]> git.ipfire.org Git - thirdparty/git.git/blob - object-store.h
Merge branch 'bc/sha-256-part-2'
[thirdparty/git.git] / object-store.h
1 #ifndef OBJECT_STORE_H
2 #define OBJECT_STORE_H
3
4 #include "cache.h"
5 #include "oidmap.h"
6 #include "list.h"
7 #include "oid-array.h"
8 #include "strbuf.h"
9 #include "thread-utils.h"
10
11 struct object_directory {
12 struct object_directory *next;
13
14 /*
15 * Used to store the results of readdir(3) calls when we are OK
16 * sacrificing accuracy due to races for speed. That includes
17 * object existence with OBJECT_INFO_QUICK, as well as
18 * our search for unique abbreviated hashes. Don't use it for tasks
19 * requiring greater accuracy!
20 *
21 * Be sure to call odb_load_loose_cache() before using.
22 */
23 char loose_objects_subdir_seen[256];
24 struct oid_array loose_objects_cache[256];
25
26 /*
27 * Path to the alternative object store. If this is a relative path,
28 * it is relative to the current working directory.
29 */
30 char *path;
31 };
32
33 void prepare_alt_odb(struct repository *r);
34 char *compute_alternate_path(const char *path, struct strbuf *err);
35 typedef int alt_odb_fn(struct object_directory *, void *);
36 int foreach_alt_odb(alt_odb_fn, void*);
37 typedef void alternate_ref_fn(const struct object_id *oid, void *);
38 void for_each_alternate_ref(alternate_ref_fn, void *);
39
40 /*
41 * Add the directory to the on-disk alternates file; the new entry will also
42 * take effect in the current process.
43 */
44 void add_to_alternates_file(const char *dir);
45
46 /*
47 * Add the directory to the in-memory list of alternates (along with any
48 * recursive alternates it points to), but do not modify the on-disk alternates
49 * file.
50 */
51 void add_to_alternates_memory(const char *dir);
52
53 /*
54 * Populate and return the loose object cache array corresponding to the
55 * given object ID.
56 */
57 struct oid_array *odb_loose_cache(struct object_directory *odb,
58 const struct object_id *oid);
59
60 /* Empty the loose object cache for the specified object directory. */
61 void odb_clear_loose_cache(struct object_directory *odb);
62
63 struct packed_git {
64 struct hashmap_entry packmap_ent;
65 struct packed_git *next;
66 struct list_head mru;
67 struct pack_window *windows;
68 off_t pack_size;
69 const void *index_data;
70 size_t index_size;
71 uint32_t num_objects;
72 uint32_t num_bad_objects;
73 uint32_t crc_offset;
74 unsigned char *bad_object_sha1;
75 int index_version;
76 time_t mtime;
77 int pack_fd;
78 int index; /* for builtin/pack-objects.c */
79 unsigned pack_local:1,
80 pack_keep:1,
81 pack_keep_in_core:1,
82 freshened:1,
83 do_not_close:1,
84 pack_promisor:1,
85 multi_pack_index:1;
86 unsigned char hash[GIT_MAX_RAWSZ];
87 struct revindex_entry *revindex;
88 /* something like ".git/objects/pack/xxxxx.pack" */
89 char pack_name[FLEX_ARRAY]; /* more */
90 };
91
92 struct multi_pack_index;
93
94 static inline int pack_map_entry_cmp(const void *unused_cmp_data,
95 const struct hashmap_entry *entry,
96 const struct hashmap_entry *entry2,
97 const void *keydata)
98 {
99 const char *key = keydata;
100 const struct packed_git *pg1, *pg2;
101
102 pg1 = container_of(entry, const struct packed_git, packmap_ent);
103 pg2 = container_of(entry2, const struct packed_git, packmap_ent);
104
105 return strcmp(pg1->pack_name, key ? key : pg2->pack_name);
106 }
107
108 struct raw_object_store {
109 /*
110 * Set of all object directories; the main directory is first (and
111 * cannot be NULL after initialization). Subsequent directories are
112 * alternates.
113 */
114 struct object_directory *odb;
115 struct object_directory **odb_tail;
116 int loaded_alternates;
117
118 /*
119 * A list of alternate object directories loaded from the environment;
120 * this should not generally need to be accessed directly, but will
121 * populate the "odb" list when prepare_alt_odb() is run.
122 */
123 char *alternate_db;
124
125 /*
126 * Objects that should be substituted by other objects
127 * (see git-replace(1)).
128 */
129 struct oidmap *replace_map;
130 unsigned replace_map_initialized : 1;
131 pthread_mutex_t replace_mutex; /* protect object replace functions */
132
133 struct commit_graph *commit_graph;
134 unsigned commit_graph_attempted : 1; /* if loading has been attempted */
135
136 /*
137 * private data
138 *
139 * should only be accessed directly by packfile.c and midx.c
140 */
141 struct multi_pack_index *multi_pack_index;
142
143 /*
144 * private data
145 *
146 * should only be accessed directly by packfile.c
147 */
148
149 struct packed_git *packed_git;
150 /* A most-recently-used ordered version of the packed_git list. */
151 struct list_head packed_git_mru;
152
153 /*
154 * A map of packfiles to packed_git structs for tracking which
155 * packs have been loaded already.
156 */
157 struct hashmap pack_map;
158
159 /*
160 * A fast, rough count of the number of objects in the repository.
161 * These two fields are not meant for direct access. Use
162 * approximate_object_count() instead.
163 */
164 unsigned long approximate_object_count;
165 unsigned approximate_object_count_valid : 1;
166
167 /*
168 * Whether packed_git has already been populated with this repository's
169 * packs.
170 */
171 unsigned packed_git_initialized : 1;
172 };
173
174 struct raw_object_store *raw_object_store_new(void);
175 void raw_object_store_clear(struct raw_object_store *o);
176
177 /*
178 * Put in `buf` the name of the file in the local object database that
179 * would be used to store a loose object with the specified oid.
180 */
181 const char *loose_object_path(struct repository *r, struct strbuf *buf,
182 const struct object_id *oid);
183
184 void *map_loose_object(struct repository *r, const struct object_id *oid,
185 unsigned long *size);
186
187 void *read_object_file_extended(struct repository *r,
188 const struct object_id *oid,
189 enum object_type *type,
190 unsigned long *size, int lookup_replace);
191 static inline void *repo_read_object_file(struct repository *r,
192 const struct object_id *oid,
193 enum object_type *type,
194 unsigned long *size)
195 {
196 return read_object_file_extended(r, oid, type, size, 1);
197 }
198 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
199 #define read_object_file(oid, type, size) repo_read_object_file(the_repository, oid, type, size)
200 #endif
201
202 /* Read and unpack an object file into memory, write memory to an object file */
203 int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
204
205 int hash_object_file(const struct git_hash_algo *algo, const void *buf,
206 unsigned long len, const char *type,
207 struct object_id *oid);
208
209 int write_object_file(const void *buf, unsigned long len,
210 const char *type, struct object_id *oid);
211
212 int hash_object_file_literally(const void *buf, unsigned long len,
213 const char *type, struct object_id *oid,
214 unsigned flags);
215
216 /*
217 * Add an object file to the in-memory object store, without writing it
218 * to disk.
219 *
220 * Callers are responsible for calling write_object_file to record the
221 * object in persistent storage before writing any other new objects
222 * that reference it.
223 */
224 int pretend_object_file(void *, unsigned long, enum object_type,
225 struct object_id *oid);
226
227 int force_object_loose(const struct object_id *oid, time_t mtime);
228
229 /*
230 * Open the loose object at path, check its hash, and return the contents,
231 * type, and size. If the object is a blob, then "contents" may return NULL,
232 * to allow streaming of large blobs.
233 *
234 * Returns 0 on success, negative on error (details may be written to stderr).
235 */
236 int read_loose_object(const char *path,
237 const struct object_id *expected_oid,
238 enum object_type *type,
239 unsigned long *size,
240 void **contents);
241
242 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
243 #define has_sha1_file_with_flags(sha1, flags) repo_has_sha1_file_with_flags(the_repository, sha1, flags)
244 #define has_sha1_file(sha1) repo_has_sha1_file(the_repository, sha1)
245 #endif
246
247 /* Same as the above, except for struct object_id. */
248 int repo_has_object_file(struct repository *r, const struct object_id *oid);
249 int repo_has_object_file_with_flags(struct repository *r,
250 const struct object_id *oid, int flags);
251 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
252 #define has_object_file(oid) repo_has_object_file(the_repository, oid)
253 #define has_object_file_with_flags(oid, flags) repo_has_object_file_with_flags(the_repository, oid, flags)
254 #endif
255
256 /*
257 * Return true iff an alternate object database has a loose object
258 * with the specified name. This function does not respect replace
259 * references.
260 */
261 int has_loose_object_nonlocal(const struct object_id *);
262
263 void assert_oid_type(const struct object_id *oid, enum object_type expect);
264
265 /*
266 * Enabling the object read lock allows multiple threads to safely call the
267 * following functions in parallel: repo_read_object_file(), read_object_file(),
268 * read_object_file_extended(), read_object_with_reference(), read_object(),
269 * oid_object_info() and oid_object_info_extended().
270 *
271 * obj_read_lock() and obj_read_unlock() may also be used to protect other
272 * section which cannot execute in parallel with object reading. Since the used
273 * lock is a recursive mutex, these sections can even contain calls to object
274 * reading functions. However, beware that in these cases zlib inflation won't
275 * be performed in parallel, losing performance.
276 *
277 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
278 * any of its callees end up calling it, this recursive call won't benefit from
279 * parallel inflation.
280 */
281 void enable_obj_read_lock(void);
282 void disable_obj_read_lock(void);
283
284 extern int obj_read_use_lock;
285 extern pthread_mutex_t obj_read_mutex;
286
287 static inline void obj_read_lock(void)
288 {
289 if(obj_read_use_lock)
290 pthread_mutex_lock(&obj_read_mutex);
291 }
292
293 static inline void obj_read_unlock(void)
294 {
295 if(obj_read_use_lock)
296 pthread_mutex_unlock(&obj_read_mutex);
297 }
298
299 struct object_info {
300 /* Request */
301 enum object_type *typep;
302 unsigned long *sizep;
303 off_t *disk_sizep;
304 struct object_id *delta_base_oid;
305 struct strbuf *type_name;
306 void **contentp;
307
308 /* Response */
309 enum {
310 OI_CACHED,
311 OI_LOOSE,
312 OI_PACKED,
313 OI_DBCACHED
314 } whence;
315 union {
316 /*
317 * struct {
318 * ... Nothing to expose in this case
319 * } cached;
320 * struct {
321 * ... Nothing to expose in this case
322 * } loose;
323 */
324 struct {
325 struct packed_git *pack;
326 off_t offset;
327 unsigned int is_delta;
328 } packed;
329 } u;
330 };
331
332 /*
333 * Initializer for a "struct object_info" that wants no items. You may
334 * also memset() the memory to all-zeroes.
335 */
336 #define OBJECT_INFO_INIT {NULL}
337
338 /* Invoke lookup_replace_object() on the given hash */
339 #define OBJECT_INFO_LOOKUP_REPLACE 1
340 /* Allow reading from a loose object file of unknown/bogus type */
341 #define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
342 /* Do not retry packed storage after checking packed and loose storage */
343 #define OBJECT_INFO_QUICK 8
344 /* Do not check loose object */
345 #define OBJECT_INFO_IGNORE_LOOSE 16
346 /*
347 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
348 * nonzero).
349 */
350 #define OBJECT_INFO_SKIP_FETCH_OBJECT 32
351 /*
352 * This is meant for bulk prefetching of missing blobs in a partial
353 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
354 */
355 #define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
356
357 int oid_object_info_extended(struct repository *r,
358 const struct object_id *,
359 struct object_info *, unsigned flags);
360
361 /*
362 * Iterate over the files in the loose-object parts of the object
363 * directory "path", triggering the following callbacks:
364 *
365 * - loose_object is called for each loose object we find.
366 *
367 * - loose_cruft is called for any files that do not appear to be
368 * loose objects. Note that we only look in the loose object
369 * directories "objects/[0-9a-f]{2}/", so we will not report
370 * "objects/foobar" as cruft.
371 *
372 * - loose_subdir is called for each top-level hashed subdirectory
373 * of the object directory (e.g., "$OBJDIR/f0"). It is called
374 * after the objects in the directory are processed.
375 *
376 * Any callback that is NULL will be ignored. Callbacks returning non-zero
377 * will end the iteration.
378 *
379 * In the "buf" variant, "path" is a strbuf which will also be used as a
380 * scratch buffer, but restored to its original contents before
381 * the function returns.
382 */
383 typedef int each_loose_object_fn(const struct object_id *oid,
384 const char *path,
385 void *data);
386 typedef int each_loose_cruft_fn(const char *basename,
387 const char *path,
388 void *data);
389 typedef int each_loose_subdir_fn(unsigned int nr,
390 const char *path,
391 void *data);
392 int for_each_file_in_obj_subdir(unsigned int subdir_nr,
393 struct strbuf *path,
394 each_loose_object_fn obj_cb,
395 each_loose_cruft_fn cruft_cb,
396 each_loose_subdir_fn subdir_cb,
397 void *data);
398 int for_each_loose_file_in_objdir(const char *path,
399 each_loose_object_fn obj_cb,
400 each_loose_cruft_fn cruft_cb,
401 each_loose_subdir_fn subdir_cb,
402 void *data);
403 int for_each_loose_file_in_objdir_buf(struct strbuf *path,
404 each_loose_object_fn obj_cb,
405 each_loose_cruft_fn cruft_cb,
406 each_loose_subdir_fn subdir_cb,
407 void *data);
408
409 /* Flags for for_each_*_object() below. */
410 enum for_each_object_flags {
411 /* Iterate only over local objects, not alternates. */
412 FOR_EACH_OBJECT_LOCAL_ONLY = (1<<0),
413
414 /* Only iterate over packs obtained from the promisor remote. */
415 FOR_EACH_OBJECT_PROMISOR_ONLY = (1<<1),
416
417 /*
418 * Visit objects within a pack in packfile order rather than .idx order
419 */
420 FOR_EACH_OBJECT_PACK_ORDER = (1<<2),
421 };
422
423 /*
424 * Iterate over all accessible loose objects without respect to
425 * reachability. By default, this includes both local and alternate objects.
426 * The order in which objects are visited is unspecified.
427 *
428 * Any flags specific to packs are ignored.
429 */
430 int for_each_loose_object(each_loose_object_fn, void *,
431 enum for_each_object_flags flags);
432
433 /*
434 * Iterate over all accessible packed objects without respect to reachability.
435 * By default, this includes both local and alternate packs.
436 *
437 * Note that some objects may appear twice if they are found in multiple packs.
438 * Each pack is visited in an unspecified order. By default, objects within a
439 * pack are visited in pack-idx order (i.e., sorted by oid).
440 */
441 typedef int each_packed_object_fn(const struct object_id *oid,
442 struct packed_git *pack,
443 uint32_t pos,
444 void *data);
445 int for_each_object_in_pack(struct packed_git *p,
446 each_packed_object_fn, void *data,
447 enum for_each_object_flags flags);
448 int for_each_packed_object(each_packed_object_fn, void *,
449 enum for_each_object_flags flags);
450
451 #endif /* OBJECT_STORE_H */