]> git.ipfire.org Git - thirdparty/git.git/blob - object-store.h
Merge branch 'tb/diffstat-with-utf8-strwidth'
[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 #include "khash.h"
11 #include "dir.h"
12 #include "oidtree.h"
13 #include "oidset.h"
14
15 struct object_directory {
16 struct object_directory *next;
17
18 /*
19 * Used to store the results of readdir(3) calls when we are OK
20 * sacrificing accuracy due to races for speed. That includes
21 * object existence with OBJECT_INFO_QUICK, as well as
22 * our search for unique abbreviated hashes. Don't use it for tasks
23 * requiring greater accuracy!
24 *
25 * Be sure to call odb_load_loose_cache() before using.
26 */
27 uint32_t loose_objects_subdir_seen[8]; /* 256 bits */
28 struct oidtree *loose_objects_cache;
29
30 /*
31 * This is a temporary object store created by the tmp_objdir
32 * facility. Disable ref updates since the objects in the store
33 * might be discarded on rollback.
34 */
35 int disable_ref_updates;
36
37 /*
38 * This object store is ephemeral, so there is no need to fsync.
39 */
40 int will_destroy;
41
42 /*
43 * Path to the alternative object store. If this is a relative path,
44 * it is relative to the current working directory.
45 */
46 char *path;
47 };
48
49 struct input_stream {
50 const void *(*read)(struct input_stream *, unsigned long *len);
51 void *data;
52 int is_finished;
53 };
54
55 KHASH_INIT(odb_path_map, const char * /* key: odb_path */,
56 struct object_directory *, 1, fspathhash, fspatheq)
57
58 void prepare_alt_odb(struct repository *r);
59 char *compute_alternate_path(const char *path, struct strbuf *err);
60 struct object_directory *find_odb(struct repository *r, const char *obj_dir);
61 typedef int alt_odb_fn(struct object_directory *, void *);
62 int foreach_alt_odb(alt_odb_fn, void*);
63 typedef void alternate_ref_fn(const struct object_id *oid, void *);
64 void for_each_alternate_ref(alternate_ref_fn, void *);
65
66 /*
67 * Add the directory to the on-disk alternates file; the new entry will also
68 * take effect in the current process.
69 */
70 void add_to_alternates_file(const char *dir);
71
72 /*
73 * Add the directory to the in-memory list of alternates (along with any
74 * recursive alternates it points to), but do not modify the on-disk alternates
75 * file.
76 */
77 void add_to_alternates_memory(const char *dir);
78
79 /*
80 * Replace the current writable object directory with the specified temporary
81 * object directory; returns the former primary object directory.
82 */
83 struct object_directory *set_temporary_primary_odb(const char *dir, int will_destroy);
84
85 /*
86 * Restore a previous ODB replaced by set_temporary_main_odb.
87 */
88 void restore_primary_odb(struct object_directory *restore_odb, const char *old_path);
89
90 /*
91 * Populate and return the loose object cache array corresponding to the
92 * given object ID.
93 */
94 struct oidtree *odb_loose_cache(struct object_directory *odb,
95 const struct object_id *oid);
96
97 /* Empty the loose object cache for the specified object directory. */
98 void odb_clear_loose_cache(struct object_directory *odb);
99
100 /* Clear and free the specified object directory */
101 void free_object_directory(struct object_directory *odb);
102
103 struct packed_git {
104 struct hashmap_entry packmap_ent;
105 struct packed_git *next;
106 struct list_head mru;
107 struct pack_window *windows;
108 off_t pack_size;
109 const void *index_data;
110 size_t index_size;
111 uint32_t num_objects;
112 uint32_t crc_offset;
113 struct oidset bad_objects;
114 int index_version;
115 time_t mtime;
116 int pack_fd;
117 int index; /* for builtin/pack-objects.c */
118 unsigned pack_local:1,
119 pack_keep:1,
120 pack_keep_in_core:1,
121 freshened:1,
122 do_not_close:1,
123 pack_promisor:1,
124 multi_pack_index:1,
125 is_cruft:1;
126 unsigned char hash[GIT_MAX_RAWSZ];
127 struct revindex_entry *revindex;
128 const uint32_t *revindex_data;
129 const uint32_t *revindex_map;
130 size_t revindex_size;
131 /*
132 * mtimes_map points at the beginning of the memory mapped region of
133 * this pack's corresponding .mtimes file, and mtimes_size is the size
134 * of that .mtimes file
135 */
136 const uint32_t *mtimes_map;
137 size_t mtimes_size;
138 /* something like ".git/objects/pack/xxxxx.pack" */
139 char pack_name[FLEX_ARRAY]; /* more */
140 };
141
142 struct multi_pack_index;
143
144 static inline int pack_map_entry_cmp(const void *cmp_data UNUSED,
145 const struct hashmap_entry *entry,
146 const struct hashmap_entry *entry2,
147 const void *keydata)
148 {
149 const char *key = keydata;
150 const struct packed_git *pg1, *pg2;
151
152 pg1 = container_of(entry, const struct packed_git, packmap_ent);
153 pg2 = container_of(entry2, const struct packed_git, packmap_ent);
154
155 return strcmp(pg1->pack_name, key ? key : pg2->pack_name);
156 }
157
158 struct raw_object_store {
159 /*
160 * Set of all object directories; the main directory is first (and
161 * cannot be NULL after initialization). Subsequent directories are
162 * alternates.
163 */
164 struct object_directory *odb;
165 struct object_directory **odb_tail;
166 kh_odb_path_map_t *odb_by_path;
167
168 int loaded_alternates;
169
170 /*
171 * A list of alternate object directories loaded from the environment;
172 * this should not generally need to be accessed directly, but will
173 * populate the "odb" list when prepare_alt_odb() is run.
174 */
175 char *alternate_db;
176
177 /*
178 * Objects that should be substituted by other objects
179 * (see git-replace(1)).
180 */
181 struct oidmap *replace_map;
182 unsigned replace_map_initialized : 1;
183 pthread_mutex_t replace_mutex; /* protect object replace functions */
184
185 struct commit_graph *commit_graph;
186 unsigned commit_graph_attempted : 1; /* if loading has been attempted */
187
188 /*
189 * private data
190 *
191 * should only be accessed directly by packfile.c and midx.c
192 */
193 struct multi_pack_index *multi_pack_index;
194
195 /*
196 * private data
197 *
198 * should only be accessed directly by packfile.c
199 */
200
201 struct packed_git *packed_git;
202 /* A most-recently-used ordered version of the packed_git list. */
203 struct list_head packed_git_mru;
204
205 struct {
206 struct packed_git **packs;
207 unsigned flags;
208 } kept_pack_cache;
209
210 /*
211 * A map of packfiles to packed_git structs for tracking which
212 * packs have been loaded already.
213 */
214 struct hashmap pack_map;
215
216 /*
217 * A fast, rough count of the number of objects in the repository.
218 * These two fields are not meant for direct access. Use
219 * approximate_object_count() instead.
220 */
221 unsigned long approximate_object_count;
222 unsigned approximate_object_count_valid : 1;
223
224 /*
225 * Whether packed_git has already been populated with this repository's
226 * packs.
227 */
228 unsigned packed_git_initialized : 1;
229 };
230
231 struct raw_object_store *raw_object_store_new(void);
232 void raw_object_store_clear(struct raw_object_store *o);
233
234 /*
235 * Put in `buf` the name of the file in the local object database that
236 * would be used to store a loose object with the specified oid.
237 */
238 const char *loose_object_path(struct repository *r, struct strbuf *buf,
239 const struct object_id *oid);
240
241 void *map_loose_object(struct repository *r, const struct object_id *oid,
242 unsigned long *size);
243
244 void *read_object_file_extended(struct repository *r,
245 const struct object_id *oid,
246 enum object_type *type,
247 unsigned long *size, int lookup_replace);
248 static inline void *repo_read_object_file(struct repository *r,
249 const struct object_id *oid,
250 enum object_type *type,
251 unsigned long *size)
252 {
253 return read_object_file_extended(r, oid, type, size, 1);
254 }
255 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
256 #define read_object_file(oid, type, size) repo_read_object_file(the_repository, oid, type, size)
257 #endif
258
259 /* Read and unpack an object file into memory, write memory to an object file */
260 int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
261
262 void hash_object_file(const struct git_hash_algo *algo, const void *buf,
263 unsigned long len, enum object_type type,
264 struct object_id *oid);
265
266 int write_object_file_flags(const void *buf, unsigned long len,
267 enum object_type type, struct object_id *oid,
268 unsigned flags);
269 static inline int write_object_file(const void *buf, unsigned long len,
270 enum object_type type, struct object_id *oid)
271 {
272 return write_object_file_flags(buf, len, type, oid, 0);
273 }
274
275 int write_object_file_literally(const void *buf, unsigned long len,
276 const char *type, struct object_id *oid,
277 unsigned flags);
278 int stream_loose_object(struct input_stream *in_stream, size_t len,
279 struct object_id *oid);
280
281 /*
282 * Add an object file to the in-memory object store, without writing it
283 * to disk.
284 *
285 * Callers are responsible for calling write_object_file to record the
286 * object in persistent storage before writing any other new objects
287 * that reference it.
288 */
289 int pretend_object_file(void *, unsigned long, enum object_type,
290 struct object_id *oid);
291
292 int force_object_loose(const struct object_id *oid, time_t mtime);
293
294 /*
295 * Open the loose object at path, check its hash, and return the contents,
296 * use the "oi" argument to assert things about the object, or e.g. populate its
297 * type, and size. If the object is a blob, then "contents" may return NULL,
298 * to allow streaming of large blobs.
299 *
300 * Returns 0 on success, negative on error (details may be written to stderr).
301 */
302 int read_loose_object(const char *path,
303 const struct object_id *expected_oid,
304 struct object_id *real_oid,
305 void **contents,
306 struct object_info *oi);
307
308 /* Retry packed storage after checking packed and loose storage */
309 #define HAS_OBJECT_RECHECK_PACKED 1
310
311 /*
312 * Returns 1 if the object exists. This function will not lazily fetch objects
313 * in a partial clone.
314 */
315 int has_object(struct repository *r, const struct object_id *oid,
316 unsigned flags);
317
318 /*
319 * These macros and functions are deprecated. If checking existence for an
320 * object that is likely to be missing and/or whose absence is relatively
321 * inconsequential (or is consequential but the caller is prepared to handle
322 * it), use has_object(), which has better defaults (no lazy fetch in a partial
323 * clone and no rechecking of packed storage). In the unlikely event that a
324 * caller needs to assert existence of an object that it fully expects to
325 * exist, and wants to trigger a lazy fetch in a partial clone, use
326 * oid_object_info_extended() with a NULL struct object_info.
327 *
328 * These functions can be removed once all callers have migrated to
329 * has_object() and/or oid_object_info_extended().
330 */
331 int repo_has_object_file(struct repository *r, const struct object_id *oid);
332 int repo_has_object_file_with_flags(struct repository *r,
333 const struct object_id *oid, int flags);
334 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
335 #define has_object_file(oid) repo_has_object_file(the_repository, oid)
336 #define has_object_file_with_flags(oid, flags) repo_has_object_file_with_flags(the_repository, oid, flags)
337 #endif
338
339 /*
340 * Return true iff an alternate object database has a loose object
341 * with the specified name. This function does not respect replace
342 * references.
343 */
344 int has_loose_object_nonlocal(const struct object_id *);
345
346 int has_loose_object(const struct object_id *);
347
348 /**
349 * format_object_header() is a thin wrapper around s xsnprintf() that
350 * writes the initial "<type> <obj-len>" part of the loose object
351 * header. It returns the size that snprintf() returns + 1.
352 */
353 int format_object_header(char *str, size_t size, enum object_type type,
354 size_t objsize);
355
356 void assert_oid_type(const struct object_id *oid, enum object_type expect);
357
358 /*
359 * Enabling the object read lock allows multiple threads to safely call the
360 * following functions in parallel: repo_read_object_file(), read_object_file(),
361 * read_object_file_extended(), read_object_with_reference(), read_object(),
362 * oid_object_info() and oid_object_info_extended().
363 *
364 * obj_read_lock() and obj_read_unlock() may also be used to protect other
365 * section which cannot execute in parallel with object reading. Since the used
366 * lock is a recursive mutex, these sections can even contain calls to object
367 * reading functions. However, beware that in these cases zlib inflation won't
368 * be performed in parallel, losing performance.
369 *
370 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
371 * any of its callees end up calling it, this recursive call won't benefit from
372 * parallel inflation.
373 */
374 void enable_obj_read_lock(void);
375 void disable_obj_read_lock(void);
376
377 extern int obj_read_use_lock;
378 extern pthread_mutex_t obj_read_mutex;
379
380 static inline void obj_read_lock(void)
381 {
382 if(obj_read_use_lock)
383 pthread_mutex_lock(&obj_read_mutex);
384 }
385
386 static inline void obj_read_unlock(void)
387 {
388 if(obj_read_use_lock)
389 pthread_mutex_unlock(&obj_read_mutex);
390 }
391
392 struct object_info {
393 /* Request */
394 enum object_type *typep;
395 unsigned long *sizep;
396 off_t *disk_sizep;
397 struct object_id *delta_base_oid;
398 struct strbuf *type_name;
399 void **contentp;
400
401 /* Response */
402 enum {
403 OI_CACHED,
404 OI_LOOSE,
405 OI_PACKED,
406 OI_DBCACHED
407 } whence;
408 union {
409 /*
410 * struct {
411 * ... Nothing to expose in this case
412 * } cached;
413 * struct {
414 * ... Nothing to expose in this case
415 * } loose;
416 */
417 struct {
418 struct packed_git *pack;
419 off_t offset;
420 unsigned int is_delta;
421 } packed;
422 } u;
423 };
424
425 /*
426 * Initializer for a "struct object_info" that wants no items. You may
427 * also memset() the memory to all-zeroes.
428 */
429 #define OBJECT_INFO_INIT { 0 }
430
431 /* Invoke lookup_replace_object() on the given hash */
432 #define OBJECT_INFO_LOOKUP_REPLACE 1
433 /* Allow reading from a loose object file of unknown/bogus type */
434 #define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
435 /* Do not retry packed storage after checking packed and loose storage */
436 #define OBJECT_INFO_QUICK 8
437 /* Do not check loose object */
438 #define OBJECT_INFO_IGNORE_LOOSE 16
439 /*
440 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
441 * nonzero).
442 */
443 #define OBJECT_INFO_SKIP_FETCH_OBJECT 32
444 /*
445 * This is meant for bulk prefetching of missing blobs in a partial
446 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
447 */
448 #define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
449
450 int oid_object_info_extended(struct repository *r,
451 const struct object_id *,
452 struct object_info *, unsigned flags);
453
454 /*
455 * Iterate over the files in the loose-object parts of the object
456 * directory "path", triggering the following callbacks:
457 *
458 * - loose_object is called for each loose object we find.
459 *
460 * - loose_cruft is called for any files that do not appear to be
461 * loose objects. Note that we only look in the loose object
462 * directories "objects/[0-9a-f]{2}/", so we will not report
463 * "objects/foobar" as cruft.
464 *
465 * - loose_subdir is called for each top-level hashed subdirectory
466 * of the object directory (e.g., "$OBJDIR/f0"). It is called
467 * after the objects in the directory are processed.
468 *
469 * Any callback that is NULL will be ignored. Callbacks returning non-zero
470 * will end the iteration.
471 *
472 * In the "buf" variant, "path" is a strbuf which will also be used as a
473 * scratch buffer, but restored to its original contents before
474 * the function returns.
475 */
476 typedef int each_loose_object_fn(const struct object_id *oid,
477 const char *path,
478 void *data);
479 typedef int each_loose_cruft_fn(const char *basename,
480 const char *path,
481 void *data);
482 typedef int each_loose_subdir_fn(unsigned int nr,
483 const char *path,
484 void *data);
485 int for_each_file_in_obj_subdir(unsigned int subdir_nr,
486 struct strbuf *path,
487 each_loose_object_fn obj_cb,
488 each_loose_cruft_fn cruft_cb,
489 each_loose_subdir_fn subdir_cb,
490 void *data);
491 int for_each_loose_file_in_objdir(const char *path,
492 each_loose_object_fn obj_cb,
493 each_loose_cruft_fn cruft_cb,
494 each_loose_subdir_fn subdir_cb,
495 void *data);
496 int for_each_loose_file_in_objdir_buf(struct strbuf *path,
497 each_loose_object_fn obj_cb,
498 each_loose_cruft_fn cruft_cb,
499 each_loose_subdir_fn subdir_cb,
500 void *data);
501
502 /* Flags for for_each_*_object() below. */
503 enum for_each_object_flags {
504 /* Iterate only over local objects, not alternates. */
505 FOR_EACH_OBJECT_LOCAL_ONLY = (1<<0),
506
507 /* Only iterate over packs obtained from the promisor remote. */
508 FOR_EACH_OBJECT_PROMISOR_ONLY = (1<<1),
509
510 /*
511 * Visit objects within a pack in packfile order rather than .idx order
512 */
513 FOR_EACH_OBJECT_PACK_ORDER = (1<<2),
514
515 /* Only iterate over packs that are not marked as kept in-core. */
516 FOR_EACH_OBJECT_SKIP_IN_CORE_KEPT_PACKS = (1<<3),
517
518 /* Only iterate over packs that do not have .keep files. */
519 FOR_EACH_OBJECT_SKIP_ON_DISK_KEPT_PACKS = (1<<4),
520 };
521
522 /*
523 * Iterate over all accessible loose objects without respect to
524 * reachability. By default, this includes both local and alternate objects.
525 * The order in which objects are visited is unspecified.
526 *
527 * Any flags specific to packs are ignored.
528 */
529 int for_each_loose_object(each_loose_object_fn, void *,
530 enum for_each_object_flags flags);
531
532 /*
533 * Iterate over all accessible packed objects without respect to reachability.
534 * By default, this includes both local and alternate packs.
535 *
536 * Note that some objects may appear twice if they are found in multiple packs.
537 * Each pack is visited in an unspecified order. By default, objects within a
538 * pack are visited in pack-idx order (i.e., sorted by oid).
539 */
540 typedef int each_packed_object_fn(const struct object_id *oid,
541 struct packed_git *pack,
542 uint32_t pos,
543 void *data);
544 int for_each_object_in_pack(struct packed_git *p,
545 each_packed_object_fn, void *data,
546 enum for_each_object_flags flags);
547 int for_each_packed_object(each_packed_object_fn, void *,
548 enum for_each_object_flags flags);
549
550 #endif /* OBJECT_STORE_H */