.tarFd < 0) {
1169                 error_msg( "Error opening '%s': %s\n", tarName, strerror(errno));
1170                 freeHardLinkInfo(&tbInfo.hlInfoHead);
1171                 return ( FALSE);
1172         }
1173         tbInfo.excludeList=excludeList;
1174         /* Store the stat info for the tarball's file, so
1175          * can avoid including the tarball into itself....  */
1176         if (fstat(tbInfo.tarFd, &tbInfo.statBuf) < 0)
1177                 error_msg_and_die(io_error, tarName, strerror(errno)); 
1178
1179         /* Set the umask for this process so it doesn't 
1180          * screw up permission setting for us later. */
1181         umask(0);
1182
1183         /* Read the directory/files and iterate over them one at a time */
1184         while (*argv != NULL) {
1185                 if (recursive_action(*argv++, TRUE, FALSE, FALSE,
1186                                         writeFileToTarball, writeFileToTarball, 
1187                                         (void*) &tbInfo) == FALSE) {
1188                         errorFlag = TRUE;
1189                 }
1190         }
1191         /* Write two empty blocks to the end of the archive */
1192         for (size=0; size<(2*TAR_BLOCK_SIZE); size++) {
1193                 write(tbInfo.tarFd, "\0", 1);
1194         }
1195
1196         /* To be pedantically correct, we would check if the tarball
1197          * is smaller than 20 tar blocks, and pad it if it was smaller,
1198          * but that isn't necessary for GNU tar interoperability, and
1199          * so is considered a waste of space */
1200
1201         /* Hang up the tools, close up shop, head home */
1202         close(tarFd);
1203         if (errorFlag == TRUE) {
1204                 error_msg("Error exit delayed from previous errors\n");
1205                 freeHardLinkInfo(&tbInfo.hlInfoHead);
1206                 return(FALSE);
1207         }
1208         freeHardLinkInfo(&tbInfo.hlInfoHead);
1209         return( TRUE);
1210 }
1211
1212
1213 #endif
1214