tar: fix a bug where autodetection messes up -z on extract
[oweals/busybox.git] / archival / tar.c
index c12331c832146feb2b8a2fe5574f137ef3faa0d9..1e3cecf44cc2e7c7c006225a06b87ccf11785b98 100644 (file)
  *  Copyright (c) 1999 by David I. Bell
  *  Permission is granted to use, distribute, or modify this source,
  *  provided that this copyright notice remains intact.
- *  Permission to distribute sash derived code under the GPL has been granted.
+ *  Permission to distribute sash derived code under GPL has been granted.
  *
  * Based in part on the tar implementation from busybox-0.28
  *  Copyright (C) 1995 Bruce Perens
  *
- * Licensed under GPLv2 or later, see file LICENSE in this tarball for details.
+ * Licensed under GPLv2 or later, see file LICENSE in this source tree.
+ */
+
+/* TODO: security with -C DESTDIR option can be enhanced.
+ * Consider tar file created via:
+ * $ tar cvf bug.tar anything.txt
+ * $ ln -s /tmp symlink
+ * $ tar --append -f bug.tar symlink
+ * $ rm symlink
+ * $ mkdir symlink
+ * $ tar --append -f bug.tar symlink/evil.py
+ *
+ * This will result in an archive which contains:
+ * $ tar --list -f bug.tar
+ * anything.txt
+ * symlink
+ * symlink/evil.py
+ *
+ * Untarring it puts evil.py in '/tmp' even if the -C DESTDIR is given.
+ * This doesn't feel right, and IIRC GNU tar doesn't do that.
  */
 
 #include <fnmatch.h>
 #include "libbb.h"
-#include "unarchive.h"
+#include "archive.h"
 /* FIXME: Stop using this non-standard feature */
 #ifndef FNM_LEADING_DIR
 # define FNM_LEADING_DIR 0
 
 #if ENABLE_FEATURE_TAR_CREATE
 
-/* Tar file constants  */
-
-#define TAR_BLOCK_SIZE         512
-
-/* POSIX tar Header Block, from POSIX 1003.1-1990  */
-#define NAME_SIZE      100
-#define NAME_SIZE_STR "100"
-typedef struct TarHeader {       /* byte offset */
-       char name[NAME_SIZE];     /*   0-99 */
-       char mode[8];             /* 100-107 */
-       char uid[8];              /* 108-115 */
-       char gid[8];              /* 116-123 */
-       char size[12];            /* 124-135 */
-       char mtime[12];           /* 136-147 */
-       char chksum[8];           /* 148-155 */
-       char typeflag;            /* 156-156 */
-       char linkname[NAME_SIZE]; /* 157-256 */
-       /* POSIX:   "ustar" NUL "00" */
-       /* GNU tar: "ustar  " NUL */
-       /* Normally it's defined as magic[6] followed by
-        * version[2], but we put them together to save code.
-        */
-       char magic[8];            /* 257-264 */
-       char uname[32];           /* 265-296 */
-       char gname[32];           /* 297-328 */
-       char devmajor[8];         /* 329-336 */
-       char devminor[8];         /* 337-344 */
-       char prefix[155];         /* 345-499 */
-       char padding[12];         /* 500-512 (pad to exactly TAR_BLOCK_SIZE) */
-} TarHeader;
-
 /*
 ** writeTarFile(), writeFileToTarball(), and writeTarHeader() are
 ** the only functions that deal with the HardLinkInfo structure.
@@ -193,7 +181,7 @@ static void putOctal(char *cp, int len, off_t value)
 }
 #define PUT_OCTAL(a, b) putOctal((a), sizeof(a), (b))
 
-static void chksum_and_xwrite(int fd, struct TarHeader* hp)
+static void chksum_and_xwrite(int fd, struct tar_header_t* hp)
 {
        /* POSIX says that checksum is done on unsigned bytes
         * (Sun and HP-UX gets it wrong... more details in
@@ -235,7 +223,7 @@ static void writeLongname(int fd, int type, const char *name, int dir)
                "00000000000",
                "00000000000",
        };
-       struct TarHeader header;
+       struct tar_header_t header;
        int size;
 
        dir = !!dir; /* normalize: 0/1 */
@@ -262,16 +250,12 @@ static void writeLongname(int fd, int type, const char *name, int dir)
 #endif
 
 /* Write out a tar header for the specified file/directory/whatever */
-void BUG_tar_header_size(void);
 static int writeTarHeader(struct TarBallInfo *tbInfo,
                const char *header_name, const char *fileName, struct stat *statbuf)
 {
-       struct TarHeader header;
+       struct tar_header_t header;
 
-       if (sizeof(header) != 512)
-               BUG_tar_header_size();
-
-       memset(&header, 0, sizeof(struct TarHeader));
+       memset(&header, 0, sizeof(header));
 
        strncpy(header.name, header_name, sizeof(header.name));
 
@@ -336,7 +320,7 @@ static int writeTarHeader(struct TarBallInfo *tbInfo,
                 && statbuf->st_size > (off_t)0777777777777LL
                ) {
                        bb_error_msg_and_die("can't store file '%s' "
-                               "of size %"OFF_FMT"d, aborting",
+                               "of size %"OFF_FMT"u, aborting",
                                fileName, statbuf->st_size);
                }
                header.typeflag = REGTYPE;
@@ -549,9 +533,7 @@ static void NOINLINE vfork_compressor(int tar_fd, int gzip)
        (void) &zip_exec;
 # endif
 
-       gzipPid = vfork();
-       if (gzipPid < 0)
-               bb_perror_msg_and_die("vfork");
+       gzipPid = xvfork();
 
        if (gzipPid == 0) {
                /* child */
@@ -609,8 +591,7 @@ static NOINLINE int writeTarFile(int tar_fd, int verboseFlag,
 
        /* Store the stat info for the tarball's file, so
         * can avoid including the tarball into itself....  */
-       if (fstat(tbInfo.tarFd, &tbInfo.tarFileStatBuf) < 0)
-               bb_perror_msg_and_die("can't stat tar file");
+       xfstat(tbInfo.tarFd, &tbInfo.tarFileStatBuf, "can't stat tar file");
 
 #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2
        if (gzip)
@@ -674,7 +655,7 @@ static llist_t *append_file_list_to_list(llist_t *list)
        llist_t *newlist = NULL;
 
        while (list) {
-               src_stream = xfopen_for_read(llist_pop(&list));
+               src_stream = xfopen_stdin(llist_pop(&list));
                while ((line = xmalloc_fgetline(src_stream)) != NULL) {
                        /* kill trailing '/' unless the string is just "/" */
                        char *cp = last_char_is(line, '/');
@@ -729,7 +710,7 @@ static void handle_SIGCHLD(int status)
                /* wait failed?! I'm confused... */
                return;
 
-       if (WIFEXITED(status) && WEXITSTATUS(status)==0)
+       if (WIFEXITED(status) && WEXITSTATUS(status) == 0)
                /* child exited with 0 */
                return;
        /* Cannot happen?
@@ -738,6 +719,73 @@ static void handle_SIGCHLD(int status)
 }
 #endif
 
+//usage:#define tar_trivial_usage
+//usage:       "-[" IF_FEATURE_TAR_CREATE("c") "xt"
+//usage:       IF_FEATURE_SEAMLESS_Z("Z")
+//usage:       IF_FEATURE_SEAMLESS_GZ("z")
+//usage:       IF_FEATURE_SEAMLESS_BZ2("j")
+//usage:       IF_FEATURE_SEAMLESS_LZMA("a")
+//usage:       IF_FEATURE_TAR_CREATE("h")
+//usage:       IF_FEATURE_TAR_NOPRESERVE_TIME("m")
+//usage:       "vO] "
+//usage:       IF_FEATURE_TAR_FROM("[-X FILE] [-T FILE] ")
+//usage:       "[-f TARFILE] [-C DIR] [FILE]..."
+//usage:#define tar_full_usage "\n\n"
+//usage:       IF_FEATURE_TAR_CREATE("Create, extract, ")
+//usage:       IF_NOT_FEATURE_TAR_CREATE("Extract ")
+//usage:       "or list files from a tar file\n"
+//usage:     "\nOperation:"
+//usage:       IF_FEATURE_TAR_CREATE(
+//usage:     "\n       c       Create"
+//usage:       )
+//usage:     "\n       x       Extract"
+//usage:     "\n       t       List"
+//usage:     "\nOptions:"
+//usage:     "\n       f       Name of TARFILE ('-' for stdin/out)"
+//usage:     "\n       C       Change to DIR before operation"
+//usage:     "\n       v       Verbose"
+//usage:       IF_FEATURE_SEAMLESS_Z(
+//usage:     "\n       Z       (De)compress using compress"
+//usage:       )
+//usage:       IF_FEATURE_SEAMLESS_GZ(
+//usage:     "\n       z       (De)compress using gzip"
+//usage:       )
+//usage:       IF_FEATURE_SEAMLESS_BZ2(
+//usage:     "\n       j       (De)compress using bzip2"
+//usage:       )
+//usage:       IF_FEATURE_SEAMLESS_LZMA(
+//usage:     "\n       a       (De)compress using lzma"
+//usage:       )
+//usage:     "\n       O       Extract to stdout"
+//usage:       IF_FEATURE_TAR_CREATE(
+//usage:     "\n       h       Follow symlinks"
+//usage:       )
+//usage:       IF_FEATURE_TAR_NOPRESERVE_TIME(
+//usage:     "\n       m       Don't restore mtime"
+//usage:       )
+//usage:       IF_FEATURE_TAR_FROM(
+//usage:       IF_FEATURE_TAR_LONG_OPTIONS(
+//usage:     "\n       exclude File to exclude"
+//usage:       )
+//usage:     "\n       X       File with names to exclude"
+//usage:     "\n       T       File with names to include"
+//usage:       )
+//usage:
+//usage:#define tar_example_usage
+//usage:       "$ zcat /tmp/tarball.tar.gz | tar -xf -\n"
+//usage:       "$ tar -cf /tmp/tarball.tar /usr/local\n"
+
+// Supported but aren't in --help:
+//     o       no-same-owner
+//     p       same-permissions
+//     k       keep-old
+//     numeric-owner
+//     no-same-permissions
+//     overwrite
+//IF_FEATURE_TAR_TO_COMMAND(
+//     to-command
+//)
+
 enum {
        OPTBIT_KEEP_OLD = 8,
        IF_FEATURE_TAR_CREATE(   OPTBIT_CREATE      ,)
@@ -748,15 +796,19 @@ enum {
        IF_FEATURE_TAR_FROM(     OPTBIT_EXCLUDE_FROM,)
        IF_FEATURE_SEAMLESS_GZ(  OPTBIT_GZIP        ,)
        IF_FEATURE_SEAMLESS_Z(   OPTBIT_COMPRESS    ,) // 16th bit
+       IF_FEATURE_TAR_NOPRESERVE_TIME(OPTBIT_NOPRESERVE_TIME,)
+#if ENABLE_FEATURE_TAR_LONG_OPTIONS
+       IF_FEATURE_TAR_TO_COMMAND(OPTBIT_2COMMAND   ,)
        OPTBIT_NUMERIC_OWNER,
-       OPTBIT_NOPRESERVE_OWNER,
        OPTBIT_NOPRESERVE_PERM,
+       OPTBIT_OVERWRITE,
+#endif
        OPT_TEST         = 1 << 0, // t
        OPT_EXTRACT      = 1 << 1, // x
        OPT_BASEDIR      = 1 << 2, // C
        OPT_TARNAME      = 1 << 3, // f
        OPT_2STDOUT      = 1 << 4, // O
-       OPT_NOPRESERVE_OWNER = 1 << 5, // no-same-owner
+       OPT_NOPRESERVE_OWNER = 1 << 5, // o == no-same-owner
        OPT_P            = 1 << 6, // p
        OPT_VERBOSE      = 1 << 7, // v
        OPT_KEEP_OLD     = 1 << 8, // k
@@ -768,8 +820,11 @@ enum {
        OPT_EXCLUDE_FROM = IF_FEATURE_TAR_FROM(     (1 << OPTBIT_EXCLUDE_FROM)) + 0, // X
        OPT_GZIP         = IF_FEATURE_SEAMLESS_GZ(  (1 << OPTBIT_GZIP        )) + 0, // z
        OPT_COMPRESS     = IF_FEATURE_SEAMLESS_Z(   (1 << OPTBIT_COMPRESS    )) + 0, // Z
-       OPT_NUMERIC_OWNER = 1 << OPTBIT_NUMERIC_OWNER,
-       OPT_NOPRESERVE_PERM = 1 << OPTBIT_NOPRESERVE_PERM, // no-same-permissions
+       OPT_NOPRESERVE_TIME = IF_FEATURE_TAR_NOPRESERVE_TIME((1 << OPTBIT_NOPRESERVE_TIME)) + 0, // m
+       OPT_2COMMAND        = IF_FEATURE_TAR_TO_COMMAND(  (1 << OPTBIT_2COMMAND       )) + 0, // to-command
+       OPT_NUMERIC_OWNER   = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NUMERIC_OWNER  )) + 0, // numeric-owner
+       OPT_NOPRESERVE_PERM = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NOPRESERVE_PERM)) + 0, // no-same-permissions
+       OPT_OVERWRITE       = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_OVERWRITE      )) + 0, // overwrite
 };
 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
 static const char tar_longopts[] ALIGN1 =
@@ -804,14 +859,21 @@ static const char tar_longopts[] ALIGN1 =
 # endif
 # if ENABLE_FEATURE_SEAMLESS_Z
        "compress\0"            No_argument       "Z"
+# endif
+# if ENABLE_FEATURE_TAR_NOPRESERVE_TIME
+       "touch\0"               No_argument       "m"
+# endif
+# if ENABLE_FEATURE_TAR_TO_COMMAND
+       "to-command\0"          Required_argument "\xfb"
 # endif
        /* use numeric uid/gid from tar header, not textual */
-       "numeric-owner\0"       No_argument       "\xfd"
+       "numeric-owner\0"       No_argument       "\xfc"
        /* do not restore mode */
-       "no-same-permissions\0" No_argument       "\xfe"
+       "no-same-permissions\0" No_argument       "\xfd"
+       /* on unpack, open with O_TRUNC and !O_EXCL */
+       "overwrite\0"           No_argument       "\xfe"
        /* --exclude takes next bit position in option mask, */
-       /* therefore we have to either put it _after_ --no-same-perm */
-       /* or add OPT[BIT]_EXCLUDE before OPT[BIT]_NOPRESERVE_OWNER */
+       /* therefore we have to put it _after_ --no-same-permissions */
 # if ENABLE_FEATURE_TAR_FROM
        "exclude\0"             Required_argument "\xff"
 # endif
@@ -844,7 +906,6 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
        /* Prepend '-' to the first argument if required */
        opt_complementary = "--:" // first arg is options
                "tt:vv:" // count -t,-v
-               "?:" // bail out with usage instead of error return
                "X::T::" // cumulative lists
 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
                "\xff::" // cumulative lists for --exclude
@@ -856,7 +917,7 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
        applet_long_options = tar_longopts;
 #endif
 #if ENABLE_DESKTOP
-       if (argv[1][0] != '-') {
+       if (argv[1] && argv[1][0] != '-') {
                /* Compat:
                 * 1st argument without dash handles options with parameters
                 * differently from dashed one: it takes *next argv[i]*
@@ -890,10 +951,12 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
                IF_FEATURE_TAR_FROM(     "T:X:")
                IF_FEATURE_SEAMLESS_GZ(  "z"   )
                IF_FEATURE_SEAMLESS_Z(   "Z"   )
+               IF_FEATURE_TAR_NOPRESERVE_TIME("m")
                , &base_dir // -C dir
                , &tar_filename // -f filename
                IF_FEATURE_TAR_FROM(, &(tar_handle->accept)) // T
                IF_FEATURE_TAR_FROM(, &(tar_handle->reject)) // X
+               IF_FEATURE_TAR_TO_COMMAND(, &(tar_handle->tar__to_command)) // --to-command
 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
                , &excludes // --exclude
 #endif
@@ -912,6 +975,12 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
        if (opt & OPT_2STDOUT)
                tar_handle->action_data = data_extract_to_stdout;
 
+       if (opt & OPT_2COMMAND) {
+               putenv((char*)"TAR_FILETYPE=f");
+               signal(SIGPIPE, SIG_IGN);
+               tar_handle->action_data = data_extract_to_command;
+       }
+
        if (opt & OPT_KEEP_OLD)
                tar_handle->ah_flags &= ~ARCHIVE_UNLINK_OLD;
 
@@ -924,6 +993,11 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
        if (opt & OPT_NOPRESERVE_PERM)
                tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_PERM;
 
+       if (opt & OPT_OVERWRITE) {
+               tar_handle->ah_flags &= ~ARCHIVE_UNLINK_OLD;
+               tar_handle->ah_flags |= ARCHIVE_O_TRUNC;
+       }
+
        if (opt & OPT_GZIP)
                get_header_ptr = get_header_tar_gz;
 
@@ -936,9 +1010,12 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
        if (opt & OPT_COMPRESS)
                get_header_ptr = get_header_tar_Z;
 
+       if (opt & OPT_NOPRESERVE_TIME)
+               tar_handle->ah_flags &= ~ARCHIVE_RESTORE_DATE;
+
 #if ENABLE_FEATURE_TAR_FROM
        tar_handle->reject = append_file_list_to_list(tar_handle->reject);
-#if ENABLE_FEATURE_TAR_LONG_OPTIONS
+# if ENABLE_FEATURE_TAR_LONG_OPTIONS
        /* Append excludes to reject */
        while (excludes) {
                llist_t *next = excludes->link;
@@ -946,12 +1023,12 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
                tar_handle->reject = excludes;
                excludes = next;
        }
-#endif
+# endif
        tar_handle->accept = append_file_list_to_list(tar_handle->accept);
 #endif
 
        /* Setup an array of filenames to work with */
-       /* TODO: This is the same as in ar, separate function ? */
+       /* TODO: This is the same as in ar, make a separate function? */
        while (*argv) {
                /* kill trailing '/' unless the string is just "/" */
                char *cp = last_char_is(*argv, '/');
@@ -970,7 +1047,7 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
                int flags = O_RDONLY;
 
                if (opt & OPT_CREATE) {
-                       /* Make sure there is at least one file to tar up */
+                       /* Make sure there is at least one file to tar up */
                        if (tar_handle->accept == NULL)
                                bb_error_msg_and_die("empty archive");
 
@@ -983,8 +1060,10 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
                        tar_handle->src_fd = tar_fd;
                        tar_handle->seek = seek_by_read;
                } else {
-                       if (ENABLE_FEATURE_TAR_AUTODETECT && flags == O_RDONLY) {
-                               get_header_ptr = get_header_tar;
+                       if (ENABLE_FEATURE_TAR_AUTODETECT
+                        && flags == O_RDONLY
+                        && get_header_ptr == get_header_tar
+                       ) {
                                tar_handle->src_fd = open_zipped(tar_filename);
                                if (tar_handle->src_fd < 0)
                                        bb_perror_msg_and_die("can't open '%s'", tar_filename);
@@ -1002,7 +1081,7 @@ int tar_main(int argc UNUSED_PARAM, char **argv)
        signal(SIGCHLD, handle_SIGCHLD);
 #endif
 
-       /* create an archive */
+       /* Create an archive */
        if (opt & OPT_CREATE) {
 #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2
                int zipMode = 0;