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;
135 /* Node for the parent's red-black tree of child dentries, sorted by
136 * case insensitive long name. */
137 struct rb_node rb_node_case_insensitive;
139 /* List of dentries in a directory that have different case sensitive
140 * long names but share the same case insensitive long name */
141 struct list_head case_insensitive_conflict_list;
144 /* Length of UTF-16LE encoded short filename, in bytes, not including
145 * the terminating zero wide-character. */
146 u16 short_name_nbytes;
148 /* Length of UTF-16LE encoded "long" file name, in bytes, not including
149 * the terminating null character. */
150 u16 file_name_nbytes;
152 /* Length of full path name encoded using "tchars", in bytes, not
153 * including the terminating null character. */
154 u32 full_path_nbytes;
156 /* For extraction operations, a subtree of dentries will have this flag
157 * set so we can keep track of which dentries still need to be
158 * extracted. Otherwise this will always be 0. */
159 u8 needs_extraction : 1;
161 /* For extraction operations, this flag will be set when a dentry in the
162 * tree being extracted is not being extracted for some reason (file
163 * type not supported by target filesystem or contains invalid
164 * characters). Otherwise this will always be 0. */
165 u8 not_extracted : 1;
167 /* When capturing from a NTFS volume using NTFS-3g, this flag is set on
168 * dentries that were created from a filename in the WIN32 or WIN32+DOS
169 * namespaces rather than the POSIX namespace. Otherwise this will
171 u8 is_win32_name : 1;
173 /* When verifying the dentry tree after reading it into memory, this
174 * flag will be set on all dentries in a hard link group that have a
175 * nonempty DOS name except one. This is because it is supposed to be
176 * illegal (on NTFS, at least) for a single inode to have multiple DOS
178 u8 dos_name_invalid : 1;
180 /* Temporary list field used to make lists of dentries in a few places.
182 struct list_head tmp_list;
184 /* Linked list node that places this dentry in the list of aliases for
185 * its inode (d_inode) */
186 struct list_head d_alias;
188 /* The parent of this directory entry. */
189 struct wim_dentry *parent;
191 /* 'length' and 'subdir_offset' are only used while reading and writing
192 * this dentry; see the corresponding field in
193 * `struct wim_dentry_on_disk' for explanation. */
197 /* These correspond to the two unused fields in the on-disk WIM dentry;
198 * we read them into memory so we can write them unchanged. These
199 * fields are set to 0 on new dentries. */
203 /* Pointer to the UTF-16LE short filename (malloc()ed buffer), or NULL
204 * if this dentry has no short name. */
205 utf16lechar *short_name;
207 /* Pointer to the UTF-16LE filename (malloc()ed buffer), or NULL if this
208 * dentry has no filename. */
209 utf16lechar *file_name;
211 /* Full path to this dentry in the WIM, in platform-dependent tchars
212 * that can be printed without conversion. By default this field will
213 * be NULL and will only be calculated on-demand by the
214 * calculate_dentry_full_path() or dentry_full_path() functions. */
217 /* (Extraction only) Actual name to extract this dentry as, along with
218 * its length in tchars excluding the NULL terminator. This usually
219 * will be the same as file_name, with the character encoding converted
220 * if needed. But if file_name contains characters not accepted on the
221 * current platform, then this may be set slightly differently from
222 * file_name. This will be either NULL or a malloc()ed buffer that may
223 * alias file_name. */
224 tchar *extraction_name;
225 size_t extraction_name_nchars;
227 /* (Extraction only) List head for building a list of dentries that
228 * contain a certain stream. */
229 struct list_head extraction_stream_list;
232 #define rbnode_dentry(node) container_of(node, struct wim_dentry, rb_node)
237 * As mentioned in the comment above `struct wim_dentry', in the WIM file that
238 * is no on-disk analogue of a real inode, as most of these fields are
239 * duplicated in the dentries. Instead, a `struct wim_inode' is something we
240 * create ourselves to simplify the handling of hard links.
243 /* If i_resolved == 0:
244 * SHA1 message digest of the contents of the unnamed-data stream
247 * If i_resolved == 1:
248 * Pointer to the lookup table entry for the unnamed data stream
249 * of this inode, or NULL.
251 * i_hash corresponds to the 'unnamed_stream_hash' field of the `struct
252 * wim_dentry_on_disk' and the additional caveats documented about that
253 * field apply here (for example, the quirks regarding all-zero hashes).
256 u8 i_hash[SHA1_HASH_SIZE];
257 struct wim_lookup_table_entry *i_lte;
260 /* Corresponds to the 'attributes' field of `struct wim_dentry_on_disk';
261 * bitwise OR of the FILE_ATTRIBUTE_* flags that give the attributes of
265 /* Root of a red-black tree storing the child dentries of this inode, if
266 * any. Keyed by wim_dentry->file_name, case sensitively. */
267 struct rb_root i_children;
270 /* Root of a red-black tree storing the children of this inode, if any.
271 * Keyed by wim_dentry->file_name, case insensitively. */
272 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 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 /* For NTFS-3g extraction: Set after the DOS name for this inode has
317 u8 i_dos_name_extracted : 1;
319 /* Pointer to a malloc()ed array of i_num_ads alternate data stream
320 * entries for this inode. */
321 struct wim_ads_entry *i_ads_entries;
323 /* Creation time, last access time, and last write time for this inode, in
324 * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601. They
325 * should correspond to the times gotten by calling GetFileTime() on
328 u64 i_last_access_time;
329 u64 i_last_write_time;
331 /* Corresponds to 'security_id' in `struct wim_dentry_on_disk': The
332 * index of this inode's security descriptor in the WIM image's table of
333 * security descriptors, or -1. Note: in verify_inode(), called
334 * whenever a WIM image is loaded, out-of-bounds indices are set to -1,
335 * so the extraction code does not need to do bounds checks. */
336 int32_t i_security_id;
338 /* Identity of a reparse point. See
339 * http://msdn.microsoft.com/en-us/library/windows/desktop/aa365503(v=vs.85).aspx
340 * for what a reparse point is. */
343 /* Unused/unknown fields that we just read into memory so we can
344 * re-write them unchanged. */
348 /* Corresponds to not_rpfixed in `struct wim_dentry_on_disk': Set to 0
349 * if reparse point fixups have been done. Otherwise set to 1. Note:
350 * this actually may reflect the SYMBOLIC_LINK_RELATIVE flag.
354 /* Inode number; corresponds to hard_link_group_id in the `struct
355 * wim_dentry_on_disk'. */
359 /* Device number, used only during image capture, so we can
360 * identify hard linked files by the combination of inode number
361 * and device number (rather than just inode number, which could
362 * be ambigious if the captured tree spans a mountpoint). Set
366 /* Used only during image extraction: pointer to the first path
367 * (malloc()ed buffer) at which this inode has been extracted.
368 * Freed and set to NULL after the extraction is done (either
369 * success or failure). */
370 tchar *i_extracted_file;
373 /* Next alternate data stream ID to be assigned */
374 u32 i_next_stream_id;
377 /* wimfs file descriptors table for the inode */
378 u16 i_num_opened_fds;
379 u16 i_num_allocated_fds;
380 struct wimfs_fd **i_fds;
384 #define inode_for_each_dentry(dentry, inode) \
385 list_for_each_entry((dentry), &(inode)->i_dentry, d_alias)
387 #define inode_add_dentry(dentry, inode) \
388 list_add_tail(&(dentry)->d_alias, &(inode)->i_dentry)
390 #define inode_first_dentry(inode) \
391 container_of(inode->i_dentry.next, struct wim_dentry, d_alias)
393 #define inode_first_full_path(inode) \
394 dentry_full_path(inode_first_dentry(inode))
397 dentry_is_first_in_inode(const struct wim_dentry *dentry)
399 return inode_first_dentry(dentry->d_inode) == dentry;
403 dentry_correct_total_length(const struct wim_dentry *dentry);
406 for_dentry_in_tree(struct wim_dentry *root,
407 int (*visitor)(struct wim_dentry*, void*),
411 for_dentry_in_rbtree(struct rb_node *node,
412 int (*visitor)(struct wim_dentry *, void *),
416 for_dentry_child(const struct wim_dentry *dentry,
417 int (*visitor)(struct wim_dentry *, void *),
420 return for_dentry_in_rbtree(dentry->d_inode->i_children.rb_node,
426 for_dentry_in_tree_depth(struct wim_dentry *root,
427 int (*visitor)(struct wim_dentry*, void*),
431 calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p);
434 set_dentry_name(struct wim_dentry *dentry, const tchar *new_name);
436 extern struct wim_dentry *
437 get_dentry(struct WIMStruct *w, const tchar *path);
439 extern struct wim_inode *
440 wim_pathname_to_inode(struct WIMStruct *w, const tchar *path);
442 extern struct wim_dentry *
443 get_dentry_child_with_name(const struct wim_dentry *dentry,
446 extern struct wim_dentry *
447 get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry,
448 const utf16lechar *name,
451 extern struct wim_dentry *
452 get_parent_dentry(struct WIMStruct *w, const tchar *path);
455 print_dentry(struct wim_dentry *dentry, void *lookup_table);
458 print_dentry_full_path(struct wim_dentry *entry, void *ignore);
461 calculate_dentry_full_path(struct wim_dentry *dentry);
464 calculate_dentry_tree_full_paths(struct wim_dentry *root);
467 dentry_full_path(struct wim_dentry *dentry);
469 extern struct wim_inode *
470 new_timeless_inode(void) _malloc_attribute;
473 new_dentry(const tchar *name, struct wim_dentry **dentry_ret);
476 new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret);
479 new_dentry_with_timeless_inode(const tchar *name, struct wim_dentry **dentry_ret);
482 new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret);
485 free_inode(struct wim_inode *inode);
488 free_dentry(struct wim_dentry *dentry);
491 put_dentry(struct wim_dentry *dentry);
494 free_dentry_tree(struct wim_dentry *root,
495 struct wim_lookup_table *lookup_table);
498 unlink_dentry(struct wim_dentry *dentry);
500 extern struct wim_dentry *
501 dentry_add_child(struct wim_dentry * restrict parent,
502 struct wim_dentry * restrict child);
504 extern struct wim_ads_entry *
505 inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name,
508 extern struct wim_ads_entry *
509 inode_add_ads_utf16le(struct wim_inode *inode,
510 const utf16lechar *stream_name,
511 size_t stream_name_nbytes);
513 extern struct wim_ads_entry *
514 inode_add_ads(struct wim_inode *dentry, const tchar *stream_name);
517 inode_add_ads_with_data(struct wim_inode *inode, const tchar *name,
518 const void *value, size_t size,
519 struct wim_lookup_table *lookup_table);
522 inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len,
523 struct wim_lookup_table *lookup_table);
526 inode_remove_ads(struct wim_inode *inode, u16 idx,
527 struct wim_lookup_table *lookup_table);
530 #define WIMLIB_UNIX_DATA_TAG "$$__wimlib_UNIX_data"
531 #define WIMLIB_UNIX_DATA_TAG_NBYTES (sizeof(WIMLIB_UNIX_DATA_TAG) - 1)
533 #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"
534 #define WIMLIB_UNIX_DATA_TAG_UTF16LE_NBYTES (sizeof(WIMLIB_UNIX_DATA_TAG_UTF16LE) - 1)
536 /* Format for special alternate data stream entries to store UNIX data for files
537 * and directories (see: WIMLIB_ADD_FLAG_UNIX_DATA) */
538 struct wimlib_unix_data {
539 u16 version; /* Must be 0 */
547 #define NO_UNIX_DATA (-1)
548 #define BAD_UNIX_DATA (-2)
550 inode_get_unix_data(const struct wim_inode *inode,
551 struct wimlib_unix_data *unix_data,
552 u16 *stream_idx_ret);
554 #define UNIX_DATA_UID 0x1
555 #define UNIX_DATA_GID 0x2
556 #define UNIX_DATA_MODE 0x4
557 #define UNIX_DATA_ALL (UNIX_DATA_UID | UNIX_DATA_GID | UNIX_DATA_MODE)
558 #define UNIX_DATA_CREATE 0x8
560 inode_set_unix_data(struct wim_inode *inode, uid_t uid, gid_t gid, mode_t mode,
561 struct wim_lookup_table *lookup_table, int which);
565 read_dentry(const u8 * restrict metadata_resource,
566 u64 metadata_resource_len, u64 offset,
567 struct wim_dentry * restrict dentry);
570 read_dentry_tree(const u8 * restrict metadata_resource,
571 u64 metadata_resource_len,
572 struct wim_dentry * restrict dentry);
575 write_dentry_tree(const struct wim_dentry * restrict tree,
579 dentry_is_root(const struct wim_dentry *dentry)
581 return dentry->parent == dentry;
585 inode_is_directory(const struct wim_inode *inode)
587 return (inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
588 FILE_ATTRIBUTE_REPARSE_POINT))
589 == FILE_ATTRIBUTE_DIRECTORY;
593 dentry_is_directory(const struct wim_dentry *dentry)
595 return inode_is_directory(dentry->d_inode);
598 /* For our purposes, we consider "real" symlinks and "junction points" to both
601 inode_is_symlink(const struct wim_inode *inode)
603 return (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
604 && (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
605 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT);
609 inode_has_children(const struct wim_inode *inode)
611 return inode->i_children.rb_node != NULL;
615 dentry_has_children(const struct wim_dentry *dentry)
617 return inode_has_children(dentry->d_inode);
621 dentry_has_short_name(const struct wim_dentry *dentry)
623 return dentry->short_name_nbytes != 0;
627 dentry_has_long_name(const struct wim_dentry *dentry)
629 return dentry->file_name_nbytes != 0;
633 inode_ref_streams(struct wim_inode *inode);
636 dentry_tree_fix_inodes(struct wim_dentry *root, struct list_head *inode_list);
639 verify_inode(struct wim_inode *inode, const struct wim_security_data *sd);
641 #endif /* _WIMLIB_DENTRY_H */