/* Create a tar archive.
- Copyright (C) 1985, 1992 Free Software Foundation
-This file is part of GNU Tar.
+ Copyright (C) 1985, 1992, 1993, 1994, 1996, 1997, 1999, 2000, 2001,
+ 2003, 2004, 2005, 2006, 2007, 2009, 2010 Free Software Foundation, Inc.
-GNU Tar is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2, or (at your option)
-any later version.
+ Written by John Gilmore, on 1985-08-25.
-GNU Tar is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
-GNU General Public License for more details.
+ This program is free software; you can redistribute it and/or modify it
+ under the terms of the GNU General Public License as published by the
+ Free Software Foundation; either version 3, or (at your option) any later
+ version.
-You should have received a copy of the GNU General Public License
-along with GNU Tar; see the file COPYING. If not, write to
-the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */
+ This program is distributed in the hope that it will be useful, but
+ WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
+ Public License for more details.
-/*
- * Create a tar archive.
- *
- * Written 25 Aug 1985 by John Gilmore, ihnp4!hoptoad!gnu.
- */
+ You should have received a copy of the GNU General Public License along
+ with this program; if not, write to the Free Software Foundation, Inc.,
+ 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */
-#ifdef _AIX
- #pragma alloca
-#endif
-#include <sys/types.h>
-#include <stdio.h>
-#include <errno.h>
-#ifndef STDC_HEADERS
-extern int errno;
-#endif
+#include <system.h>
-#ifdef BSD42
-#include <sys/file.h>
-#else
-#ifndef V7
-#include <fcntl.h>
-#endif
-#endif
+#include <quotearg.h>
-#include "tar.h"
-#include "port.h"
+#include "common.h"
+#include <hash.h>
-#ifndef __MSDOS__
-#include <pwd.h>
-#include <grp.h>
-#endif
+struct link
+ {
+ dev_t dev;
+ ino_t ino;
+ nlink_t nlink;
+ char name[1];
+ };
-#if defined(_POSIX_VERSION) || defined(DIRENT)
-#include <dirent.h>
-#ifdef direct
-#undef direct
-#endif /* direct */
-#define direct dirent
-#define DP_NAMELEN(x) strlen((x)->d_name)
-#endif /* _POSIX_VERSION or DIRENT */
-#if !defined(_POSIX_VERSION) && !defined(DIRENT) && defined(BSD42)
-#include <sys/dir.h>
-#define DP_NAMELEN(x) (x)->d_namlen
-#endif /* not _POSIX_VERSION and BSD42 */
-#ifdef __MSDOS__
-#include "msd_dir.h"
-#define DP_NAMELEN(x) (x)->d_namlen
-#define direct dirent
-#endif
-#if defined(USG) && !defined(_POSIX_VERSION) && !defined(DIRENT)
-#include <ndir.h>
-#define DP_NAMELEN(x) strlen((x)->d_name)
-#endif /* USG and not _POSIX_VERSION and not DIRENT */
+struct exclusion_tag
+{
+ const char *name;
+ size_t length;
+ enum exclusion_tag_type type;
+ bool (*predicate) (const char *name);
+ struct exclusion_tag *next;
+};
-extern struct stat hstat; /* Stat struct corresponding */
+static struct exclusion_tag *exclusion_tags;
-#ifndef __MSDOS__
-extern dev_t ar_dev;
-extern ino_t ar_ino;
-#endif
+void
+add_exclusion_tag (const char *name, enum exclusion_tag_type type,
+ bool (*predicate) (const char *name))
+{
+ struct exclusion_tag *tag = xmalloc (sizeof tag[0]);
+ tag->next = exclusion_tags;
+ tag->name = name;
+ tag->type = type;
+ tag->predicate = predicate;
+ tag->length = strlen (name);
+ exclusion_tags = tag;
+}
-/* JF */
-extern struct name *gnu_list_name;
+void
+exclusion_tag_warning (const char *dirname, const char *tagname,
+ const char *message)
+{
+ if (verbose_option)
+ WARNOPT (WARN_CACHEDIR,
+ (0, 0,
+ _("%s: contains a cache directory tag %s; %s"),
+ quotearg_colon (dirname),
+ quotearg_n (1, tagname),
+ message));
+}
-/*
- * If there are no symbolic links, there is no lstat(). Use stat().
- */
-#ifndef S_ISLNK
-#define lstat stat
-#endif
+enum exclusion_tag_type
+check_exclusion_tags (const char *dirname, const char **tag_file_name)
+{
+ static char *tagname;
+ static size_t tagsize;
+ struct exclusion_tag *tag;
+ size_t dlen = strlen (dirname);
+ int addslash = !ISSLASH (dirname[dlen-1]);
+ size_t noff = 0;
+
+ for (tag = exclusion_tags; tag; tag = tag->next)
+ {
+ size_t size = dlen + addslash + tag->length + 1;
+ if (size > tagsize)
+ {
+ tagsize = size;
+ tagname = xrealloc (tagname, tagsize);
+ }
-extern void print_header();
-
-union record *start_header();
-void blank_name_list();
-int check_exclude();
-PTR ck_malloc();
-PTR ck_realloc();
-void clear_buffer();
-void close_archive();
-void collect_and_sort_names();
-int confirm();
-int deal_with_sparse();
-void find_new_file_size();
-void finish_header();
-int finish_sparse_file();
-void finduname();
-void findgname();
-int is_dot_or_dotdot();
-void open_archive();
-char *name_next();
-void name_close();
-void to_oct();
-void dump_file();
-void write_dir_file();
-void write_eot();
-int zero_record();
-
-/* This code moved from tar.h since create.c is the only file that cares
- about 'struct link's. This means that other files might not have to
- include sys/types.h any more. */
-
-struct link {
- struct link *next;
- dev_t dev;
- ino_t ino;
- short linkcount;
- char name[1];
-};
+ if (noff == 0)
+ {
+ strcpy (tagname, dirname);
+ noff = dlen;
+ if (addslash)
+ tagname[noff++] = '/';
+ }
+ strcpy (tagname + noff, tag->name);
+ if (access (tagname, F_OK) == 0
+ && (!tag->predicate || tag->predicate (tagname)))
+ {
+ if (tag_file_name)
+ *tag_file_name = tag->name;
+ return tag->type;
+ }
+ }
-struct link *linklist; /* Points to first link in list */
+ return exclusion_tag_none;
+}
-static nolinks; /* Gets set if we run out of RAM */
+/* Exclusion predicate to test if the named file (usually "CACHEDIR.TAG")
+ contains a valid header, as described at:
+ http://www.brynosaurus.com/cachedir
+ Applications can write this file into directories they create
+ for use as caches containing purely regenerable, non-precious data,
+ allowing us to avoid archiving them if --exclude-caches is specified. */
-/*
- * "Scratch" space to store the information about a sparse file before
- * writing the info into the header or extended header
- */
-/* struct sp_array *sparsearray;*/
+#define CACHEDIR_SIGNATURE "Signature: 8a477f597d28d172789f06886806bc55"
+#define CACHEDIR_SIGNATURE_SIZE (sizeof CACHEDIR_SIGNATURE - 1)
-/* number of elts storable in the sparsearray */
-/*int sparse_array_size = 10;*/
+bool
+cachedir_file_p (const char *name)
+{
+ bool tag_present = false;
+ int fd = open (name, O_RDONLY);
+ if (fd >= 0)
+ {
+ static char tagbuf[CACHEDIR_SIGNATURE_SIZE];
-void
-create_archive()
+ if (read (fd, tagbuf, CACHEDIR_SIGNATURE_SIZE)
+ == CACHEDIR_SIGNATURE_SIZE
+ && memcmp (tagbuf, CACHEDIR_SIGNATURE, CACHEDIR_SIGNATURE_SIZE) == 0)
+ tag_present = true;
+
+ close (fd);
+ }
+ return tag_present;
+}
+
+\f
+/* The maximum uintmax_t value that can be represented with DIGITS digits,
+ assuming that each digit is BITS_PER_DIGIT wide. */
+#define MAX_VAL_WITH_DIGITS(digits, bits_per_digit) \
+ ((digits) * (bits_per_digit) < sizeof (uintmax_t) * CHAR_BIT \
+ ? ((uintmax_t) 1 << ((digits) * (bits_per_digit))) - 1 \
+ : (uintmax_t) -1)
+
+/* The maximum uintmax_t value that can be represented with octal
+ digits and a trailing NUL in BUFFER. */
+#define MAX_OCTAL_VAL(buffer) MAX_VAL_WITH_DIGITS (sizeof (buffer) - 1, LG_8)
+
+/* Convert VALUE to an octal representation suitable for tar headers.
+ Output to buffer WHERE with size SIZE.
+ The result is undefined if SIZE is 0 or if VALUE is too large to fit. */
+
+static void
+to_octal (uintmax_t value, char *where, size_t size)
{
- register char *p;
- char *name_from_list();
-
- open_archive(0); /* Open for writing */
-
- if(f_gnudump) {
- char *buf = ck_malloc(PATH_MAX);
- char *q,*bufp;
-
- collect_and_sort_names();
-
- while(p=name_from_list())
- dump_file(p,-1, 1);
- /* if(!f_dironly) { */
- blank_name_list();
- while(p=name_from_list()) {
- strcpy(buf,p);
- if(p[strlen(p)-1]!='/')
- strcat(buf,"/");
- bufp=buf+strlen(buf);
- for(q=gnu_list_name->dir_contents;q && *q;q+=strlen(q)+1) {
- if(*q=='Y') {
- strcpy(bufp,q+1);
- dump_file(buf,-1, 1);
- }
- }
- }
- /* } */
- free(buf);
- } else {
- p = name_next(1);
- do
- dump_file(p, -1, 1);
- while (p = name_next(1));
- }
+ uintmax_t v = value;
+ size_t i = size;
- write_eot();
- close_archive();
- if(f_gnudump)
- write_dir_file();
- name_close();
+ do
+ {
+ where[--i] = '0' + (v & ((1 << LG_8) - 1));
+ v >>= LG_8;
+ }
+ while (i);
}
-/*
- * Dump a single file. If it's a directory, recurse.
- * Result is 1 for success, 0 for failure.
- * Sets global "hstat" to stat() output for this file.
- */
-void
-dump_file (p, curdev, toplevel)
- char *p; /* File name to dump */
- int curdev; /* Device our parent dir was on */
- int toplevel; /* Whether we are a toplevel call */
+/* Copy at most LEN bytes from the string SRC to DST. Terminate with
+ NUL unless SRC is LEN or more bytes long. */
+
+static void
+tar_copy_str (char *dst, const char *src, size_t len)
{
- union record *header;
- char type;
- extern char *save_name; /* JF for multi-volume support */
- extern long save_totsize;
- extern long save_sizeleft;
- union record *exhdr;
- char save_linkflag;
- extern time_t new_time;
- int critical_error = 0;
- time_t restore_times[2];
-/* int sparse_ind = 0;*/
-
-
- if(f_confirm && !confirm("add",p))
- return;
-
- /*
- * Use stat if following (rather than dumping) 4.2BSD's
- * symbolic links. Otherwise, use lstat (which, on non-4.2
- * systems, is #define'd to stat anyway.
- */
-#ifdef STX_HIDDEN /* AIX */
- if (0 != f_follow_links ?
- statx (p, &hstat, STATSIZE, STX_HIDDEN):
- statx (p, &hstat, STATSIZE, STX_HIDDEN|STX_LINK))
-#else
- if (0 != f_follow_links? stat(p, &hstat): lstat(p, &hstat))
-#endif
- {
-badperror:
- msg_perror("can't add file %s",p);
-badfile:
- if (!f_ignore_failed_read || critical_error)
- errors++;
- return;
- }
-
- restore_times[0] = hstat.st_atime;
- restore_times[1] = hstat.st_mtime;
+ size_t i;
+ for (i = 0; i < len; i++)
+ if (! (dst[i] = src[i]))
+ break;
+}
-#ifdef S_ISHIDDEN
- if (S_ISHIDDEN (hstat.st_mode)) {
- char *new = (char *)alloca (strlen (p) + 2);
- if (new) {
- strcpy (new, p);
- strcat (new, "@");
- p = new;
- }
- }
-#endif
+/* Same as tar_copy_str, but always terminate with NUL if using
+ is OLDGNU format */
- /* See if we only want new files, and check if this one is too old to
- put in the archive. */
- if( f_new_files
- && !f_gnudump
- && new_time>hstat.st_mtime
- && !S_ISDIR(hstat.st_mode)
- && (f_new_files>1 || new_time>hstat.st_ctime)) {
- if(curdev<0) {
- msg("%s: is unchanged; not dumped",p);
- }
- return;
- }
+static void
+tar_name_copy_str (char *dst, const char *src, size_t len)
+{
+ tar_copy_str (dst, src, len);
+ if (archive_format == OLDGNU_FORMAT)
+ dst[len-1] = 0;
+}
-#ifndef __MSDOS__
- /* See if we are trying to dump the archive */
- if(ar_dev && hstat.st_dev==ar_dev && hstat.st_ino==ar_ino) {
- msg("%s is the archive; not dumped",p);
- return;
- }
-#endif
- /*
- * Check for multiple links.
- *
- * We maintain a list of all such files that we've written so
- * far. Any time we see another, we check the list and
- * avoid dumping the data again if we've done it once already.
- */
- if (hstat.st_nlink > 1
- && (S_ISREG(hstat.st_mode)
-#ifdef S_ISCTG
- || S_ISCTG(hstat.st_mode)
-#endif
-#ifdef S_ISCHR
- || S_ISCHR(hstat.st_mode)
-#endif
-#ifdef S_ISBLK
- || S_ISBLK(hstat.st_mode)
-#endif
-#ifdef S_ISFIFO
- || S_ISFIFO(hstat.st_mode)
-#endif
- )) {
- register struct link *lp;
-
- /* First quick and dirty. Hashing, etc later FIXME */
- for (lp = linklist; lp; lp = lp->next) {
- if (lp->ino == hstat.st_ino &&
- lp->dev == hstat.st_dev) {
- char *link_name = lp->name;
-
- /* We found a link. */
- while(!f_absolute_paths && *link_name == '/') {
- static int link_warn = 0;
-
- if (!link_warn) {
- msg("Removing leading / from absolute links");
- link_warn++;
- }
- link_name++;
- }
- if (link_name - lp->name >= NAMSIZ)
- write_long (link_name, LF_LONGLINK);
-
- hstat.st_size = 0;
- header = start_header(p, &hstat);
- if (header == NULL)
- {
- critical_error = 1;
- goto badfile;
- }
- strncpy(header->header.arch_linkname,
- link_name,NAMSIZ);
-
- /* Force null truncated */
- header->header.arch_linkname [NAMSIZ-1] = 0;
-
- header->header.linkflag = LF_LINK;
- finish_header(header);
- /* FIXME: Maybe remove from list after all links found? */
- if (f_remove_files)
- {
- if (unlink (p) == -1)
- msg_perror ("cannot remove %s", p);
- }
- return; /* We dumped it */
- }
- }
+/* Convert NEGATIVE VALUE to a base-256 representation suitable for
+ tar headers. NEGATIVE is 1 if VALUE was negative before being cast
+ to uintmax_t, 0 otherwise. Output to buffer WHERE with size SIZE.
+ The result is undefined if SIZE is 0 or if VALUE is too large to
+ fit. */
- /* Not found. Add it to the list of possible links. */
- lp = (struct link *)malloc((unsigned)(sizeof(struct link)+strlen(p)));
- if (!lp) {
- if (!nolinks) {
- msg(
- "no memory for links, they will be dumped as separate files");
- nolinks++;
- }
- }
- lp->ino = hstat.st_ino;
- lp->dev = hstat.st_dev;
- strcpy(lp->name, p);
- lp->next = linklist;
- linklist = lp;
- }
+static void
+to_base256 (int negative, uintmax_t value, char *where, size_t size)
+{
+ uintmax_t v = value;
+ uintmax_t propagated_sign_bits =
+ ((uintmax_t) - negative << (CHAR_BIT * sizeof v - LG_256));
+ size_t i = size;
- /*
- * This is not a link to a previously dumped file, so dump it.
- */
- if (S_ISREG(hstat.st_mode)
-#ifdef S_ISCTG
- || S_ISCTG(hstat.st_mode)
-#endif
- )
- {
- int f; /* File descriptor */
- long bufsize, count;
- long sizeleft;
- register union record *start;
- int header_moved;
- char isextended = 0;
- int upperbound;
-/* int end_nulls = 0; */
-
- header_moved = 0;
-
-#ifdef BSD42
- if (f_sparse_files) {
- /*
- * JK - This is the test for sparseness: whether the
- * "size" of the file matches the number of blocks
- * allocated for it. If there is a smaller number
- * of blocks that would be necessary to accommodate
- * a file of this size, we have a sparse file, i.e.,
- * at least one of those records in the file is just
- * a useless hole.
- */
-#ifdef hpux /* Nice of HPUX to gratuitiously change it, huh? - mib */
- if (hstat.st_size - (hstat.st_blocks * 1024) > 1024 )
-#else
- if (hstat.st_size - (hstat.st_blocks * RECORDSIZE) > RECORDSIZE)
-#endif
- {
- int filesize = hstat.st_size;
- register int i;
-
- header = start_header(p, &hstat);
- if (header == NULL)
- {
- critical_error = 1;
- goto badfile;
- }
- header->header.linkflag = LF_SPARSE;
- header_moved++;
-
- /*
- * Call the routine that figures out the
- * layout of the sparse file in question.
- * UPPERBOUND is the index of the last
- * element of the "sparsearray," i.e.,
- * the number of elements it needed to
- * describe the file.
- */
-
- upperbound = deal_with_sparse(p, header);
-
- /*
- * See if we'll need an extended header
- * later
- */
- if (upperbound > SPARSE_IN_HDR-1)
- header->header.isextended++;
- /*
- * We store the "real" file size so
- * we can show that in case someone wants
- * to list the archive, i.e., tar tvf <file>.
- * It might be kind of disconcerting if the
- * shrunken file size was the one that showed
- * up.
- */
- to_oct((long) hstat.st_size, 1+12,
- header->header.realsize);
-
- /*
- * This will be the new "size" of the
- * file, i.e., the size of the file
- * minus the records of holes that we're
- * skipping over.
- */
-
- find_new_file_size(&filesize, upperbound);
- hstat.st_size = filesize;
- to_oct((long) filesize, 1+12,
- header->header.size);
-/* to_oct((long) end_nulls, 1+12,
- header->header.ending_blanks);*/
-
- for (i = 0; i < SPARSE_IN_HDR; i++) {
- if (!sparsearray[i].numbytes)
- break;
- to_oct(sparsearray[i].offset, 1+12,
- header->header.sp[i].offset);
- to_oct(sparsearray[i].numbytes, 1+12,
- header->header.sp[i].numbytes);
- }
-
- }
- }
-#else
- upperbound=SPARSE_IN_HDR-1;
-#endif
-
- sizeleft = hstat.st_size;
- /* Don't bother opening empty, world readable files. */
- if (sizeleft > 0 || 0444 != (0444 & hstat.st_mode)) {
- f = open(p, O_RDONLY|O_BINARY);
- if (f < 0) goto badperror;
- } else {
- f = -1;
- }
-
- /* If the file is sparse, we've already taken care of this */
- if (!header_moved) {
- header = start_header(p, &hstat);
- if (header == NULL) {
- if(f>=0)
- (void)close(f);
- critical_error = 1;
- goto badfile;
- }
- }
-#ifdef S_ISCTG
- /* Mark contiguous files, if we support them */
- if (f_standard && S_ISCTG(hstat.st_mode)) {
- header->header.linkflag = LF_CONTIG;
- }
-#endif
- isextended = header->header.isextended;
- save_linkflag = header->header.linkflag;
- finish_header(header);
- if (isextended) {
-/* int sum = 0;*/
- register int i;
-/* register union record *exhdr;*/
-/* int arraybound = SPARSE_EXT_HDR;*/
- /* static */ int index_offset = SPARSE_IN_HDR;
-
- extend: exhdr = findrec();
-
- if (exhdr == NULL)
- {
- critical_error = 1;
- goto badfile;
- }
- bzero(exhdr->charptr, RECORDSIZE);
- for (i = 0; i < SPARSE_EXT_HDR; i++) {
- if (i+index_offset > upperbound)
- break;
- to_oct((long) sparsearray[i+index_offset].numbytes,
- 1+12,
- exhdr->ext_hdr.sp[i].numbytes);
- to_oct((long) sparsearray[i+index_offset].offset,
- 1+12,
- exhdr->ext_hdr.sp[i].offset);
- }
- userec(exhdr);
-/* sum += i;
- if (sum < upperbound)
- goto extend;*/
- if (index_offset+i < upperbound) {
- index_offset += i;
- exhdr->ext_hdr.isextended++;
- goto extend;
- }
-
- }
- if (save_linkflag == LF_SPARSE) {
- if (finish_sparse_file(f, &sizeleft, hstat.st_size, p))
- goto padit;
- }
- else
- while (sizeleft > 0) {
-
- if(f_multivol) {
- save_name = p;
- save_sizeleft = sizeleft;
- save_totsize = hstat.st_size;
- }
- start = findrec();
-
- bufsize = endofrecs()->charptr - start->charptr;
-
- if (sizeleft < bufsize) {
- /* Last read -- zero out area beyond */
- bufsize = (int)sizeleft;
- count = bufsize % RECORDSIZE;
- if (count)
- bzero(start->charptr + sizeleft,
- (int)(RECORDSIZE - count));
- }
- count = read(f, start->charptr, bufsize);
- if (count < 0) {
- msg_perror("read error at byte %ld, reading\
- %d bytes, in file %s", hstat.st_size - sizeleft, bufsize,p);
- goto padit;
- }
- sizeleft -= count;
-
- /* This is nonportable (the type of userec's arg). */
- userec(start+(count-1)/RECORDSIZE);
-
- if (count == bufsize) continue;
- msg( "file %s shrunk by %d bytes, padding with zeros.", p, sizeleft);
- goto padit; /* Short read */
- }
+ do
+ {
+ where[--i] = v & ((1 << LG_256) - 1);
+ v = propagated_sign_bits | (v >> LG_256);
+ }
+ while (i);
+}
- if(f_multivol)
- save_name = 0;
+#define GID_TO_CHARS(val, where) gid_to_chars (val, where, sizeof (where))
+#define MAJOR_TO_CHARS(val, where) major_to_chars (val, where, sizeof (where))
+#define MINOR_TO_CHARS(val, where) minor_to_chars (val, where, sizeof (where))
+#define MODE_TO_CHARS(val, where) mode_to_chars (val, where, sizeof (where))
+#define UID_TO_CHARS(val, where) uid_to_chars (val, where, sizeof (where))
+#define UINTMAX_TO_CHARS(val, where) uintmax_to_chars (val, where, sizeof (where))
+#define UNAME_TO_CHARS(name,buf) string_to_chars (name, buf, sizeof(buf))
+#define GNAME_TO_CHARS(name,buf) string_to_chars (name, buf, sizeof(buf))
+
+static bool
+to_chars (int negative, uintmax_t value, size_t valsize,
+ uintmax_t (*substitute) (int *),
+ char *where, size_t size, const char *type);
+
+static bool
+to_chars_subst (int negative, int gnu_format, uintmax_t value, size_t valsize,
+ uintmax_t (*substitute) (int *),
+ char *where, size_t size, const char *type)
+{
+ uintmax_t maxval = (gnu_format
+ ? MAX_VAL_WITH_DIGITS (size - 1, LG_256)
+ : MAX_VAL_WITH_DIGITS (size - 1, LG_8));
+ char valbuf[UINTMAX_STRSIZE_BOUND + 1];
+ char maxbuf[UINTMAX_STRSIZE_BOUND];
+ char minbuf[UINTMAX_STRSIZE_BOUND + 1];
+ char const *minval_string;
+ char const *maxval_string = STRINGIFY_BIGINT (maxval, maxbuf);
+ char const *value_string;
+
+ if (gnu_format)
+ {
+ uintmax_t m = maxval + 1 ? maxval + 1 : maxval / 2 + 1;
+ char *p = STRINGIFY_BIGINT (m, minbuf + 1);
+ *--p = '-';
+ minval_string = p;
+ }
+ else
+ minval_string = "0";
- if (f >= 0)
- (void)close(f);
+ if (negative)
+ {
+ char *p = STRINGIFY_BIGINT (- value, valbuf + 1);
+ *--p = '-';
+ value_string = p;
+ }
+ else
+ value_string = STRINGIFY_BIGINT (value, valbuf);
- if (f_remove_files)
- {
- if (unlink (p) == -1)
- msg_perror ("cannot remove %s", p);
- }
- if (f_atime_preserve)
- utime (p, restore_times);
- return;
-
- /*
- * File shrunk or gave error, pad out tape to match
- * the size we specified in the header.
- */
- padit:
- while(sizeleft>0) {
- save_sizeleft=sizeleft;
- start=findrec();
- bzero(start->charptr,RECORDSIZE);
- userec(start);
- sizeleft-=RECORDSIZE;
- }
- if(f_multivol)
- save_name=0;
- if(f>=0)
- (void)close(f);
- if (f_atime_preserve)
- utime (p, restore_times);
- return;
- }
+ if (substitute)
+ {
+ int negsub;
+ uintmax_t sub = substitute (&negsub) & maxval;
+ /* NOTE: This is one of the few places where GNU_FORMAT differs from
+ OLDGNU_FORMAT. The actual differences are:
+
+ 1. In OLDGNU_FORMAT all strings in a tar header end in \0
+ 2. Incremental archives use oldgnu_header.
+
+ Apart from this they are completely identical. */
+ uintmax_t s = (negsub &= archive_format == GNU_FORMAT) ? - sub : sub;
+ char subbuf[UINTMAX_STRSIZE_BOUND + 1];
+ char *sub_string = STRINGIFY_BIGINT (s, subbuf + 1);
+ if (negsub)
+ *--sub_string = '-';
+ WARN ((0, 0, _("value %s out of %s range %s..%s; substituting %s"),
+ value_string, type, minval_string, maxval_string,
+ sub_string));
+ return to_chars (negsub, s, valsize, 0, where, size, type);
+ }
+ else
+ ERROR ((0, 0, _("value %s out of %s range %s..%s"),
+ value_string, type, minval_string, maxval_string));
+ return false;
+}
-#ifdef S_ISLNK
- else if(S_ISLNK(hstat.st_mode))
+/* Convert NEGATIVE VALUE (which was originally of size VALSIZE) to
+ external form, using SUBSTITUTE (...) if VALUE won't fit. Output
+ to buffer WHERE with size SIZE. NEGATIVE is 1 iff VALUE was
+ negative before being cast to uintmax_t; its original bitpattern
+ can be deduced from VALSIZE, its original size before casting.
+ TYPE is the kind of value being output (useful for diagnostics).
+ Prefer the POSIX format of SIZE - 1 octal digits (with leading zero
+ digits), followed by '\0'. If this won't work, and if GNU or
+ OLDGNU format is allowed, use '\200' followed by base-256, or (if
+ NEGATIVE is nonzero) '\377' followed by two's complement base-256.
+ If neither format works, use SUBSTITUTE (...) instead. Pass to
+ SUBSTITUTE the address of an 0-or-1 flag recording whether the
+ substitute value is negative. */
+
+static bool
+to_chars (int negative, uintmax_t value, size_t valsize,
+ uintmax_t (*substitute) (int *),
+ char *where, size_t size, const char *type)
+{
+ int gnu_format = (archive_format == GNU_FORMAT
+ || archive_format == OLDGNU_FORMAT);
+
+ /* Generate the POSIX octal representation if the number fits. */
+ if (! negative && value <= MAX_VAL_WITH_DIGITS (size - 1, LG_8))
+ {
+ where[size - 1] = '\0';
+ to_octal (value, where, size - 1);
+ return true;
+ }
+ else if (gnu_format)
+ {
+ /* Try to cope with the number by using traditional GNU format
+ methods */
+
+ /* Generate the base-256 representation if the number fits. */
+ if (((negative ? -1 - value : value)
+ <= MAX_VAL_WITH_DIGITS (size - 1, LG_256)))
{
- int size;
- char *buf = alloca (PATH_MAX + 1);
-
- size = readlink (p, buf, PATH_MAX + 1);
- if (size < 0)
- goto badperror;
- buf[size] = '\0';
- if (size >= NAMSIZ)
- write_long (buf, LF_LONGLINK);
-
- buf[NAMSIZ - 1] = '\0';
- if (size >= NAMSIZ)
- size = NAMSIZ - 1;
- hstat.st_size = 0; /* Force 0 size on symlink */
- header = start_header(p, &hstat);
- if (header == NULL)
- {
- critical_error = 1;
- goto badfile;
- }
- strcpy (header->header.arch_linkname, buf);
- header->header.linkflag = LF_SYMLINK;
- finish_header(header); /* Nothing more to do to it */
- if (f_remove_files)
- {
- if (unlink (p) == -1)
- msg_perror ("cannot remove %s", p);
- }
- return;
+ where[0] = negative ? -1 : 1 << (LG_256 - 1);
+ to_base256 (negative, value, where + 1, size - 1);
+ return true;
}
-#endif
- else if (S_ISDIR(hstat.st_mode))
+ /* Otherwise, if the number is negative, and if it would not cause
+ ambiguity on this host by confusing positive with negative
+ values, then generate the POSIX octal representation of the value
+ modulo 2**(field bits). The resulting tar file is
+ machine-dependent, since it depends on the host word size. Yuck!
+ But this is the traditional behavior. */
+ else if (negative && valsize * CHAR_BIT <= (size - 1) * LG_8)
{
- register DIR *dirp;
- register struct direct *d;
- char *namebuf;
- int buflen;
- register int len;
- int our_device = hstat.st_dev;
-
- /* Build new prototype name */
- len = strlen(p);
- buflen=len+NAMSIZ;
- namebuf=ck_malloc(buflen+1);
- strncpy(namebuf, p, buflen);
- while (len >= 1 && '/' == namebuf[len-1])
- len--; /* Delete trailing slashes */
- namebuf[len++] = '/'; /* Now add exactly one back */
- namebuf[len] = '\0'; /* Make sure null-terminated */
-
- /*
- * Output directory header record with permissions
- * FIXME, do this AFTER files, to avoid R/O dir problems?
- * If old archive format, don't write record at all.
- */
- if (!f_oldarch) {
- hstat.st_size = 0; /* Force 0 size on dir */
- /*
- * If people could really read standard archives,
- * this should be: (FIXME)
- header = start_header(f_standard? p: namebuf, &hstat);
- * but since they'd interpret LF_DIR records as
- * regular files, we'd better put the / on the name.
- */
- header = start_header(namebuf, &hstat);
- if (header == NULL)
- {
- critical_error = 1;
- goto badfile; /* eg name too long */
- }
-
- if (f_gnudump)
- header->header.linkflag = LF_DUMPDIR;
- else if (f_standard)
- header->header.linkflag = LF_DIR;
-
- /* If we're gnudumping, we aren't done yet so don't close it. */
- if(!f_gnudump)
- finish_header(header); /* Done with directory header */
- }
+ static int warned_once;
+ if (! warned_once)
+ {
+ warned_once = 1;
+ WARN ((0, 0, _("Generating negative octal headers")));
+ }
+ where[size - 1] = '\0';
+ to_octal (value & MAX_VAL_WITH_DIGITS (valsize * CHAR_BIT, 1),
+ where, size - 1);
+ return true;
+ }
+ /* Otherwise fall back to substitution, if possible: */
+ }
+ else
+ substitute = NULL; /* No substitution for formats, other than GNU */
- if(f_gnudump) {
- int sizeleft;
- int totsize;
- int bufsize;
- union record *start;
- int count;
- char *buf,*p_buf;
-
- buf=gnu_list_name->dir_contents; /* FOO */
- totsize=0;
- for(p_buf=buf;p_buf && *p_buf;) {
- int tmp;
-
- tmp=strlen(p_buf)+1;
- totsize+=tmp;
- p_buf+=tmp;
- }
- totsize++;
- to_oct((long)totsize,1+12,header->header.size);
- finish_header(header);
- p_buf=buf;
- sizeleft=totsize;
- while(sizeleft>0) {
- if(f_multivol) {
- save_name=p;
- save_sizeleft=sizeleft;
- save_totsize=totsize;
- }
- start=findrec();
- bufsize=endofrecs()->charptr - start->charptr;
- if(sizeleft<bufsize) {
- bufsize=sizeleft;
- count=bufsize%RECORDSIZE;
- if(count)
- bzero(start->charptr+sizeleft,RECORDSIZE-count);
- }
- bcopy(p_buf,start->charptr,bufsize);
- sizeleft-=bufsize;
- p_buf+=bufsize;
- userec(start+(bufsize-1)/RECORDSIZE);
- }
- if(f_multivol)
- save_name = 0;
- if (f_atime_preserve)
- utime (p, restore_times);
- return;
- }
+ return to_chars_subst (negative, gnu_format, value, valsize, substitute,
+ where, size, type);
+}
- /* Now output all the files in the directory */
-#if 0
- if (f_dironly)
- return; /* Unless the cmdline said not to */
+static uintmax_t
+gid_substitute (int *negative)
+{
+ gid_t r;
+#ifdef GID_NOBODY
+ r = GID_NOBODY;
+#else
+ static gid_t gid_nobody;
+ if (!gid_nobody && !gname_to_gid ("nobody", &gid_nobody))
+ gid_nobody = -2;
+ r = gid_nobody;
#endif
- /*
- * See if we are crossing from one file system to another,
- * and avoid doing so if the user only wants to dump one file system.
- */
- if (f_local_filesys && !toplevel && curdev != hstat.st_dev) {
- if(f_verbose)
- msg("%s: is on a different filesystem; not dumped",p);
- return;
- }
+ *negative = r < 0;
+ return r;
+}
+static bool
+gid_to_chars (gid_t v, char *p, size_t s)
+{
+ return to_chars (v < 0, (uintmax_t) v, sizeof v, gid_substitute, p, s, "gid_t");
+}
- errno = 0;
- dirp = opendir(p);
- if (!dirp) {
- if (errno) {
- msg_perror ("can't open directory %s",p);
- } else {
- msg("error opening directory %s",
- p);
- }
- return;
- }
+static bool
+major_to_chars (major_t v, char *p, size_t s)
+{
+ return to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "major_t");
+}
- /* Hack to remove "./" from the front of all the file names */
- if (len == 2 && namebuf[0] == '.' && namebuf[1]=='/')
- len = 0;
-
- /* Should speed this up by cd-ing into the dir, FIXME */
- while (NULL != (d=readdir(dirp))) {
- /* Skip . and .. */
- if(is_dot_or_dotdot(d->d_name))
- continue;
-
- if (DP_NAMELEN(d) + len >= buflen) {
- buflen=len+DP_NAMELEN(d);
- namebuf=ck_realloc(namebuf,buflen+1);
- /* namebuf[len]='\0';
- msg("file name %s%s too long",
- namebuf, d->d_name);
- continue; */
- }
- strcpy(namebuf+len, d->d_name);
- if(f_exclude && check_exclude(namebuf))
- continue;
- dump_file(namebuf, our_device, 0);
- }
+static bool
+minor_to_chars (minor_t v, char *p, size_t s)
+{
+ return to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "minor_t");
+}
- closedir(dirp);
- free(namebuf);
- if (f_atime_preserve)
- utime (p, restore_times);
- return;
- }
+static bool
+mode_to_chars (mode_t v, char *p, size_t s)
+{
+ /* In the common case where the internal and external mode bits are the same,
+ and we are not using POSIX or GNU format,
+ propagate all unknown bits to the external mode.
+ This matches historical practice.
+ Otherwise, just copy the bits we know about. */
+ int negative;
+ uintmax_t u;
+ if (S_ISUID == TSUID && S_ISGID == TSGID && S_ISVTX == TSVTX
+ && S_IRUSR == TUREAD && S_IWUSR == TUWRITE && S_IXUSR == TUEXEC
+ && S_IRGRP == TGREAD && S_IWGRP == TGWRITE && S_IXGRP == TGEXEC
+ && S_IROTH == TOREAD && S_IWOTH == TOWRITE && S_IXOTH == TOEXEC
+ && archive_format != POSIX_FORMAT
+ && archive_format != USTAR_FORMAT
+ && archive_format != GNU_FORMAT)
+ {
+ negative = v < 0;
+ u = v;
+ }
+ else
+ {
+ negative = 0;
+ u = ((v & S_ISUID ? TSUID : 0)
+ | (v & S_ISGID ? TSGID : 0)
+ | (v & S_ISVTX ? TSVTX : 0)
+ | (v & S_IRUSR ? TUREAD : 0)
+ | (v & S_IWUSR ? TUWRITE : 0)
+ | (v & S_IXUSR ? TUEXEC : 0)
+ | (v & S_IRGRP ? TGREAD : 0)
+ | (v & S_IWGRP ? TGWRITE : 0)
+ | (v & S_IXGRP ? TGEXEC : 0)
+ | (v & S_IROTH ? TOREAD : 0)
+ | (v & S_IWOTH ? TOWRITE : 0)
+ | (v & S_IXOTH ? TOEXEC : 0));
+ }
+ return to_chars (negative, u, sizeof v, 0, p, s, "mode_t");
+}
-#ifdef S_ISCHR
- else if (S_ISCHR(hstat.st_mode)) {
- type = LF_CHR;
- }
-#endif
+bool
+off_to_chars (off_t v, char *p, size_t s)
+{
+ return to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "off_t");
+}
-#ifdef S_ISBLK
- else if (S_ISBLK(hstat.st_mode)) {
- type = LF_BLK;
- }
-#endif
+bool
+size_to_chars (size_t v, char *p, size_t s)
+{
+ return to_chars (0, (uintmax_t) v, sizeof v, 0, p, s, "size_t");
+}
-/* Avoid screwy apollo lossage where S_IFIFO == S_IFSOCK */
-#if (_ISP__M68K == 0) && (_ISP__A88K == 0) && defined(S_ISFIFO)
- else if (S_ISFIFO(hstat.st_mode)) {
- type = LF_FIFO;
- }
-#endif
+bool
+time_to_chars (time_t v, char *p, size_t s)
+{
+ return to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "time_t");
+}
-#ifdef S_ISSOCK
- else if (S_ISSOCK(hstat.st_mode)) {
- type = LF_FIFO;
- }
+static uintmax_t
+uid_substitute (int *negative)
+{
+ uid_t r;
+#ifdef UID_NOBODY
+ r = UID_NOBODY;
+#else
+ static uid_t uid_nobody;
+ if (!uid_nobody && !uname_to_uid ("nobody", &uid_nobody))
+ uid_nobody = -2;
+ r = uid_nobody;
#endif
- else
- goto unknown;
+ *negative = r < 0;
+ return r;
+}
- if (!f_standard) goto unknown;
+static bool
+uid_to_chars (uid_t v, char *p, size_t s)
+{
+ return to_chars (v < 0, (uintmax_t) v, sizeof v, uid_substitute, p, s, "uid_t");
+}
- hstat.st_size = 0; /* Force 0 size */
- header = start_header(p, &hstat);
- if (header == NULL)
- {
- critical_error = 1;
- goto badfile; /* eg name too long */
- }
+static bool
+uintmax_to_chars (uintmax_t v, char *p, size_t s)
+{
+ return to_chars (0, v, sizeof v, 0, p, s, "uintmax_t");
+}
- header->header.linkflag = type;
-#if defined(S_IFBLK) || defined(S_IFCHR)
- if (type != LF_FIFO) {
- to_oct((long) major(hstat.st_rdev), 8,
- header->header.devmajor);
- to_oct((long) minor(hstat.st_rdev), 8,
- header->header.devminor);
- }
-#endif
+static void
+string_to_chars (char const *str, char *p, size_t s)
+{
+ tar_copy_str (p, str, s);
+ p[s - 1] = '\0';
+}
- finish_header(header);
- if (f_remove_files)
- {
- if (unlink (p) == -1)
- msg_perror ("cannot remove %s", p);
- }
- return;
+\f
+/* A file is considered dumpable if it is sparse and both --sparse and --totals
+ are specified.
+ Otherwise, it is dumpable unless any of the following conditions occur:
+
+ a) it is empty *and* world-readable, or
+ b) current archive is /dev/null */
- unknown:
- msg("%s: Unknown file type; file ignored.", p);
+static bool
+file_dumpable_p (struct tar_stat_info *st)
+{
+ if (dev_null_output)
+ return totals_option && sparse_option && ST_IS_SPARSE (st->stat);
+ return !(st->archive_file_size == 0
+ && (st->stat.st_mode & MODE_R) == MODE_R);
}
-int
-finish_sparse_file(fd, sizeleft, fullsize, name)
- int fd;
- long *sizeleft,
- fullsize;
- char *name;
+\f
+/* Writing routines. */
+
+/* Write the EOT block(s). Zero at least two blocks, through the end
+ of the record. Old tar, as previous versions of GNU tar, writes
+ garbage after two zeroed blocks. */
+void
+write_eot (void)
{
- union record *start;
- char tempbuf[RECORDSIZE];
- int bufsize,
- sparse_ind = 0,
- count;
- long pos;
- long nwritten = 0;
-
-
- while (*sizeleft > 0) {
- start = findrec();
- bzero(start->charptr, RECORDSIZE);
- bufsize = sparsearray[sparse_ind].numbytes;
- if (!bufsize) { /* we blew it, maybe */
- msg("Wrote %ld of %ld bytes to file %s",
- fullsize - *sizeleft, fullsize, name);
- break;
- }
- pos = lseek(fd, sparsearray[sparse_ind++].offset, 0);
- /*
- * If the number of bytes to be written here exceeds
- * the size of the temporary buffer, do it in steps.
- */
- while (bufsize > RECORDSIZE) {
-/* if (amt_read) {
- count = read(fd, start->charptr+amt_read, RECORDSIZE-amt_read);
- bufsize -= RECORDSIZE - amt_read;
- amt_read = 0;
- userec(start);
- start = findrec();
- bzero(start->charptr, RECORDSIZE);
- }*/
- /* store the data */
- count = read(fd, start->charptr, RECORDSIZE);
- if (count < 0) {
- msg_perror("read error at byte %ld, reading %d bytes, in file %s",
- fullsize - *sizeleft, bufsize, name);
- return 1;
- }
- bufsize -= count;
- *sizeleft -= count;
- userec(start);
- nwritten += RECORDSIZE; /* XXX */
- start = findrec();
- bzero(start->charptr, RECORDSIZE);
- }
+ union block *pointer = find_next_block ();
+ memset (pointer->buffer, 0, BLOCKSIZE);
+ set_next_block_after (pointer);
+ pointer = find_next_block ();
+ memset (pointer->buffer, 0, available_space_after (pointer));
+ set_next_block_after (pointer);
+}
+/* Write a "private" header */
+union block *
+start_private_header (const char *name, size_t size, time_t t)
+{
+ union block *header = find_next_block ();
+
+ memset (header->buffer, 0, sizeof (union block));
+
+ tar_name_copy_str (header->header.name, name, NAME_FIELD_SIZE);
+ OFF_TO_CHARS (size, header->header.size);
+
+ TIME_TO_CHARS (t, header->header.mtime);
+ MODE_TO_CHARS (S_IFREG|S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH, header->header.mode);
+ UID_TO_CHARS (getuid (), header->header.uid);
+ GID_TO_CHARS (getgid (), header->header.gid);
+ MAJOR_TO_CHARS (0, header->header.devmajor);
+ MINOR_TO_CHARS (0, header->header.devminor);
+ strncpy (header->header.magic, TMAGIC, TMAGLEN);
+ strncpy (header->header.version, TVERSION, TVERSLEN);
+ return header;
+}
- clear_buffer(tempbuf);
- count = read(fd, tempbuf, bufsize);
- bcopy(tempbuf, start->charptr, RECORDSIZE);
- if (count < 0) {
- msg_perror("read error at byte %ld, reading %d bytes, in file %s",
- fullsize - *sizeleft, bufsize, name);
- return 1;
- }
-/* if (amt_read >= RECORDSIZE) {
- amt_read = 0;
- userec(start+(count-1)/RECORDSIZE);
- if (count != bufsize) {
- msg("file %s shrunk by %d bytes, padding with zeros.", name, sizeleft);
- return 1;
- }
- start = findrec();
- } else
- amt_read += bufsize;*/
- nwritten += count; /* XXX */
- *sizeleft -= count;
- userec(start);
+/* Create a new header and store there at most NAME_FIELD_SIZE bytes of
+ the file name */
- }
- free(sparsearray);
-/* printf ("Amount actually written is (I hope) %d.\n", nwritten); */
-/* userec(start+(count-1)/RECORDSIZE);*/
- return 0;
+static union block *
+write_short_name (struct tar_stat_info *st)
+{
+ union block *header = find_next_block ();
+ memset (header->buffer, 0, sizeof (union block));
+ tar_name_copy_str (header->header.name, st->file_name, NAME_FIELD_SIZE);
+ return header;
+}
+#define FILL(field,byte) do { \
+ memset(field, byte, sizeof(field)-1); \
+ (field)[sizeof(field)-1] = 0; \
+} while (0)
+
+/* Write a GNUTYPE_LONGLINK or GNUTYPE_LONGNAME block. */
+static void
+write_gnu_long_link (struct tar_stat_info *st, const char *p, char type)
+{
+ size_t size = strlen (p) + 1;
+ size_t bufsize;
+ union block *header;
+ char *tmpname;
+
+ header = start_private_header ("././@LongLink", size, time (NULL));
+ FILL (header->header.mtime, '0');
+ FILL (header->header.mode, '0');
+ FILL (header->header.uid, '0');
+ FILL (header->header.gid, '0');
+ FILL (header->header.devmajor, 0);
+ FILL (header->header.devminor, 0);
+ uid_to_uname (0, &tmpname);
+ UNAME_TO_CHARS (tmpname, header->header.uname);
+ free (tmpname);
+ gid_to_gname (0, &tmpname);
+ GNAME_TO_CHARS (tmpname, header->header.gname);
+ free (tmpname);
+
+ strcpy (header->header.magic, OLDGNU_MAGIC);
+ header->header.typeflag = type;
+ finish_header (st, header, -1);
+
+ header = find_next_block ();
+
+ bufsize = available_space_after (header);
+
+ while (bufsize < size)
+ {
+ memcpy (header->buffer, p, bufsize);
+ p += bufsize;
+ size -= bufsize;
+ set_next_block_after (header + (bufsize - 1) / BLOCKSIZE);
+ header = find_next_block ();
+ bufsize = available_space_after (header);
+ }
+ memcpy (header->buffer, p, size);
+ memset (header->buffer + size, 0, bufsize - size);
+ set_next_block_after (header + (size - 1) / BLOCKSIZE);
}
-void
-init_sparsearray()
+static size_t
+split_long_name (const char *name, size_t length)
{
- register int i;
-
- sp_array_size = 10;
- /*
- * Make room for our scratch space -- initially is 10 elts long
- */
- sparsearray = (struct sp_array *) malloc(sp_array_size * sizeof(struct sp_array));
- for (i = 0; i < sp_array_size; i++) {
- sparsearray[i].offset = 0;
- sparsearray[i].numbytes = 0;
- }
+ size_t i;
+
+ if (length > PREFIX_FIELD_SIZE + 1)
+ length = PREFIX_FIELD_SIZE + 1;
+ else if (ISSLASH (name[length - 1]))
+ length--;
+ for (i = length - 1; i > 0; i--)
+ if (ISSLASH (name[i]))
+ break;
+ return i;
}
+static union block *
+write_ustar_long_name (const char *name)
+{
+ size_t length = strlen (name);
+ size_t i, nlen;
+ union block *header;
+ if (length > PREFIX_FIELD_SIZE + NAME_FIELD_SIZE + 1)
+ {
+ ERROR ((0, 0, _("%s: file name is too long (max %d); not dumped"),
+ quotearg_colon (name),
+ PREFIX_FIELD_SIZE + NAME_FIELD_SIZE + 1));
+ return NULL;
+ }
-/*
- * Okay, we've got a sparse file on our hands -- now, what we need to do is
- * make a pass through the file and carefully note where any data is, i.e.,
- * we want to find how far into the file each instance of data is, and how
- * many bytes are there. We store this information in the sparsearray,
- * which will later be translated into header information. For now, we use
- * the sparsearray as convenient storage.
- *
- * As a side note, this routine is a mess. If I could have found a cleaner
- * way to do it, I would have. If anyone wants to find a nicer way to do
- * this, feel free.
- */
-
-/* There is little point in trimming small amounts of null data at the */
- /* head and tail of blocks -- it's ok if we only avoid dumping blocks */
- /* of complete null data */
-int
-deal_with_sparse(name, header, nulls_at_end)
- char *name;
- union record *header;
- int nulls_at_end;
+ i = split_long_name (name, length);
+ if (i == 0 || (nlen = length - i - 1) > NAME_FIELD_SIZE || nlen == 0)
+ {
+ ERROR ((0, 0,
+ _("%s: file name is too long (cannot be split); not dumped"),
+ quotearg_colon (name)));
+ return NULL;
+ }
+
+ header = find_next_block ();
+ memset (header->buffer, 0, sizeof (header->buffer));
+ memcpy (header->header.prefix, name, i);
+ memcpy (header->header.name, name + i + 1, length - i - 1);
+
+ return header;
+}
+
+/* Write a long link name, depending on the current archive format */
+static void
+write_long_link (struct tar_stat_info *st)
{
- long numbytes = 0;
- long offset = 0;
-/* long save_offset;*/
- int fd;
-/* int current_size = hstat.st_size;*/
- int sparse_ind = 0,
- cc;
- char buf[RECORDSIZE];
-#if 0
- int read_last_data = 0; /* did we just read the last record? */
-#endif
- int amidst_data = 0;
-
- header->header.isextended = 0;
- /*
- * Can't open the file -- this problem will be caught later on,
- * so just return.
- */
- if ((fd = open(name, O_RDONLY)) < 0)
- return 0;
-
- init_sparsearray();
- clear_buffer(buf);
-
- while ((cc = read(fd, buf, sizeof buf)) != 0) {
-
- if (sparse_ind > sp_array_size-1) {
-
- /*
- * realloc the scratch area, since we've run out of room --
- */
- sparsearray = (struct sp_array *)
- realloc(sparsearray,
- 2 * sp_array_size * (sizeof(struct sp_array)));
- sp_array_size *= 2;
- }
- if (cc == sizeof buf) {
- if (zero_record(buf)) {
- if (amidst_data) {
- sparsearray[sparse_ind++].numbytes
- = numbytes;
- amidst_data = 0;
- }
- } else { /* !zero_record(buf) */
- if (amidst_data)
- numbytes += cc;
- else {
- amidst_data = 1;
- numbytes = cc;
- sparsearray[sparse_ind].offset
- = offset;
- }
- }
- } else if (cc < sizeof buf) {
- /* This has to be the last bit of the file, so this */
- /* is somewhat shorter than the above. */
- if (!zero_record(buf)) {
- if (!amidst_data) {
- amidst_data = 1;
- numbytes = cc;
- sparsearray[sparse_ind].offset
- = offset;
- } else
- numbytes += cc;
- }
- }
- offset += cc;
- clear_buffer(buf);
+ switch (archive_format)
+ {
+ case POSIX_FORMAT:
+ xheader_store ("linkpath", st, NULL);
+ break;
+
+ case V7_FORMAT: /* old V7 tar format */
+ case USTAR_FORMAT:
+ case STAR_FORMAT:
+ ERROR ((0, 0,
+ _("%s: link name is too long; not dumped"),
+ quotearg_colon (st->link_name)));
+ break;
+
+ case OLDGNU_FORMAT:
+ case GNU_FORMAT:
+ write_gnu_long_link (st, st->link_name, GNUTYPE_LONGLINK);
+ break;
+
+ default:
+ abort(); /*FIXME*/
+ }
+}
+
+static union block *
+write_long_name (struct tar_stat_info *st)
+{
+ switch (archive_format)
+ {
+ case POSIX_FORMAT:
+ xheader_store ("path", st, NULL);
+ break;
+
+ case V7_FORMAT:
+ if (strlen (st->file_name) > NAME_FIELD_SIZE-1)
+ {
+ ERROR ((0, 0, _("%s: file name is too long (max %d); not dumped"),
+ quotearg_colon (st->file_name),
+ NAME_FIELD_SIZE - 1));
+ return NULL;
}
- if (amidst_data)
- sparsearray[sparse_ind++].numbytes = numbytes;
- close(fd);
+ break;
+
+ case USTAR_FORMAT:
+ case STAR_FORMAT:
+ return write_ustar_long_name (st->file_name);
- return sparse_ind - 1;
+ case OLDGNU_FORMAT:
+ case GNU_FORMAT:
+ write_gnu_long_link (st, st->file_name, GNUTYPE_LONGNAME);
+ break;
+
+ default:
+ abort(); /*FIXME*/
+ }
+ return write_short_name (st);
}
-/*
- * Just zeroes out the buffer so we don't confuse ourselves with leftover
- * data.
- */
-void
-clear_buffer(buf)
- char *buf;
+union block *
+write_extended (bool global, struct tar_stat_info *st, union block *old_header)
{
- register int i;
+ union block *header, hp;
+ char *p;
+ int type;
+ time_t t;
- for (i = 0; i < RECORDSIZE; i++)
- buf[i] = '\0';
+ if (st->xhdr.buffer || st->xhdr.stk == NULL)
+ return old_header;
+
+ xheader_finish (&st->xhdr);
+ memcpy (hp.buffer, old_header, sizeof (hp));
+ if (global)
+ {
+ type = XGLTYPE;
+ p = xheader_ghdr_name ();
+ time (&t);
+ }
+ else
+ {
+ type = XHDTYPE;
+ p = xheader_xhdr_name (st);
+ t = st->stat.st_mtime;
+ }
+ xheader_write (type, p, t, &st->xhdr);
+ free (p);
+ header = find_next_block ();
+ memcpy (header, &hp.buffer, sizeof (hp.buffer));
+ return header;
}
-#if 0 /* I'm leaving this as a monument to Joy Kendall, who wrote it -mib */
-/*
- * JK -
- * This routine takes a character array, and tells where within that array
- * the data can be found. It skips over any zeros, and sets the first
- * non-zero point in the array to be the "start", and continues until it
- * finds non-data again, which is marked as the "end." This routine is
- * mainly for 1) seeing how far into a file we must lseek to data, given
- * that we have a sparse file, and 2) determining the "real size" of the
- * file, i.e., the number of bytes in the sparse file that are data, as
- * opposed to the zeros we are trying to skip.
- */
-where_is_data(from, to, buffer)
- int *from,
- *to;
- char *buffer;
+static union block *
+write_header_name (struct tar_stat_info *st)
{
- register int i = 0;
- register int save_to = *to;
- int amidst_data = 0;
-
-
- while (!buffer[i])
- i++;
- *from = i;
-
- if (*from < 16) /* don't bother */
- *from = 0;
- /* keep going to make sure there isn't more real
- data in this record */
- while (i < RECORDSIZE) {
- if (!buffer[i]) {
- if (amidst_data) {
- save_to = i;
- amidst_data = 0;
- }
- i++;
- }
- else if (buffer[i]) {
- if (!amidst_data)
- amidst_data = 1;
- i++;
- }
- }
- if (i == RECORDSIZE)
- *to = i;
- else
- *to = save_to;
-
+ if (archive_format == POSIX_FORMAT && !string_ascii_p (st->file_name))
+ {
+ xheader_store ("path", st, NULL);
+ return write_short_name (st);
+ }
+ else if (NAME_FIELD_SIZE - (archive_format == OLDGNU_FORMAT)
+ < strlen (st->file_name))
+ return write_long_name (st);
+ else
+ return write_short_name (st);
}
-#endif
-/* Note that this routine is only called if zero_record returned true */
-#if 0 /* But we actually don't need it at all. */
-where_is_data (from, to, buffer)
- int *from, *to;
- char *buffer;
+\f
+/* Header handling. */
+
+/* Make a header block for the file whose stat info is st,
+ and return its address. */
+
+union block *
+start_header (struct tar_stat_info *st)
{
- char *fp, *tp;
-
- for (fp = buffer; ! *fp; fp++)
- ;
- for (tp = buffer + RECORDSIZE - 1; ! *tp; tp--)
- ;
- *from = fp - buffer;
- *to = tp - buffer + 1;
-}
-#endif
+ union block *header;
+
+ header = write_header_name (st);
+ if (!header)
+ return NULL;
+
+ /* Override some stat fields, if requested to do so. */
+
+ if (owner_option != (uid_t) -1)
+ st->stat.st_uid = owner_option;
+ if (group_option != (gid_t) -1)
+ st->stat.st_gid = group_option;
+ if (mode_option)
+ st->stat.st_mode =
+ ((st->stat.st_mode & ~MODE_ALL)
+ | mode_adjust (st->stat.st_mode, S_ISDIR (st->stat.st_mode) != 0,
+ initial_umask, mode_option, NULL));
+
+ /* Paul Eggert tried the trivial test ($WRITER cf a b; $READER tvf a)
+ for a few tars and came up with the following interoperability
+ matrix:
+
+ WRITER
+ 1 2 3 4 5 6 7 8 9 READER
+ . . . . . . . . . 1 = SunOS 4.2 tar
+ # . . # # . . # # 2 = NEC SVR4.0.2 tar
+ . . . # # . . # . 3 = Solaris 2.1 tar
+ . . . . . . . . . 4 = GNU tar 1.11.1
+ . . . . . . . . . 5 = HP-UX 8.07 tar
+ . . . . . . . . . 6 = Ultrix 4.1
+ . . . . . . . . . 7 = AIX 3.2
+ . . . . . . . . . 8 = Hitachi HI-UX 1.03
+ . . . . . . . . . 9 = Omron UNIOS-B 4.3BSD 1.60Beta
+
+ . = works
+ # = ``impossible file type''
+
+ The following mask for old archive removes the `#'s in column 4
+ above, thus making GNU tar both a universal donor and a universal
+ acceptor for Paul's test. */
+
+ if (archive_format == V7_FORMAT || archive_format == USTAR_FORMAT)
+ MODE_TO_CHARS (st->stat.st_mode & MODE_ALL, header->header.mode);
+ else
+ MODE_TO_CHARS (st->stat.st_mode, header->header.mode);
+
+ {
+ uid_t uid = st->stat.st_uid;
+ if (archive_format == POSIX_FORMAT
+ && MAX_OCTAL_VAL (header->header.uid) < uid)
+ {
+ xheader_store ("uid", st, NULL);
+ uid = 0;
+ }
+ if (!UID_TO_CHARS (uid, header->header.uid))
+ return NULL;
+ }
+
+ {
+ gid_t gid = st->stat.st_gid;
+ if (archive_format == POSIX_FORMAT
+ && MAX_OCTAL_VAL (header->header.gid) < gid)
+ {
+ xheader_store ("gid", st, NULL);
+ gid = 0;
+ }
+ if (!GID_TO_CHARS (gid, header->header.gid))
+ return NULL;
+ }
+
+ {
+ off_t size = st->stat.st_size;
+ if (archive_format == POSIX_FORMAT
+ && MAX_OCTAL_VAL (header->header.size) < size)
+ {
+ xheader_store ("size", st, NULL);
+ size = 0;
+ }
+ if (!OFF_TO_CHARS (size, header->header.size))
+ return NULL;
+ }
+
+ {
+ struct timespec mtime = set_mtime_option ? mtime_option : st->mtime;
+ if (archive_format == POSIX_FORMAT)
+ {
+ if (MAX_OCTAL_VAL (header->header.mtime) < mtime.tv_sec
+ || mtime.tv_nsec != 0)
+ xheader_store ("mtime", st, &mtime);
+ if (MAX_OCTAL_VAL (header->header.mtime) < mtime.tv_sec)
+ mtime.tv_sec = 0;
+ }
+ if (!TIME_TO_CHARS (mtime.tv_sec, header->header.mtime))
+ return NULL;
+ }
+
+ /* FIXME */
+ if (S_ISCHR (st->stat.st_mode)
+ || S_ISBLK (st->stat.st_mode))
+ {
+ major_t devmajor = major (st->stat.st_rdev);
+ minor_t devminor = minor (st->stat.st_rdev);
+
+ if (archive_format == POSIX_FORMAT
+ && MAX_OCTAL_VAL (header->header.devmajor) < devmajor)
+ {
+ xheader_store ("devmajor", st, NULL);
+ devmajor = 0;
+ }
+ if (!MAJOR_TO_CHARS (devmajor, header->header.devmajor))
+ return NULL;
+
+ if (archive_format == POSIX_FORMAT
+ && MAX_OCTAL_VAL (header->header.devminor) < devminor)
+ {
+ xheader_store ("devminor", st, NULL);
+ devminor = 0;
+ }
+ if (!MINOR_TO_CHARS (devminor, header->header.devminor))
+ return NULL;
+ }
+ else if (archive_format != GNU_FORMAT && archive_format != OLDGNU_FORMAT)
+ {
+ if (!(MAJOR_TO_CHARS (0, header->header.devmajor)
+ && MINOR_TO_CHARS (0, header->header.devminor)))
+ return NULL;
+ }
+
+ if (archive_format == POSIX_FORMAT)
+ {
+ xheader_store ("atime", st, NULL);
+ xheader_store ("ctime", st, NULL);
+ }
+ else if (incremental_option)
+ if (archive_format == OLDGNU_FORMAT || archive_format == GNU_FORMAT)
+ {
+ TIME_TO_CHARS (st->atime.tv_sec, header->oldgnu_header.atime);
+ TIME_TO_CHARS (st->ctime.tv_sec, header->oldgnu_header.ctime);
+ }
+
+ header->header.typeflag = archive_format == V7_FORMAT ? AREGTYPE : REGTYPE;
+ switch (archive_format)
+ {
+ case V7_FORMAT:
+ break;
+
+ case OLDGNU_FORMAT:
+ case GNU_FORMAT: /*FIXME?*/
+ /* Overwrite header->header.magic and header.version in one blow. */
+ strcpy (header->header.magic, OLDGNU_MAGIC);
+ break;
+
+ case POSIX_FORMAT:
+ case USTAR_FORMAT:
+ strncpy (header->header.magic, TMAGIC, TMAGLEN);
+ strncpy (header->header.version, TVERSION, TVERSLEN);
+ break;
+
+ default:
+ abort ();
+ }
+
+ if (archive_format == V7_FORMAT || numeric_owner_option)
+ {
+ /* header->header.[ug]name are left as the empty string. */
+ }
+ else
+ {
+ uid_to_uname (st->stat.st_uid, &st->uname);
+ gid_to_gname (st->stat.st_gid, &st->gname);
+
+ if (archive_format == POSIX_FORMAT
+ && (strlen (st->uname) > UNAME_FIELD_SIZE
+ || !string_ascii_p (st->uname)))
+ xheader_store ("uname", st, NULL);
+ UNAME_TO_CHARS (st->uname, header->header.uname);
+
+ if (archive_format == POSIX_FORMAT
+ && (strlen (st->gname) > GNAME_FIELD_SIZE
+ || !string_ascii_p (st->gname)))
+ xheader_store ("gname", st, NULL);
+ GNAME_TO_CHARS (st->gname, header->header.gname);
+ }
+ return header;
+}
-/*
- * Takes a recordful of data and basically cruises through it to see if
- * it's made *entirely* of zeros, returning a 0 the instant it finds
- * something that is a non-zero, i.e., useful data.
- */
-int
-zero_record(buffer)
- char *buffer;
+void
+simple_finish_header (union block *header)
{
- register int i;
+ size_t i;
+ int sum;
+ char *p;
+
+ memcpy (header->header.chksum, CHKBLANKS, sizeof header->header.chksum);
+
+ sum = 0;
+ p = header->buffer;
+ for (i = sizeof *header; i-- != 0; )
+ /* We can't use unsigned char here because of old compilers, e.g. V7. */
+ sum += 0xFF & *p++;
+
+ /* Fill in the checksum field. It's formatted differently from the
+ other fields: it has [6] digits, a null, then a space -- rather than
+ digits, then a null. We use to_chars.
+ The final space is already there, from
+ checksumming, and to_chars doesn't modify it.
- for (i = 0; i < RECORDSIZE; i++)
- if (buffer[i] != '\000')
- return 0;
- return 1;
+ This is a fast way to do:
+
+ sprintf(header->header.chksum, "%6o", sum); */
+
+ uintmax_to_chars ((uintmax_t) sum, header->header.chksum, 7);
+
+ set_next_block_after (header);
}
+/* Finish off a filled-in header block and write it out. We also
+ print the file name and/or full info if verbose is on. If BLOCK_ORDINAL
+ is not negative, is the block ordinal of the first record for this
+ file, which may be a preceding long name or long link record. */
void
-find_new_file_size(filesize, highest_index)
- int *filesize;
- int highest_index;
+finish_header (struct tar_stat_info *st,
+ union block *header, off_t block_ordinal)
{
- register int i;
+ /* Note: It is important to do this before the call to write_extended(),
+ so that the actual ustar header is printed */
+ if (verbose_option
+ && header->header.typeflag != GNUTYPE_LONGLINK
+ && header->header.typeflag != GNUTYPE_LONGNAME
+ && header->header.typeflag != XHDTYPE
+ && header->header.typeflag != XGLTYPE)
+ {
+ /* FIXME: This global is used in print_header, sigh. */
+ current_format = archive_format;
+ print_header (st, header, block_ordinal);
+ }
- *filesize = 0;
- for (i = 0; sparsearray[i].numbytes && i <= highest_index; i++)
- *filesize += sparsearray[i].numbytes;
+ header = write_extended (false, st, header);
+ simple_finish_header (header);
}
-
-/*
- * Make a header block for the file name whose stat info is st .
- * Return header pointer for success, NULL if the name is too long.
- */
-union record *
-start_header(name, st)
- char *name;
- register struct stat *st;
+\f
+
+void
+pad_archive (off_t size_left)
{
- register union record *header;
-
- if (strlen (name) >= NAMSIZ)
- write_long (name, LF_LONGNAME);
-
- header = (union record *) findrec();
- bzero(header->charptr, sizeof(*header)); /* XXX speed up */
-
- /*
- * Check the file name and put it in the record.
- */
- if(!f_absolute_paths) {
- static int warned_once = 0;
-#ifdef __MSDOS__
- if(name[1]==':') {
- name+=2;
- if(!warned_once++)
- msg("Removing drive spec from names in the archive");
- }
-#endif
- while ('/' == *name) {
- name++; /* Force relative path */
- if (!warned_once++)
- msg("Removing leading / from absolute path names in the archive.");
- }
+ union block *blk;
+ while (size_left > 0)
+ {
+ blk = find_next_block ();
+ memset (blk->buffer, 0, BLOCKSIZE);
+ set_next_block_after (blk);
+ size_left -= BLOCKSIZE;
+ }
+}
+
+static enum dump_status
+dump_regular_file (int fd, struct tar_stat_info *st)
+{
+ off_t size_left = st->stat.st_size;
+ off_t block_ordinal;
+ union block *blk;
+
+ block_ordinal = current_block_ordinal ();
+ blk = start_header (st);
+ if (!blk)
+ return dump_status_fail;
+
+ /* Mark contiguous files, if we support them. */
+ if (archive_format != V7_FORMAT && S_ISCTG (st->stat.st_mode))
+ blk->header.typeflag = CONTTYPE;
+
+ finish_header (st, blk, block_ordinal);
+
+ mv_begin_write (st->file_name, st->stat.st_size, st->stat.st_size);
+ while (size_left > 0)
+ {
+ size_t bufsize, count;
+
+ blk = find_next_block ();
+
+ bufsize = available_space_after (blk);
+
+ if (size_left < bufsize)
+ {
+ /* Last read -- zero out area beyond. */
+ bufsize = size_left;
+ count = bufsize % BLOCKSIZE;
+ if (count)
+ memset (blk->buffer + size_left, 0, BLOCKSIZE - count);
}
- strncpy(header->header.arch_name, name, NAMSIZ);
- header->header.arch_name[NAMSIZ-1] = '\0';
-
- to_oct((long) (st->st_mode & 07777),
- 8, header->header.mode);
- to_oct((long) st->st_uid, 8, header->header.uid);
- to_oct((long) st->st_gid, 8, header->header.gid);
- to_oct((long) st->st_size, 1+12, header->header.size);
- to_oct((long) st->st_mtime, 1+12, header->header.mtime);
- /* header->header.linkflag is left as null */
- if(f_gnudump) {
- to_oct((long) st->st_atime, 1+12, header->header.atime);
- to_oct((long) st->st_ctime, 1+12, header->header.ctime);
+
+ count = (fd < 0) ? bufsize : safe_read (fd, blk->buffer, bufsize);
+ if (count == SAFE_READ_ERROR)
+ {
+ read_diag_details (st->orig_file_name,
+ st->stat.st_size - size_left, bufsize);
+ pad_archive (size_left);
+ return dump_status_short;
}
+ size_left -= count;
+ set_next_block_after (blk + (bufsize - 1) / BLOCKSIZE);
-#ifndef NONAMES
- /* Fill in new Unix Standard fields if desired. */
- if (f_standard) {
- header->header.linkflag = LF_NORMAL; /* New default */
- strcpy(header->header.magic, TMAGIC); /* Mark as Unix Std */
- finduname(header->header.uname, st->st_uid);
- findgname(header->header.gname, st->st_gid);
+ if (count != bufsize)
+ {
+ char buf[UINTMAX_STRSIZE_BOUND];
+ memset (blk->buffer + count, 0, bufsize - count);
+ WARNOPT (WARN_FILE_SHRANK,
+ (0, 0,
+ ngettext ("%s: File shrank by %s byte; padding with zeros",
+ "%s: File shrank by %s bytes; padding with zeros",
+ size_left),
+ quotearg_colon (st->orig_file_name),
+ STRINGIFY_BIGINT (size_left, buf)));
+ if (! ignore_failed_read_option)
+ set_exit_status (TAREXIT_DIFFERS);
+ pad_archive (size_left - (bufsize - count));
+ return dump_status_short;
}
-#endif
- return header;
+ }
+ return dump_status_ok;
}
-/*
- * Finish off a filled-in header block and write it out.
- * We also print the file name and/or full info if verbose is on.
- */
-void
-finish_header(header)
- register union record *header;
+\f
+static void
+dump_dir0 (char *directory,
+ struct tar_stat_info *st, bool top_level, dev_t parent_device)
{
- register int i, sum;
- register char *p;
-
- bcopy(CHKBLANKS, header->header.chksum, sizeof(header->header.chksum));
-
- sum = 0;
- p = header->charptr;
- for (i = sizeof(*header); --i >= 0; ) {
- /*
- * We can't use unsigned char here because of old compilers,
- * e.g. V7.
- */
- sum += 0xFF & *p++;
+ dev_t our_device = st->stat.st_dev;
+ const char *tag_file_name;
+ union block *blk = NULL;
+ off_t block_ordinal = current_block_ordinal ();
+
+ st->stat.st_size = 0; /* force 0 size on dir */
+
+ blk = start_header (st);
+ if (!blk)
+ return;
+
+ if (incremental_option && archive_format != POSIX_FORMAT)
+ blk->header.typeflag = GNUTYPE_DUMPDIR;
+ else /* if (standard_option) */
+ blk->header.typeflag = DIRTYPE;
+
+ /* If we're gnudumping, we aren't done yet so don't close it. */
+
+ if (!incremental_option)
+ finish_header (st, blk, block_ordinal);
+ else if (gnu_list_name->directory)
+ {
+ if (archive_format == POSIX_FORMAT)
+ {
+ xheader_store ("GNU.dumpdir", st,
+ safe_directory_contents (gnu_list_name->directory));
+ finish_header (st, blk, block_ordinal);
}
+ else
+ {
+ off_t size_left;
+ off_t totsize;
+ size_t bufsize;
+ ssize_t count;
+ const char *buffer, *p_buffer;
+
+ block_ordinal = current_block_ordinal ();
+ buffer = safe_directory_contents (gnu_list_name->directory);
+ totsize = dumpdir_size (buffer);
+ OFF_TO_CHARS (totsize, blk->header.size);
+ finish_header (st, blk, block_ordinal);
+ p_buffer = buffer;
+ size_left = totsize;
+
+ mv_begin_write (st->file_name, totsize, totsize);
+ while (size_left > 0)
+ {
+ blk = find_next_block ();
+ bufsize = available_space_after (blk);
+ if (size_left < bufsize)
+ {
+ bufsize = size_left;
+ count = bufsize % BLOCKSIZE;
+ if (count)
+ memset (blk->buffer + size_left, 0, BLOCKSIZE - count);
+ }
+ memcpy (blk->buffer, p_buffer, bufsize);
+ size_left -= bufsize;
+ p_buffer += bufsize;
+ set_next_block_after (blk + (bufsize - 1) / BLOCKSIZE);
+ }
+ }
+ return;
+ }
- /*
- * Fill in the checksum field. It's formatted differently
- * from the other fields: it has [6] digits, a null, then a
- * space -- rather than digits, a space, then a null.
- * We use to_oct then write the null in over to_oct's space.
- * The final space is already there, from checksumming, and
- * to_oct doesn't modify it.
- *
- * This is a fast way to do:
- * (void) sprintf(header->header.chksum, "%6o", sum);
- */
- to_oct((long) sum, 8, header->header.chksum);
- header->header.chksum[6] = '\0'; /* Zap the space */
-
- userec(header);
-
- if (f_verbose) {
- extern union record *head; /* Points to current tape header */
- extern int head_standard; /* Tape header is in ANSI format */
-
- /* These globals are parameters to print_header, sigh */
- head = header;
- /* hstat is already set up */
- head_standard = f_standard;
- print_header();
+ if (!recursion_option)
+ return;
+
+ if (one_file_system_option
+ && !top_level
+ && parent_device != st->stat.st_dev)
+ {
+ if (verbose_option)
+ WARNOPT (WARN_XDEV,
+ (0, 0,
+ _("%s: file is on a different filesystem; not dumped"),
+ quotearg_colon (st->orig_file_name)));
+ }
+ else
+ {
+ char *name_buf;
+ size_t name_size;
+
+ switch (check_exclusion_tags (st->orig_file_name, &tag_file_name))
+ {
+ case exclusion_tag_all:
+ /* Handled in dump_file0 */
+ break;
+
+ case exclusion_tag_none:
+ {
+ char const *entry;
+ size_t entry_len;
+ size_t name_len;
+
+ name_buf = xstrdup (st->orig_file_name);
+ name_size = name_len = strlen (name_buf);
+
+ /* Now output all the files in the directory. */
+ /* FIXME: Should speed this up by cd-ing into the dir. */
+ for (entry = directory; (entry_len = strlen (entry)) != 0;
+ entry += entry_len + 1)
+ {
+ if (name_size < name_len + entry_len)
+ {
+ name_size = name_len + entry_len;
+ name_buf = xrealloc (name_buf, name_size + 1);
+ }
+ strcpy (name_buf + name_len, entry);
+ if (!excluded_name (name_buf))
+ dump_file (name_buf, false, our_device);
+ }
+
+ free (name_buf);
+ }
+ break;
+
+ case exclusion_tag_contents:
+ exclusion_tag_warning (st->orig_file_name, tag_file_name,
+ _("contents not dumped"));
+ name_size = strlen (st->orig_file_name) + strlen (tag_file_name) + 1;
+ name_buf = xmalloc (name_size);
+ strcpy (name_buf, st->orig_file_name);
+ strcat (name_buf, tag_file_name);
+ dump_file (name_buf, false, our_device);
+ free (name_buf);
+ break;
+
+ case exclusion_tag_under:
+ exclusion_tag_warning (st->orig_file_name, tag_file_name,
+ _("contents not dumped"));
+ break;
}
+ }
+}
- return;
+/* Ensure exactly one trailing slash. */
+static void
+ensure_slash (char **pstr)
+{
+ size_t len = strlen (*pstr);
+ while (len >= 1 && ISSLASH ((*pstr)[len - 1]))
+ len--;
+ if (!ISSLASH ((*pstr)[len]))
+ *pstr = xrealloc (*pstr, len + 2);
+ (*pstr)[len++] = '/';
+ (*pstr)[len] = '\0';
+}
+
+static bool
+dump_dir (int fd, struct tar_stat_info *st, bool top_level,
+ dev_t parent_device)
+{
+ char *directory = fdsavedir (fd);
+ if (!directory)
+ {
+ savedir_diag (st->orig_file_name);
+ return false;
+ }
+
+ dump_dir0 (directory, st, top_level, parent_device);
+
+ free (directory);
+ return true;
}
+\f
+/* Number of links a file can have without having to be entered into
+ the link table. Typically this is 1, but in trickier circumstances
+ it is 0. */
+static nlink_t trivial_link_count;
+
+\f
+/* Main functions of this module. */
-/*
- * Quick and dirty octal conversion.
- * Converts long "value" into a "digs"-digit field at "where",
- * including a trailing space and room for a null. "digs"==3 means
- * 1 digit, a space, and room for a null.
- *
- * We assume the trailing null is already there and don't fill it in.
- * This fact is used by start_header and finish_header, so don't change it!
- *
- * This should be equivalent to:
- * (void) sprintf(where, "%*lo ", digs-2, value);
- * except that sprintf fills in the trailing null and we don't.
- */
void
-to_oct(value, digs, where)
- register long value;
- register int digs;
- register char *where;
+create_archive (void)
+{
+ struct name const *p;
+
+ trivial_link_count = name_count <= 1 && ! dereference_option;
+
+ open_archive (ACCESS_WRITE);
+ buffer_write_global_xheader ();
+
+ if (incremental_option)
+ {
+ size_t buffer_size = 1000;
+ char *buffer = xmalloc (buffer_size);
+ const char *q;
+
+ collect_and_sort_names ();
+
+ while ((p = name_from_list ()) != NULL)
+ if (!excluded_name (p->name))
+ dump_file (p->name, p->cmdline, (dev_t) 0);
+
+ blank_name_list ();
+ while ((p = name_from_list ()) != NULL)
+ if (!excluded_name (p->name))
+ {
+ size_t plen = strlen (p->name);
+ if (buffer_size <= plen)
+ {
+ while ((buffer_size *= 2) <= plen)
+ continue;
+ buffer = xrealloc (buffer, buffer_size);
+ }
+ memcpy (buffer, p->name, plen);
+ if (! ISSLASH (buffer[plen - 1]))
+ buffer[plen++] = DIRECTORY_SEPARATOR;
+ q = directory_contents (gnu_list_name->directory);
+ if (q)
+ while (*q)
+ {
+ size_t qlen = strlen (q);
+ if (*q == 'Y')
+ {
+ if (buffer_size < plen + qlen)
+ {
+ while ((buffer_size *=2 ) < plen + qlen)
+ continue;
+ buffer = xrealloc (buffer, buffer_size);
+ }
+ strcpy (buffer + plen, q + 1);
+ dump_file (buffer, false, (dev_t) 0);
+ }
+ q += qlen + 1;
+ }
+ }
+ free (buffer);
+ }
+ else
+ {
+ const char *name;
+ while ((name = name_next (1)) != NULL)
+ if (!excluded_name (name))
+ dump_file (name, true, (dev_t) 0);
+ }
+
+ write_eot ();
+ close_archive ();
+ finish_deferred_unlinks ();
+ if (listed_incremental_option)
+ write_directory_file ();
+}
+
+
+/* Calculate the hash of a link. */
+static size_t
+hash_link (void const *entry, size_t n_buckets)
+{
+ struct link const *l = entry;
+ uintmax_t num = l->dev ^ l->ino;
+ return num % n_buckets;
+}
+
+/* Compare two links for equality. */
+static bool
+compare_links (void const *entry1, void const *entry2)
+{
+ struct link const *link1 = entry1;
+ struct link const *link2 = entry2;
+ return ((link1->dev ^ link2->dev) | (link1->ino ^ link2->ino)) == 0;
+}
+
+static void
+unknown_file_error (char const *p)
+{
+ WARNOPT (WARN_FILE_IGNORED,
+ (0, 0, _("%s: Unknown file type; file ignored"),
+ quotearg_colon (p)));
+ if (!ignore_failed_read_option)
+ set_exit_status (TAREXIT_FAILURE);
+}
+
+\f
+/* Handling of hard links */
+
+/* Table of all non-directories that we've written so far. Any time
+ we see another, we check the table and avoid dumping the data
+ again if we've done it once already. */
+static Hash_table *link_table;
+
+/* Try to dump stat as a hard link to another file in the archive.
+ Return true if successful. */
+static bool
+dump_hard_link (struct tar_stat_info *st)
{
-
- --digs; /* Trailing null slot is left alone */
- where[--digs] = ' '; /* Put in the space, though */
+ if (link_table
+ && (trivial_link_count < st->stat.st_nlink || remove_files_option))
+ {
+ struct link lp;
+ struct link *duplicate;
+ off_t block_ordinal;
+ union block *blk;
+
+ lp.ino = st->stat.st_ino;
+ lp.dev = st->stat.st_dev;
+
+ if ((duplicate = hash_lookup (link_table, &lp)))
+ {
+ /* We found a link. */
+ char const *link_name = safer_name_suffix (duplicate->name, true,
+ absolute_names_option);
+
+ duplicate->nlink--;
+
+ block_ordinal = current_block_ordinal ();
+ assign_string (&st->link_name, link_name);
+ if (NAME_FIELD_SIZE - (archive_format == OLDGNU_FORMAT)
+ < strlen (link_name))
+ write_long_link (st);
- /* Produce the digits -- at least one */
- do {
- where[--digs] = '0' + (char)(value & 7); /* one octal digit */
- value >>= 3;
- } while (digs > 0 && value != 0);
+ st->stat.st_size = 0;
+ blk = start_header (st);
+ if (!blk)
+ return false;
+ tar_copy_str (blk->header.linkname, link_name, NAME_FIELD_SIZE);
- /* Leading spaces, if necessary */
- while (digs > 0)
- where[--digs] = ' ';
+ blk->header.typeflag = LNKTYPE;
+ finish_header (st, blk, block_ordinal);
+ if (remove_files_option)
+ queue_deferred_unlink (st->orig_file_name, false);
+
+ return true;
+ }
+ }
+ return false;
}
+static void
+file_count_links (struct tar_stat_info *st)
+{
+ if (hard_dereference_option)
+ return;
+ if (trivial_link_count < st->stat.st_nlink)
+ {
+ struct link *duplicate;
+ char *linkname = NULL;
+ struct link *lp;
+
+ assign_string (&linkname, st->orig_file_name);
+ transform_name (&linkname, XFORM_LINK);
+
+ lp = xmalloc (offsetof (struct link, name)
+ + strlen (linkname) + 1);
+ lp->ino = st->stat.st_ino;
+ lp->dev = st->stat.st_dev;
+ lp->nlink = st->stat.st_nlink;
+ strcpy (lp->name, linkname);
+ free (linkname);
+
+ if (! ((link_table
+ || (link_table = hash_initialize (0, 0, hash_link,
+ compare_links, 0)))
+ && (duplicate = hash_insert (link_table, lp))))
+ xalloc_die ();
+
+ if (duplicate != lp)
+ abort ();
+ lp->nlink--;
+ }
+}
-/*
- * Write the EOT record(s).
- * We actually zero at least one record, through the end of the block.
- * Old tar writes garbage after two zeroed records -- and PDtar used to.
- */
+/* For each dumped file, check if all its links were dumped. Emit
+ warnings if it is not so. */
void
-write_eot()
+check_links (void)
{
- union record *p;
- int bufsize;
+ struct link *lp;
- p = findrec();
- if (p)
- {
- bufsize = endofrecs()->charptr - p->charptr;
- bzero(p->charptr, bufsize);
- userec(p);
- }
+ if (!link_table)
+ return;
+
+ for (lp = hash_get_first (link_table); lp;
+ lp = hash_get_next (link_table, lp))
+ {
+ if (lp->nlink)
+ {
+ WARN ((0, 0, _("Missing links to %s."), quote (lp->name)));
+ }
+ }
}
-/* Write a LF_LONGLINK or LF_LONGNAME record. */
-write_long (p, type)
- char *p;
- char type;
+/* Dump a single file, recursing on directories. P is the file name
+ to dump. TOP_LEVEL tells whether this is a top-level call; zero
+ means no, positive means yes, and negative means the top level
+ of an incremental dump. PARENT_DEVICE is the device of P's
+ parent directory; it is examined only if TOP_LEVEL is zero. */
+
+/* FIXME: One should make sure that for *every* path leading to setting
+ exit_status to failure, a clear diagnostic has been issued. */
+
+static void
+dump_file0 (struct tar_stat_info *st, const char *p,
+ bool top_level, dev_t parent_device)
{
- int size = strlen (p) + 1;
- int bufsize;
- union record *header;
-
- /* Link name won't fit, so we write
- an LF_LONGLINK record. */
- hstat.st_size = size;
- header = start_header ("././@LongLink", &hstat);
- header->header.linkflag = type;
- finish_header (header);
-
- header = findrec ();
-
- bufsize = endofrecs ()->charptr - header->charptr;
-
- while (bufsize < size)
+ union block *header;
+ char type;
+ off_t original_size;
+ struct timespec original_ctime;
+ struct timespec restore_times[2];
+ off_t block_ordinal = -1;
+ bool is_dir;
+
+ if (interactive_option && !confirm ("add", p))
+ return;
+
+ assign_string (&st->orig_file_name, p);
+ assign_string (&st->file_name,
+ safer_name_suffix (p, false, absolute_names_option));
+
+ transform_name (&st->file_name, XFORM_REGFILE);
+
+ if (deref_stat (dereference_option, p, &st->stat) != 0)
{
- bcopy (p, header->charptr, bufsize);
- p += bufsize;
- size -= bufsize;
- userec (header + (bufsize - 1)/RECORDSIZE);
- header = findrec ();
- bufsize = endofrecs ()->charptr - header->charptr;
+ file_removed_diag (p, top_level, stat_diag);
+ return;
}
- bcopy (p, header->charptr, size);
- bzero (header->charptr + size, bufsize - size);
- userec (header + (size - 1)/RECORDSIZE);
+ st->archive_file_size = original_size = st->stat.st_size;
+ st->atime = restore_times[0] = get_stat_atime (&st->stat);
+ st->mtime = restore_times[1] = get_stat_mtime (&st->stat);
+ st->ctime = original_ctime = get_stat_ctime (&st->stat);
+
+#ifdef S_ISHIDDEN
+ if (S_ISHIDDEN (st->stat.st_mode))
+ {
+ char *new = (char *) alloca (strlen (p) + 2);
+ if (new)
+ {
+ strcpy (new, p);
+ strcat (new, "@");
+ p = new;
+ }
+ }
+#endif
+
+ /* See if we want only new files, and check if this one is too old to
+ put in the archive.
+
+ This check is omitted if incremental_option is set *and* the
+ requested file is not explicitely listed in the command line. */
+
+ if (!(incremental_option && !is_individual_file (p))
+ && !S_ISDIR (st->stat.st_mode)
+ && OLDER_TAR_STAT_TIME (*st, m)
+ && (!after_date_option || OLDER_TAR_STAT_TIME (*st, c)))
+ {
+ if (!incremental_option && verbose_option)
+ WARNOPT (WARN_FILE_UNCHANGED,
+ (0, 0, _("%s: file is unchanged; not dumped"),
+ quotearg_colon (p)));
+ return;
+ }
+
+ /* See if we are trying to dump the archive. */
+ if (sys_file_is_archive (st))
+ {
+ WARNOPT (WARN_IGNORE_ARCHIVE,
+ (0, 0, _("%s: file is the archive; not dumped"),
+ quotearg_colon (p)));
+ return;
+ }
+
+ is_dir = S_ISDIR (st->stat.st_mode) != 0;
+
+ if (!is_dir && dump_hard_link (st))
+ return;
+
+ if (is_dir || S_ISREG (st->stat.st_mode) || S_ISCTG (st->stat.st_mode))
+ {
+ bool ok;
+ int fd = -1;
+ struct stat final_stat;
+
+ if (is_dir || file_dumpable_p (st))
+ {
+ fd = open (p,
+ (O_RDONLY | O_BINARY
+ | (is_dir ? O_DIRECTORY | O_NONBLOCK : 0)
+ | (atime_preserve_option == system_atime_preserve
+ ? O_NOATIME
+ : 0)));
+ if (fd < 0)
+ {
+ file_removed_diag (p, top_level, open_diag);
+ return;
+ }
+ }
+
+ if (is_dir)
+ {
+ const char *tag_file_name;
+ ensure_slash (&st->orig_file_name);
+ ensure_slash (&st->file_name);
+
+ if (check_exclusion_tags (st->orig_file_name, &tag_file_name)
+ == exclusion_tag_all)
+ {
+ exclusion_tag_warning (st->orig_file_name, tag_file_name,
+ _("directory not dumped"));
+ if (fd >= 0)
+ close (fd);
+ return;
+ }
+
+ ok = dump_dir (fd, st, top_level, parent_device);
+
+ /* dump_dir consumes FD if successful. */
+ if (ok)
+ fd = -1;
+ }
+ else
+ {
+ enum dump_status status;
+
+ if (fd != -1 && sparse_option && ST_IS_SPARSE (st->stat))
+ {
+ status = sparse_dump_file (fd, st);
+ if (status == dump_status_not_implemented)
+ status = dump_regular_file (fd, st);
+ }
+ else
+ status = dump_regular_file (fd, st);
+
+ switch (status)
+ {
+ case dump_status_ok:
+ case dump_status_short:
+ file_count_links (st);
+ break;
+
+ case dump_status_fail:
+ break;
+
+ case dump_status_not_implemented:
+ abort ();
+ }
+
+ ok = status == dump_status_ok;
+ }
+
+ if (ok)
+ {
+ /* If possible, reopen a directory if we are preserving
+ atimes, so that we can set just the atime on systems with
+ _FIOSATIME. */
+ if (fd < 0 && is_dir
+ && atime_preserve_option == replace_atime_preserve)
+ fd = open (p, O_RDONLY | O_BINARY | O_DIRECTORY | O_NONBLOCK);
+
+ if ((fd < 0
+ ? deref_stat (dereference_option, p, &final_stat)
+ : fstat (fd, &final_stat))
+ != 0)
+ {
+ file_removed_diag (p, top_level, stat_diag);
+ ok = false;
+ }
+ }
+
+ if (ok)
+ {
+ if ((timespec_cmp (get_stat_ctime (&final_stat), original_ctime) != 0
+ /* Original ctime will change if the file is a directory and
+ --remove-files is given */
+ && !(remove_files_option && is_dir))
+ || original_size < final_stat.st_size)
+ {
+ WARNOPT (WARN_FILE_CHANGED,
+ (0, 0, _("%s: file changed as we read it"),
+ quotearg_colon (p)));
+ set_exit_status (TAREXIT_DIFFERS);
+ }
+ else if (atime_preserve_option == replace_atime_preserve
+ && set_file_atime (fd, p, restore_times) != 0)
+ utime_error (p);
+ }
+
+ if (0 <= fd && close (fd) != 0)
+ {
+ close_diag (p);
+ ok = false;
+ }
+
+ if (ok && remove_files_option)
+ queue_deferred_unlink (p, is_dir);
+
+ return;
+ }
+#ifdef HAVE_READLINK
+ else if (S_ISLNK (st->stat.st_mode))
+ {
+ char *buffer;
+ int size;
+ size_t linklen = st->stat.st_size;
+ if (linklen != st->stat.st_size || linklen + 1 == 0)
+ xalloc_die ();
+ buffer = (char *) alloca (linklen + 1);
+ size = readlink (p, buffer, linklen + 1);
+ if (size < 0)
+ {
+ file_removed_diag (p, top_level, readlink_diag);
+ return;
+ }
+ buffer[size] = '\0';
+ assign_string (&st->link_name, buffer);
+ transform_name (&st->link_name, XFORM_SYMLINK);
+ if (NAME_FIELD_SIZE - (archive_format == OLDGNU_FORMAT) < size)
+ write_long_link (st);
+
+ block_ordinal = current_block_ordinal ();
+ st->stat.st_size = 0; /* force 0 size on symlink */
+ header = start_header (st);
+ if (!header)
+ return;
+ tar_copy_str (header->header.linkname, st->link_name, NAME_FIELD_SIZE);
+ header->header.typeflag = SYMTYPE;
+ finish_header (st, header, block_ordinal);
+ /* nothing more to do to it */
+
+ if (remove_files_option)
+ queue_deferred_unlink (p, false);
+
+ file_count_links (st);
+ return;
+ }
+#endif
+ else if (S_ISCHR (st->stat.st_mode))
+ type = CHRTYPE;
+ else if (S_ISBLK (st->stat.st_mode))
+ type = BLKTYPE;
+ else if (S_ISFIFO (st->stat.st_mode))
+ type = FIFOTYPE;
+ else if (S_ISSOCK (st->stat.st_mode))
+ {
+ WARNOPT (WARN_FILE_IGNORED,
+ (0, 0, _("%s: socket ignored"), quotearg_colon (p)));
+ return;
+ }
+ else if (S_ISDOOR (st->stat.st_mode))
+ {
+ WARNOPT (WARN_FILE_IGNORED,
+ (0, 0, _("%s: door ignored"), quotearg_colon (p)));
+ return;
+ }
+ else
+ {
+ unknown_file_error (p);
+ return;
+ }
+
+ if (archive_format == V7_FORMAT)
+ {
+ unknown_file_error (p);
+ return;
+ }
+
+ block_ordinal = current_block_ordinal ();
+ st->stat.st_size = 0; /* force 0 size */
+ header = start_header (st);
+ if (!header)
+ return;
+ header->header.typeflag = type;
+
+ if (type != FIFOTYPE)
+ {
+ MAJOR_TO_CHARS (major (st->stat.st_rdev),
+ header->header.devmajor);
+ MINOR_TO_CHARS (minor (st->stat.st_rdev),
+ header->header.devminor);
+ }
+
+ finish_header (st, header, block_ordinal);
+ if (remove_files_option)
+ queue_deferred_unlink (p, false);
+}
+
+void
+dump_file (const char *p, bool top_level, dev_t parent_device)
+{
+ struct tar_stat_info st;
+ tar_stat_init (&st);
+ dump_file0 (&st, p, top_level, parent_device);
+ if (listed_incremental_option)
+ update_parent_directory (p);
+ tar_stat_destroy (&st);
}