1#include "cache.h"
2#include "delta.h"
3#include "pack.h"
4#include "csum-file.h"
5#include "blob.h"
6#include "commit.h"
7#include "tag.h"
8#include "tree.h"
9
10static const char index_pack_usage[] =
11"git-index-pack [-o index-file] pack-file";
12
13struct object_entry
14{
15 unsigned long offset;
16 unsigned long size;
17 unsigned int hdr_size;
18 enum object_type type;
19 enum object_type real_type;
20 unsigned char sha1[20];
21};
22
23union delta_base {
24 unsigned char sha1[20];
25 unsigned long offset;
26};
27
28/*
29 * Even if sizeof(union delta_base) == 24 on 64-bit archs, we really want
30 * to memcmp() only the first 20 bytes.
31 */
32#define UNION_BASE_SZ 20
33
34struct delta_entry
35{
36 struct object_entry *obj;
37 union delta_base base;
38};
39
40static const char *pack_name;
41static struct object_entry *objects;
42static struct delta_entry *deltas;
43static int nr_objects;
44static int nr_deltas;
45
46/* We always read in 4kB chunks. */
47static unsigned char input_buffer[4096];
48static unsigned long input_offset, input_len, consumed_bytes;
49static SHA_CTX input_ctx;
50static int input_fd;
51
52/*
53 * Make sure at least "min" bytes are available in the buffer, and
54 * return the pointer to the buffer.
55 */
56static void *fill(int min)
57{
58 if (min <= input_len)
59 return input_buffer + input_offset;
60 if (min > sizeof(input_buffer))
61 die("cannot fill %d bytes", min);
62 if (input_offset) {
63 SHA1_Update(&input_ctx, input_buffer, input_offset);
64 memmove(input_buffer, input_buffer + input_offset, input_len);
65 input_offset = 0;
66 }
67 do {
68 int ret = xread(input_fd, input_buffer + input_len,
69 sizeof(input_buffer) - input_len);
70 if (ret <= 0) {
71 if (!ret)
72 die("early EOF");
73 die("read error on input: %s", strerror(errno));
74 }
75 input_len += ret;
76 } while (input_len < min);
77 return input_buffer;
78}
79
80static void use(int bytes)
81{
82 if (bytes > input_len)
83 die("used more bytes than were available");
84 input_len -= bytes;
85 input_offset += bytes;
86 consumed_bytes += bytes;
87}
88
89static void open_pack_file(void)
90{
91 input_fd = open(pack_name, O_RDONLY);
92 if (input_fd < 0)
93 die("cannot open packfile '%s': %s", pack_name,
94 strerror(errno));
95 SHA1_Init(&input_ctx);
96}
97
98static void parse_pack_header(void)
99{
100 struct pack_header *hdr = fill(sizeof(struct pack_header));
101
102 /* Header consistency check */
103 if (hdr->hdr_signature != htonl(PACK_SIGNATURE))
104 die("packfile '%s' signature mismatch", pack_name);
105 if (!pack_version_ok(hdr->hdr_version))
106 die("packfile '%s' version %d unsupported",
107 pack_name, ntohl(hdr->hdr_version));
108
109 nr_objects = ntohl(hdr->hdr_entries);
110 use(sizeof(struct pack_header));
111 /*fprintf(stderr, "Indexing %d objects\n", nr_objects);*/
112}
113
114static void bad_object(unsigned long offset, const char *format,
115 ...) NORETURN __attribute__((format (printf, 2, 3)));
116
117static void bad_object(unsigned long offset, const char *format, ...)
118{
119 va_list params;
120 char buf[1024];
121
122 va_start(params, format);
123 vsnprintf(buf, sizeof(buf), format, params);
124 va_end(params);
125 die("packfile '%s': bad object at offset %lu: %s",
126 pack_name, offset, buf);
127}
128
129static void *unpack_entry_data(unsigned long offset, unsigned long size)
130{
131 z_stream stream;
132 void *buf = xmalloc(size);
133
134 memset(&stream, 0, sizeof(stream));
135 stream.next_out = buf;
136 stream.avail_out = size;
137 stream.next_in = fill(1);
138 stream.avail_in = input_len;
139 inflateInit(&stream);
140
141 for (;;) {
142 int ret = inflate(&stream, 0);
143 use(input_len - stream.avail_in);
144 if (stream.total_out == size && ret == Z_STREAM_END)
145 break;
146 if (ret != Z_OK)
147 bad_object(offset, "inflate returned %d", ret);
148 stream.next_in = fill(1);
149 stream.avail_in = input_len;
150 }
151 inflateEnd(&stream);
152 return buf;
153}
154
155static void *unpack_raw_entry(struct object_entry *obj, union delta_base *delta_base)
156{
157 unsigned char *p, c;
158 unsigned long size, base_offset;
159 unsigned shift;
160
161 obj->offset = consumed_bytes;
162
163 p = fill(1);
164 c = *p;
165 use(1);
166 obj->type = (c >> 4) & 7;
167 size = (c & 15);
168 shift = 4;
169 while (c & 0x80) {
170 p = fill(1);
171 c = *p;
172 use(1);
173 size += (c & 0x7fUL) << shift;
174 shift += 7;
175 }
176 obj->size = size;
177
178 switch (obj->type) {
179 case OBJ_REF_DELTA:
180 hashcpy(delta_base->sha1, fill(20));
181 use(20);
182 break;
183 case OBJ_OFS_DELTA:
184 memset(delta_base, 0, sizeof(*delta_base));
185 p = fill(1);
186 c = *p;
187 use(1);
188 base_offset = c & 127;
189 while (c & 128) {
190 base_offset += 1;
191 if (!base_offset || base_offset & ~(~0UL >> 7))
192 bad_object(obj->offset, "offset value overflow for delta base object");
193 p = fill(1);
194 c = *p;
195 use(1);
196 base_offset = (base_offset << 7) + (c & 127);
197 }
198 delta_base->offset = obj->offset - base_offset;
199 if (delta_base->offset >= obj->offset)
200 bad_object(obj->offset, "delta base offset is out of bound");
201 break;
202 case OBJ_COMMIT:
203 case OBJ_TREE:
204 case OBJ_BLOB:
205 case OBJ_TAG:
206 break;
207 default:
208 bad_object(obj->offset, "bad object type %d", obj->type);
209 }
210 obj->hdr_size = consumed_bytes - obj->offset;
211
212 return unpack_entry_data(obj->offset, obj->size);
213}
214
215static void * get_data_from_pack(struct object_entry *obj)
216{
217 unsigned long from = obj[0].offset + obj[0].hdr_size;
218 unsigned long len = obj[1].offset - from;
219 unsigned pg_offset = from % getpagesize();
220 unsigned char *map, *data;
221 z_stream stream;
222 int st;
223
224 map = mmap(NULL, len + pg_offset, PROT_READ, MAP_PRIVATE,
225 input_fd, from - pg_offset);
226 if (map == MAP_FAILED)
227 die("cannot mmap packfile '%s': %s", pack_name, strerror(errno));
228 data = xmalloc(obj->size);
229 memset(&stream, 0, sizeof(stream));
230 stream.next_out = data;
231 stream.avail_out = obj->size;
232 stream.next_in = map + pg_offset;
233 stream.avail_in = len;
234 inflateInit(&stream);
235 while ((st = inflate(&stream, Z_FINISH)) == Z_OK);
236 inflateEnd(&stream);
237 if (st != Z_STREAM_END || stream.total_out != obj->size)
238 die("serious inflate inconsistency");
239 munmap(map, len + pg_offset);
240 return data;
241}
242
243static int find_delta(const union delta_base *base)
244{
245 int first = 0, last = nr_deltas;
246
247 while (first < last) {
248 int next = (first + last) / 2;
249 struct delta_entry *delta = &deltas[next];
250 int cmp;
251
252 cmp = memcmp(base, &delta->base, UNION_BASE_SZ);
253 if (!cmp)
254 return next;
255 if (cmp < 0) {
256 last = next;
257 continue;
258 }
259 first = next+1;
260 }
261 return -first-1;
262}
263
264static int find_delta_childs(const union delta_base *base,
265 int *first_index, int *last_index)
266{
267 int first = find_delta(base);
268 int last = first;
269 int end = nr_deltas - 1;
270
271 if (first < 0)
272 return -1;
273 while (first > 0 && !memcmp(&deltas[first - 1].base, base, UNION_BASE_SZ))
274 --first;
275 while (last < end && !memcmp(&deltas[last + 1].base, base, UNION_BASE_SZ))
276 ++last;
277 *first_index = first;
278 *last_index = last;
279 return 0;
280}
281
282static void sha1_object(const void *data, unsigned long size,
283 enum object_type type, unsigned char *sha1)
284{
285 SHA_CTX ctx;
286 char header[50];
287 int header_size;
288 const char *type_str;
289
290 switch (type) {
291 case OBJ_COMMIT: type_str = commit_type; break;
292 case OBJ_TREE: type_str = tree_type; break;
293 case OBJ_BLOB: type_str = blob_type; break;
294 case OBJ_TAG: type_str = tag_type; break;
295 default:
296 die("bad type %d", type);
297 }
298
299 header_size = sprintf(header, "%s %lu", type_str, size) + 1;
300
301 SHA1_Init(&ctx);
302 SHA1_Update(&ctx, header, header_size);
303 SHA1_Update(&ctx, data, size);
304 SHA1_Final(sha1, &ctx);
305}
306
307static void resolve_delta(struct delta_entry *delta, void *base_data,
308 unsigned long base_size, enum object_type type)
309{
310 struct object_entry *obj = delta->obj;
311 void *delta_data;
312 unsigned long delta_size;
313 void *result;
314 unsigned long result_size;
315 union delta_base delta_base;
316 int j, first, last;
317
318 obj->real_type = type;
319 delta_data = get_data_from_pack(obj);
320 delta_size = obj->size;
321 result = patch_delta(base_data, base_size, delta_data, delta_size,
322 &result_size);
323 free(delta_data);
324 if (!result)
325 bad_object(obj->offset, "failed to apply delta");
326 sha1_object(result, result_size, type, obj->sha1);
327
328 hashcpy(delta_base.sha1, obj->sha1);
329 if (!find_delta_childs(&delta_base, &first, &last)) {
330 for (j = first; j <= last; j++)
331 if (deltas[j].obj->type == OBJ_REF_DELTA)
332 resolve_delta(&deltas[j], result, result_size, type);
333 }
334
335 memset(&delta_base, 0, sizeof(delta_base));
336 delta_base.offset = obj->offset;
337 if (!find_delta_childs(&delta_base, &first, &last)) {
338 for (j = first; j <= last; j++)
339 if (deltas[j].obj->type == OBJ_OFS_DELTA)
340 resolve_delta(&deltas[j], result, result_size, type);
341 }
342
343 free(result);
344}
345
346static int compare_delta_entry(const void *a, const void *b)
347{
348 const struct delta_entry *delta_a = a;
349 const struct delta_entry *delta_b = b;
350 return memcmp(&delta_a->base, &delta_b->base, UNION_BASE_SZ);
351}
352
353/* Parse all objects and return the pack content SHA1 hash */
354static void parse_pack_objects(unsigned char *sha1)
355{
356 int i;
357 struct delta_entry *delta = deltas;
358 void *data;
359 struct stat st;
360
361 /*
362 * First pass:
363 * - find locations of all objects;
364 * - calculate SHA1 of all non-delta objects;
365 * - remember base SHA1 for all deltas.
366 */
367 for (i = 0; i < nr_objects; i++) {
368 struct object_entry *obj = &objects[i];
369 data = unpack_raw_entry(obj, &delta->base);
370 obj->real_type = obj->type;
371 if (obj->type == OBJ_REF_DELTA || obj->type == OBJ_OFS_DELTA) {
372 nr_deltas++;
373 delta->obj = obj;
374 delta++;
375 } else
376 sha1_object(data, obj->size, obj->type, obj->sha1);
377 free(data);
378 }
379 objects[i].offset = consumed_bytes;
380
381 /* Check pack integrity */
382 SHA1_Update(&input_ctx, input_buffer, input_offset);
383 SHA1_Final(sha1, &input_ctx);
384 if (hashcmp(fill(20), sha1))
385 die("packfile '%s' SHA1 mismatch", pack_name);
386 use(20);
387
388 /* If input_fd is a file, we should have reached its end now. */
389 if (fstat(input_fd, &st))
390 die("cannot fstat packfile '%s': %s", pack_name, strerror(errno));
391 if (S_ISREG(st.st_mode) && st.st_size != consumed_bytes)
392 die("packfile '%s' has junk at the end", pack_name);
393
394 /* Sort deltas by base SHA1/offset for fast searching */
395 qsort(deltas, nr_deltas, sizeof(struct delta_entry),
396 compare_delta_entry);
397
398 /*
399 * Second pass:
400 * - for all non-delta objects, look if it is used as a base for
401 * deltas;
402 * - if used as a base, uncompress the object and apply all deltas,
403 * recursively checking if the resulting object is used as a base
404 * for some more deltas.
405 */
406 for (i = 0; i < nr_objects; i++) {
407 struct object_entry *obj = &objects[i];
408 union delta_base base;
409 int j, ref, ref_first, ref_last, ofs, ofs_first, ofs_last;
410
411 if (obj->type == OBJ_REF_DELTA || obj->type == OBJ_OFS_DELTA)
412 continue;
413 hashcpy(base.sha1, obj->sha1);
414 ref = !find_delta_childs(&base, &ref_first, &ref_last);
415 memset(&base, 0, sizeof(base));
416 base.offset = obj->offset;
417 ofs = !find_delta_childs(&base, &ofs_first, &ofs_last);
418 if (!ref && !ofs)
419 continue;
420 data = get_data_from_pack(obj);
421 if (ref)
422 for (j = ref_first; j <= ref_last; j++)
423 if (deltas[j].obj->type == OBJ_REF_DELTA)
424 resolve_delta(&deltas[j], data,
425 obj->size, obj->type);
426 if (ofs)
427 for (j = ofs_first; j <= ofs_last; j++)
428 if (deltas[j].obj->type == OBJ_OFS_DELTA)
429 resolve_delta(&deltas[j], data,
430 obj->size, obj->type);
431 free(data);
432 }
433
434 /* Check for unresolved deltas */
435 for (i = 0; i < nr_deltas; i++) {
436 if (deltas[i].obj->real_type == OBJ_REF_DELTA ||
437 deltas[i].obj->real_type == OBJ_OFS_DELTA)
438 die("packfile '%s' has unresolved deltas", pack_name);
439 }
440}
441
442static int sha1_compare(const void *_a, const void *_b)
443{
444 struct object_entry *a = *(struct object_entry **)_a;
445 struct object_entry *b = *(struct object_entry **)_b;
446 return hashcmp(a->sha1, b->sha1);
447}
448
449/*
450 * On entry *sha1 contains the pack content SHA1 hash, on exit it is
451 * the SHA1 hash of sorted object names.
452 */
453static void write_index_file(const char *index_name, unsigned char *sha1)
454{
455 struct sha1file *f;
456 struct object_entry **sorted_by_sha, **list, **last;
457 unsigned int array[256];
458 int i;
459 SHA_CTX ctx;
460
461 if (nr_objects) {
462 sorted_by_sha =
463 xcalloc(nr_objects, sizeof(struct object_entry *));
464 list = sorted_by_sha;
465 last = sorted_by_sha + nr_objects;
466 for (i = 0; i < nr_objects; ++i)
467 sorted_by_sha[i] = &objects[i];
468 qsort(sorted_by_sha, nr_objects, sizeof(sorted_by_sha[0]),
469 sha1_compare);
470
471 }
472 else
473 sorted_by_sha = list = last = NULL;
474
475 unlink(index_name);
476 f = sha1create("%s", index_name);
477
478 /*
479 * Write the first-level table (the list is sorted,
480 * but we use a 256-entry lookup to be able to avoid
481 * having to do eight extra binary search iterations).
482 */
483 for (i = 0; i < 256; i++) {
484 struct object_entry **next = list;
485 while (next < last) {
486 struct object_entry *obj = *next;
487 if (obj->sha1[0] != i)
488 break;
489 next++;
490 }
491 array[i] = htonl(next - sorted_by_sha);
492 list = next;
493 }
494 sha1write(f, array, 256 * sizeof(int));
495
496 /* recompute the SHA1 hash of sorted object names.
497 * currently pack-objects does not do this, but that
498 * can be fixed.
499 */
500 SHA1_Init(&ctx);
501 /*
502 * Write the actual SHA1 entries..
503 */
504 list = sorted_by_sha;
505 for (i = 0; i < nr_objects; i++) {
506 struct object_entry *obj = *list++;
507 unsigned int offset = htonl(obj->offset);
508 sha1write(f, &offset, 4);
509 sha1write(f, obj->sha1, 20);
510 SHA1_Update(&ctx, obj->sha1, 20);
511 }
512 sha1write(f, sha1, 20);
513 sha1close(f, NULL, 1);
514 free(sorted_by_sha);
515 SHA1_Final(sha1, &ctx);
516}
517
518int main(int argc, char **argv)
519{
520 int i;
521 char *index_name = NULL;
522 char *index_name_buf = NULL;
523 unsigned char sha1[20];
524
525 for (i = 1; i < argc; i++) {
526 const char *arg = argv[i];
527
528 if (*arg == '-') {
529 if (!strcmp(arg, "-o")) {
530 if (index_name || (i+1) >= argc)
531 usage(index_pack_usage);
532 index_name = argv[++i];
533 } else
534 usage(index_pack_usage);
535 continue;
536 }
537
538 if (pack_name)
539 usage(index_pack_usage);
540 pack_name = arg;
541 }
542
543 if (!pack_name)
544 usage(index_pack_usage);
545 if (!index_name) {
546 int len = strlen(pack_name);
547 if (!has_extension(pack_name, ".pack"))
548 die("packfile name '%s' does not end with '.pack'",
549 pack_name);
550 index_name_buf = xmalloc(len);
551 memcpy(index_name_buf, pack_name, len - 5);
552 strcpy(index_name_buf + len - 5, ".idx");
553 index_name = index_name_buf;
554 }
555
556 open_pack_file();
557 parse_pack_header();
558 objects = xcalloc(nr_objects + 1, sizeof(struct object_entry));
559 deltas = xcalloc(nr_objects, sizeof(struct delta_entry));
560 parse_pack_objects(sha1);
561 free(deltas);
562 write_index_file(index_name, sha1);
563 free(objects);
564 free(index_name_buf);
565
566 printf("%s\n", sha1_to_hex(sha1));
567
568 return 0;
569}