4 #include "object-store.h"
12 #include "tree-walk.h"
17 static int dry_run, quiet, recover, has_errors, strict;
18 static const char unpack_usage[] = "git unpack-objects [-n] [-q] [-r] [--strict]";
20 /* We always read in 4kB chunks. */
21 static unsigned char buffer[4096];
22 static unsigned int offset, len;
23 static off_t consumed_bytes;
24 static off_t max_input_size;
25 static git_hash_ctx ctx;
26 static struct fsck_options fsck_options = FSCK_OPTIONS_STRICT;
27 static struct progress *progress;
30 * When running under --strict mode, objects whose reachability are
31 * suspect are kept in core without getting written in the object
39 static struct decoration obj_decorate;
41 static struct obj_buffer *lookup_object_buffer(struct object *base)
43 return lookup_decoration(&obj_decorate, base);
46 static void add_object_buffer(struct object *object, char *buffer, unsigned long size)
48 struct obj_buffer *obj;
52 if (add_decoration(&obj_decorate, object, obj))
53 die("object %s tried to add buffer twice!", oid_to_hex(&object->oid));
57 * Make sure at least "min" bytes are available in the buffer, and
58 * return the pointer to the buffer.
60 static void *fill(int min)
63 return buffer + offset;
64 if (min > sizeof(buffer))
65 die("cannot fill %d bytes", min);
67 the_hash_algo->update_fn(&ctx, buffer, offset);
68 memmove(buffer, buffer + offset, len);
72 ssize_t ret = xread(0, buffer + len, sizeof(buffer) - len);
76 die_errno("read error on input");
83 static void use(int bytes)
86 die("used more bytes than were available");
90 /* make sure off_t is sufficiently large not to wrap */
91 if (signed_add_overflows(consumed_bytes, bytes))
92 die("pack too large for current definition of off_t");
93 consumed_bytes += bytes;
94 if (max_input_size && consumed_bytes > max_input_size)
95 die(_("pack exceeds maximum allowed size"));
96 display_throughput(progress, consumed_bytes);
99 static void *get_data(unsigned long size)
102 void *buf = xmallocz(size);
104 memset(&stream, 0, sizeof(stream));
106 stream.next_out = buf;
107 stream.avail_out = size;
108 stream.next_in = fill(1);
109 stream.avail_in = len;
110 git_inflate_init(&stream);
113 int ret = git_inflate(&stream, 0);
114 use(len - stream.avail_in);
115 if (stream.total_out == size && ret == Z_STREAM_END)
118 error("inflate returned %d", ret);
125 stream.next_in = fill(1);
126 stream.avail_in = len;
128 git_inflate_end(&stream);
133 struct object_id base_oid;
138 struct delta_info *next;
141 static struct delta_info *delta_list;
143 static void add_delta_to_list(unsigned nr, const struct object_id *base_oid,
145 void *delta, unsigned long size)
147 struct delta_info *info = xmalloc(sizeof(*info));
149 oidcpy(&info->base_oid, base_oid);
150 info->base_offset = base_offset;
154 info->next = delta_list;
160 struct object_id oid;
164 /* Remember to update object flag allocation in object.h */
165 #define FLAG_OPEN (1u<<20)
166 #define FLAG_WRITTEN (1u<<21)
168 static struct obj_info *obj_list;
169 static unsigned nr_objects;
172 * Called only from check_object() after it verified this object
175 static void write_cached_object(struct object *obj, struct obj_buffer *obj_buf)
177 struct object_id oid;
179 if (write_object_file(obj_buf->buffer, obj_buf->size,
180 type_name(obj->type), &oid) < 0)
181 die("failed to write object %s", oid_to_hex(&obj->oid));
182 obj->flags |= FLAG_WRITTEN;
186 * At the very end of the processing, write_rest() scans the objects
187 * that have reachability requirements and calls this function.
188 * Verify its reachability and validity recursively and write it out.
190 static int check_object(struct object *obj, enum object_type type,
191 void *data, struct fsck_options *options)
193 struct obj_buffer *obj_buf;
198 if (obj->flags & FLAG_WRITTEN)
201 if (type != OBJ_ANY && obj->type != type)
202 die("object type mismatch");
204 if (!(obj->flags & FLAG_OPEN)) {
206 int type = oid_object_info(the_repository, &obj->oid, &size);
207 if (type != obj->type || type <= 0)
208 die("object of unexpected type");
209 obj->flags |= FLAG_WRITTEN;
213 obj_buf = lookup_object_buffer(obj);
215 die("Whoops! Cannot find object '%s'", oid_to_hex(&obj->oid));
216 if (fsck_object(obj, obj_buf->buffer, obj_buf->size, &fsck_options))
217 die("fsck error in packed object");
218 fsck_options.walk = check_object;
219 if (fsck_walk(obj, NULL, &fsck_options))
220 die("Error on reachable objects of %s", oid_to_hex(&obj->oid));
221 write_cached_object(obj, obj_buf);
225 static void write_rest(void)
228 for (i = 0; i < nr_objects; i++) {
230 check_object(obj_list[i].obj, OBJ_ANY, NULL, NULL);
234 static void added_object(unsigned nr, enum object_type type,
235 void *data, unsigned long size);
238 * Write out nr-th object from the list, now we know the contents
239 * of it. Under --strict, this buffers structured objects in-core,
240 * to be checked at the end.
242 static void write_object(unsigned nr, enum object_type type,
243 void *buf, unsigned long size)
246 if (write_object_file(buf, size, type_name(type),
247 &obj_list[nr].oid) < 0)
248 die("failed to write object");
249 added_object(nr, type, buf, size);
251 obj_list[nr].obj = NULL;
252 } else if (type == OBJ_BLOB) {
254 if (write_object_file(buf, size, type_name(type),
255 &obj_list[nr].oid) < 0)
256 die("failed to write object");
257 added_object(nr, type, buf, size);
260 blob = lookup_blob(the_repository, &obj_list[nr].oid);
262 blob->object.flags |= FLAG_WRITTEN;
264 die("invalid blob object");
265 obj_list[nr].obj = NULL;
269 hash_object_file(the_hash_algo, buf, size, type_name(type),
271 added_object(nr, type, buf, size);
272 obj = parse_object_buffer(the_repository, &obj_list[nr].oid,
276 die("invalid %s", type_name(type));
277 add_object_buffer(obj, buf, size);
278 obj->flags |= FLAG_OPEN;
279 obj_list[nr].obj = obj;
283 static void resolve_delta(unsigned nr, enum object_type type,
284 void *base, unsigned long base_size,
285 void *delta, unsigned long delta_size)
288 unsigned long result_size;
290 result = patch_delta(base, base_size,
294 die("failed to apply delta");
296 write_object(nr, type, result, result_size);
300 * We now know the contents of an object (which is nr-th in the pack);
301 * resolve all the deltified objects that are based on it.
303 static void added_object(unsigned nr, enum object_type type,
304 void *data, unsigned long size)
306 struct delta_info **p = &delta_list;
307 struct delta_info *info;
309 while ((info = *p) != NULL) {
310 if (oideq(&info->base_oid, &obj_list[nr].oid) ||
311 info->base_offset == obj_list[nr].offset) {
314 resolve_delta(info->nr, type, data, size,
315 info->delta, info->size);
323 static void unpack_non_delta_entry(enum object_type type, unsigned long size,
326 void *buf = get_data(size);
329 write_object(nr, type, buf, size);
334 static int resolve_against_held(unsigned nr, const struct object_id *base,
335 void *delta_data, unsigned long delta_size)
338 struct obj_buffer *obj_buffer;
339 obj = lookup_object(the_repository, base);
342 obj_buffer = lookup_object_buffer(obj);
345 resolve_delta(nr, obj->type, obj_buffer->buffer,
346 obj_buffer->size, delta_data, delta_size);
350 static void unpack_delta_entry(enum object_type type, unsigned long delta_size,
353 void *delta_data, *base;
354 unsigned long base_size;
355 struct object_id base_oid;
357 if (type == OBJ_REF_DELTA) {
358 hashcpy(base_oid.hash, fill(the_hash_algo->rawsz));
359 use(the_hash_algo->rawsz);
360 delta_data = get_data(delta_size);
361 if (dry_run || !delta_data) {
365 if (has_object_file(&base_oid))
366 ; /* Ok we have this one */
367 else if (resolve_against_held(nr, &base_oid,
368 delta_data, delta_size))
369 return; /* we are done */
371 /* cannot resolve yet --- queue it */
372 oidclr(&obj_list[nr].oid);
373 add_delta_to_list(nr, &base_oid, 0, delta_data, delta_size);
377 unsigned base_found = 0;
378 unsigned char *pack, c;
380 unsigned lo, mid, hi;
385 base_offset = c & 127;
388 if (!base_offset || MSB(base_offset, 7))
389 die("offset value overflow for delta base object");
393 base_offset = (base_offset << 7) + (c & 127);
395 base_offset = obj_list[nr].offset - base_offset;
396 if (base_offset <= 0 || base_offset >= obj_list[nr].offset)
397 die("offset value out of bound for delta base object");
399 delta_data = get_data(delta_size);
400 if (dry_run || !delta_data) {
407 mid = lo + (hi - lo) / 2;
408 if (base_offset < obj_list[mid].offset) {
410 } else if (base_offset > obj_list[mid].offset) {
413 oidcpy(&base_oid, &obj_list[mid].oid);
414 base_found = !is_null_oid(&base_oid);
420 * The delta base object is itself a delta that
421 * has not been resolved yet.
423 oidclr(&obj_list[nr].oid);
424 add_delta_to_list(nr, &null_oid, base_offset, delta_data, delta_size);
429 if (resolve_against_held(nr, &base_oid, delta_data, delta_size))
432 base = read_object_file(&base_oid, &type, &base_size);
434 error("failed to read delta-pack base object %s",
435 oid_to_hex(&base_oid));
441 resolve_delta(nr, type, base, base_size, delta_data, delta_size);
445 static void unpack_one(unsigned nr)
449 unsigned long size, c;
450 enum object_type type;
452 obj_list[nr].offset = consumed_bytes;
464 size += (c & 0x7f) << shift;
473 unpack_non_delta_entry(type, size, nr);
477 unpack_delta_entry(type, size, nr);
480 error("bad object type %d", type);
488 static void unpack_all(void)
491 struct pack_header *hdr = fill(sizeof(struct pack_header));
493 nr_objects = ntohl(hdr->hdr_entries);
495 if (ntohl(hdr->hdr_signature) != PACK_SIGNATURE)
496 die("bad pack file");
497 if (!pack_version_ok(hdr->hdr_version))
498 die("unknown pack file version %"PRIu32,
499 ntohl(hdr->hdr_version));
500 use(sizeof(struct pack_header));
503 progress = start_progress(_("Unpacking objects"), nr_objects);
504 CALLOC_ARRAY(obj_list, nr_objects);
505 for (i = 0; i < nr_objects; i++) {
507 display_progress(progress, i + 1);
509 stop_progress(&progress);
512 die("unresolved deltas left after unpacking");
515 int cmd_unpack_objects(int argc, const char **argv, const char *prefix)
518 struct object_id oid;
520 read_replace_refs = 0;
522 git_config(git_default_config, NULL);
526 for (i = 1 ; i < argc; i++) {
527 const char *arg = argv[i];
530 if (!strcmp(arg, "-n")) {
534 if (!strcmp(arg, "-q")) {
538 if (!strcmp(arg, "-r")) {
542 if (!strcmp(arg, "--strict")) {
546 if (skip_prefix(arg, "--strict=", &arg)) {
548 fsck_set_msg_types(&fsck_options, arg);
551 if (starts_with(arg, "--pack_header=")) {
552 struct pack_header *hdr;
555 hdr = (struct pack_header *)buffer;
556 hdr->hdr_signature = htonl(PACK_SIGNATURE);
557 hdr->hdr_version = htonl(strtoul(arg + 14, &c, 10));
560 hdr->hdr_entries = htonl(strtoul(c + 1, &c, 10));
566 if (skip_prefix(arg, "--max-input-size=", &arg)) {
567 max_input_size = strtoumax(arg, NULL, 10);
573 /* We don't take any non-flag arguments now.. Maybe some day */
576 the_hash_algo->init_fn(&ctx);
578 the_hash_algo->update_fn(&ctx, buffer, offset);
579 the_hash_algo->final_fn(oid.hash, &ctx);
582 if (fsck_finish(&fsck_options))
583 die(_("fsck error in pack objects"));
585 if (!hasheq(fill(the_hash_algo->rawsz), oid.hash))
586 die("final sha1 did not match");
587 use(the_hash_algo->rawsz);
589 /* Write the last part of the buffer to stdout */
591 int ret = xwrite(1, buffer + offset, len);