1 #ifndef _WIMLIB_DENTRY_H
2 #define _WIMLIB_DENTRY_H
4 #include "wimlib/list.h"
5 #include "wimlib/rbtree.h"
6 #include "wimlib/sha1.h"
7 #include "wimlib/types.h"
10 #include <sys/types.h> /* uid_t, gid_t */
16 struct wim_lookup_table;
17 struct wim_lookup_table_entry;
21 /* Size of the struct wim_dentry up to and including the file_name_len. */
22 #define WIM_DENTRY_DISK_SIZE 102
24 /* Size of on-disk WIM alternate data stream entry, in bytes, up to and
25 * including the stream length field (see below). */
26 #define WIM_ADS_ENTRY_DISK_SIZE 38
29 * Reparse tags documented at
30 * http://msdn.microsoft.com/en-us/library/dd541667(v=prot.10).aspx
32 #define WIM_IO_REPARSE_TAG_RESERVED_ZERO 0x00000000
33 #define WIM_IO_REPARSE_TAG_RESERVED_ONE 0x00000001
34 #define WIM_IO_REPARSE_TAG_MOUNT_POINT 0xA0000003
35 #define WIM_IO_REPARSE_TAG_HSM 0xC0000004
36 #define WIM_IO_REPARSE_TAG_HSM2 0x80000006
37 #define WIM_IO_REPARSE_TAG_DRIVER_EXTENDER 0x80000005
38 #define WIM_IO_REPARSE_TAG_SIS 0x80000007
39 #define WIM_IO_REPARSE_TAG_DFS 0x8000000A
40 #define WIM_IO_REPARSE_TAG_DFSR 0x80000012
41 #define WIM_IO_REPARSE_TAG_FILTER_MANAGER 0x8000000B
42 #define WIM_IO_REPARSE_TAG_SYMLINK 0xA000000C
44 #define FILE_ATTRIBUTE_READONLY 0x00000001
45 #define FILE_ATTRIBUTE_HIDDEN 0x00000002
46 #define FILE_ATTRIBUTE_SYSTEM 0x00000004
47 #define FILE_ATTRIBUTE_DIRECTORY 0x00000010
48 #define FILE_ATTRIBUTE_ARCHIVE 0x00000020
49 #define FILE_ATTRIBUTE_DEVICE 0x00000040
50 #define FILE_ATTRIBUTE_NORMAL 0x00000080
51 #define FILE_ATTRIBUTE_TEMPORARY 0x00000100
52 #define FILE_ATTRIBUTE_SPARSE_FILE 0x00000200
53 #define FILE_ATTRIBUTE_REPARSE_POINT 0x00000400
54 #define FILE_ATTRIBUTE_COMPRESSED 0x00000800
55 #define FILE_ATTRIBUTE_OFFLINE 0x00001000
56 #define FILE_ATTRIBUTE_NOT_CONTENT_INDEXED 0x00002000
57 #define FILE_ATTRIBUTE_ENCRYPTED 0x00004000
58 #define FILE_ATTRIBUTE_VIRTUAL 0x00010000
61 /* Alternate data stream entry.
63 * We read this from disk in the read_ads_entries() function; see that function
64 * for more explanation. */
65 struct wim_ads_entry {
67 /* SHA-1 message digest of stream contents */
68 u8 hash[SHA1_HASH_SIZE];
70 /* The corresponding lookup table entry (only for resolved
72 struct wim_lookup_table_entry *lte;
75 /* Length of UTF16-encoded stream name, in bytes, not including the
76 * terminating null character. */
77 u16 stream_name_nbytes;
79 /* Number to identify an alternate data stream even after it's possibly
80 * been moved or renamed. */
83 /* Stream name (UTF-16LE) */
84 utf16lechar *stream_name;
86 /* Unused field. We read it into memory so we can write it out
93 ads_entries_have_same_name(const struct wim_ads_entry *entry_1,
94 const struct wim_ads_entry *entry_2)
96 return entry_1->stream_name_nbytes == entry_2->stream_name_nbytes &&
97 memcmp(entry_1->stream_name, entry_2->stream_name,
98 entry_1->stream_name_nbytes) == 0;
102 * In-memory structure for a WIM directory entry (dentry). There is a directory
103 * tree for each image in the WIM.
105 * Note that this is a directory entry and not an inode. Since NTFS allows hard
106 * links, it's possible for a NTFS inode to correspond to multiple WIM dentries.
107 * The hard link group ID field of the on-disk WIM dentry tells us the number of
108 * the NTFS inode that the dentry corresponds to (and this gets placed in
111 * Unfortunately, WIM files do not have an analogue to an inode; instead certain
112 * information, such as file attributes, the security descriptor, and file
113 * streams is replicated in each hard-linked dentry, even though this
114 * information really is associated with an inode. In-memory, we fix up this
115 * flaw by allocating a `struct wim_inode' for each dentry that contains some of
116 * this duplicated information, then combining the inodes for each hard link
119 * Confusingly, it's possible for stream information to be missing from a dentry
120 * in a hard link set, in which case the stream information needs to be gotten
121 * from one of the other dentries in the hard link set. In addition, it is
122 * possible for dentries to have inconsistent security IDs, file attributes, or
123 * file streams when they share the same hard link ID (don't even ask. I hope
124 * that Microsoft may have fixed this problem, since I've only noticed it in the
125 * 'install.wim' for Windows 7). For those dentries, we have to use the
126 * conflicting fields to split up the hard link groups. (See
127 * dentry_tree_fix_inodes() in hardlink.c).
130 /* The inode for this dentry */
131 struct wim_inode *d_inode;
133 /* Red-black tree of sibling dentries */
134 struct rb_node rb_node;
136 /* Length of UTF-16LE encoded short filename, in bytes, not including
137 * the terminating zero wide-character. */
138 u16 short_name_nbytes;
140 /* Length of UTF-16LE encoded "long" file name, in bytes, not including
141 * the terminating null character. */
142 u16 file_name_nbytes;
144 /* Length of full path name encoded using "tchars", in bytes, not
145 * including the terminating null character. */
146 u32 full_path_nbytes;
148 /* Does this dentry need to be extracted? */
149 u8 needs_extraction : 1;
151 /* Only used during NTFS capture */
152 u8 is_win32_name : 1;
155 struct list_head tmp_list;
157 /* List of dentries in the inode (hard link set) */
158 struct list_head d_alias;
160 /* The parent of this directory entry. */
161 struct wim_dentry *parent;
164 * Size of directory entry on disk, in bytes. Typical size is around
167 * It is possible for the length field to be 0. This situation, which
168 * is undocumented, indicates the end of a list of sibling nodes in a
169 * directory. It also means the real length is 8, because the dentry
170 * included only the length field, but that takes up 8 bytes.
172 * The length here includes the base directory entry on disk as well as
173 * the long and short filenames. It does NOT include any alternate
174 * stream entries that may follow the directory entry, even though the
175 * size of those needs to be considered. The length SHOULD be 8-byte
176 * aligned, although we don't require it to be. We do require the
177 * length to be large enough to hold the file name(s) of the dentry;
178 * additionally, a warning is issued if this field is larger than the
183 /* The offset, from the start of the uncompressed WIM metadata resource
184 * for this image, of this dentry's child dentries. 0 if the directory
185 * entry has no children, which is the case for regular files or reparse
189 /* Pointer to the UTF-16LE short filename (malloc()ed buffer) */
190 utf16lechar *short_name;
192 /* Pointer to the UTF-16LE filename (malloc()ed buffer). */
193 utf16lechar *file_name;
195 /* Full path of this dentry */
199 #define rbnode_dentry(node) container_of(node, struct wim_dentry, rb_node)
204 * As mentioned above, in the WIM file that is no on-disk analogue of a real
205 * inode, as most of these fields are duplicated in the dentries.
208 /* Timestamps for the inode. The timestamps are the number of
209 * 100-nanosecond intervals that have elapsed since 12:00 A.M., January
210 * 1st, 1601, UTC. This is the same format used in NTFS inodes. */
212 u64 i_last_access_time;
213 u64 i_last_write_time;
215 /* The file attributes associated with this inode. This is a bitwise OR
216 * of the FILE_ATTRIBUTE_* flags. */
219 /* The index of the security descriptor in the WIM image's table of
220 * security descriptors that contains this file's security information.
221 * If -1, no security information exists for this file. */
222 int32_t i_security_id;
224 /* %true iff the inode's lookup table entries has been resolved (i.e.
225 * the @lte field is valid, but the @hash field is not valid)
227 * (This is not an on-disk field.) */
230 /* %true iff verify_inode() has run on this inode. */
235 /* Used only in NTFS-mode extraction */
236 u8 i_dos_name_extracted : 1;
238 /* Set to 0 if reparse point fixups have been done. Otherwise set to 1.
240 * Note: this actually may reflect the SYMBOLIC_LINK_RELATIVE flag. */
243 /* Number of alternate data streams associated with this inode */
246 /* Unused/unknown fields that we just read into memory so we can
247 * re-write them unchanged. */
253 /* A hash of the file's contents, or a pointer to the lookup table entry
254 * for this dentry if the lookup table entries have been resolved.
256 * More specifically, this is for the un-named default file stream, as
257 * opposed to the alternate (named) file streams, which may have their
258 * own lookup table entries. */
260 u8 i_hash[SHA1_HASH_SIZE];
261 struct wim_lookup_table_entry *i_lte;
264 /* Identity of a reparse point. See
265 * http://msdn.microsoft.com/en-us/library/windows/desktop/aa365503(v=vs.85).aspx
266 * for what a reparse point is. */
269 /* Number of dentries that reference this inode */
272 /* Alternate data stream entries. */
273 struct wim_ads_entry *i_ads_entries;
278 /* Device number, used only during image capture */
281 /* List of dentries that reference this inode (there should be
282 * link_count of them) */
283 struct list_head i_dentry;
286 struct hlist_node i_hlist;
287 struct list_head i_list;
290 tchar *i_extracted_file;
292 /* Root of a red-black tree storing the children of this inode (if
293 * non-empty, implies the inode is a directory, although that is also
294 * noted in the @attributes field.) */
295 struct rb_root i_children;
297 /* Next alternate data stream ID to be assigned */
298 u32 i_next_stream_id;
301 /* wimfs file descriptors table for the inode */
302 u16 i_num_opened_fds;
303 u16 i_num_allocated_fds;
304 struct wimfs_fd **i_fds;
305 /* This mutex protects the inode's file descriptors table during
306 * read-only mounts. Read-write mounts are still restricted to 1
308 pthread_mutex_t i_mutex;
312 #define inode_for_each_dentry(dentry, inode) \
313 list_for_each_entry((dentry), &(inode)->i_dentry, d_alias)
315 #define inode_add_dentry(dentry, inode) \
316 list_add_tail(&(dentry)->d_alias, &(inode)->i_dentry)
318 #define inode_first_dentry(inode) \
319 container_of(inode->i_dentry.next, struct wim_dentry, d_alias)
322 dentry_is_first_in_inode(const struct wim_dentry *dentry)
324 return inode_first_dentry(dentry->d_inode) == dentry;
328 dentry_correct_total_length(const struct wim_dentry *dentry);
331 for_dentry_in_tree(struct wim_dentry *root,
332 int (*visitor)(struct wim_dentry*, void*),
336 for_dentry_in_rbtree(struct rb_node *node,
337 int (*visitor)(struct wim_dentry *, void *),
341 for_dentry_child(const struct wim_dentry *dentry,
342 int (*visitor)(struct wim_dentry *, void *),
345 return for_dentry_in_rbtree(dentry->d_inode->i_children.rb_node,
351 for_dentry_in_tree_depth(struct wim_dentry *root,
352 int (*visitor)(struct wim_dentry*, void*),
356 calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p);
359 set_dentry_name(struct wim_dentry *dentry, const tchar *new_name);
361 extern struct wim_dentry *
362 get_dentry(struct WIMStruct *w, const tchar *path);
364 extern struct wim_inode *
365 wim_pathname_to_inode(struct WIMStruct *w, const tchar *path);
367 extern struct wim_dentry *
368 get_dentry_child_with_name(const struct wim_dentry *dentry,
371 extern struct wim_dentry *
372 get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry,
373 const utf16lechar *name,
376 extern struct wim_dentry *
377 get_parent_dentry(struct WIMStruct *w, const tchar *path);
380 print_dentry(struct wim_dentry *dentry, void *lookup_table);
383 print_dentry_full_path(struct wim_dentry *entry, void *ignore);
386 calculate_dentry_tree_full_paths(struct wim_dentry *root);
389 dentry_full_path(struct wim_dentry *dentry);
391 extern struct wim_inode *
392 new_timeless_inode(void);
395 new_dentry(const tchar *name, struct wim_dentry **dentry_ret);
398 new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret);
401 new_dentry_with_timeless_inode(const tchar *name, struct wim_dentry **dentry_ret);
404 new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret);
407 free_inode(struct wim_inode *inode);
410 free_dentry(struct wim_dentry *dentry);
413 put_dentry(struct wim_dentry *dentry);
416 free_dentry_tree(struct wim_dentry *root,
417 struct wim_lookup_table *lookup_table);
420 unlink_dentry(struct wim_dentry *dentry);
422 extern struct wim_dentry *
423 dentry_add_child(struct wim_dentry * restrict parent,
424 struct wim_dentry * restrict child);
426 extern struct wim_ads_entry *
427 inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name,
430 extern struct wim_ads_entry *
431 inode_add_ads_utf16le(struct wim_inode *inode,
432 const utf16lechar *stream_name,
433 size_t stream_name_nbytes);
435 extern struct wim_ads_entry *
436 inode_add_ads(struct wim_inode *dentry, const tchar *stream_name);
439 inode_add_ads_with_data(struct wim_inode *inode, const tchar *name,
440 const void *value, size_t size,
441 struct wim_lookup_table *lookup_table);
444 inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len,
445 struct wim_lookup_table *lookup_table);
448 inode_remove_ads(struct wim_inode *inode, u16 idx,
449 struct wim_lookup_table *lookup_table);
452 #define WIMLIB_UNIX_DATA_TAG "$$__wimlib_UNIX_data"
453 #define WIMLIB_UNIX_DATA_TAG_NBYTES (sizeof(WIMLIB_UNIX_DATA_TAG) - 1)
455 #define WIMLIB_UNIX_DATA_TAG_UTF16LE "$\0$\0_\0_\0w\0i\0m\0l\0i\0b\0_\0U\0N\0I\0X\0_\0d\0a\0t\0a\0"
456 #define WIMLIB_UNIX_DATA_TAG_UTF16LE_NBYTES (sizeof(WIMLIB_UNIX_DATA_TAG_UTF16LE) - 1)
458 /* Format for special alternate data stream entries to store UNIX data for files
459 * and directories (see: WIMLIB_ADD_FLAG_UNIX_DATA) */
460 struct wimlib_unix_data {
461 u16 version; /* Must be 0 */
469 #define NO_UNIX_DATA (-1)
470 #define BAD_UNIX_DATA (-2)
472 inode_get_unix_data(const struct wim_inode *inode,
473 struct wimlib_unix_data *unix_data,
474 u16 *stream_idx_ret);
476 #define UNIX_DATA_UID 0x1
477 #define UNIX_DATA_GID 0x2
478 #define UNIX_DATA_MODE 0x4
479 #define UNIX_DATA_ALL (UNIX_DATA_UID | UNIX_DATA_GID | UNIX_DATA_MODE)
480 #define UNIX_DATA_CREATE 0x8
482 inode_set_unix_data(struct wim_inode *inode, uid_t uid, gid_t gid, mode_t mode,
483 struct wim_lookup_table *lookup_table, int which);
487 read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
488 u64 offset, struct wim_dentry *dentry);
492 read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
493 struct wim_dentry *dentry);
496 write_dentry_tree(const struct wim_dentry *tree, u8 *p);
499 dentry_is_root(const struct wim_dentry *dentry)
501 return dentry->parent == dentry;
505 inode_is_directory(const struct wim_inode *inode)
507 return (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
508 && !(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
512 dentry_is_directory(const struct wim_dentry *dentry)
514 return inode_is_directory(dentry->d_inode);
517 /* For our purposes, we consider "real" symlinks and "junction points" to both
520 inode_is_symlink(const struct wim_inode *inode)
522 return (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
523 && (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
524 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT);
528 inode_is_regular_file(const struct wim_inode *inode)
530 return !inode_is_directory(inode) && !inode_is_symlink(inode);
534 dentry_is_regular_file(const struct wim_dentry *dentry)
536 return inode_is_regular_file(dentry->d_inode);
540 inode_has_children(const struct wim_inode *inode)
542 return inode->i_children.rb_node != NULL;
546 dentry_has_children(const struct wim_dentry *dentry)
548 return inode_has_children(dentry->d_inode);
552 dentry_has_short_name(const struct wim_dentry *dentry)
554 return dentry->short_name_nbytes != 0;
558 dentry_has_long_name(const struct wim_dentry *dentry)
560 return dentry->file_name_nbytes != 0;
564 inode_ref_streams(struct wim_inode *inode);
567 dentry_tree_fix_inodes(struct wim_dentry *root, struct list_head *inode_list);
570 verify_dentry(struct wim_dentry *dentry, void *wim);