diff options
Diffstat (limited to 'Utilities/cmlibarchive/libarchive/archive_read_disk_posix.c')
| -rw-r--r-- | Utilities/cmlibarchive/libarchive/archive_read_disk_posix.c | 2722 |
1 files changed, 2722 insertions, 0 deletions
diff --git a/Utilities/cmlibarchive/libarchive/archive_read_disk_posix.c b/Utilities/cmlibarchive/libarchive/archive_read_disk_posix.c new file mode 100644 index 0000000..52fec7b --- /dev/null +++ b/Utilities/cmlibarchive/libarchive/archive_read_disk_posix.c @@ -0,0 +1,2722 @@ +/*- + * 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 <sys/param.h> +#endif +#ifdef HAVE_SYS_MOUNT_H +#include <sys/mount.h> +#endif +#ifdef HAVE_SYS_STAT_H +#include <sys/stat.h> +#endif +#ifdef HAVE_SYS_STATFS_H +#include <sys/statfs.h> +#endif +#ifdef HAVE_SYS_STATVFS_H +#include <sys/statvfs.h> +#endif +#ifdef HAVE_SYS_TIME_H +#include <sys/time.h> +#endif +#ifdef HAVE_LINUX_MAGIC_H +#include <linux/magic.h> +#endif +#ifdef HAVE_LINUX_FS_H +#include <linux/fs.h> +#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 <linux/ext2_fs.h> /* for Linux file flags */ +#endif +#if defined(HAVE_EXT2FS_EXT2_FS_H) && !defined(__CYGWIN__) +#include <ext2fs/ext2_fs.h> /* Linux file flags, broken on Cygwin */ +#endif +#ifdef HAVE_DIRECT_H +#include <direct.h> +#endif +#ifdef HAVE_DIRENT_H +#include <dirent.h> +#endif +#ifdef HAVE_ERRNO_H +#include <errno.h> +#endif +#ifdef HAVE_FCNTL_H +#include <fcntl.h> +#endif +#ifdef HAVE_LIMITS_H +#include <limits.h> +#endif +#ifdef HAVE_STDLIB_H +#include <stdlib.h> +#endif +#ifdef HAVE_STRING_H +#include <string.h> +#endif +#ifdef HAVE_UNISTD_H +#include <unistd.h> +#endif +#ifdef HAVE_SYS_IOCTL_H +#include <sys/ioctl.h> +#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(USE_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(USE_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_filesystem; + + 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, la_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, la_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, la_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, la_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->flags = ARCHIVE_READDISK_MAC_COPYFILE; + 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) +{ + 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->flags |= ARCHIVE_READDISK_RESTORE_ATIME; + if (a->tree != NULL) + a->tree->flags |= needsRestoreTimes; + return (ARCHIVE_OK); +#else + /* Display warning and unset flag */ + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "Cannot restore access time on this system"); + a->flags &= ~ARCHIVE_READDISK_RESTORE_ATIME; + 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"); + + a->flags = flags; + + if (flags & ARCHIVE_READDISK_RESTORE_ATIME) + r = archive_read_disk_set_atime_restored(_a); + else { + if (a->tree != NULL) + a->tree->flags &= ~needsRestoreTimes; + } + 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 increment 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; + int64_t sparse_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; +#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; + } +#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; + } + sparse_bytes = t->current_sparse->offset - t->entry_total; + t->entry_remaining_bytes -= sparse_bytes; + t->entry_total += sparse_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 delayed, delayed_errno, descend, r; + struct archive_string delayed_str; + + delayed = ARCHIVE_OK; + delayed_errno = 0; + archive_string_init(&delayed_str); + + 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) { + if (errno == ENOENT && t->depth > 0) { + delayed = ARCHIVE_WARN; + delayed_errno = errno; + if (delayed_str.length == 0) { + archive_string_sprintf(&delayed_str, + "%s", tree_current_path(t)); + } else { + archive_string_sprintf(&delayed_str, + " %s", tree_current_path(t)); + } + } else { + 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->flags & ARCHIVE_READDISK_MAC_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, + "Failed : %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->flags & ARCHIVE_READDISK_NO_TRAVERSE_MOUNTS) { + 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->flags & ARCHIVE_READDISK_HONOR_NODUMP) { +#if defined(HAVE_STRUCT_STAT_ST_FLAGS) && defined(UF_NODUMP) + if (st->st_flags & UF_NODUMP) + return (ARCHIVE_RETRY); +#elif (defined(FS_IOC_GETFLAGS) && defined(FS_NODUMP_FL) && \ + defined(HAVE_WORKING_FS_IOC_GETFLAGS)) || \ + (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, +#ifdef FS_IOC_GETFLAGS + FS_IOC_GETFLAGS, +#else + EXT2_IOC_GETFLAGS, +#endif + &stflags); +#ifdef FS_NODUMP_FL + if (r == 0 && (stflags & FS_NODUMP_FL) != 0) +#else + if (r == 0 && (stflags & EXT2_NODUMP_FL) != 0) +#endif + 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, invoking 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, + "Failed : %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, + "Failed : %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); + + if (r == ARCHIVE_OK) { + r = delayed; + if (r != ARCHIVE_OK) { + archive_string_sprintf(&delayed_str, ": %s", + "File removed before we read it"); + archive_set_error(&(a->archive), delayed_errno, + "%s", delayed_str.s); + } + } + archive_string_free(&delayed_str); + + 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; + } + + archive_entry_clear(entry); + + 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); + + /* + * We must not treat the initial specified path as a physical dir, + * because if we do then we will try and ascend out of it by opening + * ".." which is (a) wrong and (b) causes spurious permissions errors + * if ".." is not readable by us. Instead, treat it as if it were a + * symlink. (This uses an extra fd, but it can only happen once at the + * top level of a traverse.) But we can't necessarily assume t->st is + * valid here (though t->lst is), which complicates the logic a + * little. + */ + 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); + if (t->stack->parent->parent != NULL) + t->stack->flags |= isDir; + else + t->stack->flags |= isDirLink; + } 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->flags & ARCHIVE_READDISK_RESTORE_ATIME); + else + a->tree = tree_open(pathname, a->symlink_mode, + a->flags & ARCHIVE_READDISK_RESTORE_ATIME); + 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 filesystem ID we've already generated. */ + t->current_filesystem_id = i; + t->current_filesystem = &(t->filesystem_table[i]); + return (ARCHIVE_OK); + } + } + + /* + * This is the new filesystem which we have to generate a new ID for. + */ + fid = t->max_filesystem_id++; + if (t->max_filesystem_id > t->allocated_filesystem) { + 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_filesystem = 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) +/* TODO: configure should set GETVFSBYNAME_ARG_TYPE to make + * this accurate; some platforms have both and we need the one that's + * used by getvfsbyname() + * + * Then the following would become: + * #if defined(GETVFSBYNAME_ARG_TYPE) + * GETVFSBYNAME_ARG_TYPE vfc; + * #endif + */ +# if defined(HAVE_STRUCT_XVFSCONF) + struct xvfsconf vfc; +# else + struct vfsconf 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(USE_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 /* USE_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) { + /* Usually 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(USE_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(USE_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(USE_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(USE_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 systems (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 /* USE_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 = calloc(1, 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 = calloc(1, sizeof(*t))) == NULL) + return (NULL); + 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) +{ +#if defined(O_PATH) + /* Linux */ + const int o_flag = O_PATH; +#elif defined(O_SEARCH) + /* SunOS */ + const int o_flag = O_SEARCH; +#elif defined(__FreeBSD__) && defined(O_EXEC) + /* FreeBSD */ + const int o_flag = O_EXEC; +#endif + + t->flags = (restore_time != 0)?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); +#if defined(O_PATH) || defined(O_SEARCH) || \ + (defined(__FreeBSD__) && defined(O_EXEC)) + /* + * Most likely reason to fail opening "." is that it's not readable, + * so try again for execute. The consequences of not opening this are + * unhelpful and unnecessary errors later. + */ + if (t->initial_dir_fd < 0) + t->initial_dir_fd = open(".", o_flag | O_CLOEXEC); +#endif + __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(USE_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(USE_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 /* USE_READDIR_R */ + } + for (;;) { + errno = 0; +#if defined(USE_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 (la_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(USE_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 |