]> git.ipfire.org Git - thirdparty/git.git/blame - object-store.h
sparse-index.h: move declarations for sparse-index.c from cache.h
[thirdparty/git.git] / object-store.h
CommitLineData
90c62155
SB
1#ifndef OBJECT_STORE_H
2#define OBJECT_STORE_H
3
1c028400 4#include "object.h"
14727b7f 5#include "list.h"
31877c9a 6#include "thread-utils.h"
cf2dc1c2
EW
7#include "khash.h"
8#include "dir.h"
09ef6617 9#include "oidset.h"
d88f9fdf 10
e3d2f20e
EN
11struct oidmap;
12struct oidtree;
13struct strbuf;
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
2b6070ac
HX
49struct input_stream {
50 const void *(*read)(struct input_stream *, unsigned long *len);
51 void *data;
52 int is_finished;
53};
54
cf2dc1c2 55KHASH_INIT(odb_path_map, const char * /* key: odb_path */,
dd3c8a72 56 struct object_directory *, 1, fspathhash, fspatheq)
cf2dc1c2 57
13068bf0 58void prepare_alt_odb(struct repository *r);
d85cd187 59int has_alt_odb(struct repository *r);
0d4a1321 60char *compute_alternate_path(const char *path, struct strbuf *err);
f57a7396 61struct object_directory *find_odb(struct repository *r, const char *obj_dir);
263db403 62typedef int alt_odb_fn(struct object_directory *, void *);
0d4a1321 63int foreach_alt_odb(alt_odb_fn, void*);
709dfa69
JK
64typedef void alternate_ref_fn(const struct object_id *oid, void *);
65void for_each_alternate_ref(alternate_ref_fn, void *);
0d4a1321 66
0d4a1321
SB
67/*
68 * Add the directory to the on-disk alternates file; the new entry will also
69 * take effect in the current process.
70 */
71void add_to_alternates_file(const char *dir);
72
73/*
74 * Add the directory to the in-memory list of alternates (along with any
75 * recursive alternates it points to), but do not modify the on-disk alternates
76 * file.
77 */
78void add_to_alternates_memory(const char *dir);
79
b3cecf49
NS
80/*
81 * Replace the current writable object directory with the specified temporary
82 * object directory; returns the former primary object directory.
83 */
84struct object_directory *set_temporary_primary_odb(const char *dir, int will_destroy);
85
86/*
87 * Restore a previous ODB replaced by set_temporary_main_odb.
88 */
89void restore_primary_odb(struct object_directory *restore_odb, const char *old_path);
90
0000d654
RS
91/*
92 * Populate and return the loose object cache array corresponding to the
93 * given object ID.
94 */
92d8ed8a 95struct oidtree *odb_loose_cache(struct object_directory *odb,
0000d654
RS
96 const struct object_id *oid);
97
d4e19e51
RS
98/* Empty the loose object cache for the specified object directory. */
99void odb_clear_loose_cache(struct object_directory *odb);
100
b3cecf49
NS
101/* Clear and free the specified object directory */
102void free_object_directory(struct object_directory *odb);
103
a80d72db 104struct packed_git {
ec48540f 105 struct hashmap_entry packmap_ent;
a80d72db
SB
106 struct packed_git *next;
107 struct list_head mru;
108 struct pack_window *windows;
109 off_t pack_size;
110 const void *index_data;
111 size_t index_size;
112 uint32_t num_objects;
629dffc4 113 uint32_t crc_offset;
09ef6617 114 struct oidset bad_objects;
a80d72db
SB
115 int index_version;
116 time_t mtime;
117 int pack_fd;
43fa44fa 118 int index; /* for builtin/pack-objects.c */
a80d72db
SB
119 unsigned pack_local:1,
120 pack_keep:1,
ed7e5fc3 121 pack_keep_in_core:1,
a80d72db
SB
122 freshened:1,
123 do_not_close:1,
af96fe33 124 pack_promisor:1,
94cd775a
TB
125 multi_pack_index:1,
126 is_cruft:1;
538b1523 127 unsigned char hash[GIT_MAX_RAWSZ];
a80d72db 128 struct revindex_entry *revindex;
2f4ba2a8
TB
129 const uint32_t *revindex_data;
130 const uint32_t *revindex_map;
131 size_t revindex_size;
94cd775a
TB
132 /*
133 * mtimes_map points at the beginning of the memory mapped region of
134 * this pack's corresponding .mtimes file, and mtimes_size is the size
135 * of that .mtimes file
136 */
137 const uint32_t *mtimes_map;
138 size_t mtimes_size;
a80d72db
SB
139 /* something like ".git/objects/pack/xxxxx.pack" */
140 char pack_name[FLEX_ARRAY]; /* more */
141};
142
4d80560c
DS
143struct multi_pack_index;
144
5cf88fd8 145static inline int pack_map_entry_cmp(const void *cmp_data UNUSED,
ec48540f
CS
146 const struct hashmap_entry *entry,
147 const struct hashmap_entry *entry2,
148 const void *keydata)
149{
150 const char *key = keydata;
151 const struct packed_git *pg1, *pg2;
152
153 pg1 = container_of(entry, const struct packed_git, packmap_ent);
154 pg2 = container_of(entry2, const struct packed_git, packmap_ent);
155
156 return strcmp(pg1->pack_name, key ? key : pg2->pack_name);
157}
158
90c62155
SB
159struct raw_object_store {
160 /*
f0eaf638
JK
161 * Set of all object directories; the main directory is first (and
162 * cannot be NULL after initialization). Subsequent directories are
163 * alternates.
90c62155 164 */
f0eaf638
JK
165 struct object_directory *odb;
166 struct object_directory **odb_tail;
cf2dc1c2
EW
167 kh_odb_path_map_t *odb_by_path;
168
f0eaf638 169 int loaded_alternates;
90c62155 170
f0eaf638
JK
171 /*
172 * A list of alternate object directories loaded from the environment;
173 * this should not generally need to be accessed directly, but will
174 * populate the "odb" list when prepare_alt_odb() is run.
175 */
90c62155 176 char *alternate_db;
031dc927 177
d88f9fdf
SB
178 /*
179 * Objects that should be substituted by other objects
180 * (see git-replace(1)).
181 */
c1274495 182 struct oidmap *replace_map;
b1fc9da1
MT
183 unsigned replace_map_initialized : 1;
184 pthread_mutex_t replace_mutex; /* protect object replace functions */
d88f9fdf 185
85277506
JT
186 struct commit_graph *commit_graph;
187 unsigned commit_graph_attempted : 1; /* if loading has been attempted */
188
c4d25228
DS
189 /*
190 * private data
191 *
192 * should only be accessed directly by packfile.c and midx.c
193 */
194 struct multi_pack_index *multi_pack_index;
195
a80d72db
SB
196 /*
197 * private data
198 *
199 * should only be accessed directly by packfile.c
200 */
201
202 struct packed_git *packed_git;
203 /* A most-recently-used ordered version of the packed_git list. */
204 struct list_head packed_git_mru;
5508f693 205
20b031fe
JK
206 struct {
207 struct packed_git **packs;
208 unsigned flags;
209 } kept_pack_cache;
210
ec48540f
CS
211 /*
212 * A map of packfiles to packed_git structs for tracking which
213 * packs have been loaded already.
214 */
215 struct hashmap pack_map;
216
9a00580d
SB
217 /*
218 * A fast, rough count of the number of objects in the repository.
219 * These two fields are not meant for direct access. Use
c7c33f50 220 * repo_approximate_object_count() instead.
9a00580d
SB
221 */
222 unsigned long approximate_object_count;
223 unsigned approximate_object_count_valid : 1;
224
5508f693
SB
225 /*
226 * Whether packed_git has already been populated with this repository's
227 * packs.
228 */
229 unsigned packed_git_initialized : 1;
90c62155
SB
230};
231
232struct raw_object_store *raw_object_store_new(void);
233void raw_object_store_clear(struct raw_object_store *o);
234
cf78ae4f
SB
235/*
236 * Put in `buf` the name of the file in the local object database that
514c5fdd 237 * would be used to store a loose object with the specified oid.
cf78ae4f 238 */
514c5fdd
JK
239const char *loose_object_path(struct repository *r, struct strbuf *buf,
240 const struct object_id *oid);
cf78ae4f 241
514c5fdd
JK
242void *map_loose_object(struct repository *r, const struct object_id *oid,
243 unsigned long *size);
e35454fa 244
0ba05cf2
JK
245void *repo_read_object_file(struct repository *r,
246 const struct object_id *oid,
247 enum object_type *type,
248 unsigned long *size);
cbd53a21
SB
249
250/* Read and unpack an object file into memory, write memory to an object file */
251int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
252
63e05f90 253void hash_object_file(const struct git_hash_algo *algo, const void *buf,
44439c1c 254 unsigned long len, enum object_type type,
63e05f90 255 struct object_id *oid);
cbd53a21 256
4ef91a2d 257int write_object_file_flags(const void *buf, unsigned long len,
c80d226a 258 enum object_type type, struct object_id *oid,
4ef91a2d
ÆAB
259 unsigned flags);
260static inline int write_object_file(const void *buf, unsigned long len,
c80d226a 261 enum object_type type, struct object_id *oid)
4ef91a2d
ÆAB
262{
263 return write_object_file_flags(buf, len, type, oid, 0);
264}
cbd53a21 265
0ff7b4f9
ÆAB
266int write_object_file_literally(const void *buf, unsigned long len,
267 const char *type, struct object_id *oid,
268 unsigned flags);
2b6070ac
HX
269int stream_loose_object(struct input_stream *in_stream, size_t len,
270 struct object_id *oid);
cbd53a21 271
60440d72
JN
272/*
273 * Add an object file to the in-memory object store, without writing it
274 * to disk.
275 *
276 * Callers are responsible for calling write_object_file to record the
277 * object in persistent storage before writing any other new objects
278 * that reference it.
279 */
55454427 280int pretend_object_file(void *, unsigned long, enum object_type,
ad6dad09 281 struct object_id *oid);
cbd53a21 282
55454427 283int force_object_loose(const struct object_id *oid, time_t mtime);
cbd53a21 284
1c028400
EN
285struct object_info {
286 /* Request */
287 enum object_type *typep;
288 unsigned long *sizep;
289 off_t *disk_sizep;
290 struct object_id *delta_base_oid;
291 struct strbuf *type_name;
292 void **contentp;
293
294 /* Response */
295 enum {
296 OI_CACHED,
297 OI_LOOSE,
298 OI_PACKED,
299 OI_DBCACHED
300 } whence;
301 union {
302 /*
303 * struct {
304 * ... Nothing to expose in this case
305 * } cached;
306 * struct {
307 * ... Nothing to expose in this case
308 * } loose;
309 */
310 struct {
311 struct packed_git *pack;
312 off_t offset;
313 unsigned int is_delta;
314 } packed;
315 } u;
316};
317
318/*
319 * Initializer for a "struct object_info" that wants no items. You may
320 * also memset() the memory to all-zeroes.
321 */
322#define OBJECT_INFO_INIT { 0 }
323
324/* Invoke lookup_replace_object() on the given hash */
325#define OBJECT_INFO_LOOKUP_REPLACE 1
326/* Allow reading from a loose object file of unknown/bogus type */
327#define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
328/* Do not retry packed storage after checking packed and loose storage */
329#define OBJECT_INFO_QUICK 8
330/*
331 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
332 * nonzero).
333 */
334#define OBJECT_INFO_SKIP_FETCH_OBJECT 16
335/*
336 * This is meant for bulk prefetching of missing blobs in a partial
337 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
338 */
339#define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
340
341/* Die if object corruption (not just an object being missing) was detected. */
342#define OBJECT_INFO_DIE_IF_CORRUPT 32
343
344int oid_object_info_extended(struct repository *r,
345 const struct object_id *,
346 struct object_info *, unsigned flags);
347
cbd53a21
SB
348/*
349 * Open the loose object at path, check its hash, and return the contents,
31deb28f 350 * use the "oi" argument to assert things about the object, or e.g. populate its
cbd53a21
SB
351 * type, and size. If the object is a blob, then "contents" may return NULL,
352 * to allow streaming of large blobs.
353 *
354 * Returns 0 on success, negative on error (details may be written to stderr).
355 */
356int read_loose_object(const char *path,
357 const struct object_id *expected_oid,
96e41f58 358 struct object_id *real_oid,
31deb28f
ÆAB
359 void **contents,
360 struct object_info *oi);
cbd53a21 361
1d8d9cb6
JT
362/* Retry packed storage after checking packed and loose storage */
363#define HAS_OBJECT_RECHECK_PACKED 1
364
365/*
366 * Returns 1 if the object exists. This function will not lazily fetch objects
367 * in a partial clone.
368 */
369int has_object(struct repository *r, const struct object_id *oid,
370 unsigned flags);
371
372/*
373 * These macros and functions are deprecated. If checking existence for an
374 * object that is likely to be missing and/or whose absence is relatively
375 * inconsequential (or is consequential but the caller is prepared to handle
376 * it), use has_object(), which has better defaults (no lazy fetch in a partial
377 * clone and no rechecking of packed storage). In the unlikely event that a
378 * caller needs to assert existence of an object that it fully expects to
379 * exist, and wants to trigger a lazy fetch in a partial clone, use
380 * oid_object_info_extended() with a NULL struct object_info.
381 *
382 * These functions can be removed once all callers have migrated to
383 * has_object() and/or oid_object_info_extended().
384 */
9b45f499
SB
385int repo_has_object_file(struct repository *r, const struct object_id *oid);
386int repo_has_object_file_with_flags(struct repository *r,
387 const struct object_id *oid, int flags);
cbd53a21
SB
388
389/*
390 * Return true iff an alternate object database has a loose object
391 * with the specified name. This function does not respect replace
392 * references.
393 */
55454427 394int has_loose_object_nonlocal(const struct object_id *);
cbd53a21 395
b7573536
TB
396int has_loose_object(const struct object_id *);
397
b04cdea4
ÆAB
398/**
399 * format_object_header() is a thin wrapper around s xsnprintf() that
400 * writes the initial "<type> <obj-len>" part of the loose object
401 * header. It returns the size that snprintf() returns + 1.
402 */
403int format_object_header(char *str, size_t size, enum object_type type,
404 size_t objsize);
405
55454427 406void assert_oid_type(const struct object_id *oid, enum object_type expect);
cbd53a21 407
31877c9a
MT
408/*
409 * Enabling the object read lock allows multiple threads to safely call the
c7c33f50 410 * following functions in parallel: repo_read_object_file(),
0ba05cf2 411 * read_object_with_reference(), oid_object_info() and oid_object_info_extended().
31877c9a
MT
412 *
413 * obj_read_lock() and obj_read_unlock() may also be used to protect other
414 * section which cannot execute in parallel with object reading. Since the used
415 * lock is a recursive mutex, these sections can even contain calls to object
416 * reading functions. However, beware that in these cases zlib inflation won't
417 * be performed in parallel, losing performance.
418 *
419 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
420 * any of its callees end up calling it, this recursive call won't benefit from
421 * parallel inflation.
422 */
423void enable_obj_read_lock(void);
424void disable_obj_read_lock(void);
425
426extern int obj_read_use_lock;
427extern pthread_mutex_t obj_read_mutex;
428
429static inline void obj_read_lock(void)
430{
431 if(obj_read_use_lock)
432 pthread_mutex_lock(&obj_read_mutex);
433}
434
435static inline void obj_read_unlock(void)
436{
437 if(obj_read_use_lock)
438 pthread_mutex_unlock(&obj_read_mutex);
439}
440
0889aae1
JK
441/*
442 * Iterate over the files in the loose-object parts of the object
443 * directory "path", triggering the following callbacks:
444 *
445 * - loose_object is called for each loose object we find.
446 *
447 * - loose_cruft is called for any files that do not appear to be
448 * loose objects. Note that we only look in the loose object
449 * directories "objects/[0-9a-f]{2}/", so we will not report
450 * "objects/foobar" as cruft.
451 *
452 * - loose_subdir is called for each top-level hashed subdirectory
453 * of the object directory (e.g., "$OBJDIR/f0"). It is called
454 * after the objects in the directory are processed.
455 *
456 * Any callback that is NULL will be ignored. Callbacks returning non-zero
457 * will end the iteration.
458 *
459 * In the "buf" variant, "path" is a strbuf which will also be used as a
460 * scratch buffer, but restored to its original contents before
461 * the function returns.
462 */
463typedef int each_loose_object_fn(const struct object_id *oid,
464 const char *path,
465 void *data);
466typedef int each_loose_cruft_fn(const char *basename,
467 const char *path,
468 void *data);
469typedef int each_loose_subdir_fn(unsigned int nr,
470 const char *path,
471 void *data);
472int for_each_file_in_obj_subdir(unsigned int subdir_nr,
473 struct strbuf *path,
474 each_loose_object_fn obj_cb,
475 each_loose_cruft_fn cruft_cb,
476 each_loose_subdir_fn subdir_cb,
477 void *data);
478int for_each_loose_file_in_objdir(const char *path,
479 each_loose_object_fn obj_cb,
480 each_loose_cruft_fn cruft_cb,
481 each_loose_subdir_fn subdir_cb,
482 void *data);
483int for_each_loose_file_in_objdir_buf(struct strbuf *path,
484 each_loose_object_fn obj_cb,
485 each_loose_cruft_fn cruft_cb,
486 each_loose_subdir_fn subdir_cb,
487 void *data);
488
489/* Flags for for_each_*_object() below. */
490enum for_each_object_flags {
491 /* Iterate only over local objects, not alternates. */
492 FOR_EACH_OBJECT_LOCAL_ONLY = (1<<0),
493
494 /* Only iterate over packs obtained from the promisor remote. */
495 FOR_EACH_OBJECT_PROMISOR_ONLY = (1<<1),
496
497 /*
498 * Visit objects within a pack in packfile order rather than .idx order
499 */
500 FOR_EACH_OBJECT_PACK_ORDER = (1<<2),
a241878a
TB
501
502 /* Only iterate over packs that are not marked as kept in-core. */
503 FOR_EACH_OBJECT_SKIP_IN_CORE_KEPT_PACKS = (1<<3),
504
505 /* Only iterate over packs that do not have .keep files. */
506 FOR_EACH_OBJECT_SKIP_ON_DISK_KEPT_PACKS = (1<<4),
0889aae1
JK
507};
508
509/*
510 * Iterate over all accessible loose objects without respect to
511 * reachability. By default, this includes both local and alternate objects.
512 * The order in which objects are visited is unspecified.
513 *
514 * Any flags specific to packs are ignored.
515 */
516int for_each_loose_object(each_loose_object_fn, void *,
517 enum for_each_object_flags flags);
518
519/*
520 * Iterate over all accessible packed objects without respect to reachability.
521 * By default, this includes both local and alternate packs.
522 *
523 * Note that some objects may appear twice if they are found in multiple packs.
524 * Each pack is visited in an unspecified order. By default, objects within a
525 * pack are visited in pack-idx order (i.e., sorted by oid).
526 */
527typedef int each_packed_object_fn(const struct object_id *oid,
528 struct packed_git *pack,
529 uint32_t pos,
530 void *data);
531int for_each_object_in_pack(struct packed_git *p,
532 each_packed_object_fn, void *data,
533 enum for_each_object_flags flags);
534int for_each_packed_object(each_packed_object_fn, void *,
535 enum for_each_object_flags flags);
536
90c62155 537#endif /* OBJECT_STORE_H */