Index: stable/10/contrib/libarchive/libarchive/archive_read_disk_posix.c =================================================================== --- stable/10/contrib/libarchive/libarchive/archive_read_disk_posix.c (revision 302424) +++ stable/10/contrib/libarchive/libarchive/archive_read_disk_posix.c (revision 302425) @@ -1,2660 +1,2664 @@ /*- * Copyright (c) 2003-2009 Tim Kientzle * Copyright (c) 2010-2012 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ /* This is the tree-walking code for POSIX systems. */ #if !defined(_WIN32) || defined(__CYGWIN__) #include "archive_platform.h" __FBSDID("$FreeBSD$"); #ifdef HAVE_SYS_PARAM_H #include #endif #ifdef HAVE_SYS_MOUNT_H #include #endif #ifdef HAVE_SYS_STAT_H #include #endif #ifdef HAVE_SYS_STATFS_H #include #endif #ifdef HAVE_SYS_STATVFS_H #include #endif #ifdef HAVE_SYS_TIME_H #include #endif #ifdef HAVE_LINUX_MAGIC_H #include #endif #ifdef HAVE_LINUX_FS_H #include #endif /* * Some Linux distributions have both linux/ext2_fs.h and ext2fs/ext2_fs.h. * As the include guards don't agree, the order of include is important. */ #ifdef HAVE_LINUX_EXT2_FS_H #include /* for Linux file flags */ #endif #if defined(HAVE_EXT2FS_EXT2_FS_H) && !defined(__CYGWIN__) #include /* Linux file flags, broken on Cygwin */ #endif #ifdef HAVE_DIRECT_H #include #endif #ifdef HAVE_DIRENT_H #include #endif #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_FCNTL_H #include #endif #ifdef HAVE_LIMITS_H #include #endif #ifdef HAVE_STDLIB_H #include #endif #ifdef HAVE_STRING_H #include #endif #ifdef HAVE_UNISTD_H #include #endif #ifdef HAVE_SYS_IOCTL_H #include #endif #include "archive.h" #include "archive_string.h" #include "archive_entry.h" #include "archive_private.h" #include "archive_read_disk_private.h" #ifndef HAVE_FCHDIR #error fchdir function required. #endif #ifndef O_BINARY #define O_BINARY 0 #endif #ifndef O_CLOEXEC #define O_CLOEXEC 0 #endif /*- * This is a new directory-walking system that addresses a number * of problems I've had with fts(3). In particular, it has no * pathname-length limits (other than the size of 'int'), handles * deep logical traversals, uses considerably less memory, and has * an opaque interface (easier to modify in the future). * * Internally, it keeps a single list of "tree_entry" items that * represent filesystem objects that require further attention. * Non-directories are not kept in memory: they are pulled from * readdir(), returned to the client, then freed as soon as possible. * Any directory entry to be traversed gets pushed onto the stack. * * There is surprisingly little information that needs to be kept for * each item on the stack. Just the name, depth (represented here as the * string length of the parent directory's pathname), and some markers * indicating how to get back to the parent (via chdir("..") for a * regular dir or via fchdir(2) for a symlink). */ /* * TODO: * 1) Loop checking. * 3) Arbitrary logical traversals by closing/reopening intermediate fds. */ struct restore_time { const char *name; time_t mtime; long mtime_nsec; time_t atime; long atime_nsec; mode_t filetype; int noatime; }; struct tree_entry { int depth; struct tree_entry *next; struct tree_entry *parent; struct archive_string name; size_t dirname_length; int64_t dev; int64_t ino; int flags; int filesystem_id; /* How to return back to the parent of a symlink. */ int symlink_parent_fd; /* How to restore time of a directory. */ struct restore_time restore_time; }; struct filesystem { int64_t dev; int synthetic; int remote; int noatime; #if defined(HAVE_READDIR_R) size_t name_max; #endif long incr_xfer_size; long max_xfer_size; long min_xfer_size; long xfer_align; /* * Buffer used for reading file contents. */ /* Exactly allocated memory pointer. */ unsigned char *allocation_ptr; /* Pointer adjusted to the filesystem alignment . */ unsigned char *buff; size_t buff_size; }; /* Definitions for tree_entry.flags bitmap. */ #define isDir 1 /* This entry is a regular directory. */ #define isDirLink 2 /* This entry is a symbolic link to a directory. */ #define needsFirstVisit 4 /* This is an initial entry. */ #define needsDescent 8 /* This entry needs to be previsited. */ #define needsOpen 16 /* This is a directory that needs to be opened. */ #define needsAscent 32 /* This entry needs to be postvisited. */ /* * Local data for this package. */ struct tree { struct tree_entry *stack; struct tree_entry *current; DIR *d; #define INVALID_DIR_HANDLE NULL struct dirent *de; #if defined(HAVE_READDIR_R) struct dirent *dirent; size_t dirent_allocated; #endif int flags; int visit_type; /* Error code from last failed operation. */ int tree_errno; /* Dynamically-sized buffer for holding path */ struct archive_string path; /* Last path element */ const char *basename; /* Leading dir length */ size_t dirname_length; int depth; int openCount; int maxOpenCount; int initial_dir_fd; int working_dir_fd; struct stat lst; struct stat st; int descend; int nlink; /* How to restore time of a file. */ struct restore_time restore_time; struct entry_sparse { int64_t length; int64_t offset; } *sparse_list, *current_sparse; int sparse_count; int sparse_list_size; char initial_symlink_mode; char symlink_mode; struct filesystem *current_filesystem; struct filesystem *filesystem_table; int initial_filesystem_id; int current_filesystem_id; int max_filesystem_id; int allocated_filesytem; int entry_fd; int entry_eof; int64_t entry_remaining_bytes; int64_t entry_total; unsigned char *entry_buff; size_t entry_buff_size; }; /* Definitions for tree.flags bitmap. */ #define hasStat 16 /* The st entry is valid. */ #define hasLstat 32 /* The lst entry is valid. */ #define onWorkingDir 64 /* We are on the working dir where we are * reading directory entry at this time. */ #define needsRestoreTimes 128 #define onInitialDir 256 /* We are on the initial dir. */ static int tree_dir_next_posix(struct tree *t); #ifdef HAVE_DIRENT_D_NAMLEN /* BSD extension; avoids need for a strlen() call. */ #define D_NAMELEN(dp) (dp)->d_namlen #else #define D_NAMELEN(dp) (strlen((dp)->d_name)) #endif /* Initiate/terminate a tree traversal. */ static struct tree *tree_open(const char *, int, int); static struct tree *tree_reopen(struct tree *, const char *, int); static void tree_close(struct tree *); static void tree_free(struct tree *); static void tree_push(struct tree *, const char *, int, int64_t, int64_t, struct restore_time *); static int tree_enter_initial_dir(struct tree *); static int tree_enter_working_dir(struct tree *); static int tree_current_dir_fd(struct tree *); /* * tree_next() returns Zero if there is no next entry, non-zero if * there is. Note that directories are visited three times. * Directories are always visited first as part of enumerating their * parent; that is a "regular" visit. If tree_descend() is invoked at * that time, the directory is added to a work list and will * subsequently be visited two more times: once just after descending * into the directory ("postdescent") and again just after ascending * back to the parent ("postascent"). * * TREE_ERROR_DIR is returned if the descent failed (because the * directory couldn't be opened, for instance). This is returned * instead of TREE_POSTDESCENT/TREE_POSTASCENT. TREE_ERROR_DIR is not a * fatal error, but it does imply that the relevant subtree won't be * visited. TREE_ERROR_FATAL is returned for an error that left the * traversal completely hosed. Right now, this is only returned for * chdir() failures during ascent. */ #define TREE_REGULAR 1 #define TREE_POSTDESCENT 2 #define TREE_POSTASCENT 3 #define TREE_ERROR_DIR -1 #define TREE_ERROR_FATAL -2 static int tree_next(struct tree *); /* * Return information about the current entry. */ /* * The current full pathname, length of the full pathname, and a name * that can be used to access the file. Because tree does use chdir * extensively, the access path is almost never the same as the full * current path. * * TODO: On platforms that support it, use openat()-style operations * to eliminate the chdir() operations entirely while still supporting * arbitrarily deep traversals. This makes access_path troublesome to * support, of course, which means we'll need a rich enough interface * that clients can function without it. (In particular, we'll need * tree_current_open() that returns an open file descriptor.) * */ static const char *tree_current_path(struct tree *); static const char *tree_current_access_path(struct tree *); /* * Request the lstat() or stat() data for the current path. Since the * tree package needs to do some of this anyway, and caches the * results, you should take advantage of it here if you need it rather * than make a redundant stat() or lstat() call of your own. */ static const struct stat *tree_current_stat(struct tree *); static const struct stat *tree_current_lstat(struct tree *); static int tree_current_is_symblic_link_target(struct tree *); /* The following functions use tricks to avoid a certain number of * stat()/lstat() calls. */ /* "is_physical_dir" is equivalent to S_ISDIR(tree_current_lstat()->st_mode) */ static int tree_current_is_physical_dir(struct tree *); /* "is_dir" is equivalent to S_ISDIR(tree_current_stat()->st_mode) */ static int tree_current_is_dir(struct tree *); static int update_current_filesystem(struct archive_read_disk *a, int64_t dev); static int setup_current_filesystem(struct archive_read_disk *); static int tree_target_is_same_as_parent(struct tree *, const struct stat *); static int _archive_read_disk_open(struct archive *, const char *); static int _archive_read_free(struct archive *); static int _archive_read_close(struct archive *); static int _archive_read_data_block(struct archive *, const void **, size_t *, int64_t *); static int _archive_read_next_header(struct archive *, struct archive_entry **); static int _archive_read_next_header2(struct archive *, struct archive_entry *); static const char *trivial_lookup_gname(void *, int64_t gid); static const char *trivial_lookup_uname(void *, int64_t uid); static int setup_sparse(struct archive_read_disk *, struct archive_entry *); static int close_and_restore_time(int fd, struct tree *, struct restore_time *); static int open_on_current_dir(struct tree *, const char *, int); static int tree_dup(int); static struct archive_vtable * archive_read_disk_vtable(void) { static struct archive_vtable av; static int inited = 0; if (!inited) { av.archive_free = _archive_read_free; av.archive_close = _archive_read_close; av.archive_read_data_block = _archive_read_data_block; av.archive_read_next_header = _archive_read_next_header; av.archive_read_next_header2 = _archive_read_next_header2; inited = 1; } return (&av); } const char * archive_read_disk_gname(struct archive *_a, int64_t gid) { struct archive_read_disk *a = (struct archive_read_disk *)_a; if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_gname")) return (NULL); if (a->lookup_gname == NULL) return (NULL); return ((*a->lookup_gname)(a->lookup_gname_data, gid)); } const char * archive_read_disk_uname(struct archive *_a, int64_t uid) { struct archive_read_disk *a = (struct archive_read_disk *)_a; if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_uname")) return (NULL); if (a->lookup_uname == NULL) return (NULL); return ((*a->lookup_uname)(a->lookup_uname_data, uid)); } int archive_read_disk_set_gname_lookup(struct archive *_a, void *private_data, const char * (*lookup_gname)(void *private, int64_t gid), void (*cleanup_gname)(void *private)) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_gname_lookup"); if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL) (a->cleanup_gname)(a->lookup_gname_data); a->lookup_gname = lookup_gname; a->cleanup_gname = cleanup_gname; a->lookup_gname_data = private_data; return (ARCHIVE_OK); } int archive_read_disk_set_uname_lookup(struct archive *_a, void *private_data, const char * (*lookup_uname)(void *private, int64_t uid), void (*cleanup_uname)(void *private)) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_uname_lookup"); if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL) (a->cleanup_uname)(a->lookup_uname_data); a->lookup_uname = lookup_uname; a->cleanup_uname = cleanup_uname; a->lookup_uname_data = private_data; return (ARCHIVE_OK); } /* * Create a new archive_read_disk object and initialize it with global state. */ struct archive * archive_read_disk_new(void) { struct archive_read_disk *a; a = (struct archive_read_disk *)calloc(1, sizeof(*a)); if (a == NULL) return (NULL); a->archive.magic = ARCHIVE_READ_DISK_MAGIC; a->archive.state = ARCHIVE_STATE_NEW; a->archive.vtable = archive_read_disk_vtable(); a->entry = archive_entry_new2(&a->archive); a->lookup_uname = trivial_lookup_uname; a->lookup_gname = trivial_lookup_gname; a->enable_copyfile = 1; a->traverse_mount_points = 1; a->open_on_current_dir = open_on_current_dir; a->tree_current_dir_fd = tree_current_dir_fd; a->tree_enter_working_dir = tree_enter_working_dir; return (&a->archive); } static int _archive_read_free(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; int r; if (_a == NULL) return (ARCHIVE_OK); archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free"); if (a->archive.state != ARCHIVE_STATE_CLOSED) r = _archive_read_close(&a->archive); else r = ARCHIVE_OK; tree_free(a->tree); if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL) (a->cleanup_gname)(a->lookup_gname_data); if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL) (a->cleanup_uname)(a->lookup_uname_data); archive_string_free(&a->archive.error_string); archive_entry_free(a->entry); a->archive.magic = 0; __archive_clean(&a->archive); free(a); return (r); } static int _archive_read_close(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close"); if (a->archive.state != ARCHIVE_STATE_FATAL) a->archive.state = ARCHIVE_STATE_CLOSED; tree_close(a->tree); return (ARCHIVE_OK); } static void setup_symlink_mode(struct archive_read_disk *a, char symlink_mode, int follow_symlinks) { a->symlink_mode = symlink_mode; a->follow_symlinks = follow_symlinks; if (a->tree != NULL) { a->tree->initial_symlink_mode = a->symlink_mode; a->tree->symlink_mode = a->symlink_mode; } } int archive_read_disk_set_symlink_logical(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_logical"); setup_symlink_mode(a, 'L', 1); return (ARCHIVE_OK); } int archive_read_disk_set_symlink_physical(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_physical"); setup_symlink_mode(a, 'P', 0); return (ARCHIVE_OK); } int archive_read_disk_set_symlink_hybrid(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_hybrid"); setup_symlink_mode(a, 'H', 1);/* Follow symlinks initially. */ return (ARCHIVE_OK); } int archive_read_disk_set_atime_restored(struct archive *_a) { #ifndef HAVE_UTIMES static int warning_done = 0; #endif struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_restore_atime"); #ifdef HAVE_UTIMES a->restore_time = 1; if (a->tree != NULL) a->tree->flags |= needsRestoreTimes; return (ARCHIVE_OK); #else if (warning_done) /* Warning was already emitted; suppress further warnings. */ return (ARCHIVE_OK); archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Cannot restore access time on this system"); warning_done = 1; return (ARCHIVE_WARN); #endif } int archive_read_disk_set_behavior(struct archive *_a, int flags) { struct archive_read_disk *a = (struct archive_read_disk *)_a; int r = ARCHIVE_OK; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_honor_nodump"); if (flags & ARCHIVE_READDISK_RESTORE_ATIME) r = archive_read_disk_set_atime_restored(_a); else { a->restore_time = 0; if (a->tree != NULL) a->tree->flags &= ~needsRestoreTimes; } if (flags & ARCHIVE_READDISK_HONOR_NODUMP) a->honor_nodump = 1; else a->honor_nodump = 0; if (flags & ARCHIVE_READDISK_MAC_COPYFILE) a->enable_copyfile = 1; else a->enable_copyfile = 0; if (flags & ARCHIVE_READDISK_NO_TRAVERSE_MOUNTS) a->traverse_mount_points = 0; else a->traverse_mount_points = 1; if (flags & ARCHIVE_READDISK_NO_XATTR) a->suppress_xattr = 1; else a->suppress_xattr = 0; return (r); } /* * Trivial implementations of gname/uname lookup functions. * These are normally overridden by the client, but these stub * versions ensure that we always have something that works. */ static const char * trivial_lookup_gname(void *private_data, int64_t gid) { (void)private_data; /* UNUSED */ (void)gid; /* UNUSED */ return (NULL); } static const char * trivial_lookup_uname(void *private_data, int64_t uid) { (void)private_data; /* UNUSED */ (void)uid; /* UNUSED */ return (NULL); } /* * Allocate memory for the reading buffer adjusted to the filesystem * alignment. */ static int setup_suitable_read_buffer(struct archive_read_disk *a) { struct tree *t = a->tree; struct filesystem *cf = t->current_filesystem; size_t asize; size_t s; if (cf->allocation_ptr == NULL) { /* If we couldn't get a filesystem alignment, * we use 4096 as default value but we won't use * O_DIRECT to open() and openat() operations. */ long xfer_align = (cf->xfer_align == -1)?4096:cf->xfer_align; if (cf->max_xfer_size != -1) asize = cf->max_xfer_size + xfer_align; else { long incr = cf->incr_xfer_size; /* Some platform does not set a proper value to * incr_xfer_size.*/ if (incr < 0) incr = cf->min_xfer_size; if (cf->min_xfer_size < 0) { incr = xfer_align; asize = xfer_align; } else asize = cf->min_xfer_size; /* Increase a buffer size up to 64K bytes in * a proper incremant size. */ while (asize < 1024*64) asize += incr; /* Take a margin to adjust to the filesystem * alignment. */ asize += xfer_align; } cf->allocation_ptr = malloc(asize); if (cf->allocation_ptr == NULL) { archive_set_error(&a->archive, ENOMEM, "Couldn't allocate memory"); a->archive.state = ARCHIVE_STATE_FATAL; return (ARCHIVE_FATAL); } /* * Calculate proper address for the filesystem. */ s = (uintptr_t)cf->allocation_ptr; s %= xfer_align; if (s > 0) s = xfer_align - s; /* * Set a read buffer pointer in the proper alignment of * the current filesystem. */ cf->buff = cf->allocation_ptr + s; cf->buff_size = asize - xfer_align; } return (ARCHIVE_OK); } static int _archive_read_data_block(struct archive *_a, const void **buff, size_t *size, int64_t *offset) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct tree *t = a->tree; int r; ssize_t bytes; size_t buffbytes; int empty_sparse_region = 0; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_data_block"); if (t->entry_eof || t->entry_remaining_bytes <= 0) { r = ARCHIVE_EOF; goto abort_read_data; } /* * Open the current file. */ if (t->entry_fd < 0) { int flags = O_RDONLY | O_BINARY | O_CLOEXEC; /* * Eliminate or reduce cache effects if we can. * * Carefully consider this to be enabled. */ #if defined(O_DIRECT) && 0/* Disabled for now */ if (t->current_filesystem->xfer_align != -1 && t->nlink == 1) flags |= O_DIRECT; #endif #if defined(O_NOATIME) /* * Linux has O_NOATIME flag; use it if we need. */ if ((t->flags & needsRestoreTimes) != 0 && t->restore_time.noatime == 0) flags |= O_NOATIME; do { #endif t->entry_fd = open_on_current_dir(t, tree_current_access_path(t), flags); __archive_ensure_cloexec_flag(t->entry_fd); #if defined(O_NOATIME) /* * When we did open the file with O_NOATIME flag, * if successful, set 1 to t->restore_time.noatime * not to restore an atime of the file later. * if failed by EPERM, retry it without O_NOATIME flag. */ if (flags & O_NOATIME) { if (t->entry_fd >= 0) t->restore_time.noatime = 1; else if (errno == EPERM) { flags &= ~O_NOATIME; continue; } } } while (0); #endif if (t->entry_fd < 0) { archive_set_error(&a->archive, errno, "Couldn't open %s", tree_current_path(t)); r = ARCHIVE_FAILED; tree_enter_initial_dir(t); goto abort_read_data; } tree_enter_initial_dir(t); } /* * Allocate read buffer if not allocated. */ if (t->current_filesystem->allocation_ptr == NULL) { r = setup_suitable_read_buffer(a); if (r != ARCHIVE_OK) { a->archive.state = ARCHIVE_STATE_FATAL; goto abort_read_data; } } t->entry_buff = t->current_filesystem->buff; t->entry_buff_size = t->current_filesystem->buff_size; buffbytes = t->entry_buff_size; if ((int64_t)buffbytes > t->current_sparse->length) buffbytes = t->current_sparse->length; if (t->current_sparse->length == 0) empty_sparse_region = 1; /* * Skip hole. * TODO: Should we consider t->current_filesystem->xfer_align? */ if (t->current_sparse->offset > t->entry_total) { if (lseek(t->entry_fd, (off_t)t->current_sparse->offset, SEEK_SET) < 0) { archive_set_error(&a->archive, errno, "Seek error"); r = ARCHIVE_FATAL; a->archive.state = ARCHIVE_STATE_FATAL; goto abort_read_data; } bytes = t->current_sparse->offset - t->entry_total; t->entry_remaining_bytes -= bytes; t->entry_total += bytes; } /* * Read file contents. */ if (buffbytes > 0) { bytes = read(t->entry_fd, t->entry_buff, buffbytes); if (bytes < 0) { archive_set_error(&a->archive, errno, "Read error"); r = ARCHIVE_FATAL; a->archive.state = ARCHIVE_STATE_FATAL; goto abort_read_data; } } else bytes = 0; /* * Return an EOF unless we've read a leading empty sparse region, which * is used to represent fully-sparse files. */ if (bytes == 0 && !empty_sparse_region) { /* Get EOF */ t->entry_eof = 1; r = ARCHIVE_EOF; goto abort_read_data; } *buff = t->entry_buff; *size = bytes; *offset = t->entry_total; t->entry_total += bytes; t->entry_remaining_bytes -= bytes; if (t->entry_remaining_bytes == 0) { /* Close the current file descriptor */ close_and_restore_time(t->entry_fd, t, &t->restore_time); t->entry_fd = -1; t->entry_eof = 1; } t->current_sparse->offset += bytes; t->current_sparse->length -= bytes; if (t->current_sparse->length == 0 && !t->entry_eof) t->current_sparse++; return (ARCHIVE_OK); abort_read_data: *buff = NULL; *size = 0; *offset = t->entry_total; if (t->entry_fd >= 0) { /* Close the current file descriptor */ close_and_restore_time(t->entry_fd, t, &t->restore_time); t->entry_fd = -1; } return (r); } static int next_entry(struct archive_read_disk *a, struct tree *t, struct archive_entry *entry) { const struct stat *st; /* info to use for this entry */ const struct stat *lst;/* lstat() information */ const char *name; int descend, r; st = NULL; lst = NULL; t->descend = 0; do { switch (tree_next(t)) { case TREE_ERROR_FATAL: archive_set_error(&a->archive, t->tree_errno, "%s: Unable to continue traversing directory tree", tree_current_path(t)); a->archive.state = ARCHIVE_STATE_FATAL; tree_enter_initial_dir(t); return (ARCHIVE_FATAL); case TREE_ERROR_DIR: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "%s: Couldn't visit directory", tree_current_path(t)); tree_enter_initial_dir(t); return (ARCHIVE_FAILED); case 0: tree_enter_initial_dir(t); return (ARCHIVE_EOF); case TREE_POSTDESCENT: case TREE_POSTASCENT: break; case TREE_REGULAR: lst = tree_current_lstat(t); if (lst == NULL) { archive_set_error(&a->archive, errno, "%s: Cannot stat", tree_current_path(t)); tree_enter_initial_dir(t); return (ARCHIVE_FAILED); } break; } } while (lst == NULL); #ifdef __APPLE__ if (a->enable_copyfile) { /* If we're using copyfile(), ignore "._XXX" files. */ const char *bname = strrchr(tree_current_path(t), '/'); if (bname == NULL) bname = tree_current_path(t); else ++bname; if (bname[0] == '.' && bname[1] == '_') return (ARCHIVE_RETRY); } #endif archive_entry_copy_pathname(entry, tree_current_path(t)); /* * Perform path matching. */ if (a->matching) { r = archive_match_path_excluded(a->matching, entry); if (r < 0) { archive_set_error(&(a->archive), errno, "Faild : %s", archive_error_string(a->matching)); return (r); } if (r) { if (a->excluded_cb_func) a->excluded_cb_func(&(a->archive), a->excluded_cb_data, entry); return (ARCHIVE_RETRY); } } /* * Distinguish 'L'/'P'/'H' symlink following. */ switch(t->symlink_mode) { case 'H': /* 'H': After the first item, rest like 'P'. */ t->symlink_mode = 'P'; /* 'H': First item (from command line) like 'L'. */ /* FALLTHROUGH */ case 'L': /* 'L': Do descend through a symlink to dir. */ descend = tree_current_is_dir(t); /* 'L': Follow symlinks to files. */ a->symlink_mode = 'L'; a->follow_symlinks = 1; /* 'L': Archive symlinks as targets, if we can. */ st = tree_current_stat(t); if (st != NULL && !tree_target_is_same_as_parent(t, st)) break; /* If stat fails, we have a broken symlink; * in that case, don't follow the link. */ /* FALLTHROUGH */ default: /* 'P': Don't descend through a symlink to dir. */ descend = tree_current_is_physical_dir(t); /* 'P': Don't follow symlinks to files. */ a->symlink_mode = 'P'; a->follow_symlinks = 0; /* 'P': Archive symlinks as symlinks. */ st = lst; break; } if (update_current_filesystem(a, st->st_dev) != ARCHIVE_OK) { a->archive.state = ARCHIVE_STATE_FATAL; tree_enter_initial_dir(t); return (ARCHIVE_FATAL); } if (t->initial_filesystem_id == -1) t->initial_filesystem_id = t->current_filesystem_id; if (!a->traverse_mount_points) { if (t->initial_filesystem_id != t->current_filesystem_id) descend = 0; } t->descend = descend; /* * Honor nodump flag. * If the file is marked with nodump flag, do not return this entry. */ if (a->honor_nodump) { #if defined(HAVE_STRUCT_STAT_ST_FLAGS) && defined(UF_NODUMP) if (st->st_flags & UF_NODUMP) return (ARCHIVE_RETRY); #elif defined(EXT2_IOC_GETFLAGS) && defined(EXT2_NODUMP_FL) &&\ defined(HAVE_WORKING_EXT2_IOC_GETFLAGS) if (S_ISREG(st->st_mode) || S_ISDIR(st->st_mode)) { int stflags; t->entry_fd = open_on_current_dir(t, tree_current_access_path(t), O_RDONLY | O_NONBLOCK | O_CLOEXEC); __archive_ensure_cloexec_flag(t->entry_fd); if (t->entry_fd >= 0) { r = ioctl(t->entry_fd, EXT2_IOC_GETFLAGS, &stflags); if (r == 0 && (stflags & EXT2_NODUMP_FL) != 0) return (ARCHIVE_RETRY); } } #endif } archive_entry_copy_stat(entry, st); /* Save the times to be restored. This must be in before * calling archive_read_disk_descend() or any chance of it, * especially, invokng a callback. */ t->restore_time.mtime = archive_entry_mtime(entry); t->restore_time.mtime_nsec = archive_entry_mtime_nsec(entry); t->restore_time.atime = archive_entry_atime(entry); t->restore_time.atime_nsec = archive_entry_atime_nsec(entry); t->restore_time.filetype = archive_entry_filetype(entry); t->restore_time.noatime = t->current_filesystem->noatime; /* * Perform time matching. */ if (a->matching) { r = archive_match_time_excluded(a->matching, entry); if (r < 0) { archive_set_error(&(a->archive), errno, "Faild : %s", archive_error_string(a->matching)); return (r); } if (r) { if (a->excluded_cb_func) a->excluded_cb_func(&(a->archive), a->excluded_cb_data, entry); return (ARCHIVE_RETRY); } } /* Lookup uname/gname */ name = archive_read_disk_uname(&(a->archive), archive_entry_uid(entry)); if (name != NULL) archive_entry_copy_uname(entry, name); name = archive_read_disk_gname(&(a->archive), archive_entry_gid(entry)); if (name != NULL) archive_entry_copy_gname(entry, name); /* * Perform owner matching. */ if (a->matching) { r = archive_match_owner_excluded(a->matching, entry); if (r < 0) { archive_set_error(&(a->archive), errno, "Faild : %s", archive_error_string(a->matching)); return (r); } if (r) { if (a->excluded_cb_func) a->excluded_cb_func(&(a->archive), a->excluded_cb_data, entry); return (ARCHIVE_RETRY); } } /* * Invoke a meta data filter callback. */ if (a->metadata_filter_func) { if (!a->metadata_filter_func(&(a->archive), a->metadata_filter_data, entry)) return (ARCHIVE_RETRY); } /* * Populate the archive_entry with metadata from the disk. */ archive_entry_copy_sourcepath(entry, tree_current_access_path(t)); r = archive_read_disk_entry_from_file(&(a->archive), entry, t->entry_fd, st); return (r); } static int _archive_read_next_header(struct archive *_a, struct archive_entry **entryp) { int ret; struct archive_read_disk *a = (struct archive_read_disk *)_a; *entryp = NULL; ret = _archive_read_next_header2(_a, a->entry); *entryp = a->entry; return ret; } static int _archive_read_next_header2(struct archive *_a, struct archive_entry *entry) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct tree *t; int r; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_read_next_header2"); t = a->tree; if (t->entry_fd >= 0) { close_and_restore_time(t->entry_fd, t, &t->restore_time); t->entry_fd = -1; } for (;;) { r = next_entry(a, t, entry); if (t->entry_fd >= 0) { close(t->entry_fd); t->entry_fd = -1; } if (r == ARCHIVE_RETRY) { archive_entry_clear(entry); continue; } break; } /* Return to the initial directory. */ tree_enter_initial_dir(t); /* * EOF and FATAL are persistent at this layer. By * modifying the state, we guarantee that future calls to * read a header or read data will fail. */ switch (r) { case ARCHIVE_EOF: a->archive.state = ARCHIVE_STATE_EOF; break; case ARCHIVE_OK: case ARCHIVE_WARN: /* Overwrite the sourcepath based on the initial directory. */ archive_entry_copy_sourcepath(entry, tree_current_path(t)); t->entry_total = 0; if (archive_entry_filetype(entry) == AE_IFREG) { t->nlink = archive_entry_nlink(entry); t->entry_remaining_bytes = archive_entry_size(entry); t->entry_eof = (t->entry_remaining_bytes == 0)? 1: 0; if (!t->entry_eof && setup_sparse(a, entry) != ARCHIVE_OK) return (ARCHIVE_FATAL); } else { t->entry_remaining_bytes = 0; t->entry_eof = 1; } a->archive.state = ARCHIVE_STATE_DATA; break; case ARCHIVE_RETRY: break; case ARCHIVE_FATAL: a->archive.state = ARCHIVE_STATE_FATAL; break; } __archive_reset_read_data(&a->archive); return (r); } static int setup_sparse(struct archive_read_disk *a, struct archive_entry *entry) { struct tree *t = a->tree; int64_t length, offset; int i; t->sparse_count = archive_entry_sparse_reset(entry); if (t->sparse_count+1 > t->sparse_list_size) { free(t->sparse_list); t->sparse_list_size = t->sparse_count + 1; t->sparse_list = malloc(sizeof(t->sparse_list[0]) * t->sparse_list_size); if (t->sparse_list == NULL) { t->sparse_list_size = 0; archive_set_error(&a->archive, ENOMEM, "Can't allocate data"); a->archive.state = ARCHIVE_STATE_FATAL; return (ARCHIVE_FATAL); } } for (i = 0; i < t->sparse_count; i++) { archive_entry_sparse_next(entry, &offset, &length); t->sparse_list[i].offset = offset; t->sparse_list[i].length = length; } if (i == 0) { t->sparse_list[i].offset = 0; t->sparse_list[i].length = archive_entry_size(entry); } else { t->sparse_list[i].offset = archive_entry_size(entry); t->sparse_list[i].length = 0; } t->current_sparse = t->sparse_list; return (ARCHIVE_OK); } int archive_read_disk_set_matching(struct archive *_a, struct archive *_ma, void (*_excluded_func)(struct archive *, void *, struct archive_entry *), void *_client_data) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_matching"); a->matching = _ma; a->excluded_cb_func = _excluded_func; a->excluded_cb_data = _client_data; return (ARCHIVE_OK); } int archive_read_disk_set_metadata_filter_callback(struct archive *_a, int (*_metadata_filter_func)(struct archive *, void *, struct archive_entry *), void *_client_data) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_metadata_filter_callback"); a->metadata_filter_func = _metadata_filter_func; a->metadata_filter_data = _client_data; return (ARCHIVE_OK); } int archive_read_disk_can_descend(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct tree *t = a->tree; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_read_disk_can_descend"); return (t->visit_type == TREE_REGULAR && t->descend); } /* * Called by the client to mark the directory just returned from * tree_next() as needing to be visited. */ int archive_read_disk_descend(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct tree *t = a->tree; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_read_disk_descend"); if (t->visit_type != TREE_REGULAR || !t->descend) return (ARCHIVE_OK); if (tree_current_is_physical_dir(t)) { tree_push(t, t->basename, t->current_filesystem_id, t->lst.st_dev, t->lst.st_ino, &t->restore_time); t->stack->flags |= isDir; } else if (tree_current_is_dir(t)) { tree_push(t, t->basename, t->current_filesystem_id, t->st.st_dev, t->st.st_ino, &t->restore_time); t->stack->flags |= isDirLink; } t->descend = 0; return (ARCHIVE_OK); } int archive_read_disk_open(struct archive *_a, const char *pathname) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED, "archive_read_disk_open"); archive_clear_error(&a->archive); return (_archive_read_disk_open(_a, pathname)); } int archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct archive_string path; int ret; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED, "archive_read_disk_open_w"); archive_clear_error(&a->archive); /* Make a char string from a wchar_t string. */ archive_string_init(&path); if (archive_string_append_from_wcs(&path, pathname, wcslen(pathname)) != 0) { if (errno == ENOMEM) archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); else archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Can't convert a path to a char string"); a->archive.state = ARCHIVE_STATE_FATAL; ret = ARCHIVE_FATAL; } else ret = _archive_read_disk_open(_a, path.s); archive_string_free(&path); return (ret); } static int _archive_read_disk_open(struct archive *_a, const char *pathname) { struct archive_read_disk *a = (struct archive_read_disk *)_a; if (a->tree != NULL) a->tree = tree_reopen(a->tree, pathname, a->restore_time); else a->tree = tree_open(pathname, a->symlink_mode, a->restore_time); if (a->tree == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate tar data"); a->archive.state = ARCHIVE_STATE_FATAL; return (ARCHIVE_FATAL); } a->archive.state = ARCHIVE_STATE_HEADER; return (ARCHIVE_OK); } /* * Return a current filesystem ID which is index of the filesystem entry * you've visited through archive_read_disk. */ int archive_read_disk_current_filesystem(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_disk_current_filesystem"); return (a->tree->current_filesystem_id); } static int update_current_filesystem(struct archive_read_disk *a, int64_t dev) { struct tree *t = a->tree; int i, fid; if (t->current_filesystem != NULL && t->current_filesystem->dev == dev) return (ARCHIVE_OK); for (i = 0; i < t->max_filesystem_id; i++) { if (t->filesystem_table[i].dev == dev) { /* There is the filesytem ID we've already generated. */ t->current_filesystem_id = i; t->current_filesystem = &(t->filesystem_table[i]); return (ARCHIVE_OK); } } /* * This is the new filesytem which we have to generate a new ID for. */ fid = t->max_filesystem_id++; if (t->max_filesystem_id > t->allocated_filesytem) { size_t s; void *p; s = t->max_filesystem_id * 2; p = realloc(t->filesystem_table, s * sizeof(*t->filesystem_table)); if (p == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate tar data"); return (ARCHIVE_FATAL); } t->filesystem_table = (struct filesystem *)p; t->allocated_filesytem = s; } t->current_filesystem_id = fid; t->current_filesystem = &(t->filesystem_table[fid]); t->current_filesystem->dev = dev; t->current_filesystem->allocation_ptr = NULL; t->current_filesystem->buff = NULL; /* Setup the current filesystem properties which depend on * platform specific. */ return (setup_current_filesystem(a)); } /* * Returns 1 if current filesystem is generated filesystem, 0 if it is not * or -1 if it is unknown. */ int archive_read_disk_current_filesystem_is_synthetic(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_disk_current_filesystem"); return (a->tree->current_filesystem->synthetic); } /* * Returns 1 if current filesystem is remote filesystem, 0 if it is not * or -1 if it is unknown. */ int archive_read_disk_current_filesystem_is_remote(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_disk_current_filesystem"); return (a->tree->current_filesystem->remote); } #if defined(_PC_REC_INCR_XFER_SIZE) && defined(_PC_REC_MAX_XFER_SIZE) &&\ defined(_PC_REC_MIN_XFER_SIZE) && defined(_PC_REC_XFER_ALIGN) static int get_xfer_size(struct tree *t, int fd, const char *path) { t->current_filesystem->xfer_align = -1; errno = 0; if (fd >= 0) { t->current_filesystem->incr_xfer_size = fpathconf(fd, _PC_REC_INCR_XFER_SIZE); t->current_filesystem->max_xfer_size = fpathconf(fd, _PC_REC_MAX_XFER_SIZE); t->current_filesystem->min_xfer_size = fpathconf(fd, _PC_REC_MIN_XFER_SIZE); t->current_filesystem->xfer_align = fpathconf(fd, _PC_REC_XFER_ALIGN); } else if (path != NULL) { t->current_filesystem->incr_xfer_size = pathconf(path, _PC_REC_INCR_XFER_SIZE); t->current_filesystem->max_xfer_size = pathconf(path, _PC_REC_MAX_XFER_SIZE); t->current_filesystem->min_xfer_size = pathconf(path, _PC_REC_MIN_XFER_SIZE); t->current_filesystem->xfer_align = pathconf(path, _PC_REC_XFER_ALIGN); } /* At least we need an alignment size. */ if (t->current_filesystem->xfer_align == -1) return ((errno == EINVAL)?1:-1); else return (0); } #else static int get_xfer_size(struct tree *t, int fd, const char *path) { (void)t; /* UNUSED */ (void)fd; /* UNUSED */ (void)path; /* UNUSED */ return (1);/* Not supported */ } #endif #if defined(HAVE_STATFS) && defined(HAVE_FSTATFS) && defined(MNT_LOCAL) \ && !defined(ST_LOCAL) /* * Gather current filesystem properties on FreeBSD, OpenBSD and Mac OS X. */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; struct statfs sfs; #if defined(HAVE_GETVFSBYNAME) && defined(VFCF_SYNTHETIC) +# if defined(HAVE_STRUCT_VFSCONF) + struct vfsconf vfc; +# else struct xvfsconf vfc; +# endif #endif int r, xr = 0; #if !defined(HAVE_STRUCT_STATFS_F_NAMEMAX) long nm; #endif t->current_filesystem->synthetic = -1; t->current_filesystem->remote = -1; if (tree_current_is_symblic_link_target(t)) { #if defined(HAVE_OPENAT) /* * Get file system statistics on any directory * where current is. */ int fd = openat(tree_current_dir_fd(t), tree_current_access_path(t), O_RDONLY | O_CLOEXEC); __archive_ensure_cloexec_flag(fd); if (fd < 0) { archive_set_error(&a->archive, errno, "openat failed"); return (ARCHIVE_FAILED); } r = fstatfs(fd, &sfs); if (r == 0) xr = get_xfer_size(t, fd, NULL); close(fd); #else if (tree_enter_working_dir(t) != 0) { archive_set_error(&a->archive, errno, "fchdir failed"); return (ARCHIVE_FAILED); } r = statfs(tree_current_access_path(t), &sfs); if (r == 0) xr = get_xfer_size(t, -1, tree_current_access_path(t)); #endif } else { r = fstatfs(tree_current_dir_fd(t), &sfs); if (r == 0) xr = get_xfer_size(t, tree_current_dir_fd(t), NULL); } if (r == -1 || xr == -1) { archive_set_error(&a->archive, errno, "statfs failed"); return (ARCHIVE_FAILED); } else if (xr == 1) { /* pathconf(_PC_REX_*) operations are not supported. */ t->current_filesystem->xfer_align = sfs.f_bsize; t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = sfs.f_iosize; t->current_filesystem->incr_xfer_size = sfs.f_iosize; } if (sfs.f_flags & MNT_LOCAL) t->current_filesystem->remote = 0; else t->current_filesystem->remote = 1; #if defined(HAVE_GETVFSBYNAME) && defined(VFCF_SYNTHETIC) r = getvfsbyname(sfs.f_fstypename, &vfc); if (r == -1) { archive_set_error(&a->archive, errno, "getvfsbyname failed"); return (ARCHIVE_FAILED); } if (vfc.vfc_flags & VFCF_SYNTHETIC) t->current_filesystem->synthetic = 1; else t->current_filesystem->synthetic = 0; #endif #if defined(MNT_NOATIME) if (sfs.f_flags & MNT_NOATIME) t->current_filesystem->noatime = 1; else #endif t->current_filesystem->noatime = 0; #if defined(HAVE_READDIR_R) /* Set maximum filename length. */ #if defined(HAVE_STRUCT_STATFS_F_NAMEMAX) t->current_filesystem->name_max = sfs.f_namemax; #else # if defined(_PC_NAME_MAX) /* Mac OS X does not have f_namemax in struct statfs. */ if (tree_current_is_symblic_link_target(t)) { if (tree_enter_working_dir(t) != 0) { archive_set_error(&a->archive, errno, "fchdir failed"); return (ARCHIVE_FAILED); } nm = pathconf(tree_current_access_path(t), _PC_NAME_MAX); } else nm = fpathconf(tree_current_dir_fd(t), _PC_NAME_MAX); # else nm = -1; # endif if (nm == -1) t->current_filesystem->name_max = NAME_MAX; else t->current_filesystem->name_max = nm; #endif #endif /* HAVE_READDIR_R */ return (ARCHIVE_OK); } #elif (defined(HAVE_STATVFS) || defined(HAVE_FSTATVFS)) && defined(ST_LOCAL) /* * Gather current filesystem properties on NetBSD */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; struct statvfs sfs; int r, xr = 0; t->current_filesystem->synthetic = -1; if (tree_enter_working_dir(t) != 0) { archive_set_error(&a->archive, errno, "fchdir failed"); return (ARCHIVE_FAILED); } if (tree_current_is_symblic_link_target(t)) { r = statvfs(tree_current_access_path(t), &sfs); if (r == 0) xr = get_xfer_size(t, -1, tree_current_access_path(t)); } else { #ifdef HAVE_FSTATVFS r = fstatvfs(tree_current_dir_fd(t), &sfs); if (r == 0) xr = get_xfer_size(t, tree_current_dir_fd(t), NULL); #else r = statvfs(".", &sfs); if (r == 0) xr = get_xfer_size(t, -1, "."); #endif } if (r == -1 || xr == -1) { t->current_filesystem->remote = -1; archive_set_error(&a->archive, errno, "statvfs failed"); return (ARCHIVE_FAILED); } else if (xr == 1) { /* Usuall come here unless NetBSD supports _PC_REC_XFER_ALIGN * for pathconf() function. */ t->current_filesystem->xfer_align = sfs.f_frsize; t->current_filesystem->max_xfer_size = -1; #if defined(HAVE_STRUCT_STATVFS_F_IOSIZE) t->current_filesystem->min_xfer_size = sfs.f_iosize; t->current_filesystem->incr_xfer_size = sfs.f_iosize; #else t->current_filesystem->min_xfer_size = sfs.f_bsize; t->current_filesystem->incr_xfer_size = sfs.f_bsize; #endif } if (sfs.f_flag & ST_LOCAL) t->current_filesystem->remote = 0; else t->current_filesystem->remote = 1; #if defined(ST_NOATIME) if (sfs.f_flag & ST_NOATIME) t->current_filesystem->noatime = 1; else #endif t->current_filesystem->noatime = 0; /* Set maximum filename length. */ t->current_filesystem->name_max = sfs.f_namemax; return (ARCHIVE_OK); } #elif defined(HAVE_SYS_STATFS_H) && defined(HAVE_LINUX_MAGIC_H) &&\ defined(HAVE_STATFS) && defined(HAVE_FSTATFS) /* * Note: statfs is deprecated since LSB 3.2 */ #ifndef CIFS_SUPER_MAGIC #define CIFS_SUPER_MAGIC 0xFF534D42 #endif #ifndef DEVFS_SUPER_MAGIC #define DEVFS_SUPER_MAGIC 0x1373 #endif /* * Gather current filesystem properties on Linux */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; struct statfs sfs; #if defined(HAVE_STATVFS) struct statvfs svfs; #endif int r, vr = 0, xr = 0; if (tree_current_is_symblic_link_target(t)) { #if defined(HAVE_OPENAT) /* * Get file system statistics on any directory * where current is. */ int fd = openat(tree_current_dir_fd(t), tree_current_access_path(t), O_RDONLY | O_CLOEXEC); __archive_ensure_cloexec_flag(fd); if (fd < 0) { archive_set_error(&a->archive, errno, "openat failed"); return (ARCHIVE_FAILED); } #if defined(HAVE_FSTATVFS) vr = fstatvfs(fd, &svfs);/* for f_flag, mount flags */ #endif r = fstatfs(fd, &sfs); if (r == 0) xr = get_xfer_size(t, fd, NULL); close(fd); #else if (tree_enter_working_dir(t) != 0) { archive_set_error(&a->archive, errno, "fchdir failed"); return (ARCHIVE_FAILED); } #if defined(HAVE_STATVFS) vr = statvfs(tree_current_access_path(t), &svfs); #endif r = statfs(tree_current_access_path(t), &sfs); if (r == 0) xr = get_xfer_size(t, -1, tree_current_access_path(t)); #endif } else { #ifdef HAVE_FSTATFS #if defined(HAVE_FSTATVFS) vr = fstatvfs(tree_current_dir_fd(t), &svfs); #endif r = fstatfs(tree_current_dir_fd(t), &sfs); if (r == 0) xr = get_xfer_size(t, tree_current_dir_fd(t), NULL); #else if (tree_enter_working_dir(t) != 0) { archive_set_error(&a->archive, errno, "fchdir failed"); return (ARCHIVE_FAILED); } #if defined(HAVE_STATVFS) vr = statvfs(".", &svfs); #endif r = statfs(".", &sfs); if (r == 0) xr = get_xfer_size(t, -1, "."); #endif } if (r == -1 || xr == -1 || vr == -1) { t->current_filesystem->synthetic = -1; t->current_filesystem->remote = -1; archive_set_error(&a->archive, errno, "statfs failed"); return (ARCHIVE_FAILED); } else if (xr == 1) { /* pathconf(_PC_REX_*) operations are not supported. */ #if defined(HAVE_STATVFS) t->current_filesystem->xfer_align = svfs.f_frsize; t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = svfs.f_bsize; t->current_filesystem->incr_xfer_size = svfs.f_bsize; #else t->current_filesystem->xfer_align = sfs.f_frsize; t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = sfs.f_bsize; t->current_filesystem->incr_xfer_size = sfs.f_bsize; #endif } switch (sfs.f_type) { case AFS_SUPER_MAGIC: case CIFS_SUPER_MAGIC: case CODA_SUPER_MAGIC: case NCP_SUPER_MAGIC:/* NetWare */ case NFS_SUPER_MAGIC: case SMB_SUPER_MAGIC: t->current_filesystem->remote = 1; t->current_filesystem->synthetic = 0; break; case DEVFS_SUPER_MAGIC: case PROC_SUPER_MAGIC: case USBDEVICE_SUPER_MAGIC: t->current_filesystem->remote = 0; t->current_filesystem->synthetic = 1; break; default: t->current_filesystem->remote = 0; t->current_filesystem->synthetic = 0; break; } #if defined(ST_NOATIME) #if defined(HAVE_STATVFS) if (svfs.f_flag & ST_NOATIME) #else if (sfs.f_flag & ST_NOATIME) #endif t->current_filesystem->noatime = 1; else #endif t->current_filesystem->noatime = 0; #if defined(HAVE_READDIR_R) /* Set maximum filename length. */ t->current_filesystem->name_max = sfs.f_namelen; #endif return (ARCHIVE_OK); } #elif defined(HAVE_SYS_STATVFS_H) &&\ (defined(HAVE_STATVFS) || defined(HAVE_FSTATVFS)) /* * Gather current filesystem properties on other posix platform. */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; struct statvfs sfs; int r, xr = 0; t->current_filesystem->synthetic = -1;/* Not supported */ t->current_filesystem->remote = -1;/* Not supported */ if (tree_current_is_symblic_link_target(t)) { #if defined(HAVE_OPENAT) /* * Get file system statistics on any directory * where current is. */ int fd = openat(tree_current_dir_fd(t), tree_current_access_path(t), O_RDONLY | O_CLOEXEC); __archive_ensure_cloexec_flag(fd); if (fd < 0) { archive_set_error(&a->archive, errno, "openat failed"); return (ARCHIVE_FAILED); } r = fstatvfs(fd, &sfs); if (r == 0) xr = get_xfer_size(t, fd, NULL); close(fd); #else if (tree_enter_working_dir(t) != 0) { archive_set_error(&a->archive, errno, "fchdir failed"); return (ARCHIVE_FAILED); } r = statvfs(tree_current_access_path(t), &sfs); if (r == 0) xr = get_xfer_size(t, -1, tree_current_access_path(t)); #endif } else { #ifdef HAVE_FSTATVFS r = fstatvfs(tree_current_dir_fd(t), &sfs); if (r == 0) xr = get_xfer_size(t, tree_current_dir_fd(t), NULL); #else if (tree_enter_working_dir(t) != 0) { archive_set_error(&a->archive, errno, "fchdir failed"); return (ARCHIVE_FAILED); } r = statvfs(".", &sfs); if (r == 0) xr = get_xfer_size(t, -1, "."); #endif } if (r == -1 || xr == -1) { t->current_filesystem->synthetic = -1; t->current_filesystem->remote = -1; archive_set_error(&a->archive, errno, "statvfs failed"); return (ARCHIVE_FAILED); } else if (xr == 1) { /* pathconf(_PC_REX_*) operations are not supported. */ t->current_filesystem->xfer_align = sfs.f_frsize; t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = sfs.f_bsize; t->current_filesystem->incr_xfer_size = sfs.f_bsize; } #if defined(ST_NOATIME) if (sfs.f_flag & ST_NOATIME) t->current_filesystem->noatime = 1; else #endif t->current_filesystem->noatime = 0; #if defined(HAVE_READDIR_R) /* Set maximum filename length. */ t->current_filesystem->name_max = sfs.f_namemax; #endif return (ARCHIVE_OK); } #else /* * Generic: Gather current filesystem properties. * TODO: Is this generic function really needed? */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; #if defined(_PC_NAME_MAX) && defined(HAVE_READDIR_R) long nm; #endif t->current_filesystem->synthetic = -1;/* Not supported */ t->current_filesystem->remote = -1;/* Not supported */ t->current_filesystem->noatime = 0; (void)get_xfer_size(t, -1, ".");/* Dummy call to avoid build error. */ t->current_filesystem->xfer_align = -1;/* Unknown */ t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = -1; t->current_filesystem->incr_xfer_size = -1; #if defined(HAVE_READDIR_R) /* Set maximum filename length. */ # if defined(_PC_NAME_MAX) if (tree_current_is_symblic_link_target(t)) { if (tree_enter_working_dir(t) != 0) { archive_set_error(&a->archive, errno, "fchdir failed"); return (ARCHIVE_FAILED); } nm = pathconf(tree_current_access_path(t), _PC_NAME_MAX); } else nm = fpathconf(tree_current_dir_fd(t), _PC_NAME_MAX); if (nm == -1) # endif /* _PC_NAME_MAX */ /* * Some sysmtes (HP-UX or others?) incorrectly defined * NAME_MAX macro to be a smaller value. */ # if defined(NAME_MAX) && NAME_MAX >= 255 t->current_filesystem->name_max = NAME_MAX; # else /* No way to get a trusted value of maximum filename * length. */ t->current_filesystem->name_max = PATH_MAX; # endif /* NAME_MAX */ # if defined(_PC_NAME_MAX) else t->current_filesystem->name_max = nm; # endif /* _PC_NAME_MAX */ #endif /* HAVE_READDIR_R */ return (ARCHIVE_OK); } #endif static int close_and_restore_time(int fd, struct tree *t, struct restore_time *rt) { #ifndef HAVE_UTIMES (void)t; /* UNUSED */ (void)rt; /* UNUSED */ return (close(fd)); #else #if defined(HAVE_FUTIMENS) && !defined(__CYGWIN__) struct timespec timespecs[2]; #endif struct timeval times[2]; if ((t->flags & needsRestoreTimes) == 0 || rt->noatime) { if (fd >= 0) return (close(fd)); else return (0); } #if defined(HAVE_FUTIMENS) && !defined(__CYGWIN__) timespecs[1].tv_sec = rt->mtime; timespecs[1].tv_nsec = rt->mtime_nsec; timespecs[0].tv_sec = rt->atime; timespecs[0].tv_nsec = rt->atime_nsec; /* futimens() is defined in POSIX.1-2008. */ if (futimens(fd, timespecs) == 0) return (close(fd)); #endif times[1].tv_sec = rt->mtime; times[1].tv_usec = rt->mtime_nsec / 1000; times[0].tv_sec = rt->atime; times[0].tv_usec = rt->atime_nsec / 1000; #if !defined(HAVE_FUTIMENS) && defined(HAVE_FUTIMES) && !defined(__CYGWIN__) if (futimes(fd, times) == 0) return (close(fd)); #endif close(fd); #if defined(HAVE_FUTIMESAT) if (futimesat(tree_current_dir_fd(t), rt->name, times) == 0) return (0); #endif #ifdef HAVE_LUTIMES if (lutimes(rt->name, times) != 0) #else if (AE_IFLNK != rt->filetype && utimes(rt->name, times) != 0) #endif return (-1); #endif return (0); } static int open_on_current_dir(struct tree *t, const char *path, int flags) { #ifdef HAVE_OPENAT return (openat(tree_current_dir_fd(t), path, flags)); #else if (tree_enter_working_dir(t) != 0) return (-1); return (open(path, flags)); #endif } static int tree_dup(int fd) { int new_fd; #ifdef F_DUPFD_CLOEXEC static volatile int can_dupfd_cloexec = 1; if (can_dupfd_cloexec) { new_fd = fcntl(fd, F_DUPFD_CLOEXEC, 0); if (new_fd != -1) return (new_fd); /* Linux 2.6.18 - 2.6.23 declare F_DUPFD_CLOEXEC, * but it cannot be used. So we have to try dup(). */ /* We won't try F_DUPFD_CLOEXEC. */ can_dupfd_cloexec = 0; } #endif /* F_DUPFD_CLOEXEC */ new_fd = dup(fd); __archive_ensure_cloexec_flag(new_fd); return (new_fd); } /* * Add a directory path to the current stack. */ static void tree_push(struct tree *t, const char *path, int filesystem_id, int64_t dev, int64_t ino, struct restore_time *rt) { struct tree_entry *te; te = malloc(sizeof(*te)); memset(te, 0, sizeof(*te)); te->next = t->stack; te->parent = t->current; if (te->parent) te->depth = te->parent->depth + 1; t->stack = te; archive_string_init(&te->name); te->symlink_parent_fd = -1; archive_strcpy(&te->name, path); te->flags = needsDescent | needsOpen | needsAscent; te->filesystem_id = filesystem_id; te->dev = dev; te->ino = ino; te->dirname_length = t->dirname_length; te->restore_time.name = te->name.s; if (rt != NULL) { te->restore_time.mtime = rt->mtime; te->restore_time.mtime_nsec = rt->mtime_nsec; te->restore_time.atime = rt->atime; te->restore_time.atime_nsec = rt->atime_nsec; te->restore_time.filetype = rt->filetype; te->restore_time.noatime = rt->noatime; } } /* * Append a name to the current dir path. */ static void tree_append(struct tree *t, const char *name, size_t name_length) { size_t size_needed; t->path.s[t->dirname_length] = '\0'; t->path.length = t->dirname_length; /* Strip trailing '/' from name, unless entire name is "/". */ while (name_length > 1 && name[name_length - 1] == '/') name_length--; /* Resize pathname buffer as needed. */ size_needed = name_length + t->dirname_length + 2; archive_string_ensure(&t->path, size_needed); /* Add a separating '/' if it's needed. */ if (t->dirname_length > 0 && t->path.s[archive_strlen(&t->path)-1] != '/') archive_strappend_char(&t->path, '/'); t->basename = t->path.s + archive_strlen(&t->path); archive_strncat(&t->path, name, name_length); t->restore_time.name = t->basename; } /* * Open a directory tree for traversal. */ static struct tree * tree_open(const char *path, int symlink_mode, int restore_time) { struct tree *t; if ((t = malloc(sizeof(*t))) == NULL) return (NULL); memset(t, 0, sizeof(*t)); archive_string_init(&t->path); archive_string_ensure(&t->path, 31); t->initial_symlink_mode = symlink_mode; return (tree_reopen(t, path, restore_time)); } static struct tree * tree_reopen(struct tree *t, const char *path, int restore_time) { t->flags = (restore_time)?needsRestoreTimes:0; t->flags |= onInitialDir; t->visit_type = 0; t->tree_errno = 0; t->dirname_length = 0; t->depth = 0; t->descend = 0; t->current = NULL; t->d = INVALID_DIR_HANDLE; t->symlink_mode = t->initial_symlink_mode; archive_string_empty(&t->path); t->entry_fd = -1; t->entry_eof = 0; t->entry_remaining_bytes = 0; t->initial_filesystem_id = -1; /* First item is set up a lot like a symlink traversal. */ tree_push(t, path, 0, 0, 0, NULL); t->stack->flags = needsFirstVisit; t->maxOpenCount = t->openCount = 1; t->initial_dir_fd = open(".", O_RDONLY | O_CLOEXEC); __archive_ensure_cloexec_flag(t->initial_dir_fd); t->working_dir_fd = tree_dup(t->initial_dir_fd); return (t); } static int tree_descent(struct tree *t) { int flag, new_fd, r = 0; t->dirname_length = archive_strlen(&t->path); flag = O_RDONLY | O_CLOEXEC; #if defined(O_DIRECTORY) flag |= O_DIRECTORY; #endif new_fd = open_on_current_dir(t, t->stack->name.s, flag); __archive_ensure_cloexec_flag(new_fd); if (new_fd < 0) { t->tree_errno = errno; r = TREE_ERROR_DIR; } else { t->depth++; /* If it is a link, set up fd for the ascent. */ if (t->stack->flags & isDirLink) { t->stack->symlink_parent_fd = t->working_dir_fd; t->openCount++; if (t->openCount > t->maxOpenCount) t->maxOpenCount = t->openCount; } else close(t->working_dir_fd); /* Renew the current working directory. */ t->working_dir_fd = new_fd; t->flags &= ~onWorkingDir; } return (r); } /* * We've finished a directory; ascend back to the parent. */ static int tree_ascend(struct tree *t) { struct tree_entry *te; int new_fd, r = 0, prev_dir_fd; te = t->stack; prev_dir_fd = t->working_dir_fd; if (te->flags & isDirLink) new_fd = te->symlink_parent_fd; else { new_fd = open_on_current_dir(t, "..", O_RDONLY | O_CLOEXEC); __archive_ensure_cloexec_flag(new_fd); } if (new_fd < 0) { t->tree_errno = errno; r = TREE_ERROR_FATAL; } else { /* Renew the current working directory. */ t->working_dir_fd = new_fd; t->flags &= ~onWorkingDir; /* Current directory has been changed, we should * close an fd of previous working directory. */ close_and_restore_time(prev_dir_fd, t, &te->restore_time); if (te->flags & isDirLink) { t->openCount--; te->symlink_parent_fd = -1; } t->depth--; } return (r); } /* * Return to the initial directory where tree_open() was performed. */ static int tree_enter_initial_dir(struct tree *t) { int r = 0; if ((t->flags & onInitialDir) == 0) { r = fchdir(t->initial_dir_fd); if (r == 0) { t->flags &= ~onWorkingDir; t->flags |= onInitialDir; } } return (r); } /* * Restore working directory of directory traversals. */ static int tree_enter_working_dir(struct tree *t) { int r = 0; /* * Change the current directory if really needed. * Sometimes this is unneeded when we did not do * descent. */ if (t->depth > 0 && (t->flags & onWorkingDir) == 0) { r = fchdir(t->working_dir_fd); if (r == 0) { t->flags &= ~onInitialDir; t->flags |= onWorkingDir; } } return (r); } static int tree_current_dir_fd(struct tree *t) { return (t->working_dir_fd); } /* * Pop the working stack. */ static void tree_pop(struct tree *t) { struct tree_entry *te; t->path.s[t->dirname_length] = '\0'; t->path.length = t->dirname_length; if (t->stack == t->current && t->current != NULL) t->current = t->current->parent; te = t->stack; t->stack = te->next; t->dirname_length = te->dirname_length; t->basename = t->path.s + t->dirname_length; while (t->basename[0] == '/') t->basename++; archive_string_free(&te->name); free(te); } /* * Get the next item in the tree traversal. */ static int tree_next(struct tree *t) { int r; while (t->stack != NULL) { /* If there's an open dir, get the next entry from there. */ if (t->d != INVALID_DIR_HANDLE) { r = tree_dir_next_posix(t); if (r == 0) continue; return (r); } if (t->stack->flags & needsFirstVisit) { /* Top stack item needs a regular visit. */ t->current = t->stack; tree_append(t, t->stack->name.s, archive_strlen(&(t->stack->name))); /* t->dirname_length = t->path_length; */ /* tree_pop(t); */ t->stack->flags &= ~needsFirstVisit; return (t->visit_type = TREE_REGULAR); } else if (t->stack->flags & needsDescent) { /* Top stack item is dir to descend into. */ t->current = t->stack; tree_append(t, t->stack->name.s, archive_strlen(&(t->stack->name))); t->stack->flags &= ~needsDescent; r = tree_descent(t); if (r != 0) { tree_pop(t); t->visit_type = r; } else t->visit_type = TREE_POSTDESCENT; return (t->visit_type); } else if (t->stack->flags & needsOpen) { t->stack->flags &= ~needsOpen; r = tree_dir_next_posix(t); if (r == 0) continue; return (r); } else if (t->stack->flags & needsAscent) { /* Top stack item is dir and we're done with it. */ r = tree_ascend(t); tree_pop(t); t->visit_type = r != 0 ? r : TREE_POSTASCENT; return (t->visit_type); } else { /* Top item on stack is dead. */ tree_pop(t); t->flags &= ~hasLstat; t->flags &= ~hasStat; } } return (t->visit_type = 0); } static int tree_dir_next_posix(struct tree *t) { int r; const char *name; size_t namelen; if (t->d == NULL) { #if defined(HAVE_READDIR_R) size_t dirent_size; #endif #if defined(HAVE_FDOPENDIR) t->d = fdopendir(tree_dup(t->working_dir_fd)); #else /* HAVE_FDOPENDIR */ if (tree_enter_working_dir(t) == 0) { t->d = opendir("."); #if HAVE_DIRFD || defined(dirfd) __archive_ensure_cloexec_flag(dirfd(t->d)); #endif } #endif /* HAVE_FDOPENDIR */ if (t->d == NULL) { r = tree_ascend(t); /* Undo "chdir" */ tree_pop(t); t->tree_errno = errno; t->visit_type = r != 0 ? r : TREE_ERROR_DIR; return (t->visit_type); } #if defined(HAVE_READDIR_R) dirent_size = offsetof(struct dirent, d_name) + t->filesystem_table[t->current->filesystem_id].name_max + 1; if (t->dirent == NULL || t->dirent_allocated < dirent_size) { free(t->dirent); t->dirent = malloc(dirent_size); if (t->dirent == NULL) { closedir(t->d); t->d = INVALID_DIR_HANDLE; (void)tree_ascend(t); tree_pop(t); t->tree_errno = ENOMEM; t->visit_type = TREE_ERROR_DIR; return (t->visit_type); } t->dirent_allocated = dirent_size; } #endif /* HAVE_READDIR_R */ } for (;;) { errno = 0; #if defined(HAVE_READDIR_R) r = readdir_r(t->d, t->dirent, &t->de); #ifdef _AIX /* Note: According to the man page, return value 9 indicates * that the readdir_r was not successful and the error code * is set to the global errno variable. And then if the end * of directory entries was reached, the return value is 9 * and the third parameter is set to NULL and errno is * unchanged. */ if (r == 9) r = errno; #endif /* _AIX */ if (r != 0 || t->de == NULL) { #else t->de = readdir(t->d); if (t->de == NULL) { r = errno; #endif closedir(t->d); t->d = INVALID_DIR_HANDLE; if (r != 0) { t->tree_errno = r; t->visit_type = TREE_ERROR_DIR; return (t->visit_type); } else return (0); } name = t->de->d_name; namelen = D_NAMELEN(t->de); t->flags &= ~hasLstat; t->flags &= ~hasStat; if (name[0] == '.' && name[1] == '\0') continue; if (name[0] == '.' && name[1] == '.' && name[2] == '\0') continue; tree_append(t, name, namelen); return (t->visit_type = TREE_REGULAR); } } /* * Get the stat() data for the entry just returned from tree_next(). */ static const struct stat * tree_current_stat(struct tree *t) { if (!(t->flags & hasStat)) { #ifdef HAVE_FSTATAT if (fstatat(tree_current_dir_fd(t), tree_current_access_path(t), &t->st, 0) != 0) #else if (tree_enter_working_dir(t) != 0) return NULL; if (stat(tree_current_access_path(t), &t->st) != 0) #endif return NULL; t->flags |= hasStat; } return (&t->st); } /* * Get the lstat() data for the entry just returned from tree_next(). */ static const struct stat * tree_current_lstat(struct tree *t) { if (!(t->flags & hasLstat)) { #ifdef HAVE_FSTATAT if (fstatat(tree_current_dir_fd(t), tree_current_access_path(t), &t->lst, AT_SYMLINK_NOFOLLOW) != 0) #else if (tree_enter_working_dir(t) != 0) return NULL; if (lstat(tree_current_access_path(t), &t->lst) != 0) #endif return NULL; t->flags |= hasLstat; } return (&t->lst); } /* * Test whether current entry is a dir or link to a dir. */ static int tree_current_is_dir(struct tree *t) { const struct stat *st; /* * If we already have lstat() info, then try some * cheap tests to determine if this is a dir. */ if (t->flags & hasLstat) { /* If lstat() says it's a dir, it must be a dir. */ st = tree_current_lstat(t); if (st == NULL) return 0; if (S_ISDIR(st->st_mode)) return 1; /* Not a dir; might be a link to a dir. */ /* If it's not a link, then it's not a link to a dir. */ if (!S_ISLNK(st->st_mode)) return 0; /* * It's a link, but we don't know what it's a link to, * so we'll have to use stat(). */ } st = tree_current_stat(t); /* If we can't stat it, it's not a dir. */ if (st == NULL) return 0; /* Use the definitive test. Hopefully this is cached. */ return (S_ISDIR(st->st_mode)); } /* * Test whether current entry is a physical directory. Usually, we * already have at least one of stat() or lstat() in memory, so we * use tricks to try to avoid an extra trip to the disk. */ static int tree_current_is_physical_dir(struct tree *t) { const struct stat *st; /* * If stat() says it isn't a dir, then it's not a dir. * If stat() data is cached, this check is free, so do it first. */ if (t->flags & hasStat) { st = tree_current_stat(t); if (st == NULL) return (0); if (!S_ISDIR(st->st_mode)) return (0); } /* * Either stat() said it was a dir (in which case, we have * to determine whether it's really a link to a dir) or * stat() info wasn't available. So we use lstat(), which * hopefully is already cached. */ st = tree_current_lstat(t); /* If we can't stat it, it's not a dir. */ if (st == NULL) return 0; /* Use the definitive test. Hopefully this is cached. */ return (S_ISDIR(st->st_mode)); } /* * Test whether the same file has been in the tree as its parent. */ static int tree_target_is_same_as_parent(struct tree *t, const struct stat *st) { struct tree_entry *te; for (te = t->current->parent; te != NULL; te = te->parent) { if (te->dev == (int64_t)st->st_dev && te->ino == (int64_t)st->st_ino) return (1); } return (0); } /* * Test whether the current file is symbolic link target and * on the other filesystem. */ static int tree_current_is_symblic_link_target(struct tree *t) { static const struct stat *lst, *st; lst = tree_current_lstat(t); st = tree_current_stat(t); return (st != NULL && lst != NULL && (int64_t)st->st_dev == t->current_filesystem->dev && st->st_dev != lst->st_dev); } /* * Return the access path for the entry just returned from tree_next(). */ static const char * tree_current_access_path(struct tree *t) { return (t->basename); } /* * Return the full path for the entry just returned from tree_next(). */ static const char * tree_current_path(struct tree *t) { return (t->path.s); } /* * Terminate the traversal. */ static void tree_close(struct tree *t) { if (t == NULL) return; if (t->entry_fd >= 0) { close_and_restore_time(t->entry_fd, t, &t->restore_time); t->entry_fd = -1; } /* Close the handle of readdir(). */ if (t->d != INVALID_DIR_HANDLE) { closedir(t->d); t->d = INVALID_DIR_HANDLE; } /* Release anything remaining in the stack. */ while (t->stack != NULL) { if (t->stack->flags & isDirLink) close(t->stack->symlink_parent_fd); tree_pop(t); } if (t->working_dir_fd >= 0) { close(t->working_dir_fd); t->working_dir_fd = -1; } if (t->initial_dir_fd >= 0) { close(t->initial_dir_fd); t->initial_dir_fd = -1; } } /* * Release any resources. */ static void tree_free(struct tree *t) { int i; if (t == NULL) return; archive_string_free(&t->path); #if defined(HAVE_READDIR_R) free(t->dirent); #endif free(t->sparse_list); for (i = 0; i < t->max_filesystem_id; i++) free(t->filesystem_table[i].allocation_ptr); free(t->filesystem_table); free(t); } #endif Index: stable/10/contrib/libarchive/libarchive/archive_string.c =================================================================== --- stable/10/contrib/libarchive/libarchive/archive_string.c (revision 302424) +++ stable/10/contrib/libarchive/libarchive/archive_string.c (revision 302425) @@ -1,4195 +1,4198 @@ /*- * Copyright (c) 2003-2011 Tim Kientzle * Copyright (c) 2011-2012 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "archive_platform.h" __FBSDID("$FreeBSD$"); /* * Basic resizable string support, to simplify manipulating arbitrary-sized * strings while minimizing heap activity. * * In particular, the buffer used by a string object is only grown, it * never shrinks, so you can clear and reuse the same string object * without incurring additional memory allocations. */ #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_ICONV_H #include #endif #ifdef HAVE_LANGINFO_H #include #endif #ifdef HAVE_LOCALCHARSET_H #include #endif #ifdef HAVE_STDLIB_H #include #endif #ifdef HAVE_STRING_H #include #endif #ifdef HAVE_WCHAR_H #include #endif #if defined(_WIN32) && !defined(__CYGWIN__) #include #include #endif #include "archive_endian.h" #include "archive_private.h" #include "archive_string.h" #include "archive_string_composition.h" #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy) #define wmemcpy(a,b,i) (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t)) #endif #if !defined(HAVE_WMEMMOVE) && !defined(wmemmove) #define wmemmove(a,b,i) (wchar_t *)memmove((a), (b), (i) * sizeof(wchar_t)) #endif struct archive_string_conv { struct archive_string_conv *next; char *from_charset; char *to_charset; unsigned from_cp; unsigned to_cp; /* Set 1 if from_charset and to_charset are the same. */ int same; int flag; #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified * charset. */ #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from * specified charset. */ #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */ #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting * MBS. */ #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive * 2.x in the wrong assumption. */ #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C. * Before UTF-8 characters are actually * processed. */ #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D. * Before UTF-8 characters are actually * processed. * Currently this only for MAC OS X. */ #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */ #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */ #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */ #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */ #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */ #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */ #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE) #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE) #if HAVE_ICONV iconv_t cd; iconv_t cd_w;/* Use at archive_mstring on * Windows. */ #endif /* A temporary buffer for normalization. */ struct archive_string utftmp; int (*converter[2])(struct archive_string *, const void *, size_t, struct archive_string_conv *); int nconverter; }; #define CP_C_LOCALE 0 /* "C" locale only for this file. */ #define CP_UTF16LE 1200 #define CP_UTF16BE 1201 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF) #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF) #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF) #define UNICODE_MAX 0x10FFFF #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */ /* Set U+FFFD(Replacement character) in UTF-8. */ static const char utf8_replacement_char[] = {0xef, 0xbf, 0xbd}; static struct archive_string_conv *find_sconv_object(struct archive *, const char *, const char *); static void add_sconv_object(struct archive *, struct archive_string_conv *); static struct archive_string_conv *create_sconv_object(const char *, const char *, unsigned, int); static void free_sconv_object(struct archive_string_conv *); static struct archive_string_conv *get_sconv_object(struct archive *, const char *, const char *, int); static unsigned make_codepage_from_charset(const char *); static unsigned get_current_codepage(void); static unsigned get_current_oemcp(void); static size_t mbsnbytes(const void *, size_t); static size_t utf16nbytes(const void *, size_t); #if defined(_WIN32) && !defined(__CYGWIN__) static int archive_wstring_append_from_mbs_in_codepage( struct archive_wstring *, const char *, size_t, struct archive_string_conv *); static int archive_string_append_from_wcs_in_codepage(struct archive_string *, const wchar_t *, size_t, struct archive_string_conv *); static int is_big_endian(void); static int strncat_in_codepage(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int win_strncat_from_utf16be(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int win_strncat_from_utf16le(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int win_strncat_to_utf16be(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int win_strncat_to_utf16le(struct archive_string *, const void *, size_t, struct archive_string_conv *); #endif static int best_effort_strncat_from_utf16be(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int best_effort_strncat_from_utf16le(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int best_effort_strncat_to_utf16be(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int best_effort_strncat_to_utf16le(struct archive_string *, const void *, size_t, struct archive_string_conv *); #if defined(HAVE_ICONV) static int iconv_strncat_in_locale(struct archive_string *, const void *, size_t, struct archive_string_conv *); #endif static int best_effort_strncat_in_locale(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int _utf8_to_unicode(uint32_t *, const char *, size_t); static int utf8_to_unicode(uint32_t *, const char *, size_t); static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t); static int cesu8_to_unicode(uint32_t *, const char *, size_t); static size_t unicode_to_utf8(char *, size_t, uint32_t); static int utf16_to_unicode(uint32_t *, const char *, size_t, int); static size_t unicode_to_utf16be(char *, size_t, uint32_t); static size_t unicode_to_utf16le(char *, size_t, uint32_t); static int strncat_from_utf8_libarchive2(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int strncat_from_utf8_to_utf8(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int archive_string_normalize_C(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int archive_string_normalize_D(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int archive_string_append_unicode(struct archive_string *, const void *, size_t, struct archive_string_conv *); static struct archive_string * archive_string_append(struct archive_string *as, const char *p, size_t s) { if (archive_string_ensure(as, as->length + s + 1) == NULL) return (NULL); memmove(as->s + as->length, p, s); as->length += s; as->s[as->length] = 0; return (as); } static struct archive_wstring * archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s) { if (archive_wstring_ensure(as, as->length + s + 1) == NULL) return (NULL); wmemmove(as->s + as->length, p, s); as->length += s; as->s[as->length] = 0; return (as); } void archive_string_concat(struct archive_string *dest, struct archive_string *src) { if (archive_string_append(dest, src->s, src->length) == NULL) __archive_errx(1, "Out of memory"); } void archive_wstring_concat(struct archive_wstring *dest, struct archive_wstring *src) { if (archive_wstring_append(dest, src->s, src->length) == NULL) __archive_errx(1, "Out of memory"); } void archive_string_free(struct archive_string *as) { as->length = 0; as->buffer_length = 0; free(as->s); as->s = NULL; } void archive_wstring_free(struct archive_wstring *as) { as->length = 0; as->buffer_length = 0; free(as->s); as->s = NULL; } struct archive_wstring * archive_wstring_ensure(struct archive_wstring *as, size_t s) { return (struct archive_wstring *) archive_string_ensure((struct archive_string *)as, s * sizeof(wchar_t)); } /* Returns NULL on any allocation failure. */ struct archive_string * archive_string_ensure(struct archive_string *as, size_t s) { char *p; size_t new_length; /* If buffer is already big enough, don't reallocate. */ if (as->s && (s <= as->buffer_length)) return (as); /* * Growing the buffer at least exponentially ensures that * append operations are always linear in the number of * characters appended. Using a smaller growth rate for * larger buffers reduces memory waste somewhat at the cost of * a larger constant factor. */ if (as->buffer_length < 32) /* Start with a minimum 32-character buffer. */ new_length = 32; else if (as->buffer_length < 8192) /* Buffers under 8k are doubled for speed. */ new_length = as->buffer_length + as->buffer_length; else { /* Buffers 8k and over grow by at least 25% each time. */ new_length = as->buffer_length + as->buffer_length / 4; /* Be safe: If size wraps, fail. */ if (new_length < as->buffer_length) { /* On failure, wipe the string and return NULL. */ archive_string_free(as); errno = ENOMEM;/* Make sure errno has ENOMEM. */ return (NULL); } } /* * The computation above is a lower limit to how much we'll * grow the buffer. In any case, we have to grow it enough to * hold the request. */ if (new_length < s) new_length = s; /* Now we can reallocate the buffer. */ p = (char *)realloc(as->s, new_length); if (p == NULL) { /* On failure, wipe the string and return NULL. */ archive_string_free(as); errno = ENOMEM;/* Make sure errno has ENOMEM. */ return (NULL); } as->s = p; as->buffer_length = new_length; return (as); } /* * TODO: See if there's a way to avoid scanning * the source string twice. Then test to see * if it actually helps (remember that we're almost * always called with pretty short arguments, so * such an optimization might not help). */ struct archive_string * archive_strncat(struct archive_string *as, const void *_p, size_t n) { size_t s; const char *p, *pp; p = (const char *)_p; /* Like strlen(p), except won't examine positions beyond p[n]. */ s = 0; pp = p; while (s < n && *pp) { pp++; s++; } if ((as = archive_string_append(as, p, s)) == NULL) __archive_errx(1, "Out of memory"); return (as); } struct archive_wstring * archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n) { size_t s; const wchar_t *pp; /* Like strlen(p), except won't examine positions beyond p[n]. */ s = 0; pp = p; while (s < n && *pp) { pp++; s++; } if ((as = archive_wstring_append(as, p, s)) == NULL) __archive_errx(1, "Out of memory"); return (as); } struct archive_string * archive_strcat(struct archive_string *as, const void *p) { /* strcat is just strncat without an effective limit. * Assert that we'll never get called with a source * string over 16MB. * TODO: Review all uses of strcat in the source * and try to replace them with strncat(). */ return archive_strncat(as, p, 0x1000000); } struct archive_wstring * archive_wstrcat(struct archive_wstring *as, const wchar_t *p) { /* Ditto. */ return archive_wstrncat(as, p, 0x1000000); } struct archive_string * archive_strappend_char(struct archive_string *as, char c) { if ((as = archive_string_append(as, &c, 1)) == NULL) __archive_errx(1, "Out of memory"); return (as); } struct archive_wstring * archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c) { if ((as = archive_wstring_append(as, &c, 1)) == NULL) __archive_errx(1, "Out of memory"); return (as); } /* * Get the "current character set" name to use with iconv. * On FreeBSD, the empty character set name "" chooses * the correct character encoding for the current locale, * so this isn't necessary. * But iconv on Mac OS 10.6 doesn't seem to handle this correctly; * on that system, we have to explicitly call nl_langinfo() * to get the right name. Not sure about other platforms. * * NOTE: GNU libiconv does not recognize the character-set name * which some platform nl_langinfo(CODESET) returns, so we should * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv. */ static const char * default_iconv_charset(const char *charset) { if (charset != NULL && charset[0] != '\0') return charset; #if HAVE_LOCALE_CHARSET && !defined(__APPLE__) /* locale_charset() is broken on Mac OS */ return locale_charset(); #elif HAVE_NL_LANGINFO return nl_langinfo(CODESET); #else return ""; #endif } #if defined(_WIN32) && !defined(__CYGWIN__) /* * Convert MBS to WCS. * Note: returns -1 if conversion fails. */ int archive_wstring_append_from_mbs(struct archive_wstring *dest, const char *p, size_t len) { return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL); } static int archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest, const char *s, size_t length, struct archive_string_conv *sc) { int count, ret = 0; UINT from_cp; if (sc != NULL) from_cp = sc->from_cp; else from_cp = get_current_codepage(); if (from_cp == CP_C_LOCALE) { /* * "C" locale special process. */ wchar_t *ws; const unsigned char *mp; if (NULL == archive_wstring_ensure(dest, dest->length + length + 1)) return (-1); ws = dest->s + dest->length; mp = (const unsigned char *)s; count = 0; while (count < (int)length && *mp) { *ws++ = (wchar_t)*mp++; count++; } } else if (sc != NULL && (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) { /* * Normalize UTF-8 and UTF-16BE and convert it directly * to UTF-16 as wchar_t. */ struct archive_string u16; int saved_flag = sc->flag;/* save current flag. */ if (is_big_endian()) sc->flag |= SCONV_TO_UTF16BE; else sc->flag |= SCONV_TO_UTF16LE; if (sc->flag & SCONV_FROM_UTF16) { /* * UTF-16BE/LE NFD ===> UTF-16 NFC * UTF-16BE/LE NFC ===> UTF-16 NFD */ count = (int)utf16nbytes(s, length); } else { /* * UTF-8 NFD ===> UTF-16 NFC * UTF-8 NFC ===> UTF-16 NFD */ count = (int)mbsnbytes(s, length); } u16.s = (char *)dest->s; u16.length = dest->length << 1;; u16.buffer_length = dest->buffer_length; if (sc->flag & SCONV_NORMALIZATION_C) ret = archive_string_normalize_C(&u16, s, count, sc); else ret = archive_string_normalize_D(&u16, s, count, sc); dest->s = (wchar_t *)u16.s; dest->length = u16.length >> 1; dest->buffer_length = u16.buffer_length; sc->flag = saved_flag;/* restore the saved flag. */ return (ret); } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) { count = (int)utf16nbytes(s, length); count >>= 1; /* to be WCS length */ /* Allocate memory for WCS. */ if (NULL == archive_wstring_ensure(dest, dest->length + count + 1)) return (-1); wmemcpy(dest->s + dest->length, (const wchar_t *)s, count); if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) { uint16_t *u16 = (uint16_t *)(dest->s + dest->length); int b; for (b = 0; b < count; b++) { uint16_t val = archive_le16dec(u16+b); archive_be16enc(u16+b, val); } } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) { uint16_t *u16 = (uint16_t *)(dest->s + dest->length); int b; for (b = 0; b < count; b++) { uint16_t val = archive_be16dec(u16+b); archive_le16enc(u16+b, val); } } } else { DWORD mbflag; size_t buffsize; if (sc == NULL) mbflag = 0; else if (sc->flag & SCONV_FROM_CHARSET) { /* Do not trust the length which comes from * an archive file. */ length = mbsnbytes(s, length); mbflag = 0; } else mbflag = MB_PRECOMPOSED; buffsize = dest->length + length + 1; do { /* Allocate memory for WCS. */ if (NULL == archive_wstring_ensure(dest, buffsize)) return (-1); /* Convert MBS to WCS. */ count = MultiByteToWideChar(from_cp, mbflag, s, (int)length, dest->s + dest->length, (int)(dest->buffer_length >> 1) -1); if (count == 0 && GetLastError() == ERROR_INSUFFICIENT_BUFFER) { /* Expand the WCS buffer. */ buffsize = dest->buffer_length << 1; continue; } if (count == 0 && length != 0) ret = -1; - } while (0); + break; + } while (1); } dest->length += count; dest->s[dest->length] = L'\0'; return (ret); } #else /* * Convert MBS to WCS. * Note: returns -1 if conversion fails. */ int archive_wstring_append_from_mbs(struct archive_wstring *dest, const char *p, size_t len) { size_t r; int ret_val = 0; /* * No single byte will be more than one wide character, * so this length estimate will always be big enough. */ size_t wcs_length = len; size_t mbs_length = len; const char *mbs = p; wchar_t *wcs; #if HAVE_MBRTOWC mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); #endif if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1)) return (-1); wcs = dest->s + dest->length; /* * We cannot use mbsrtowcs/mbstowcs here because those may convert * extra MBS when strlen(p) > len and one wide character consis of * multi bytes. */ while (*mbs && mbs_length > 0) { if (wcs_length == 0) { dest->length = wcs - dest->s; dest->s[dest->length] = L'\0'; wcs_length = mbs_length; if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1)) return (-1); wcs = dest->s + dest->length; } #if HAVE_MBRTOWC r = mbrtowc(wcs, mbs, wcs_length, &shift_state); #else r = mbtowc(wcs, mbs, wcs_length); #endif if (r == (size_t)-1 || r == (size_t)-2) { ret_val = -1; if (errno == EILSEQ) { ++mbs; --mbs_length; continue; } else break; } if (r == 0 || r > mbs_length) break; wcs++; wcs_length--; mbs += r; mbs_length -= r; } dest->length = wcs - dest->s; dest->s[dest->length] = L'\0'; return (ret_val); } #endif #if defined(_WIN32) && !defined(__CYGWIN__) /* * WCS ==> MBS. * Note: returns -1 if conversion fails. * * Win32 builds use WideCharToMultiByte from the Windows API. * (Maybe Cygwin should too? WideCharToMultiByte will know a * lot more about local character encodings than the wcrtomb() * wrapper is going to know.) */ int archive_string_append_from_wcs(struct archive_string *as, const wchar_t *w, size_t len) { return archive_string_append_from_wcs_in_codepage(as, w, len, NULL); } static int archive_string_append_from_wcs_in_codepage(struct archive_string *as, const wchar_t *ws, size_t len, struct archive_string_conv *sc) { BOOL defchar_used, *dp; int count, ret = 0; UINT to_cp; int wslen = (int)len; if (sc != NULL) to_cp = sc->to_cp; else to_cp = get_current_codepage(); if (to_cp == CP_C_LOCALE) { /* * "C" locale special process. */ const wchar_t *wp = ws; char *p; if (NULL == archive_string_ensure(as, as->length + wslen +1)) return (-1); p = as->s + as->length; count = 0; defchar_used = 0; while (count < wslen && *wp) { if (*wp > 255) { *p++ = '?'; wp++; defchar_used = 1; } else *p++ = (char)*wp++; count++; } } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) { uint16_t *u16; if (NULL == archive_string_ensure(as, as->length + len * 2 + 2)) return (-1); u16 = (uint16_t *)(as->s + as->length); count = 0; defchar_used = 0; if (sc->flag & SCONV_TO_UTF16BE) { while (count < (int)len && *ws) { archive_be16enc(u16+count, *ws); ws++; count++; } } else { while (count < (int)len && *ws) { archive_le16enc(u16+count, *ws); ws++; count++; } } count <<= 1; /* to be byte size */ } else { /* Make sure the MBS buffer has plenty to set. */ if (NULL == archive_string_ensure(as, as->length + len * 2 + 1)) return (-1); do { defchar_used = 0; if (to_cp == CP_UTF8 || sc == NULL) dp = NULL; else dp = &defchar_used; count = WideCharToMultiByte(to_cp, 0, ws, wslen, as->s + as->length, (int)as->buffer_length-1, NULL, dp); if (count == 0 && GetLastError() == ERROR_INSUFFICIENT_BUFFER) { /* Expand the MBS buffer and retry. */ if (NULL == archive_string_ensure(as, as->buffer_length + len)) return (-1); continue; } if (count == 0) ret = -1; break; } while (1); } as->length += count; as->s[as->length] = '\0'; return (defchar_used?-1:ret); } #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB) /* * Translates a wide character string into current locale character set * and appends to the archive_string. Note: returns -1 if conversion * fails. */ int archive_string_append_from_wcs(struct archive_string *as, const wchar_t *w, size_t len) { /* We cannot use the standard wcstombs() here because it * cannot tell us how big the output buffer should be. So * I've built a loop around wcrtomb() or wctomb() that * converts a character at a time and resizes the string as * needed. We prefer wcrtomb() when it's available because * it's thread-safe. */ int n, ret_val = 0; char *p; char *end; #if HAVE_WCRTOMB mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); #else /* Clear the shift state before starting. */ wctomb(NULL, L'\0'); #endif /* * Allocate buffer for MBS. * We need this allocation here since it is possible that * as->s is still NULL. */ if (archive_string_ensure(as, as->length + len + 1) == NULL) return (-1); p = as->s + as->length; end = as->s + as->buffer_length - MB_CUR_MAX -1; while (*w != L'\0' && len > 0) { if (p >= end) { as->length = p - as->s; as->s[as->length] = '\0'; /* Re-allocate buffer for MBS. */ if (archive_string_ensure(as, as->length + len * 2 + 1) == NULL) return (-1); p = as->s + as->length; end = as->s + as->buffer_length - MB_CUR_MAX -1; } #if HAVE_WCRTOMB n = wcrtomb(p, *w++, &shift_state); #else n = wctomb(p, *w++); #endif if (n == -1) { if (errno == EILSEQ) { /* Skip an illegal wide char. */ *p++ = '?'; ret_val = -1; } else { ret_val = -1; break; } } else p += n; len--; } as->length = p - as->s; as->s[as->length] = '\0'; return (ret_val); } #else /* HAVE_WCTOMB || HAVE_WCRTOMB */ /* * TODO: Test if __STDC_ISO_10646__ is defined. * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion * one character at a time. If a non-Windows platform doesn't have * either of these, fall back to the built-in UTF8 conversion. */ int archive_string_append_from_wcs(struct archive_string *as, const wchar_t *w, size_t len) { (void)as;/* UNUSED */ (void)w;/* UNUSED */ (void)len;/* UNUSED */ errno = ENOSYS; return (-1); } #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */ /* * Find a string conversion object by a pair of 'from' charset name * and 'to' charset name from an archive object. * Return NULL if not found. */ static struct archive_string_conv * find_sconv_object(struct archive *a, const char *fc, const char *tc) { struct archive_string_conv *sc; if (a == NULL) return (NULL); for (sc = a->sconv; sc != NULL; sc = sc->next) { if (strcmp(sc->from_charset, fc) == 0 && strcmp(sc->to_charset, tc) == 0) break; } return (sc); } /* * Register a string object to an archive object. */ static void add_sconv_object(struct archive *a, struct archive_string_conv *sc) { struct archive_string_conv **psc; /* Add a new sconv to sconv list. */ psc = &(a->sconv); while (*psc != NULL) psc = &((*psc)->next); *psc = sc; } static void add_converter(struct archive_string_conv *sc, int (*converter) (struct archive_string *, const void *, size_t, struct archive_string_conv *)) { if (sc == NULL || sc->nconverter >= 2) __archive_errx(1, "Programing error"); sc->converter[sc->nconverter++] = converter; } static void setup_converter(struct archive_string_conv *sc) { /* Reset. */ sc->nconverter = 0; /* * Perform special sequence for the incorrect UTF-8 filenames * made by libarchive2.x. */ if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) { add_converter(sc, strncat_from_utf8_libarchive2); return; } /* * Convert a string to UTF-16BE/LE. */ if (sc->flag & SCONV_TO_UTF16) { /* * If the current locale is UTF-8, we can translate * a UTF-8 string into a UTF-16BE string. */ if (sc->flag & SCONV_FROM_UTF8) { add_converter(sc, archive_string_append_unicode); return; } #if defined(_WIN32) && !defined(__CYGWIN__) if (sc->flag & SCONV_WIN_CP) { if (sc->flag & SCONV_TO_UTF16BE) add_converter(sc, win_strncat_to_utf16be); else add_converter(sc, win_strncat_to_utf16le); return; } #endif #if defined(HAVE_ICONV) if (sc->cd != (iconv_t)-1) { add_converter(sc, iconv_strncat_in_locale); return; } #endif if (sc->flag & SCONV_BEST_EFFORT) { if (sc->flag & SCONV_TO_UTF16BE) add_converter(sc, best_effort_strncat_to_utf16be); else add_converter(sc, best_effort_strncat_to_utf16le); } else /* Make sure we have no converter. */ sc->nconverter = 0; return; } /* * Convert a string from UTF-16BE/LE. */ if (sc->flag & SCONV_FROM_UTF16) { /* * At least we should normalize a UTF-16BE string. */ if (sc->flag & SCONV_NORMALIZATION_D) add_converter(sc,archive_string_normalize_D); else if (sc->flag & SCONV_NORMALIZATION_C) add_converter(sc, archive_string_normalize_C); if (sc->flag & SCONV_TO_UTF8) { /* * If the current locale is UTF-8, we can translate * a UTF-16BE/LE string into a UTF-8 string directly. */ if (!(sc->flag & (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) add_converter(sc, archive_string_append_unicode); return; } #if defined(_WIN32) && !defined(__CYGWIN__) if (sc->flag & SCONV_WIN_CP) { if (sc->flag & SCONV_FROM_UTF16BE) add_converter(sc, win_strncat_from_utf16be); else add_converter(sc, win_strncat_from_utf16le); return; } #endif #if defined(HAVE_ICONV) if (sc->cd != (iconv_t)-1) { add_converter(sc, iconv_strncat_in_locale); return; } #endif if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) add_converter(sc, best_effort_strncat_from_utf16be); else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) add_converter(sc, best_effort_strncat_from_utf16le); else /* Make sure we have no converter. */ sc->nconverter = 0; return; } if (sc->flag & SCONV_FROM_UTF8) { /* * At least we should normalize a UTF-8 string. */ if (sc->flag & SCONV_NORMALIZATION_D) add_converter(sc,archive_string_normalize_D); else if (sc->flag & SCONV_NORMALIZATION_C) add_converter(sc, archive_string_normalize_C); /* * Copy UTF-8 string with a check of CESU-8. * Apparently, iconv does not check surrogate pairs in UTF-8 * when both from-charset and to-charset are UTF-8, and then * we use our UTF-8 copy code. */ if (sc->flag & SCONV_TO_UTF8) { /* * If the current locale is UTF-8, we can translate * a UTF-16BE string into a UTF-8 string directly. */ if (!(sc->flag & (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) add_converter(sc, strncat_from_utf8_to_utf8); return; } } #if defined(_WIN32) && !defined(__CYGWIN__) /* * On Windows we can use Windows API for a string conversion. */ if (sc->flag & SCONV_WIN_CP) { add_converter(sc, strncat_in_codepage); return; } #endif #if HAVE_ICONV if (sc->cd != (iconv_t)-1) { add_converter(sc, iconv_strncat_in_locale); /* * iconv generally does not support UTF-8-MAC and so * we have to the output of iconv from NFC to NFD if * need. */ if ((sc->flag & SCONV_FROM_CHARSET) && (sc->flag & SCONV_TO_UTF8)) { if (sc->flag & SCONV_NORMALIZATION_D) add_converter(sc, archive_string_normalize_D); } return; } #endif /* * Try conversion in the best effort or no conversion. */ if ((sc->flag & SCONV_BEST_EFFORT) || sc->same) add_converter(sc, best_effort_strncat_in_locale); else /* Make sure we have no converter. */ sc->nconverter = 0; } /* * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE * and CP932 which are referenced in create_sconv_object(). */ static const char * canonical_charset_name(const char *charset) { char cs[16]; char *p; const char *s; if (charset == NULL || charset[0] == '\0' || strlen(charset) > 15) return (charset); /* Copy name to uppercase. */ p = cs; s = charset; while (*s) { char c = *s++; if (c >= 'a' && c <= 'z') c -= 'a' - 'A'; *p++ = c; } *p++ = '\0'; if (strcmp(cs, "UTF-8") == 0 || strcmp(cs, "UTF8") == 0) return ("UTF-8"); if (strcmp(cs, "UTF-16BE") == 0 || strcmp(cs, "UTF16BE") == 0) return ("UTF-16BE"); if (strcmp(cs, "UTF-16LE") == 0 || strcmp(cs, "UTF16LE") == 0) return ("UTF-16LE"); if (strcmp(cs, "CP932") == 0) return ("CP932"); return (charset); } /* * Create a string conversion object. */ static struct archive_string_conv * create_sconv_object(const char *fc, const char *tc, unsigned current_codepage, int flag) { struct archive_string_conv *sc; sc = calloc(1, sizeof(*sc)); if (sc == NULL) return (NULL); sc->next = NULL; sc->from_charset = strdup(fc); if (sc->from_charset == NULL) { free(sc); return (NULL); } sc->to_charset = strdup(tc); if (sc->to_charset == NULL) { free(sc->from_charset); free(sc); return (NULL); } archive_string_init(&sc->utftmp); if (flag & SCONV_TO_CHARSET) { /* * Convert characters from the current locale charset to * a specified charset. */ sc->from_cp = current_codepage; sc->to_cp = make_codepage_from_charset(tc); #if defined(_WIN32) && !defined(__CYGWIN__) if (IsValidCodePage(sc->to_cp)) flag |= SCONV_WIN_CP; #endif } else if (flag & SCONV_FROM_CHARSET) { /* * Convert characters from a specified charset to * the current locale charset. */ sc->to_cp = current_codepage; sc->from_cp = make_codepage_from_charset(fc); #if defined(_WIN32) && !defined(__CYGWIN__) if (IsValidCodePage(sc->from_cp)) flag |= SCONV_WIN_CP; #endif } /* * Check if "from charset" and "to charset" are the same. */ if (strcmp(fc, tc) == 0 || (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp)) sc->same = 1; else sc->same = 0; /* * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE. */ if (strcmp(tc, "UTF-8") == 0) flag |= SCONV_TO_UTF8; else if (strcmp(tc, "UTF-16BE") == 0) flag |= SCONV_TO_UTF16BE; else if (strcmp(tc, "UTF-16LE") == 0) flag |= SCONV_TO_UTF16LE; if (strcmp(fc, "UTF-8") == 0) flag |= SCONV_FROM_UTF8; else if (strcmp(fc, "UTF-16BE") == 0) flag |= SCONV_FROM_UTF16BE; else if (strcmp(fc, "UTF-16LE") == 0) flag |= SCONV_FROM_UTF16LE; #if defined(_WIN32) && !defined(__CYGWIN__) if (sc->to_cp == CP_UTF8) flag |= SCONV_TO_UTF8; else if (sc->to_cp == CP_UTF16BE) flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP; else if (sc->to_cp == CP_UTF16LE) flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP; if (sc->from_cp == CP_UTF8) flag |= SCONV_FROM_UTF8; else if (sc->from_cp == CP_UTF16BE) flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP; else if (sc->from_cp == CP_UTF16LE) flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP; #endif /* * Set a flag for Unicode NFD. Usually iconv cannot correctly * handle it. So we have to translate NFD characters to NFC ones * ourselves before iconv handles. Another reason is to prevent * that the same sight of two filenames, one is NFC and other * is NFD, would be in its directory. * On Mac OS X, although its filesystem layer automatically * convert filenames to NFD, it would be useful for filename * comparing to find out the same filenames that we normalize * that to be NFD ourselves. */ if ((flag & SCONV_FROM_CHARSET) && (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) { #if defined(__APPLE__) if (flag & SCONV_TO_UTF8) flag |= SCONV_NORMALIZATION_D; else #endif flag |= SCONV_NORMALIZATION_C; } #if defined(__APPLE__) /* * In case writing an archive file, make sure that a filename * going to be passed to iconv is a Unicode NFC string since * a filename in HFS Plus filesystem is a Unicode NFD one and * iconv cannot handle it with "UTF-8" charset. It is simpler * than a use of "UTF-8-MAC" charset. */ if ((flag & SCONV_TO_CHARSET) && (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) flag |= SCONV_NORMALIZATION_C; /* * In case reading an archive file. make sure that a filename * will be passed to users is a Unicode NFD string in order to * correctly compare the filename with other one which comes * from HFS Plus filesystem. */ if ((flag & SCONV_FROM_CHARSET) && !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && (flag & SCONV_TO_UTF8)) flag |= SCONV_NORMALIZATION_D; #endif #if defined(HAVE_ICONV) sc->cd_w = (iconv_t)-1; /* * Create an iconv object. */ if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) && (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) || (flag & SCONV_WIN_CP)) { /* This case we won't use iconv. */ sc->cd = (iconv_t)-1; } else { sc->cd = iconv_open(tc, fc); if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) { /* * Unfortunaly, all of iconv implements do support * "CP932" character-set, so we should use "SJIS" * instead if iconv_open failed. */ if (strcmp(tc, "CP932") == 0) sc->cd = iconv_open("SJIS", fc); else if (strcmp(fc, "CP932") == 0) sc->cd = iconv_open(tc, "SJIS"); } #if defined(_WIN32) && !defined(__CYGWIN__) /* * archive_mstring on Windows directly convert multi-bytes * into archive_wstring in order not to depend on locale * so that you can do a I18N programing. This will be * used only in archive_mstring_copy_mbs_len_l so far. */ if (flag & SCONV_FROM_CHARSET) { sc->cd_w = iconv_open("UTF-8", fc); if (sc->cd_w == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) { if (strcmp(fc, "CP932") == 0) sc->cd_w = iconv_open("UTF-8", "SJIS"); } } #endif /* _WIN32 && !__CYGWIN__ */ } #endif /* HAVE_ICONV */ sc->flag = flag; /* * Set up converters. */ setup_converter(sc); return (sc); } /* * Free a string conversion object. */ static void free_sconv_object(struct archive_string_conv *sc) { free(sc->from_charset); free(sc->to_charset); archive_string_free(&sc->utftmp); #if HAVE_ICONV if (sc->cd != (iconv_t)-1) iconv_close(sc->cd); if (sc->cd_w != (iconv_t)-1) iconv_close(sc->cd_w); #endif free(sc); } #if defined(_WIN32) && !defined(__CYGWIN__) static unsigned my_atoi(const char *p) { unsigned cp; cp = 0; while (*p) { if (*p >= '0' && *p <= '9') cp = cp * 10 + (*p - '0'); else return (-1); p++; } return (cp); } /* * Translate Charset name (as used by iconv) into CodePage (as used by Windows) * Return -1 if failed. * * Note: This translation code may be insufficient. */ static struct charset { const char *name; unsigned cp; } charsets[] = { /* MUST BE SORTED! */ {"ASCII", 1252}, {"ASMO-708", 708}, {"BIG5", 950}, {"CHINESE", 936}, {"CP367", 1252}, {"CP819", 1252}, {"CP1025", 21025}, {"DOS-720", 720}, {"DOS-862", 862}, {"EUC-CN", 51936}, {"EUC-JP", 51932}, {"EUC-KR", 949}, {"EUCCN", 51936}, {"EUCJP", 51932}, {"EUCKR", 949}, {"GB18030", 54936}, {"GB2312", 936}, {"HEBREW", 1255}, {"HZ-GB-2312", 52936}, {"IBM273", 20273}, {"IBM277", 20277}, {"IBM278", 20278}, {"IBM280", 20280}, {"IBM284", 20284}, {"IBM285", 20285}, {"IBM290", 20290}, {"IBM297", 20297}, {"IBM367", 1252}, {"IBM420", 20420}, {"IBM423", 20423}, {"IBM424", 20424}, {"IBM819", 1252}, {"IBM871", 20871}, {"IBM880", 20880}, {"IBM905", 20905}, {"IBM924", 20924}, {"ISO-8859-1", 28591}, {"ISO-8859-13", 28603}, {"ISO-8859-15", 28605}, {"ISO-8859-2", 28592}, {"ISO-8859-3", 28593}, {"ISO-8859-4", 28594}, {"ISO-8859-5", 28595}, {"ISO-8859-6", 28596}, {"ISO-8859-7", 28597}, {"ISO-8859-8", 28598}, {"ISO-8859-9", 28599}, {"ISO8859-1", 28591}, {"ISO8859-13", 28603}, {"ISO8859-15", 28605}, {"ISO8859-2", 28592}, {"ISO8859-3", 28593}, {"ISO8859-4", 28594}, {"ISO8859-5", 28595}, {"ISO8859-6", 28596}, {"ISO8859-7", 28597}, {"ISO8859-8", 28598}, {"ISO8859-9", 28599}, {"JOHAB", 1361}, {"KOI8-R", 20866}, {"KOI8-U", 21866}, {"KS_C_5601-1987", 949}, {"LATIN1", 1252}, {"LATIN2", 28592}, {"MACINTOSH", 10000}, {"SHIFT-JIS", 932}, {"SHIFT_JIS", 932}, {"SJIS", 932}, {"US", 1252}, {"US-ASCII", 1252}, {"UTF-16", 1200}, {"UTF-16BE", 1201}, {"UTF-16LE", 1200}, {"UTF-8", CP_UTF8}, {"X-EUROPA", 29001}, {"X-MAC-ARABIC", 10004}, {"X-MAC-CE", 10029}, {"X-MAC-CHINESEIMP", 10008}, {"X-MAC-CHINESETRAD", 10002}, {"X-MAC-CROATIAN", 10082}, {"X-MAC-CYRILLIC", 10007}, {"X-MAC-GREEK", 10006}, {"X-MAC-HEBREW", 10005}, {"X-MAC-ICELANDIC", 10079}, {"X-MAC-JAPANESE", 10001}, {"X-MAC-KOREAN", 10003}, {"X-MAC-ROMANIAN", 10010}, {"X-MAC-THAI", 10021}, {"X-MAC-TURKISH", 10081}, {"X-MAC-UKRAINIAN", 10017}, }; static unsigned make_codepage_from_charset(const char *charset) { char cs[16]; char *p; unsigned cp; int a, b; if (charset == NULL || strlen(charset) > 15) return -1; /* Copy name to uppercase. */ p = cs; while (*charset) { char c = *charset++; if (c >= 'a' && c <= 'z') c -= 'a' - 'A'; *p++ = c; } *p++ = '\0'; cp = -1; /* Look it up in the table first, so that we can easily * override CP367, which we map to 1252 instead of 367. */ a = 0; b = sizeof(charsets)/sizeof(charsets[0]); while (b > a) { int c = (b + a) / 2; int r = strcmp(charsets[c].name, cs); if (r < 0) a = c + 1; else if (r > 0) b = c; else return charsets[c].cp; } /* If it's not in the table, try to parse it. */ switch (*cs) { case 'C': if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') { cp = my_atoi(cs + 2); } else if (strcmp(cs, "CP_ACP") == 0) cp = get_current_codepage(); else if (strcmp(cs, "CP_OEMCP") == 0) cp = get_current_oemcp(); break; case 'I': if (cs[1] == 'B' && cs[2] == 'M' && cs[3] >= '0' && cs[3] <= '9') { cp = my_atoi(cs + 3); } break; case 'W': if (strncmp(cs, "WINDOWS-", 8) == 0) { cp = my_atoi(cs + 8); if (cp != 874 && (cp < 1250 || cp > 1258)) cp = -1;/* This may invalid code. */ } break; } return (cp); } /* * Return ANSI Code Page of current locale set by setlocale(). */ static unsigned get_current_codepage(void) { char *locale, *p; unsigned cp; locale = setlocale(LC_CTYPE, NULL); if (locale == NULL) return (GetACP()); if (locale[0] == 'C' && locale[1] == '\0') return (CP_C_LOCALE); p = strrchr(locale, '.'); if (p == NULL) return (GetACP()); cp = my_atoi(p+1); if (cp <= 0) return (GetACP()); return (cp); } /* * Translation table between Locale Name and ACP/OEMCP. */ static struct { unsigned acp; unsigned ocp; const char *locale; } acp_ocp_map[] = { { 950, 950, "Chinese_Taiwan" }, { 936, 936, "Chinese_People's Republic of China" }, { 950, 950, "Chinese_Taiwan" }, { 1250, 852, "Czech_Czech Republic" }, { 1252, 850, "Danish_Denmark" }, { 1252, 850, "Dutch_Netherlands" }, { 1252, 850, "Dutch_Belgium" }, { 1252, 437, "English_United States" }, { 1252, 850, "English_Australia" }, { 1252, 850, "English_Canada" }, { 1252, 850, "English_New Zealand" }, { 1252, 850, "English_United Kingdom" }, { 1252, 437, "English_United States" }, { 1252, 850, "Finnish_Finland" }, { 1252, 850, "French_France" }, { 1252, 850, "French_Belgium" }, { 1252, 850, "French_Canada" }, { 1252, 850, "French_Switzerland" }, { 1252, 850, "German_Germany" }, { 1252, 850, "German_Austria" }, { 1252, 850, "German_Switzerland" }, { 1253, 737, "Greek_Greece" }, { 1250, 852, "Hungarian_Hungary" }, { 1252, 850, "Icelandic_Iceland" }, { 1252, 850, "Italian_Italy" }, { 1252, 850, "Italian_Switzerland" }, { 932, 932, "Japanese_Japan" }, { 949, 949, "Korean_Korea" }, { 1252, 850, "Norwegian (BokmOl)_Norway" }, { 1252, 850, "Norwegian (BokmOl)_Norway" }, { 1252, 850, "Norwegian-Nynorsk_Norway" }, { 1250, 852, "Polish_Poland" }, { 1252, 850, "Portuguese_Portugal" }, { 1252, 850, "Portuguese_Brazil" }, { 1251, 866, "Russian_Russia" }, { 1250, 852, "Slovak_Slovakia" }, { 1252, 850, "Spanish_Spain" }, { 1252, 850, "Spanish_Mexico" }, { 1252, 850, "Spanish_Spain" }, { 1252, 850, "Swedish_Sweden" }, { 1254, 857, "Turkish_Turkey" }, { 0, 0, NULL} }; /* * Return OEM Code Page of current locale set by setlocale(). */ static unsigned get_current_oemcp(void) { int i; char *locale, *p; size_t len; locale = setlocale(LC_CTYPE, NULL); if (locale == NULL) return (GetOEMCP()); if (locale[0] == 'C' && locale[1] == '\0') return (CP_C_LOCALE); p = strrchr(locale, '.'); if (p == NULL) return (GetOEMCP()); len = p - locale; for (i = 0; acp_ocp_map[i].acp; i++) { if (strncmp(acp_ocp_map[i].locale, locale, len) == 0) return (acp_ocp_map[i].ocp); } return (GetOEMCP()); } #else /* * POSIX platform does not use CodePage. */ static unsigned get_current_codepage(void) { return (-1);/* Unknown */ } static unsigned make_codepage_from_charset(const char *charset) { (void)charset; /* UNUSED */ return (-1);/* Unknown */ } static unsigned get_current_oemcp(void) { return (-1);/* Unknown */ } #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ /* * Return a string conversion object. */ static struct archive_string_conv * get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag) { struct archive_string_conv *sc; unsigned current_codepage; /* Check if we have made the sconv object. */ sc = find_sconv_object(a, fc, tc); if (sc != NULL) return (sc); if (a == NULL) current_codepage = get_current_codepage(); else current_codepage = a->current_codepage; sc = create_sconv_object(canonical_charset_name(fc), canonical_charset_name(tc), current_codepage, flag); if (sc == NULL) { if (a != NULL) archive_set_error(a, ENOMEM, "Could not allocate memory for " "a string conversion object"); return (NULL); } /* * If there is no converter for current string conversion object, * we cannot handle this conversion. */ if (sc->nconverter == 0) { if (a != NULL) { #if HAVE_ICONV archive_set_error(a, ARCHIVE_ERRNO_MISC, "iconv_open failed : Cannot handle ``%s''", (flag & SCONV_TO_CHARSET)?tc:fc); #else archive_set_error(a, ARCHIVE_ERRNO_MISC, "A character-set conversion not fully supported " "on this platform"); #endif } /* Failed; free a sconv object. */ free_sconv_object(sc); return (NULL); } /* * Success! */ if (a != NULL) add_sconv_object(a, sc); return (sc); } static const char * get_current_charset(struct archive *a) { const char *cur_charset; if (a == NULL) cur_charset = default_iconv_charset(""); else { cur_charset = default_iconv_charset(a->current_code); if (a->current_code == NULL) { a->current_code = strdup(cur_charset); a->current_codepage = get_current_codepage(); a->current_oemcp = get_current_oemcp(); } } return (cur_charset); } /* * Make and Return a string conversion object. * Return NULL if the platform does not support the specified conversion * and best_effort is 0. * If best_effort is set, A string conversion object must be returned * unless memory allocation for the object fails, but the conversion * might fail when non-ASCII code is found. */ struct archive_string_conv * archive_string_conversion_to_charset(struct archive *a, const char *charset, int best_effort) { int flag = SCONV_TO_CHARSET; if (best_effort) flag |= SCONV_BEST_EFFORT; return (get_sconv_object(a, get_current_charset(a), charset, flag)); } struct archive_string_conv * archive_string_conversion_from_charset(struct archive *a, const char *charset, int best_effort) { int flag = SCONV_FROM_CHARSET; if (best_effort) flag |= SCONV_BEST_EFFORT; return (get_sconv_object(a, charset, get_current_charset(a), flag)); } /* * archive_string_default_conversion_*_archive() are provided for Windows * platform because other archiver application use CP_OEMCP for * MultiByteToWideChar() and WideCharToMultiByte() for the filenames * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP). * So we should make a string conversion between CP_ACP and CP_OEMCP * for compatibillty. */ #if defined(_WIN32) && !defined(__CYGWIN__) struct archive_string_conv * archive_string_default_conversion_for_read(struct archive *a) { const char *cur_charset = get_current_charset(a); char oemcp[16]; /* NOTE: a check of cur_charset is unneeded but we need * that get_current_charset() has been surely called at * this time whatever C compiler optimized. */ if (cur_charset != NULL && (a->current_codepage == CP_C_LOCALE || a->current_codepage == a->current_oemcp)) return (NULL);/* no conversion. */ _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); /* Make sure a null termination must be set. */ oemcp[sizeof(oemcp)-1] = '\0'; return (get_sconv_object(a, oemcp, cur_charset, SCONV_FROM_CHARSET)); } struct archive_string_conv * archive_string_default_conversion_for_write(struct archive *a) { const char *cur_charset = get_current_charset(a); char oemcp[16]; /* NOTE: a check of cur_charset is unneeded but we need * that get_current_charset() has been surely called at * this time whatever C compiler optimized. */ if (cur_charset != NULL && (a->current_codepage == CP_C_LOCALE || a->current_codepage == a->current_oemcp)) return (NULL);/* no conversion. */ _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); /* Make sure a null termination must be set. */ oemcp[sizeof(oemcp)-1] = '\0'; return (get_sconv_object(a, cur_charset, oemcp, SCONV_TO_CHARSET)); } #else struct archive_string_conv * archive_string_default_conversion_for_read(struct archive *a) { (void)a; /* UNUSED */ return (NULL); } struct archive_string_conv * archive_string_default_conversion_for_write(struct archive *a) { (void)a; /* UNUSED */ return (NULL); } #endif /* * Dispose of all character conversion objects in the archive object. */ void archive_string_conversion_free(struct archive *a) { struct archive_string_conv *sc; struct archive_string_conv *sc_next; for (sc = a->sconv; sc != NULL; sc = sc_next) { sc_next = sc->next; free_sconv_object(sc); } a->sconv = NULL; free(a->current_code); a->current_code = NULL; } /* * Return a conversion charset name. */ const char * archive_string_conversion_charset_name(struct archive_string_conv *sc) { if (sc->flag & SCONV_TO_CHARSET) return (sc->to_charset); else return (sc->from_charset); } /* * Change the behavior of a string conversion. */ void archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt) { switch (opt) { /* * A filename in UTF-8 was made with libarchive 2.x in a wrong * assumption that wchar_t was Unicode. * This option enables simulating the assumption in order to read * that filname correctly. */ case SCONV_SET_OPT_UTF8_LIBARCHIVE2X: #if (defined(_WIN32) && !defined(__CYGWIN__)) \ || defined(__STDC_ISO_10646__) || defined(__APPLE__) /* * Nothing to do for it since wchar_t on these platforms * is really Unicode. */ (void)sc; /* UNUSED */ #else if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) { sc->flag |= SCONV_UTF8_LIBARCHIVE_2; /* Set up string converters. */ setup_converter(sc); } #endif break; case SCONV_SET_OPT_NORMALIZATION_C: if ((sc->flag & SCONV_NORMALIZATION_C) == 0) { sc->flag |= SCONV_NORMALIZATION_C; sc->flag &= ~SCONV_NORMALIZATION_D; /* Set up string converters. */ setup_converter(sc); } break; case SCONV_SET_OPT_NORMALIZATION_D: #if defined(HAVE_ICONV) /* * If iconv will take the string, do not change the * setting of the normalization. */ if (!(sc->flag & SCONV_WIN_CP) && (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) break; #endif if ((sc->flag & SCONV_NORMALIZATION_D) == 0) { sc->flag |= SCONV_NORMALIZATION_D; sc->flag &= ~SCONV_NORMALIZATION_C; /* Set up string converters. */ setup_converter(sc); } break; default: break; } } /* * * Copy one archive_string to another in locale conversion. * * archive_strncat_l(); * archive_strncpy_l(); * */ static size_t mbsnbytes(const void *_p, size_t n) { size_t s; const char *p, *pp; if (_p == NULL) return (0); p = (const char *)_p; /* Like strlen(p), except won't examine positions beyond p[n]. */ s = 0; pp = p; while (s < n && *pp) { pp++; s++; } return (s); } static size_t utf16nbytes(const void *_p, size_t n) { size_t s; const char *p, *pp; if (_p == NULL) return (0); p = (const char *)_p; /* Like strlen(p), except won't examine positions beyond p[n]. */ s = 0; pp = p; n >>= 1; while (s < n && (pp[0] || pp[1])) { pp += 2; s++; } return (s<<1); } int archive_strncpy_l(struct archive_string *as, const void *_p, size_t n, struct archive_string_conv *sc) { as->length = 0; return (archive_strncat_l(as, _p, n, sc)); } int archive_strncat_l(struct archive_string *as, const void *_p, size_t n, struct archive_string_conv *sc) { const void *s; size_t length; int i, r = 0, r2; /* We must allocate memory even if there is no data for conversion * or copy. This simulates archive_string_append behavior. */ if (_p == NULL || n == 0) { int tn = 1; if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) tn = 2; if (archive_string_ensure(as, as->length + tn) == NULL) return (-1); as->s[as->length] = 0; if (tn == 2) as->s[as->length+1] = 0; return (0); } /* * If sc is NULL, we just make a copy. */ if (sc == NULL) { length = mbsnbytes(_p, n); if (archive_string_append(as, _p, length) == NULL) return (-1);/* No memory */ return (0); } if (sc->flag & SCONV_FROM_UTF16) length = utf16nbytes(_p, n); else length = mbsnbytes(_p, n); s = _p; i = 0; if (sc->nconverter > 1) { sc->utftmp.length = 0; r2 = sc->converter[0](&(sc->utftmp), s, length, sc); if (r2 != 0 && errno == ENOMEM) return (r2); if (r > r2) r = r2; s = sc->utftmp.s; length = sc->utftmp.length; ++i; } r2 = sc->converter[i](as, s, length, sc); if (r > r2) r = r2; return (r); } #if HAVE_ICONV /* * Return -1 if conversion failes. */ static int iconv_strncat_in_locale(struct archive_string *as, const void *_p, size_t length, struct archive_string_conv *sc) { ICONV_CONST char *itp; size_t remaining; iconv_t cd; char *outp; size_t avail, bs; int return_value = 0; /* success */ int to_size, from_size; if (sc->flag & SCONV_TO_UTF16) to_size = 2; else to_size = 1; if (sc->flag & SCONV_FROM_UTF16) from_size = 2; else from_size = 1; if (archive_string_ensure(as, as->length + length*2+to_size) == NULL) return (-1); cd = sc->cd; itp = (char *)(uintptr_t)_p; remaining = length; outp = as->s + as->length; avail = as->buffer_length - as->length - to_size; while (remaining >= (size_t)from_size) { size_t result = iconv(cd, &itp, &remaining, &outp, &avail); if (result != (size_t)-1) break; /* Conversion completed. */ if (errno == EILSEQ || errno == EINVAL) { /* * If an output charset is UTF-8 or UTF-16BE/LE, * unknown character should be U+FFFD * (replacement character). */ if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) { size_t rbytes; if (sc->flag & SCONV_TO_UTF8) rbytes = sizeof(utf8_replacement_char); else rbytes = 2; if (avail < rbytes) { as->length = outp - as->s; bs = as->buffer_length + (remaining * to_size) + rbytes; if (NULL == archive_string_ensure(as, bs)) return (-1); outp = as->s + as->length; avail = as->buffer_length - as->length - to_size; } if (sc->flag & SCONV_TO_UTF8) memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char)); else if (sc->flag & SCONV_TO_UTF16BE) archive_be16enc(outp, UNICODE_R_CHAR); else archive_le16enc(outp, UNICODE_R_CHAR); outp += rbytes; avail -= rbytes; } else { /* Skip the illegal input bytes. */ *outp++ = '?'; avail--; } itp += from_size; remaining -= from_size; return_value = -1; /* failure */ } else { /* E2BIG no output buffer, * Increase an output buffer. */ as->length = outp - as->s; bs = as->buffer_length + remaining * 2; if (NULL == archive_string_ensure(as, bs)) return (-1); outp = as->s + as->length; avail = as->buffer_length - as->length - to_size; } } as->length = outp - as->s; as->s[as->length] = 0; if (to_size == 2) as->s[as->length+1] = 0; return (return_value); } #endif /* HAVE_ICONV */ #if defined(_WIN32) && !defined(__CYGWIN__) /* * Translate a string from a some CodePage to an another CodePage by * Windows APIs, and copy the result. Return -1 if conversion failes. */ static int strncat_in_codepage(struct archive_string *as, const void *_p, size_t length, struct archive_string_conv *sc) { const char *s = (const char *)_p; struct archive_wstring aws; size_t l; int r, saved_flag; archive_string_init(&aws); saved_flag = sc->flag; sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C); r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc); sc->flag = saved_flag; if (r != 0) { archive_wstring_free(&aws); if (errno != ENOMEM) archive_string_append(as, s, length); return (-1); } l = as->length; r = archive_string_append_from_wcs_in_codepage( as, aws.s, aws.length, sc); if (r != 0 && errno != ENOMEM && l == as->length) archive_string_append(as, s, length); archive_wstring_free(&aws); return (r); } /* * Test whether MBS ==> WCS is okay. */ static int invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) { const char *p = (const char *)_p; unsigned codepage; DWORD mbflag = MB_ERR_INVALID_CHARS; if (sc->flag & SCONV_FROM_CHARSET) codepage = sc->to_cp; else codepage = sc->from_cp; if (codepage == CP_C_LOCALE) return (0); if (codepage != CP_UTF8) mbflag |= MB_PRECOMPOSED; if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0) return (-1); /* Invalid */ return (0); /* Okay */ } #else /* * Test whether MBS ==> WCS is okay. */ static int invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) { const char *p = (const char *)_p; size_t r; #if HAVE_MBRTOWC mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); #else /* Clear the shift state before starting. */ mbtowc(NULL, NULL, 0); #endif while (n) { wchar_t wc; #if HAVE_MBRTOWC r = mbrtowc(&wc, p, n, &shift_state); #else r = mbtowc(&wc, p, n); #endif if (r == (size_t)-1 || r == (size_t)-2) return (-1);/* Invalid. */ if (r == 0) break; p += r; n -= r; } (void)sc; /* UNUSED */ return (0); /* All Okey. */ } #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ /* * Basically returns -1 because we cannot make a conversion of charset * without iconv but in some cases this would return 0. * Returns 0 if all copied characters are ASCII. * Returns 0 if both from-locale and to-locale are the same and those * can be WCS with no error. */ static int best_effort_strncat_in_locale(struct archive_string *as, const void *_p, size_t length, struct archive_string_conv *sc) { size_t remaining; const uint8_t *itp; int return_value = 0; /* success */ /* * If both from-locale and to-locale is the same, this makes a copy. * And then this checks all copied MBS can be WCS if so returns 0. */ if (sc->same) { if (archive_string_append(as, _p, length) == NULL) return (-1);/* No memory */ return (invalid_mbs(_p, length, sc)); } /* * If a character is ASCII, this just copies it. If not, this * assigns '?' charater instead but in UTF-8 locale this assigns * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD, * a Replacement Character in Unicode. */ remaining = length; itp = (const uint8_t *)_p; while (*itp && remaining > 0) { if (*itp > 127) { // Non-ASCII: Substitute with suitable replacement if (sc->flag & SCONV_TO_UTF8) { if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) { __archive_errx(1, "Out of memory"); } } else { archive_strappend_char(as, '?'); } return_value = -1; } else { archive_strappend_char(as, *itp); } ++itp; } return (return_value); } /* * Unicode conversion functions. * - UTF-8 <===> UTF-8 in removing surrogate pairs. * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs. * - UTF-8 made by libarchive 2.x ===> UTF-8. * - UTF-16BE <===> UTF-8. * */ /* * Utility to convert a single UTF-8 sequence. * * Usually return used bytes, return used byte in negative value when * a unicode character is replaced with U+FFFD. * See also http://unicode.org/review/pr-121.html Public Review Issue #121 * Recommended Practice for Replacement Characters. */ static int _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) { static const char utf8_count[256] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ }; int ch, i; int cnt; uint32_t wc; /* Sanity check. */ if (n == 0) return (0); /* * Decode 1-4 bytes depending on the value of the first byte. */ ch = (unsigned char)*s; if (ch == 0) return (0); /* Standard: return 0 for end-of-string. */ cnt = utf8_count[ch]; /* Invalide sequence or there are not plenty bytes. */ if ((int)n < cnt) { cnt = (int)n; for (i = 1; i < cnt; i++) { if ((s[i] & 0xc0) != 0x80) { cnt = i; break; } } goto invalid_sequence; } /* Make a Unicode code point from a single UTF-8 sequence. */ switch (cnt) { case 1: /* 1 byte sequence. */ *pwc = ch & 0x7f; return (cnt); case 2: /* 2 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) { cnt = 1; goto invalid_sequence; } *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); return (cnt); case 3: /* 3 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) { cnt = 1; goto invalid_sequence; } if ((s[2] & 0xc0) != 0x80) { cnt = 2; goto invalid_sequence; } wc = ((ch & 0x0f) << 12) | ((s[1] & 0x3f) << 6) | (s[2] & 0x3f); if (wc < 0x800) goto invalid_sequence;/* Overlong sequence. */ break; case 4: /* 4 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) { cnt = 1; goto invalid_sequence; } if ((s[2] & 0xc0) != 0x80) { cnt = 2; goto invalid_sequence; } if ((s[3] & 0xc0) != 0x80) { cnt = 3; goto invalid_sequence; } wc = ((ch & 0x07) << 18) | ((s[1] & 0x3f) << 12) | ((s[2] & 0x3f) << 6) | (s[3] & 0x3f); if (wc < 0x10000) goto invalid_sequence;/* Overlong sequence. */ break; default: /* Others are all invalid sequence. */ if (ch == 0xc0 || ch == 0xc1) cnt = 2; else if (ch >= 0xf5 && ch <= 0xf7) cnt = 4; else if (ch >= 0xf8 && ch <= 0xfb) cnt = 5; else if (ch == 0xfc || ch == 0xfd) cnt = 6; else cnt = 1; if ((int)n < cnt) cnt = (int)n; for (i = 1; i < cnt; i++) { if ((s[i] & 0xc0) != 0x80) { cnt = i; break; } } goto invalid_sequence; } /* The code point larger than 0x10FFFF is not leagal * Unicode values. */ if (wc > UNICODE_MAX) goto invalid_sequence; /* Correctly gets a Unicode, returns used bytes. */ *pwc = wc; return (cnt); invalid_sequence: *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ return (cnt * -1); } static int utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) { int cnt; cnt = _utf8_to_unicode(pwc, s, n); /* Any of Surrogate pair is not leagal Unicode values. */ if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc)) return (-3); return (cnt); } static inline uint32_t combine_surrogate_pair(uint32_t uc, uint32_t uc2) { uc -= 0xD800; uc *= 0x400; uc += uc2 - 0xDC00; uc += 0x10000; return (uc); } /* * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in * removing surrogate pairs. * * CESU-8: The Compatibility Encoding Scheme for UTF-16. * * Usually return used bytes, return used byte in negative value when * a unicode character is replaced with U+FFFD. */ static int cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n) { uint32_t wc = 0; int cnt; cnt = _utf8_to_unicode(&wc, s, n); if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) { uint32_t wc2 = 0; if (n - 3 < 3) { /* Invalid byte sequence. */ goto invalid_sequence; } cnt = _utf8_to_unicode(&wc2, s+3, n-3); if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) { /* Invalid byte sequence. */ goto invalid_sequence; } wc = combine_surrogate_pair(wc, wc2); cnt = 6; } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) { /* Invalid byte sequence. */ goto invalid_sequence; } *pwc = wc; return (cnt); invalid_sequence: *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ if (cnt > 0) cnt *= -1; return (cnt); } /* * Convert a Unicode code point to a single UTF-8 sequence. * * NOTE:This function does not check if the Unicode is leagal or not. * Please you definitely check it before calling this. */ static size_t unicode_to_utf8(char *p, size_t remaining, uint32_t uc) { char *_p = p; /* Invalid Unicode char maps to Replacement character */ if (uc > UNICODE_MAX) uc = UNICODE_R_CHAR; /* Translate code point to UTF8 */ if (uc <= 0x7f) { if (remaining == 0) return (0); *p++ = (char)uc; } else if (uc <= 0x7ff) { if (remaining < 2) return (0); *p++ = 0xc0 | ((uc >> 6) & 0x1f); *p++ = 0x80 | (uc & 0x3f); } else if (uc <= 0xffff) { if (remaining < 3) return (0); *p++ = 0xe0 | ((uc >> 12) & 0x0f); *p++ = 0x80 | ((uc >> 6) & 0x3f); *p++ = 0x80 | (uc & 0x3f); } else { if (remaining < 4) return (0); *p++ = 0xf0 | ((uc >> 18) & 0x07); *p++ = 0x80 | ((uc >> 12) & 0x3f); *p++ = 0x80 | ((uc >> 6) & 0x3f); *p++ = 0x80 | (uc & 0x3f); } return (p - _p); } static int utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n) { return (utf16_to_unicode(pwc, s, n, 1)); } static int utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n) { return (utf16_to_unicode(pwc, s, n, 0)); } static int utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be) { const char *utf16 = s; unsigned uc; if (n == 0) return (0); if (n == 1) { /* set the Replacement Character instead. */ *pwc = UNICODE_R_CHAR; return (-1); } if (be) uc = archive_be16dec(utf16); else uc = archive_le16dec(utf16); utf16 += 2; /* If this is a surrogate pair, assemble the full code point.*/ if (IS_HIGH_SURROGATE_LA(uc)) { unsigned uc2; if (n >= 4) { if (be) uc2 = archive_be16dec(utf16); else uc2 = archive_le16dec(utf16); } else uc2 = 0; if (IS_LOW_SURROGATE_LA(uc2)) { uc = combine_surrogate_pair(uc, uc2); utf16 += 2; } else { /* Undescribed code point should be U+FFFD * (replacement character). */ *pwc = UNICODE_R_CHAR; return (-2); } } /* * Surrogate pair values(0xd800 through 0xdfff) are only * used by UTF-16, so, after above culculation, the code * must not be surrogate values, and Unicode has no codes * larger than 0x10ffff. Thus, those are not leagal Unicode * values. */ if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) { /* Undescribed code point should be U+FFFD * (replacement character). */ *pwc = UNICODE_R_CHAR; return (((int)(utf16 - s)) * -1); } *pwc = uc; return ((int)(utf16 - s)); } static size_t unicode_to_utf16be(char *p, size_t remaining, uint32_t uc) { char *utf16 = p; if (uc > 0xffff) { /* We have a code point that won't fit into a * wchar_t; convert it to a surrogate pair. */ if (remaining < 4) return (0); uc -= 0x10000; archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00); return (4); } else { if (remaining < 2) return (0); archive_be16enc(utf16, uc); return (2); } } static size_t unicode_to_utf16le(char *p, size_t remaining, uint32_t uc) { char *utf16 = p; if (uc > 0xffff) { /* We have a code point that won't fit into a * wchar_t; convert it to a surrogate pair. */ if (remaining < 4) return (0); uc -= 0x10000; archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00); return (4); } else { if (remaining < 2) return (0); archive_le16enc(utf16, uc); return (2); } } /* * Copy UTF-8 string in checking surrogate pair. * If any surrogate pair are found, it would be canonicalized. */ static int strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s; char *p, *endp; int n, ret = 0; (void)sc; /* UNUSED */ if (archive_string_ensure(as, as->length + len + 1) == NULL) return (-1); s = (const char *)_p; p = as->s + as->length; endp = as->s + as->buffer_length -1; do { uint32_t uc; const char *ss = s; size_t w; /* * Forward byte sequence until a conversion of that is needed. */ while ((n = utf8_to_unicode(&uc, s, len)) > 0) { s += n; len -= n; } if (ss < s) { if (p + (s - ss) > endp) { as->length = p - as->s; if (archive_string_ensure(as, as->buffer_length + len + 1) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length -1; } memcpy(p, ss, s - ss); p += s - ss; } /* * If n is negative, current byte sequence needs a replacement. */ if (n < 0) { if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) { /* Current byte sequence may be CESU-8. */ n = cesu8_to_unicode(&uc, s, len); } if (n < 0) { ret = -1; n *= -1;/* Use a replaced unicode character. */ } /* Rebuild UTF-8 byte sequence. */ while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) { as->length = p - as->s; if (archive_string_ensure(as, as->buffer_length + len + 1) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length -1; } p += w; s += n; len -= n; } } while (n > 0); as->length = p - as->s; as->s[as->length] = '\0'; return (ret); } static int archive_string_append_unicode(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s; char *p, *endp; uint32_t uc; size_t w; int n, ret = 0, ts, tm; int (*parse)(uint32_t *, const char *, size_t); size_t (*unparse)(char *, size_t, uint32_t); if (sc->flag & SCONV_TO_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; } else if (sc->flag & SCONV_TO_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; } else if (sc->flag & SCONV_TO_UTF8) { unparse = unicode_to_utf8; ts = 1; } else { /* * This case is going to be converted to another * character-set through iconv. */ if (sc->flag & SCONV_FROM_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; } else if (sc->flag & SCONV_FROM_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; } else { unparse = unicode_to_utf8; ts = 1; } } if (sc->flag & SCONV_FROM_UTF16BE) { parse = utf16be_to_unicode; tm = 1; } else if (sc->flag & SCONV_FROM_UTF16LE) { parse = utf16le_to_unicode; tm = 1; } else { parse = cesu8_to_unicode; tm = ts; } if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) return (-1); s = (const char *)_p; p = as->s + as->length; endp = as->s + as->buffer_length - ts; while ((n = parse(&uc, s, len)) != 0) { if (n < 0) { /* Use a replaced unicode character. */ n *= -1; ret = -1; } s += n; len -= n; while ((w = unparse(p, endp - p, uc)) == 0) { /* There is not enough output buffer so * we have to expand it. */ as->length = p - as->s; if (archive_string_ensure(as, as->buffer_length + len * tm + ts) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length - ts; } p += w; } as->length = p - as->s; as->s[as->length] = '\0'; if (ts == 2) as->s[as->length+1] = '\0'; return (ret); } /* * Following Constants for Hangul compositions this information comes from * Unicode Standard Annex #15 http://unicode.org/reports/tr15/ */ #define HC_SBASE 0xAC00 #define HC_LBASE 0x1100 #define HC_VBASE 0x1161 #define HC_TBASE 0x11A7 #define HC_LCOUNT 19 #define HC_VCOUNT 21 #define HC_TCOUNT 28 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT) #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT) static uint32_t get_nfc(uint32_t uc, uint32_t uc2) { int t, b; t = 0; b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1; while (b >= t) { int m = (t + b) / 2; if (u_composition_table[m].cp1 < uc) t = m + 1; else if (u_composition_table[m].cp1 > uc) b = m - 1; else if (u_composition_table[m].cp2 < uc2) t = m + 1; else if (u_composition_table[m].cp2 > uc2) b = m - 1; else return (u_composition_table[m].nfc); } return (0); } #define FDC_MAX 10 /* The maximum number of Following Decomposable * Characters. */ /* * Update first code point. */ #define UPDATE_UC(new_uc) do { \ uc = new_uc; \ ucptr = NULL; \ } while (0) /* * Replace first code point with second code point. */ #define REPLACE_UC_WITH_UC2() do { \ uc = uc2; \ ucptr = uc2ptr; \ n = n2; \ } while (0) #define EXPAND_BUFFER() do { \ as->length = p - as->s; \ if (archive_string_ensure(as, \ as->buffer_length + len * tm + ts) == NULL)\ return (-1); \ p = as->s + as->length; \ endp = as->s + as->buffer_length - ts; \ } while (0) #define UNPARSE(p, endp, uc) do { \ while ((w = unparse(p, (endp) - (p), uc)) == 0) {\ EXPAND_BUFFER(); \ } \ p += w; \ } while (0) /* * Write first code point. * If the code point has not be changed from its original code, * this just copies it from its original buffer pointer. * If not, this converts it to UTF-8 byte sequence and copies it. */ #define WRITE_UC() do { \ if (ucptr) { \ if (p + n > endp) \ EXPAND_BUFFER(); \ switch (n) { \ case 4: \ *p++ = *ucptr++; \ /* FALL THROUGH */ \ case 3: \ *p++ = *ucptr++; \ /* FALL THROUGH */ \ case 2: \ *p++ = *ucptr++; \ /* FALL THROUGH */ \ case 1: \ *p++ = *ucptr; \ break; \ } \ ucptr = NULL; \ } else { \ UNPARSE(p, endp, uc); \ } \ } while (0) /* * Collect following decomposable code points. */ #define COLLECT_CPS(start) do { \ int _i; \ for (_i = start; _i < FDC_MAX ; _i++) { \ nx = parse(&ucx[_i], s, len); \ if (nx <= 0) \ break; \ cx = CCC(ucx[_i]); \ if (cl >= cx && cl != 228 && cx != 228)\ break; \ s += nx; \ len -= nx; \ cl = cx; \ ccx[_i] = cx; \ } \ if (_i >= FDC_MAX) { \ ret = -1; \ ucx_size = FDC_MAX; \ } else \ ucx_size = _i; \ } while (0) /* * Normalize UTF-8/UTF-16BE characters to Form C and copy the result. * * TODO: Convert composition exclusions,which are never converted * from NFC,NFD,NFKC and NFKD, to Form C. */ static int archive_string_normalize_C(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s = (const char *)_p; char *p, *endp; uint32_t uc, uc2; size_t w; int always_replace, n, n2, ret = 0, spair, ts, tm; int (*parse)(uint32_t *, const char *, size_t); size_t (*unparse)(char *, size_t, uint32_t); always_replace = 1; ts = 1;/* text size. */ if (sc->flag & SCONV_TO_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; if (sc->flag & SCONV_FROM_UTF16BE) always_replace = 0; } else if (sc->flag & SCONV_TO_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; if (sc->flag & SCONV_FROM_UTF16LE) always_replace = 0; } else if (sc->flag & SCONV_TO_UTF8) { unparse = unicode_to_utf8; if (sc->flag & SCONV_FROM_UTF8) always_replace = 0; } else { /* * This case is going to be converted to another * character-set through iconv. */ always_replace = 0; if (sc->flag & SCONV_FROM_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; } else if (sc->flag & SCONV_FROM_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; } else { unparse = unicode_to_utf8; } } if (sc->flag & SCONV_FROM_UTF16BE) { parse = utf16be_to_unicode; tm = 1; spair = 4;/* surrogate pair size in UTF-16. */ } else if (sc->flag & SCONV_FROM_UTF16LE) { parse = utf16le_to_unicode; tm = 1; spair = 4;/* surrogate pair size in UTF-16. */ } else { parse = cesu8_to_unicode; tm = ts; spair = 6;/* surrogate pair size in UTF-8. */ } if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length - ts; while ((n = parse(&uc, s, len)) != 0) { const char *ucptr, *uc2ptr; if (n < 0) { /* Use a replaced unicode character. */ UNPARSE(p, endp, uc); s += n*-1; len -= n*-1; ret = -1; continue; } else if (n == spair || always_replace) /* uc is converted from a surrogate pair. * this should be treated as a changed code. */ ucptr = NULL; else ucptr = s; s += n; len -= n; /* Read second code point. */ while ((n2 = parse(&uc2, s, len)) > 0) { uint32_t ucx[FDC_MAX]; int ccx[FDC_MAX]; int cl, cx, i, nx, ucx_size; int LIndex,SIndex; uint32_t nfc; if (n2 == spair || always_replace) /* uc2 is converted from a surrogate pair. * this should be treated as a changed code. */ uc2ptr = NULL; else uc2ptr = s; s += n2; len -= n2; /* * If current second code point is out of decomposable * code points, finding compositions is unneeded. */ if (!IS_DECOMPOSABLE_BLOCK(uc2)) { WRITE_UC(); REPLACE_UC_WITH_UC2(); continue; } /* * Try to combine current code points. */ /* * We have to combine Hangul characters according to * http://uniicode.org/reports/tr15/#Hangul */ if (0 <= (LIndex = uc - HC_LBASE) && LIndex < HC_LCOUNT) { /* * Hangul Composition. * 1. Two current code points are L and V. */ int VIndex = uc2 - HC_VBASE; if (0 <= VIndex && VIndex < HC_VCOUNT) { /* Make syllable of form LV. */ UPDATE_UC(HC_SBASE + (LIndex * HC_VCOUNT + VIndex) * HC_TCOUNT); } else { WRITE_UC(); REPLACE_UC_WITH_UC2(); } continue; } else if (0 <= (SIndex = uc - HC_SBASE) && SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) { /* * Hangul Composition. * 2. Two current code points are LV and T. */ int TIndex = uc2 - HC_TBASE; if (0 < TIndex && TIndex < HC_TCOUNT) { /* Make syllable of form LVT. */ UPDATE_UC(uc + TIndex); } else { WRITE_UC(); REPLACE_UC_WITH_UC2(); } continue; } else if ((nfc = get_nfc(uc, uc2)) != 0) { /* A composition to current code points * is found. */ UPDATE_UC(nfc); continue; } else if ((cl = CCC(uc2)) == 0) { /* Clearly 'uc2' the second code point is not * a decomposable code. */ WRITE_UC(); REPLACE_UC_WITH_UC2(); continue; } /* * Collect following decomposable code points. */ cx = 0; ucx[0] = uc2; ccx[0] = cl; COLLECT_CPS(1); /* * Find a composed code in the collected code points. */ i = 1; while (i < ucx_size) { int j; if ((nfc = get_nfc(uc, ucx[i])) == 0) { i++; continue; } /* * nfc is composed of uc and ucx[i]. */ UPDATE_UC(nfc); /* * Remove ucx[i] by shifting * following code points. */ for (j = i; j+1 < ucx_size; j++) { ucx[j] = ucx[j+1]; ccx[j] = ccx[j+1]; } ucx_size --; /* * Collect following code points blocked * by ucx[i] the removed code point. */ if (ucx_size > 0 && i == ucx_size && nx > 0 && cx == cl) { cl = ccx[ucx_size-1]; COLLECT_CPS(ucx_size); } /* * Restart finding a composed code with * the updated uc from the top of the * collected code points. */ i = 0; } /* * Apparently the current code points are not * decomposed characters or already composed. */ WRITE_UC(); for (i = 0; i < ucx_size; i++) UNPARSE(p, endp, ucx[i]); /* * Flush out remaining canonical combining characters. */ if (nx > 0 && cx == cl && len > 0) { while ((nx = parse(&ucx[0], s, len)) > 0) { cx = CCC(ucx[0]); if (cl > cx) break; s += nx; len -= nx; cl = cx; UNPARSE(p, endp, ucx[0]); } } break; } if (n2 < 0) { WRITE_UC(); /* Use a replaced unicode character. */ UNPARSE(p, endp, uc2); s += n2*-1; len -= n2*-1; ret = -1; continue; } else if (n2 == 0) { WRITE_UC(); break; } } as->length = p - as->s; as->s[as->length] = '\0'; if (ts == 2) as->s[as->length+1] = '\0'; return (ret); } static int get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc) { int t, b; /* * These are not converted to NFD on Mac OS. */ if ((uc >= 0x2000 && uc <= 0x2FFF) || (uc >= 0xF900 && uc <= 0xFAFF) || (uc >= 0x2F800 && uc <= 0x2FAFF)) return (0); /* * Those code points are not converted to NFD on Mac OS. * I do not know the reason because it is undocumented. * NFC NFD * 1109A ==> 11099 110BA * 1109C ==> 1109B 110BA * 110AB ==> 110A5 110BA */ if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB) return (0); t = 0; b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1; while (b >= t) { int m = (t + b) / 2; if (u_decomposition_table[m].nfc < uc) t = m + 1; else if (u_decomposition_table[m].nfc > uc) b = m - 1; else { *cp1 = u_decomposition_table[m].cp1; *cp2 = u_decomposition_table[m].cp2; return (1); } } return (0); } #define REPLACE_UC_WITH(cp) do { \ uc = cp; \ ucptr = NULL; \ } while (0) /* * Normalize UTF-8 characters to Form D and copy the result. */ static int archive_string_normalize_D(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s = (const char *)_p; char *p, *endp; uint32_t uc, uc2; size_t w; int always_replace, n, n2, ret = 0, spair, ts, tm; int (*parse)(uint32_t *, const char *, size_t); size_t (*unparse)(char *, size_t, uint32_t); always_replace = 1; ts = 1;/* text size. */ if (sc->flag & SCONV_TO_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; if (sc->flag & SCONV_FROM_UTF16BE) always_replace = 0; } else if (sc->flag & SCONV_TO_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; if (sc->flag & SCONV_FROM_UTF16LE) always_replace = 0; } else if (sc->flag & SCONV_TO_UTF8) { unparse = unicode_to_utf8; if (sc->flag & SCONV_FROM_UTF8) always_replace = 0; } else { /* * This case is going to be converted to another * character-set through iconv. */ always_replace = 0; if (sc->flag & SCONV_FROM_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; } else if (sc->flag & SCONV_FROM_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; } else { unparse = unicode_to_utf8; } } if (sc->flag & SCONV_FROM_UTF16BE) { parse = utf16be_to_unicode; tm = 1; spair = 4;/* surrogate pair size in UTF-16. */ } else if (sc->flag & SCONV_FROM_UTF16LE) { parse = utf16le_to_unicode; tm = 1; spair = 4;/* surrogate pair size in UTF-16. */ } else { parse = cesu8_to_unicode; tm = ts; spair = 6;/* surrogate pair size in UTF-8. */ } if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length - ts; while ((n = parse(&uc, s, len)) != 0) { const char *ucptr; uint32_t cp1, cp2; int SIndex; struct { uint32_t uc; int ccc; } fdc[FDC_MAX]; int fdi, fdj; int ccc; check_first_code: if (n < 0) { /* Use a replaced unicode character. */ UNPARSE(p, endp, uc); s += n*-1; len -= n*-1; ret = -1; continue; } else if (n == spair || always_replace) /* uc is converted from a surrogate pair. * this should be treated as a changed code. */ ucptr = NULL; else ucptr = s; s += n; len -= n; /* Hangul Decomposition. */ if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) { int L = HC_LBASE + SIndex / HC_NCOUNT; int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT; int T = HC_TBASE + SIndex % HC_TCOUNT; REPLACE_UC_WITH(L); WRITE_UC(); REPLACE_UC_WITH(V); WRITE_UC(); if (T != HC_TBASE) { REPLACE_UC_WITH(T); WRITE_UC(); } continue; } if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) { WRITE_UC(); continue; } fdi = 0; while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) { int k; for (k = fdi; k > 0; k--) fdc[k] = fdc[k-1]; fdc[0].ccc = CCC(cp2); fdc[0].uc = cp2; fdi++; REPLACE_UC_WITH(cp1); } /* Read following code points. */ while ((n2 = parse(&uc2, s, len)) > 0 && (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) { int j, k; s += n2; len -= n2; for (j = 0; j < fdi; j++) { if (fdc[j].ccc > ccc) break; } if (j < fdi) { for (k = fdi; k > j; k--) fdc[k] = fdc[k-1]; fdc[j].ccc = ccc; fdc[j].uc = uc2; } else { fdc[fdi].ccc = ccc; fdc[fdi].uc = uc2; } fdi++; } WRITE_UC(); for (fdj = 0; fdj < fdi; fdj++) { REPLACE_UC_WITH(fdc[fdj].uc); WRITE_UC(); } if (n2 == 0) break; REPLACE_UC_WITH(uc2); n = n2; goto check_first_code; } as->length = p - as->s; as->s[as->length] = '\0'; if (ts == 2) as->s[as->length+1] = '\0'; return (ret); } /* * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption * that WCS is Unicode. It is true for several platforms but some are false. * And then people who did not use UTF-8 locale on the non Unicode WCS * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those * now cannot get right filename from libarchive 3.x and later since we * fixed the wrong assumption and it is incompatible to older its versions. * So we provide special option, "compat-2x.x", for resolving it. * That option enable the string conversion of libarchive 2.x. * * Translates the wrong UTF-8 string made by libarchive 2.x into current * locale character set and appends to the archive_string. * Note: returns -1 if conversion fails. */ static int strncat_from_utf8_libarchive2(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s; int n; char *p; char *end; uint32_t unicode; #if HAVE_WCRTOMB mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); #else /* Clear the shift state before starting. */ wctomb(NULL, L'\0'); #endif (void)sc; /* UNUSED */ /* * Allocate buffer for MBS. * We need this allocation here since it is possible that * as->s is still NULL. */ if (archive_string_ensure(as, as->length + len + 1) == NULL) return (-1); s = (const char *)_p; p = as->s + as->length; end = as->s + as->buffer_length - MB_CUR_MAX -1; while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) { wchar_t wc; if (p >= end) { as->length = p - as->s; /* Re-allocate buffer for MBS. */ if (archive_string_ensure(as, as->length + len * 2 + 1) == NULL) return (-1); p = as->s + as->length; end = as->s + as->buffer_length - MB_CUR_MAX -1; } /* * As libarchie 2.x, translates the UTF-8 characters into * wide-characters in the assumption that WCS is Unicode. */ if (n < 0) { n *= -1; wc = L'?'; } else wc = (wchar_t)unicode; s += n; len -= n; /* * Translates the wide-character into the current locale MBS. */ #if HAVE_WCRTOMB n = (int)wcrtomb(p, wc, &shift_state); #else n = (int)wctomb(p, wc); #endif if (n == -1) return (-1); p += n; } as->length = p - as->s; as->s[as->length] = '\0'; return (0); } /* * Conversion functions between current locale dependent MBS and UTF-16BE. * strncat_from_utf16be() : UTF-16BE --> MBS * strncat_to_utf16be() : MBS --> UTF16BE */ #if defined(_WIN32) && !defined(__CYGWIN__) /* * Convert a UTF-16BE/LE string to current locale and copy the result. * Return -1 if conversion failes. */ static int win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc, int be) { struct archive_string tmp; const char *u16; int ll; BOOL defchar; char *mbs; size_t mbs_size, b; int ret = 0; bytes &= ~1; if (archive_string_ensure(as, as->length + bytes +1) == NULL) return (-1); mbs = as->s + as->length; mbs_size = as->buffer_length - as->length -1; if (sc->to_cp == CP_C_LOCALE) { /* * "C" locale special process. */ u16 = _p; ll = 0; for (b = 0; b < bytes; b += 2) { uint16_t val; if (be) val = archive_be16dec(u16+b); else val = archive_le16dec(u16+b); if (val > 255) { *mbs++ = '?'; ret = -1; } else *mbs++ = (char)(val&0xff); ll++; } as->length += ll; as->s[as->length] = '\0'; return (ret); } archive_string_init(&tmp); if (be) { if (is_big_endian()) { u16 = _p; } else { if (archive_string_ensure(&tmp, bytes+2) == NULL) return (-1); memcpy(tmp.s, _p, bytes); for (b = 0; b < bytes; b += 2) { uint16_t val = archive_be16dec(tmp.s+b); archive_le16enc(tmp.s+b, val); } u16 = tmp.s; } } else { if (!is_big_endian()) { u16 = _p; } else { if (archive_string_ensure(&tmp, bytes+2) == NULL) return (-1); memcpy(tmp.s, _p, bytes); for (b = 0; b < bytes; b += 2) { uint16_t val = archive_le16dec(tmp.s+b); archive_be16enc(tmp.s+b, val); } u16 = tmp.s; } } do { defchar = 0; ll = WideCharToMultiByte(sc->to_cp, 0, (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size, NULL, &defchar); - if (ll == 0 && - GetLastError() == ERROR_INSUFFICIENT_BUFFER) { - /* Need more buffer for MBS. */ - ll = WideCharToMultiByte(sc->to_cp, 0, - (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL); - if (archive_string_ensure(as, ll +1) == NULL) - return (-1); - mbs = as->s + as->length; - mbs_size = as->buffer_length - as->length -1; - continue; + /* Exit loop if we succeeded */ + if (ll != 0 || + GetLastError() != ERROR_INSUFFICIENT_BUFFER) { + break; } - } while (0); + /* Else expand buffer and loop to try again. */ + ll = WideCharToMultiByte(sc->to_cp, 0, + (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL); + if (archive_string_ensure(as, ll +1) == NULL) + return (-1); + mbs = as->s + as->length; + mbs_size = as->buffer_length - as->length -1; + } while (1); archive_string_free(&tmp); as->length += ll; as->s[as->length] = '\0'; if (ll == 0 || defchar) ret = -1; return (ret); } static int win_strncat_from_utf16be(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc) { return (win_strncat_from_utf16(as, _p, bytes, sc, 1)); } static int win_strncat_from_utf16le(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc) { return (win_strncat_from_utf16(as, _p, bytes, sc, 0)); } static int is_big_endian(void) { uint16_t d = 1; return (archive_be16dec(&d) == 1); } /* * Convert a current locale string to UTF-16BE/LE and copy the result. * Return -1 if conversion failes. */ static int win_strncat_to_utf16(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc, int bigendian) { const char *s = (const char *)_p; char *u16; size_t count, avail; if (archive_string_ensure(as16, as16->length + (length + 1) * 2) == NULL) return (-1); u16 = as16->s + as16->length; avail = as16->buffer_length - 2; if (sc->from_cp == CP_C_LOCALE) { /* * "C" locale special process. */ count = 0; while (count < length && *s) { if (bigendian) archive_be16enc(u16, *s); else archive_le16enc(u16, *s); u16 += 2; s++; count++; } as16->length += count << 1; as16->s[as16->length] = 0; as16->s[as16->length+1] = 0; return (0); } do { count = MultiByteToWideChar(sc->from_cp, MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1); - if (count == 0 && - GetLastError() == ERROR_INSUFFICIENT_BUFFER) { - /* Need more buffer for UTF-16 string */ - count = MultiByteToWideChar(sc->from_cp, - MB_PRECOMPOSED, s, (int)length, NULL, 0); - if (archive_string_ensure(as16, (count +1) * 2) - == NULL) - return (-1); - u16 = as16->s + as16->length; - avail = as16->buffer_length - 2; - continue; + /* Exit loop if we succeeded */ + if (count != 0 || + GetLastError() != ERROR_INSUFFICIENT_BUFFER) { + break; } - } while (0); + /* Expand buffer and try again */ + count = MultiByteToWideChar(sc->from_cp, + MB_PRECOMPOSED, s, (int)length, NULL, 0); + if (archive_string_ensure(as16, (count +1) * 2) + == NULL) + return (-1); + u16 = as16->s + as16->length; + avail = as16->buffer_length - 2; + } while (1); as16->length += count * 2; as16->s[as16->length] = 0; as16->s[as16->length+1] = 0; if (count == 0) return (-1); if (is_big_endian()) { if (!bigendian) { while (count > 0) { uint16_t v = archive_be16dec(u16); archive_le16enc(u16, v); u16 += 2; count--; } } } else { if (bigendian) { while (count > 0) { uint16_t v = archive_le16dec(u16); archive_be16enc(u16, v); u16 += 2; count--; } } } return (0); } static int win_strncat_to_utf16be(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc) { return (win_strncat_to_utf16(as16, _p, length, sc, 1)); } static int win_strncat_to_utf16le(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc) { return (win_strncat_to_utf16(as16, _p, length, sc, 0)); } #endif /* _WIN32 && !__CYGWIN__ */ /* * Do the best effort for conversions. * We cannot handle UTF-16BE character-set without such iconv, * but there is a chance if a string consists just ASCII code or * a current locale is UTF-8. */ /* * Convert a UTF-16BE string to current locale and copy the result. * Return -1 if conversion failes. */ static int best_effort_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc, int be) { const char *utf16 = (const char *)_p; char *mbs; uint32_t uc; int n, ret; (void)sc; /* UNUSED */ /* * Other case, we should do the best effort. * If all character are ASCII(<0x7f), we can convert it. * if not , we set a alternative character and return -1. */ ret = 0; if (archive_string_ensure(as, as->length + bytes +1) == NULL) return (-1); mbs = as->s + as->length; while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) { if (n < 0) { n *= -1; ret = -1; } bytes -= n; utf16 += n; if (uc > 127) { /* We cannot handle it. */ *mbs++ = '?'; ret = -1; } else *mbs++ = (char)uc; } as->length = mbs - as->s; as->s[as->length] = '\0'; return (ret); } static int best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc) { return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1)); } static int best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc) { return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0)); } /* * Convert a current locale string to UTF-16BE/LE and copy the result. * Return -1 if conversion failes. */ static int best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc, int bigendian) { const char *s = (const char *)_p; char *utf16; size_t remaining; int ret; (void)sc; /* UNUSED */ /* * Other case, we should do the best effort. * If all character are ASCII(<0x7f), we can convert it. * if not , we set a alternative character and return -1. */ ret = 0; remaining = length; if (archive_string_ensure(as16, as16->length + (length + 1) * 2) == NULL) return (-1); utf16 = as16->s + as16->length; while (remaining--) { unsigned c = *s++; if (c > 127) { /* We cannot handle it. */ c = UNICODE_R_CHAR; ret = -1; } if (bigendian) archive_be16enc(utf16, c); else archive_le16enc(utf16, c); utf16 += 2; } as16->length = utf16 - as16->s; as16->s[as16->length] = 0; as16->s[as16->length+1] = 0; return (ret); } static int best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc) { return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1)); } static int best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc) { return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0)); } /* * Multistring operations. */ void archive_mstring_clean(struct archive_mstring *aes) { archive_wstring_free(&(aes->aes_wcs)); archive_string_free(&(aes->aes_mbs)); archive_string_free(&(aes->aes_utf8)); archive_string_free(&(aes->aes_mbs_in_locale)); aes->aes_set = 0; } void archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src) { dest->aes_set = src->aes_set; archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs)); archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8)); archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs)); } int archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes, const char **p) { struct archive_string_conv *sc; int r; /* If we already have a UTF8 form, return that immediately. */ if (aes->aes_set & AES_SET_UTF8) { *p = aes->aes_utf8.s; return (0); } *p = NULL; if (aes->aes_set & AES_SET_MBS) { sc = archive_string_conversion_to_charset(a, "UTF-8", 1); if (sc == NULL) return (-1);/* Couldn't allocate memory for sc. */ r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s, aes->aes_mbs.length, sc); if (a == NULL) free_sconv_object(sc); if (r == 0) { aes->aes_set |= AES_SET_UTF8; *p = aes->aes_utf8.s; return (0);/* success. */ } else return (-1);/* failure. */ } return (0);/* success. */ } int archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes, const char **p) { int r, ret = 0; (void)a; /* UNUSED */ /* If we already have an MBS form, return that immediately. */ if (aes->aes_set & AES_SET_MBS) { *p = aes->aes_mbs.s; return (ret); } *p = NULL; /* If there's a WCS form, try converting with the native locale. */ if (aes->aes_set & AES_SET_WCS) { archive_string_empty(&(aes->aes_mbs)); r = archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s, aes->aes_wcs.length); *p = aes->aes_mbs.s; if (r == 0) { aes->aes_set |= AES_SET_MBS; return (ret); } else ret = -1; } /* * Only a UTF-8 form cannot avail because its conversion already * failed at archive_mstring_update_utf8(). */ return (ret); } int archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes, const wchar_t **wp) { int r, ret = 0; (void)a;/* UNUSED */ /* Return WCS form if we already have it. */ if (aes->aes_set & AES_SET_WCS) { *wp = aes->aes_wcs.s; return (ret); } *wp = NULL; /* Try converting MBS to WCS using native locale. */ if (aes->aes_set & AES_SET_MBS) { archive_wstring_empty(&(aes->aes_wcs)); r = archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, aes->aes_mbs.length); if (r == 0) { aes->aes_set |= AES_SET_WCS; *wp = aes->aes_wcs.s; } else ret = -1;/* failure. */ } return (ret); } int archive_mstring_get_mbs_l(struct archive_mstring *aes, const char **p, size_t *length, struct archive_string_conv *sc) { int r, ret = 0; #if defined(_WIN32) && !defined(__CYGWIN__) /* * Internationalization programing on Windows must use Wide * characters because Windows platform cannot make locale UTF-8. */ if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) { archive_string_empty(&(aes->aes_mbs_in_locale)); r = archive_string_append_from_wcs_in_codepage( &(aes->aes_mbs_in_locale), aes->aes_wcs.s, aes->aes_wcs.length, sc); if (r == 0) { *p = aes->aes_mbs_in_locale.s; if (length != NULL) *length = aes->aes_mbs_in_locale.length; return (0); } else if (errno == ENOMEM) return (-1); else ret = -1; } #endif /* If there is not an MBS form but is a WCS form, try converting * with the native locale to be used for translating it to specified * character-set. */ if ((aes->aes_set & AES_SET_MBS) == 0 && (aes->aes_set & AES_SET_WCS) != 0) { archive_string_empty(&(aes->aes_mbs)); r = archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s, aes->aes_wcs.length); if (r == 0) aes->aes_set |= AES_SET_MBS; else if (errno == ENOMEM) return (-1); else ret = -1; } /* If we already have an MBS form, use it to be translated to * specified character-set. */ if (aes->aes_set & AES_SET_MBS) { if (sc == NULL) { /* Conversion is unneeded. */ *p = aes->aes_mbs.s; if (length != NULL) *length = aes->aes_mbs.length; return (0); } ret = archive_strncpy_l(&(aes->aes_mbs_in_locale), aes->aes_mbs.s, aes->aes_mbs.length, sc); *p = aes->aes_mbs_in_locale.s; if (length != NULL) *length = aes->aes_mbs_in_locale.length; } else { *p = NULL; if (length != NULL) *length = 0; } return (ret); } int archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs) { if (mbs == NULL) { aes->aes_set = 0; return (0); } return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs))); } int archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs, size_t len) { if (mbs == NULL) { aes->aes_set = 0; return (0); } aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ archive_strncpy(&(aes->aes_mbs), mbs, len); archive_string_empty(&(aes->aes_utf8)); archive_wstring_empty(&(aes->aes_wcs)); return (0); } int archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs) { return archive_mstring_copy_wcs_len(aes, wcs, wcs == NULL ? 0 : wcslen(wcs)); } int archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8) { if (utf8 == NULL) { aes->aes_set = 0; } aes->aes_set = AES_SET_UTF8; archive_string_empty(&(aes->aes_mbs)); archive_string_empty(&(aes->aes_wcs)); archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8)); return (int)strlen(utf8); } int archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs, size_t len) { if (wcs == NULL) { aes->aes_set = 0; } aes->aes_set = AES_SET_WCS; /* Only WCS form set. */ archive_string_empty(&(aes->aes_mbs)); archive_string_empty(&(aes->aes_utf8)); archive_wstrncpy(&(aes->aes_wcs), wcs, len); return (0); } int archive_mstring_copy_mbs_len_l(struct archive_mstring *aes, const char *mbs, size_t len, struct archive_string_conv *sc) { int r; if (mbs == NULL) { aes->aes_set = 0; return (0); } archive_string_empty(&(aes->aes_mbs)); archive_wstring_empty(&(aes->aes_wcs)); archive_string_empty(&(aes->aes_utf8)); #if defined(_WIN32) && !defined(__CYGWIN__) /* * Internationalization programing on Windows must use Wide * characters because Windows platform cannot make locale UTF-8. */ if (sc == NULL) { if (archive_string_append(&(aes->aes_mbs), mbs, mbsnbytes(mbs, len)) == NULL) { aes->aes_set = 0; r = -1; } else { aes->aes_set = AES_SET_MBS; r = 0; } #if defined(HAVE_ICONV) } else if (sc != NULL && sc->cd_w != (iconv_t)-1) { /* * This case happens only when MultiByteToWideChar() cannot * handle sc->from_cp, and we have to iconv in order to * translate character-set to wchar_t,UTF-16. */ iconv_t cd = sc->cd; unsigned from_cp; int flag; /* * Translate multi-bytes from some character-set to UTF-8. */ sc->cd = sc->cd_w; r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc); sc->cd = cd; if (r != 0) { aes->aes_set = 0; return (r); } aes->aes_set = AES_SET_UTF8; /* * Append the UTF-8 string into wstring. */ flag = sc->flag; sc->flag &= ~(SCONV_NORMALIZATION_C | SCONV_TO_UTF16| SCONV_FROM_UTF16); from_cp = sc->from_cp; sc->from_cp = CP_UTF8; r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs), aes->aes_utf8.s, aes->aes_utf8.length, sc); sc->flag = flag; sc->from_cp = from_cp; if (r == 0) aes->aes_set |= AES_SET_WCS; #endif } else { r = archive_wstring_append_from_mbs_in_codepage( &(aes->aes_wcs), mbs, len, sc); if (r == 0) aes->aes_set = AES_SET_WCS; else aes->aes_set = 0; } #else r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc); if (r == 0) aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ else aes->aes_set = 0; #endif return (r); } /* * The 'update' form tries to proactively update all forms of * this string (WCS and MBS) and returns an error if any of * them fail. This is used by the 'pax' handler, for instance, * to detect and report character-conversion failures early while * still allowing clients to get potentially useful values from * the more tolerant lazy conversions. (get_mbs and get_wcs will * strive to give the user something useful, so you can get hopefully * usable values even if some of the character conversions are failing.) */ int archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes, const char *utf8) { struct archive_string_conv *sc; int r; if (utf8 == NULL) { aes->aes_set = 0; return (0); /* Succeeded in clearing everything. */ } /* Save the UTF8 string. */ archive_strcpy(&(aes->aes_utf8), utf8); /* Empty the mbs and wcs strings. */ archive_string_empty(&(aes->aes_mbs)); archive_wstring_empty(&(aes->aes_wcs)); aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */ /* Try converting UTF-8 to MBS, return false on failure. */ sc = archive_string_conversion_from_charset(a, "UTF-8", 1); if (sc == NULL) return (-1);/* Couldn't allocate memory for sc. */ r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc); if (a == NULL) free_sconv_object(sc); if (r != 0) return (-1); aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */ /* Try converting MBS to WCS, return false on failure. */ if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, aes->aes_mbs.length)) return (-1); aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS; /* All conversions succeeded. */ return (0); } Index: stable/10/contrib/libarchive/libarchive/archive_write_disk_posix.c =================================================================== --- stable/10/contrib/libarchive/libarchive/archive_write_disk_posix.c (revision 302424) +++ stable/10/contrib/libarchive/libarchive/archive_write_disk_posix.c (revision 302425) @@ -1,3903 +1,3907 @@ /*- * Copyright (c) 2003-2010 Tim Kientzle * Copyright (c) 2012 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "archive_platform.h" __FBSDID("$FreeBSD$"); #if !defined(_WIN32) || defined(__CYGWIN__) #ifdef HAVE_SYS_TYPES_H #include #endif #ifdef HAVE_SYS_ACL_H #include #endif #ifdef HAVE_SYS_EXTATTR_H #include #endif #if defined(HAVE_SYS_XATTR_H) #include #elif defined(HAVE_ATTR_XATTR_H) #include #endif #ifdef HAVE_SYS_EA_H #include #endif #ifdef HAVE_SYS_IOCTL_H #include #endif #ifdef HAVE_SYS_STAT_H #include #endif #ifdef HAVE_SYS_TIME_H #include #endif #ifdef HAVE_SYS_UTIME_H #include #endif #ifdef HAVE_COPYFILE_H #include #endif #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_FCNTL_H #include #endif #ifdef HAVE_GRP_H #include #endif #ifdef HAVE_LANGINFO_H #include #endif #ifdef HAVE_LINUX_FS_H #include /* for Linux file flags */ #endif /* * Some Linux distributions have both linux/ext2_fs.h and ext2fs/ext2_fs.h. * As the include guards don't agree, the order of include is important. */ #ifdef HAVE_LINUX_EXT2_FS_H #include /* for Linux file flags */ #endif #if defined(HAVE_EXT2FS_EXT2_FS_H) && !defined(__CYGWIN__) #include /* Linux file flags, broken on Cygwin */ #endif #ifdef HAVE_LIMITS_H #include #endif #ifdef HAVE_PWD_H #include #endif #include #ifdef HAVE_STDLIB_H #include #endif #ifdef HAVE_STRING_H #include #endif #ifdef HAVE_UNISTD_H #include #endif #ifdef HAVE_UTIME_H #include #endif #ifdef F_GETTIMES /* Tru64 specific */ #include #endif #if __APPLE__ #include #if TARGET_OS_MAC && !TARGET_OS_EMBEDDED && HAVE_QUARANTINE_H #include #define HAVE_QUARANTINE 1 #endif #endif #ifdef HAVE_ZLIB_H #include #endif /* TODO: Support Mac OS 'quarantine' feature. This is really just a * standard tag to mark files that have been downloaded as "tainted". * On Mac OS, we should mark the extracted files as tainted if the * archive being read was tainted. Windows has a similar feature; we * should investigate ways to support this generically. */ #include "archive.h" #include "archive_acl_private.h" #include "archive_string.h" #include "archive_endian.h" #include "archive_entry.h" #include "archive_private.h" #include "archive_write_disk_private.h" #ifndef O_BINARY #define O_BINARY 0 #endif #ifndef O_CLOEXEC #define O_CLOEXEC 0 #endif struct fixup_entry { struct fixup_entry *next; struct archive_acl acl; mode_t mode; int64_t atime; int64_t birthtime; int64_t mtime; int64_t ctime; unsigned long atime_nanos; unsigned long birthtime_nanos; unsigned long mtime_nanos; unsigned long ctime_nanos; unsigned long fflags_set; size_t mac_metadata_size; void *mac_metadata; int fixup; /* bitmask of what needs fixing */ char *name; }; /* * We use a bitmask to track which operations remain to be done for * this file. In particular, this helps us avoid unnecessary * operations when it's possible to take care of one step as a * side-effect of another. For example, mkdir() can specify the mode * for the newly-created object but symlink() cannot. This means we * can skip chmod() if mkdir() succeeded, but we must explicitly * chmod() if we're trying to create a directory that already exists * (mkdir() failed) or if we're restoring a symlink. Similarly, we * need to verify UID/GID before trying to restore SUID/SGID bits; * that verification can occur explicitly through a stat() call or * implicitly because of a successful chown() call. */ #define TODO_MODE_FORCE 0x40000000 #define TODO_MODE_BASE 0x20000000 #define TODO_SUID 0x10000000 #define TODO_SUID_CHECK 0x08000000 #define TODO_SGID 0x04000000 #define TODO_SGID_CHECK 0x02000000 #define TODO_APPLEDOUBLE 0x01000000 #define TODO_MODE (TODO_MODE_BASE|TODO_SUID|TODO_SGID) #define TODO_TIMES ARCHIVE_EXTRACT_TIME #define TODO_OWNER ARCHIVE_EXTRACT_OWNER #define TODO_FFLAGS ARCHIVE_EXTRACT_FFLAGS #define TODO_ACLS ARCHIVE_EXTRACT_ACL #define TODO_XATTR ARCHIVE_EXTRACT_XATTR #define TODO_MAC_METADATA ARCHIVE_EXTRACT_MAC_METADATA #define TODO_HFS_COMPRESSION ARCHIVE_EXTRACT_HFS_COMPRESSION_FORCED struct archive_write_disk { struct archive archive; mode_t user_umask; struct fixup_entry *fixup_list; struct fixup_entry *current_fixup; int64_t user_uid; int skip_file_set; int64_t skip_file_dev; int64_t skip_file_ino; time_t start_time; int64_t (*lookup_gid)(void *private, const char *gname, int64_t gid); void (*cleanup_gid)(void *private); void *lookup_gid_data; int64_t (*lookup_uid)(void *private, const char *uname, int64_t uid); void (*cleanup_uid)(void *private); void *lookup_uid_data; /* * Full path of last file to satisfy symlink checks. */ struct archive_string path_safe; /* * Cached stat data from disk for the current entry. * If this is valid, pst points to st. Otherwise, * pst is null. */ struct stat st; struct stat *pst; /* Information about the object being restored right now. */ struct archive_entry *entry; /* Entry being extracted. */ char *name; /* Name of entry, possibly edited. */ struct archive_string _name_data; /* backing store for 'name' */ /* Tasks remaining for this object. */ int todo; /* Tasks deferred until end-of-archive. */ int deferred; /* Options requested by the client. */ int flags; /* Handle for the file we're restoring. */ int fd; /* Current offset for writing data to the file. */ int64_t offset; /* Last offset actually written to disk. */ int64_t fd_offset; /* Total bytes actually written to files. */ int64_t total_bytes_written; /* Maximum size of file, -1 if unknown. */ int64_t filesize; /* Dir we were in before this restore; only for deep paths. */ int restore_pwd; /* Mode we should use for this entry; affected by _PERM and umask. */ mode_t mode; /* UID/GID to use in restoring this entry. */ int64_t uid; int64_t gid; /* * HFS+ Compression. */ /* Xattr "com.apple.decmpfs". */ uint32_t decmpfs_attr_size; unsigned char *decmpfs_header_p; /* ResourceFork set options used for fsetxattr. */ int rsrc_xattr_options; /* Xattr "com.apple.ResourceFork". */ unsigned char *resource_fork; size_t resource_fork_allocated_size; unsigned int decmpfs_block_count; uint32_t *decmpfs_block_info; /* Buffer for compressed data. */ unsigned char *compressed_buffer; size_t compressed_buffer_size; size_t compressed_buffer_remaining; /* The offset of the ResourceFork where compressed data will * be placed. */ uint32_t compressed_rsrc_position; uint32_t compressed_rsrc_position_v; /* Buffer for uncompressed data. */ char *uncompressed_buffer; size_t block_remaining_bytes; size_t file_remaining_bytes; #ifdef HAVE_ZLIB_H z_stream stream; int stream_valid; int decmpfs_compression_level; #endif }; /* * Default mode for dirs created automatically (will be modified by umask). * Note that POSIX specifies 0777 for implicitly-created dirs, "modified * by the process' file creation mask." */ #define DEFAULT_DIR_MODE 0777 /* * Dir modes are restored in two steps: During the extraction, the permissions * in the archive are modified to match the following limits. During * the post-extract fixup pass, the permissions from the archive are * applied. */ #define MINIMUM_DIR_MODE 0700 #define MAXIMUM_DIR_MODE 0775 /* * Maxinum uncompressed size of a decmpfs block. */ #define MAX_DECMPFS_BLOCK_SIZE (64 * 1024) /* * HFS+ compression type. */ #define CMP_XATTR 3/* Compressed data in xattr. */ #define CMP_RESOURCE_FORK 4/* Compressed data in resource fork. */ /* * HFS+ compression resource fork. */ #define RSRC_H_SIZE 260 /* Base size of Resource fork header. */ #define RSRC_F_SIZE 50 /* Size of Resource fork footer. */ /* Size to write compressed data to resource fork. */ #define COMPRESSED_W_SIZE (64 * 1024) /* decmpfs difinitions. */ #define MAX_DECMPFS_XATTR_SIZE 3802 #ifndef DECMPFS_XATTR_NAME #define DECMPFS_XATTR_NAME "com.apple.decmpfs" #endif #define DECMPFS_MAGIC 0x636d7066 #define DECMPFS_COMPRESSION_MAGIC 0 #define DECMPFS_COMPRESSION_TYPE 4 #define DECMPFS_UNCOMPRESSED_SIZE 8 #define DECMPFS_HEADER_SIZE 16 #define HFS_BLOCKS(s) ((s) >> 12) static int check_symlinks(struct archive_write_disk *); static int create_filesystem_object(struct archive_write_disk *); static struct fixup_entry *current_fixup(struct archive_write_disk *, const char *pathname); #if defined(HAVE_FCHDIR) && defined(PATH_MAX) static void edit_deep_directories(struct archive_write_disk *ad); #endif static int cleanup_pathname(struct archive_write_disk *); static int create_dir(struct archive_write_disk *, char *); static int create_parent_dir(struct archive_write_disk *, char *); static ssize_t hfs_write_data_block(struct archive_write_disk *, const char *, size_t); static int fixup_appledouble(struct archive_write_disk *, const char *); static int older(struct stat *, struct archive_entry *); static int restore_entry(struct archive_write_disk *); static int set_mac_metadata(struct archive_write_disk *, const char *, const void *, size_t); static int set_xattrs(struct archive_write_disk *); static int clear_nochange_fflags(struct archive_write_disk *); static int set_fflags(struct archive_write_disk *); static int set_fflags_platform(struct archive_write_disk *, int fd, const char *name, mode_t mode, unsigned long fflags_set, unsigned long fflags_clear); static int set_ownership(struct archive_write_disk *); static int set_mode(struct archive_write_disk *, int mode); static int set_time(int, int, const char *, time_t, long, time_t, long); static int set_times(struct archive_write_disk *, int, int, const char *, time_t, long, time_t, long, time_t, long, time_t, long); static int set_times_from_entry(struct archive_write_disk *); static struct fixup_entry *sort_dir_list(struct fixup_entry *p); static ssize_t write_data_block(struct archive_write_disk *, const char *, size_t); static struct archive_vtable *archive_write_disk_vtable(void); static int _archive_write_disk_close(struct archive *); static int _archive_write_disk_free(struct archive *); static int _archive_write_disk_header(struct archive *, struct archive_entry *); static int64_t _archive_write_disk_filter_bytes(struct archive *, int); static int _archive_write_disk_finish_entry(struct archive *); static ssize_t _archive_write_disk_data(struct archive *, const void *, size_t); static ssize_t _archive_write_disk_data_block(struct archive *, const void *, size_t, int64_t); static int lazy_stat(struct archive_write_disk *a) { if (a->pst != NULL) { /* Already have stat() data available. */ return (ARCHIVE_OK); } #ifdef HAVE_FSTAT if (a->fd >= 0 && fstat(a->fd, &a->st) == 0) { a->pst = &a->st; return (ARCHIVE_OK); } #endif /* * XXX At this point, symlinks should not be hit, otherwise * XXX a race occurred. Do we want to check explicitly for that? */ if (lstat(a->name, &a->st) == 0) { a->pst = &a->st; return (ARCHIVE_OK); } archive_set_error(&a->archive, errno, "Couldn't stat file"); return (ARCHIVE_WARN); } static struct archive_vtable * archive_write_disk_vtable(void) { static struct archive_vtable av; static int inited = 0; if (!inited) { av.archive_close = _archive_write_disk_close; av.archive_filter_bytes = _archive_write_disk_filter_bytes; av.archive_free = _archive_write_disk_free; av.archive_write_header = _archive_write_disk_header; av.archive_write_finish_entry = _archive_write_disk_finish_entry; av.archive_write_data = _archive_write_disk_data; av.archive_write_data_block = _archive_write_disk_data_block; inited = 1; } return (&av); } static int64_t _archive_write_disk_filter_bytes(struct archive *_a, int n) { struct archive_write_disk *a = (struct archive_write_disk *)_a; (void)n; /* UNUSED */ if (n == -1 || n == 0) return (a->total_bytes_written); return (-1); } int archive_write_disk_set_options(struct archive *_a, int flags) { struct archive_write_disk *a = (struct archive_write_disk *)_a; a->flags = flags; return (ARCHIVE_OK); } /* * Extract this entry to disk. * * TODO: Validate hardlinks. According to the standards, we're * supposed to check each extracted hardlink and squawk if it refers * to a file that we didn't restore. I'm not entirely convinced this * is a good idea, but more importantly: Is there any way to validate * hardlinks without keeping a complete list of filenames from the * entire archive?? Ugh. * */ static int _archive_write_disk_header(struct archive *_a, struct archive_entry *entry) { struct archive_write_disk *a = (struct archive_write_disk *)_a; struct fixup_entry *fe; int ret, r; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_write_disk_header"); archive_clear_error(&a->archive); if (a->archive.state & ARCHIVE_STATE_DATA) { r = _archive_write_disk_finish_entry(&a->archive); if (r == ARCHIVE_FATAL) return (r); } /* Set up for this particular entry. */ a->pst = NULL; a->current_fixup = NULL; a->deferred = 0; if (a->entry) { archive_entry_free(a->entry); a->entry = NULL; } a->entry = archive_entry_clone(entry); a->fd = -1; a->fd_offset = 0; a->offset = 0; a->restore_pwd = -1; a->uid = a->user_uid; a->mode = archive_entry_mode(a->entry); if (archive_entry_size_is_set(a->entry)) a->filesize = archive_entry_size(a->entry); else a->filesize = -1; archive_strcpy(&(a->_name_data), archive_entry_pathname(a->entry)); a->name = a->_name_data.s; archive_clear_error(&a->archive); /* * Clean up the requested path. This is necessary for correct * dir restores; the dir restore logic otherwise gets messed * up by nonsense like "dir/.". */ ret = cleanup_pathname(a); if (ret != ARCHIVE_OK) return (ret); /* * Query the umask so we get predictable mode settings. * This gets done on every call to _write_header in case the * user edits their umask during the extraction for some * reason. */ umask(a->user_umask = umask(0)); /* Figure out what we need to do for this entry. */ a->todo = TODO_MODE_BASE; if (a->flags & ARCHIVE_EXTRACT_PERM) { a->todo |= TODO_MODE_FORCE; /* Be pushy about permissions. */ /* * SGID requires an extra "check" step because we * cannot easily predict the GID that the system will * assign. (Different systems assign GIDs to files * based on a variety of criteria, including process * credentials and the gid of the enclosing * directory.) We can only restore the SGID bit if * the file has the right GID, and we only know the * GID if we either set it (see set_ownership) or if * we've actually called stat() on the file after it * was restored. Since there are several places at * which we might verify the GID, we need a TODO bit * to keep track. */ if (a->mode & S_ISGID) a->todo |= TODO_SGID | TODO_SGID_CHECK; /* * Verifying the SUID is simpler, but can still be * done in multiple ways, hence the separate "check" bit. */ if (a->mode & S_ISUID) a->todo |= TODO_SUID | TODO_SUID_CHECK; } else { /* * User didn't request full permissions, so don't * restore SUID, SGID bits and obey umask. */ a->mode &= ~S_ISUID; a->mode &= ~S_ISGID; a->mode &= ~S_ISVTX; a->mode &= ~a->user_umask; } if (a->flags & ARCHIVE_EXTRACT_OWNER) a->todo |= TODO_OWNER; if (a->flags & ARCHIVE_EXTRACT_TIME) a->todo |= TODO_TIMES; if (a->flags & ARCHIVE_EXTRACT_ACL) { if (archive_entry_filetype(a->entry) == AE_IFDIR) a->deferred |= TODO_ACLS; else a->todo |= TODO_ACLS; } if (a->flags & ARCHIVE_EXTRACT_MAC_METADATA) { if (archive_entry_filetype(a->entry) == AE_IFDIR) a->deferred |= TODO_MAC_METADATA; else a->todo |= TODO_MAC_METADATA; } #if defined(__APPLE__) && defined(UF_COMPRESSED) && defined(HAVE_ZLIB_H) if ((a->flags & ARCHIVE_EXTRACT_NO_HFS_COMPRESSION) == 0) { unsigned long set, clear; archive_entry_fflags(a->entry, &set, &clear); if ((set & ~clear) & UF_COMPRESSED) { a->todo |= TODO_HFS_COMPRESSION; a->decmpfs_block_count = (unsigned)-1; } } if ((a->flags & ARCHIVE_EXTRACT_HFS_COMPRESSION_FORCED) != 0 && (a->mode & AE_IFMT) == AE_IFREG && a->filesize > 0) { a->todo |= TODO_HFS_COMPRESSION; a->decmpfs_block_count = (unsigned)-1; } { const char *p; /* Check if the current file name is a type of the * resource fork file. */ p = strrchr(a->name, '/'); if (p == NULL) p = a->name; else p++; if (p[0] == '.' && p[1] == '_') { /* Do not compress "._XXX" files. */ a->todo &= ~TODO_HFS_COMPRESSION; if (a->filesize > 0) a->todo |= TODO_APPLEDOUBLE; } } #endif if (a->flags & ARCHIVE_EXTRACT_XATTR) a->todo |= TODO_XATTR; if (a->flags & ARCHIVE_EXTRACT_FFLAGS) a->todo |= TODO_FFLAGS; if (a->flags & ARCHIVE_EXTRACT_SECURE_SYMLINKS) { ret = check_symlinks(a); if (ret != ARCHIVE_OK) return (ret); } #if defined(HAVE_FCHDIR) && defined(PATH_MAX) /* If path exceeds PATH_MAX, shorten the path. */ edit_deep_directories(a); #endif ret = restore_entry(a); #if defined(__APPLE__) && defined(UF_COMPRESSED) && defined(HAVE_ZLIB_H) /* * Check if the filesystem the file is restoring on supports * HFS+ Compression. If not, cancel HFS+ Compression. */ if (a->todo | TODO_HFS_COMPRESSION) { /* * NOTE: UF_COMPRESSED is ignored even if the filesystem * supports HFS+ Compression because the file should * have at least an extended attriute "com.apple.decmpfs" * before the flag is set to indicate that the file have * been compressed. If hte filesystem does not support * HFS+ Compression the system call will fail. */ if (a->fd < 0 || fchflags(a->fd, UF_COMPRESSED) != 0) a->todo &= ~TODO_HFS_COMPRESSION; } #endif /* * TODO: There are rumours that some extended attributes must * be restored before file data is written. If this is true, * then we either need to write all extended attributes both * before and after restoring the data, or find some rule for * determining which must go first and which last. Due to the * many ways people are using xattrs, this may prove to be an * intractable problem. */ #ifdef HAVE_FCHDIR /* If we changed directory above, restore it here. */ if (a->restore_pwd >= 0) { r = fchdir(a->restore_pwd); if (r != 0) { archive_set_error(&a->archive, errno, "chdir() failure"); ret = ARCHIVE_FATAL; } close(a->restore_pwd); a->restore_pwd = -1; } #endif /* * Fixup uses the unedited pathname from archive_entry_pathname(), * because it is relative to the base dir and the edited path * might be relative to some intermediate dir as a result of the * deep restore logic. */ if (a->deferred & TODO_MODE) { fe = current_fixup(a, archive_entry_pathname(entry)); if (fe == NULL) return (ARCHIVE_FATAL); fe->fixup |= TODO_MODE_BASE; fe->mode = a->mode; } if ((a->deferred & TODO_TIMES) && (archive_entry_mtime_is_set(entry) || archive_entry_atime_is_set(entry))) { fe = current_fixup(a, archive_entry_pathname(entry)); if (fe == NULL) return (ARCHIVE_FATAL); fe->mode = a->mode; fe->fixup |= TODO_TIMES; if (archive_entry_atime_is_set(entry)) { fe->atime = archive_entry_atime(entry); fe->atime_nanos = archive_entry_atime_nsec(entry); } else { /* If atime is unset, use start time. */ fe->atime = a->start_time; fe->atime_nanos = 0; } if (archive_entry_mtime_is_set(entry)) { fe->mtime = archive_entry_mtime(entry); fe->mtime_nanos = archive_entry_mtime_nsec(entry); } else { /* If mtime is unset, use start time. */ fe->mtime = a->start_time; fe->mtime_nanos = 0; } if (archive_entry_birthtime_is_set(entry)) { fe->birthtime = archive_entry_birthtime(entry); fe->birthtime_nanos = archive_entry_birthtime_nsec(entry); } else { /* If birthtime is unset, use mtime. */ fe->birthtime = fe->mtime; fe->birthtime_nanos = fe->mtime_nanos; } } if (a->deferred & TODO_ACLS) { fe = current_fixup(a, archive_entry_pathname(entry)); if (fe == NULL) return (ARCHIVE_FATAL); fe->fixup |= TODO_ACLS; archive_acl_copy(&fe->acl, archive_entry_acl(entry)); } if (a->deferred & TODO_MAC_METADATA) { const void *metadata; size_t metadata_size; metadata = archive_entry_mac_metadata(a->entry, &metadata_size); if (metadata != NULL && metadata_size > 0) { fe = current_fixup(a, archive_entry_pathname(entry)); if (fe == NULL) return (ARCHIVE_FATAL); fe->mac_metadata = malloc(metadata_size); if (fe->mac_metadata != NULL) { memcpy(fe->mac_metadata, metadata, metadata_size); fe->mac_metadata_size = metadata_size; fe->fixup |= TODO_MAC_METADATA; } } } if (a->deferred & TODO_FFLAGS) { fe = current_fixup(a, archive_entry_pathname(entry)); if (fe == NULL) return (ARCHIVE_FATAL); fe->fixup |= TODO_FFLAGS; /* TODO: Complete this.. defer fflags from below. */ } /* We've created the object and are ready to pour data into it. */ if (ret >= ARCHIVE_WARN) a->archive.state = ARCHIVE_STATE_DATA; /* * If it's not open, tell our client not to try writing. * In particular, dirs, links, etc, don't get written to. */ if (a->fd < 0) { archive_entry_set_size(entry, 0); a->filesize = 0; } return (ret); } int archive_write_disk_set_skip_file(struct archive *_a, int64_t d, int64_t i) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_set_skip_file"); a->skip_file_set = 1; a->skip_file_dev = d; a->skip_file_ino = i; return (ARCHIVE_OK); } static ssize_t write_data_block(struct archive_write_disk *a, const char *buff, size_t size) { uint64_t start_size = size; ssize_t bytes_written = 0; ssize_t block_size = 0, bytes_to_write; if (size == 0) return (ARCHIVE_OK); if (a->filesize == 0 || a->fd < 0) { archive_set_error(&a->archive, 0, "Attempt to write to an empty file"); return (ARCHIVE_WARN); } if (a->flags & ARCHIVE_EXTRACT_SPARSE) { #if HAVE_STRUCT_STAT_ST_BLKSIZE int r; if ((r = lazy_stat(a)) != ARCHIVE_OK) return (r); block_size = a->pst->st_blksize; #else /* XXX TODO XXX Is there a more appropriate choice here ? */ /* This needn't match the filesystem allocation size. */ block_size = 16*1024; #endif } /* If this write would run beyond the file size, truncate it. */ if (a->filesize >= 0 && (int64_t)(a->offset + size) > a->filesize) start_size = size = (size_t)(a->filesize - a->offset); /* Write the data. */ while (size > 0) { if (block_size == 0) { bytes_to_write = size; } else { /* We're sparsifying the file. */ const char *p, *end; int64_t block_end; /* Skip leading zero bytes. */ for (p = buff, end = buff + size; p < end; ++p) { if (*p != '\0') break; } a->offset += p - buff; size -= p - buff; buff = p; if (size == 0) break; /* Calculate next block boundary after offset. */ block_end = (a->offset / block_size + 1) * block_size; /* If the adjusted write would cross block boundary, * truncate it to the block boundary. */ bytes_to_write = size; if (a->offset + bytes_to_write > block_end) bytes_to_write = block_end - a->offset; } /* Seek if necessary to the specified offset. */ if (a->offset != a->fd_offset) { if (lseek(a->fd, a->offset, SEEK_SET) < 0) { archive_set_error(&a->archive, errno, "Seek failed"); return (ARCHIVE_FATAL); } a->fd_offset = a->offset; } bytes_written = write(a->fd, buff, bytes_to_write); if (bytes_written < 0) { archive_set_error(&a->archive, errno, "Write failed"); return (ARCHIVE_WARN); } buff += bytes_written; size -= bytes_written; a->total_bytes_written += bytes_written; a->offset += bytes_written; a->fd_offset = a->offset; } return (start_size - size); } #if defined(__APPLE__) && defined(UF_COMPRESSED) && defined(HAVE_SYS_XATTR_H)\ && defined(HAVE_ZLIB_H) /* * Set UF_COMPRESSED file flag. * This have to be called after hfs_write_decmpfs() because if the * file does not have "com.apple.decmpfs" xattr the flag is ignored. */ static int hfs_set_compressed_fflag(struct archive_write_disk *a) { int r; if ((r = lazy_stat(a)) != ARCHIVE_OK) return (r); a->st.st_flags |= UF_COMPRESSED; if (fchflags(a->fd, a->st.st_flags) != 0) { archive_set_error(&a->archive, errno, "Failed to set UF_COMPRESSED file flag"); return (ARCHIVE_WARN); } return (ARCHIVE_OK); } /* * HFS+ Compression decmpfs * * +------------------------------+ +0 * | Magic(LE 4 bytes) | * +------------------------------+ * | Type(LE 4 bytes) | * +------------------------------+ * | Uncompressed size(LE 8 bytes)| * +------------------------------+ +16 * | | * | Compressed data | * | (Placed only if Type == 3) | * | | * +------------------------------+ +3802 = MAX_DECMPFS_XATTR_SIZE * * Type is 3: decmpfs has compressed data. * Type is 4: Resource Fork has compressed data. */ /* * Write "com.apple.decmpfs" */ static int hfs_write_decmpfs(struct archive_write_disk *a) { int r; uint32_t compression_type; r = fsetxattr(a->fd, DECMPFS_XATTR_NAME, a->decmpfs_header_p, a->decmpfs_attr_size, 0, 0); if (r < 0) { archive_set_error(&a->archive, errno, "Cannot restore xattr:%s", DECMPFS_XATTR_NAME); compression_type = archive_le32dec( &a->decmpfs_header_p[DECMPFS_COMPRESSION_TYPE]); if (compression_type == CMP_RESOURCE_FORK) fremovexattr(a->fd, XATTR_RESOURCEFORK_NAME, XATTR_SHOWCOMPRESSION); return (ARCHIVE_WARN); } return (ARCHIVE_OK); } /* * HFS+ Compression Resource Fork * * +-----------------------------+ * | Header(260 bytes) | * +-----------------------------+ * | Block count(LE 4 bytes) | * +-----------------------------+ --+ * +-- | Offset (LE 4 bytes) | | * | | [distance from Block count] | | Block 0 * | +-----------------------------+ | * | | Compressed size(LE 4 bytes) | | * | +-----------------------------+ --+ * | | | * | | .................. | * | | | * | +-----------------------------+ --+ * | | Offset (LE 4 bytes) | | * | +-----------------------------+ | Block (Block count -1) * | | Compressed size(LE 4 bytes) | | * +-> +-----------------------------+ --+ * | Compressed data(n bytes) | Block 0 * +-----------------------------+ * | | * | .................. | * | | * +-----------------------------+ * | Compressed data(n bytes) | Block (Block count -1) * +-----------------------------+ * | Footer(50 bytes) | * +-----------------------------+ * */ /* * Write the header of "com.apple.ResourceFork" */ static int hfs_write_resource_fork(struct archive_write_disk *a, unsigned char *buff, size_t bytes, uint32_t position) { int ret; ret = fsetxattr(a->fd, XATTR_RESOURCEFORK_NAME, buff, bytes, position, a->rsrc_xattr_options); if (ret < 0) { archive_set_error(&a->archive, errno, "Cannot restore xattr: %s at %u pos %u bytes", XATTR_RESOURCEFORK_NAME, (unsigned)position, (unsigned)bytes); return (ARCHIVE_WARN); } a->rsrc_xattr_options &= ~XATTR_CREATE; return (ARCHIVE_OK); } static int hfs_write_compressed_data(struct archive_write_disk *a, size_t bytes_compressed) { int ret; ret = hfs_write_resource_fork(a, a->compressed_buffer, bytes_compressed, a->compressed_rsrc_position); if (ret == ARCHIVE_OK) a->compressed_rsrc_position += bytes_compressed; return (ret); } static int hfs_write_resource_fork_header(struct archive_write_disk *a) { unsigned char *buff; uint32_t rsrc_bytes; uint32_t rsrc_header_bytes; /* * Write resource fork header + block info. */ buff = a->resource_fork; rsrc_bytes = a->compressed_rsrc_position - RSRC_F_SIZE; rsrc_header_bytes = RSRC_H_SIZE + /* Header base size. */ 4 + /* Block count. */ (a->decmpfs_block_count * 8);/* Block info */ archive_be32enc(buff, 0x100); archive_be32enc(buff + 4, rsrc_bytes); archive_be32enc(buff + 8, rsrc_bytes - 256); archive_be32enc(buff + 12, 0x32); memset(buff + 16, 0, 240); archive_be32enc(buff + 256, rsrc_bytes - 260); return hfs_write_resource_fork(a, buff, rsrc_header_bytes, 0); } static size_t hfs_set_resource_fork_footer(unsigned char *buff, size_t buff_size) { static const char rsrc_footer[RSRC_F_SIZE] = { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x32, 0x00, 0x00, 'c', 'm', 'p', 'f', 0x00, 0x00, 0x00, 0x0a, 0x00, 0x01, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 }; if (buff_size < sizeof(rsrc_footer)) return (0); memcpy(buff, rsrc_footer, sizeof(rsrc_footer)); return (sizeof(rsrc_footer)); } static int hfs_reset_compressor(struct archive_write_disk *a) { int ret; if (a->stream_valid) ret = deflateReset(&a->stream); else ret = deflateInit(&a->stream, a->decmpfs_compression_level); if (ret != Z_OK) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Failed to initialize compressor"); return (ARCHIVE_FATAL); } else a->stream_valid = 1; return (ARCHIVE_OK); } static int hfs_decompress(struct archive_write_disk *a) { uint32_t *block_info; unsigned int block_count; uint32_t data_pos, data_size; ssize_t r; ssize_t bytes_written, bytes_to_write; unsigned char *b; block_info = (uint32_t *)(a->resource_fork + RSRC_H_SIZE); block_count = archive_le32dec(block_info++); while (block_count--) { data_pos = RSRC_H_SIZE + archive_le32dec(block_info++); data_size = archive_le32dec(block_info++); r = fgetxattr(a->fd, XATTR_RESOURCEFORK_NAME, a->compressed_buffer, data_size, data_pos, 0); if (r != data_size) { archive_set_error(&a->archive, (r < 0)?errno:ARCHIVE_ERRNO_MISC, "Failed to read resource fork"); return (ARCHIVE_WARN); } if (a->compressed_buffer[0] == 0xff) { bytes_to_write = data_size -1; b = a->compressed_buffer + 1; } else { uLong dest_len = MAX_DECMPFS_BLOCK_SIZE; int zr; zr = uncompress((Bytef *)a->uncompressed_buffer, &dest_len, a->compressed_buffer, data_size); if (zr != Z_OK) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Failed to decompress resource fork"); return (ARCHIVE_WARN); } bytes_to_write = dest_len; b = (unsigned char *)a->uncompressed_buffer; } do { bytes_written = write(a->fd, b, bytes_to_write); if (bytes_written < 0) { archive_set_error(&a->archive, errno, "Write failed"); return (ARCHIVE_WARN); } bytes_to_write -= bytes_written; b += bytes_written; } while (bytes_to_write > 0); } r = fremovexattr(a->fd, XATTR_RESOURCEFORK_NAME, 0); if (r == -1) { archive_set_error(&a->archive, errno, "Failed to remove resource fork"); return (ARCHIVE_WARN); } return (ARCHIVE_OK); } static int hfs_drive_compressor(struct archive_write_disk *a, const char *buff, size_t size) { unsigned char *buffer_compressed; size_t bytes_compressed; size_t bytes_used; int ret; ret = hfs_reset_compressor(a); if (ret != ARCHIVE_OK) return (ret); if (a->compressed_buffer == NULL) { size_t block_size; block_size = COMPRESSED_W_SIZE + RSRC_F_SIZE + + compressBound(MAX_DECMPFS_BLOCK_SIZE); a->compressed_buffer = malloc(block_size); if (a->compressed_buffer == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for Resource Fork"); return (ARCHIVE_FATAL); } a->compressed_buffer_size = block_size; a->compressed_buffer_remaining = block_size; } buffer_compressed = a->compressed_buffer + a->compressed_buffer_size - a->compressed_buffer_remaining; a->stream.next_in = (Bytef *)(uintptr_t)(const void *)buff; a->stream.avail_in = size; a->stream.next_out = buffer_compressed; a->stream.avail_out = a->compressed_buffer_remaining; do { ret = deflate(&a->stream, Z_FINISH); switch (ret) { case Z_OK: case Z_STREAM_END: break; default: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Failed to compress data"); return (ARCHIVE_FAILED); } } while (ret == Z_OK); bytes_compressed = a->compressed_buffer_remaining - a->stream.avail_out; /* * If the compressed size is larger than the original size, * throw away compressed data, use uncompressed data instead. */ if (bytes_compressed > size) { buffer_compressed[0] = 0xFF;/* uncompressed marker. */ memcpy(buffer_compressed + 1, buff, size); bytes_compressed = size + 1; } a->compressed_buffer_remaining -= bytes_compressed; /* * If the compressed size is smaller than MAX_DECMPFS_XATTR_SIZE * and the block count in the file is only one, store compressed * data to decmpfs xattr instead of the resource fork. */ if (a->decmpfs_block_count == 1 && (a->decmpfs_attr_size + bytes_compressed) <= MAX_DECMPFS_XATTR_SIZE) { archive_le32enc(&a->decmpfs_header_p[DECMPFS_COMPRESSION_TYPE], CMP_XATTR); memcpy(a->decmpfs_header_p + DECMPFS_HEADER_SIZE, buffer_compressed, bytes_compressed); a->decmpfs_attr_size += bytes_compressed; a->compressed_buffer_remaining = a->compressed_buffer_size; /* * Finish HFS+ Compression. * - Write the decmpfs xattr. * - Set the UF_COMPRESSED file flag. */ ret = hfs_write_decmpfs(a); if (ret == ARCHIVE_OK) ret = hfs_set_compressed_fflag(a); return (ret); } /* Update block info. */ archive_le32enc(a->decmpfs_block_info++, a->compressed_rsrc_position_v - RSRC_H_SIZE); archive_le32enc(a->decmpfs_block_info++, bytes_compressed); a->compressed_rsrc_position_v += bytes_compressed; /* * Write the compressed data to the resource fork. */ bytes_used = a->compressed_buffer_size - a->compressed_buffer_remaining; while (bytes_used >= COMPRESSED_W_SIZE) { ret = hfs_write_compressed_data(a, COMPRESSED_W_SIZE); if (ret != ARCHIVE_OK) return (ret); bytes_used -= COMPRESSED_W_SIZE; if (bytes_used > COMPRESSED_W_SIZE) memmove(a->compressed_buffer, a->compressed_buffer + COMPRESSED_W_SIZE, bytes_used); else memcpy(a->compressed_buffer, a->compressed_buffer + COMPRESSED_W_SIZE, bytes_used); } a->compressed_buffer_remaining = a->compressed_buffer_size - bytes_used; /* * If the current block is the last block, write the remaining * compressed data and the resource fork footer. */ if (a->file_remaining_bytes == 0) { size_t rsrc_size; int64_t bk; /* Append the resource footer. */ rsrc_size = hfs_set_resource_fork_footer( a->compressed_buffer + bytes_used, a->compressed_buffer_remaining); ret = hfs_write_compressed_data(a, bytes_used + rsrc_size); a->compressed_buffer_remaining = a->compressed_buffer_size; /* If the compressed size is not enouph smaller than * the uncompressed size. cancel HFS+ compression. * TODO: study a behavior of ditto utility and improve * the condition to fall back into no HFS+ compression. */ bk = HFS_BLOCKS(a->compressed_rsrc_position); bk += bk >> 7; if (bk > HFS_BLOCKS(a->filesize)) return hfs_decompress(a); /* * Write the resourcefork header. */ if (ret == ARCHIVE_OK) ret = hfs_write_resource_fork_header(a); /* * Finish HFS+ Compression. * - Write the decmpfs xattr. * - Set the UF_COMPRESSED file flag. */ if (ret == ARCHIVE_OK) ret = hfs_write_decmpfs(a); if (ret == ARCHIVE_OK) ret = hfs_set_compressed_fflag(a); } return (ret); } static ssize_t hfs_write_decmpfs_block(struct archive_write_disk *a, const char *buff, size_t size) { const char *buffer_to_write; size_t bytes_to_write; int ret; if (a->decmpfs_block_count == (unsigned)-1) { void *new_block; size_t new_size; unsigned int block_count; if (a->decmpfs_header_p == NULL) { new_block = malloc(MAX_DECMPFS_XATTR_SIZE + sizeof(uint32_t)); if (new_block == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for decmpfs"); return (ARCHIVE_FATAL); } a->decmpfs_header_p = new_block; } a->decmpfs_attr_size = DECMPFS_HEADER_SIZE; archive_le32enc(&a->decmpfs_header_p[DECMPFS_COMPRESSION_MAGIC], DECMPFS_MAGIC); archive_le32enc(&a->decmpfs_header_p[DECMPFS_COMPRESSION_TYPE], CMP_RESOURCE_FORK); archive_le64enc(&a->decmpfs_header_p[DECMPFS_UNCOMPRESSED_SIZE], a->filesize); /* Calculate a block count of the file. */ block_count = (a->filesize + MAX_DECMPFS_BLOCK_SIZE -1) / MAX_DECMPFS_BLOCK_SIZE; /* * Allocate buffer for resource fork. * Set up related pointers; */ new_size = RSRC_H_SIZE + /* header */ 4 + /* Block count */ (block_count * sizeof(uint32_t) * 2) + RSRC_F_SIZE; /* footer */ if (new_size > a->resource_fork_allocated_size) { new_block = realloc(a->resource_fork, new_size); if (new_block == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for ResourceFork"); return (ARCHIVE_FATAL); } a->resource_fork_allocated_size = new_size; a->resource_fork = new_block; } /* Allocate uncompressed buffer */ if (a->uncompressed_buffer == NULL) { new_block = malloc(MAX_DECMPFS_BLOCK_SIZE); if (new_block == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for decmpfs"); return (ARCHIVE_FATAL); } a->uncompressed_buffer = new_block; } a->block_remaining_bytes = MAX_DECMPFS_BLOCK_SIZE; a->file_remaining_bytes = a->filesize; a->compressed_buffer_remaining = a->compressed_buffer_size; /* * Set up a resource fork. */ a->rsrc_xattr_options = XATTR_CREATE; /* Get the position where we are going to set a bunch * of block info. */ a->decmpfs_block_info = (uint32_t *)(a->resource_fork + RSRC_H_SIZE); /* Set the block count to the resource fork. */ archive_le32enc(a->decmpfs_block_info++, block_count); /* Get the position where we are goint to set compressed * data. */ a->compressed_rsrc_position = RSRC_H_SIZE + 4 + (block_count * 8); a->compressed_rsrc_position_v = a->compressed_rsrc_position; a->decmpfs_block_count = block_count; } /* Ignore redundant bytes. */ if (a->file_remaining_bytes == 0) return ((ssize_t)size); /* Do not overrun a block size. */ if (size > a->block_remaining_bytes) bytes_to_write = a->block_remaining_bytes; else bytes_to_write = size; /* Do not overrun the file size. */ if (bytes_to_write > a->file_remaining_bytes) bytes_to_write = a->file_remaining_bytes; /* For efficiency, if a copy length is full of the uncompressed * buffer size, do not copy writing data to it. */ if (bytes_to_write == MAX_DECMPFS_BLOCK_SIZE) buffer_to_write = buff; else { memcpy(a->uncompressed_buffer + MAX_DECMPFS_BLOCK_SIZE - a->block_remaining_bytes, buff, bytes_to_write); buffer_to_write = a->uncompressed_buffer; } a->block_remaining_bytes -= bytes_to_write; a->file_remaining_bytes -= bytes_to_write; if (a->block_remaining_bytes == 0 || a->file_remaining_bytes == 0) { ret = hfs_drive_compressor(a, buffer_to_write, MAX_DECMPFS_BLOCK_SIZE - a->block_remaining_bytes); if (ret < 0) return (ret); a->block_remaining_bytes = MAX_DECMPFS_BLOCK_SIZE; } /* Ignore redundant bytes. */ if (a->file_remaining_bytes == 0) return ((ssize_t)size); return (bytes_to_write); } static ssize_t hfs_write_data_block(struct archive_write_disk *a, const char *buff, size_t size) { uint64_t start_size = size; ssize_t bytes_written = 0; ssize_t bytes_to_write; if (size == 0) return (ARCHIVE_OK); if (a->filesize == 0 || a->fd < 0) { archive_set_error(&a->archive, 0, "Attempt to write to an empty file"); return (ARCHIVE_WARN); } /* If this write would run beyond the file size, truncate it. */ if (a->filesize >= 0 && (int64_t)(a->offset + size) > a->filesize) start_size = size = (size_t)(a->filesize - a->offset); /* Write the data. */ while (size > 0) { bytes_to_write = size; /* Seek if necessary to the specified offset. */ if (a->offset < a->fd_offset) { /* Can't support backword move. */ archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Seek failed"); return (ARCHIVE_FATAL); } else if (a->offset > a->fd_offset) { int64_t skip = a->offset - a->fd_offset; char nullblock[1024]; memset(nullblock, 0, sizeof(nullblock)); while (skip > 0) { if (skip > (int64_t)sizeof(nullblock)) bytes_written = hfs_write_decmpfs_block( a, nullblock, sizeof(nullblock)); else bytes_written = hfs_write_decmpfs_block( a, nullblock, skip); if (bytes_written < 0) { archive_set_error(&a->archive, errno, "Write failed"); return (ARCHIVE_WARN); } skip -= bytes_written; } a->fd_offset = a->offset; } bytes_written = hfs_write_decmpfs_block(a, buff, bytes_to_write); if (bytes_written < 0) return (bytes_written); buff += bytes_written; size -= bytes_written; a->total_bytes_written += bytes_written; a->offset += bytes_written; a->fd_offset = a->offset; } return (start_size - size); } #else static ssize_t hfs_write_data_block(struct archive_write_disk *a, const char *buff, size_t size) { return (write_data_block(a, buff, size)); } #endif static ssize_t _archive_write_disk_data_block(struct archive *_a, const void *buff, size_t size, int64_t offset) { struct archive_write_disk *a = (struct archive_write_disk *)_a; ssize_t r; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_write_data_block"); a->offset = offset; if (a->todo & TODO_HFS_COMPRESSION) r = hfs_write_data_block(a, buff, size); else r = write_data_block(a, buff, size); if (r < ARCHIVE_OK) return (r); if ((size_t)r < size) { archive_set_error(&a->archive, 0, "Too much data: Truncating file at %ju bytes", (uintmax_t)a->filesize); return (ARCHIVE_WARN); } #if ARCHIVE_VERSION_NUMBER < 3999000 return (ARCHIVE_OK); #else return (size); #endif } static ssize_t _archive_write_disk_data(struct archive *_a, const void *buff, size_t size) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_write_data"); if (a->todo & TODO_HFS_COMPRESSION) return (hfs_write_data_block(a, buff, size)); return (write_data_block(a, buff, size)); } static int _archive_write_disk_finish_entry(struct archive *_a) { struct archive_write_disk *a = (struct archive_write_disk *)_a; int ret = ARCHIVE_OK; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_write_finish_entry"); if (a->archive.state & ARCHIVE_STATE_HEADER) return (ARCHIVE_OK); archive_clear_error(&a->archive); /* Pad or truncate file to the right size. */ if (a->fd < 0) { /* There's no file. */ } else if (a->filesize < 0) { /* File size is unknown, so we can't set the size. */ } else if (a->fd_offset == a->filesize) { /* Last write ended at exactly the filesize; we're done. */ /* Hopefully, this is the common case. */ #if defined(__APPLE__) && defined(UF_COMPRESSED) && defined(HAVE_ZLIB_H) } else if (a->todo & TODO_HFS_COMPRESSION) { char null_d[1024]; ssize_t r; if (a->file_remaining_bytes) memset(null_d, 0, sizeof(null_d)); while (a->file_remaining_bytes) { if (a->file_remaining_bytes > sizeof(null_d)) r = hfs_write_data_block( a, null_d, sizeof(null_d)); else r = hfs_write_data_block( a, null_d, a->file_remaining_bytes); if (r < 0) return ((int)r); } #endif } else { #if HAVE_FTRUNCATE if (ftruncate(a->fd, a->filesize) == -1 && a->filesize == 0) { archive_set_error(&a->archive, errno, "File size could not be restored"); return (ARCHIVE_FAILED); } #endif /* * Not all platforms implement the XSI option to * extend files via ftruncate. Stat() the file again * to see what happened. */ a->pst = NULL; if ((ret = lazy_stat(a)) != ARCHIVE_OK) return (ret); /* We can use lseek()/write() to extend the file if * ftruncate didn't work or isn't available. */ if (a->st.st_size < a->filesize) { const char nul = '\0'; if (lseek(a->fd, a->filesize - 1, SEEK_SET) < 0) { archive_set_error(&a->archive, errno, "Seek failed"); return (ARCHIVE_FATAL); } if (write(a->fd, &nul, 1) < 0) { archive_set_error(&a->archive, errno, "Write to restore size failed"); return (ARCHIVE_FATAL); } a->pst = NULL; } } /* Restore metadata. */ /* * This is specific to Mac OS X. * If the current file is an AppleDouble file, it should be * linked with the data fork file and remove it. */ if (a->todo & TODO_APPLEDOUBLE) { int r2 = fixup_appledouble(a, a->name); if (r2 == ARCHIVE_EOF) { /* The current file has been successfully linked * with the data fork file and removed. So there * is nothing to do on the current file. */ goto finish_metadata; } if (r2 < ret) ret = r2; } /* * Look up the "real" UID only if we're going to need it. * TODO: the TODO_SGID condition can be dropped here, can't it? */ if (a->todo & (TODO_OWNER | TODO_SUID | TODO_SGID)) { a->uid = archive_write_disk_uid(&a->archive, archive_entry_uname(a->entry), archive_entry_uid(a->entry)); } /* Look up the "real" GID only if we're going to need it. */ /* TODO: the TODO_SUID condition can be dropped here, can't it? */ if (a->todo & (TODO_OWNER | TODO_SGID | TODO_SUID)) { a->gid = archive_write_disk_gid(&a->archive, archive_entry_gname(a->entry), archive_entry_gid(a->entry)); } /* * Restore ownership before set_mode tries to restore suid/sgid * bits. If we set the owner, we know what it is and can skip * a stat() call to examine the ownership of the file on disk. */ if (a->todo & TODO_OWNER) { int r2 = set_ownership(a); if (r2 < ret) ret = r2; } /* * set_mode must precede ACLs on systems such as Solaris and * FreeBSD where setting the mode implicitly clears extended ACLs */ if (a->todo & TODO_MODE) { int r2 = set_mode(a, a->mode); if (r2 < ret) ret = r2; } /* * Security-related extended attributes (such as * security.capability on Linux) have to be restored last, * since they're implicitly removed by other file changes. */ if (a->todo & TODO_XATTR) { int r2 = set_xattrs(a); if (r2 < ret) ret = r2; } /* * Some flags prevent file modification; they must be restored after * file contents are written. */ if (a->todo & TODO_FFLAGS) { int r2 = set_fflags(a); if (r2 < ret) ret = r2; } /* * Time must follow most other metadata; * otherwise atime will get changed. */ if (a->todo & TODO_TIMES) { int r2 = set_times_from_entry(a); if (r2 < ret) ret = r2; } /* * Mac extended metadata includes ACLs. */ if (a->todo & TODO_MAC_METADATA) { const void *metadata; size_t metadata_size; metadata = archive_entry_mac_metadata(a->entry, &metadata_size); if (metadata != NULL && metadata_size > 0) { int r2 = set_mac_metadata(a, archive_entry_pathname( a->entry), metadata, metadata_size); if (r2 < ret) ret = r2; } } /* * ACLs must be restored after timestamps because there are * ACLs that prevent attribute changes (including time). */ if (a->todo & TODO_ACLS) { int r2 = archive_write_disk_set_acls(&a->archive, a->fd, archive_entry_pathname(a->entry), archive_entry_acl(a->entry)); if (r2 < ret) ret = r2; } finish_metadata: /* If there's an fd, we can close it now. */ if (a->fd >= 0) { close(a->fd); a->fd = -1; } /* If there's an entry, we can release it now. */ if (a->entry) { archive_entry_free(a->entry); a->entry = NULL; } a->archive.state = ARCHIVE_STATE_HEADER; return (ret); } int archive_write_disk_set_group_lookup(struct archive *_a, void *private_data, int64_t (*lookup_gid)(void *private, const char *gname, int64_t gid), void (*cleanup_gid)(void *private)) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_set_group_lookup"); if (a->cleanup_gid != NULL && a->lookup_gid_data != NULL) (a->cleanup_gid)(a->lookup_gid_data); a->lookup_gid = lookup_gid; a->cleanup_gid = cleanup_gid; a->lookup_gid_data = private_data; return (ARCHIVE_OK); } int archive_write_disk_set_user_lookup(struct archive *_a, void *private_data, int64_t (*lookup_uid)(void *private, const char *uname, int64_t uid), void (*cleanup_uid)(void *private)) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_set_user_lookup"); if (a->cleanup_uid != NULL && a->lookup_uid_data != NULL) (a->cleanup_uid)(a->lookup_uid_data); a->lookup_uid = lookup_uid; a->cleanup_uid = cleanup_uid; a->lookup_uid_data = private_data; return (ARCHIVE_OK); } int64_t archive_write_disk_gid(struct archive *_a, const char *name, int64_t id) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_gid"); if (a->lookup_gid) return (a->lookup_gid)(a->lookup_gid_data, name, id); return (id); } int64_t archive_write_disk_uid(struct archive *_a, const char *name, int64_t id) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_uid"); if (a->lookup_uid) return (a->lookup_uid)(a->lookup_uid_data, name, id); return (id); } /* * Create a new archive_write_disk object and initialize it with global state. */ struct archive * archive_write_disk_new(void) { struct archive_write_disk *a; a = (struct archive_write_disk *)malloc(sizeof(*a)); if (a == NULL) return (NULL); memset(a, 0, sizeof(*a)); a->archive.magic = ARCHIVE_WRITE_DISK_MAGIC; /* We're ready to write a header immediately. */ a->archive.state = ARCHIVE_STATE_HEADER; a->archive.vtable = archive_write_disk_vtable(); a->start_time = time(NULL); /* Query and restore the umask. */ umask(a->user_umask = umask(0)); #ifdef HAVE_GETEUID a->user_uid = geteuid(); #endif /* HAVE_GETEUID */ if (archive_string_ensure(&a->path_safe, 512) == NULL) { free(a); return (NULL); } #ifdef HAVE_ZLIB_H a->decmpfs_compression_level = 5; #endif return (&a->archive); } /* * If pathname is longer than PATH_MAX, chdir to a suitable * intermediate dir and edit the path down to a shorter suffix. Note * that this routine never returns an error; if the chdir() attempt * fails for any reason, we just go ahead with the long pathname. The * object creation is likely to fail, but any error will get handled * at that time. */ #if defined(HAVE_FCHDIR) && defined(PATH_MAX) static void edit_deep_directories(struct archive_write_disk *a) { int ret; char *tail = a->name; /* If path is short, avoid the open() below. */ if (strlen(tail) <= PATH_MAX) return; /* Try to record our starting dir. */ a->restore_pwd = open(".", O_RDONLY | O_BINARY | O_CLOEXEC); __archive_ensure_cloexec_flag(a->restore_pwd); if (a->restore_pwd < 0) return; /* As long as the path is too long... */ while (strlen(tail) > PATH_MAX) { /* Locate a dir prefix shorter than PATH_MAX. */ tail += PATH_MAX - 8; while (tail > a->name && *tail != '/') tail--; /* Exit if we find a too-long path component. */ if (tail <= a->name) return; /* Create the intermediate dir and chdir to it. */ *tail = '\0'; /* Terminate dir portion */ ret = create_dir(a, a->name); if (ret == ARCHIVE_OK && chdir(a->name) != 0) ret = ARCHIVE_FAILED; *tail = '/'; /* Restore the / we removed. */ if (ret != ARCHIVE_OK) return; tail++; /* The chdir() succeeded; we've now shortened the path. */ a->name = tail; } return; } #endif /* * The main restore function. */ static int restore_entry(struct archive_write_disk *a) { int ret = ARCHIVE_OK, en; if (a->flags & ARCHIVE_EXTRACT_UNLINK && !S_ISDIR(a->mode)) { /* * TODO: Fix this. Apparently, there are platforms * that still allow root to hose the entire filesystem * by unlinking a dir. The S_ISDIR() test above * prevents us from using unlink() here if the new * object is a dir, but that doesn't mean the old * object isn't a dir. */ if (a->flags & ARCHIVE_EXTRACT_CLEAR_NOCHANGE_FFLAGS) (void)clear_nochange_fflags(a); if (unlink(a->name) == 0) { /* We removed it, reset cached stat. */ a->pst = NULL; } else if (errno == ENOENT) { /* File didn't exist, that's just as good. */ } else if (rmdir(a->name) == 0) { /* It was a dir, but now it's gone. */ a->pst = NULL; } else { /* We tried, but couldn't get rid of it. */ archive_set_error(&a->archive, errno, "Could not unlink"); return(ARCHIVE_FAILED); } } /* Try creating it first; if this fails, we'll try to recover. */ en = create_filesystem_object(a); if ((en == ENOTDIR || en == ENOENT) && !(a->flags & ARCHIVE_EXTRACT_NO_AUTODIR)) { /* If the parent dir doesn't exist, try creating it. */ create_parent_dir(a, a->name); /* Now try to create the object again. */ en = create_filesystem_object(a); } if ((en == ENOENT) && (archive_entry_hardlink(a->entry) != NULL)) { archive_set_error(&a->archive, en, "Hard-link target '%s' does not exist.", archive_entry_hardlink(a->entry)); return (ARCHIVE_FAILED); } if ((en == EISDIR || en == EEXIST) && (a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE)) { /* If we're not overwriting, we're done. */ archive_entry_unset_size(a->entry); return (ARCHIVE_OK); } /* * Some platforms return EISDIR if you call * open(O_WRONLY | O_EXCL | O_CREAT) on a directory, some * return EEXIST. POSIX is ambiguous, requiring EISDIR * for open(O_WRONLY) on a dir and EEXIST for open(O_EXCL | O_CREAT) * on an existing item. */ if (en == EISDIR) { /* A dir is in the way of a non-dir, rmdir it. */ if (rmdir(a->name) != 0) { archive_set_error(&a->archive, errno, "Can't remove already-existing dir"); return (ARCHIVE_FAILED); } a->pst = NULL; /* Try again. */ en = create_filesystem_object(a); } else if (en == EEXIST) { /* * We know something is in the way, but we don't know what; * we need to find out before we go any further. */ int r = 0; /* * The SECURE_SYMLINKS logic has already removed a * symlink to a dir if the client wants that. So * follow the symlink if we're creating a dir. */ if (S_ISDIR(a->mode)) r = stat(a->name, &a->st); /* * If it's not a dir (or it's a broken symlink), * then don't follow it. */ if (r != 0 || !S_ISDIR(a->mode)) r = lstat(a->name, &a->st); if (r != 0) { archive_set_error(&a->archive, errno, "Can't stat existing object"); return (ARCHIVE_FAILED); } /* * NO_OVERWRITE_NEWER doesn't apply to directories. */ if ((a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE_NEWER) && !S_ISDIR(a->st.st_mode)) { if (!older(&(a->st), a->entry)) { archive_entry_unset_size(a->entry); return (ARCHIVE_OK); } } /* If it's our archive, we're done. */ if (a->skip_file_set && a->st.st_dev == (dev_t)a->skip_file_dev && a->st.st_ino == (ino_t)a->skip_file_ino) { archive_set_error(&a->archive, 0, "Refusing to overwrite archive"); return (ARCHIVE_FAILED); } if (!S_ISDIR(a->st.st_mode)) { /* A non-dir is in the way, unlink it. */ if (a->flags & ARCHIVE_EXTRACT_CLEAR_NOCHANGE_FFLAGS) (void)clear_nochange_fflags(a); if (unlink(a->name) != 0) { archive_set_error(&a->archive, errno, "Can't unlink already-existing object"); return (ARCHIVE_FAILED); } a->pst = NULL; /* Try again. */ en = create_filesystem_object(a); } else if (!S_ISDIR(a->mode)) { /* A dir is in the way of a non-dir, rmdir it. */ if (a->flags & ARCHIVE_EXTRACT_CLEAR_NOCHANGE_FFLAGS) (void)clear_nochange_fflags(a); if (rmdir(a->name) != 0) { archive_set_error(&a->archive, errno, "Can't replace existing directory with non-directory"); return (ARCHIVE_FAILED); } /* Try again. */ en = create_filesystem_object(a); } else { /* * There's a dir in the way of a dir. Don't * waste time with rmdir()/mkdir(), just fix * up the permissions on the existing dir. * Note that we don't change perms on existing * dirs unless _EXTRACT_PERM is specified. */ if ((a->mode != a->st.st_mode) && (a->todo & TODO_MODE_FORCE)) a->deferred |= (a->todo & TODO_MODE); /* Ownership doesn't need deferred fixup. */ en = 0; /* Forget the EEXIST. */ } } if (en) { /* Everything failed; give up here. */ archive_set_error(&a->archive, en, "Can't create '%s'", a->name); return (ARCHIVE_FAILED); } a->pst = NULL; /* Cached stat data no longer valid. */ return (ret); } /* * Returns 0 if creation succeeds, or else returns errno value from * the failed system call. Note: This function should only ever perform * a single system call. */ static int create_filesystem_object(struct archive_write_disk *a) { /* Create the entry. */ const char *linkname; mode_t final_mode, mode; int r; /* We identify hard/symlinks according to the link names. */ /* Since link(2) and symlink(2) don't handle modes, we're done here. */ linkname = archive_entry_hardlink(a->entry); if (linkname != NULL) { #if !HAVE_LINK return (EPERM); #else r = link(linkname, a->name) ? errno : 0; /* * New cpio and pax formats allow hardlink entries * to carry data, so we may have to open the file * for hardlink entries. * * If the hardlink was successfully created and * the archive doesn't have carry data for it, * consider it to be non-authoritative for meta data. * This is consistent with GNU tar and BSD pax. * If the hardlink does carry data, let the last * archive entry decide ownership. */ if (r == 0 && a->filesize <= 0) { a->todo = 0; a->deferred = 0; } else if (r == 0 && a->filesize > 0) { a->fd = open(a->name, O_WRONLY | O_TRUNC | O_BINARY | O_CLOEXEC); __archive_ensure_cloexec_flag(a->fd); if (a->fd < 0) r = errno; } return (r); #endif } linkname = archive_entry_symlink(a->entry); if (linkname != NULL) { #if HAVE_SYMLINK return symlink(linkname, a->name) ? errno : 0; #else return (EPERM); #endif } /* * The remaining system calls all set permissions, so let's * try to take advantage of that to avoid an extra chmod() * call. (Recall that umask is set to zero right now!) */ /* Mode we want for the final restored object (w/o file type bits). */ final_mode = a->mode & 07777; /* * The mode that will actually be restored in this step. Note * that SUID, SGID, etc, require additional work to ensure * security, so we never restore them at this point. */ mode = final_mode & 0777 & ~a->user_umask; switch (a->mode & AE_IFMT) { default: /* POSIX requires that we fall through here. */ /* FALLTHROUGH */ case AE_IFREG: a->fd = open(a->name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY | O_CLOEXEC, mode); __archive_ensure_cloexec_flag(a->fd); r = (a->fd < 0); break; case AE_IFCHR: #ifdef HAVE_MKNOD /* Note: we use AE_IFCHR for the case label, and * S_IFCHR for the mknod() call. This is correct. */ r = mknod(a->name, mode | S_IFCHR, archive_entry_rdev(a->entry)); break; #else /* TODO: Find a better way to warn about our inability * to restore a char device node. */ return (EINVAL); #endif /* HAVE_MKNOD */ case AE_IFBLK: #ifdef HAVE_MKNOD r = mknod(a->name, mode | S_IFBLK, archive_entry_rdev(a->entry)); break; #else /* TODO: Find a better way to warn about our inability * to restore a block device node. */ return (EINVAL); #endif /* HAVE_MKNOD */ case AE_IFDIR: mode = (mode | MINIMUM_DIR_MODE) & MAXIMUM_DIR_MODE; r = mkdir(a->name, mode); if (r == 0) { /* Defer setting dir times. */ a->deferred |= (a->todo & TODO_TIMES); a->todo &= ~TODO_TIMES; /* Never use an immediate chmod(). */ /* We can't avoid the chmod() entirely if EXTRACT_PERM * because of SysV SGID inheritance. */ if ((mode != final_mode) || (a->flags & ARCHIVE_EXTRACT_PERM)) a->deferred |= (a->todo & TODO_MODE); a->todo &= ~TODO_MODE; } break; case AE_IFIFO: #ifdef HAVE_MKFIFO r = mkfifo(a->name, mode); break; #else /* TODO: Find a better way to warn about our inability * to restore a fifo. */ return (EINVAL); #endif /* HAVE_MKFIFO */ } /* All the system calls above set errno on failure. */ if (r) return (errno); /* If we managed to set the final mode, we've avoided a chmod(). */ if (mode == final_mode) a->todo &= ~TODO_MODE; return (0); } /* * Cleanup function for archive_extract. Mostly, this involves processing * the fixup list, which is used to address a number of problems: * * Dir permissions might prevent us from restoring a file in that * dir, so we restore the dir with minimum 0700 permissions first, * then correct the mode at the end. * * Similarly, the act of restoring a file touches the directory * and changes the timestamp on the dir, so we have to touch-up dir * timestamps at the end as well. * * Some file flags can interfere with the restore by, for example, * preventing the creation of hardlinks to those files. * * Mac OS extended metadata includes ACLs, so must be deferred on dirs. * * Note that tar/cpio do not require that archives be in a particular * order; there is no way to know when the last file has been restored * within a directory, so there's no way to optimize the memory usage * here by fixing up the directory any earlier than the * end-of-archive. * * XXX TODO: Directory ACLs should be restored here, for the same * reason we set directory perms here. XXX */ static int _archive_write_disk_close(struct archive *_a) { struct archive_write_disk *a = (struct archive_write_disk *)_a; struct fixup_entry *next, *p; int ret; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_write_disk_close"); ret = _archive_write_disk_finish_entry(&a->archive); /* Sort dir list so directories are fixed up in depth-first order. */ p = sort_dir_list(a->fixup_list); while (p != NULL) { a->pst = NULL; /* Mark stat cache as out-of-date. */ if (p->fixup & TODO_TIMES) { set_times(a, -1, p->mode, p->name, p->atime, p->atime_nanos, p->birthtime, p->birthtime_nanos, p->mtime, p->mtime_nanos, p->ctime, p->ctime_nanos); } if (p->fixup & TODO_MODE_BASE) chmod(p->name, p->mode); if (p->fixup & TODO_ACLS) archive_write_disk_set_acls(&a->archive, -1, p->name, &p->acl); if (p->fixup & TODO_FFLAGS) set_fflags_platform(a, -1, p->name, p->mode, p->fflags_set, 0); if (p->fixup & TODO_MAC_METADATA) set_mac_metadata(a, p->name, p->mac_metadata, p->mac_metadata_size); next = p->next; archive_acl_clear(&p->acl); free(p->mac_metadata); free(p->name); free(p); p = next; } a->fixup_list = NULL; return (ret); } static int _archive_write_disk_free(struct archive *_a) { struct archive_write_disk *a; int ret; if (_a == NULL) return (ARCHIVE_OK); archive_check_magic(_a, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_write_disk_free"); a = (struct archive_write_disk *)_a; ret = _archive_write_disk_close(&a->archive); archive_write_disk_set_group_lookup(&a->archive, NULL, NULL, NULL); archive_write_disk_set_user_lookup(&a->archive, NULL, NULL, NULL); if (a->entry) archive_entry_free(a->entry); archive_string_free(&a->_name_data); archive_string_free(&a->archive.error_string); archive_string_free(&a->path_safe); a->archive.magic = 0; __archive_clean(&a->archive); free(a->decmpfs_header_p); free(a->resource_fork); free(a->compressed_buffer); free(a->uncompressed_buffer); #if defined(__APPLE__) && defined(UF_COMPRESSED) && defined(HAVE_SYS_XATTR_H)\ && defined(HAVE_ZLIB_H) if (a->stream_valid) { switch (deflateEnd(&a->stream)) { case Z_OK: break; default: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Failed to clean up compressor"); ret = ARCHIVE_FATAL; break; } } #endif free(a); return (ret); } /* * Simple O(n log n) merge sort to order the fixup list. In * particular, we want to restore dir timestamps depth-first. */ static struct fixup_entry * sort_dir_list(struct fixup_entry *p) { struct fixup_entry *a, *b, *t; if (p == NULL) return (NULL); /* A one-item list is already sorted. */ if (p->next == NULL) return (p); /* Step 1: split the list. */ t = p; a = p->next->next; while (a != NULL) { /* Step a twice, t once. */ a = a->next; if (a != NULL) a = a->next; t = t->next; } /* Now, t is at the mid-point, so break the list here. */ b = t->next; t->next = NULL; a = p; /* Step 2: Recursively sort the two sub-lists. */ a = sort_dir_list(a); b = sort_dir_list(b); /* Step 3: Merge the returned lists. */ /* Pick the first element for the merged list. */ if (strcmp(a->name, b->name) > 0) { t = p = a; a = a->next; } else { t = p = b; b = b->next; } /* Always put the later element on the list first. */ while (a != NULL && b != NULL) { if (strcmp(a->name, b->name) > 0) { t->next = a; a = a->next; } else { t->next = b; b = b->next; } t = t->next; } /* Only one list is non-empty, so just splice it on. */ if (a != NULL) t->next = a; if (b != NULL) t->next = b; return (p); } /* * Returns a new, initialized fixup entry. * * TODO: Reduce the memory requirements for this list by using a tree * structure rather than a simple list of names. */ static struct fixup_entry * new_fixup(struct archive_write_disk *a, const char *pathname) { struct fixup_entry *fe; fe = (struct fixup_entry *)calloc(1, sizeof(struct fixup_entry)); if (fe == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for a fixup"); return (NULL); } fe->next = a->fixup_list; a->fixup_list = fe; fe->fixup = 0; fe->name = strdup(pathname); return (fe); } /* * Returns a fixup structure for the current entry. */ static struct fixup_entry * current_fixup(struct archive_write_disk *a, const char *pathname) { if (a->current_fixup == NULL) a->current_fixup = new_fixup(a, pathname); return (a->current_fixup); } /* TODO: Make this work. */ /* * TODO: The deep-directory support bypasses this; disable deep directory * support if we're doing symlink checks. */ /* * TODO: Someday, integrate this with the deep dir support; they both * scan the path and both can be optimized by comparing against other * recent paths. */ /* TODO: Extend this to support symlinks on Windows Vista and later. */ static int check_symlinks(struct archive_write_disk *a) { #if !defined(HAVE_LSTAT) /* Platform doesn't have lstat, so we can't look for symlinks. */ (void)a; /* UNUSED */ return (ARCHIVE_OK); #else char *pn; char c; int r; struct stat st; /* * Guard against symlink tricks. Reject any archive entry whose * destination would be altered by a symlink. */ /* Whatever we checked last time doesn't need to be re-checked. */ pn = a->name; if (archive_strlen(&(a->path_safe)) > 0) { char *p = a->path_safe.s; while ((*pn != '\0') && (*p == *pn)) ++p, ++pn; } /* Skip the root directory if the path is absolute. */ if(pn == a->name && pn[0] == '/') ++pn; c = pn[0]; /* Keep going until we've checked the entire name. */ while (pn[0] != '\0' && (pn[0] != '/' || pn[1] != '\0')) { /* Skip the next path element. */ while (*pn != '\0' && *pn != '/') ++pn; c = pn[0]; pn[0] = '\0'; /* Check that we haven't hit a symlink. */ r = lstat(a->name, &st); if (r != 0) { /* We've hit a dir that doesn't exist; stop now. */ if (errno == ENOENT) break; } else if (S_ISLNK(st.st_mode)) { if (c == '\0') { /* * Last element is symlink; remove it * so we can overwrite it with the * item being extracted. */ if (unlink(a->name)) { archive_set_error(&a->archive, errno, "Could not remove symlink %s", a->name); pn[0] = c; return (ARCHIVE_FAILED); } a->pst = NULL; /* * Even if we did remove it, a warning * is in order. The warning is silly, * though, if we're just replacing one * symlink with another symlink. */ if (!S_ISLNK(a->mode)) { archive_set_error(&a->archive, 0, "Removing symlink %s", a->name); } /* Symlink gone. No more problem! */ pn[0] = c; return (0); } else if (a->flags & ARCHIVE_EXTRACT_UNLINK) { /* User asked us to remove problems. */ if (unlink(a->name) != 0) { archive_set_error(&a->archive, 0, "Cannot remove intervening symlink %s", a->name); pn[0] = c; return (ARCHIVE_FAILED); } a->pst = NULL; } else { archive_set_error(&a->archive, 0, "Cannot extract through symlink %s", a->name); pn[0] = c; return (ARCHIVE_FAILED); } } pn[0] = c; if (pn[0] != '\0') pn++; /* Advance to the next segment. */ } pn[0] = c; /* We've checked and/or cleaned the whole path, so remember it. */ archive_strcpy(&a->path_safe, a->name); return (ARCHIVE_OK); #endif } #if defined(__CYGWIN__) /* * 1. Convert a path separator from '\' to '/' . * We shouldn't check multibyte character directly because some * character-set have been using the '\' character for a part of * its multibyte character code. * 2. Replace unusable characters in Windows with underscore('_'). * See also : http://msdn.microsoft.com/en-us/library/aa365247.aspx */ static void cleanup_pathname_win(struct archive_write_disk *a) { wchar_t wc; char *p; size_t alen, l; int mb, complete, utf8; alen = 0; mb = 0; complete = 1; utf8 = (strcmp(nl_langinfo(CODESET), "UTF-8") == 0)? 1: 0; for (p = a->name; *p != '\0'; p++) { ++alen; if (*p == '\\') { /* If previous byte is smaller than 128, * this is not second byte of multibyte characters, * so we can replace '\' with '/'. */ if (utf8 || !mb) *p = '/'; else complete = 0;/* uncompleted. */ } else if (*(unsigned char *)p > 127) mb = 1; else mb = 0; /* Rewrite the path name if its next character is unusable. */ if (*p == ':' || *p == '*' || *p == '?' || *p == '"' || *p == '<' || *p == '>' || *p == '|') *p = '_'; } if (complete) return; /* * Convert path separator in wide-character. */ p = a->name; while (*p != '\0' && alen) { l = mbtowc(&wc, p, alen); if (l == (size_t)-1) { while (*p != '\0') { if (*p == '\\') *p = '/'; ++p; } break; } if (l == 1 && wc == L'\\') *p = '/'; p += l; alen -= l; } } #endif /* * Canonicalize the pathname. In particular, this strips duplicate * '/' characters, '.' elements, and trailing '/'. It also raises an * error for an empty path, a trailing '..', (if _SECURE_NODOTDOT is * set) any '..' in the path or (if ARCHIVE_EXTRACT_SECURE_NOABSOLUTEPATHS * is set) if the path is absolute. */ static int cleanup_pathname(struct archive_write_disk *a) { char *dest, *src; char separator = '\0'; dest = src = a->name; if (*src == '\0') { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Invalid empty pathname"); return (ARCHIVE_FAILED); } #if defined(__CYGWIN__) cleanup_pathname_win(a); #endif /* Skip leading '/'. */ if (*src == '/') { if (a->flags & ARCHIVE_EXTRACT_SECURE_NOABSOLUTEPATHS) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Path is absolute"); return (ARCHIVE_FAILED); } separator = *src++; } /* Scan the pathname one element at a time. */ for (;;) { /* src points to first char after '/' */ if (src[0] == '\0') { break; } else if (src[0] == '/') { /* Found '//', ignore second one. */ src++; continue; } else if (src[0] == '.') { if (src[1] == '\0') { /* Ignore trailing '.' */ break; } else if (src[1] == '/') { /* Skip './'. */ src += 2; continue; } else if (src[1] == '.') { if (src[2] == '/' || src[2] == '\0') { /* Conditionally warn about '..' */ if (a->flags & ARCHIVE_EXTRACT_SECURE_NODOTDOT) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Path contains '..'"); return (ARCHIVE_FAILED); } } /* * Note: Under no circumstances do we * remove '..' elements. In * particular, restoring * '/foo/../bar/' should create the * 'foo' dir as a side-effect. */ } } /* Copy current element, including leading '/'. */ if (separator) *dest++ = '/'; while (*src != '\0' && *src != '/') { *dest++ = *src++; } if (*src == '\0') break; /* Skip '/' separator. */ separator = *src++; } /* * We've just copied zero or more path elements, not including the * final '/'. */ if (dest == a->name) { /* * Nothing got copied. The path must have been something * like '.' or '/' or './' or '/././././/./'. */ if (separator) *dest++ = '/'; else *dest++ = '.'; } /* Terminate the result. */ *dest = '\0'; return (ARCHIVE_OK); } /* * Create the parent directory of the specified path, assuming path * is already in mutable storage. */ static int create_parent_dir(struct archive_write_disk *a, char *path) { char *slash; int r; /* Remove tail element to obtain parent name. */ slash = strrchr(path, '/'); if (slash == NULL) return (ARCHIVE_OK); *slash = '\0'; r = create_dir(a, path); *slash = '/'; return (r); } /* * Create the specified dir, recursing to create parents as necessary. * * Returns ARCHIVE_OK if the path exists when we're done here. * Otherwise, returns ARCHIVE_FAILED. * Assumes path is in mutable storage; path is unchanged on exit. */ static int create_dir(struct archive_write_disk *a, char *path) { struct stat st; struct fixup_entry *le; char *slash, *base; mode_t mode_final, mode; int r; /* Check for special names and just skip them. */ slash = strrchr(path, '/'); if (slash == NULL) base = path; else base = slash + 1; if (base[0] == '\0' || (base[0] == '.' && base[1] == '\0') || (base[0] == '.' && base[1] == '.' && base[2] == '\0')) { /* Don't bother trying to create null path, '.', or '..'. */ if (slash != NULL) { *slash = '\0'; r = create_dir(a, path); *slash = '/'; return (r); } return (ARCHIVE_OK); } /* * Yes, this should be stat() and not lstat(). Using lstat() * here loses the ability to extract through symlinks. Also note * that this should not use the a->st cache. */ if (stat(path, &st) == 0) { if (S_ISDIR(st.st_mode)) return (ARCHIVE_OK); if ((a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE)) { archive_set_error(&a->archive, EEXIST, "Can't create directory '%s'", path); return (ARCHIVE_FAILED); } if (unlink(path) != 0) { archive_set_error(&a->archive, errno, "Can't create directory '%s': " "Conflicting file cannot be removed", path); return (ARCHIVE_FAILED); } } else if (errno != ENOENT && errno != ENOTDIR) { /* Stat failed? */ archive_set_error(&a->archive, errno, "Can't test directory '%s'", path); return (ARCHIVE_FAILED); } else if (slash != NULL) { *slash = '\0'; r = create_dir(a, path); *slash = '/'; if (r != ARCHIVE_OK) return (r); } /* * Mode we want for the final restored directory. Per POSIX, * implicitly-created dirs must be created obeying the umask. * There's no mention whether this is different for privileged * restores (which the rest of this code handles by pretending * umask=0). I've chosen here to always obey the user's umask for * implicit dirs, even if _EXTRACT_PERM was specified. */ mode_final = DEFAULT_DIR_MODE & ~a->user_umask; /* Mode we want on disk during the restore process. */ mode = mode_final; mode |= MINIMUM_DIR_MODE; mode &= MAXIMUM_DIR_MODE; if (mkdir(path, mode) == 0) { if (mode != mode_final) { le = new_fixup(a, path); if (le == NULL) return (ARCHIVE_FATAL); le->fixup |=TODO_MODE_BASE; le->mode = mode_final; } return (ARCHIVE_OK); } /* * Without the following check, a/b/../b/c/d fails at the * second visit to 'b', so 'd' can't be created. Note that we * don't add it to the fixup list here, as it's already been * added. */ if (stat(path, &st) == 0 && S_ISDIR(st.st_mode)) return (ARCHIVE_OK); archive_set_error(&a->archive, errno, "Failed to create dir '%s'", path); return (ARCHIVE_FAILED); } /* * Note: Although we can skip setting the user id if the desired user * id matches the current user, we cannot skip setting the group, as * many systems set the gid based on the containing directory. So * we have to perform a chown syscall if we want to set the SGID * bit. (The alternative is to stat() and then possibly chown(); it's * more efficient to skip the stat() and just always chown().) Note * that a successful chown() here clears the TODO_SGID_CHECK bit, which * allows set_mode to skip the stat() check for the GID. */ static int set_ownership(struct archive_write_disk *a) { #ifndef __CYGWIN__ /* unfortunately, on win32 there is no 'root' user with uid 0, so we just have to try the chown and see if it works */ /* If we know we can't change it, don't bother trying. */ if (a->user_uid != 0 && a->user_uid != a->uid) { archive_set_error(&a->archive, errno, "Can't set UID=%jd", (intmax_t)a->uid); return (ARCHIVE_WARN); } #endif #ifdef HAVE_FCHOWN /* If we have an fd, we can avoid a race. */ if (a->fd >= 0 && fchown(a->fd, a->uid, a->gid) == 0) { /* We've set owner and know uid/gid are correct. */ a->todo &= ~(TODO_OWNER | TODO_SGID_CHECK | TODO_SUID_CHECK); return (ARCHIVE_OK); } #endif /* We prefer lchown() but will use chown() if that's all we have. */ /* Of course, if we have neither, this will always fail. */ #ifdef HAVE_LCHOWN if (lchown(a->name, a->uid, a->gid) == 0) { /* We've set owner and know uid/gid are correct. */ a->todo &= ~(TODO_OWNER | TODO_SGID_CHECK | TODO_SUID_CHECK); return (ARCHIVE_OK); } #elif HAVE_CHOWN if (!S_ISLNK(a->mode) && chown(a->name, a->uid, a->gid) == 0) { /* We've set owner and know uid/gid are correct. */ a->todo &= ~(TODO_OWNER | TODO_SGID_CHECK | TODO_SUID_CHECK); return (ARCHIVE_OK); } #endif archive_set_error(&a->archive, errno, "Can't set user=%jd/group=%jd for %s", (intmax_t)a->uid, (intmax_t)a->gid, a->name); return (ARCHIVE_WARN); } /* * Note: Returns 0 on success, non-zero on failure. */ static int set_time(int fd, int mode, const char *name, time_t atime, long atime_nsec, time_t mtime, long mtime_nsec) { /* Select the best implementation for this platform. */ #if defined(HAVE_UTIMENSAT) && defined(HAVE_FUTIMENS) /* * utimensat() and futimens() are defined in * POSIX.1-2008. They support ns resolution and setting times * on fds and symlinks. */ struct timespec ts[2]; (void)mode; /* UNUSED */ ts[0].tv_sec = atime; ts[0].tv_nsec = atime_nsec; ts[1].tv_sec = mtime; ts[1].tv_nsec = mtime_nsec; if (fd >= 0) return futimens(fd, ts); return utimensat(AT_FDCWD, name, ts, AT_SYMLINK_NOFOLLOW); #elif HAVE_UTIMES /* * The utimes()-family functions support µs-resolution and * setting times fds and symlinks. utimes() is documented as * LEGACY by POSIX, futimes() and lutimes() are not described * in POSIX. */ struct timeval times[2]; times[0].tv_sec = atime; times[0].tv_usec = atime_nsec / 1000; times[1].tv_sec = mtime; times[1].tv_usec = mtime_nsec / 1000; #ifdef HAVE_FUTIMES if (fd >= 0) return (futimes(fd, times)); #else (void)fd; /* UNUSED */ #endif #ifdef HAVE_LUTIMES (void)mode; /* UNUSED */ return (lutimes(name, times)); #else if (S_ISLNK(mode)) return (0); return (utimes(name, times)); #endif #elif defined(HAVE_UTIME) /* * utime() is POSIX-standard but only supports 1s resolution and * does not support fds or symlinks. */ struct utimbuf times; (void)fd; /* UNUSED */ (void)name; /* UNUSED */ (void)atime_nsec; /* UNUSED */ (void)mtime_nsec; /* UNUSED */ times.actime = atime; times.modtime = mtime; if (S_ISLNK(mode)) return (ARCHIVE_OK); return (utime(name, ×)); #else /* * We don't know how to set the time on this platform. */ (void)fd; /* UNUSED */ (void)mode; /* UNUSED */ (void)name; /* UNUSED */ (void)atime_nsec; /* UNUSED */ (void)mtime_nsec; /* UNUSED */ return (ARCHIVE_WARN); #endif } #ifdef F_SETTIMES static int set_time_tru64(int fd, int mode, const char *name, time_t atime, long atime_nsec, time_t mtime, long mtime_nsec, time_t ctime, long ctime_nsec) { struct attr_timbuf tstamp; tstamp.atime.tv_sec = atime; tstamp.mtime.tv_sec = mtime; tstamp.ctime.tv_sec = ctime; #if defined (__hpux) && defined (__ia64) tstamp.atime.tv_nsec = atime_nsec; tstamp.mtime.tv_nsec = mtime_nsec; tstamp.ctime.tv_nsec = ctime_nsec; #else tstamp.atime.tv_usec = atime_nsec / 1000; tstamp.mtime.tv_usec = mtime_nsec / 1000; tstamp.ctime.tv_usec = ctime_nsec / 1000; #endif return (fcntl(fd,F_SETTIMES,&tstamp)); } #endif /* F_SETTIMES */ static int set_times(struct archive_write_disk *a, int fd, int mode, const char *name, time_t atime, long atime_nanos, time_t birthtime, long birthtime_nanos, time_t mtime, long mtime_nanos, time_t cctime, long ctime_nanos) { /* Note: set_time doesn't use libarchive return conventions! * It uses syscall conventions. So 0 here instead of ARCHIVE_OK. */ int r1 = 0, r2 = 0; #ifdef F_SETTIMES /* * on Tru64 try own fcntl first which can restore even the * ctime, fall back to default code path below if it fails * or if we are not running as root */ if (a->user_uid == 0 && set_time_tru64(fd, mode, name, atime, atime_nanos, mtime, mtime_nanos, cctime, ctime_nanos) == 0) { return (ARCHIVE_OK); } #else /* Tru64 */ (void)cctime; /* UNUSED */ (void)ctime_nanos; /* UNUSED */ #endif /* Tru64 */ #ifdef HAVE_STRUCT_STAT_ST_BIRTHTIME /* * If you have struct stat.st_birthtime, we assume BSD * birthtime semantics, in which {f,l,}utimes() updates * birthtime to earliest mtime. So we set the time twice, * first using the birthtime, then using the mtime. If * birthtime == mtime, this isn't necessary, so we skip it. * If birthtime > mtime, then this won't work, so we skip it. */ if (birthtime < mtime || (birthtime == mtime && birthtime_nanos < mtime_nanos)) r1 = set_time(fd, mode, name, atime, atime_nanos, birthtime, birthtime_nanos); #else (void)birthtime; /* UNUSED */ (void)birthtime_nanos; /* UNUSED */ #endif r2 = set_time(fd, mode, name, atime, atime_nanos, mtime, mtime_nanos); if (r1 != 0 || r2 != 0) { archive_set_error(&a->archive, errno, "Can't restore time"); return (ARCHIVE_WARN); } return (ARCHIVE_OK); } static int set_times_from_entry(struct archive_write_disk *a) { time_t atime, birthtime, mtime, cctime; long atime_nsec, birthtime_nsec, mtime_nsec, ctime_nsec; /* Suitable defaults. */ atime = birthtime = mtime = cctime = a->start_time; atime_nsec = birthtime_nsec = mtime_nsec = ctime_nsec = 0; /* If no time was provided, we're done. */ if (!archive_entry_atime_is_set(a->entry) #if HAVE_STRUCT_STAT_ST_BIRTHTIME && !archive_entry_birthtime_is_set(a->entry) #endif && !archive_entry_mtime_is_set(a->entry)) return (ARCHIVE_OK); if (archive_entry_atime_is_set(a->entry)) { atime = archive_entry_atime(a->entry); atime_nsec = archive_entry_atime_nsec(a->entry); } if (archive_entry_birthtime_is_set(a->entry)) { birthtime = archive_entry_birthtime(a->entry); birthtime_nsec = archive_entry_birthtime_nsec(a->entry); } if (archive_entry_mtime_is_set(a->entry)) { mtime = archive_entry_mtime(a->entry); mtime_nsec = archive_entry_mtime_nsec(a->entry); } if (archive_entry_ctime_is_set(a->entry)) { cctime = archive_entry_ctime(a->entry); ctime_nsec = archive_entry_ctime_nsec(a->entry); } return set_times(a, a->fd, a->mode, a->name, atime, atime_nsec, birthtime, birthtime_nsec, mtime, mtime_nsec, cctime, ctime_nsec); } static int set_mode(struct archive_write_disk *a, int mode) { int r = ARCHIVE_OK; mode &= 07777; /* Strip off file type bits. */ if (a->todo & TODO_SGID_CHECK) { /* * If we don't know the GID is right, we must stat() * to verify it. We can't just check the GID of this * process, since systems sometimes set GID from * the enclosing dir or based on ACLs. */ if ((r = lazy_stat(a)) != ARCHIVE_OK) return (r); if (a->pst->st_gid != a->gid) { mode &= ~ S_ISGID; if (a->flags & ARCHIVE_EXTRACT_OWNER) { /* * This is only an error if you * requested owner restore. If you * didn't, we'll try to restore * sgid/suid, but won't consider it a * problem if we can't. */ archive_set_error(&a->archive, -1, "Can't restore SGID bit"); r = ARCHIVE_WARN; } } /* While we're here, double-check the UID. */ if (a->pst->st_uid != a->uid && (a->todo & TODO_SUID)) { mode &= ~ S_ISUID; if (a->flags & ARCHIVE_EXTRACT_OWNER) { archive_set_error(&a->archive, -1, "Can't restore SUID bit"); r = ARCHIVE_WARN; } } a->todo &= ~TODO_SGID_CHECK; a->todo &= ~TODO_SUID_CHECK; } else if (a->todo & TODO_SUID_CHECK) { /* * If we don't know the UID is right, we can just check * the user, since all systems set the file UID from * the process UID. */ if (a->user_uid != a->uid) { mode &= ~ S_ISUID; if (a->flags & ARCHIVE_EXTRACT_OWNER) { archive_set_error(&a->archive, -1, "Can't make file SUID"); r = ARCHIVE_WARN; } } a->todo &= ~TODO_SUID_CHECK; } if (S_ISLNK(a->mode)) { #ifdef HAVE_LCHMOD /* * If this is a symlink, use lchmod(). If the * platform doesn't support lchmod(), just skip it. A * platform that doesn't provide a way to set * permissions on symlinks probably ignores * permissions on symlinks, so a failure here has no * impact. */ if (lchmod(a->name, mode) != 0) { switch (errno) { case ENOTSUP: case ENOSYS: #if ENOTSUP != EOPNOTSUPP case EOPNOTSUPP: #endif /* * if lchmod is defined but the platform * doesn't support it, silently ignore * error */ break; default: archive_set_error(&a->archive, errno, "Can't set permissions to 0%o", (int)mode); r = ARCHIVE_WARN; } } #endif } else if (!S_ISDIR(a->mode)) { /* * If it's not a symlink and not a dir, then use * fchmod() or chmod(), depending on whether we have * an fd. Dirs get their perms set during the * post-extract fixup, which is handled elsewhere. */ #ifdef HAVE_FCHMOD if (a->fd >= 0) { if (fchmod(a->fd, mode) != 0) { archive_set_error(&a->archive, errno, "Can't set permissions to 0%o", (int)mode); r = ARCHIVE_WARN; } } else #endif /* If this platform lacks fchmod(), then * we'll just use chmod(). */ if (chmod(a->name, mode) != 0) { archive_set_error(&a->archive, errno, "Can't set permissions to 0%o", (int)mode); r = ARCHIVE_WARN; } } return (r); } static int set_fflags(struct archive_write_disk *a) { struct fixup_entry *le; unsigned long set, clear; int r; int critical_flags; mode_t mode = archive_entry_mode(a->entry); /* * Make 'critical_flags' hold all file flags that can't be * immediately restored. For example, on BSD systems, * SF_IMMUTABLE prevents hardlinks from being created, so * should not be set until after any hardlinks are created. To * preserve some semblance of portability, this uses #ifdef * extensively. Ugly, but it works. * * Yes, Virginia, this does create a security race. It's mitigated * somewhat by the practice of creating dirs 0700 until the extract * is done, but it would be nice if we could do more than that. * People restoring critical file systems should be wary of * other programs that might try to muck with files as they're * being restored. */ /* Hopefully, the compiler will optimize this mess into a constant. */ critical_flags = 0; #ifdef SF_IMMUTABLE critical_flags |= SF_IMMUTABLE; #endif #ifdef UF_IMMUTABLE critical_flags |= UF_IMMUTABLE; #endif #ifdef SF_APPEND critical_flags |= SF_APPEND; #endif #ifdef UF_APPEND critical_flags |= UF_APPEND; #endif #ifdef EXT2_APPEND_FL critical_flags |= EXT2_APPEND_FL; #endif #ifdef EXT2_IMMUTABLE_FL critical_flags |= EXT2_IMMUTABLE_FL; #endif if (a->todo & TODO_FFLAGS) { archive_entry_fflags(a->entry, &set, &clear); /* * The first test encourages the compiler to eliminate * all of this if it's not necessary. */ if ((critical_flags != 0) && (set & critical_flags)) { le = current_fixup(a, a->name); if (le == NULL) return (ARCHIVE_FATAL); le->fixup |= TODO_FFLAGS; le->fflags_set = set; /* Store the mode if it's not already there. */ if ((le->fixup & TODO_MODE) == 0) le->mode = mode; } else { r = set_fflags_platform(a, a->fd, a->name, mode, set, clear); if (r != ARCHIVE_OK) return (r); } } return (ARCHIVE_OK); } static int clear_nochange_fflags(struct archive_write_disk *a) { int nochange_flags; mode_t mode = archive_entry_mode(a->entry); /* Hopefully, the compiler will optimize this mess into a constant. */ nochange_flags = 0; #ifdef SF_IMMUTABLE nochange_flags |= SF_IMMUTABLE; #endif #ifdef UF_IMMUTABLE nochange_flags |= UF_IMMUTABLE; #endif #ifdef SF_APPEND nochange_flags |= SF_APPEND; #endif #ifdef UF_APPEND nochange_flags |= UF_APPEND; #endif #ifdef EXT2_APPEND_FL nochange_flags |= EXT2_APPEND_FL; #endif #ifdef EXT2_IMMUTABLE_FL nochange_flags |= EXT2_IMMUTABLE_FL; #endif return (set_fflags_platform(a, a->fd, a->name, mode, 0, nochange_flags)); } #if ( defined(HAVE_LCHFLAGS) || defined(HAVE_CHFLAGS) || defined(HAVE_FCHFLAGS) ) && defined(HAVE_STRUCT_STAT_ST_FLAGS) /* * BSD reads flags using stat() and sets them with one of {f,l,}chflags() */ static int set_fflags_platform(struct archive_write_disk *a, int fd, const char *name, mode_t mode, unsigned long set, unsigned long clear) { int r; (void)mode; /* UNUSED */ if (set == 0 && clear == 0) return (ARCHIVE_OK); /* * XXX Is the stat here really necessary? Or can I just use * the 'set' flags directly? In particular, I'm not sure * about the correct approach if we're overwriting an existing * file that already has flags on it. XXX */ if ((r = lazy_stat(a)) != ARCHIVE_OK) return (r); a->st.st_flags &= ~clear; a->st.st_flags |= set; #ifdef HAVE_FCHFLAGS /* If platform has fchflags() and we were given an fd, use it. */ if (fd >= 0 && fchflags(fd, a->st.st_flags) == 0) return (ARCHIVE_OK); #endif /* * If we can't use the fd to set the flags, we'll use the * pathname to set flags. We prefer lchflags() but will use * chflags() if we must. */ #ifdef HAVE_LCHFLAGS if (lchflags(name, a->st.st_flags) == 0) return (ARCHIVE_OK); #elif defined(HAVE_CHFLAGS) if (S_ISLNK(a->st.st_mode)) { archive_set_error(&a->archive, errno, "Can't set file flags on symlink."); return (ARCHIVE_WARN); } if (chflags(name, a->st.st_flags) == 0) return (ARCHIVE_OK); #endif archive_set_error(&a->archive, errno, "Failed to set file flags"); return (ARCHIVE_WARN); } #elif defined(EXT2_IOC_GETFLAGS) && defined(EXT2_IOC_SETFLAGS) && defined(HAVE_WORKING_EXT2_IOC_GETFLAGS) /* * Linux uses ioctl() to read and write file flags. */ static int set_fflags_platform(struct archive_write_disk *a, int fd, const char *name, mode_t mode, unsigned long set, unsigned long clear) { int ret; int myfd = fd; int newflags, oldflags; int sf_mask = 0; if (set == 0 && clear == 0) return (ARCHIVE_OK); /* Only regular files and dirs can have flags. */ if (!S_ISREG(mode) && !S_ISDIR(mode)) return (ARCHIVE_OK); /* If we weren't given an fd, open it ourselves. */ if (myfd < 0) { myfd = open(name, O_RDONLY | O_NONBLOCK | O_BINARY | O_CLOEXEC); __archive_ensure_cloexec_flag(myfd); } if (myfd < 0) return (ARCHIVE_OK); /* * Linux has no define for the flags that are only settable by * the root user. This code may seem a little complex, but * there seem to be some Linux systems that lack these * defines. (?) The code below degrades reasonably gracefully * if sf_mask is incomplete. */ #ifdef EXT2_IMMUTABLE_FL sf_mask |= EXT2_IMMUTABLE_FL; #endif #ifdef EXT2_APPEND_FL sf_mask |= EXT2_APPEND_FL; #endif /* * XXX As above, this would be way simpler if we didn't have * to read the current flags from disk. XXX */ ret = ARCHIVE_OK; /* Read the current file flags. */ if (ioctl(myfd, EXT2_IOC_GETFLAGS, &oldflags) < 0) goto fail; /* Try setting the flags as given. */ newflags = (oldflags & ~clear) | set; if (ioctl(myfd, EXT2_IOC_SETFLAGS, &newflags) >= 0) goto cleanup; if (errno != EPERM) goto fail; /* If we couldn't set all the flags, try again with a subset. */ newflags &= ~sf_mask; oldflags &= sf_mask; newflags |= oldflags; if (ioctl(myfd, EXT2_IOC_SETFLAGS, &newflags) >= 0) goto cleanup; /* We couldn't set the flags, so report the failure. */ fail: archive_set_error(&a->archive, errno, "Failed to set file flags"); ret = ARCHIVE_WARN; cleanup: if (fd < 0) close(myfd); return (ret); } #else /* * Of course, some systems have neither BSD chflags() nor Linux' flags * support through ioctl(). */ static int set_fflags_platform(struct archive_write_disk *a, int fd, const char *name, mode_t mode, unsigned long set, unsigned long clear) { (void)a; /* UNUSED */ (void)fd; /* UNUSED */ (void)name; /* UNUSED */ (void)mode; /* UNUSED */ (void)set; /* UNUSED */ (void)clear; /* UNUSED */ return (ARCHIVE_OK); } #endif /* __linux */ #ifndef HAVE_COPYFILE_H /* Default is to simply drop Mac extended metadata. */ static int set_mac_metadata(struct archive_write_disk *a, const char *pathname, const void *metadata, size_t metadata_size) { (void)a; /* UNUSED */ (void)pathname; /* UNUSED */ (void)metadata; /* UNUSED */ (void)metadata_size; /* UNUSED */ return (ARCHIVE_OK); } static int fixup_appledouble(struct archive_write_disk *a, const char *pathname) { (void)a; /* UNUSED */ (void)pathname; /* UNUSED */ return (ARCHIVE_OK); } #else /* * On Mac OS, we use copyfile() to unpack the metadata and * apply it to the target file. */ #if defined(HAVE_SYS_XATTR_H) static int copy_xattrs(struct archive_write_disk *a, int tmpfd, int dffd) { ssize_t xattr_size; char *xattr_names = NULL, *xattr_val = NULL; int ret = ARCHIVE_OK, xattr_i; xattr_size = flistxattr(tmpfd, NULL, 0, 0); if (xattr_size == -1) { archive_set_error(&a->archive, errno, "Failed to read metadata(xattr)"); ret = ARCHIVE_WARN; goto exit_xattr; } xattr_names = malloc(xattr_size); if (xattr_names == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for metadata(xattr)"); ret = ARCHIVE_FATAL; goto exit_xattr; } xattr_size = flistxattr(tmpfd, xattr_names, xattr_size, 0); if (xattr_size == -1) { archive_set_error(&a->archive, errno, "Failed to read metadata(xattr)"); ret = ARCHIVE_WARN; goto exit_xattr; } for (xattr_i = 0; xattr_i < xattr_size; xattr_i += strlen(xattr_names + xattr_i) + 1) { char *xattr_val_saved; ssize_t s; int f; s = fgetxattr(tmpfd, xattr_names + xattr_i, NULL, 0, 0, 0); if (s == -1) { archive_set_error(&a->archive, errno, "Failed to get metadata(xattr)"); ret = ARCHIVE_WARN; goto exit_xattr; } xattr_val_saved = xattr_val; xattr_val = realloc(xattr_val, s); if (xattr_val == NULL) { archive_set_error(&a->archive, ENOMEM, "Failed to get metadata(xattr)"); ret = ARCHIVE_WARN; free(xattr_val_saved); goto exit_xattr; } s = fgetxattr(tmpfd, xattr_names + xattr_i, xattr_val, s, 0, 0); if (s == -1) { archive_set_error(&a->archive, errno, "Failed to get metadata(xattr)"); ret = ARCHIVE_WARN; goto exit_xattr; } f = fsetxattr(dffd, xattr_names + xattr_i, xattr_val, s, 0, 0); if (f == -1) { archive_set_error(&a->archive, errno, "Failed to get metadata(xattr)"); ret = ARCHIVE_WARN; goto exit_xattr; } } exit_xattr: free(xattr_names); free(xattr_val); return (ret); } #endif static int copy_acls(struct archive_write_disk *a, int tmpfd, int dffd) { +#ifndef HAVE_SYS_ACL_H + return 0; +#else acl_t acl, dfacl = NULL; int acl_r, ret = ARCHIVE_OK; acl = acl_get_fd(tmpfd); if (acl == NULL) { if (errno == ENOENT) /* There are not any ACLs. */ return (ret); archive_set_error(&a->archive, errno, "Failed to get metadata(acl)"); ret = ARCHIVE_WARN; goto exit_acl; } dfacl = acl_dup(acl); acl_r = acl_set_fd(dffd, dfacl); if (acl_r == -1) { archive_set_error(&a->archive, errno, "Failed to get metadata(acl)"); ret = ARCHIVE_WARN; goto exit_acl; } exit_acl: if (acl) acl_free(acl); if (dfacl) acl_free(dfacl); return (ret); +#endif } static int create_tempdatafork(struct archive_write_disk *a, const char *pathname) { struct archive_string tmpdatafork; int tmpfd; archive_string_init(&tmpdatafork); archive_strcpy(&tmpdatafork, "tar.md.XXXXXX"); tmpfd = mkstemp(tmpdatafork.s); if (tmpfd < 0) { archive_set_error(&a->archive, errno, "Failed to mkstemp"); archive_string_free(&tmpdatafork); return (-1); } if (copyfile(pathname, tmpdatafork.s, 0, COPYFILE_UNPACK | COPYFILE_NOFOLLOW | COPYFILE_ACL | COPYFILE_XATTR) < 0) { archive_set_error(&a->archive, errno, "Failed to restore metadata"); close(tmpfd); tmpfd = -1; } unlink(tmpdatafork.s); archive_string_free(&tmpdatafork); return (tmpfd); } static int copy_metadata(struct archive_write_disk *a, const char *metadata, const char *datafork, int datafork_compressed) { int ret = ARCHIVE_OK; if (datafork_compressed) { int dffd, tmpfd; tmpfd = create_tempdatafork(a, metadata); if (tmpfd == -1) return (ARCHIVE_WARN); /* * Do not open the data fork compressed by HFS+ compression * with at least a writing mode(O_RDWR or O_WRONLY). it * makes the data fork uncompressed. */ dffd = open(datafork, 0); if (dffd == -1) { archive_set_error(&a->archive, errno, "Failed to open the data fork for metadata"); close(tmpfd); return (ARCHIVE_WARN); } #if defined(HAVE_SYS_XATTR_H) ret = copy_xattrs(a, tmpfd, dffd); if (ret == ARCHIVE_OK) #endif ret = copy_acls(a, tmpfd, dffd); close(tmpfd); close(dffd); } else { if (copyfile(metadata, datafork, 0, COPYFILE_UNPACK | COPYFILE_NOFOLLOW | COPYFILE_ACL | COPYFILE_XATTR) < 0) { archive_set_error(&a->archive, errno, "Failed to restore metadata"); ret = ARCHIVE_WARN; } } return (ret); } static int set_mac_metadata(struct archive_write_disk *a, const char *pathname, const void *metadata, size_t metadata_size) { struct archive_string tmp; ssize_t written; int fd; int ret = ARCHIVE_OK; /* This would be simpler if copyfile() could just accept the * metadata as a block of memory; then we could sidestep this * silly dance of writing the data to disk just so that * copyfile() can read it back in again. */ archive_string_init(&tmp); archive_strcpy(&tmp, pathname); archive_strcat(&tmp, ".XXXXXX"); fd = mkstemp(tmp.s); if (fd < 0) { archive_set_error(&a->archive, errno, "Failed to restore metadata"); archive_string_free(&tmp); return (ARCHIVE_WARN); } written = write(fd, metadata, metadata_size); close(fd); if ((size_t)written != metadata_size) { archive_set_error(&a->archive, errno, "Failed to restore metadata"); ret = ARCHIVE_WARN; } else { int compressed; #if defined(UF_COMPRESSED) if ((a->todo & TODO_HFS_COMPRESSION) != 0 && (ret = lazy_stat(a)) == ARCHIVE_OK) compressed = a->st.st_flags & UF_COMPRESSED; else #endif compressed = 0; ret = copy_metadata(a, tmp.s, pathname, compressed); } unlink(tmp.s); archive_string_free(&tmp); return (ret); } static int fixup_appledouble(struct archive_write_disk *a, const char *pathname) { char buff[8]; struct stat st; const char *p; struct archive_string datafork; int fd = -1, ret = ARCHIVE_OK; archive_string_init(&datafork); /* Check if the current file name is a type of the resource * fork file. */ p = strrchr(pathname, '/'); if (p == NULL) p = pathname; else p++; if (p[0] != '.' || p[1] != '_') goto skip_appledouble; /* * Check if the data fork file exists. * * TODO: Check if this write disk object has handled it. */ archive_strncpy(&datafork, pathname, p - pathname); archive_strcat(&datafork, p + 2); if (lstat(datafork.s, &st) == -1 || (st.st_mode & AE_IFMT) != AE_IFREG) goto skip_appledouble; /* * Check if the file is in the AppleDouble form. */ fd = open(pathname, O_RDONLY | O_BINARY | O_CLOEXEC); __archive_ensure_cloexec_flag(fd); if (fd == -1) { archive_set_error(&a->archive, errno, "Failed to open a restoring file"); ret = ARCHIVE_WARN; goto skip_appledouble; } if (read(fd, buff, 8) == -1) { archive_set_error(&a->archive, errno, "Failed to read a restoring file"); close(fd); ret = ARCHIVE_WARN; goto skip_appledouble; } close(fd); /* Check AppleDouble Magic Code. */ if (archive_be32dec(buff) != 0x00051607) goto skip_appledouble; /* Check AppleDouble Version. */ if (archive_be32dec(buff+4) != 0x00020000) goto skip_appledouble; ret = copy_metadata(a, pathname, datafork.s, #if defined(UF_COMPRESSED) st.st_flags & UF_COMPRESSED); #else 0); #endif if (ret == ARCHIVE_OK) { unlink(pathname); ret = ARCHIVE_EOF; } skip_appledouble: archive_string_free(&datafork); return (ret); } #endif #if HAVE_LSETXATTR || HAVE_LSETEA /* * Restore extended attributes - Linux and AIX implementations: * AIX' ea interface is syntaxwise identical to the Linux xattr interface. */ static int set_xattrs(struct archive_write_disk *a) { struct archive_entry *entry = a->entry; static int warning_done = 0; int ret = ARCHIVE_OK; int i = archive_entry_xattr_reset(entry); while (i--) { const char *name; const void *value; size_t size; archive_entry_xattr_next(entry, &name, &value, &size); if (name != NULL && strncmp(name, "xfsroot.", 8) != 0 && strncmp(name, "system.", 7) != 0) { int e; #if HAVE_FSETXATTR if (a->fd >= 0) e = fsetxattr(a->fd, name, value, size, 0); else #elif HAVE_FSETEA if (a->fd >= 0) e = fsetea(a->fd, name, value, size, 0); else #endif { #if HAVE_LSETXATTR e = lsetxattr(archive_entry_pathname(entry), name, value, size, 0); #elif HAVE_LSETEA e = lsetea(archive_entry_pathname(entry), name, value, size, 0); #endif } if (e == -1) { if (errno == ENOTSUP || errno == ENOSYS) { if (!warning_done) { warning_done = 1; archive_set_error(&a->archive, errno, "Cannot restore extended " "attributes on this file " "system"); } } else archive_set_error(&a->archive, errno, "Failed to set extended attribute"); ret = ARCHIVE_WARN; } } else { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid extended attribute encountered"); ret = ARCHIVE_WARN; } } return (ret); } #elif HAVE_EXTATTR_SET_FILE && HAVE_DECL_EXTATTR_NAMESPACE_USER /* * Restore extended attributes - FreeBSD implementation */ static int set_xattrs(struct archive_write_disk *a) { struct archive_entry *entry = a->entry; static int warning_done = 0; int ret = ARCHIVE_OK; int i = archive_entry_xattr_reset(entry); while (i--) { const char *name; const void *value; size_t size; archive_entry_xattr_next(entry, &name, &value, &size); if (name != NULL) { ssize_t e; int namespace; if (strncmp(name, "user.", 5) == 0) { /* "user." attributes go to user namespace */ name += 5; namespace = EXTATTR_NAMESPACE_USER; } else { /* Warn about other extended attributes. */ archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Can't restore extended attribute ``%s''", name); ret = ARCHIVE_WARN; continue; } errno = 0; #if HAVE_EXTATTR_SET_FD if (a->fd >= 0) e = extattr_set_fd(a->fd, namespace, name, value, size); else #endif /* TODO: should we use extattr_set_link() instead? */ { e = extattr_set_file(archive_entry_pathname(entry), namespace, name, value, size); } if (e != (ssize_t)size) { if (errno == ENOTSUP || errno == ENOSYS) { if (!warning_done) { warning_done = 1; archive_set_error(&a->archive, errno, "Cannot restore extended " "attributes on this file " "system"); } } else { archive_set_error(&a->archive, errno, "Failed to set extended attribute"); } ret = ARCHIVE_WARN; } } } return (ret); } #else /* * Restore extended attributes - stub implementation for unsupported systems */ static int set_xattrs(struct archive_write_disk *a) { static int warning_done = 0; /* If there aren't any extended attributes, then it's okay not * to extract them, otherwise, issue a single warning. */ if (archive_entry_xattr_count(a->entry) != 0 && !warning_done) { warning_done = 1; archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Cannot restore extended attributes on this system"); return (ARCHIVE_WARN); } /* Warning was already emitted; suppress further warnings. */ return (ARCHIVE_OK); } #endif /* * Test if file on disk is older than entry. */ static int older(struct stat *st, struct archive_entry *entry) { /* First, test the seconds and return if we have a definite answer. */ /* Definitely older. */ if (st->st_mtime < archive_entry_mtime(entry)) return (1); /* Definitely younger. */ if (st->st_mtime > archive_entry_mtime(entry)) return (0); /* If this platform supports fractional seconds, try those. */ #if HAVE_STRUCT_STAT_ST_MTIMESPEC_TV_NSEC /* Definitely older. */ if (st->st_mtimespec.tv_nsec < archive_entry_mtime_nsec(entry)) return (1); #elif HAVE_STRUCT_STAT_ST_MTIM_TV_NSEC /* Definitely older. */ if (st->st_mtim.tv_nsec < archive_entry_mtime_nsec(entry)) return (1); #elif HAVE_STRUCT_STAT_ST_MTIME_N /* older. */ if (st->st_mtime_n < archive_entry_mtime_nsec(entry)) return (1); #elif HAVE_STRUCT_STAT_ST_UMTIME /* older. */ if (st->st_umtime * 1000 < archive_entry_mtime_nsec(entry)) return (1); #elif HAVE_STRUCT_STAT_ST_MTIME_USEC /* older. */ if (st->st_mtime_usec * 1000 < archive_entry_mtime_nsec(entry)) return (1); #else /* This system doesn't have high-res timestamps. */ #endif /* Same age or newer, so not older. */ return (0); } #endif /* !_WIN32 || __CYGWIN__ */ Index: stable/10/contrib/libarchive/libarchive/test/main.c =================================================================== --- stable/10/contrib/libarchive/libarchive/test/main.c (revision 302424) +++ stable/10/contrib/libarchive/libarchive/test/main.c (revision 302425) @@ -1,2991 +1,2996 @@ /* * Copyright (c) 2003-2009 Tim Kientzle * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "test.h" #include "test_utils.h" #ifdef HAVE_SYS_IOCTL_H #include #endif #ifdef HAVE_SYS_TIME_H #include #endif #include #ifdef HAVE_ICONV_H #include #endif /* * Some Linux distributions have both linux/ext2_fs.h and ext2fs/ext2_fs.h. * As the include guards don't agree, the order of include is important. */ #ifdef HAVE_LINUX_EXT2_FS_H #include /* for Linux file flags */ #endif #if defined(HAVE_EXT2FS_EXT2_FS_H) && !defined(__CYGWIN__) #include /* Linux file flags, broken on Cygwin */ #endif #include #include #ifdef HAVE_SIGNAL_H #include #endif #include #include /* * This same file is used pretty much verbatim for all test harnesses. * * The next few lines are the only differences. * TODO: Move this into a separate configuration header, have all test * suites share one copy of this file. */ __FBSDID("$FreeBSD$"); #define KNOWNREF "test_compat_gtar_1.tar.uu" #define ENVBASE "LIBARCHIVE" /* Prefix for environment variables. */ #undef PROGRAM /* Testing a library, not a program. */ #define LIBRARY "libarchive" #define EXTRA_DUMP(x) archive_error_string((struct archive *)(x)) #define EXTRA_ERRNO(x) archive_errno((struct archive *)(x)) #define EXTRA_VERSION archive_version_details() /* * * Windows support routines * * Note: Configuration is a tricky issue. Using HAVE_* feature macros * in the test harness is dangerous because they cover up * configuration errors. The classic example of this is omitting a * configure check. If libarchive and libarchive_test both look for * the same feature macro, such errors are hard to detect. Platform * macros (e.g., _WIN32 or __GNUC__) are a little better, but can * easily lead to very messy code. It's best to limit yourself * to only the most generic programming techniques in the test harness * and thus avoid conditionals altogether. Where that's not possible, * try to minimize conditionals by grouping platform-specific tests in * one place (e.g., test_acl_freebsd) or by adding new assert() * functions (e.g., assertMakeHardlink()) to cover up platform * differences. Platform-specific coding in libarchive_test is often * a symptom that some capability is missing from libarchive itself. */ #if defined(_WIN32) && !defined(__CYGWIN__) #include #include #include #ifndef F_OK #define F_OK (0) #endif #ifndef S_ISDIR #define S_ISDIR(m) ((m) & _S_IFDIR) #endif #ifndef S_ISREG #define S_ISREG(m) ((m) & _S_IFREG) #endif #if !defined(__BORLANDC__) #define access _access #undef chdir #define chdir _chdir #endif #ifndef fileno #define fileno _fileno #endif /*#define fstat _fstat64*/ #if !defined(__BORLANDC__) #define getcwd _getcwd #endif #define lstat stat /*#define lstat _stat64*/ /*#define stat _stat64*/ #define rmdir _rmdir #if !defined(__BORLANDC__) #define strdup _strdup #define umask _umask #endif #define int64_t __int64 #endif #if defined(HAVE__CrtSetReportMode) # include #endif /* Path to working directory for current test */ const char *testworkdir; #ifdef PROGRAM /* Pathname of exe to be tested. */ const char *testprogfile; /* Name of exe to use in printf-formatted command strings. */ /* On Windows, this includes leading/trailing quotes. */ const char *testprog; #endif #if defined(_WIN32) && !defined(__CYGWIN__) static void *GetFunctionKernel32(const char *); static int my_CreateSymbolicLinkA(const char *, const char *, int); static int my_CreateHardLinkA(const char *, const char *); static int my_GetFileInformationByName(const char *, BY_HANDLE_FILE_INFORMATION *); static void * GetFunctionKernel32(const char *name) { static HINSTANCE lib; static int set; if (!set) { set = 1; lib = LoadLibrary("kernel32.dll"); } if (lib == NULL) { fprintf(stderr, "Can't load kernel32.dll?!\n"); exit(1); } return (void *)GetProcAddress(lib, name); } static int my_CreateSymbolicLinkA(const char *linkname, const char *target, int flags) { static BOOLEAN (WINAPI *f)(LPCSTR, LPCSTR, DWORD); static int set; if (!set) { set = 1; f = GetFunctionKernel32("CreateSymbolicLinkA"); } return f == NULL ? 0 : (*f)(linkname, target, flags); } static int my_CreateHardLinkA(const char *linkname, const char *target) { static BOOLEAN (WINAPI *f)(LPCSTR, LPCSTR, LPSECURITY_ATTRIBUTES); static int set; if (!set) { set = 1; f = GetFunctionKernel32("CreateHardLinkA"); } return f == NULL ? 0 : (*f)(linkname, target, NULL); } static int my_GetFileInformationByName(const char *path, BY_HANDLE_FILE_INFORMATION *bhfi) { HANDLE h; int r; memset(bhfi, 0, sizeof(*bhfi)); h = CreateFile(path, FILE_READ_ATTRIBUTES, 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) return (0); r = GetFileInformationByHandle(h, bhfi); CloseHandle(h); return (r); } #endif #if defined(HAVE__CrtSetReportMode) && !defined(__WATCOMC__) static void invalid_parameter_handler(const wchar_t * expression, const wchar_t * function, const wchar_t * file, unsigned int line, uintptr_t pReserved) { /* nop */ } #endif /* * * OPTIONS FLAGS * */ /* Enable core dump on failure. */ static int dump_on_failure = 0; /* Default is to remove temp dirs and log data for successful tests. */ static int keep_temp_files = 0; /* Default is to run the specified tests once and report errors. */ static int until_failure = 0; /* Default is to just report pass/fail for each test. */ static int verbosity = 0; #define VERBOSITY_SUMMARY_ONLY -1 /* -q */ #define VERBOSITY_PASSFAIL 0 /* Default */ #define VERBOSITY_LIGHT_REPORT 1 /* -v */ #define VERBOSITY_FULL 2 /* -vv */ /* A few places generate even more output for verbosity > VERBOSITY_FULL, * mostly for debugging the test harness itself. */ /* Cumulative count of assertion failures. */ static int failures = 0; /* Cumulative count of reported skips. */ static int skips = 0; /* Cumulative count of assertions checked. */ static int assertions = 0; /* Directory where uuencoded reference files can be found. */ static const char *refdir; /* * Report log information selectively to console and/or disk log. */ static int log_console = 0; static FILE *logfile; static void vlogprintf(const char *fmt, va_list ap) { #ifdef va_copy va_list lfap; va_copy(lfap, ap); #endif if (log_console) vfprintf(stdout, fmt, ap); if (logfile != NULL) #ifdef va_copy vfprintf(logfile, fmt, lfap); va_end(lfap); #else vfprintf(logfile, fmt, ap); #endif } static void logprintf(const char *fmt, ...) { va_list ap; va_start(ap, fmt); vlogprintf(fmt, ap); va_end(ap); } /* Set up a message to display only if next assertion fails. */ static char msgbuff[4096]; static const char *msg, *nextmsg; void failure(const char *fmt, ...) { va_list ap; if (fmt == NULL) { nextmsg = NULL; } else { va_start(ap, fmt); vsprintf(msgbuff, fmt, ap); va_end(ap); nextmsg = msgbuff; } } /* * Copy arguments into file-local variables. * This was added to permit vararg assert() functions without needing * variadic wrapper macros. Turns out that the vararg capability is almost * never used, so almost all of the vararg assertions can be simplified * by removing the vararg capability and reworking the wrapper macro to * pass __FILE__, __LINE__ directly into the function instead of using * this hook. I suspect this machinery is used so rarely that we * would be better off just removing it entirely. That would simplify * the code here noticeably. */ static const char *skipping_filename; static int skipping_line; void skipping_setup(const char *filename, int line) { skipping_filename = filename; skipping_line = line; } /* Called at the beginning of each assert() function. */ static void assertion_count(const char *file, int line) { (void)file; /* UNUSED */ (void)line; /* UNUSED */ ++assertions; /* Proper handling of "failure()" message. */ msg = nextmsg; nextmsg = NULL; /* Uncomment to print file:line after every assertion. * Verbose, but occasionally useful in tracking down crashes. */ /* printf("Checked %s:%d\n", file, line); */ } /* * For each test source file, we remember how many times each * assertion was reported. Cleared before each new test, * used by test_summarize(). */ static struct line { int count; int skip; } failed_lines[10000]; const char *failed_filename; /* Count this failure, setup up log destination and handle initial report. */ static void failure_start(const char *filename, int line, const char *fmt, ...) { va_list ap; /* Record another failure for this line. */ ++failures; failed_filename = filename; failed_lines[line].count++; /* Determine whether to log header to console. */ switch (verbosity) { case VERBOSITY_LIGHT_REPORT: log_console = (failed_lines[line].count < 2); break; default: log_console = (verbosity >= VERBOSITY_FULL); } /* Log file:line header for this failure */ va_start(ap, fmt); #if _MSC_VER logprintf("%s(%d): ", filename, line); #else logprintf("%s:%d: ", filename, line); #endif vlogprintf(fmt, ap); va_end(ap); logprintf("\n"); if (msg != NULL && msg[0] != '\0') { logprintf(" Description: %s\n", msg); msg = NULL; } /* Determine whether to log details to console. */ if (verbosity == VERBOSITY_LIGHT_REPORT) log_console = 0; } /* Complete reporting of failed tests. */ /* * The 'extra' hook here is used by libarchive to include libarchive * error messages with assertion failures. It could also be used * to add strerror() output, for example. Just define the EXTRA_DUMP() * macro appropriately. */ static void failure_finish(void *extra) { (void)extra; /* UNUSED (maybe) */ #ifdef EXTRA_DUMP if (extra != NULL) { logprintf(" errno: %d\n", EXTRA_ERRNO(extra)); logprintf(" detail: %s\n", EXTRA_DUMP(extra)); } #endif if (dump_on_failure) { fprintf(stderr, " *** forcing core dump so failure can be debugged ***\n"); abort(); } } /* Inform user that we're skipping some checks. */ void test_skipping(const char *fmt, ...) { char buff[1024]; va_list ap; va_start(ap, fmt); vsprintf(buff, fmt, ap); va_end(ap); /* Use failure() message if set. */ msg = nextmsg; nextmsg = NULL; /* failure_start() isn't quite right, but is awfully convenient. */ failure_start(skipping_filename, skipping_line, "SKIPPING: %s", buff); --failures; /* Undo failures++ in failure_start() */ /* Don't failure_finish() here. */ /* Mark as skip, so doesn't count as failed test. */ failed_lines[skipping_line].skip = 1; ++skips; } /* * * ASSERTIONS * */ /* Generic assert() just displays the failed condition. */ int assertion_assert(const char *file, int line, int value, const char *condition, void *extra) { assertion_count(file, line); if (!value) { failure_start(file, line, "Assertion failed: %s", condition); failure_finish(extra); } return (value); } /* chdir() and report any errors */ int assertion_chdir(const char *file, int line, const char *pathname) { assertion_count(file, line); if (chdir(pathname) == 0) return (1); failure_start(file, line, "chdir(\"%s\")", pathname); failure_finish(NULL); return (0); } /* Verify two integers are equal. */ int assertion_equal_int(const char *file, int line, long long v1, const char *e1, long long v2, const char *e2, void *extra) { assertion_count(file, line); if (v1 == v2) return (1); failure_start(file, line, "%s != %s", e1, e2); logprintf(" %s=%lld (0x%llx, 0%llo)\n", e1, v1, v1, v1); logprintf(" %s=%lld (0x%llx, 0%llo)\n", e2, v2, v2, v2); failure_finish(extra); return (0); } /* * Utility to convert a single UTF-8 sequence. */ static int _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) { static const char utf8_count[256] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ }; int ch; int cnt; uint32_t wc; *pwc = 0; /* Sanity check. */ if (n == 0) return (0); /* * Decode 1-4 bytes depending on the value of the first byte. */ ch = (unsigned char)*s; if (ch == 0) return (0); /* Standard: return 0 for end-of-string. */ cnt = utf8_count[ch]; /* Invalide sequence or there are not plenty bytes. */ if (n < (size_t)cnt) return (-1); /* Make a Unicode code point from a single UTF-8 sequence. */ switch (cnt) { case 1: /* 1 byte sequence. */ *pwc = ch & 0x7f; return (cnt); case 2: /* 2 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) return (-1); *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); return (cnt); case 3: /* 3 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) return (-1); if ((s[2] & 0xc0) != 0x80) return (-1); wc = ((ch & 0x0f) << 12) | ((s[1] & 0x3f) << 6) | (s[2] & 0x3f); if (wc < 0x800) return (-1);/* Overlong sequence. */ break; case 4: /* 4 bytes sequence. */ if (n < 4) return (-1); if ((s[1] & 0xc0) != 0x80) return (-1); if ((s[2] & 0xc0) != 0x80) return (-1); if ((s[3] & 0xc0) != 0x80) return (-1); wc = ((ch & 0x07) << 18) | ((s[1] & 0x3f) << 12) | ((s[2] & 0x3f) << 6) | (s[3] & 0x3f); if (wc < 0x10000) return (-1);/* Overlong sequence. */ break; default: return (-1); } /* The code point larger than 0x10FFFF is not leagal * Unicode values. */ if (wc > 0x10FFFF) return (-1); /* Correctly gets a Unicode, returns used bytes. */ *pwc = wc; return (cnt); } static void strdump(const char *e, const char *p, int ewidth, int utf8) { const char *q = p; logprintf(" %*s = ", ewidth, e); if (p == NULL) { logprintf("NULL\n"); return; } logprintf("\""); while (*p != '\0') { unsigned int c = 0xff & *p++; switch (c) { case '\a': logprintf("\\a"); break; case '\b': logprintf("\\b"); break; case '\n': logprintf("\\n"); break; case '\r': logprintf("\\r"); break; default: if (c >= 32 && c < 127) logprintf("%c", c); else logprintf("\\x%02X", c); } } logprintf("\""); logprintf(" (length %d)", q == NULL ? -1 : (int)strlen(q)); /* * If the current string is UTF-8, dump its code points. */ if (utf8) { size_t len; uint32_t uc; int n; int cnt = 0; p = q; len = strlen(p); logprintf(" ["); while ((n = _utf8_to_unicode(&uc, p, len)) > 0) { if (p != q) logprintf(" "); logprintf("%04X", uc); p += n; len -= n; cnt++; } logprintf("]"); logprintf(" (count %d", cnt); if (n < 0) { logprintf(",unknown %d bytes", len); } logprintf(")"); } logprintf("\n"); } /* Verify two strings are equal, dump them if not. */ int assertion_equal_string(const char *file, int line, const char *v1, const char *e1, const char *v2, const char *e2, void *extra, int utf8) { int l1, l2; assertion_count(file, line); if (v1 == v2 || (v1 != NULL && v2 != NULL && strcmp(v1, v2) == 0)) return (1); failure_start(file, line, "%s != %s", e1, e2); l1 = (int)strlen(e1); l2 = (int)strlen(e2); if (l1 < l2) l1 = l2; strdump(e1, v1, l1, utf8); strdump(e2, v2, l1, utf8); failure_finish(extra); return (0); } static void wcsdump(const char *e, const wchar_t *w) { logprintf(" %s = ", e); if (w == NULL) { logprintf("(null)"); return; } logprintf("\""); while (*w != L'\0') { unsigned int c = *w++; if (c >= 32 && c < 127) logprintf("%c", c); else if (c < 256) logprintf("\\x%02X", c); else if (c < 0x10000) logprintf("\\u%04X", c); else logprintf("\\U%08X", c); } logprintf("\"\n"); } #ifndef HAVE_WCSCMP static int wcscmp(const wchar_t *s1, const wchar_t *s2) { while (*s1 == *s2++) { if (*s1++ == L'\0') return 0; } if (*s1 > *--s2) return 1; else return -1; } #endif /* Verify that two wide strings are equal, dump them if not. */ int assertion_equal_wstring(const char *file, int line, const wchar_t *v1, const char *e1, const wchar_t *v2, const char *e2, void *extra) { assertion_count(file, line); if (v1 == v2) return (1); if (v1 != NULL && v2 != NULL && wcscmp(v1, v2) == 0) return (1); failure_start(file, line, "%s != %s", e1, e2); wcsdump(e1, v1); wcsdump(e2, v2); failure_finish(extra); return (0); } /* * Pretty standard hexdump routine. As a bonus, if ref != NULL, then * any bytes in p that differ from ref will be highlighted with '_' * before and after the hex value. */ static void hexdump(const char *p, const char *ref, size_t l, size_t offset) { size_t i, j; char sep; if (p == NULL) { logprintf("(null)\n"); return; } for(i=0; i < l; i+=16) { logprintf("%04x", (unsigned)(i + offset)); sep = ' '; for (j = 0; j < 16 && i + j < l; j++) { if (ref != NULL && p[i + j] != ref[i + j]) sep = '_'; logprintf("%c%02x", sep, 0xff & (int)p[i+j]); if (ref != NULL && p[i + j] == ref[i + j]) sep = ' '; } for (; j < 16; j++) { logprintf("%c ", sep); sep = ' '; } logprintf("%c", sep); for (j=0; j < 16 && i + j < l; j++) { int c = p[i + j]; if (c >= ' ' && c <= 126) logprintf("%c", c); else logprintf("."); } logprintf("\n"); } } /* Verify that two blocks of memory are the same, display the first * block of differences if they're not. */ int assertion_equal_mem(const char *file, int line, const void *_v1, const char *e1, const void *_v2, const char *e2, size_t l, const char *ld, void *extra) { const char *v1 = (const char *)_v1; const char *v2 = (const char *)_v2; size_t offset; assertion_count(file, line); if (v1 == v2 || (v1 != NULL && v2 != NULL && memcmp(v1, v2, l) == 0)) return (1); if (v1 == NULL || v2 == NULL) return (0); failure_start(file, line, "%s != %s", e1, e2); logprintf(" size %s = %d\n", ld, (int)l); /* Dump 48 bytes (3 lines) so that the first difference is * in the second line. */ offset = 0; while (l > 64 && memcmp(v1, v2, 32) == 0) { /* Two lines agree, so step forward one line. */ v1 += 16; v2 += 16; l -= 16; offset += 16; } logprintf(" Dump of %s\n", e1); hexdump(v1, v2, l < 128 ? l : 128, offset); logprintf(" Dump of %s\n", e2); hexdump(v2, v1, l < 128 ? l : 128, offset); logprintf("\n"); failure_finish(extra); return (0); } /* Verify that a block of memory is filled with the specified byte. */ int assertion_memory_filled_with(const char *file, int line, const void *_v1, const char *vd, size_t l, const char *ld, char b, const char *bd, void *extra) { const char *v1 = (const char *)_v1; size_t c = 0; size_t i; (void)ld; /* UNUSED */ assertion_count(file, line); for (i = 0; i < l; ++i) { if (v1[i] == b) { ++c; } } if (c == l) return (1); failure_start(file, line, "%s (size %d) not filled with %s", vd, (int)l, bd); logprintf(" Only %d bytes were correct\n", (int)c); failure_finish(extra); return (0); } /* Verify that the named file exists and is empty. */ int assertion_empty_file(const char *filename, int line, const char *f1) { char buff[1024]; struct stat st; ssize_t s; FILE *f; assertion_count(filename, line); if (stat(f1, &st) != 0) { failure_start(filename, line, "Stat failed: %s", f1); failure_finish(NULL); return (0); } if (st.st_size == 0) return (1); failure_start(filename, line, "File should be empty: %s", f1); logprintf(" File size: %d\n", (int)st.st_size); logprintf(" Contents:\n"); f = fopen(f1, "rb"); if (f == NULL) { logprintf(" Unable to open %s\n", f1); } else { s = ((off_t)sizeof(buff) < st.st_size) ? (ssize_t)sizeof(buff) : (ssize_t)st.st_size; s = fread(buff, 1, s, f); hexdump(buff, NULL, s, 0); fclose(f); } failure_finish(NULL); return (0); } /* Verify that the named file exists and is not empty. */ int assertion_non_empty_file(const char *filename, int line, const char *f1) { struct stat st; assertion_count(filename, line); if (stat(f1, &st) != 0) { failure_start(filename, line, "Stat failed: %s", f1); failure_finish(NULL); return (0); } if (st.st_size == 0) { failure_start(filename, line, "File empty: %s", f1); failure_finish(NULL); return (0); } return (1); } /* Verify that two files have the same contents. */ /* TODO: hexdump the first bytes that actually differ. */ int assertion_equal_file(const char *filename, int line, const char *fn1, const char *fn2) { char buff1[1024]; char buff2[1024]; FILE *f1, *f2; int n1, n2; assertion_count(filename, line); f1 = fopen(fn1, "rb"); f2 = fopen(fn2, "rb"); if (f1 == NULL || f2 == NULL) { if (f1) fclose(f1); if (f2) fclose(f2); return (0); } for (;;) { n1 = (int)fread(buff1, 1, sizeof(buff1), f1); n2 = (int)fread(buff2, 1, sizeof(buff2), f2); if (n1 != n2) break; if (n1 == 0 && n2 == 0) { fclose(f1); fclose(f2); return (1); } if (memcmp(buff1, buff2, n1) != 0) break; } fclose(f1); fclose(f2); failure_start(filename, line, "Files not identical"); logprintf(" file1=\"%s\"\n", fn1); logprintf(" file2=\"%s\"\n", fn2); failure_finish(NULL); return (0); } /* Verify that the named file does exist. */ int assertion_file_exists(const char *filename, int line, const char *f) { assertion_count(filename, line); #if defined(_WIN32) && !defined(__CYGWIN__) if (!_access(f, 0)) return (1); #else if (!access(f, F_OK)) return (1); #endif failure_start(filename, line, "File should exist: %s", f); failure_finish(NULL); return (0); } /* Verify that the named file doesn't exist. */ int assertion_file_not_exists(const char *filename, int line, const char *f) { assertion_count(filename, line); #if defined(_WIN32) && !defined(__CYGWIN__) if (_access(f, 0)) return (1); #else if (access(f, F_OK)) return (1); #endif failure_start(filename, line, "File should not exist: %s", f); failure_finish(NULL); return (0); } /* Compare the contents of a file to a block of memory. */ int assertion_file_contents(const char *filename, int line, const void *buff, int s, const char *fn) { char *contents; FILE *f; int n; assertion_count(filename, line); f = fopen(fn, "rb"); if (f == NULL) { failure_start(filename, line, "File should exist: %s", fn); failure_finish(NULL); return (0); } contents = malloc(s * 2); n = (int)fread(contents, 1, s * 2, f); fclose(f); if (n == s && memcmp(buff, contents, s) == 0) { free(contents); return (1); } failure_start(filename, line, "File contents don't match"); logprintf(" file=\"%s\"\n", fn); if (n > 0) hexdump(contents, buff, n > 512 ? 512 : n, 0); else { logprintf(" File empty, contents should be:\n"); hexdump(buff, NULL, s > 512 ? 512 : s, 0); } failure_finish(NULL); free(contents); return (0); } /* Check the contents of a text file, being tolerant of line endings. */ int assertion_text_file_contents(const char *filename, int line, const char *buff, const char *fn) { char *contents; const char *btxt, *ftxt; FILE *f; int n, s; assertion_count(filename, line); f = fopen(fn, "r"); if (f == NULL) { failure_start(filename, line, "File doesn't exist: %s", fn); failure_finish(NULL); return (0); } s = (int)strlen(buff); contents = malloc(s * 2 + 128); n = (int)fread(contents, 1, s * 2 + 128 - 1, f); if (n >= 0) contents[n] = '\0'; fclose(f); /* Compare texts. */ btxt = buff; ftxt = (const char *)contents; while (*btxt != '\0' && *ftxt != '\0') { if (*btxt == *ftxt) { ++btxt; ++ftxt; continue; } if (btxt[0] == '\n' && ftxt[0] == '\r' && ftxt[1] == '\n') { /* Pass over different new line characters. */ ++btxt; ftxt += 2; continue; } break; } if (*btxt == '\0' && *ftxt == '\0') { free(contents); return (1); } failure_start(filename, line, "Contents don't match"); logprintf(" file=\"%s\"\n", fn); if (n > 0) { hexdump(contents, buff, n, 0); logprintf(" expected\n", fn); hexdump(buff, contents, s, 0); } else { logprintf(" File empty, contents should be:\n"); hexdump(buff, NULL, s, 0); } failure_finish(NULL); free(contents); return (0); } /* Verify that a text file contains the specified lines, regardless of order */ /* This could be more efficient if we sorted both sets of lines, etc, but * since this is used only for testing and only ever deals with a dozen or so * lines at a time, this relatively crude approach is just fine. */ int assertion_file_contains_lines_any_order(const char *file, int line, const char *pathname, const char *lines[]) { char *buff; size_t buff_size; size_t expected_count, actual_count, i, j; char **expected = NULL; char *p, **actual = NULL; char c; int expected_failure = 0, actual_failure = 0; assertion_count(file, line); buff = slurpfile(&buff_size, "%s", pathname); if (buff == NULL) { failure_start(pathname, line, "Can't read file: %s", pathname); failure_finish(NULL); return (0); } /* Make a copy of the provided lines and count up the expected * file size. */ for (i = 0; lines[i] != NULL; ++i) { } expected_count = i; if (expected_count) { expected = malloc(sizeof(char *) * expected_count); if (expected == NULL) { failure_start(pathname, line, "Can't allocate memory"); failure_finish(NULL); free(expected); return (0); } for (i = 0; lines[i] != NULL; ++i) { expected[i] = strdup(lines[i]); } } /* Break the file into lines */ actual_count = 0; for (c = '\0', p = buff; p < buff + buff_size; ++p) { if (*p == '\x0d' || *p == '\x0a') *p = '\0'; if (c == '\0' && *p != '\0') ++actual_count; c = *p; } if (actual_count) { actual = calloc(sizeof(char *), actual_count); if (actual == NULL) { failure_start(pathname, line, "Can't allocate memory"); failure_finish(NULL); free(expected); return (0); } for (j = 0, p = buff; p < buff + buff_size; p += 1 + strlen(p)) { if (*p != '\0') { actual[j] = p; ++j; } } } /* Erase matching lines from both lists */ for (i = 0; i < expected_count; ++i) { if (expected[i] == NULL) continue; for (j = 0; j < actual_count; ++j) { if (actual[j] == NULL) continue; if (strcmp(expected[i], actual[j]) == 0) { free(expected[i]); expected[i] = NULL; actual[j] = NULL; break; } } } /* If there's anything left, it's a failure */ for (i = 0; i < expected_count; ++i) { if (expected[i] != NULL) ++expected_failure; } for (j = 0; j < actual_count; ++j) { if (actual[j] != NULL) ++actual_failure; } if (expected_failure == 0 && actual_failure == 0) { free(buff); free(expected); free(actual); return (1); } failure_start(file, line, "File doesn't match: %s", pathname); for (i = 0; i < expected_count; ++i) { if (expected[i] != NULL) { logprintf(" Expected but not present: %s\n", expected[i]); free(expected[i]); } } for (j = 0; j < actual_count; ++j) { if (actual[j] != NULL) logprintf(" Present but not expected: %s\n", actual[j]); } failure_finish(NULL); free(buff); free(expected); free(actual); return (0); } /* Test that two paths point to the same file. */ /* As a side-effect, asserts that both files exist. */ static int is_hardlink(const char *file, int line, const char *path1, const char *path2) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi1, bhfi2; int r; assertion_count(file, line); r = my_GetFileInformationByName(path1, &bhfi1); if (r == 0) { failure_start(file, line, "File %s can't be inspected?", path1); failure_finish(NULL); return (0); } r = my_GetFileInformationByName(path2, &bhfi2); if (r == 0) { failure_start(file, line, "File %s can't be inspected?", path2); failure_finish(NULL); return (0); } return (bhfi1.dwVolumeSerialNumber == bhfi2.dwVolumeSerialNumber && bhfi1.nFileIndexHigh == bhfi2.nFileIndexHigh && bhfi1.nFileIndexLow == bhfi2.nFileIndexLow); #else struct stat st1, st2; int r; assertion_count(file, line); r = lstat(path1, &st1); if (r != 0) { failure_start(file, line, "File should exist: %s", path1); failure_finish(NULL); return (0); } r = lstat(path2, &st2); if (r != 0) { failure_start(file, line, "File should exist: %s", path2); failure_finish(NULL); return (0); } return (st1.st_ino == st2.st_ino && st1.st_dev == st2.st_dev); #endif } int assertion_is_hardlink(const char *file, int line, const char *path1, const char *path2) { if (is_hardlink(file, line, path1, path2)) return (1); failure_start(file, line, "Files %s and %s are not hardlinked", path1, path2); failure_finish(NULL); return (0); } int assertion_is_not_hardlink(const char *file, int line, const char *path1, const char *path2) { if (!is_hardlink(file, line, path1, path2)) return (1); failure_start(file, line, "Files %s and %s should not be hardlinked", path1, path2); failure_finish(NULL); return (0); } /* Verify a/b/mtime of 'pathname'. */ /* If 'recent', verify that it's within last 10 seconds. */ static int assertion_file_time(const char *file, int line, const char *pathname, long t, long nsec, char type, int recent) { long long filet, filet_nsec; int r; #if defined(_WIN32) && !defined(__CYGWIN__) #define EPOC_TIME (116444736000000000ULL) FILETIME fxtime, fbirthtime, fatime, fmtime; ULARGE_INTEGER wintm; HANDLE h; fxtime.dwLowDateTime = 0; fxtime.dwHighDateTime = 0; assertion_count(file, line); /* Note: FILE_FLAG_BACKUP_SEMANTICS applies to open * a directory file. If not, CreateFile() will fail when * the pathname is a directory. */ h = CreateFile(pathname, FILE_READ_ATTRIBUTES, 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) { failure_start(file, line, "Can't access %s\n", pathname); failure_finish(NULL); return (0); } r = GetFileTime(h, &fbirthtime, &fatime, &fmtime); switch (type) { case 'a': fxtime = fatime; break; case 'b': fxtime = fbirthtime; break; case 'm': fxtime = fmtime; break; } CloseHandle(h); if (r == 0) { failure_start(file, line, "Can't GetFileTime %s\n", pathname); failure_finish(NULL); return (0); } wintm.LowPart = fxtime.dwLowDateTime; wintm.HighPart = fxtime.dwHighDateTime; filet = (wintm.QuadPart - EPOC_TIME) / 10000000; filet_nsec = ((wintm.QuadPart - EPOC_TIME) % 10000000) * 100; nsec = (nsec / 100) * 100; /* Round the request */ #else struct stat st; assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Can't stat %s\n", pathname); failure_finish(NULL); return (0); } switch (type) { case 'a': filet = st.st_atime; break; case 'm': filet = st.st_mtime; break; case 'b': filet = 0; break; default: fprintf(stderr, "INTERNAL: Bad type %c for file time", type); exit(1); } #if defined(__FreeBSD__) switch (type) { case 'a': filet_nsec = st.st_atimespec.tv_nsec; break; case 'b': filet = st.st_birthtime; + /* FreeBSD filesystems that don't support birthtime + * (e.g., UFS1) always return -1 here. */ + if (filet == -1) { + return (1); + } filet_nsec = st.st_birthtimespec.tv_nsec; break; case 'm': filet_nsec = st.st_mtimespec.tv_nsec; break; default: fprintf(stderr, "INTERNAL: Bad type %c for file time", type); exit(1); } /* FreeBSD generally only stores to microsecond res, so round. */ filet_nsec = (filet_nsec / 1000) * 1000; nsec = (nsec / 1000) * 1000; #else filet_nsec = nsec = 0; /* Generic POSIX only has whole seconds. */ if (type == 'b') return (1); /* Generic POSIX doesn't have birthtime */ #if defined(__HAIKU__) if (type == 'a') return (1); /* Haiku doesn't have atime. */ #endif #endif #endif if (recent) { /* Check that requested time is up-to-date. */ time_t now = time(NULL); if (filet < now - 10 || filet > now + 1) { failure_start(file, line, "File %s has %ctime %lld, %lld seconds ago\n", pathname, type, filet, now - filet); failure_finish(NULL); return (0); } } else if (filet != t || filet_nsec != nsec) { failure_start(file, line, "File %s has %ctime %lld.%09lld, expected %lld.%09lld", pathname, type, filet, filet_nsec, t, nsec); failure_finish(NULL); return (0); } return (1); } /* Verify atime of 'pathname'. */ int assertion_file_atime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'a', 0); } /* Verify atime of 'pathname' is up-to-date. */ int assertion_file_atime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'a', 1); } /* Verify birthtime of 'pathname'. */ int assertion_file_birthtime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'b', 0); } /* Verify birthtime of 'pathname' is up-to-date. */ int assertion_file_birthtime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'b', 1); } /* Verify mtime of 'pathname'. */ int assertion_file_mtime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'm', 0); } /* Verify mtime of 'pathname' is up-to-date. */ int assertion_file_mtime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'm', 1); } /* Verify number of links to 'pathname'. */ int assertion_file_nlinks(const char *file, int line, const char *pathname, int nlinks) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi; int r; assertion_count(file, line); r = my_GetFileInformationByName(pathname, &bhfi); if (r != 0 && bhfi.nNumberOfLinks == (DWORD)nlinks) return (1); failure_start(file, line, "File %s has %d links, expected %d", pathname, bhfi.nNumberOfLinks, nlinks); failure_finish(NULL); return (0); #else struct stat st; int r; assertion_count(file, line); r = lstat(pathname, &st); if (r == 0 && (int)st.st_nlink == nlinks) return (1); failure_start(file, line, "File %s has %d links, expected %d", pathname, st.st_nlink, nlinks); failure_finish(NULL); return (0); #endif } /* Verify size of 'pathname'. */ int assertion_file_size(const char *file, int line, const char *pathname, long size) { int64_t filesize; int r; assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) { BY_HANDLE_FILE_INFORMATION bhfi; r = !my_GetFileInformationByName(pathname, &bhfi); filesize = ((int64_t)bhfi.nFileSizeHigh << 32) + bhfi.nFileSizeLow; } #else { struct stat st; r = lstat(pathname, &st); filesize = st.st_size; } #endif if (r == 0 && filesize == size) return (1); failure_start(file, line, "File %s has size %ld, expected %ld", pathname, (long)filesize, (long)size); failure_finish(NULL); return (0); } /* Assert that 'pathname' is a dir. If mode >= 0, verify that too. */ int assertion_is_dir(const char *file, int line, const char *pathname, int mode) { struct stat st; int r; #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ #endif assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Dir should exist: %s", pathname); failure_finish(NULL); return (0); } if (!S_ISDIR(st.st_mode)) { failure_start(file, line, "%s is not a dir", pathname); failure_finish(NULL); return (0); } #if !defined(_WIN32) || defined(__CYGWIN__) /* Windows doesn't handle permissions the same way as POSIX, * so just ignore the mode tests. */ /* TODO: Can we do better here? */ if (mode >= 0 && (mode_t)mode != (st.st_mode & 07777)) { failure_start(file, line, "Dir %s has wrong mode", pathname); logprintf(" Expected: 0%3o\n", mode); logprintf(" Found: 0%3o\n", st.st_mode & 07777); failure_finish(NULL); return (0); } #endif return (1); } /* Verify that 'pathname' is a regular file. If 'mode' is >= 0, * verify that too. */ int assertion_is_reg(const char *file, int line, const char *pathname, int mode) { struct stat st; int r; #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ #endif assertion_count(file, line); r = lstat(pathname, &st); if (r != 0 || !S_ISREG(st.st_mode)) { failure_start(file, line, "File should exist: %s", pathname); failure_finish(NULL); return (0); } #if !defined(_WIN32) || defined(__CYGWIN__) /* Windows doesn't handle permissions the same way as POSIX, * so just ignore the mode tests. */ /* TODO: Can we do better here? */ if (mode >= 0 && (mode_t)mode != (st.st_mode & 07777)) { failure_start(file, line, "File %s has wrong mode", pathname); logprintf(" Expected: 0%3o\n", mode); logprintf(" Found: 0%3o\n", st.st_mode & 07777); failure_finish(NULL); return (0); } #endif return (1); } /* Check whether 'pathname' is a symbolic link. If 'contents' is * non-NULL, verify that the symlink has those contents. */ static int is_symlink(const char *file, int line, const char *pathname, const char *contents) { #if defined(_WIN32) && !defined(__CYGWIN__) (void)pathname; /* UNUSED */ (void)contents; /* UNUSED */ assertion_count(file, line); /* Windows sort-of has real symlinks, but they're only usable * by privileged users and are crippled even then, so there's * really not much point in bothering with this. */ return (0); #else char buff[300]; struct stat st; ssize_t linklen; int r; assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Symlink should exist: %s", pathname); failure_finish(NULL); return (0); } if (!S_ISLNK(st.st_mode)) return (0); if (contents == NULL) return (1); linklen = readlink(pathname, buff, sizeof(buff)); if (linklen < 0) { failure_start(file, line, "Can't read symlink %s", pathname); failure_finish(NULL); return (0); } buff[linklen] = '\0'; if (strcmp(buff, contents) != 0) return (0); return (1); #endif } /* Assert that path is a symlink that (optionally) contains contents. */ int assertion_is_symlink(const char *file, int line, const char *path, const char *contents) { if (is_symlink(file, line, path, contents)) return (1); if (contents) failure_start(file, line, "File %s is not a symlink to %s", path, contents); else failure_start(file, line, "File %s is not a symlink", path); failure_finish(NULL); return (0); } /* Create a directory and report any errors. */ int assertion_make_dir(const char *file, int line, const char *dirname, int mode) { assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ if (0 == _mkdir(dirname)) return (1); #else if (0 == mkdir(dirname, mode)) return (1); #endif failure_start(file, line, "Could not create directory %s", dirname); failure_finish(NULL); return(0); } /* Create a file with the specified contents and report any failures. */ int assertion_make_file(const char *file, int line, const char *path, int mode, int csize, const void *contents) { #if defined(_WIN32) && !defined(__CYGWIN__) /* TODO: Rework this to set file mode as well. */ FILE *f; (void)mode; /* UNUSED */ assertion_count(file, line); f = fopen(path, "wb"); if (f == NULL) { failure_start(file, line, "Could not create file %s", path); failure_finish(NULL); return (0); } if (contents != NULL) { size_t wsize; if (csize < 0) wsize = strlen(contents); else wsize = (size_t)csize; if (wsize != fwrite(contents, 1, wsize, f)) { fclose(f); failure_start(file, line, "Could not write file %s", path); failure_finish(NULL); return (0); } } fclose(f); return (1); #else int fd; assertion_count(file, line); fd = open(path, O_CREAT | O_WRONLY, mode >= 0 ? mode : 0644); if (fd < 0) { failure_start(file, line, "Could not create %s", path); failure_finish(NULL); return (0); } if (contents != NULL) { ssize_t wsize; if (csize < 0) wsize = (ssize_t)strlen(contents); else wsize = (ssize_t)csize; if (wsize != write(fd, contents, wsize)) { close(fd); failure_start(file, line, "Could not write to %s", path); failure_finish(NULL); return (0); } } close(fd); return (1); #endif } /* Create a hardlink and report any failures. */ int assertion_make_hardlink(const char *file, int line, const char *newpath, const char *linkto) { int succeeded; assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) succeeded = my_CreateHardLinkA(newpath, linkto); #elif HAVE_LINK succeeded = !link(linkto, newpath); #else succeeded = 0; #endif if (succeeded) return (1); failure_start(file, line, "Could not create hardlink"); logprintf(" New link: %s\n", newpath); logprintf(" Old name: %s\n", linkto); failure_finish(NULL); return(0); } /* Create a symlink and report any failures. */ int assertion_make_symlink(const char *file, int line, const char *newpath, const char *linkto) { #if defined(_WIN32) && !defined(__CYGWIN__) int targetIsDir = 0; /* TODO: Fix this */ assertion_count(file, line); if (my_CreateSymbolicLinkA(newpath, linkto, targetIsDir)) return (1); #elif HAVE_SYMLINK assertion_count(file, line); if (0 == symlink(linkto, newpath)) return (1); #endif failure_start(file, line, "Could not create symlink"); logprintf(" New link: %s\n", newpath); logprintf(" Old name: %s\n", linkto); failure_finish(NULL); return(0); } /* Set umask, report failures. */ int assertion_umask(const char *file, int line, int mask) { assertion_count(file, line); (void)file; /* UNUSED */ (void)line; /* UNUSED */ umask(mask); return (1); } /* Set times, report failures. */ int assertion_utimes(const char *file, int line, const char *pathname, long at, long at_nsec, long mt, long mt_nsec) { int r; #if defined(_WIN32) && !defined(__CYGWIN__) #define WINTIME(sec, nsec) ((Int32x32To64(sec, 10000000) + EPOC_TIME)\ + (((nsec)/1000)*10)) HANDLE h; ULARGE_INTEGER wintm; FILETIME fatime, fmtime; FILETIME *pat, *pmt; assertion_count(file, line); h = CreateFileA(pathname,GENERIC_READ | GENERIC_WRITE, FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) { failure_start(file, line, "Can't access %s\n", pathname); failure_finish(NULL); return (0); } if (at > 0 || at_nsec > 0) { wintm.QuadPart = WINTIME(at, at_nsec); fatime.dwLowDateTime = wintm.LowPart; fatime.dwHighDateTime = wintm.HighPart; pat = &fatime; } else pat = NULL; if (mt > 0 || mt_nsec > 0) { wintm.QuadPart = WINTIME(mt, mt_nsec); fmtime.dwLowDateTime = wintm.LowPart; fmtime.dwHighDateTime = wintm.HighPart; pmt = &fmtime; } else pmt = NULL; if (pat != NULL || pmt != NULL) r = SetFileTime(h, NULL, pat, pmt); else r = 1; CloseHandle(h); if (r == 0) { failure_start(file, line, "Can't SetFileTime %s\n", pathname); failure_finish(NULL); return (0); } return (1); #else /* defined(_WIN32) && !defined(__CYGWIN__) */ struct stat st; struct timeval times[2]; #if !defined(__FreeBSD__) mt_nsec = at_nsec = 0; /* Generic POSIX only has whole seconds. */ #endif if (mt == 0 && mt_nsec == 0 && at == 0 && at_nsec == 0) return (1); r = lstat(pathname, &st); if (r < 0) { failure_start(file, line, "Can't stat %s\n", pathname); failure_finish(NULL); return (0); } if (mt == 0 && mt_nsec == 0) { mt = st.st_mtime; #if defined(__FreeBSD__) mt_nsec = st.st_mtimespec.tv_nsec; /* FreeBSD generally only stores to microsecond res, so round. */ mt_nsec = (mt_nsec / 1000) * 1000; #endif } if (at == 0 && at_nsec == 0) { at = st.st_atime; #if defined(__FreeBSD__) at_nsec = st.st_atimespec.tv_nsec; /* FreeBSD generally only stores to microsecond res, so round. */ at_nsec = (at_nsec / 1000) * 1000; #endif } times[1].tv_sec = mt; times[1].tv_usec = mt_nsec / 1000; times[0].tv_sec = at; times[0].tv_usec = at_nsec / 1000; #ifdef HAVE_LUTIMES r = lutimes(pathname, times); #else r = utimes(pathname, times); #endif if (r < 0) { failure_start(file, line, "Can't utimes %s\n", pathname); failure_finish(NULL); return (0); } return (1); #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ } /* Set nodump, report failures. */ int assertion_nodump(const char *file, int line, const char *pathname) { #if defined(HAVE_STRUCT_STAT_ST_FLAGS) && defined(UF_NODUMP) int r; assertion_count(file, line); r = chflags(pathname, UF_NODUMP); if (r < 0) { failure_start(file, line, "Can't set nodump %s\n", pathname); failure_finish(NULL); return (0); } #elif defined(EXT2_IOC_GETFLAGS) && defined(HAVE_WORKING_EXT2_IOC_GETFLAGS)\ && defined(EXT2_NODUMP_FL) int fd, r, flags; assertion_count(file, line); fd = open(pathname, O_RDONLY | O_NONBLOCK); if (fd < 0) { failure_start(file, line, "Can't open %s\n", pathname); failure_finish(NULL); return (0); } r = ioctl(fd, EXT2_IOC_GETFLAGS, &flags); if (r < 0) { failure_start(file, line, "Can't get flags %s\n", pathname); failure_finish(NULL); return (0); } flags |= EXT2_NODUMP_FL; r = ioctl(fd, EXT2_IOC_SETFLAGS, &flags); if (r < 0) { failure_start(file, line, "Can't set nodump %s\n", pathname); failure_finish(NULL); return (0); } close(fd); #else (void)pathname; /* UNUSED */ assertion_count(file, line); #endif return (1); } /* * * UTILITIES for use by tests. * */ /* * Check whether platform supports symlinks. This is intended * for tests to use in deciding whether to bother testing symlink * support; if the platform doesn't support symlinks, there's no point * in checking whether the program being tested can create them. * * Note that the first time this test is called, we actually go out to * disk to create and verify a symlink. This is necessary because * symlink support is actually a property of a particular filesystem * and can thus vary between directories on a single system. After * the first call, this returns the cached result from memory, so it's * safe to call it as often as you wish. */ int canSymlink(void) { /* Remember the test result */ static int value = 0, tested = 0; if (tested) return (value); ++tested; assertion_make_file(__FILE__, __LINE__, "canSymlink.0", 0644, 1, "a"); /* Note: Cygwin has its own symlink() emulation that does not * use the Win32 CreateSymbolicLink() function. */ #if defined(_WIN32) && !defined(__CYGWIN__) value = my_CreateSymbolicLinkA("canSymlink.1", "canSymlink.0", 0) && is_symlink(__FILE__, __LINE__, "canSymlink.1", "canSymlink.0"); #elif HAVE_SYMLINK value = (0 == symlink("canSymlink.0", "canSymlink.1")) && is_symlink(__FILE__, __LINE__, "canSymlink.1","canSymlink.0"); #endif return (value); } /* Platform-dependent options for hiding the output of a subcommand. */ #if defined(_WIN32) && !defined(__CYGWIN__) static const char *redirectArgs = ">NUL 2>NUL"; /* Win32 cmd.exe */ #else static const char *redirectArgs = ">/dev/null 2>/dev/null"; /* POSIX 'sh' */ #endif /* * Can this platform run the bzip2 program? */ int canBzip2(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("bzip2 -d -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the grzip program? */ int canGrzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("grzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the gzip program? */ int canGzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("gzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the lrzip program? */ int canRunCommand(const char *cmd) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("%s %s", cmd, redirectArgs) == 0) value = 1; } return (value); } int canLrzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("lrzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the lz4 program? */ int canLz4(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("lz4 -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the lzip program? */ int canLzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("lzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the lzma program? */ int canLzma(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("lzma -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the lzop program? */ int canLzop(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("lzop -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the xz program? */ int canXz(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("xz -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this filesystem handle nodump flags. */ #if defined(HAVE_STRUCT_STAT_ST_FLAGS) && defined(UF_NODUMP) int canNodump(void) { const char *path = "cannodumptest"; struct stat sb; assertion_make_file(__FILE__, __LINE__, path, 0644, 0, NULL); if (chflags(path, UF_NODUMP) < 0) return (0); if (stat(path, &sb) < 0) return (0); if (sb.st_flags & UF_NODUMP) return (1); return (0); } #elif defined(EXT2_IOC_GETFLAGS) && defined(HAVE_WORKING_EXT2_IOC_GETFLAGS)\ && defined(EXT2_NODUMP_FL) int canNodump(void) { const char *path = "cannodumptest"; int fd, r, flags; assertion_make_file(__FILE__, __LINE__, path, 0644, 0, NULL); fd = open(path, O_RDONLY | O_NONBLOCK); if (fd < 0) return (0); r = ioctl(fd, EXT2_IOC_GETFLAGS, &flags); if (r < 0) return (0); flags |= EXT2_NODUMP_FL; r = ioctl(fd, EXT2_IOC_SETFLAGS, &flags); if (r < 0) return (0); close(fd); fd = open(path, O_RDONLY | O_NONBLOCK); if (fd < 0) return (0); r = ioctl(fd, EXT2_IOC_GETFLAGS, &flags); if (r < 0) return (0); close(fd); if (flags & EXT2_NODUMP_FL) return (1); return (0); } #else int canNodump() { return (0); } #endif /* * Sleep as needed; useful for verifying disk timestamp changes by * ensuring that the wall-clock time has actually changed before we * go back to re-read something from disk. */ void sleepUntilAfter(time_t t) { while (t >= time(NULL)) #if defined(_WIN32) && !defined(__CYGWIN__) Sleep(500); #else sleep(1); #endif } /* * Call standard system() call, but build up the command line using * sprintf() conventions. */ int systemf(const char *fmt, ...) { char buff[8192]; va_list ap; int r; va_start(ap, fmt); vsprintf(buff, fmt, ap); if (verbosity > VERBOSITY_FULL) logprintf("Cmd: %s\n", buff); r = system(buff); va_end(ap); return (r); } /* * Slurp a file into memory for ease of comparison and testing. * Returns size of file in 'sizep' if non-NULL, null-terminates * data in memory for ease of use. */ char * slurpfile(size_t * sizep, const char *fmt, ...) { char filename[8192]; struct stat st; va_list ap; char *p; ssize_t bytes_read; FILE *f; int r; va_start(ap, fmt); vsprintf(filename, fmt, ap); va_end(ap); f = fopen(filename, "rb"); if (f == NULL) { /* Note: No error; non-existent file is okay here. */ return (NULL); } r = fstat(fileno(f), &st); if (r != 0) { logprintf("Can't stat file %s\n", filename); fclose(f); return (NULL); } p = malloc((size_t)st.st_size + 1); if (p == NULL) { logprintf("Can't allocate %ld bytes of memory to read file %s\n", (long int)st.st_size, filename); fclose(f); return (NULL); } bytes_read = fread(p, 1, (size_t)st.st_size, f); if (bytes_read < st.st_size) { logprintf("Can't read file %s\n", filename); fclose(f); free(p); return (NULL); } p[st.st_size] = '\0'; if (sizep != NULL) *sizep = (size_t)st.st_size; fclose(f); return (p); } /* * Slurp a file into memory for ease of comparison and testing. * Returns size of file in 'sizep' if non-NULL, null-terminates * data in memory for ease of use. */ void dumpfile(const char *filename, void *data, size_t len) { ssize_t bytes_written; FILE *f; f = fopen(filename, "wb"); if (f == NULL) { logprintf("Can't open file %s for writing\n", filename); return; } bytes_written = fwrite(data, 1, len, f); if (bytes_written < (ssize_t)len) logprintf("Can't write file %s\n", filename); fclose(f); } /* Read a uuencoded file from the reference directory, decode, and * write the result into the current directory. */ #define VALID_UUDECODE(c) (c >= 32 && c <= 96) #define UUDECODE(c) (((c) - 0x20) & 0x3f) void extract_reference_file(const char *name) { char buff[1024]; FILE *in, *out; sprintf(buff, "%s/%s.uu", refdir, name); in = fopen(buff, "r"); failure("Couldn't open reference file %s", buff); assert(in != NULL); if (in == NULL) return; /* Read up to and including the 'begin' line. */ for (;;) { if (fgets(buff, sizeof(buff), in) == NULL) { /* TODO: This is a failure. */ return; } if (memcmp(buff, "begin ", 6) == 0) break; } /* Now, decode the rest and write it. */ out = fopen(name, "wb"); while (fgets(buff, sizeof(buff), in) != NULL) { char *p = buff; int bytes; if (memcmp(buff, "end", 3) == 0) break; bytes = UUDECODE(*p++); while (bytes > 0) { int n = 0; /* Write out 1-3 bytes from that. */ if (bytes > 0) { assert(VALID_UUDECODE(p[0])); assert(VALID_UUDECODE(p[1])); n = UUDECODE(*p++) << 18; n |= UUDECODE(*p++) << 12; fputc(n >> 16, out); --bytes; } if (bytes > 0) { assert(VALID_UUDECODE(p[0])); n |= UUDECODE(*p++) << 6; fputc((n >> 8) & 0xFF, out); --bytes; } if (bytes > 0) { assert(VALID_UUDECODE(p[0])); n |= UUDECODE(*p++); fputc(n & 0xFF, out); --bytes; } } } fclose(out); fclose(in); } void copy_reference_file(const char *name) { char buff[1024]; FILE *in, *out; size_t rbytes; sprintf(buff, "%s/%s", refdir, name); in = fopen(buff, "rb"); failure("Couldn't open reference file %s", buff); assert(in != NULL); if (in == NULL) return; /* Now, decode the rest and write it. */ /* Not a lot of error checking here; the input better be right. */ out = fopen(name, "wb"); while ((rbytes = fread(buff, 1, sizeof(buff), in)) > 0) { if (fwrite(buff, 1, rbytes, out) != rbytes) { logprintf("Error: fwrite\n"); break; } } fclose(out); fclose(in); } int is_LargeInode(const char *file) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi; int r; r = my_GetFileInformationByName(file, &bhfi); if (r != 0) return (0); return (bhfi.nFileIndexHigh & 0x0000FFFFUL); #else struct stat st; int64_t ino; if (stat(file, &st) < 0) return (0); ino = (int64_t)st.st_ino; return (ino > 0xffffffff); #endif } void extract_reference_files(const char **names) { while (names && *names) extract_reference_file(*names++); } /* * * TEST management * */ /* * "list.h" is simply created by "grep DEFINE_TEST test_*.c"; it has * a line like * DEFINE_TEST(test_function) * for each test. */ /* Use "list.h" to declare all of the test functions. */ #undef DEFINE_TEST #define DEFINE_TEST(name) void name(void); #include "list.h" /* Use "list.h" to create a list of all tests (functions and names). */ #undef DEFINE_TEST #define DEFINE_TEST(n) { n, #n, 0 }, struct test_list_t tests[] = { #include "list.h" }; /* * Summarize repeated failures in the just-completed test. */ static void test_summarize(int failed, int skips_num) { unsigned int i; switch (verbosity) { case VERBOSITY_SUMMARY_ONLY: printf(failed ? "E" : "."); fflush(stdout); break; case VERBOSITY_PASSFAIL: printf(failed ? "FAIL\n" : skips_num ? "ok (S)\n" : "ok\n"); break; } log_console = (verbosity == VERBOSITY_LIGHT_REPORT); for (i = 0; i < sizeof(failed_lines)/sizeof(failed_lines[0]); i++) { if (failed_lines[i].count > 1 && !failed_lines[i].skip) logprintf("%s:%d: Summary: Failed %d times\n", failed_filename, i, failed_lines[i].count); } /* Clear the failure history for the next file. */ failed_filename = NULL; memset(failed_lines, 0, sizeof(failed_lines)); } /* * Actually run a single test, with appropriate setup and cleanup. */ static int test_run(int i, const char *tmpdir) { char workdir[1024]; char logfilename[64]; int failures_before = failures; int skips_before = skips; int oldumask; switch (verbosity) { case VERBOSITY_SUMMARY_ONLY: /* No per-test reports at all */ break; case VERBOSITY_PASSFAIL: /* rest of line will include ok/FAIL marker */ printf("%3d: %-64s", i, tests[i].name); fflush(stdout); break; default: /* Title of test, details will follow */ printf("%3d: %s\n", i, tests[i].name); } /* Chdir to the top-level work directory. */ if (!assertChdir(tmpdir)) { fprintf(stderr, "ERROR: Can't chdir to top work dir %s\n", tmpdir); exit(1); } /* Create a log file for this test. */ sprintf(logfilename, "%s.log", tests[i].name); logfile = fopen(logfilename, "w"); fprintf(logfile, "%s\n\n", tests[i].name); /* Chdir() to a work dir for this specific test. */ snprintf(workdir, sizeof(workdir), "%s/%s", tmpdir, tests[i].name); testworkdir = workdir; if (!assertMakeDir(testworkdir, 0755) || !assertChdir(testworkdir)) { fprintf(stderr, "ERROR: Can't chdir to work dir %s\n", testworkdir); exit(1); } /* Explicitly reset the locale before each test. */ setlocale(LC_ALL, "C"); /* Record the umask before we run the test. */ umask(oldumask = umask(0)); /* * Run the actual test. */ (*tests[i].func)(); /* * Clean up and report afterwards. */ testworkdir = NULL; /* Restore umask */ umask(oldumask); /* Reset locale. */ setlocale(LC_ALL, "C"); /* Reset directory. */ if (!assertChdir(tmpdir)) { fprintf(stderr, "ERROR: Couldn't chdir to temp dir %s\n", tmpdir); exit(1); } /* Report per-test summaries. */ tests[i].failures = failures - failures_before; test_summarize(tests[i].failures, skips - skips_before); /* Close the per-test log file. */ fclose(logfile); logfile = NULL; /* If there were no failures, we can remove the work dir and logfile. */ if (tests[i].failures == 0) { if (!keep_temp_files && assertChdir(tmpdir)) { #if defined(_WIN32) && !defined(__CYGWIN__) /* Make sure not to leave empty directories. * Sometimes a processing of closing files used by tests * is not done, then rmdir will be failed and it will * leave a empty test directory. So we should wait a few * seconds and retry rmdir. */ int r, t; for (t = 0; t < 10; t++) { if (t > 0) Sleep(1000); r = systemf("rmdir /S /Q %s", tests[i].name); if (r == 0) break; } systemf("del %s", logfilename); #else systemf("rm -rf %s", tests[i].name); systemf("rm %s", logfilename); #endif } } /* Return appropriate status. */ return (tests[i].failures); } /* * * * MAIN and support routines. * * */ static void usage(const char *program) { static const int limit = sizeof(tests) / sizeof(tests[0]); int i; printf("Usage: %s [options] ...\n", program); printf("Default is to run all tests.\n"); printf("Otherwise, specify the numbers of the tests you wish to run.\n"); printf("Options:\n"); printf(" -d Dump core after any failure, for debugging.\n"); printf(" -k Keep all temp files.\n"); printf(" Default: temp files for successful tests deleted.\n"); #ifdef PROGRAM printf(" -p Path to executable to be tested.\n"); printf(" Default: path taken from " ENVBASE " environment variable.\n"); #endif printf(" -q Quiet.\n"); printf(" -r Path to dir containing reference files.\n"); printf(" Default: Current directory.\n"); printf(" -u Keep running specifies tests until one fails.\n"); printf(" -v Verbose.\n"); printf("Available tests:\n"); for (i = 0; i < limit; i++) printf(" %d: %s\n", i, tests[i].name); exit(1); } static char * get_refdir(const char *d) { size_t tried_size, buff_size; char *buff, *tried, *pwd = NULL, *p = NULL; #ifdef PATH_MAX buff_size = PATH_MAX; #else buff_size = 8192; #endif buff = calloc(buff_size, 1); if (buff == NULL) { fprintf(stderr, "Unable to allocate memory\n"); exit(1); } /* Allocate a buffer to hold the various directories we checked. */ tried_size = buff_size * 2; tried = calloc(tried_size, 1); if (tried == NULL) { fprintf(stderr, "Unable to allocate memory\n"); exit(1); } /* If a dir was specified, try that */ if (d != NULL) { pwd = NULL; snprintf(buff, buff_size, "%s", d); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, tried_size - strlen(tried) - 1); strncat(tried, "\n", tried_size - strlen(tried) - 1); goto failure; } /* Get the current dir. */ #ifdef PATH_MAX pwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else pwd = getcwd(NULL, 0); #endif while (pwd[strlen(pwd) - 1] == '\n') pwd[strlen(pwd) - 1] = '\0'; /* Look for a known file. */ snprintf(buff, buff_size, "%s", pwd); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, tried_size - strlen(tried) - 1); strncat(tried, "\n", tried_size - strlen(tried) - 1); snprintf(buff, buff_size, "%s/test", pwd); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, tried_size - strlen(tried) - 1); strncat(tried, "\n", tried_size - strlen(tried) - 1); #if defined(LIBRARY) snprintf(buff, buff_size, "%s/%s/test", pwd, LIBRARY); #else snprintf(buff, buff_size, "%s/%s/test", pwd, PROGRAM); #endif p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, tried_size - strlen(tried) - 1); strncat(tried, "\n", tried_size - strlen(tried) - 1); #if defined(PROGRAM_ALIAS) snprintf(buff, buff_size, "%s/%s/test", pwd, PROGRAM_ALIAS); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, tried_size - strlen(tried) - 1); strncat(tried, "\n", tried_size - strlen(tried) - 1); #endif if (memcmp(pwd, "/usr/obj", 8) == 0) { snprintf(buff, buff_size, "%s", pwd + 8); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, tried_size - strlen(tried) - 1); strncat(tried, "\n", tried_size - strlen(tried) - 1); snprintf(buff, buff_size, "%s/test", pwd + 8); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, tried_size - strlen(tried) - 1); strncat(tried, "\n", tried_size - strlen(tried) - 1); } failure: printf("Unable to locate known reference file %s\n", KNOWNREF); printf(" Checked following directories:\n%s\n", tried); printf("Use -r option to specify full path to reference directory\n"); #if defined(_WIN32) && !defined(__CYGWIN__) && defined(_DEBUG) DebugBreak(); #endif exit(1); success: free(p); free(pwd); free(tried); /* Copy result into a fresh buffer to reduce memory usage. */ p = strdup(buff); free(buff); return p; } int main(int argc, char **argv) { static const int limit = sizeof(tests) / sizeof(tests[0]); int test_set[sizeof(tests) / sizeof(tests[0])]; int i = 0, j = 0, tests_run = 0, tests_failed = 0, option; time_t now; char *refdir_alloc = NULL; const char *progname; char **saved_argv; const char *tmp, *option_arg, *p; char tmpdir[256], *pwd, *testprogdir, *tmp2 = NULL, *vlevel = NULL; char tmpdir_timestamp[256]; (void)argc; /* UNUSED */ /* Get the current dir. */ #ifdef PATH_MAX pwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else pwd = getcwd(NULL, 0); #endif while (pwd[strlen(pwd) - 1] == '\n') pwd[strlen(pwd) - 1] = '\0'; #if defined(HAVE__CrtSetReportMode) && !defined(__WATCOMC__) /* To stop to run the default invalid parameter handler. */ _set_invalid_parameter_handler(invalid_parameter_handler); /* Disable annoying assertion message box. */ _CrtSetReportMode(_CRT_ASSERT, 0); #endif /* * Name of this program, used to build root of our temp directory * tree. */ progname = p = argv[0]; if ((testprogdir = (char *)malloc(strlen(progname) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(testprogdir, progname); while (*p != '\0') { /* Support \ or / dir separators for Windows compat. */ if (*p == '/' || *p == '\\') { progname = p + 1; i = j; } ++p; j++; } testprogdir[i] = '\0'; #if defined(_WIN32) && !defined(__CYGWIN__) if (testprogdir[0] != '/' && testprogdir[0] != '\\' && !(((testprogdir[0] >= 'a' && testprogdir[0] <= 'z') || (testprogdir[0] >= 'A' && testprogdir[0] <= 'Z')) && testprogdir[1] == ':' && (testprogdir[2] == '/' || testprogdir[2] == '\\'))) #else if (testprogdir[0] != '/') #endif { /* Fixup path for relative directories. */ if ((testprogdir = (char *)realloc(testprogdir, strlen(pwd) + 1 + strlen(testprogdir) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } memmove(testprogdir + strlen(pwd) + 1, testprogdir, strlen(testprogdir) + 1); memcpy(testprogdir, pwd, strlen(pwd)); testprogdir[strlen(pwd)] = '/'; } #ifdef PROGRAM /* Get the target program from environment, if available. */ testprogfile = getenv(ENVBASE); #endif if (getenv("TMPDIR") != NULL) tmp = getenv("TMPDIR"); else if (getenv("TMP") != NULL) tmp = getenv("TMP"); else if (getenv("TEMP") != NULL) tmp = getenv("TEMP"); else if (getenv("TEMPDIR") != NULL) tmp = getenv("TEMPDIR"); else tmp = "/tmp"; /* Allow -d to be controlled through the environment. */ if (getenv(ENVBASE "_DEBUG") != NULL) dump_on_failure = 1; /* Allow -v to be controlled through the environment. */ if (getenv("_VERBOSITY_LEVEL") != NULL) { vlevel = getenv("_VERBOSITY_LEVEL"); verbosity = atoi(vlevel); if (verbosity < VERBOSITY_SUMMARY_ONLY || verbosity > VERBOSITY_FULL) { /* Unsupported verbosity levels are silently ignored */ vlevel = NULL; verbosity = VERBOSITY_PASSFAIL; } } /* Get the directory holding test files from environment. */ refdir = getenv(ENVBASE "_TEST_FILES"); /* * Parse options, without using getopt(), which isn't available * on all platforms. */ ++argv; /* Skip program name */ while (*argv != NULL) { if (**argv != '-') break; p = *argv++; ++p; /* Skip '-' */ while (*p != '\0') { option = *p++; option_arg = NULL; /* If 'opt' takes an argument, parse that. */ if (option == 'p' || option == 'r') { if (*p != '\0') option_arg = p; else if (*argv == NULL) { fprintf(stderr, "Option -%c requires argument.\n", option); usage(progname); } else option_arg = *argv++; p = ""; /* End of this option word. */ } /* Now, handle the option. */ switch (option) { case 'd': dump_on_failure = 1; break; case 'k': keep_temp_files = 1; break; case 'p': #ifdef PROGRAM testprogfile = option_arg; #else fprintf(stderr, "-p option not permitted\n"); usage(progname); #endif break; case 'q': if (!vlevel) verbosity--; break; case 'r': refdir = option_arg; break; case 'u': until_failure++; break; case 'v': if (!vlevel) verbosity++; break; default: fprintf(stderr, "Unrecognized option '%c'\n", option); usage(progname); } } } /* * Sanity-check that our options make sense. */ #ifdef PROGRAM if (testprogfile == NULL) { if ((tmp2 = (char *)malloc(strlen(testprogdir) + 1 + strlen(PROGRAM) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(tmp2, testprogdir); strcat(tmp2, "/"); strcat(tmp2, PROGRAM); testprogfile = tmp2; } { char *testprg; #if defined(_WIN32) && !defined(__CYGWIN__) /* Command.com sometimes rejects '/' separators. */ testprg = strdup(testprogfile); for (i = 0; testprg[i] != '\0'; i++) { if (testprg[i] == '/') testprg[i] = '\\'; } testprogfile = testprg; #endif /* Quote the name that gets put into shell command lines. */ testprg = malloc(strlen(testprogfile) + 3); strcpy(testprg, "\""); strcat(testprg, testprogfile); strcat(testprg, "\""); testprog = testprg; } #endif #if !defined(_WIN32) && defined(SIGPIPE) { /* Ignore SIGPIPE signals */ struct sigaction sa; sa.sa_handler = SIG_IGN; sigemptyset(&sa.sa_mask); sa.sa_flags = 0; sigaction(SIGPIPE, &sa, NULL); } #endif /* * Create a temp directory for the following tests. * Include the time the tests started as part of the name, * to make it easier to track the results of multiple tests. */ now = time(NULL); for (i = 0; ; i++) { strftime(tmpdir_timestamp, sizeof(tmpdir_timestamp), "%Y-%m-%dT%H.%M.%S", localtime(&now)); sprintf(tmpdir, "%s/%s.%s-%03d", tmp, progname, tmpdir_timestamp, i); if (assertMakeDir(tmpdir,0755)) break; if (i >= 999) { fprintf(stderr, "ERROR: Unable to create temp directory %s\n", tmpdir); exit(1); } } /* * If the user didn't specify a directory for locating * reference files, try to find the reference files in * the "usual places." */ refdir = refdir_alloc = get_refdir(refdir); /* * Banner with basic information. */ printf("\n"); printf("If tests fail or crash, details will be in:\n"); printf(" %s\n", tmpdir); printf("\n"); if (verbosity > VERBOSITY_SUMMARY_ONLY) { printf("Reference files will be read from: %s\n", refdir); #ifdef PROGRAM printf("Running tests on: %s\n", testprog); #endif printf("Exercising: "); fflush(stdout); printf("%s\n", EXTRA_VERSION); } else { printf("Running "); fflush(stdout); } /* * Run some or all of the individual tests. */ saved_argv = argv; do { argv = saved_argv; do { int test_num; test_num = get_test_set(test_set, limit, *argv, tests); if (test_num < 0) { printf("*** INVALID Test %s\n", *argv); free(refdir_alloc); free(testprogdir); usage(progname); return (1); } for (i = 0; i < test_num; i++) { tests_run++; if (test_run(test_set[i], tmpdir)) { tests_failed++; if (until_failure) goto finish; } } if (*argv != NULL) argv++; } while (*argv != NULL); } while (until_failure); finish: /* Must be freed after all tests run */ free(tmp2); free(testprogdir); free(pwd); /* * Report summary statistics. */ if (verbosity > VERBOSITY_SUMMARY_ONLY) { printf("\n"); printf("Totals:\n"); printf(" Tests run: %8d\n", tests_run); printf(" Tests failed: %8d\n", tests_failed); printf(" Assertions checked:%8d\n", assertions); printf(" Assertions failed: %8d\n", failures); printf(" Skips reported: %8d\n", skips); } if (failures) { printf("\n"); printf("Failing tests:\n"); for (i = 0; i < limit; ++i) { if (tests[i].failures) printf(" %d: %s (%d failures)\n", i, tests[i].name, tests[i].failures); } printf("\n"); printf("Details for failing tests: %s\n", tmpdir); printf("\n"); } else { if (verbosity == VERBOSITY_SUMMARY_ONLY) printf("\n"); printf("%d tests passed, no failures\n", tests_run); } free(refdir_alloc); /* If the final tmpdir is empty, we can remove it. */ /* This should be the usual case when all tests succeed. */ assertChdir(".."); rmdir(tmpdir); return (tests_failed ? 1 : 0); } Index: stable/10/contrib/libarchive/libarchive/test/test_write_format_gnutar_filenames.c =================================================================== --- stable/10/contrib/libarchive/libarchive/test/test_write_format_gnutar_filenames.c (revision 302424) +++ stable/10/contrib/libarchive/libarchive/test/test_write_format_gnutar_filenames.c (revision 302425) @@ -1,145 +1,147 @@ /*- * Copyright (c) 2016 Tim Kientzle * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "test.h" __FBSDID("$FreeBSD$"); /* * Inspired by Github issue #682, which reported that gnutar filenames * of exactly 512 bytes weren't getting written correctly. * * This writes a filename of every length from 1 to 2000 bytes and * reads back to verify it. */ static char filename[2048]; DEFINE_TEST(test_write_format_gnutar_filenames) { size_t buffsize = 1000000; char *buff; struct archive_entry *ae, *template; struct archive *a; size_t used; + int i; buff = malloc(buffsize); /* million bytes of work area */ assert(buff != NULL); /* Create a template entry. */ assert((template = archive_entry_new()) != NULL); archive_entry_set_atime(template, 2, 20); archive_entry_set_birthtime(template, 3, 30); archive_entry_set_ctime(template, 4, 40); archive_entry_set_mtime(template, 5, 50); archive_entry_set_mode(template, S_IFREG | 0755); archive_entry_set_size(template, 8); - for (int i = 0; i < 2000; ++i) { + for (i = 0; i < 2000; ++i) { filename[i] = 'a'; filename[i + 1] = '\0'; archive_entry_copy_pathname(template, filename); /* Write a one-item gnutar format archive. */ assert((a = archive_write_new()) != NULL); assertA(0 == archive_write_set_format_gnutar(a)); assertA(0 == archive_write_add_filter_none(a)); assertA(0 == archive_write_open_memory(a, buff, buffsize, &used)); assertEqualIntA(a, ARCHIVE_OK, archive_write_header(a, template)); assertEqualIntA(a, 8, archive_write_data(a, "12345678", 9)); assertEqualIntA(a, ARCHIVE_OK, archive_write_close(a)); assertEqualIntA(a, ARCHIVE_OK, archive_write_free(a)); /* Read back and verify the filename. */ assert((a = archive_read_new()) != NULL); assertEqualIntA(a, 0, archive_read_support_format_all(a)); assertEqualIntA(a, 0, archive_read_support_filter_all(a)); assertEqualIntA(a, 0, archive_read_open_memory(a, buff, used)); assertEqualIntA(a, 0, archive_read_next_header(a, &ae)); assertEqualString(filename, archive_entry_pathname(ae)); assertEqualIntA(a, ARCHIVE_EOF, archive_read_next_header(a, &ae)); assertEqualIntA(a, ARCHIVE_OK, archive_read_close(a)); assertEqualIntA(a, ARCHIVE_OK, archive_read_free(a)); } archive_entry_free(template); free(buff); } DEFINE_TEST(test_write_format_gnutar_linknames) { size_t buffsize = 1000000; char *buff; struct archive_entry *ae, *template; struct archive *a; size_t used; + int i; buff = malloc(buffsize); /* million bytes of work area */ assert(buff != NULL); /* Create a template entry. */ assert((template = archive_entry_new()) != NULL); archive_entry_set_atime(template, 2, 20); archive_entry_set_birthtime(template, 3, 30); archive_entry_set_ctime(template, 4, 40); archive_entry_set_mtime(template, 5, 50); archive_entry_set_mode(template, S_IFLNK | 0755); archive_entry_copy_pathname(template, "link"); - for (int i = 0; i < 2000; ++i) { + for (i = 0; i < 2000; ++i) { filename[i] = 'a'; filename[i + 1] = '\0'; archive_entry_copy_symlink(template, filename); /* Write a one-item gnutar format archive. */ assert((a = archive_write_new()) != NULL); assertA(0 == archive_write_set_format_gnutar(a)); assertA(0 == archive_write_add_filter_none(a)); assertA(0 == archive_write_open_memory(a, buff, buffsize, &used)); assertEqualIntA(a, ARCHIVE_OK, archive_write_header(a, template)); assertEqualIntA(a, ARCHIVE_OK, archive_write_close(a)); assertEqualIntA(a, ARCHIVE_OK, archive_write_free(a)); /* Read back and verify the filename. */ assert((a = archive_read_new()) != NULL); assertEqualIntA(a, 0, archive_read_support_format_all(a)); assertEqualIntA(a, 0, archive_read_support_filter_all(a)); assertEqualIntA(a, 0, archive_read_open_memory(a, buff, used)); assertEqualIntA(a, 0, archive_read_next_header(a, &ae)); assertEqualString("link", archive_entry_pathname(ae)); assertEqualString(filename, archive_entry_symlink(ae)); assertEqualIntA(a, ARCHIVE_EOF, archive_read_next_header(a, &ae)); assertEqualIntA(a, ARCHIVE_OK, archive_read_close(a)); assertEqualIntA(a, ARCHIVE_OK, archive_read_free(a)); } archive_entry_free(template); free(buff); } Index: stable/10/contrib/libarchive/libarchive_fe/passphrase.c =================================================================== --- stable/10/contrib/libarchive/libarchive_fe/passphrase.c (revision 302424) +++ stable/10/contrib/libarchive/libarchive_fe/passphrase.c (revision 302425) @@ -1,322 +1,326 @@ /*- * Copyright (c) 2014 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ /* $OpenBSD: readpassphrase.c,v 1.22 2010/01/13 10:20:54 dtucker Exp $ */ /* * Copyright (c) 2000-2002, 2007 Todd C. Miller * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above * copyright notice and this permission notice appear in all copies. * * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. * * Sponsored in part by the Defense Advanced Research Projects * Agency (DARPA) and Air Force Research Laboratory, Air Force * Materiel Command, USAF, under agreement number F39502-99-1-0512. */ /* OPENBSD ORIGINAL: lib/libc/gen/readpassphrase.c */ #include "lafe_platform.h" __FBSDID("$FreeBSD$"); #include #ifdef HAVE_STDLIB_H #include #endif #ifdef HAVE_UNISTD_H #include #endif #ifdef HAVE_READPASSPHRASE_H #include #endif #include "err.h" #include "passphrase.h" #ifndef HAVE_READPASSPHRASE #define RPP_ECHO_OFF 0x00 /* Turn off echo (default). */ #define RPP_ECHO_ON 0x01 /* Leave echo on. */ #define RPP_REQUIRE_TTY 0x02 /* Fail if there is no tty. */ #define RPP_FORCELOWER 0x04 /* Force input to lower case. */ #define RPP_FORCEUPPER 0x08 /* Force input to upper case. */ #define RPP_SEVENBIT 0x10 /* Strip the high bit from input. */ #define RPP_STDIN 0x20 /* Read from stdin, not /dev/tty */ #if defined(_WIN32) && !defined(__CYGWIN__) #include static char * readpassphrase(const char *prompt, char *buf, size_t bufsiz, int flags) { HANDLE hStdin, hStdout; DWORD mode, rbytes; BOOL success; (void)flags; hStdin = GetStdHandle(STD_INPUT_HANDLE); if (hStdin == INVALID_HANDLE_VALUE) return (NULL); hStdout = GetStdHandle(STD_OUTPUT_HANDLE); if (hStdout == INVALID_HANDLE_VALUE) return (NULL); success = GetConsoleMode(hStdin, &mode); if (!success) return (NULL); mode &= ~ENABLE_ECHO_INPUT; mode |= ENABLE_LINE_INPUT | ENABLE_PROCESSED_INPUT; success = SetConsoleMode(hStdin, mode); if (!success) return (NULL); success = WriteFile(hStdout, prompt, (DWORD)strlen(prompt), NULL, NULL); if (!success) return (NULL); success = ReadFile(hStdin, buf, (DWORD)bufsiz - 1, &rbytes, NULL); if (!success) return (NULL); WriteFile(hStdout, "\r\n", 2, NULL, NULL); buf[rbytes] = '\0'; /* Remove trailing carriage return(s). */ if (rbytes > 2 && buf[rbytes - 2] == '\r' && buf[rbytes - 1] == '\n') buf[rbytes - 2] = '\0'; return (buf); } #else /* _WIN32 && !__CYGWIN__ */ #include #include #include #ifdef HAVE_PATHS_H #include #endif #include #include #include #include +#ifndef _PATH_TTY +#define _PATH_TTY "/dev/tty" +#endif + #ifdef TCSASOFT # define _T_FLUSH (TCSAFLUSH|TCSASOFT) #else # define _T_FLUSH (TCSAFLUSH) #endif /* SunOS 4.x which lacks _POSIX_VDISABLE, but has VDISABLE */ #if !defined(_POSIX_VDISABLE) && defined(VDISABLE) # define _POSIX_VDISABLE VDISABLE #endif #define M(a,b) (a > b ? a : b) #define MAX_SIGNO M(M(M(SIGALRM, SIGHUP), \ M(SIGINT, SIGPIPE)), \ M(M(SIGQUIT, SIGTERM), \ M(M(SIGTSTP, SIGTTIN), SIGTTOU))) static volatile sig_atomic_t signo[MAX_SIGNO + 1]; static void handler(int s) { assert(s <= MAX_SIGNO); signo[s] = 1; } static char * readpassphrase(const char *prompt, char *buf, size_t bufsiz, int flags) { ssize_t nr; int input, output, save_errno, i, need_restart; char ch, *p, *end; struct termios term, oterm; struct sigaction sa, savealrm, saveint, savehup, savequit, saveterm; struct sigaction savetstp, savettin, savettou, savepipe; /* I suppose we could alloc on demand in this case (XXX). */ if (bufsiz == 0) { errno = EINVAL; return(NULL); } restart: for (i = 0; i <= MAX_SIGNO; i++) signo[i] = 0; nr = -1; save_errno = 0; need_restart = 0; /* * Read and write to /dev/tty if available. If not, read from * stdin and write to stderr unless a tty is required. */ if ((flags & RPP_STDIN) || (input = output = open(_PATH_TTY, O_RDWR)) == -1) { if (flags & RPP_REQUIRE_TTY) { errno = ENOTTY; return(NULL); } input = STDIN_FILENO; output = STDERR_FILENO; } /* * Catch signals that would otherwise cause the user to end * up with echo turned off in the shell. Don't worry about * things like SIGXCPU and SIGVTALRM for now. */ sigemptyset(&sa.sa_mask); sa.sa_flags = 0; /* don't restart system calls */ sa.sa_handler = handler; /* Keep this list in sync with MAX_SIGNO! */ (void)sigaction(SIGALRM, &sa, &savealrm); (void)sigaction(SIGHUP, &sa, &savehup); (void)sigaction(SIGINT, &sa, &saveint); (void)sigaction(SIGPIPE, &sa, &savepipe); (void)sigaction(SIGQUIT, &sa, &savequit); (void)sigaction(SIGTERM, &sa, &saveterm); (void)sigaction(SIGTSTP, &sa, &savetstp); (void)sigaction(SIGTTIN, &sa, &savettin); (void)sigaction(SIGTTOU, &sa, &savettou); /* Turn off echo if possible. */ if (input != STDIN_FILENO && tcgetattr(input, &oterm) == 0) { memcpy(&term, &oterm, sizeof(term)); if (!(flags & RPP_ECHO_ON)) term.c_lflag &= ~(ECHO | ECHONL); #ifdef VSTATUS if (term.c_cc[VSTATUS] != _POSIX_VDISABLE) term.c_cc[VSTATUS] = _POSIX_VDISABLE; #endif (void)tcsetattr(input, _T_FLUSH, &term); } else { memset(&term, 0, sizeof(term)); term.c_lflag |= ECHO; memset(&oterm, 0, sizeof(oterm)); oterm.c_lflag |= ECHO; } /* No I/O if we are already backgrounded. */ if (signo[SIGTTOU] != 1 && signo[SIGTTIN] != 1) { if (!(flags & RPP_STDIN)) { int r = write(output, prompt, strlen(prompt)); (void)r; } end = buf + bufsiz - 1; p = buf; while ((nr = read(input, &ch, 1)) == 1 && ch != '\n' && ch != '\r') { if (p < end) { if ((flags & RPP_SEVENBIT)) ch &= 0x7f; if (isalpha((unsigned char)ch)) { if ((flags & RPP_FORCELOWER)) ch = (char)tolower((unsigned char)ch); if ((flags & RPP_FORCEUPPER)) ch = (char)toupper((unsigned char)ch); } *p++ = ch; } } *p = '\0'; save_errno = errno; if (!(term.c_lflag & ECHO)) { int r = write(output, "\n", 1); (void)r; } } /* Restore old terminal settings and signals. */ if (memcmp(&term, &oterm, sizeof(term)) != 0) { while (tcsetattr(input, _T_FLUSH, &oterm) == -1 && errno == EINTR) continue; } (void)sigaction(SIGALRM, &savealrm, NULL); (void)sigaction(SIGHUP, &savehup, NULL); (void)sigaction(SIGINT, &saveint, NULL); (void)sigaction(SIGQUIT, &savequit, NULL); (void)sigaction(SIGPIPE, &savepipe, NULL); (void)sigaction(SIGTERM, &saveterm, NULL); (void)sigaction(SIGTSTP, &savetstp, NULL); (void)sigaction(SIGTTIN, &savettin, NULL); (void)sigaction(SIGTTOU, &savettou, NULL); if (input != STDIN_FILENO) (void)close(input); /* * If we were interrupted by a signal, resend it to ourselves * now that we have restored the signal handlers. */ for (i = 0; i <= MAX_SIGNO; i++) { if (signo[i]) { kill(getpid(), i); switch (i) { case SIGTSTP: case SIGTTIN: case SIGTTOU: need_restart = 1; } } } if (need_restart) goto restart; if (save_errno) errno = save_errno; return(nr == -1 ? NULL : buf); } #endif /* _WIN32 && !__CYGWIN__ */ #endif /* HAVE_READPASSPHRASE */ char * lafe_readpassphrase(const char *prompt, char *buf, size_t bufsiz) { char *p; p = readpassphrase(prompt, buf, bufsiz, RPP_ECHO_OFF); if (p == NULL) { switch (errno) { case EINTR: break; default: lafe_errc(1, errno, "Couldn't read passphrase"); break; } } return (p); } Index: stable/10 =================================================================== --- stable/10 (revision 302424) +++ stable/10 (revision 302425) Property changes on: stable/10 ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head:r302294