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 */
17 struct wim_lookup_table;
18 struct wim_lookup_table_entry;
21 struct wim_security_data;
23 /* Size of the struct wim_dentry up to and including the file_name_len. */
24 #define WIM_DENTRY_DISK_SIZE 102
27 * Reparse tags documented at
28 * http://msdn.microsoft.com/en-us/library/dd541667(v=prot.10).aspx
30 #define WIM_IO_REPARSE_TAG_RESERVED_ZERO 0x00000000
31 #define WIM_IO_REPARSE_TAG_RESERVED_ONE 0x00000001
32 #define WIM_IO_REPARSE_TAG_MOUNT_POINT 0xA0000003
33 #define WIM_IO_REPARSE_TAG_HSM 0xC0000004
34 #define WIM_IO_REPARSE_TAG_HSM2 0x80000006
35 #define WIM_IO_REPARSE_TAG_DRIVER_EXTENDER 0x80000005
36 #define WIM_IO_REPARSE_TAG_SIS 0x80000007
37 #define WIM_IO_REPARSE_TAG_DFS 0x8000000A
38 #define WIM_IO_REPARSE_TAG_DFSR 0x80000012
39 #define WIM_IO_REPARSE_TAG_FILTER_MANAGER 0x8000000B
40 #define WIM_IO_REPARSE_TAG_SYMLINK 0xA000000C
42 #define FILE_ATTRIBUTE_READONLY 0x00000001
43 #define FILE_ATTRIBUTE_HIDDEN 0x00000002
44 #define FILE_ATTRIBUTE_SYSTEM 0x00000004
45 #define FILE_ATTRIBUTE_DIRECTORY 0x00000010
46 #define FILE_ATTRIBUTE_ARCHIVE 0x00000020
47 #define FILE_ATTRIBUTE_DEVICE 0x00000040
48 #define FILE_ATTRIBUTE_NORMAL 0x00000080
49 #define FILE_ATTRIBUTE_TEMPORARY 0x00000100
50 #define FILE_ATTRIBUTE_SPARSE_FILE 0x00000200
51 #define FILE_ATTRIBUTE_REPARSE_POINT 0x00000400
52 #define FILE_ATTRIBUTE_COMPRESSED 0x00000800
53 #define FILE_ATTRIBUTE_OFFLINE 0x00001000
54 #define FILE_ATTRIBUTE_NOT_CONTENT_INDEXED 0x00002000
55 #define FILE_ATTRIBUTE_ENCRYPTED 0x00004000
56 #define FILE_ATTRIBUTE_VIRTUAL 0x00010000
59 /* Alternate data stream entry.
61 * We read this from disk in the read_ads_entries() function; see that function
62 * for more explanation. */
63 struct wim_ads_entry {
65 /* SHA-1 message digest of stream contents */
66 u8 hash[SHA1_HASH_SIZE];
68 /* The corresponding lookup table entry (only for resolved
70 struct wim_lookup_table_entry *lte;
73 /* Length of UTF16-encoded stream name, in bytes, not including the
74 * terminating null character; or 0 if the stream is unnamed. */
75 u16 stream_name_nbytes;
77 /* Number to identify an alternate data stream even after it's possibly
78 * been moved or renamed. */
81 /* Stream name (UTF-16LE), null-terminated, or NULL if the stream is
83 utf16lechar *stream_name;
85 /* Reserved field. We read it into memory so we can write it out
92 ads_entries_have_same_name(const struct wim_ads_entry *entry_1,
93 const struct wim_ads_entry *entry_2)
95 return entry_1->stream_name_nbytes == entry_2->stream_name_nbytes &&
96 memcmp(entry_1->stream_name, entry_2->stream_name,
97 entry_1->stream_name_nbytes) == 0;
101 * In-memory structure for a WIM directory entry (dentry). There is a directory
102 * tree for each image in the WIM.
104 * Note that this is a directory entry and not an inode. Since NTFS allows hard
105 * links, it's possible for a NTFS inode to correspond to multiple WIM dentries.
106 * The hard link group ID field of the on-disk WIM dentry tells us the number of
107 * the NTFS inode that the dentry corresponds to (and this gets placed in
110 * Unfortunately, WIM files do not have an analogue to an inode; instead certain
111 * information, such as file attributes, the security descriptor, and file
112 * streams is replicated in each hard-linked dentry, even though this
113 * information really is associated with an inode. In-memory, we fix up this
114 * flaw by allocating a `struct wim_inode' for each dentry that contains some of
115 * this duplicated information, then combining the inodes for each hard link
118 * Confusingly, it's possible for stream information to be missing from a dentry
119 * in a hard link set, in which case the stream information needs to be gotten
120 * from one of the other dentries in the hard link set. In addition, it is
121 * possible for dentries to have inconsistent security IDs, file attributes, or
122 * file streams when they share the same hard link ID (don't even ask. I hope
123 * that Microsoft may have fixed this problem, since I've only noticed it in the
124 * 'install.wim' for Windows 7). For those dentries, we have to use the
125 * conflicting fields to split up the hard link groups. (See
126 * dentry_tree_fix_inodes() in hardlink.c).
129 /* Pointer to the inode for this dentry. This will contain some
130 * information that was factored out of the on-disk WIM dentry as common
131 * to all dentries in a hard link group. */
132 struct wim_inode *d_inode;
134 /* Node for the parent's red-black tree of child dentries, sorted by
135 * case sensitive long name. */
136 struct rb_node rb_node;
139 /* Node for the parent's red-black tree of child dentries, sorted by
140 * case insensitive long name. */
141 struct rb_node rb_node_case_insensitive;
143 /* List of dentries in a directory that have different case sensitive
144 * long names but share the same case insensitive long name */
145 struct list_head case_insensitive_conflict_list;
148 /* Length of UTF-16LE encoded short filename, in bytes, not including
149 * the terminating zero wide-character. */
150 u16 short_name_nbytes;
152 /* Length of UTF-16LE encoded "long" file name, in bytes, not including
153 * the terminating null character. */
154 u16 file_name_nbytes;
156 /* Length of full path name encoded using "tchars", in bytes, not
157 * including the terminating null character. */
158 u32 full_path_nbytes;
160 /* For extraction operations, a subtree of dentries will have this flag
161 * set so we can keep track of which dentries still need to be
162 * extracted. Otherwise this will always be 0. */
163 u8 needs_extraction : 1;
165 /* For extraction operations, this flag will be set when a dentry in the
166 * tree being extracted is not being extracted for some reason (file
167 * type not supported by target filesystem or contains invalid
168 * characters). Otherwise this will always be 0. */
169 u8 not_extracted : 1;
171 /* When capturing from a NTFS volume using NTFS-3g, this flag is set on
172 * dentries that were created from a filename in the WIN32 or WIN32+DOS
173 * namespaces rather than the POSIX namespace. Otherwise this will
175 u8 is_win32_name : 1;
177 /* When verifying the dentry tree after reading it into memory, this
178 * flag will be set on all dentries in a hard link group that have a
179 * nonempty DOS name except one. This is because it is supposed to be
180 * illegal (on NTFS, at least) for a single inode to have multiple DOS
182 u8 dos_name_invalid : 1;
184 /* Temporary list field used to make lists of dentries in a few places.
186 struct list_head tmp_list;
188 /* Linked list node that places this dentry in the list of aliases for
189 * its inode (d_inode) */
190 struct list_head d_alias;
192 /* The parent of this directory entry. */
193 struct wim_dentry *parent;
195 /* 'length' and 'subdir_offset' are only used while reading and writing
196 * this dentry; see the corresponding field in
197 * `struct wim_dentry_on_disk' for explanation. */
201 /* These correspond to the two unused fields in the on-disk WIM dentry;
202 * we read them into memory so we can write them unchanged. These
203 * fields are set to 0 on new dentries. */
207 /* Pointer to the UTF-16LE short filename (malloc()ed buffer), or NULL
208 * if this dentry has no short name. */
209 utf16lechar *short_name;
211 /* Pointer to the UTF-16LE filename (malloc()ed buffer), or NULL if this
212 * dentry has no filename. */
213 utf16lechar *file_name;
215 /* Full path to this dentry in the WIM, in platform-dependent tchars
216 * that can be printed without conversion. By default this field will
217 * be NULL and will only be calculated on-demand by the
218 * calculate_dentry_full_path() or dentry_full_path() functions. */
221 /* (Extraction only) Actual name to extract this dentry as, along with
222 * its length in tchars excluding the NULL terminator. This usually
223 * will be the same as file_name, with the character encoding converted
224 * if needed. But if file_name contains characters not accepted on the
225 * current platform, then this may be set slightly differently from
226 * file_name. This will be either NULL or a malloc()ed buffer that may
227 * alias file_name. */
228 tchar *extraction_name;
229 size_t extraction_name_nchars;
231 /* (Extraction only) List head for building a list of dentries that
232 * contain a certain stream. */
233 struct list_head extraction_stream_list;
236 #define rbnode_dentry(node) container_of(node, struct wim_dentry, rb_node)
241 * As mentioned in the comment above `struct wim_dentry', in the WIM file that
242 * is no on-disk analogue of a real inode, as most of these fields are
243 * duplicated in the dentries. Instead, a `struct wim_inode' is something we
244 * create ourselves to simplify the handling of hard links.
247 /* If i_resolved == 0:
248 * SHA1 message digest of the contents of the unnamed-data stream
251 * If i_resolved == 1:
252 * Pointer to the lookup table entry for the unnamed data stream
253 * of this inode, or NULL.
255 * i_hash corresponds to the 'unnamed_stream_hash' field of the `struct
256 * wim_dentry_on_disk' and the additional caveats documented about that
257 * field apply here (for example, the quirks regarding all-zero hashes).
260 u8 i_hash[SHA1_HASH_SIZE];
261 struct wim_lookup_table_entry *i_lte;
264 /* Corresponds to the 'attributes' field of `struct wim_dentry_on_disk';
265 * bitwise OR of the FILE_ATTRIBUTE_* flags that give the attributes of
269 /* Root of a red-black tree storing the child dentries of this inode, if
270 * any. Keyed by wim_dentry->file_name, case sensitively. */
271 struct rb_root i_children;
274 /* Root of a red-black tree storing the children of this inode, if any.
275 * Keyed by wim_dentry->file_name, case insensitively. */
276 struct rb_root i_children_case_insensitive;
279 /* List of dentries that are aliases for this inode. There will be
280 * i_nlink dentries in this list. */
281 struct list_head i_dentry;
283 /* Field to place this inode into a list. */
285 /* Hash list node- used in hardlink.c when the inodes are placed
286 * into a hash table keyed by inode number and optionally device
287 * number, in order to detect dentries that are aliases for the
289 struct hlist_node i_hlist;
291 /* Normal list node- used to connect all the inodes of a WIM image
292 * into a single linked list referenced from the
293 * `struct wim_image_metadata' for that image. */
294 struct list_head i_list;
297 /* Number of dentries that are aliases for this inode. */
300 /* Number of alternate data streams associated with this inode */
303 /* Flag that indicates whether this inode's streams have been
304 * "resolved". By default, the inode starts as "unresolved", meaning
305 * that the i_hash field, along with the hash field of any associated
306 * wim_ads_entry's, are valid and should be used as keys in the WIM
307 * lookup table to find the associated `struct wim_lookup_table_entry'.
308 * But if the inode has been resolved, then each of these fields is
309 * replaced with a pointer directly to the appropriate `struct
310 * wim_lookup_table_entry', or NULL if the stream is empty. */
313 /* Flag used to mark this inode as visited; this is used when visiting
314 * all the inodes in a dentry tree exactly once. It will be 0 by
315 * default and must be cleared following the tree traversal, even in
319 /* For NTFS-3g extraction: Set after the DOS name for this inode has
321 u8 i_dos_name_extracted : 1;
323 /* Pointer to a malloc()ed array of i_num_ads alternate data stream
324 * entries for this inode. */
325 struct wim_ads_entry *i_ads_entries;
327 /* Creation time, last access time, and last write time for this inode, in
328 * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601. They
329 * should correspond to the times gotten by calling GetFileTime() on
332 u64 i_last_access_time;
333 u64 i_last_write_time;
335 /* Corresponds to 'security_id' in `struct wim_dentry_on_disk': The
336 * index of this inode's security descriptor in the WIM image's table of
337 * security descriptors, or -1. Note: in verify_inode(), called
338 * whenever a WIM image is loaded, out-of-bounds indices are set to -1,
339 * so the extraction code does not need to do bounds checks. */
340 int32_t i_security_id;
342 /* Identity of a reparse point. See
343 * http://msdn.microsoft.com/en-us/library/windows/desktop/aa365503(v=vs.85).aspx
344 * for what a reparse point is. */
347 /* Unused/unknown fields that we just read into memory so we can
348 * re-write them unchanged. */
352 /* Corresponds to not_rpfixed in `struct wim_dentry_on_disk': Set to 0
353 * if reparse point fixups have been done. Otherwise set to 1. Note:
354 * this actually may reflect the SYMBOLIC_LINK_RELATIVE flag.
358 /* Inode number; corresponds to hard_link_group_id in the `struct
359 * wim_dentry_on_disk'. */
363 /* Device number, used only during image capture, so we can
364 * identify hard linked files by the combination of inode number
365 * and device number (rather than just inode number, which could
366 * be ambigious if the captured tree spans a mountpoint). Set
370 /* Used only during image extraction: pointer to the first path
371 * (malloc()ed buffer) at which this inode has been extracted.
372 * Freed and set to NULL after the extraction is done (either
373 * success or failure). */
374 tchar *i_extracted_file;
377 /* Next alternate data stream ID to be assigned */
378 u32 i_next_stream_id;
381 /* wimfs file descriptors table for the inode */
382 u16 i_num_opened_fds;
383 u16 i_num_allocated_fds;
384 struct wimfs_fd **i_fds;
385 /* This mutex protects the inode's file descriptors table during
386 * read-only mounts. Read-write mounts are still restricted to 1
388 pthread_mutex_t i_mutex;
392 #define inode_for_each_dentry(dentry, inode) \
393 list_for_each_entry((dentry), &(inode)->i_dentry, d_alias)
395 #define inode_add_dentry(dentry, inode) \
396 list_add_tail(&(dentry)->d_alias, &(inode)->i_dentry)
398 #define inode_first_dentry(inode) \
399 container_of(inode->i_dentry.next, struct wim_dentry, d_alias)
401 #define inode_first_full_path(inode) \
402 dentry_full_path(inode_first_dentry(inode))
405 dentry_is_first_in_inode(const struct wim_dentry *dentry)
407 return inode_first_dentry(dentry->d_inode) == dentry;
411 dentry_correct_total_length(const struct wim_dentry *dentry);
414 for_dentry_in_tree(struct wim_dentry *root,
415 int (*visitor)(struct wim_dentry*, void*),
419 for_dentry_in_rbtree(struct rb_node *node,
420 int (*visitor)(struct wim_dentry *, void *),
424 for_dentry_child(const struct wim_dentry *dentry,
425 int (*visitor)(struct wim_dentry *, void *),
428 return for_dentry_in_rbtree(dentry->d_inode->i_children.rb_node,
434 for_dentry_in_tree_depth(struct wim_dentry *root,
435 int (*visitor)(struct wim_dentry*, void*),
439 calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p);
442 set_dentry_name(struct wim_dentry *dentry, const tchar *new_name);
444 extern struct wim_dentry *
445 get_dentry(struct WIMStruct *w, const tchar *path);
447 extern struct wim_inode *
448 wim_pathname_to_inode(struct WIMStruct *w, const tchar *path);
450 extern struct wim_dentry *
451 get_dentry_child_with_name(const struct wim_dentry *dentry,
454 extern struct wim_dentry *
455 get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry,
456 const utf16lechar *name,
459 extern struct wim_dentry *
460 get_parent_dentry(struct WIMStruct *w, const tchar *path);
463 print_dentry(struct wim_dentry *dentry, void *lookup_table);
466 print_dentry_full_path(struct wim_dentry *entry, void *ignore);
469 calculate_dentry_full_path(struct wim_dentry *dentry);
472 calculate_dentry_tree_full_paths(struct wim_dentry *root);
475 dentry_full_path(struct wim_dentry *dentry);
477 extern struct wim_inode *
478 new_timeless_inode(void) _malloc_attribute;
481 new_dentry(const tchar *name, struct wim_dentry **dentry_ret);
484 new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret);
487 new_dentry_with_timeless_inode(const tchar *name, struct wim_dentry **dentry_ret);
490 new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret);
493 free_inode(struct wim_inode *inode);
496 free_dentry(struct wim_dentry *dentry);
499 put_dentry(struct wim_dentry *dentry);
502 free_dentry_tree(struct wim_dentry *root,
503 struct wim_lookup_table *lookup_table);
506 unlink_dentry(struct wim_dentry *dentry);
508 extern struct wim_dentry *
509 dentry_add_child(struct wim_dentry * restrict parent,
510 struct wim_dentry * restrict child);
512 extern struct wim_ads_entry *
513 inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name,
516 extern struct wim_ads_entry *
517 inode_add_ads_utf16le(struct wim_inode *inode,
518 const utf16lechar *stream_name,
519 size_t stream_name_nbytes);
521 extern struct wim_ads_entry *
522 inode_add_ads(struct wim_inode *dentry, const tchar *stream_name);
525 inode_add_ads_with_data(struct wim_inode *inode, const tchar *name,
526 const void *value, size_t size,
527 struct wim_lookup_table *lookup_table);
530 inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len,
531 struct wim_lookup_table *lookup_table);
534 inode_remove_ads(struct wim_inode *inode, u16 idx,
535 struct wim_lookup_table *lookup_table);
538 #define WIMLIB_UNIX_DATA_TAG "$$__wimlib_UNIX_data"
539 #define WIMLIB_UNIX_DATA_TAG_NBYTES (sizeof(WIMLIB_UNIX_DATA_TAG) - 1)
541 #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"
542 #define WIMLIB_UNIX_DATA_TAG_UTF16LE_NBYTES (sizeof(WIMLIB_UNIX_DATA_TAG_UTF16LE) - 1)
544 /* Format for special alternate data stream entries to store UNIX data for files
545 * and directories (see: WIMLIB_ADD_FLAG_UNIX_DATA) */
546 struct wimlib_unix_data {
547 u16 version; /* Must be 0 */
555 #define NO_UNIX_DATA (-1)
556 #define BAD_UNIX_DATA (-2)
558 inode_get_unix_data(const struct wim_inode *inode,
559 struct wimlib_unix_data *unix_data,
560 u16 *stream_idx_ret);
562 #define UNIX_DATA_UID 0x1
563 #define UNIX_DATA_GID 0x2
564 #define UNIX_DATA_MODE 0x4
565 #define UNIX_DATA_ALL (UNIX_DATA_UID | UNIX_DATA_GID | UNIX_DATA_MODE)
566 #define UNIX_DATA_CREATE 0x8
568 inode_set_unix_data(struct wim_inode *inode, uid_t uid, gid_t gid, mode_t mode,
569 struct wim_lookup_table *lookup_table, int which);
573 read_dentry(const u8 *metadata_resource, u64 metadata_resource_len,
574 u64 offset, struct wim_dentry *dentry);
577 read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
578 struct wim_dentry *dentry);
581 write_dentry_tree(const struct wim_dentry *tree, u8 *p);
584 dentry_is_root(const struct wim_dentry *dentry)
586 return dentry->parent == dentry;
590 inode_is_directory(const struct wim_inode *inode)
592 return (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
593 && !(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
597 dentry_is_directory(const struct wim_dentry *dentry)
599 return inode_is_directory(dentry->d_inode);
602 /* For our purposes, we consider "real" symlinks and "junction points" to both
605 inode_is_symlink(const struct wim_inode *inode)
607 return (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
608 && (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
609 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT);
613 inode_is_regular_file(const struct wim_inode *inode)
615 return !inode_is_directory(inode) && !inode_is_symlink(inode);
619 dentry_is_regular_file(const struct wim_dentry *dentry)
621 return inode_is_regular_file(dentry->d_inode);
625 inode_has_children(const struct wim_inode *inode)
627 return inode->i_children.rb_node != NULL;
631 dentry_has_children(const struct wim_dentry *dentry)
633 return inode_has_children(dentry->d_inode);
637 dentry_has_short_name(const struct wim_dentry *dentry)
639 return dentry->short_name_nbytes != 0;
643 dentry_has_long_name(const struct wim_dentry *dentry)
645 return dentry->file_name_nbytes != 0;
649 inode_ref_streams(struct wim_inode *inode);
652 dentry_tree_fix_inodes(struct wim_dentry *root, struct list_head *inode_list);
655 verify_inode(struct wim_inode *inode, const struct wim_security_data *sd);