1 #ifndef _WIMLIB_DENTRY_H
2 #define _WIMLIB_DENTRY_H
4 #include "wimlib/compiler.h"
5 #include "wimlib/list.h"
6 #include "wimlib/rbtree.h"
7 #include "wimlib/sha1.h"
8 #include "wimlib/types.h"
11 #include <sys/types.h> /* uid_t, gid_t */
13 struct wim_lookup_table;
14 struct wim_lookup_table_entry;
17 struct wim_security_data;
19 /* Size of the struct wim_dentry up to and including the file_name_len. */
20 #define WIM_DENTRY_DISK_SIZE 102
23 * Reparse tags documented at
24 * http://msdn.microsoft.com/en-us/library/dd541667(v=prot.10).aspx
26 #define WIM_IO_REPARSE_TAG_RESERVED_ZERO 0x00000000
27 #define WIM_IO_REPARSE_TAG_RESERVED_ONE 0x00000001
28 #define WIM_IO_REPARSE_TAG_MOUNT_POINT 0xA0000003
29 #define WIM_IO_REPARSE_TAG_HSM 0xC0000004
30 #define WIM_IO_REPARSE_TAG_HSM2 0x80000006
31 #define WIM_IO_REPARSE_TAG_DRIVER_EXTENDER 0x80000005
32 #define WIM_IO_REPARSE_TAG_SIS 0x80000007
33 #define WIM_IO_REPARSE_TAG_DFS 0x8000000A
34 #define WIM_IO_REPARSE_TAG_DFSR 0x80000012
35 #define WIM_IO_REPARSE_TAG_FILTER_MANAGER 0x8000000B
36 #define WIM_IO_REPARSE_TAG_SYMLINK 0xA000000C
38 #define FILE_ATTRIBUTE_READONLY 0x00000001
39 #define FILE_ATTRIBUTE_HIDDEN 0x00000002
40 #define FILE_ATTRIBUTE_SYSTEM 0x00000004
41 #define FILE_ATTRIBUTE_DIRECTORY 0x00000010
42 #define FILE_ATTRIBUTE_ARCHIVE 0x00000020
43 #define FILE_ATTRIBUTE_DEVICE 0x00000040
44 #define FILE_ATTRIBUTE_NORMAL 0x00000080
45 #define FILE_ATTRIBUTE_TEMPORARY 0x00000100
46 #define FILE_ATTRIBUTE_SPARSE_FILE 0x00000200
47 #define FILE_ATTRIBUTE_REPARSE_POINT 0x00000400
48 #define FILE_ATTRIBUTE_COMPRESSED 0x00000800
49 #define FILE_ATTRIBUTE_OFFLINE 0x00001000
50 #define FILE_ATTRIBUTE_NOT_CONTENT_INDEXED 0x00002000
51 #define FILE_ATTRIBUTE_ENCRYPTED 0x00004000
52 #define FILE_ATTRIBUTE_VIRTUAL 0x00010000
55 /* Alternate data stream entry.
57 * We read this from disk in the read_ads_entries() function; see that function
58 * for more explanation. */
59 struct wim_ads_entry {
61 /* SHA-1 message digest of stream contents */
62 u8 hash[SHA1_HASH_SIZE];
64 /* The corresponding lookup table entry (only for resolved
66 struct wim_lookup_table_entry *lte;
69 /* Length of UTF16-encoded stream name, in bytes, not including the
70 * terminating null character; or 0 if the stream is unnamed. */
71 u16 stream_name_nbytes;
73 /* Number to identify an alternate data stream even after it's possibly
74 * been moved or renamed. */
77 /* Stream name (UTF-16LE), null-terminated, or NULL if the stream is
79 utf16lechar *stream_name;
81 /* Reserved field. We read it into memory so we can write it out
88 ads_entries_have_same_name(const struct wim_ads_entry *entry_1,
89 const struct wim_ads_entry *entry_2)
91 return entry_1->stream_name_nbytes == entry_2->stream_name_nbytes &&
92 memcmp(entry_1->stream_name, entry_2->stream_name,
93 entry_1->stream_name_nbytes) == 0;
97 * In-memory structure for a WIM directory entry (dentry). There is a directory
98 * tree for each image in the WIM.
100 * Note that this is a directory entry and not an inode. Since NTFS allows hard
101 * links, it's possible for a NTFS inode to correspond to multiple WIM dentries.
102 * The hard link group ID field of the on-disk WIM dentry tells us the number of
103 * the NTFS inode that the dentry corresponds to (and this gets placed in
106 * Unfortunately, WIM files do not have an analogue to an inode; instead certain
107 * information, such as file attributes, the security descriptor, and file
108 * streams is replicated in each hard-linked dentry, even though this
109 * information really is associated with an inode. In-memory, we fix up this
110 * flaw by allocating a `struct wim_inode' for each dentry that contains some of
111 * this duplicated information, then combining the inodes for each hard link
114 * Confusingly, it's possible for stream information to be missing from a dentry
115 * in a hard link set, in which case the stream information needs to be gotten
116 * from one of the other dentries in the hard link set. In addition, it is
117 * possible for dentries to have inconsistent security IDs, file attributes, or
118 * file streams when they share the same hard link ID (don't even ask. I hope
119 * that Microsoft may have fixed this problem, since I've only noticed it in the
120 * 'install.wim' for Windows 7). For those dentries, we have to use the
121 * conflicting fields to split up the hard link groups. (See
122 * dentry_tree_fix_inodes() in hardlink.c).
125 /* Pointer to the inode for this dentry. This will contain some
126 * information that was factored out of the on-disk WIM dentry as common
127 * to all dentries in a hard link group. */
128 struct wim_inode *d_inode;
130 /* Node for the parent's red-black tree of child dentries, sorted by
131 * case sensitive long name. */
132 struct rb_node rb_node;
134 /* Node for the parent's red-black tree of child dentries, sorted by
135 * case insensitive long name. */
136 struct rb_node rb_node_case_insensitive;
138 /* List of dentries in a directory that have different case sensitive
139 * long names but share the same case insensitive long name */
140 struct list_head case_insensitive_conflict_list;
142 /* Length of UTF-16LE encoded short filename, in bytes, not including
143 * the terminating zero wide-character. */
144 u16 short_name_nbytes;
146 /* Length of UTF-16LE encoded "long" file name, in bytes, not including
147 * the terminating null character. */
148 u16 file_name_nbytes;
150 /* Length of full path name encoded using "tchars", in bytes, not
151 * including the terminating null character. */
152 u32 full_path_nbytes;
154 /* For extraction operations, this flag will be set on dentries in the
155 * tree being extracted. Otherwise this will always be 0. */
156 u8 in_extraction_tree : 1;
158 /* For extraction operations, this flag will be set when a dentry in the
159 * tree being extracted is not being extracted for some reason (file
160 * type not supported by target filesystem, contains invalid characters,
161 * or not in one of the multiple sub-trees being extracted). Otherwise
162 * this will always be 0. */
163 u8 extraction_skipped : 1;
165 /* During extraction extractions, this flag will be set after the
166 * "skeleton" of the dentry has been extracted. */
167 u8 skeleton_extracted : 1;
169 /* When capturing from a NTFS volume using NTFS-3g, this flag is set on
170 * dentries that were created from a filename in the WIN32 or WIN32+DOS
171 * namespaces rather than the POSIX namespace. Otherwise this will
173 u8 is_win32_name : 1;
175 /* When verifying the dentry tree after reading it into memory, this
176 * flag will be set on all dentries in a hard link group that have a
177 * nonempty DOS name except one. This is because it is supposed to be
178 * illegal (on NTFS, at least) for a single inode to have multiple DOS
180 u8 dos_name_invalid : 1;
184 u8 was_hardlinked : 1;
186 /* Temporary list field used to make lists of dentries in a few places.
188 struct list_head tmp_list;
190 /* Linked list node that places this dentry in the list of aliases for
191 * its inode (d_inode) */
192 struct list_head d_alias;
194 /* The parent of this directory entry. */
195 struct wim_dentry *parent;
197 /* 'length' and 'subdir_offset' are only used while reading and writing
198 * this dentry; see the corresponding field in
199 * `struct wim_dentry_on_disk' for explanation. */
203 /* These correspond to the two unused fields in the on-disk WIM dentry;
204 * we read them into memory so we can write them unchanged. These
205 * fields are set to 0 on new dentries. */
209 /* Pointer to the UTF-16LE short filename (malloc()ed buffer), or NULL
210 * if this dentry has no short name. */
211 utf16lechar *short_name;
213 /* Pointer to the UTF-16LE filename (malloc()ed buffer), or NULL if this
214 * dentry has no filename. */
215 utf16lechar *file_name;
217 /* Full path to this dentry in the WIM, in platform-dependent tchars
218 * that can be printed without conversion. By default this field will
219 * be NULL and will only be calculated on-demand by the
220 * calculate_dentry_full_path() or dentry_full_path() functions. */
223 /* (Extraction only) Actual name to extract this dentry as, along with
224 * its length in tchars excluding the NULL terminator. This usually
225 * will be the same as file_name, with the character encoding converted
226 * if needed. But if file_name contains characters not accepted on the
227 * current platform, then this may be set slightly differently from
228 * file_name. This will be either NULL or a malloc()ed buffer that may
229 * alias file_name. */
230 tchar *extraction_name;
231 size_t extraction_name_nchars;
234 #define rbnode_dentry(node) container_of(node, struct wim_dentry, rb_node)
239 * As mentioned in the comment above `struct wim_dentry', in the WIM file that
240 * is no on-disk analogue of a real inode, as most of these fields are
241 * duplicated in the dentries. Instead, a `struct wim_inode' is something we
242 * create ourselves to simplify the handling of hard links.
245 /* If i_resolved == 0:
246 * SHA1 message digest of the contents of the unnamed-data stream
249 * If i_resolved == 1:
250 * Pointer to the lookup table entry for the unnamed data stream
251 * of this inode, or NULL.
253 * i_hash corresponds to the 'unnamed_stream_hash' field of the `struct
254 * wim_dentry_on_disk' and the additional caveats documented about that
255 * field apply here (for example, the quirks regarding all-zero hashes).
258 u8 i_hash[SHA1_HASH_SIZE];
259 struct wim_lookup_table_entry *i_lte;
262 /* Corresponds to the 'attributes' field of `struct wim_dentry_on_disk';
263 * bitwise OR of the FILE_ATTRIBUTE_* flags that give the attributes of
267 /* Root of a red-black tree storing the child dentries of this inode, if
268 * any. Keyed by wim_dentry->file_name, case sensitively. */
269 struct rb_root i_children;
271 /* Root of a red-black tree storing the children of this inode, if any.
272 * Keyed by wim_dentry->file_name, case insensitively. */
273 struct rb_root i_children_case_insensitive;
275 /* List of dentries that are aliases for this inode. There will be
276 * i_nlink dentries in this list. */
277 struct list_head i_dentry;
279 /* Field to place this inode into a list. */
281 /* Hash list node- used in hardlink.c when the inodes are placed
282 * into a hash table keyed by inode number and optionally device
283 * number, in order to detect dentries that are aliases for the
285 struct hlist_node i_hlist;
287 /* Normal list node- used to connect all the inodes of a WIM image
288 * into a single linked list referenced from the
289 * `struct wim_image_metadata' for that image. */
290 struct list_head i_list;
293 /* Number of dentries that are aliases for this inode. */
296 /* Number of alternate data streams (ADS) associated with this inode */
299 /* Flag that indicates whether this inode's streams have been
300 * "resolved". By default, the inode starts as "unresolved", meaning
301 * that the i_hash field, along with the hash field of any associated
302 * wim_ads_entry's, are valid and should be used as keys in the WIM
303 * lookup table to find the associated `struct wim_lookup_table_entry'.
304 * But if the inode has been resolved, then each of these fields is
305 * replaced with a pointer directly to the appropriate `struct
306 * wim_lookup_table_entry', or NULL if the stream is empty. */
309 /* Flag used to mark this inode as visited; this is used when visiting
310 * all the inodes in a dentry tree exactly once. It will be 0 by
311 * default and must be cleared following the tree traversal, even in
315 /* Set if the DOS name of an inode has already been extracted. */
316 u8 i_dos_name_extracted : 1;
318 /* 1 iff all ADS entries of this inode are named or if this inode
319 * has no ADS entries */
320 u8 i_canonical_streams : 1;
322 /* Pointer to a malloc()ed array of i_num_ads alternate data stream
323 * entries for this inode. */
324 struct wim_ads_entry *i_ads_entries;
326 /* Creation time, last access time, and last write time for this inode, in
327 * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601. They
328 * should correspond to the times gotten by calling GetFileTime() on
331 u64 i_last_access_time;
332 u64 i_last_write_time;
334 /* Corresponds to 'security_id' in `struct wim_dentry_on_disk': The
335 * index of this inode's security descriptor in the WIM image's table of
336 * security descriptors, or -1. Note: in verify_inode(), called
337 * whenever a WIM image is loaded, out-of-bounds indices are set to -1,
338 * so the extraction code does not need to do bounds checks. */
339 int32_t i_security_id;
341 /* Identity of a reparse point. See
342 * http://msdn.microsoft.com/en-us/library/windows/desktop/aa365503(v=vs.85).aspx
343 * for what a reparse point is. */
346 /* Unused/unknown fields that we just read into memory so we can
347 * re-write them unchanged. */
351 /* Corresponds to not_rpfixed in `struct wim_dentry_on_disk': Set to 0
352 * if reparse point fixups have been done. Otherwise set to 1. Note:
353 * this actually may reflect the SYMBOLIC_LINK_RELATIVE flag.
357 /* Inode number; corresponds to hard_link_group_id in the `struct
358 * wim_dentry_on_disk'. */
362 /* Device number, used only during image capture, so we can
363 * identify hard linked files by the combination of inode number
364 * and device number (rather than just inode number, which could
365 * be ambigious if the captured tree spans a mountpoint). Set
371 /* Used only during image extraction: pointer to the first path
372 * (malloc()ed buffer) at which this inode has been extracted.
373 * Freed and set to NULL after the extraction is done (either
374 * success or failure). */
375 tchar *i_extracted_file;
377 /** Used only during image extraction: "cookie" that
378 * identifies this extracted file (inode), for example
379 * an inode number. Only used if supported by the
380 * extraction mode. */
385 /* Used only during image mount: Table of file descriptors that
386 * have been opened to this inode. The table is automatically
387 * freed when the last file descriptor is closed. */
388 struct wimfs_fd **i_fds;
393 u16 i_num_opened_fds;
394 u16 i_num_allocated_fds;
397 /* Next alternate data stream ID to be assigned */
398 u32 i_next_stream_id;
401 #define inode_for_each_dentry(dentry, inode) \
402 list_for_each_entry((dentry), &(inode)->i_dentry, d_alias)
404 #define inode_add_dentry(dentry, inode) \
405 list_add_tail(&(dentry)->d_alias, &(inode)->i_dentry)
407 #define inode_first_dentry(inode) \
408 container_of(inode->i_dentry.next, struct wim_dentry, d_alias)
410 #define inode_first_full_path(inode) \
411 dentry_full_path(inode_first_dentry(inode))
414 dentry_is_first_in_inode(const struct wim_dentry *dentry)
416 return inode_first_dentry(dentry->d_inode) == dentry;
420 dentry_out_total_length(const struct wim_dentry *dentry);
423 for_dentry_in_tree(struct wim_dentry *root,
424 int (*visitor)(struct wim_dentry*, void*),
428 for_dentry_in_rbtree(struct rb_node *node,
429 int (*visitor)(struct wim_dentry *, void *),
433 for_dentry_child(const struct wim_dentry *dentry,
434 int (*visitor)(struct wim_dentry *, void *),
438 for_dentry_in_tree_depth(struct wim_dentry *root,
439 int (*visitor)(struct wim_dentry*, void*),
443 calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p);
446 set_dentry_name(struct wim_dentry *dentry, const tchar *new_name);
449 /* Note: the NTFS-3g headers define CASE_SENSITIVE, hence the WIMLIB prefix. */
451 /* Use either case-sensitive or case-insensitive search, depending on
452 * the variable @default_ignore_case. */
453 WIMLIB_CASE_PLATFORM_DEFAULT = 0,
455 /* Use case-sensitive search. */
456 WIMLIB_CASE_SENSITIVE = 1,
458 /* Use case-insensitive search. */
459 WIMLIB_CASE_INSENSITIVE = 2,
460 } CASE_SENSITIVITY_TYPE;
462 extern bool default_ignore_case;
464 extern struct wim_dentry *
465 get_dentry(struct WIMStruct *wim, const tchar *path,
466 CASE_SENSITIVITY_TYPE case_type);
468 extern struct wim_dentry *
469 get_dentry_child_with_name(const struct wim_dentry *dentry,
471 CASE_SENSITIVITY_TYPE case_type);
473 extern struct wim_dentry *
474 get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry,
475 const utf16lechar *name,
477 CASE_SENSITIVITY_TYPE case_type);
479 extern struct wim_dentry *
480 get_parent_dentry(struct WIMStruct *wim, const tchar *path,
481 CASE_SENSITIVITY_TYPE case_type);
484 print_dentry(struct wim_dentry *dentry, void *lookup_table);
487 print_dentry_full_path(struct wim_dentry *entry, void *ignore);
490 calculate_dentry_full_path(struct wim_dentry *dentry);
493 calculate_dentry_tree_full_paths(struct wim_dentry *root);
496 dentry_full_path(struct wim_dentry *dentry);
498 extern struct wim_inode *
499 new_timeless_inode(void) _malloc_attribute;
502 new_dentry(const tchar *name, struct wim_dentry **dentry_ret);
505 new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret);
508 new_dentry_with_timeless_inode(const tchar *name, struct wim_dentry **dentry_ret);
511 dentry_tree_clear_inode_visited(struct wim_dentry *root);
514 new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret);
517 free_inode(struct wim_inode *inode);
520 free_dentry(struct wim_dentry *dentry);
523 put_dentry(struct wim_dentry *dentry);
526 free_dentry_tree(struct wim_dentry *root,
527 struct wim_lookup_table *lookup_table);
530 unlink_dentry(struct wim_dentry *dentry);
532 extern struct wim_dentry *
533 dentry_add_child(struct wim_dentry * restrict parent,
534 struct wim_dentry * restrict child);
537 rename_wim_path(WIMStruct *wim, const tchar *from, const tchar *to,
538 CASE_SENSITIVITY_TYPE case_type);
540 extern struct wim_ads_entry *
541 inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name,
544 extern struct wim_ads_entry *
545 inode_add_ads_utf16le(struct wim_inode *inode,
546 const utf16lechar *stream_name,
547 size_t stream_name_nbytes);
549 extern struct wim_ads_entry *
550 inode_add_ads(struct wim_inode *dentry, const tchar *stream_name);
553 inode_add_ads_with_data(struct wim_inode *inode, const tchar *name,
554 const void *value, size_t size,
555 struct wim_lookup_table *lookup_table);
558 inode_has_named_stream(const struct wim_inode *inode);
561 inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len,
562 struct wim_lookup_table *lookup_table);
565 inode_remove_ads(struct wim_inode *inode, u16 idx,
566 struct wim_lookup_table *lookup_table);
569 #define WIMLIB_UNIX_DATA_TAG "$$__wimlib_UNIX_data"
570 #define WIMLIB_UNIX_DATA_TAG_NBYTES (sizeof(WIMLIB_UNIX_DATA_TAG) - 1)
572 #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"
573 #define WIMLIB_UNIX_DATA_TAG_UTF16LE_NBYTES (sizeof(WIMLIB_UNIX_DATA_TAG_UTF16LE) - 1)
576 ads_entry_is_unix_data(const struct wim_ads_entry *entry)
578 return (entry->stream_name_nbytes ==
579 WIMLIB_UNIX_DATA_TAG_UTF16LE_NBYTES) &&
580 !memcmp(entry->stream_name, WIMLIB_UNIX_DATA_TAG_UTF16LE,
581 WIMLIB_UNIX_DATA_TAG_UTF16LE_NBYTES);
585 ads_entry_is_named_stream(const struct wim_ads_entry *entry)
587 return entry->stream_name_nbytes != 0 && !ads_entry_is_unix_data(entry);
591 /* Format for special alternate data stream entries to store UNIX data for files
592 * and directories (see: WIMLIB_ADD_FLAG_UNIX_DATA) */
593 struct wimlib_unix_data {
594 u16 version; /* Must be 0 */
600 #define NO_UNIX_DATA (-1)
601 #define BAD_UNIX_DATA (-2)
603 inode_get_unix_data(const struct wim_inode *inode,
604 struct wimlib_unix_data *unix_data,
605 u16 *stream_idx_ret);
607 #define UNIX_DATA_UID 0x1
608 #define UNIX_DATA_GID 0x2
609 #define UNIX_DATA_MODE 0x4
610 #define UNIX_DATA_ALL (UNIX_DATA_UID | UNIX_DATA_GID | UNIX_DATA_MODE)
611 #define UNIX_DATA_CREATE 0x8
613 inode_set_unix_data(struct wim_inode *inode, uid_t uid, gid_t gid, mode_t mode,
614 struct wim_lookup_table *lookup_table, int which);
615 #endif /* !__WIN32__ */
618 inode_has_unix_data(const struct wim_inode *inode);
621 read_dentry(const u8 * restrict metadata_resource,
622 u64 metadata_resource_len, u64 offset,
623 struct wim_dentry * restrict dentry);
626 read_dentry_tree(const u8 * restrict metadata_resource,
627 u64 metadata_resource_len,
628 struct wim_dentry * restrict dentry);
631 write_dentry_tree(const struct wim_dentry * restrict tree,
635 dentry_is_root(const struct wim_dentry *dentry)
637 return dentry->parent == dentry;
641 inode_is_directory(const struct wim_inode *inode)
643 return (inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
644 FILE_ATTRIBUTE_REPARSE_POINT))
645 == FILE_ATTRIBUTE_DIRECTORY;
649 inode_is_encrypted_directory(const struct wim_inode *inode)
651 return ((inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
652 FILE_ATTRIBUTE_ENCRYPTED))
653 == (FILE_ATTRIBUTE_DIRECTORY | FILE_ATTRIBUTE_ENCRYPTED));
657 dentry_is_directory(const struct wim_dentry *dentry)
659 return inode_is_directory(dentry->d_inode);
662 /* For our purposes, we consider "real" symlinks and "junction points" to both
665 inode_is_symlink(const struct wim_inode *inode)
667 return (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
668 && (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
669 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT);
673 inode_has_children(const struct wim_inode *inode)
675 return inode->i_children.rb_node != NULL;
679 dentry_has_children(const struct wim_dentry *dentry)
681 return inode_has_children(dentry->d_inode);
685 dentry_has_short_name(const struct wim_dentry *dentry)
687 return dentry->short_name_nbytes != 0;
691 dentry_has_long_name(const struct wim_dentry *dentry)
693 return dentry->file_name_nbytes != 0;
697 inode_ref_streams(struct wim_inode *inode);
700 dentry_tree_fix_inodes(struct wim_dentry *root, struct list_head *inode_list);
703 verify_inode(struct wim_inode *inode, const struct wim_security_data *sd);
705 #endif /* _WIMLIB_DENTRY_H */