]> git.ipfire.org Git - thirdparty/git.git/blame - object-store-ll.h
Start the 2.46 cycle
[thirdparty/git.git] / object-store-ll.h
CommitLineData
a034e910
EN
1#ifndef OBJECT_STORE_LL_H
2#define OBJECT_STORE_LL_H
3
4#include "hashmap.h"
5#include "object.h"
6#include "list.h"
7#include "thread-utils.h"
8#include "oidset.h"
9
10struct oidmap;
11struct oidtree;
12struct strbuf;
13
14struct object_directory {
15 struct object_directory *next;
16
17 /*
18 * Used to store the results of readdir(3) calls when we are OK
19 * sacrificing accuracy due to races for speed. That includes
20 * object existence with OBJECT_INFO_QUICK, as well as
21 * our search for unique abbreviated hashes. Don't use it for tasks
22 * requiring greater accuracy!
23 *
24 * Be sure to call odb_load_loose_cache() before using.
25 */
26 uint32_t loose_objects_subdir_seen[8]; /* 256 bits */
27 struct oidtree *loose_objects_cache;
28
23b2c7e9 29 /* Map between object IDs for loose objects. */
30 struct loose_object_map *loose_map;
31
a034e910
EN
32 /*
33 * This is a temporary object store created by the tmp_objdir
34 * facility. Disable ref updates since the objects in the store
35 * might be discarded on rollback.
36 */
37 int disable_ref_updates;
38
39 /*
40 * This object store is ephemeral, so there is no need to fsync.
41 */
42 int will_destroy;
43
44 /*
45 * Path to the alternative object store. If this is a relative path,
46 * it is relative to the current working directory.
47 */
48 char *path;
49};
50
51struct input_stream {
52 const void *(*read)(struct input_stream *, unsigned long *len);
53 void *data;
54 int is_finished;
55};
56
57void prepare_alt_odb(struct repository *r);
58int has_alt_odb(struct repository *r);
59char *compute_alternate_path(const char *path, struct strbuf *err);
60struct object_directory *find_odb(struct repository *r, const char *obj_dir);
61typedef int alt_odb_fn(struct object_directory *, void *);
62int foreach_alt_odb(alt_odb_fn, void*);
63typedef void alternate_ref_fn(const struct object_id *oid, void *);
64void 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 */
70void 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 */
77void 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 */
83struct 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 */
88void 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 */
94struct 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. */
98void odb_clear_loose_cache(struct object_directory *odb);
99
100/* Clear and free the specified object directory */
101void free_object_directory(struct object_directory *odb);
102
103struct 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;
4488bb3b 112 size_t crc_offset;
a034e910
EN
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
142struct multi_pack_index;
143
144static 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
158struct 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 struct kh_odb_path_map *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 * repo_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
231struct raw_object_store *raw_object_store_new(void);
232void 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 */
238const char *loose_object_path(struct repository *r, struct strbuf *buf,
239 const struct object_id *oid);
240
241void *map_loose_object(struct repository *r, const struct object_id *oid,
242 unsigned long *size);
243
244void *repo_read_object_file(struct repository *r,
245 const struct object_id *oid,
246 enum object_type *type,
247 unsigned long *size);
248
249/* Read and unpack an object file into memory, write memory to an object file */
250int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
251
252void hash_object_file(const struct git_hash_algo *algo, const void *buf,
253 unsigned long len, enum object_type type,
254 struct object_id *oid);
255
256int write_object_file_flags(const void *buf, unsigned long len,
257 enum object_type type, struct object_id *oid,
c2538492 258 struct object_id *comapt_oid_in, unsigned flags);
a034e910
EN
259static inline int write_object_file(const void *buf, unsigned long len,
260 enum object_type type, struct object_id *oid)
261{
c2538492 262 return write_object_file_flags(buf, len, type, oid, NULL, 0);
a034e910
EN
263}
264
265int write_object_file_literally(const void *buf, unsigned long len,
266 const char *type, struct object_id *oid,
267 unsigned flags);
268int stream_loose_object(struct input_stream *in_stream, size_t len,
269 struct object_id *oid);
270
271/*
272 * Add an object file to the in-memory object store, without writing it
273 * to disk.
274 *
275 * Callers are responsible for calling write_object_file to record the
276 * object in persistent storage before writing any other new objects
277 * that reference it.
278 */
279int pretend_object_file(void *, unsigned long, enum object_type,
280 struct object_id *oid);
281
282int force_object_loose(const struct object_id *oid, time_t mtime);
283
284struct object_info {
285 /* Request */
286 enum object_type *typep;
287 unsigned long *sizep;
288 off_t *disk_sizep;
289 struct object_id *delta_base_oid;
290 struct strbuf *type_name;
291 void **contentp;
292
293 /* Response */
294 enum {
295 OI_CACHED,
296 OI_LOOSE,
297 OI_PACKED,
298 OI_DBCACHED
299 } whence;
300 union {
301 /*
302 * struct {
303 * ... Nothing to expose in this case
304 * } cached;
305 * struct {
306 * ... Nothing to expose in this case
307 * } loose;
308 */
309 struct {
310 struct packed_git *pack;
311 off_t offset;
312 unsigned int is_delta;
313 } packed;
314 } u;
315};
316
317/*
318 * Initializer for a "struct object_info" that wants no items. You may
319 * also memset() the memory to all-zeroes.
320 */
321#define OBJECT_INFO_INIT { 0 }
322
323/* Invoke lookup_replace_object() on the given hash */
324#define OBJECT_INFO_LOOKUP_REPLACE 1
325/* Allow reading from a loose object file of unknown/bogus type */
326#define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
327/* Do not retry packed storage after checking packed and loose storage */
328#define OBJECT_INFO_QUICK 8
329/*
330 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
331 * nonzero).
332 */
333#define OBJECT_INFO_SKIP_FETCH_OBJECT 16
334/*
335 * This is meant for bulk prefetching of missing blobs in a partial
336 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
337 */
338#define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
339
340/* Die if object corruption (not just an object being missing) was detected. */
341#define OBJECT_INFO_DIE_IF_CORRUPT 32
342
343int oid_object_info_extended(struct repository *r,
344 const struct object_id *,
345 struct object_info *, unsigned flags);
346
347/*
348 * Open the loose object at path, check its hash, and return the contents,
349 * use the "oi" argument to assert things about the object, or e.g. populate its
350 * type, and size. If the object is a blob, then "contents" may return NULL,
351 * to allow streaming of large blobs.
352 *
353 * Returns 0 on success, negative on error (details may be written to stderr).
354 */
355int read_loose_object(const char *path,
356 const struct object_id *expected_oid,
357 struct object_id *real_oid,
358 void **contents,
359 struct object_info *oi);
360
361/* Retry packed storage after checking packed and loose storage */
362#define HAS_OBJECT_RECHECK_PACKED 1
363
364/*
365 * Returns 1 if the object exists. This function will not lazily fetch objects
366 * in a partial clone.
367 */
368int has_object(struct repository *r, const struct object_id *oid,
369 unsigned flags);
370
371/*
372 * These macros and functions are deprecated. If checking existence for an
373 * object that is likely to be missing and/or whose absence is relatively
374 * inconsequential (or is consequential but the caller is prepared to handle
375 * it), use has_object(), which has better defaults (no lazy fetch in a partial
376 * clone and no rechecking of packed storage). In the unlikely event that a
377 * caller needs to assert existence of an object that it fully expects to
378 * exist, and wants to trigger a lazy fetch in a partial clone, use
379 * oid_object_info_extended() with a NULL struct object_info.
380 *
381 * These functions can be removed once all callers have migrated to
382 * has_object() and/or oid_object_info_extended().
383 */
384int repo_has_object_file(struct repository *r, const struct object_id *oid);
385int repo_has_object_file_with_flags(struct repository *r,
386 const struct object_id *oid, int flags);
387
388/*
389 * Return true iff an alternate object database has a loose object
390 * with the specified name. This function does not respect replace
391 * references.
392 */
393int has_loose_object_nonlocal(const struct object_id *);
394
395int has_loose_object(const struct object_id *);
396
397/**
398 * format_object_header() is a thin wrapper around s xsnprintf() that
399 * writes the initial "<type> <obj-len>" part of the loose object
400 * header. It returns the size that snprintf() returns + 1.
401 */
402int format_object_header(char *str, size_t size, enum object_type type,
403 size_t objsize);
404
405void assert_oid_type(const struct object_id *oid, enum object_type expect);
406
407/*
408 * Enabling the object read lock allows multiple threads to safely call the
409 * following functions in parallel: repo_read_object_file(),
410 * read_object_with_reference(), oid_object_info() and oid_object_info_extended().
411 *
412 * obj_read_lock() and obj_read_unlock() may also be used to protect other
413 * section which cannot execute in parallel with object reading. Since the used
414 * lock is a recursive mutex, these sections can even contain calls to object
415 * reading functions. However, beware that in these cases zlib inflation won't
416 * be performed in parallel, losing performance.
417 *
418 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
419 * any of its callees end up calling it, this recursive call won't benefit from
420 * parallel inflation.
421 */
422void enable_obj_read_lock(void);
423void disable_obj_read_lock(void);
424
425extern int obj_read_use_lock;
426extern pthread_mutex_t obj_read_mutex;
427
428static inline void obj_read_lock(void)
429{
430 if(obj_read_use_lock)
431 pthread_mutex_lock(&obj_read_mutex);
432}
433
434static inline void obj_read_unlock(void)
435{
436 if(obj_read_use_lock)
437 pthread_mutex_unlock(&obj_read_mutex);
438}
439
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),
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),
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
536#endif /* OBJECT_STORE_LL_H */