1 #ifndef _WIMLIB_LOOKUP_TABLE_H
2 #define _WIMLIB_LOOKUP_TABLE_H
4 #include "wimlib/assert.h"
5 #include "wimlib/dentry.h"
6 #include "wimlib/list.h"
7 #include "wimlib/sha1.h"
8 #include "wimlib/types.h"
9 #include "wimlib/wim.h"
11 #define LOOKUP_FLAG_ADS_OK 0x00000001
12 #define LOOKUP_FLAG_DIRECTORY_OK 0x00000002
15 /* The lookup table of a WIM file maps SHA1 message digests to streams of data.
16 * Here, the in-memory structure is implemented as a hash table.
18 * Given a SHA1 message digest, the mapped-to stream is specified by an offset
19 * in the WIM, an uncompressed and compressed size, and resource flags (see
20 * 'struct resource_entry'). But, we associate additional information, such as
21 * a reference count, with each stream, so the actual mapping is from SHA1
22 * message digests to 'struct wim_lookup_table_entry's, each of which contains
23 * an embedded 'struct resource_entry'.
25 * Note: Everything will break horribly if there is a SHA1 collision.
27 struct wim_lookup_table {
28 struct hlist_head *array;
31 struct list_head *unhashed_streams;
38 struct ntfs_location {
40 utf16lechar *stream_name;
41 u16 stream_name_nchars;
42 struct _ntfs_volume *ntfs_vol;
43 bool is_reparse_point;
47 /* An enumerated type that identifies where the stream corresponding to this
48 * lookup table entry is actually located.
50 * If we open a WIM and read its lookup table, the location is set to
51 * RESOURCE_IN_WIM since all the streams will initially be located in the WIM.
52 * However, to handle situations such as image capture and image mount, we allow
53 * the actual location of the stream to be somewhere else, such as an external
56 enum resource_location {
57 /* The lookup table entry does not yet correspond to a stream; this is a
58 * temporary state only. */
59 RESOURCE_NONEXISTENT = 0,
61 /* The stream is located in a resource in a WIM file identified by the
62 * `struct wim_resource_spec' pointed to by @rspec. @offset_in_res
63 * identifies the offset at which this particular stream begins in the
64 * uncompressed data of the resource; this is normally 0, but in general
65 * a WIM resource may contain multiple streams. */
68 /* The stream is located in the external file named by @file_on_disk.
69 * On Windows, @file_on_disk may actually specify a named data stream.
71 RESOURCE_IN_FILE_ON_DISK,
73 /* The stream is directly attached in the in-memory buffer pointed to by
74 * @attached_buffer. */
75 RESOURCE_IN_ATTACHED_BUFFER,
78 /* The stream is located in the external file named by
79 * @staging_file_name, located in the staging directory for a read-write
81 RESOURCE_IN_STAGING_FILE,
85 /* The stream is located in an NTFS volume. It is identified by volume,
86 * filename, data stream name, and by whether it is a reparse point or
87 * not. @ntfs_loc points to a structure containing this information.
89 RESOURCE_IN_NTFS_VOLUME,
93 /* Windows only: the stream is located in the external file named by
94 * @file_on_disk, but the file is encrypted and must be read using the
95 * appropriate Windows API. */
96 RESOURCE_WIN32_ENCRYPTED,
102 * An entry in the lookup table in the WIM file.
104 * It is used to find data streams for files in the WIM.
106 * Metadata resources and reparse point data buffers will also have lookup table
107 * entries associated with the data.
109 * The lookup_table_entry for a given dentry or alternate stream entry in the
110 * WIM is found using the SHA1 message digest field.
112 struct wim_lookup_table_entry {
114 /* List of lookup table entries in this hash bucket */
115 struct hlist_node hash_list;
117 /* Uncompressed size of the stream. */
120 /* Stream flags (WIM_RESHDR_FLAG_*). */
123 /* One of the `enum resource_location' values documented above. */
124 u32 resource_location : 5;
126 /* 1 if this stream has not had a SHA1 message digest calculated for it
130 /* Temoorary files used for writing; set as documented for
131 * prepare_stream_list_for_write(). */
133 u32 will_be_in_output_wim : 1;
135 /* Set to 1 when a metadata entry has its checksum changed; in such
136 * cases the hash is no longer valid to verify the data if the metadata
137 * resource is read again. */
138 u32 dont_check_metadata_hash : 1;
141 /* (On-disk field) SHA1 message digest of the stream referenced
142 * by this lookup table entry. */
143 u8 hash[SHA1_HASH_SIZE];
145 /* First 4 or 8 bytes of the SHA1 message digest, used for
146 * inserting the entry into the hash table. Since the SHA1
147 * message digest can be considered random, we don't really need
148 * the full 20 byte hash just to insert the entry in a hash
152 /* Unhashed entries only (unhashed == 1): these variables make
153 * it possible to find the pointer to this 'struct
154 * wim_lookup_table_entry' contained in either 'struct
155 * wim_ads_entry' or 'struct wim_inode'. There can be at most 1
156 * such pointer, as we can only join duplicate streams after
157 * they have been hashed. */
159 struct wim_inode *back_inode;
164 /* Number of times this lookup table entry is referenced by dentries in
168 /* Actual reference count to this stream (only used while verifying an
172 /* When a WIM file is written, out_refcnt starts at 0 and is incremented
173 * whenever the stream pointed to by this lookup table entry needs to be
174 * written. The stream only need to be written when out_refcnt is
175 * nonzero, since otherwise it is not referenced by any dentries. */
179 /* Number of times this stream has been opened (used only during
184 /* Pointers to somewhere where the stream is actually located. See the
185 * comments for the @resource_location field above. */
188 struct wim_resource_spec *rspec;
192 void *attached_buffer;
194 tchar *staging_file_name;
197 struct ntfs_location *ntfs_loc;
201 /* Links together streams that share the same underlying WIM resource.
202 * The head is wim_resource_spec.stream_list. */
203 struct list_head rspec_node;
205 /* This field is used for the special hardlink or symlink image
206 * extraction mode. In these mode, all identical files are linked
207 * together, and @extracted_file will be set to the filename of the
208 * first extracted file containing this stream. */
209 tchar *extracted_file;
211 /* Temporary fields */
213 /* Used temporarily during WIM file writing */
216 struct hlist_node hash_list_2;
218 u64 out_res_offset_in_wim;
219 u64 out_res_size_in_wim;
220 u64 out_res_uncompressed_size;
223 /* Links streams being written to the WIM. */
224 struct list_head write_streams_list;
226 struct wim_reshdr out_reshdr;
229 /* Used temporarily during extraction */
231 /* out_refcnt tracks number of slots filled */
232 struct wim_dentry *inline_lte_dentries[8];
234 struct wim_dentry **lte_dentries;
235 unsigned long alloc_lte_dentries;
240 /* Temporary list fields */
242 /* Links streams for writing lookup table. */
243 struct list_head lookup_table_list;
245 /* Links streams being extracted. */
246 struct list_head extraction_list;
248 /* Links streams being exported. */
249 struct list_head export_stream_list;
252 /* Links streams that are still unhashed after being been added to a
254 struct list_head unhashed_list;
258 lte_is_partial(const struct wim_lookup_table_entry * lte)
260 return lte->resource_location == RESOURCE_IN_WIM &&
261 lte->size != lte->rspec->uncompressed_size;
265 lte_filename_valid(const struct wim_lookup_table_entry *lte)
267 return lte->resource_location == RESOURCE_IN_FILE_ON_DISK
269 || lte->resource_location == RESOURCE_WIN32_ENCRYPTED
272 || lte->resource_location == RESOURCE_IN_STAGING_FILE
277 extern struct wim_lookup_table *
278 new_lookup_table(size_t capacity) _malloc_attribute;
281 read_wim_lookup_table(WIMStruct *wim);
284 write_wim_lookup_table_from_stream_list(struct list_head *stream_list,
285 struct filedes *out_fd,
287 struct wim_reshdr *out_reshdr,
288 int write_resource_flags,
289 struct wimlib_lzx_context **comp_ctx);
292 free_lookup_table(struct wim_lookup_table *table);
295 lookup_table_insert(struct wim_lookup_table *table, struct wim_lookup_table_entry *lte);
297 /* Unlinks a lookup table entry from the table; does not free it. */
299 lookup_table_unlink(struct wim_lookup_table *table, struct wim_lookup_table_entry *lte)
301 wimlib_assert(!lte->unhashed);
302 hlist_del(<e->hash_list);
303 wimlib_assert(table->num_entries != 0);
304 table->num_entries--;
307 extern struct wim_lookup_table_entry *
308 new_lookup_table_entry(void) _malloc_attribute;
310 extern struct wim_lookup_table_entry *
311 clone_lookup_table_entry(const struct wim_lookup_table_entry *lte)
315 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out);
318 free_lookup_table_entry(struct wim_lookup_table_entry *lte);
321 lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
322 struct wimlib_resource_entry *wentry);
325 for_lookup_table_entry(struct wim_lookup_table *table,
326 int (*visitor)(struct wim_lookup_table_entry *, void *),
330 sort_stream_list(struct list_head *stream_list,
331 size_t list_head_offset,
332 int (*compar)(const void *, const void*));
335 sort_stream_list_by_sequential_order(struct list_head *stream_list,
336 size_t list_head_offset);
339 for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
340 int (*visitor)(struct wim_lookup_table_entry *,
344 extern struct wim_lookup_table_entry *
345 lookup_resource(const struct wim_lookup_table *table, const u8 hash[]);
348 wim_pathname_to_stream(WIMStruct *wim, const tchar *path,
350 struct wim_dentry **dentry_ret,
351 struct wim_lookup_table_entry **lte_ret,
352 u16 *stream_idx_ret);
355 lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
356 struct wim_lookup_table *table);
359 lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte);
363 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *ignore);
366 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *ignore);
369 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *ignore);
372 lte_bind_wim_resource_spec(struct wim_lookup_table_entry *lte,
373 struct wim_resource_spec *rspec)
375 lte->resource_location = RESOURCE_IN_WIM;
377 list_add_tail(<e->rspec_node, &rspec->stream_list);
381 lte_unbind_wim_resource_spec(struct wim_lookup_table_entry *lte)
383 list_del(<e->rspec_node);
385 lte->resource_location = RESOURCE_NONEXISTENT;
389 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table,
393 resource_not_found_error(const struct wim_inode *inode, const u8 *hash);
396 inode_unresolve_ltes(struct wim_inode *inode);
398 static inline struct wim_lookup_table_entry *
399 inode_stream_lte_resolved(const struct wim_inode *inode, unsigned stream_idx)
401 wimlib_assert(inode->i_resolved);
402 wimlib_assert(stream_idx <= inode->i_num_ads);
406 return inode->i_ads_entries[stream_idx - 1].lte;
409 static inline struct wim_lookup_table_entry *
410 inode_stream_lte_unresolved(const struct wim_inode *inode, unsigned stream_idx,
411 const struct wim_lookup_table *table)
413 wimlib_assert(!inode->i_resolved);
414 wimlib_assert(stream_idx <= inode->i_num_ads);
418 return lookup_resource(table, inode->i_hash);
420 return lookup_resource(table,
421 inode->i_ads_entries[
422 stream_idx - 1].hash);
425 extern struct wim_lookup_table_entry *
426 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
427 const struct wim_lookup_table *table);
429 static inline const u8 *
430 inode_stream_hash_unresolved(const struct wim_inode *inode, unsigned stream_idx)
432 wimlib_assert(!inode->i_resolved);
433 wimlib_assert(stream_idx <= inode->i_num_ads);
435 return inode->i_hash;
437 return inode->i_ads_entries[stream_idx - 1].hash;
441 static inline const u8 *
442 inode_stream_hash_resolved(const struct wim_inode *inode, unsigned stream_idx)
444 struct wim_lookup_table_entry *lte;
445 lte = inode_stream_lte_resolved(inode, stream_idx);
453 * Returns the hash for stream @stream_idx of the inode, where stream_idx = 0
454 * means the default un-named file stream, and stream_idx >= 1 corresponds to an
455 * alternate data stream.
457 * This works for both resolved and un-resolved dentries.
459 static inline const u8 *
460 inode_stream_hash(const struct wim_inode *inode, unsigned stream_idx)
462 if (inode->i_resolved)
463 return inode_stream_hash_resolved(inode, stream_idx);
465 return inode_stream_hash_unresolved(inode, stream_idx);
469 inode_stream_name_nbytes(const struct wim_inode *inode, unsigned stream_idx)
471 wimlib_assert(stream_idx <= inode->i_num_ads);
475 return inode->i_ads_entries[stream_idx - 1].stream_name_nbytes;
478 extern struct wim_lookup_table_entry *
479 inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret);
481 extern struct wim_lookup_table_entry *
482 inode_unnamed_lte_resolved(const struct wim_inode *inode);
484 extern struct wim_lookup_table_entry *
485 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
486 const struct wim_lookup_table *table);
488 extern struct wim_lookup_table_entry *
489 inode_unnamed_lte(const struct wim_inode *inode, const struct wim_lookup_table *table);
492 inode_unnamed_stream_hash(const struct wim_inode *inode);
495 lookup_table_insert_unhashed(struct wim_lookup_table *table,
496 struct wim_lookup_table_entry *lte,
497 struct wim_inode *back_inode,
501 lte->back_inode = back_inode;
502 lte->back_stream_id = back_stream_id;
503 list_add_tail(<e->unhashed_list, table->unhashed_streams);
507 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
508 struct wim_lookup_table *lookup_table,
509 struct wim_lookup_table_entry **lte_ret);
511 extern struct wim_lookup_table_entry **
512 retrieve_lte_pointer(struct wim_lookup_table_entry *lte);
514 #endif /* _WIMLIB_LOOKUP_TABLE_H */