]> git.ipfire.org Git - thirdparty/git.git/blame - object-store.h
Sync with 2.35.8
[thirdparty/git.git] / object-store.h
CommitLineData
90c62155
SB
1#ifndef OBJECT_STORE_H
2#define OBJECT_STORE_H
3
ef3ca954 4#include "cache.h"
d88f9fdf 5#include "oidmap.h"
14727b7f 6#include "list.h"
fe299ec5 7#include "oid-array.h"
14727b7f 8#include "strbuf.h"
31877c9a 9#include "thread-utils.h"
cf2dc1c2
EW
10#include "khash.h"
11#include "dir.h"
92d8ed8a 12#include "oidtree.h"
09ef6617 13#include "oidset.h"
d88f9fdf 14
263db403
JK
15struct object_directory {
16 struct object_directory *next;
0d4a1321 17
0d4a1321 18 /*
3a2e0824
JK
19 * Used to store the results of readdir(3) calls when we are OK
20 * sacrificing accuracy due to races for speed. That includes
61c7711c 21 * object existence with OBJECT_INFO_QUICK, as well as
3a2e0824
JK
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.
0d4a1321 26 */
33f379ee 27 uint32_t loose_objects_subdir_seen[8]; /* 256 bits */
92d8ed8a 28 struct oidtree *loose_objects_cache;
0d4a1321 29
ecd81dfc
NS
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
b3cecf49
NS
37 /*
38 * This object store is ephemeral, so there is no need to fsync.
39 */
40 int will_destroy;
41
77f012e8
SB
42 /*
43 * Path to the alternative object store. If this is a relative path,
44 * it is relative to the current working directory.
45 */
f0eaf638 46 char *path;
031dc927 47};
f0eaf638 48
cf2dc1c2 49KHASH_INIT(odb_path_map, const char * /* key: odb_path */,
dd3c8a72 50 struct object_directory *, 1, fspathhash, fspatheq)
cf2dc1c2 51
13068bf0 52void prepare_alt_odb(struct repository *r);
0d4a1321 53char *compute_alternate_path(const char *path, struct strbuf *err);
f57a7396 54struct object_directory *find_odb(struct repository *r, const char *obj_dir);
263db403 55typedef int alt_odb_fn(struct object_directory *, void *);
0d4a1321 56int foreach_alt_odb(alt_odb_fn, void*);
709dfa69
JK
57typedef void alternate_ref_fn(const struct object_id *oid, void *);
58void for_each_alternate_ref(alternate_ref_fn, void *);
0d4a1321 59
0d4a1321
SB
60/*
61 * Add the directory to the on-disk alternates file; the new entry will also
62 * take effect in the current process.
63 */
64void add_to_alternates_file(const char *dir);
65
66/*
67 * Add the directory to the in-memory list of alternates (along with any
68 * recursive alternates it points to), but do not modify the on-disk alternates
69 * file.
70 */
71void add_to_alternates_memory(const char *dir);
72
b3cecf49
NS
73/*
74 * Replace the current writable object directory with the specified temporary
75 * object directory; returns the former primary object directory.
76 */
77struct object_directory *set_temporary_primary_odb(const char *dir, int will_destroy);
78
79/*
80 * Restore a previous ODB replaced by set_temporary_main_odb.
81 */
82void restore_primary_odb(struct object_directory *restore_odb, const char *old_path);
83
0000d654
RS
84/*
85 * Populate and return the loose object cache array corresponding to the
86 * given object ID.
87 */
92d8ed8a 88struct oidtree *odb_loose_cache(struct object_directory *odb,
0000d654
RS
89 const struct object_id *oid);
90
d4e19e51
RS
91/* Empty the loose object cache for the specified object directory. */
92void odb_clear_loose_cache(struct object_directory *odb);
93
b3cecf49
NS
94/* Clear and free the specified object directory */
95void free_object_directory(struct object_directory *odb);
96
a80d72db 97struct packed_git {
ec48540f 98 struct hashmap_entry packmap_ent;
a80d72db
SB
99 struct packed_git *next;
100 struct list_head mru;
101 struct pack_window *windows;
102 off_t pack_size;
103 const void *index_data;
104 size_t index_size;
105 uint32_t num_objects;
629dffc4 106 uint32_t crc_offset;
09ef6617 107 struct oidset bad_objects;
a80d72db
SB
108 int index_version;
109 time_t mtime;
110 int pack_fd;
43fa44fa 111 int index; /* for builtin/pack-objects.c */
a80d72db
SB
112 unsigned pack_local:1,
113 pack_keep:1,
ed7e5fc3 114 pack_keep_in_core:1,
a80d72db
SB
115 freshened:1,
116 do_not_close:1,
af96fe33
DS
117 pack_promisor:1,
118 multi_pack_index:1;
538b1523 119 unsigned char hash[GIT_MAX_RAWSZ];
a80d72db 120 struct revindex_entry *revindex;
2f4ba2a8
TB
121 const uint32_t *revindex_data;
122 const uint32_t *revindex_map;
123 size_t revindex_size;
a80d72db
SB
124 /* something like ".git/objects/pack/xxxxx.pack" */
125 char pack_name[FLEX_ARRAY]; /* more */
126};
127
4d80560c
DS
128struct multi_pack_index;
129
ec48540f
CS
130static inline int pack_map_entry_cmp(const void *unused_cmp_data,
131 const struct hashmap_entry *entry,
132 const struct hashmap_entry *entry2,
133 const void *keydata)
134{
135 const char *key = keydata;
136 const struct packed_git *pg1, *pg2;
137
138 pg1 = container_of(entry, const struct packed_git, packmap_ent);
139 pg2 = container_of(entry2, const struct packed_git, packmap_ent);
140
141 return strcmp(pg1->pack_name, key ? key : pg2->pack_name);
142}
143
90c62155
SB
144struct raw_object_store {
145 /*
f0eaf638
JK
146 * Set of all object directories; the main directory is first (and
147 * cannot be NULL after initialization). Subsequent directories are
148 * alternates.
90c62155 149 */
f0eaf638
JK
150 struct object_directory *odb;
151 struct object_directory **odb_tail;
cf2dc1c2
EW
152 kh_odb_path_map_t *odb_by_path;
153
f0eaf638 154 int loaded_alternates;
90c62155 155
f0eaf638
JK
156 /*
157 * A list of alternate object directories loaded from the environment;
158 * this should not generally need to be accessed directly, but will
159 * populate the "odb" list when prepare_alt_odb() is run.
160 */
90c62155 161 char *alternate_db;
031dc927 162
d88f9fdf
SB
163 /*
164 * Objects that should be substituted by other objects
165 * (see git-replace(1)).
166 */
c1274495 167 struct oidmap *replace_map;
b1fc9da1
MT
168 unsigned replace_map_initialized : 1;
169 pthread_mutex_t replace_mutex; /* protect object replace functions */
d88f9fdf 170
85277506
JT
171 struct commit_graph *commit_graph;
172 unsigned commit_graph_attempted : 1; /* if loading has been attempted */
173
c4d25228
DS
174 /*
175 * private data
176 *
177 * should only be accessed directly by packfile.c and midx.c
178 */
179 struct multi_pack_index *multi_pack_index;
180
a80d72db
SB
181 /*
182 * private data
183 *
184 * should only be accessed directly by packfile.c
185 */
186
187 struct packed_git *packed_git;
188 /* A most-recently-used ordered version of the packed_git list. */
189 struct list_head packed_git_mru;
5508f693 190
20b031fe
JK
191 struct {
192 struct packed_git **packs;
193 unsigned flags;
194 } kept_pack_cache;
195
ec48540f
CS
196 /*
197 * A map of packfiles to packed_git structs for tracking which
198 * packs have been loaded already.
199 */
200 struct hashmap pack_map;
201
9a00580d
SB
202 /*
203 * A fast, rough count of the number of objects in the repository.
204 * These two fields are not meant for direct access. Use
205 * approximate_object_count() instead.
206 */
207 unsigned long approximate_object_count;
208 unsigned approximate_object_count_valid : 1;
209
5508f693
SB
210 /*
211 * Whether packed_git has already been populated with this repository's
212 * packs.
213 */
214 unsigned packed_git_initialized : 1;
90c62155
SB
215};
216
217struct raw_object_store *raw_object_store_new(void);
218void raw_object_store_clear(struct raw_object_store *o);
219
cf78ae4f
SB
220/*
221 * Put in `buf` the name of the file in the local object database that
514c5fdd 222 * would be used to store a loose object with the specified oid.
cf78ae4f 223 */
514c5fdd
JK
224const char *loose_object_path(struct repository *r, struct strbuf *buf,
225 const struct object_id *oid);
cf78ae4f 226
514c5fdd
JK
227void *map_loose_object(struct repository *r, const struct object_id *oid,
228 unsigned long *size);
e35454fa 229
55454427 230void *read_object_file_extended(struct repository *r,
ad6dad09
DL
231 const struct object_id *oid,
232 enum object_type *type,
233 unsigned long *size, int lookup_replace);
afd69dcc
SB
234static inline void *repo_read_object_file(struct repository *r,
235 const struct object_id *oid,
236 enum object_type *type,
237 unsigned long *size)
cbd53a21 238{
afd69dcc 239 return read_object_file_extended(r, oid, type, size, 1);
cbd53a21 240}
afd69dcc
SB
241#ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
242#define read_object_file(oid, type, size) repo_read_object_file(the_repository, oid, type, size)
243#endif
cbd53a21
SB
244
245/* Read and unpack an object file into memory, write memory to an object file */
246int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
247
63e05f90 248void hash_object_file(const struct git_hash_algo *algo, const void *buf,
44439c1c 249 unsigned long len, enum object_type type,
63e05f90 250 struct object_id *oid);
cbd53a21 251
4ef91a2d 252int write_object_file_flags(const void *buf, unsigned long len,
c80d226a 253 enum object_type type, struct object_id *oid,
4ef91a2d
ÆAB
254 unsigned flags);
255static inline int write_object_file(const void *buf, unsigned long len,
c80d226a 256 enum object_type type, struct object_id *oid)
4ef91a2d
ÆAB
257{
258 return write_object_file_flags(buf, len, type, oid, 0);
259}
cbd53a21 260
0ff7b4f9
ÆAB
261int write_object_file_literally(const void *buf, unsigned long len,
262 const char *type, struct object_id *oid,
263 unsigned flags);
cbd53a21 264
60440d72
JN
265/*
266 * Add an object file to the in-memory object store, without writing it
267 * to disk.
268 *
269 * Callers are responsible for calling write_object_file to record the
270 * object in persistent storage before writing any other new objects
271 * that reference it.
272 */
55454427 273int pretend_object_file(void *, unsigned long, enum object_type,
ad6dad09 274 struct object_id *oid);
cbd53a21 275
55454427 276int force_object_loose(const struct object_id *oid, time_t mtime);
cbd53a21
SB
277
278/*
279 * Open the loose object at path, check its hash, and return the contents,
31deb28f 280 * use the "oi" argument to assert things about the object, or e.g. populate its
cbd53a21
SB
281 * type, and size. If the object is a blob, then "contents" may return NULL,
282 * to allow streaming of large blobs.
283 *
284 * Returns 0 on success, negative on error (details may be written to stderr).
285 */
286int read_loose_object(const char *path,
287 const struct object_id *expected_oid,
96e41f58 288 struct object_id *real_oid,
31deb28f
ÆAB
289 void **contents,
290 struct object_info *oi);
cbd53a21 291
1d8d9cb6
JT
292/* Retry packed storage after checking packed and loose storage */
293#define HAS_OBJECT_RECHECK_PACKED 1
294
295/*
296 * Returns 1 if the object exists. This function will not lazily fetch objects
297 * in a partial clone.
298 */
299int has_object(struct repository *r, const struct object_id *oid,
300 unsigned flags);
301
302/*
303 * These macros and functions are deprecated. If checking existence for an
304 * object that is likely to be missing and/or whose absence is relatively
305 * inconsequential (or is consequential but the caller is prepared to handle
306 * it), use has_object(), which has better defaults (no lazy fetch in a partial
307 * clone and no rechecking of packed storage). In the unlikely event that a
308 * caller needs to assert existence of an object that it fully expects to
309 * exist, and wants to trigger a lazy fetch in a partial clone, use
310 * oid_object_info_extended() with a NULL struct object_info.
311 *
312 * These functions can be removed once all callers have migrated to
313 * has_object() and/or oid_object_info_extended().
314 */
9b45f499
SB
315#ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
316#define has_sha1_file_with_flags(sha1, flags) repo_has_sha1_file_with_flags(the_repository, sha1, flags)
317#define has_sha1_file(sha1) repo_has_sha1_file(the_repository, sha1)
318#endif
9b45f499
SB
319int repo_has_object_file(struct repository *r, const struct object_id *oid);
320int repo_has_object_file_with_flags(struct repository *r,
321 const struct object_id *oid, int flags);
322#ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
323#define has_object_file(oid) repo_has_object_file(the_repository, oid)
324#define has_object_file_with_flags(oid, flags) repo_has_object_file_with_flags(the_repository, oid, flags)
325#endif
cbd53a21
SB
326
327/*
328 * Return true iff an alternate object database has a loose object
329 * with the specified name. This function does not respect replace
330 * references.
331 */
55454427 332int has_loose_object_nonlocal(const struct object_id *);
cbd53a21 333
b04cdea4
ÆAB
334/**
335 * format_object_header() is a thin wrapper around s xsnprintf() that
336 * writes the initial "<type> <obj-len>" part of the loose object
337 * header. It returns the size that snprintf() returns + 1.
338 */
339int format_object_header(char *str, size_t size, enum object_type type,
340 size_t objsize);
341
55454427 342void assert_oid_type(const struct object_id *oid, enum object_type expect);
cbd53a21 343
31877c9a
MT
344/*
345 * Enabling the object read lock allows multiple threads to safely call the
346 * following functions in parallel: repo_read_object_file(), read_object_file(),
347 * read_object_file_extended(), read_object_with_reference(), read_object(),
348 * oid_object_info() and oid_object_info_extended().
349 *
350 * obj_read_lock() and obj_read_unlock() may also be used to protect other
351 * section which cannot execute in parallel with object reading. Since the used
352 * lock is a recursive mutex, these sections can even contain calls to object
353 * reading functions. However, beware that in these cases zlib inflation won't
354 * be performed in parallel, losing performance.
355 *
356 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
357 * any of its callees end up calling it, this recursive call won't benefit from
358 * parallel inflation.
359 */
360void enable_obj_read_lock(void);
361void disable_obj_read_lock(void);
362
363extern int obj_read_use_lock;
364extern pthread_mutex_t obj_read_mutex;
365
366static inline void obj_read_lock(void)
367{
368 if(obj_read_use_lock)
369 pthread_mutex_lock(&obj_read_mutex);
370}
371
372static inline void obj_read_unlock(void)
373{
374 if(obj_read_use_lock)
375 pthread_mutex_unlock(&obj_read_mutex);
376}
377
cbd53a21
SB
378struct object_info {
379 /* Request */
380 enum object_type *typep;
381 unsigned long *sizep;
382 off_t *disk_sizep;
b99b6bcc 383 struct object_id *delta_base_oid;
cbd53a21
SB
384 struct strbuf *type_name;
385 void **contentp;
386
387 /* Response */
388 enum {
389 OI_CACHED,
390 OI_LOOSE,
391 OI_PACKED,
392 OI_DBCACHED
393 } whence;
394 union {
395 /*
396 * struct {
397 * ... Nothing to expose in this case
398 * } cached;
399 * struct {
400 * ... Nothing to expose in this case
401 * } loose;
402 */
403 struct {
404 struct packed_git *pack;
405 off_t offset;
406 unsigned int is_delta;
407 } packed;
408 } u;
409};
410
411/*
412 * Initializer for a "struct object_info" that wants no items. You may
413 * also memset() the memory to all-zeroes.
414 */
9865b6e6 415#define OBJECT_INFO_INIT { 0 }
cbd53a21
SB
416
417/* Invoke lookup_replace_object() on the given hash */
418#define OBJECT_INFO_LOOKUP_REPLACE 1
419/* Allow reading from a loose object file of unknown/bogus type */
420#define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
cbd53a21
SB
421/* Do not retry packed storage after checking packed and loose storage */
422#define OBJECT_INFO_QUICK 8
423/* Do not check loose object */
424#define OBJECT_INFO_IGNORE_LOOSE 16
0f4a4fb1
JT
425/*
426 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
31f5256c 427 * nonzero).
0f4a4fb1 428 */
31f5256c
DS
429#define OBJECT_INFO_SKIP_FETCH_OBJECT 32
430/*
431 * This is meant for bulk prefetching of missing blobs in a partial
432 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
433 */
434#define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
cbd53a21
SB
435
436int oid_object_info_extended(struct repository *r,
437 const struct object_id *,
438 struct object_info *, unsigned flags);
439
0889aae1
JK
440/*
441 * Iterate over the files in the loose-object parts of the object
442 * directory "path", triggering the following callbacks:
443 *
444 * - loose_object is called for each loose object we find.
445 *
446 * - loose_cruft is called for any files that do not appear to be
447 * loose objects. Note that we only look in the loose object
448 * directories "objects/[0-9a-f]{2}/", so we will not report
449 * "objects/foobar" as cruft.
450 *
451 * - loose_subdir is called for each top-level hashed subdirectory
452 * of the object directory (e.g., "$OBJDIR/f0"). It is called
453 * after the objects in the directory are processed.
454 *
455 * Any callback that is NULL will be ignored. Callbacks returning non-zero
456 * will end the iteration.
457 *
458 * In the "buf" variant, "path" is a strbuf which will also be used as a
459 * scratch buffer, but restored to its original contents before
460 * the function returns.
461 */
462typedef int each_loose_object_fn(const struct object_id *oid,
463 const char *path,
464 void *data);
465typedef int each_loose_cruft_fn(const char *basename,
466 const char *path,
467 void *data);
468typedef int each_loose_subdir_fn(unsigned int nr,
469 const char *path,
470 void *data);
471int for_each_file_in_obj_subdir(unsigned int subdir_nr,
472 struct strbuf *path,
473 each_loose_object_fn obj_cb,
474 each_loose_cruft_fn cruft_cb,
475 each_loose_subdir_fn subdir_cb,
476 void *data);
477int for_each_loose_file_in_objdir(const char *path,
478 each_loose_object_fn obj_cb,
479 each_loose_cruft_fn cruft_cb,
480 each_loose_subdir_fn subdir_cb,
481 void *data);
482int for_each_loose_file_in_objdir_buf(struct strbuf *path,
483 each_loose_object_fn obj_cb,
484 each_loose_cruft_fn cruft_cb,
485 each_loose_subdir_fn subdir_cb,
486 void *data);
487
488/* Flags for for_each_*_object() below. */
489enum for_each_object_flags {
490 /* Iterate only over local objects, not alternates. */
491 FOR_EACH_OBJECT_LOCAL_ONLY = (1<<0),
492
493 /* Only iterate over packs obtained from the promisor remote. */
494 FOR_EACH_OBJECT_PROMISOR_ONLY = (1<<1),
495
496 /*
497 * Visit objects within a pack in packfile order rather than .idx order
498 */
499 FOR_EACH_OBJECT_PACK_ORDER = (1<<2),
a241878a
TB
500
501 /* Only iterate over packs that are not marked as kept in-core. */
502 FOR_EACH_OBJECT_SKIP_IN_CORE_KEPT_PACKS = (1<<3),
503
504 /* Only iterate over packs that do not have .keep files. */
505 FOR_EACH_OBJECT_SKIP_ON_DISK_KEPT_PACKS = (1<<4),
0889aae1
JK
506};
507
508/*
509 * Iterate over all accessible loose objects without respect to
510 * reachability. By default, this includes both local and alternate objects.
511 * The order in which objects are visited is unspecified.
512 *
513 * Any flags specific to packs are ignored.
514 */
515int for_each_loose_object(each_loose_object_fn, void *,
516 enum for_each_object_flags flags);
517
518/*
519 * Iterate over all accessible packed objects without respect to reachability.
520 * By default, this includes both local and alternate packs.
521 *
522 * Note that some objects may appear twice if they are found in multiple packs.
523 * Each pack is visited in an unspecified order. By default, objects within a
524 * pack are visited in pack-idx order (i.e., sorted by oid).
525 */
526typedef int each_packed_object_fn(const struct object_id *oid,
527 struct packed_git *pack,
528 uint32_t pos,
529 void *data);
530int for_each_object_in_pack(struct packed_git *p,
531 each_packed_object_fn, void *data,
532 enum for_each_object_flags flags);
533int for_each_packed_object(each_packed_object_fn, void *,
534 enum for_each_object_flags flags);
535
90c62155 536#endif /* OBJECT_STORE_H */