2 * unix_capture.c: Capture a directory tree on UNIX.
6 * Copyright (C) 2012, 2013, 2014 Eric Biggers
8 * This file is free software; you can redistribute it and/or modify it under
9 * the terms of the GNU Lesser General Public License as published by the Free
10 * Software Foundation; either version 3 of the License, or (at your option) any
13 * This file is distributed in the hope that it will be useful, but WITHOUT
14 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
15 * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
18 * You should have received a copy of the GNU Lesser General Public License
19 * along with this file; if not, see http://www.gnu.org/licenses/.
31 #include <limits.h> /* for PATH_MAX */
35 #include "wimlib/blob_table.h"
36 #include "wimlib/capture.h"
37 #include "wimlib/dentry.h"
38 #include "wimlib/error.h"
39 #include "wimlib/reparse.h"
40 #include "wimlib/timestamp.h"
41 #include "wimlib/unix_data.h"
44 # define my_fdopendir(dirfd_p) fdopendir(*(dirfd_p))
47 my_fdopendir(int *dirfd_p)
52 old_pwd = open(".", O_RDONLY);
54 if (!fchdir(*dirfd_p)) {
58 *dirfd_p = dirfd(dir);
69 # define my_openat(full_path, dirfd, relpath, flags) \
70 openat((dirfd), (relpath), (flags))
72 # define my_openat(full_path, dirfd, relpath, flags) \
73 open((full_path), (flags))
76 #ifdef HAVE_READLINKAT
77 # define my_readlinkat(full_path, dirfd, relpath, buf, bufsize) \
78 readlinkat((dirfd), (relpath), (buf), (bufsize))
80 # define my_readlinkat(full_path, dirfd, relpath, buf, bufsize) \
81 readlink((full_path), (buf), (bufsize))
85 # define my_fstatat(full_path, dirfd, relpath, stbuf, flags) \
86 fstatat((dirfd), (relpath), (stbuf), (flags))
88 # define my_fstatat(full_path, dirfd, relpath, stbuf, flags) \
89 ((flags) & AT_SYMLINK_NOFOLLOW) ? \
90 lstat((full_path), (stbuf)) : \
91 stat((full_path), (stbuf))
95 # define AT_FDCWD -100
98 #ifndef AT_SYMLINK_NOFOLLOW
99 # define AT_SYMLINK_NOFOLLOW 0x100
103 unix_scan_regular_file(const char *path, u64 size, struct wim_inode *inode,
104 struct list_head *unhashed_blobs)
106 struct blob_descriptor *blob = NULL;
107 struct wim_inode_stream *strm;
109 inode->i_attributes = FILE_ATTRIBUTE_NORMAL;
112 blob = new_blob_descriptor();
115 blob->file_on_disk = STRDUP(path);
116 if (unlikely(!blob->file_on_disk))
118 blob->blob_location = BLOB_IN_FILE_ON_DISK;
120 blob->file_inode = inode;
123 strm = inode_add_stream(inode, STREAM_TYPE_DATA, NO_STREAM_NAME, blob);
127 prepare_unhashed_blob(blob, inode, strm->stream_id, unhashed_blobs);
131 free_blob_descriptor(blob);
132 return WIMLIB_ERR_NOMEM;
136 unix_build_dentry_tree_recursive(struct wim_dentry **tree_ret,
137 char *path, size_t path_len,
138 int dirfd, const char *relpath,
139 struct capture_params *params);
142 unix_scan_directory(struct wim_dentry *dir_dentry,
143 char *full_path, size_t full_path_len,
144 int parent_dirfd, const char *dir_relpath,
145 struct capture_params *params)
152 dirfd = my_openat(full_path, parent_dirfd, dir_relpath, O_RDONLY);
154 ERROR_WITH_ERRNO("\"%s\": Can't open directory", full_path);
155 return WIMLIB_ERR_OPENDIR;
158 dir_dentry->d_inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY;
159 dir = my_fdopendir(&dirfd);
161 ERROR_WITH_ERRNO("\"%s\": Can't open directory", full_path);
163 return WIMLIB_ERR_OPENDIR;
168 struct dirent *entry;
169 struct wim_dentry *child;
173 entry = readdir(dir);
176 ret = WIMLIB_ERR_READ;
177 ERROR_WITH_ERRNO("\"%s\": Error reading directory",
183 if (entry->d_name[0] == '.' &&
184 (entry->d_name[1] == '\0' ||
185 (entry->d_name[1] == '.' && entry->d_name[2] == '\0')))
188 full_path[full_path_len] = '/';
189 name_len = strlen(entry->d_name);
190 memcpy(&full_path[full_path_len + 1], entry->d_name, name_len + 1);
191 ret = unix_build_dentry_tree_recursive(&child,
193 full_path_len + 1 + name_len,
195 &full_path[full_path_len + 1],
197 full_path[full_path_len] = '\0';
201 dentry_add_child(dir_dentry, child);
207 /* Given an absolute symbolic link target @dest (UNIX-style, beginning
208 * with '/'), determine whether it points into the directory specified by
209 * @ino and @dev. If so, return the target modified to be "absolute"
210 * relative to this directory. Otherwise, return NULL. */
212 unix_fixup_abslink(char *dest, u64 ino, u64 dev)
221 /* Skip non-slashes. */
222 while (*p && *p != '/')
226 while (*p && *p == '/')
229 /* Get inode and device for this prefix. */
232 ret = stat(dest, &stbuf);
236 /* stat() failed. Assume the link points outside the
237 * directory tree being captured. */
241 if (stbuf.st_ino == ino && stbuf.st_dev == dev) {
242 /* Link points inside directory tree being captured.
243 * Return abbreviated path. */
245 while (p > dest && *(p - 1) == '/')
251 /* Link does not point inside directory tree being captured. */
256 unix_scan_symlink(const char *full_path, int dirfd, const char *relpath,
257 struct wim_inode *inode, struct capture_params *params)
259 char deref_name_buf[4096];
260 ssize_t deref_name_len;
264 inode->i_attributes = FILE_ATTRIBUTE_REPARSE_POINT;
265 inode->i_reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK;
267 /* The idea here is to call readlink() to get the UNIX target of the
268 * symbolic link, then turn the target into a reparse point data buffer
269 * that contains a relative or absolute symbolic link. */
270 deref_name_len = my_readlinkat(full_path, dirfd, relpath,
271 deref_name_buf, sizeof(deref_name_buf) - 1);
272 if (deref_name_len < 0) {
273 ERROR_WITH_ERRNO("\"%s\": Can't read target of symbolic link",
275 return WIMLIB_ERR_READLINK;
278 dest = deref_name_buf;
280 dest[deref_name_len] = '\0';
282 if ((params->add_flags & WIMLIB_ADD_FLAG_RPFIX) &&
287 /* RPFIX (reparse point fixup) mode: Change target of absolute
288 * symbolic link to be "absolute" relative to the tree being
290 fixed_dest = unix_fixup_abslink(dest,
291 params->capture_root_ino,
292 params->capture_root_dev);
293 params->progress.scan.cur_path = full_path;
294 params->progress.scan.symlink_target = deref_name_buf;
296 /* Link points inside the tree being captured, so it was
298 inode->i_not_rpfixed = 0;
300 ret = do_capture_progress(params,
301 WIMLIB_SCAN_DENTRY_FIXED_SYMLINK,
304 /* Link points outside the tree being captured, so it
306 ret = do_capture_progress(params,
307 WIMLIB_SCAN_DENTRY_NOT_FIXED_SYMLINK,
313 ret = wim_inode_set_symlink(inode, dest, params->blob_table);
317 /* Unfortunately, Windows seems to have the concept of "file" symbolic
318 * links as being different from "directory" symbolic links... so
319 * FILE_ATTRIBUTE_DIRECTORY needs to be set on the symbolic link if the
320 * *target* of the symbolic link is a directory. */
322 if (my_fstatat(full_path, dirfd, relpath, &stbuf, 0) == 0 &&
323 S_ISDIR(stbuf.st_mode))
324 inode->i_attributes |= FILE_ATTRIBUTE_DIRECTORY;
329 unix_build_dentry_tree_recursive(struct wim_dentry **tree_ret,
330 char *full_path, size_t full_path_len,
331 int dirfd, const char *relpath,
332 struct capture_params *params)
334 struct wim_dentry *tree = NULL;
335 struct wim_inode *inode = NULL;
340 ret = try_exclude(full_path, full_path_len, params);
341 if (ret < 0) /* Excluded? */
343 if (ret > 0) /* Error? */
346 if (params->add_flags & (WIMLIB_ADD_FLAG_DEREFERENCE |
347 WIMLIB_ADD_FLAG_ROOT))
350 stat_flags = AT_SYMLINK_NOFOLLOW;
352 ret = my_fstatat(full_path, dirfd, relpath, &stbuf, stat_flags);
355 ERROR_WITH_ERRNO("\"%s\": Can't read metadata", full_path);
356 ret = WIMLIB_ERR_STAT;
360 if (!(params->add_flags & WIMLIB_ADD_FLAG_UNIX_DATA)) {
361 if (unlikely(!S_ISREG(stbuf.st_mode) &&
362 !S_ISDIR(stbuf.st_mode) &&
363 !S_ISLNK(stbuf.st_mode)))
365 if (params->add_flags &
366 WIMLIB_ADD_FLAG_NO_UNSUPPORTED_EXCLUDE)
368 ERROR("\"%s\": File type is unsupported",
370 ret = WIMLIB_ERR_UNSUPPORTED_FILE;
373 params->progress.scan.cur_path = full_path;
374 ret = do_capture_progress(params,
375 WIMLIB_SCAN_DENTRY_UNSUPPORTED,
381 ret = inode_table_new_dentry(params->inode_table, relpath,
382 stbuf.st_ino, stbuf.st_dev,
383 S_ISDIR(stbuf.st_mode), &tree);
387 inode = tree->d_inode;
389 /* Already seen this inode? */
390 if (inode->i_nlink > 1)
393 #ifdef HAVE_STAT_NANOSECOND_PRECISION
394 inode->i_creation_time = timespec_to_wim_timestamp(&stbuf.st_mtim);
395 inode->i_last_write_time = timespec_to_wim_timestamp(&stbuf.st_mtim);
396 inode->i_last_access_time = timespec_to_wim_timestamp(&stbuf.st_atim);
398 inode->i_creation_time = time_t_to_wim_timestamp(stbuf.st_mtime);
399 inode->i_last_write_time = time_t_to_wim_timestamp(stbuf.st_mtime);
400 inode->i_last_access_time = time_t_to_wim_timestamp(stbuf.st_atime);
402 if (params->add_flags & WIMLIB_ADD_FLAG_UNIX_DATA) {
403 struct wimlib_unix_data unix_data;
405 unix_data.uid = stbuf.st_uid;
406 unix_data.gid = stbuf.st_gid;
407 unix_data.mode = stbuf.st_mode;
408 unix_data.rdev = stbuf.st_rdev;
409 if (!inode_set_unix_data(inode, &unix_data, UNIX_DATA_ALL)) {
410 ret = WIMLIB_ERR_NOMEM;
415 if (params->add_flags & WIMLIB_ADD_FLAG_ROOT) {
416 params->capture_root_ino = stbuf.st_ino;
417 params->capture_root_dev = stbuf.st_dev;
418 params->add_flags &= ~WIMLIB_ADD_FLAG_ROOT;
421 if (S_ISREG(stbuf.st_mode)) {
422 ret = unix_scan_regular_file(full_path, stbuf.st_size,
423 inode, params->unhashed_blobs);
424 } else if (S_ISDIR(stbuf.st_mode)) {
425 ret = unix_scan_directory(tree, full_path, full_path_len,
426 dirfd, relpath, params);
427 } else if (S_ISLNK(stbuf.st_mode)) {
428 ret = unix_scan_symlink(full_path, dirfd, relpath,
436 params->progress.scan.cur_path = full_path;
438 ret = do_capture_progress(params, WIMLIB_SCAN_DENTRY_OK, inode);
440 ret = do_capture_progress(params, WIMLIB_SCAN_DENTRY_EXCLUDED, NULL);
443 free_dentry_tree(tree, params->blob_table);
445 ret = report_capture_error(params, ret, full_path);
452 * unix_build_dentry_tree():
453 * Builds a tree of WIM dentries from an on-disk directory tree (UNIX
454 * version; no NTFS-specific data is captured).
456 * @root_ret: Place to return a pointer to the root of the dentry tree. Only
457 * modified if successful. Set to NULL if the file or directory was
458 * excluded from capture.
460 * @root_disk_path: The path to the root of the directory tree on disk.
462 * @params: See doc for `struct capture_params'.
464 * @return: 0 on success, nonzero on failure. It is a failure if any of
465 * the files cannot be `stat'ed, or if any of the needed
466 * directories cannot be opened or read. Failure to add the files
467 * to the WIM may still occur later when trying to actually read
468 * the on-disk files during a call to wimlib_write() or
469 * wimlib_overwrite().
472 unix_build_dentry_tree(struct wim_dentry **root_ret,
473 const char *root_disk_path,
474 struct capture_params *params)
481 path_len = strlen(root_disk_path);
482 path_bufsz = min(32790, PATH_MAX + 1);
484 if (path_len >= path_bufsz)
485 return WIMLIB_ERR_INVALID_PARAM;
487 path_buf = MALLOC(path_bufsz);
489 return WIMLIB_ERR_NOMEM;
490 memcpy(path_buf, root_disk_path, path_len + 1);
492 params->capture_root_nchars = path_len;
494 ret = unix_build_dentry_tree_recursive(root_ret, path_buf, path_len,
495 AT_FDCWD, path_buf, params);
500 #endif /* !__WIN32__ */