1 #ifndef _WIMLIB_INODE_H
2 #define _WIMLIB_INODE_H
3
4 #include "wimlib/assert.h"
5 #include "wimlib/list.h"
6 #include "wimlib/sha1.h"
7 #include "wimlib/types.h"
8
9 struct avl_tree_node;
10 struct blob_descriptor;
11 struct blob_table;
12 struct wim_dentry;
13 struct wim_inode_extra;
14 struct wim_security_data;
15 struct wimfs_fd;
16
17 /* Valid values for the 'stream_type' field of a 'struct wim_inode_stream' */
18 enum wim_inode_stream_type {
19
20 /* Data stream, may be unnamed (usual case) or named */
21 STREAM_TYPE_DATA,
22
23 /* Reparse point stream. This is the same as the data of the on-disk
24 * reparse point attribute, except that the first 8 bytes of the on-disk
25 * attribute are omitted. The omitted bytes contain the reparse tag
26 * (which is instead stored in the on-disk WIM dentry), the reparse data
27 * size (which is redundant with the stream size), and a reserved field
28 * that is always zero. */
29 STREAM_TYPE_REPARSE_POINT,
30
31 /* Encrypted data in the "EFSRPC raw data format" specified by [MS-EFSR]
32 * section 2.2.3. This contains metadata for the Encrypting File System
33 * as well as the encrypted data of all the file's data streams. */
34 STREAM_TYPE_EFSRPC_RAW_DATA,
35
36 /* Stream type could not be determined */
37 STREAM_TYPE_UNKNOWN,
38 };
39
40 extern const utf16lechar NO_STREAM_NAME[1];
41
42 /*
43 * 'struct wim_inode_stream' describes a "stream", which associates a blob of
44 * data with an inode. Each stream has a type and optionally a name.
45 *
46 * The most frequently seen kind of stream is the "unnamed data stream"
47 * (stream_type == STREAM_TYPE_DATA && stream_name == NO_STREAM_NAME), which is
48 * the "default file contents". Many inodes just have an unnamed data stream
49 * and no other streams. However, files on NTFS filesystems may have
50 * additional, "named" data streams, and this is supported by the WIM format.
51 *
52 * A "reparse point" is an inode with reparse data set. The reparse data is
53 * stored in a stream of type STREAM_TYPE_REPARSE_POINT. There should be only
54 * one such stream, and it should be unnamed. However, it is possible for an
55 * inode to have both a reparse point stream and an unnamed data stream, and
56 * even named data streams as well.
57 */
58 struct wim_inode_stream {
59
60 /* The name of the stream or NO_STREAM_NAME. */
61 utf16lechar *stream_name;
62
63 /*
64 * If 'stream_resolved' = 0, then 'stream_hash' is the SHA-1 message
65 * digest of the uncompressed data of this stream, or all zeroes if this
66 * stream is empty.
67 *
68 * If 'stream_resolved' = 1, then 'stream_blob' is a pointer directly to
69 * a descriptor for this stream's blob, or NULL if this stream is empty.
70 */
71 union {
72 u8 _stream_hash[SHA1_HASH_SIZE];
73 struct blob_descriptor *_stream_blob;
74 } _packed_attribute; /* union is SHA1_HASH_SIZE bytes */
75
76 /* 'stream_resolved' determines whether 'stream_hash' or 'stream_blob'
77 * is valid as described above. */
78 u32 stream_resolved : 1;
79
80 /* A unique identifier for this stream within the context of its inode.
81 * This stays constant even if the streams array is reallocated. */
82 u32 stream_id : 28;
83
84 /* The type of this stream as one of the STREAM_TYPE_* values */
85 u32 stream_type : 3;
86 };
87
88 /*
89 * WIM inode - a "file" in an image which may be accessible via multiple paths
90 *
91 * Note: in WIM files there is no true on-disk analogue of an inode; there are
92 * only directory entries, and some fields are duplicated among all links to a
93 * file. However, wimlib uses inode structures internally to simplify handling
94 * of hard links.
95 */
96 struct wim_inode {
97
98 /*
99 * The collection of streams for this inode. 'i_streams' points to
100 * either 'i_embedded_streams' or an allocated array.
101 */
102 struct wim_inode_stream *i_streams;
103 struct wim_inode_stream i_embedded_streams[1];
104 unsigned i_num_streams;
105
106 /* Windows file attribute flags (FILE_ATTRIBUTE_*). */
107 u32 i_attributes;
108
109 /* Root of a balanced binary search tree storing the child directory
110 * entries of this inode, if any, indexed by filename. If this inode is
111 * not a directory or if it has no children then this will be an empty
112 * tree (NULL). */
113 struct avl_tree_node *i_children;
114
115 /* List of dentries that are aliases for this inode. There will be
116 * i_nlink dentries in this list. */
117 struct hlist_head i_alias_list;
118
119 /* Field to place this inode into a list. While reading a WIM image or
120 * adding files to a WIM image this is owned by the inode table;
121 * otherwise this links the inodes for the WIM image. */
122 struct hlist_node i_hlist_node;
123
124 /* Number of dentries that are aliases for this inode. */
125 u32 i_nlink : 30;
126
127 /* Flag used by some code to mark this inode as visited. It will be 0
128 * by default, and it always must be cleared after use. */
129 u32 i_visited : 1;
130
131 /* Cached value */
132 u32 i_can_externally_back : 1;
133
134 /* If not NULL, a pointer to the extra data that was read from the
135 * dentry. This should be a series of tagged items, each of which
136 * represents a bit of extra metadata, such as the file's object ID.
137 * See tagged_items.c for more information. */
138 struct wim_inode_extra *i_extra;
139
140 /* Creation time, last access time, and last write time for this inode,
141 * in 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601.
142 * They should correspond to the times gotten by calling GetFileTime()
143 * on Windows. */
144 u64 i_creation_time;
145 u64 i_last_access_time;
146 u64 i_last_write_time;
147
148 /* Corresponds to 'security_id' in `struct wim_dentry_on_disk': The
149 * index of this inode's security descriptor in the WIM image's table of
150 * security descriptors, or -1 if this inode does not have a security
151 * descriptor. */
152 s32 i_security_id;
153
154 /* Unknown field that we only read into memory so we can re-write it
155 * unchanged. Probably it's actually just padding... */
156 u32 i_unknown_0x54;
157
158 /* The following fields correspond to 'reparse_tag', 'rp_reserved', and
159 * 'rp_flags' in `struct wim_dentry_on_disk'. They are only meaningful
160 * for reparse point files. */
161 u32 i_reparse_tag;
162 u16 i_rp_reserved;
163 u16 i_rp_flags;
164
165 /* Inode number; corresponds to hard_link_group_id in the `struct
166 * wim_dentry_on_disk'. */
167 u64 i_ino;
168
169 union {
170 /* Device number, used only during image capture, so we can
171 * identify hard linked files by the combination of inode number
172 * and device number (rather than just inode number, which could
173 * be ambiguous if the captured tree spans a mountpoint). Set
174 * to 0 otherwise. */
175 u64 i_devno;
176
177 /* Fields used only during extraction */
178 struct {
179 /* A singly linked list of aliases of this inode that
180 * are being extracted in the current extraction
181 * operation. This list may be shorter than the inode's
182 * full alias list. This list will be constructed
183 * regardless of whether the extraction backend supports
184 * hard links or not. */
185 struct wim_dentry *i_first_extraction_alias;
186
187 #ifdef WITH_NTFS_3G
188 /* In NTFS-3G extraction mode, this is set to the Master
189 * File Table (MFT) number of the NTFS file that was
190 * created for this inode. */
191 u64 i_mft_no;
192 #endif
193 };
194
195 /* Used during WIM writing with
196 * WIMLIB_WRITE_FLAG_SEND_DONE_WITH_FILE_MESSAGES: the number
197 * of streams this inode has that have not yet been fully read.
198 * */
199 u32 i_num_remaining_streams;
200
201 #ifdef WITH_FUSE
202 struct {
203 /* Used only during image mount: Table of file
204 * descriptors that have been opened to this inode.
205 * This table is freed when the last file descriptor is
206 * closed. */
207 struct wimfs_fd **i_fds;
208
209 /* Lower bound on the index of the next available entry
210 * in 'i_fds'. */
211 u16 i_next_fd;
212 };
213 #endif
214 };
215
216 #ifdef WITH_FUSE
217 u16 i_num_opened_fds;
218 u16 i_num_allocated_fds;
219 #endif
220
221 /* Next stream ID to be assigned */
222 u32 i_next_stream_id;
223
224 #ifdef ENABLE_TEST_SUPPORT
225 struct wim_inode *i_corresponding;
226 #endif
227 };
228
229 /* Optional extra data for a WIM inode */
230 struct wim_inode_extra {
231 size_t size; /* Size of the extra data in bytes */
232 u8 data[] _aligned_attribute(8); /* The extra data */
233 };
234
235 /*
236 * The available reparse tags are documented at
237 * http://msdn.microsoft.com/en-us/library/dd541667(v=prot.10).aspx.
238 * Here we only define the ones of interest to us.
239 */
240 #define WIM_IO_REPARSE_TAG_MOUNT_POINT 0xA0000003
241 #define WIM_IO_REPARSE_TAG_SYMLINK 0xA000000C
242 #define WIM_IO_REPARSE_TAG_DEDUP 0x80000013
243 #define WIM_IO_REPARSE_TAG_WOF 0x80000017
244
245 /* Flags for the rp_flags field. Currently the only known flag is NOT_FIXED,
246 * which indicates that the target of the absolute symbolic link or junction was
247 * not changed when it was stored. */
248 #define WIM_RP_FLAG_NOT_FIXED 0x0001
249
250 /* Windows file attribute flags */
251 #define FILE_ATTRIBUTE_READONLY 0x00000001
252 #define FILE_ATTRIBUTE_HIDDEN 0x00000002
253 #define FILE_ATTRIBUTE_SYSTEM 0x00000004
254 #define FILE_ATTRIBUTE_DIRECTORY 0x00000010
255 #define FILE_ATTRIBUTE_ARCHIVE 0x00000020
256 #define FILE_ATTRIBUTE_DEVICE 0x00000040
257 #define FILE_ATTRIBUTE_NORMAL 0x00000080
258 #define FILE_ATTRIBUTE_TEMPORARY 0x00000100
259 #define FILE_ATTRIBUTE_SPARSE_FILE 0x00000200
260 #define FILE_ATTRIBUTE_REPARSE_POINT 0x00000400
261 #define FILE_ATTRIBUTE_COMPRESSED 0x00000800
262 #define FILE_ATTRIBUTE_OFFLINE 0x00001000
263 #define FILE_ATTRIBUTE_NOT_CONTENT_INDEXED 0x00002000
264 #define FILE_ATTRIBUTE_ENCRYPTED 0x00004000
265 #define FILE_ATTRIBUTE_VIRTUAL 0x00010000
266
267 extern struct wim_inode *
268 new_inode(struct wim_dentry *dentry, bool set_timestamps);
269
270 /* Iterate through each alias of the specified inode. */
271 #define inode_for_each_dentry(dentry, inode) \
272 hlist_for_each_entry((dentry), &(inode)->i_alias_list, d_alias_node)
273
274 /* Return an alias of the specified inode. */
275 #define inode_any_dentry(inode) \
276 hlist_entry(inode->i_alias_list.first, struct wim_dentry, d_alias_node)
277
278 /* Return the full path of an alias of the specified inode, or NULL if a full
279 * path could not be determined. */
280 #define inode_any_full_path(inode) \
281 dentry_full_path(inode_any_dentry(inode))
282
283 extern void
284 d_associate(struct wim_dentry *dentry, struct wim_inode *inode);
285
286 extern void
287 d_disassociate(struct wim_dentry *dentry);
288
289 #ifdef WITH_FUSE
290 extern void
291 inode_dec_num_opened_fds(struct wim_inode *inode);
292 #endif
293
294 /* Is the inode a directory?
295 * This doesn't count directories with reparse data.
296 * wimlib only allows inodes of this type to have children.
297 */
298 static inline bool
inode_is_directory(const struct wim_inode * inode)299 inode_is_directory(const struct wim_inode *inode)
300 {
301 return (inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
302 FILE_ATTRIBUTE_REPARSE_POINT))
303 == FILE_ATTRIBUTE_DIRECTORY;
304 }
305
306 /* Is the inode a symbolic link?
307 * This returns true iff the inode is a reparse point that is either a "real"
308 * symbolic link or a junction point. */
309 static inline bool
inode_is_symlink(const struct wim_inode * inode)310 inode_is_symlink(const struct wim_inode *inode)
311 {
312 return (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
313 && (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
314 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT);
315 }
316
317 /* Does the inode have children? Currently (based on read_dentry_tree() as well
318 * as the various build-dentry-tree implementations), this can only return true
319 * for inodes for which inode_is_directory() returns true. */
320 static inline bool
inode_has_children(const struct wim_inode * inode)321 inode_has_children(const struct wim_inode *inode)
322 {
323 return inode->i_children != NULL;
324 }
325
326 /* Does the inode have a security descriptor? */
327 static inline bool
inode_has_security_descriptor(const struct wim_inode * inode)328 inode_has_security_descriptor(const struct wim_inode *inode)
329 {
330 return inode->i_security_id >= 0;
331 }
332
333 extern struct wim_inode_stream *
334 inode_get_stream(const struct wim_inode *inode, int stream_type,
335 const utf16lechar *stream_name);
336
337 extern struct wim_inode_stream *
338 inode_get_unnamed_stream(const struct wim_inode *inode, int stream_type);
339
340 static inline struct wim_inode_stream *
inode_get_unnamed_data_stream(const struct wim_inode * inode)341 inode_get_unnamed_data_stream(const struct wim_inode *inode)
342 {
343 return inode_get_unnamed_stream(inode, STREAM_TYPE_DATA);
344 }
345
346 extern struct wim_inode_stream *
347 inode_add_stream(struct wim_inode *inode, int stream_type,
348 const utf16lechar *stream_name, struct blob_descriptor *blob);
349
350 extern void
351 inode_replace_stream_blob(struct wim_inode *inode,
352 struct wim_inode_stream *strm,
353 struct blob_descriptor *new_blob,
354 struct blob_table *blob_table);
355
356 extern bool
357 inode_replace_stream_data(struct wim_inode *inode,
358 struct wim_inode_stream *strm,
359 const void *data, size_t size,
360 struct blob_table *blob_table);
361
362 extern bool
363 inode_add_stream_with_data(struct wim_inode *inode,
364 int stream_type, const utf16lechar *stream_name,
365 const void *data, size_t size,
366 struct blob_table *blob_table);
367
368 extern void
369 inode_remove_stream(struct wim_inode *inode, struct wim_inode_stream *strm,
370 struct blob_table *blob_table);
371
372 static inline struct blob_descriptor *
stream_blob_resolved(const struct wim_inode_stream * strm)373 stream_blob_resolved(const struct wim_inode_stream *strm)
374 {
375 wimlib_assert(strm->stream_resolved);
376 return strm->_stream_blob;
377 }
378
379 static inline bool
stream_is_named(const struct wim_inode_stream * strm)380 stream_is_named(const struct wim_inode_stream *strm)
381 {
382 return strm->stream_name != NO_STREAM_NAME;
383 }
384
385 static inline bool
stream_is_unnamed_data_stream(const struct wim_inode_stream * strm)386 stream_is_unnamed_data_stream(const struct wim_inode_stream *strm)
387 {
388 return strm->stream_type == STREAM_TYPE_DATA && !stream_is_named(strm);
389 }
390
391 static inline bool
stream_is_named_data_stream(const struct wim_inode_stream * strm)392 stream_is_named_data_stream(const struct wim_inode_stream *strm)
393 {
394 return strm->stream_type == STREAM_TYPE_DATA && stream_is_named(strm);
395 }
396
397 extern bool
398 inode_has_named_data_stream(const struct wim_inode *inode);
399
400 extern int
401 inode_resolve_streams(struct wim_inode *inode, struct blob_table *table,
402 bool force);
403
404 extern int
405 blob_not_found_error(const struct wim_inode *inode, const u8 *hash);
406
407 extern struct blob_descriptor *
408 stream_blob(const struct wim_inode_stream *strm, const struct blob_table *table);
409
410 extern struct blob_descriptor *
411 inode_get_blob_for_unnamed_data_stream(const struct wim_inode *inode,
412 const struct blob_table *blob_table);
413
414 extern struct blob_descriptor *
415 inode_get_blob_for_unnamed_data_stream_resolved(const struct wim_inode *inode);
416
417 extern const u8 *
418 stream_hash(const struct wim_inode_stream *strm);
419
420 extern const u8 *
421 inode_get_hash_of_unnamed_data_stream(const struct wim_inode *inode);
422
423 extern void
424 inode_ref_blobs(struct wim_inode *inode);
425
426 extern void
427 inode_unref_blobs(struct wim_inode *inode, struct blob_table *blob_table);
428
429 /* inode_fixup.c */
430 extern int
431 dentry_tree_fix_inodes(struct wim_dentry *root, struct hlist_head *inode_list);
432
433 #endif /* _WIMLIB_INODE_H */
434