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;
28 struct ntfs_location {
30 char *stream_name_utf16;
31 u16 stream_name_utf16_num_chars;
32 struct _ntfs_volume **ntfs_vol_p;
33 bool is_reparse_point;
37 /* An enumerated type that identifies where the stream corresponding to this
38 * lookup table entry is actually located.
40 * If we open a WIM and read its lookup table, the location is set to
41 * RESOURCE_IN_WIM since all the streams will initially be located in the WIM.
42 * However, to deal with problems such as image capture and image mount, we
43 * allow the actual location of the stream to be somewhere else, such as an
46 enum resource_location {
47 /* The lookup table entry does not correspond to a stream (this state
48 * should exist only temporarily) */
49 RESOURCE_NONEXISTENT = 0,
51 /* The stream resource is located in a WIM file. The WIMStruct for the
52 * WIM file will be pointed to by the @wim member. */
55 /* The stream resource is located in an external file. The name of the
56 * file will be provided by @file_on_disk member. In addition, if
57 * @file_on_disk_fp is not NULL, it will be an open FILE * to the file.
59 RESOURCE_IN_FILE_ON_DISK,
61 /* The stream resource is located in an external file in the staging
62 * directory for a read-write mount. */
63 RESOURCE_IN_STAGING_FILE,
65 /* The stream resource is directly attached in an in-memory buffer
66 * pointed to by @attached_buffer. */
67 RESOURCE_IN_ATTACHED_BUFFER,
69 /* The stream resource is located in an NTFS volume. It is identified
70 * by volume, filename, data stream name, and by whether it is a reparse
71 * point or not. @ntfs_loc points to a structure containing this
73 RESOURCE_IN_NTFS_VOLUME,
77 * An entry in the lookup table in the WIM file.
79 * It is used to find data streams for files in the WIM.
81 * Metadata resources and reparse point data buffers will also have lookup table
82 * entries associated with the data.
84 * The lookup_table_entry for a given dentry or alternate stream entry in the
85 * WIM is found using the SHA1 message digest field.
87 struct lookup_table_entry {
89 /* List of lookup table entries in this hash bucket */
90 struct hlist_node hash_list;
92 /* Location and size of the stream in the WIM, whether it is compressed
93 * or not, and whether it's a metadata resource or not. This is an
95 struct resource_entry resource_entry;
97 /* Specifies which part of the split WIM the resource is located in.
98 * This is on on-disk field.
100 * In stand-alone WIMs, this must be 1.
102 * In split WIMs, every split WIM part has its own lookup table, and in
103 * read_lookup_table() it's currently expected that the part number of
104 * each lookup table entry in a split WIM part's lookup table is the
105 * same as the part number of that split WIM part. So this makes this
106 * field redundant since we store a pointer to the corresponding
107 * WIMStruct in the lookup table entry anyway.
111 /* See enum resource_location above */
112 u16 resource_location;
115 * Number of times this lookup table entry is referenced by dentries.
116 * Unfortunately, this field is not always set correctly in Microsoft's
117 * WIMs, so we have no choice but to fix it if more references to the
118 * lookup table entry are found than stated here. */
122 /* (On-disk field) SHA1 message digest of the stream referenced
123 * by this lookup table entry */
124 u8 hash[SHA1_HASH_SIZE];
126 /* First 4 or 8 bytes of the SHA1 message digest, used for
127 * inserting the entry into the hash table. Since the SHA1
128 * message digest can be considered random, we don't really need
129 * the full 20 byte hash just to insert the entry in a hash
139 /* This field is used for the special hardlink or symlink image
140 * extraction mode. In these mode, all identical files are linked
141 * together, and @extracted_file will be set to the filename of the
142 * first extracted file containing this stream. */
143 char *extracted_file;
146 /* Pointers to somewhere where the stream is actually located. See the
147 * comments for the @resource_location field above. */
151 char *staging_file_name;
154 struct ntfs_location *ntfs_loc;
158 /* @file_on_disk_fp and @attr are both used to cache file/stream
159 * handles so we don't have re-open them on every read */
161 /* Valid iff resource_location == RESOURCE_IN_FILE_ON_DISK */
162 FILE *file_on_disk_fp;
164 /* Valid iff resource_location == RESOURCE_IN_NTFS_VOLUME */
165 struct _ntfs_attr *attr;
168 /* Pointer to inode that contains the opened file descriptors to
169 * this stream (valid iff resource_location ==
170 * RESOURCE_IN_STAGING_FILE) */
171 struct inode *lte_inode;
174 /* When a WIM file is written, out_refcnt starts at 0 and is incremented
175 * whenever the file resource pointed to by this lookup table entry
176 * needs to be written. The file resource only need to be written when
177 * out_refcnt is nonzero, since otherwise it is not referenced by any
184 /* When a WIM file is written, @output_resource_entry is filled
185 * in with the resource entry for the output WIM. This will not
186 * necessarily be the same as the @resource_entry since: - The
187 * stream may have a different offset in the new WIM - The
188 * stream may have a different compressed size in the new WIM if
189 * the compression type changed
191 struct resource_entry output_resource_entry;
193 struct list_head msg_list;
194 struct list_head inode_list;
197 /* List of lookup table entries that correspond to streams that have
198 * been extracted to the staging directory when modifying a read-write
201 * This field is also used to make other lists of lookup table entries.
203 struct list_head staging_list;
207 wim_resource_size(const struct lookup_table_entry *lte)
209 return lte->resource_entry.original_size;
213 wim_resource_chunks(const struct lookup_table_entry *lte)
215 return (wim_resource_size(lte) + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
219 wim_resource_compressed_size(const struct lookup_table_entry *lte)
221 return lte->resource_entry.size;
225 * XXX Probably should store the compression type directly in the lookup table
229 wim_resource_compression_type(const struct lookup_table_entry *lte)
231 if (!(lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
232 || lte->resource_location != RESOURCE_IN_WIM)
233 return WIMLIB_COMPRESSION_TYPE_NONE;
234 return wimlib_get_compression_type(lte->wim);
238 extern struct lookup_table *
239 new_lookup_table(size_t capacity);
242 read_lookup_table(WIMStruct *w);
245 write_lookup_table(struct lookup_table *table, FILE *out,
246 struct resource_entry *out_res_entry);
248 free_lookup_table(struct lookup_table *table);
251 lookup_table_insert(struct lookup_table *table, struct lookup_table_entry *lte);
253 /* Unlinks a lookup table entry from the table; does not free it. */
255 lookup_table_unlink(struct lookup_table *table, struct lookup_table_entry *lte)
257 hlist_del(<e->hash_list);
258 table->num_entries--;
261 extern struct lookup_table_entry *
262 new_lookup_table_entry();
264 extern struct lookup_table_entry *
265 clone_lookup_table_entry(const struct lookup_table_entry *lte);
268 print_lookup_table_entry(const struct lookup_table_entry *entry);
271 free_lookup_table_entry(struct lookup_table_entry *lte);
274 for_lookup_table_entry(struct lookup_table *table,
275 int (*visitor)(struct lookup_table_entry *, void *),
278 extern struct lookup_table_entry *
279 __lookup_resource(const struct lookup_table *table, const u8 hash[]);
282 lookup_resource(WIMStruct *w, const char *path,
283 int lookup_flags, struct dentry **dentry_ret,
284 struct lookup_table_entry **lte_ret, u16 *stream_idx_ret);
287 lte_decrement_refcnt(struct lookup_table_entry *lte,
288 struct lookup_table *table);
291 lte_decrement_num_opened_fds(struct lookup_table_entry *lte);
295 lte_zero_out_refcnt(struct lookup_table_entry *entry, void *ignore);
298 lte_zero_real_refcnt(struct lookup_table_entry *entry, void *ignore);
301 lte_free_extracted_file(struct lookup_table_entry *lte, void *ignore);
304 inode_resolve_ltes(struct inode *inode, struct lookup_table *table);
307 inode_unresolve_ltes(struct inode *inode);
310 write_lookup_table_entry(struct lookup_table_entry *lte, void *__out);
312 static inline struct resource_entry*
313 wim_metadata_resource_entry(WIMStruct *w)
315 return &w->image_metadata[
316 w->current_image - 1].metadata_lte->resource_entry;
319 static inline struct lookup_table_entry *
320 inode_stream_lte_resolved(const struct inode *inode, unsigned stream_idx)
322 wimlib_assert(inode->resolved);
323 wimlib_assert(stream_idx <= inode->num_ads);
327 return inode->ads_entries[stream_idx - 1].lte;
330 static inline struct lookup_table_entry *
331 inode_stream_lte_unresolved(const struct inode *inode, unsigned stream_idx,
332 const struct lookup_table *table)
334 wimlib_assert(!inode->resolved);
335 wimlib_assert(stream_idx <= inode->num_ads);
339 return __lookup_resource(table, inode->hash);
341 return __lookup_resource(table,
343 stream_idx - 1].hash);
346 extern struct lookup_table_entry *
347 inode_stream_lte(const struct inode *inode, unsigned stream_idx,
348 const struct lookup_table *table);
350 static inline const u8 *
351 inode_stream_hash_unresolved(const struct inode *inode, unsigned stream_idx)
353 wimlib_assert(!inode->resolved);
354 wimlib_assert(stream_idx <= inode->num_ads);
358 return inode->ads_entries[stream_idx - 1].hash;
362 static inline const u8 *
363 inode_stream_hash_resolved(const struct inode *inode, unsigned stream_idx)
365 struct lookup_table_entry *lte;
366 lte = inode_stream_lte_resolved(inode, stream_idx);
374 * Returns the hash for stream @stream_idx of the inode, where stream_idx = 0
375 * means the default un-named file stream, and stream_idx >= 1 corresponds to an
376 * alternate data stream.
378 * This works for both resolved and un-resolved dentries.
380 static inline const u8 *
381 inode_stream_hash(const struct inode *inode, unsigned stream_idx)
384 return inode_stream_hash_resolved(inode, stream_idx);
386 return inode_stream_hash_unresolved(inode, stream_idx);
390 inode_stream_name_len(const struct inode *inode, unsigned stream_idx)
392 wimlib_assert(stream_idx <= inode->num_ads);
396 return inode->ads_entries[stream_idx - 1].stream_name_len;
399 static inline struct lookup_table_entry *
400 inode_unnamed_lte_resolved(const struct inode *inode)
402 wimlib_assert(inode->resolved);
403 for (unsigned i = 0; i <= inode->num_ads; i++)
404 if (inode_stream_name_len(inode, i) == 0 &&
405 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
406 return inode_stream_lte_resolved(inode, i);
410 static inline struct lookup_table_entry *
411 inode_unnamed_lte_unresolved(const struct inode *inode,
412 const struct lookup_table *table)
414 wimlib_assert(!inode->resolved);
415 for (unsigned i = 0; i <= inode->num_ads; i++)
416 if (inode_stream_name_len(inode, i) == 0 &&
417 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
418 return inode_stream_lte_unresolved(inode, i, table);
422 extern struct lookup_table_entry *
423 inode_unnamed_lte(const struct inode *inode, const struct lookup_table *table);
426 lookup_table_total_stream_size(struct lookup_table *table);