1 #ifndef _WIMLIB_LOOKUP_TABLE_H
2 #define _WIMLIB_LOOKUP_TABLE_H
3 #include "wimlib_internal.h"
8 /* Size of each lookup table entry in the WIM file. */
9 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
11 #define LOOKUP_FLAG_ADS_OK 0x00000001
12 #define LOOKUP_FLAG_DIRECTORY_OK 0x00000002
15 //#define LOOKUP_FLAG_FOLLOW_SYMLINKS 0x00000004
18 /* A lookup table that is used to translate the hash codes of dentries into the
19 * offsets and sizes of uncompressed or compressed file resources. It is
20 * implemented as a hash table. */
22 struct hlist_head *array;
30 struct ntfs_location {
32 char *stream_name_utf16;
33 u16 stream_name_utf16_num_chars;
34 struct _ntfs_volume **ntfs_vol_p;
35 bool is_reparse_point;
40 * An entry in the lookup table in the WIM file.
42 * It is used to find data streams for files in the WIM.
44 * Metadata resources and reparse point data buffers will also have lookup table
45 * entries associated with the data.
47 * The lookup_table_entry for a given dentry or alternate stream entry in the
48 * WIM is found using the SHA1 message digest field.
50 struct lookup_table_entry {
52 /* List of lookup table entries in this hash bucket */
54 struct hlist_node hash_list;
55 struct list_head list;
58 /* Location and size of the stream in the WIM, whether it is compressed
59 * or not, and whether it's a metadata resource or not. This is an
61 struct resource_entry resource_entry;
63 /* Specifies which part of the split WIM the resource is located in.
64 * This is on on-disk field.
66 * In stand-alone WIMs, this must be 1.
68 * In split WIMs, every split WIM part has its own lookup table, and in
69 * read_lookup_table() it's currently expected that the part number of
70 * each lookup table entry in a split WIM part's lookup table is the
71 * same as the part number of that split WIM part. So this makes this
72 * field redundant since we store a pointer to the corresponding
73 * WIMStruct in the lookup table entry anyway.
77 /* An enumerated type that identifies where the stream corresponding to
78 * this lookup table entry is actually located.
80 * Obviously if we open a WIM and read its lookup table, the location is
81 * set to RESOURCE_IN_WIM since all the streams will initially be
82 * located in the WIM. However, to deal with problems such as image
83 * capture and image mount, we allow the actual location of the stream
84 * to be somewhere else, such as an external file.
87 /* The lookup table entry does not correspond to a stream (this
88 * state should exist only temporarily) */
89 RESOURCE_NONEXISTENT = 0,
91 /* The stream resource is located in a WIM file. The WIMStruct
92 * for the WIM file will be pointed to by the @wim member. */
95 /* The stream resource is located in an external file. The
96 * name of the file will be provided by @file_on_disk member.
97 * In addition, if @file_on_disk_fp is not NULL, it will be an
98 * open FILE * to the file. */
99 RESOURCE_IN_FILE_ON_DISK,
101 /* The stream resource is located in an external file in the
102 * staging directory for a read-write mount. */
103 RESOURCE_IN_STAGING_FILE,
105 /* The stream resource is directly attached in an in-memory
106 * buffer pointed to by @attached_buffer. */
107 RESOURCE_IN_ATTACHED_BUFFER,
109 /* The stream resource is located in an NTFS volume. It is
110 * identified by volume, filename, data stream name, and by
111 * whether it is a reparse point or not. @ntfs_loc points to a
112 * structure containing this information. */
113 RESOURCE_IN_NTFS_VOLUME,
117 * Number of times this lookup table entry is referenced by dentries.
118 * Unfortunately, this field is not always set correctly in Microsoft's
119 * WIMs, so we have no choice but to fix it if more references to the
120 * lookup table entry are found than stated here. */
124 /* (On-disk field) SHA1 message digest of the stream referenced
125 * by this lookup table entry */
126 u8 hash[SHA1_HASH_SIZE];
128 /* First 4 or 8 bytes of the SHA1 message digest, used for
129 * inserting the entry into the hash table. Since the SHA1
130 * message digest can be considered random, we don't really need
131 * the full 20 byte hash just to insert the entry in a hash
136 /* Pointers to somewhere where the stream is actually located. See the
137 * comments for the @resource_location field above. */
141 char *staging_file_name;
144 struct ntfs_location *ntfs_loc;
148 /* @file_on_disk_fp and @attr are both used to cache file/stream
149 * handles so we don't have re-open them on every read */
151 /* Valid iff resource_location == RESOURCE_IN_FILE_ON_DISK */
152 FILE *file_on_disk_fp;
154 /* Valid iff resource_location == RESOURCE_IN_NTFS_VOLUME */
155 struct _ntfs_attr *attr;
158 /* Pointer to inode that contains the opened file descriptors to
159 * this stream (valid iff resource_location ==
160 * RESOURCE_IN_STAGING_FILE) */
161 struct inode *lte_inode;
167 /* When a WIM file is written, out_refcnt starts at 0 and is incremented
168 * whenever the file resource pointed to by this lookup table entry
169 * needs to be written. The file resource only need to be written when
170 * out_refcnt is nonzero, since otherwise it is not referenced by any
176 /* When a WIM file is written, @output_resource_entry is filled
177 * in with the resource entry for the output WIM. This will not
178 * necessarily be the same as the @resource_entry since:
179 * - The stream may have a different offset in the new WIM
180 * - The stream may have a different compressed size in the
181 * new WIM if the compression type changed
184 struct resource_entry output_resource_entry;
185 struct list_head msg_list;
188 /* This field is used for the special hardlink or symlink image
189 * extraction mode. In these mode, all identical files are linked
190 * together, and @extracted_file will be set to the filename of the
191 * first extracted file containing this stream. */
192 char *extracted_file;
195 /* List of lookup table entries that correspond to streams that have
196 * been extracted to the staging directory when modifying a read-write
198 struct list_head staging_list;
200 /* Temporary field for creating a singly linked list. */
201 struct lookup_table_entry *next_lte_in_swm;
205 static inline u64 wim_resource_size(const struct lookup_table_entry *lte)
207 return lte->resource_entry.original_size;
210 static inline u64 wim_resource_chunks(const struct lookup_table_entry *lte)
212 return (wim_resource_size(lte) + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
216 wim_resource_compressed_size(const struct lookup_table_entry *lte)
218 return lte->resource_entry.size;
222 * XXX Probably should store the compression type directly in the lookup table
226 wim_resource_compression_type(const struct lookup_table_entry *lte)
228 if (!(lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
229 || lte->resource_location != RESOURCE_IN_WIM)
230 return WIM_COMPRESSION_TYPE_NONE;
231 return wimlib_get_compression_type(lte->wim);
235 extern struct lookup_table *new_lookup_table(size_t capacity);
237 extern void lookup_table_insert(struct lookup_table *table,
238 struct lookup_table_entry *lte);
240 /* Unlinks a lookup table entry from the table; does not free it. */
241 static inline void lookup_table_unlink(struct lookup_table *table,
242 struct lookup_table_entry *lte)
244 hlist_del(<e->hash_list);
245 table->num_entries--;
248 extern struct lookup_table_entry *new_lookup_table_entry();
250 extern struct lookup_table_entry *
251 clone_lookup_table_entry(const struct lookup_table_entry *lte);
253 extern int for_lookup_table_entry(struct lookup_table *table,
254 int (*visitor)(struct lookup_table_entry *, void *),
257 extern struct lookup_table_entry *
258 __lookup_resource(const struct lookup_table *table, const u8 hash[]);
260 extern int lookup_resource(WIMStruct *w, const char *path,
261 int lookup_flags, struct dentry **dentry_ret,
262 struct lookup_table_entry **lte_ret,
263 u16 *stream_idx_ret);
265 extern void lte_decrement_refcnt(struct lookup_table_entry *lte,
266 struct lookup_table *table);
268 extern void lte_decrement_num_opened_fds(struct lookup_table_entry *lte);
271 extern int lte_zero_out_refcnt(struct lookup_table_entry *entry, void *ignore);
272 extern int lte_zero_real_refcnt(struct lookup_table_entry *entry, void *ignore);
273 extern int lte_free_extracted_file(struct lookup_table_entry *lte, void *ignone);
275 extern void print_lookup_table_entry(const struct lookup_table_entry *entry);
277 extern int read_lookup_table(WIMStruct *w);
279 extern void free_lookup_table(struct lookup_table *table);
281 extern int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out);
283 extern void free_lookup_table_entry(struct lookup_table_entry *lte);
285 extern int dentry_resolve_ltes(struct dentry *dentry, void *__table);
286 extern int dentry_unresolve_ltes(struct dentry *dentry, void *ignore);
288 /* Writes the lookup table to the output file. */
289 static inline int write_lookup_table(struct lookup_table *table, FILE *out)
291 return for_lookup_table_entry(table, write_lookup_table_entry, out);
294 /* Unlinks and frees an entry from a lookup table. */
295 static inline void lookup_table_remove(struct lookup_table *table,
296 struct lookup_table_entry *lte)
298 lookup_table_unlink(table, lte);
299 free_lookup_table_entry(lte);
302 static inline struct resource_entry* wim_metadata_resource_entry(WIMStruct *w)
304 return &w->image_metadata[
305 w->current_image - 1].metadata_lte->resource_entry;
308 static inline struct lookup_table_entry *
309 inode_stream_lte_resolved(const struct inode *inode, unsigned stream_idx)
311 wimlib_assert(inode->resolved);
312 wimlib_assert(stream_idx <= inode->num_ads);
316 return inode->ads_entries[stream_idx - 1].lte;
319 static inline struct lookup_table_entry *
320 inode_stream_lte_unresolved(const struct inode *inode, unsigned stream_idx,
321 const struct lookup_table *table)
323 wimlib_assert(!inode->resolved);
324 wimlib_assert(stream_idx <= inode->num_ads);
328 return __lookup_resource(table, inode->hash);
330 return __lookup_resource(table,
332 stream_idx - 1].hash);
335 * Returns the lookup table entry for stream @stream_idx of the inode, where
336 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
337 * corresponds to an alternate data stream.
339 * This works for both resolved and un-resolved dentries.
341 static inline struct lookup_table_entry *
342 inode_stream_lte(const struct inode *inode, unsigned stream_idx,
343 const struct lookup_table *table)
346 return inode_stream_lte_resolved(inode, stream_idx);
348 return inode_stream_lte_unresolved(inode, stream_idx, table);
352 static inline const u8 *inode_stream_hash_unresolved(const struct inode *inode,
355 wimlib_assert(!inode->resolved);
356 wimlib_assert(stream_idx <= inode->num_ads);
360 return inode->ads_entries[stream_idx - 1].hash;
364 static inline const u8 *inode_stream_hash_resolved(const struct inode *inode,
367 struct lookup_table_entry *lte;
368 lte = inode_stream_lte_resolved(inode, stream_idx);
376 * Returns the hash for stream @stream_idx of the inode, where stream_idx = 0
377 * means the default un-named file stream, and stream_idx >= 1 corresponds to an
378 * alternate data stream.
380 * This works for both resolved and un-resolved dentries.
382 static inline const u8 *inode_stream_hash(const struct inode *inode,
386 return inode_stream_hash_resolved(inode, stream_idx);
388 return inode_stream_hash_unresolved(inode, stream_idx);
391 static inline u16 inode_stream_name_len(const struct inode *inode,
394 wimlib_assert(stream_idx <= inode->num_ads);
398 return inode->ads_entries[stream_idx - 1].stream_name_len;
401 static inline struct lookup_table_entry *
402 inode_unnamed_lte_resolved(const struct inode *inode)
404 wimlib_assert(inode->resolved);
405 for (unsigned i = 0; i <= inode->num_ads; i++)
406 if (inode_stream_name_len(inode, i) == 0 &&
407 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
408 return inode_stream_lte_resolved(inode, i);
412 static inline struct lookup_table_entry *
413 inode_unnamed_lte_unresolved(const struct inode *inode,
414 const struct lookup_table *table)
416 wimlib_assert(!inode->resolved);
417 for (unsigned i = 0; i <= inode->num_ads; i++)
418 if (inode_stream_name_len(inode, i) == 0 &&
419 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
420 return inode_stream_lte_unresolved(inode, i, table);
424 extern struct lookup_table_entry *
425 inode_unnamed_lte(const struct inode *inode,
426 const struct lookup_table *table);