[thirdparty/git.git] / pack-write.c

#include "cache.h"
#include "pack.h"
#include "csum-file.h"

void reset_pack_idx_option(struct pack_idx_option *opts)
{
	memset(opts, 0, sizeof(*opts));
	opts->version = 2;
	opts->off32_limit = 0x7fffffff;
}

static int sha1_compare(const void *_a, const void *_b)
{
	struct pack_idx_entry *a = *(struct pack_idx_entry **)_a;
	struct pack_idx_entry *b = *(struct pack_idx_entry **)_b;
	return hashcmp(a->sha1, b->sha1);
}

/*
 * On entry *sha1 contains the pack content SHA1 hash, on exit it is
 * the SHA1 hash of sorted object names. The objects array passed in
 * will be sorted by SHA1 on exit.
 */
const char *write_idx_file(const char *index_name, struct pack_idx_entry **objects,
			   int nr_objects, const struct pack_idx_option *opts,
			   unsigned char *sha1)
{
	struct sha1file *f;
	struct pack_idx_entry **sorted_by_sha, **list, **last;
	off_t last_obj_offset = 0;
	uint32_t array[256];
	int i, fd;
	git_SHA_CTX ctx;
	uint32_t index_version;

	if (nr_objects) {
		sorted_by_sha = objects;
		list = sorted_by_sha;
		last = sorted_by_sha + nr_objects;
		for (i = 0; i < nr_objects; ++i) {
			if (objects[i]->offset > last_obj_offset)
				last_obj_offset = objects[i]->offset;
		}
		qsort(sorted_by_sha, nr_objects, sizeof(sorted_by_sha[0]),
		      sha1_compare);
	}
	else
		sorted_by_sha = list = last = NULL;

	if (!index_name) {
		static char tmpfile[PATH_MAX];
		fd = odb_mkstemp(tmpfile, sizeof(tmpfile), "pack/tmp_idx_XXXXXX");
		index_name = xstrdup(tmpfile);
	} else {
		unlink(index_name);
		fd = open(index_name, O_CREAT|O_EXCL|O_WRONLY, 0600);
	}
	if (fd < 0)
		die_errno("unable to create '%s'", index_name);
	f = sha1fd(fd, index_name);

	/* if last object's offset is >= 2^31 we should use index V2 */
	index_version = (last_obj_offset >> 31) ? 2 : opts->version;

	/* index versions 2 and above need a header */
	if (index_version >= 2) {
		struct pack_idx_header hdr;
		hdr.idx_signature = htonl(PACK_IDX_SIGNATURE);
		hdr.idx_version = htonl(index_version);
		sha1write(f, &hdr, sizeof(hdr));
	}

	/*
	 * Write the first-level table (the list is sorted,
	 * but we use a 256-entry lookup to be able to avoid
	 * having to do eight extra binary search iterations).
	 */
	for (i = 0; i < 256; i++) {
		struct pack_idx_entry **next = list;
		while (next < last) {
			struct pack_idx_entry *obj = *next;
			if (obj->sha1[0] != i)
				break;
			next++;
		}
		array[i] = htonl(next - sorted_by_sha);
		list = next;
	}
	sha1write(f, array, 256 * 4);

	/* compute the SHA1 hash of sorted object names. */
	git_SHA1_Init(&ctx);

	/*
	 * Write the actual SHA1 entries..
	 */
	list = sorted_by_sha;
	for (i = 0; i < nr_objects; i++) {
		struct pack_idx_entry *obj = *list++;
		if (index_version < 2) {
			uint32_t offset = htonl(obj->offset);
			sha1write(f, &offset, 4);
		}
		sha1write(f, obj->sha1, 20);
		git_SHA1_Update(&ctx, obj->sha1, 20);
	}

	if (index_version >= 2) {
		unsigned int nr_large_offset = 0;

		/* write the crc32 table */
		list = sorted_by_sha;
		for (i = 0; i < nr_objects; i++) {
			struct pack_idx_entry *obj = *list++;
			uint32_t crc32_val = htonl(obj->crc32);
			sha1write(f, &crc32_val, 4);
		}

		/* write the 32-bit offset table */
		list = sorted_by_sha;
		for (i = 0; i < nr_objects; i++) {
			struct pack_idx_entry *obj = *list++;
			uint32_t offset = (obj->offset <= opts->off32_limit) ?
				obj->offset : (0x80000000 | nr_large_offset++);
			offset = htonl(offset);
			sha1write(f, &offset, 4);
		}

		/* write the large offset table */
		list = sorted_by_sha;
		while (nr_large_offset) {
			struct pack_idx_entry *obj = *list++;
			uint64_t offset = obj->offset;
			if (offset > opts->off32_limit) {
				uint32_t split[2];
				split[0] = htonl(offset >> 32);
				split[1] = htonl(offset & 0xffffffff);
				sha1write(f, split, 8);
				nr_large_offset--;
			}
		}
	}

	sha1write(f, sha1, 20);
	sha1close(f, NULL, CSUM_FSYNC);
	git_SHA1_Final(sha1, &ctx);
	return index_name;
}

/*
 * Update pack header with object_count and compute new SHA1 for pack data
 * associated to pack_fd, and write that SHA1 at the end.  That new SHA1
 * is also returned in new_pack_sha1.
 *
 * If partial_pack_sha1 is non null, then the SHA1 of the existing pack
 * (without the header update) is computed and validated against the
 * one provided in partial_pack_sha1.  The validation is performed at
 * partial_pack_offset bytes in the pack file.  The SHA1 of the remaining
 * data (i.e. from partial_pack_offset to the end) is then computed and
 * returned in partial_pack_sha1.
 *
 * Note that new_pack_sha1 is updated last, so both new_pack_sha1 and
 * partial_pack_sha1 can refer to the same buffer if the caller is not
 * interested in the resulting SHA1 of pack data above partial_pack_offset.
 */
void fixup_pack_header_footer(int pack_fd,
			 unsigned char *new_pack_sha1,
			 const char *pack_name,
			 uint32_t object_count,
			 unsigned char *partial_pack_sha1,
			 off_t partial_pack_offset)
{
	int aligned_sz, buf_sz = 8 * 1024;
	git_SHA_CTX old_sha1_ctx, new_sha1_ctx;
	struct pack_header hdr;
	char *buf;

	git_SHA1_Init(&old_sha1_ctx);
	git_SHA1_Init(&new_sha1_ctx);

	if (lseek(pack_fd, 0, SEEK_SET) != 0)
		die_errno("Failed seeking to start of '%s'", pack_name);
	if (read_in_full(pack_fd, &hdr, sizeof(hdr)) != sizeof(hdr))
		die_errno("Unable to reread header of '%s'", pack_name);
	if (lseek(pack_fd, 0, SEEK_SET) != 0)
		die_errno("Failed seeking to start of '%s'", pack_name);
	git_SHA1_Update(&old_sha1_ctx, &hdr, sizeof(hdr));
	hdr.hdr_entries = htonl(object_count);
	git_SHA1_Update(&new_sha1_ctx, &hdr, sizeof(hdr));
	write_or_die(pack_fd, &hdr, sizeof(hdr));
	partial_pack_offset -= sizeof(hdr);

	buf = xmalloc(buf_sz);
	aligned_sz = buf_sz - sizeof(hdr);
	for (;;) {
		ssize_t m, n;
		m = (partial_pack_sha1 && partial_pack_offset < aligned_sz) ?
			partial_pack_offset : aligned_sz;
		n = xread(pack_fd, buf, m);
		if (!n)
			break;
		if (n < 0)
			die_errno("Failed to checksum '%s'", pack_name);
		git_SHA1_Update(&new_sha1_ctx, buf, n);

		aligned_sz -= n;
		if (!aligned_sz)
			aligned_sz = buf_sz;

		if (!partial_pack_sha1)
			continue;

		git_SHA1_Update(&old_sha1_ctx, buf, n);
		partial_pack_offset -= n;
		if (partial_pack_offset == 0) {
			unsigned char sha1[20];
			git_SHA1_Final(sha1, &old_sha1_ctx);
			if (hashcmp(sha1, partial_pack_sha1) != 0)
				die("Unexpected checksum for %s "
				    "(disk corruption?)", pack_name);
			/*
			 * Now let's compute the SHA1 of the remainder of the
			 * pack, which also means making partial_pack_offset
			 * big enough not to matter anymore.
			 */
			git_SHA1_Init(&old_sha1_ctx);
			partial_pack_offset = ~partial_pack_offset;
			partial_pack_offset -= MSB(partial_pack_offset, 1);
		}
	}
	free(buf);

	if (partial_pack_sha1)
		git_SHA1_Final(partial_pack_sha1, &old_sha1_ctx);
	git_SHA1_Final(new_pack_sha1, &new_sha1_ctx);
	write_or_die(pack_fd, new_pack_sha1, 20);
	fsync_or_die(pack_fd, pack_name);
}

char *index_pack_lockfile(int ip_out)
{
	char packname[46];

	/*
	 * The first thing we expect from index-pack's output
	 * is "pack\t%40s\n" or "keep\t%40s\n" (46 bytes) where
	 * %40s is the newly created pack SHA1 name.  In the "keep"
	 * case, we need it to remove the corresponding .keep file
	 * later on.  If we don't get that then tough luck with it.
	 */
	if (read_in_full(ip_out, packname, 46) == 46 && packname[45] == '\n' &&
	    memcmp(packname, "keep\t", 5) == 0) {
		char path[PATH_MAX];
		packname[45] = 0;
		snprintf(path, sizeof(path), "%s/pack/pack-%s.keep",
			 get_object_directory(), packname + 5);
		return xstrdup(path);
	}
	return NULL;
}

/*
 * The per-object header is a pretty dense thing, which is
 *  - first byte: low four bits are "size", then three bits of "type",
 *    and the high bit is "size continues".
 *  - each byte afterwards: low seven bits are size continuation,
 *    with the high bit being "size continues"
 */
int encode_in_pack_object_header(enum object_type type, uintmax_t size, unsigned char *hdr)
{
	int n = 1;
	unsigned char c;

	if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
		die("bad type %d", type);

	c = (type << 4) | (size & 15);
	size >>= 4;
	while (size) {
		*hdr++ = c | 0x80;
		c = size & 0x7f;
		size >>= 7;
		n++;
	}
	*hdr = c;
	return n;
}
Commit	Line	Data
8b0eca7c DH	1	#include "cache.h"
8b0eca7c DH	2	#include "pack.h"
aa7e44bf GB	3	#include "csum-file.h"
aa7e44bf GB	4
ebcfb379 JH	5	void reset_pack_idx_option(struct pack_idx_option *opts)
	6	{
	7	memset(opts, 0, sizeof(*opts));
	8	opts->version = 2;
	9	opts->off32_limit = 0x7fffffff;
	10	}
aa7e44bf GB	11
	12	static int sha1_compare(const void _a, const void _b)
	13	{
	14	struct pack_idx_entry a = (struct pack_idx_entry **)_a;
	15	struct pack_idx_entry b = (struct pack_idx_entry **)_b;
	16	return hashcmp(a->sha1, b->sha1);
	17	}
	18
	19	/*
	20	* On entry *sha1 contains the pack content SHA1 hash, on exit it is
	21	* the SHA1 hash of sorted object names. The objects array passed in
	22	* will be sorted by SHA1 on exit.
	23	*/
3bb72562	24	const char write_idx_file(const char index_name, struct pack_idx_entry **objects,
ebcfb379 JH	25	int nr_objects, const struct pack_idx_option *opts,
ebcfb379 JH	26	unsigned char *sha1)
aa7e44bf GB	27	{
	28	struct sha1file *f;
	29	struct pack_idx_entry sorted_by_sha, list, **last;
	30	off_t last_obj_offset = 0;
	31	uint32_t array[256];
	32	int i, fd;
9126f009	33	git_SHA_CTX ctx;
aa7e44bf GB	34	uint32_t index_version;
	35
	36	if (nr_objects) {
	37	sorted_by_sha = objects;
	38	list = sorted_by_sha;
	39	last = sorted_by_sha + nr_objects;
	40	for (i = 0; i < nr_objects; ++i) {
	41	if (objects[i]->offset > last_obj_offset)
	42	last_obj_offset = objects[i]->offset;
	43	}
	44	qsort(sorted_by_sha, nr_objects, sizeof(sorted_by_sha[0]),
	45	sha1_compare);
	46	}
	47	else
	48	sorted_by_sha = list = last = NULL;
	49
	50	if (!index_name) {
	51	static char tmpfile[PATH_MAX];
6e180cdc	52	fd = odb_mkstemp(tmpfile, sizeof(tmpfile), "pack/tmp_idx_XXXXXX");
aa7e44bf GB	53	index_name = xstrdup(tmpfile);
	54	} else {
	55	unlink(index_name);
	56	fd = open(index_name, O_CREAT\|O_EXCL\|O_WRONLY, 0600);
	57	}
	58	if (fd < 0)
d824cbba	59	die_errno("unable to create '%s'", index_name);
aa7e44bf GB	60	f = sha1fd(fd, index_name);
	61
	62	/* if last object's offset is >= 2^31 we should use index V2 */
ebcfb379	63	index_version = (last_obj_offset >> 31) ? 2 : opts->version;
aa7e44bf GB	64
	65	/* index versions 2 and above need a header */
	66	if (index_version >= 2) {
	67	struct pack_idx_header hdr;
	68	hdr.idx_signature = htonl(PACK_IDX_SIGNATURE);
	69	hdr.idx_version = htonl(index_version);
	70	sha1write(f, &hdr, sizeof(hdr));
	71	}
	72
	73	/*
	74	* Write the first-level table (the list is sorted,
	75	* but we use a 256-entry lookup to be able to avoid
	76	* having to do eight extra binary search iterations).
	77	*/
	78	for (i = 0; i < 256; i++) {
	79	struct pack_idx_entry **next = list;
	80	while (next < last) {
	81	struct pack_idx_entry obj = next;
	82	if (obj->sha1[0] != i)
	83	break;
	84	next++;
	85	}
	86	array[i] = htonl(next - sorted_by_sha);
	87	list = next;
	88	}
	89	sha1write(f, array, 256 * 4);
	90
	91	/* compute the SHA1 hash of sorted object names. */
9126f009	92	git_SHA1_Init(&ctx);
aa7e44bf GB	93
	94	/*
	95	* Write the actual SHA1 entries..
	96	*/
	97	list = sorted_by_sha;
	98	for (i = 0; i < nr_objects; i++) {
	99	struct pack_idx_entry obj = list++;
	100	if (index_version < 2) {
	101	uint32_t offset = htonl(obj->offset);
	102	sha1write(f, &offset, 4);
	103	}
	104	sha1write(f, obj->sha1, 20);
9126f009	105	git_SHA1_Update(&ctx, obj->sha1, 20);
aa7e44bf GB	106	}
	107
	108	if (index_version >= 2) {
	109	unsigned int nr_large_offset = 0;
	110
	111	/* write the crc32 table */
	112	list = sorted_by_sha;
	113	for (i = 0; i < nr_objects; i++) {
	114	struct pack_idx_entry obj = list++;
	115	uint32_t crc32_val = htonl(obj->crc32);
	116	sha1write(f, &crc32_val, 4);
	117	}
	118
	119	/* write the 32-bit offset table */
	120	list = sorted_by_sha;
	121	for (i = 0; i < nr_objects; i++) {
	122	struct pack_idx_entry obj = list++;
ebcfb379	123	uint32_t offset = (obj->offset <= opts->off32_limit) ?
aa7e44bf GB	124	obj->offset : (0x80000000 \| nr_large_offset++);
	125	offset = htonl(offset);
	126	sha1write(f, &offset, 4);
	127	}
	128
	129	/* write the large offset table */
	130	list = sorted_by_sha;
	131	while (nr_large_offset) {
	132	struct pack_idx_entry obj = list++;
	133	uint64_t offset = obj->offset;
ebcfb379	134	if (offset > opts->off32_limit) {
aa7e44bf GB	135	uint32_t split[2];
	136	split[0] = htonl(offset >> 32);
	137	split[1] = htonl(offset & 0xffffffff);
	138	sha1write(f, split, 8);
	139	nr_large_offset--;
	140	}
	141	}
	142	}
	143
	144	sha1write(f, sha1, 20);
4c81b03e	145	sha1close(f, NULL, CSUM_FSYNC);
9126f009	146	git_SHA1_Final(sha1, &ctx);
aa7e44bf GB	147	return index_name;
aa7e44bf GB	148	}
8b0eca7c	149
abeb40e5 NP	150	/*
	151	* Update pack header with object_count and compute new SHA1 for pack data
	152	* associated to pack_fd, and write that SHA1 at the end. That new SHA1
	153	* is also returned in new_pack_sha1.
	154	*
	155	* If partial_pack_sha1 is non null, then the SHA1 of the existing pack
	156	* (without the header update) is computed and validated against the
	157	* one provided in partial_pack_sha1. The validation is performed at
	158	* partial_pack_offset bytes in the pack file. The SHA1 of the remaining
	159	* data (i.e. from partial_pack_offset to the end) is then computed and
	160	* returned in partial_pack_sha1.
	161	*
	162	* Note that new_pack_sha1 is updated last, so both new_pack_sha1 and
	163	* partial_pack_sha1 can refer to the same buffer if the caller is not
	164	* interested in the resulting SHA1 of pack data above partial_pack_offset.
	165	*/
8b0eca7c	166	void fixup_pack_header_footer(int pack_fd,
abeb40e5	167	unsigned char *new_pack_sha1,
8b0eca7c	168	const char *pack_name,
abeb40e5 NP	169	uint32_t object_count,
	170	unsigned char *partial_pack_sha1,
	171	off_t partial_pack_offset)
8b0eca7c	172	{
d35825da	173	int aligned_sz, buf_sz = 8 * 1024;
9126f009	174	git_SHA_CTX old_sha1_ctx, new_sha1_ctx;
8b0eca7c DH	175	struct pack_header hdr;
	176	char *buf;
	177
9126f009 NP	178	git_SHA1_Init(&old_sha1_ctx);
9126f009 NP	179	git_SHA1_Init(&new_sha1_ctx);
abeb40e5	180
8b0eca7c	181	if (lseek(pack_fd, 0, SEEK_SET) != 0)
d824cbba	182	die_errno("Failed seeking to start of '%s'", pack_name);
8b0eca7c	183	if (read_in_full(pack_fd, &hdr, sizeof(hdr)) != sizeof(hdr))
d824cbba	184	die_errno("Unable to reread header of '%s'", pack_name);
8b0eca7c	185	if (lseek(pack_fd, 0, SEEK_SET) != 0)
d824cbba	186	die_errno("Failed seeking to start of '%s'", pack_name);
9126f009	187	git_SHA1_Update(&old_sha1_ctx, &hdr, sizeof(hdr));
8b0eca7c	188	hdr.hdr_entries = htonl(object_count);
9126f009	189	git_SHA1_Update(&new_sha1_ctx, &hdr, sizeof(hdr));
8b0eca7c	190	write_or_die(pack_fd, &hdr, sizeof(hdr));
abeb40e5	191	partial_pack_offset -= sizeof(hdr);
8b0eca7c DH	192
8b0eca7c DH	193	buf = xmalloc(buf_sz);
d35825da	194	aligned_sz = buf_sz - sizeof(hdr);
8b0eca7c	195	for (;;) {
abeb40e5	196	ssize_t m, n;
d35825da NP	197	m = (partial_pack_sha1 && partial_pack_offset < aligned_sz) ?
d35825da NP	198	partial_pack_offset : aligned_sz;
abeb40e5	199	n = xread(pack_fd, buf, m);
8b0eca7c DH	200	if (!n)
	201	break;
	202	if (n < 0)
d824cbba	203	die_errno("Failed to checksum '%s'", pack_name);
9126f009	204	git_SHA1_Update(&new_sha1_ctx, buf, n);
abeb40e5	205
d35825da NP	206	aligned_sz -= n;
	207	if (!aligned_sz)
	208	aligned_sz = buf_sz;
	209
abeb40e5 NP	210	if (!partial_pack_sha1)
	211	continue;
	212
9126f009	213	git_SHA1_Update(&old_sha1_ctx, buf, n);
abeb40e5 NP	214	partial_pack_offset -= n;
	215	if (partial_pack_offset == 0) {
	216	unsigned char sha1[20];
9126f009	217	git_SHA1_Final(sha1, &old_sha1_ctx);
abeb40e5 NP	218	if (hashcmp(sha1, partial_pack_sha1) != 0)
	219	die("Unexpected checksum for %s "
	220	"(disk corruption?)", pack_name);
	221	/*
	222	* Now let's compute the SHA1 of the remainder of the
	223	* pack, which also means making partial_pack_offset
	224	* big enough not to matter anymore.
	225	*/
9126f009	226	git_SHA1_Init(&old_sha1_ctx);
abeb40e5 NP	227	partial_pack_offset = ~partial_pack_offset;
	228	partial_pack_offset -= MSB(partial_pack_offset, 1);
	229	}
8b0eca7c DH	230	}
	231	free(buf);
	232
abeb40e5	233	if (partial_pack_sha1)
9126f009 NP	234	git_SHA1_Final(partial_pack_sha1, &old_sha1_ctx);
9126f009 NP	235	git_SHA1_Final(new_pack_sha1, &new_sha1_ctx);
abeb40e5	236	write_or_die(pack_fd, new_pack_sha1, 20);
0c68d386	237	fsync_or_die(pack_fd, pack_name);
8b0eca7c	238	}
106764e6 SP	239
	240	char *index_pack_lockfile(int ip_out)
	241	{
106764e6 SP	242	char packname[46];
	243
	244	/*
6e180cdc	245	* The first thing we expect from index-pack's output
106764e6 SP	246	* is "pack\t%40s\n" or "keep\t%40s\n" (46 bytes) where
	247	* %40s is the newly created pack SHA1 name. In the "keep"
	248	* case, we need it to remove the corresponding .keep file
	249	* later on. If we don't get that then tough luck with it.
	250	*/
c697ad14 HO	251	if (read_in_full(ip_out, packname, 46) == 46 && packname[45] == '\n' &&
c697ad14 HO	252	memcmp(packname, "keep\t", 5) == 0) {
106764e6 SP	253	char path[PATH_MAX];
	254	packname[45] = 0;
	255	snprintf(path, sizeof(path), "%s/pack/pack-%s.keep",
	256	get_object_directory(), packname + 5);
	257	return xstrdup(path);
	258	}
	259	return NULL;
	260	}
f965c525 NP	261
	262	/*
	263	* The per-object header is a pretty dense thing, which is
	264	* - first byte: low four bits are "size", then three bits of "type",
	265	* and the high bit is "size continues".
	266	* - each byte afterwards: low seven bits are size continuation,
	267	* with the high bit being "size continues"
	268	*/
	269	int encode_in_pack_object_header(enum object_type type, uintmax_t size, unsigned char *hdr)
	270	{
	271	int n = 1;
	272	unsigned char c;
	273
	274	if (type < OBJ_COMMIT \|\| type > OBJ_REF_DELTA)
	275	die("bad type %d", type);
	276
	277	c = (type << 4) \| (size & 15);
	278	size >>= 4;
	279	while (size) {
	280	*hdr++ = c \| 0x80;
	281	c = size & 0x7f;
	282	size >>= 7;
	283	n++;
	284	}
	285	*hdr = c;
	286	return n;
	287	}