journal.c

bind 9.3结合mysql数据库

				&j->it.dctx, 0, &j->it.target));

	/*
	 * Check that the RR header is there, and parse it.
	 */
	if (isc_buffer_remaininglength(&j->it.source) < 10)
		FAIL(DNS_R_FORMERR);

	rdtype = isc_buffer_getuint16(&j->it.source);
	rdclass = isc_buffer_getuint16(&j->it.source);
	ttl = isc_buffer_getuint32(&j->it.source);
	rdlen = isc_buffer_getuint16(&j->it.source);

	/*
	 * Parse the rdata.
	 */
	isc_buffer_setactive(&j->it.source, rdlen);
	dns_rdata_reset(&j->it.rdata);
	CHECK(dns_rdata_fromwire(&j->it.rdata, rdclass,
				 rdtype, &j->it.source, &j->it.dctx,
				 0, &j->it.target));
	j->it.ttl = ttl;

	j->it.xpos += sizeof(journal_rawrrhdr_t) + rrhdr.size;
	if (rdtype == dns_rdatatype_soa) {
		/* XXX could do additional consistency checks here */
		j->it.current_serial = dns_soa_getserial(&j->it.rdata);
	}

	result = ISC_R_SUCCESS;

 failure:
	j->it.result = result;
	return (result);
}

isc_result_t
dns_journal_next_rr(dns_journal_t *j) {
	j->it.result = read_one_rr(j);
	return (j->it.result);
}

void
dns_journal_current_rr(dns_journal_t *j, dns_name_t **name, isc_uint32_t *ttl,
		   dns_rdata_t **rdata)
{
	REQUIRE(j->it.result == ISC_R_SUCCESS);
	*name = &j->it.name;
	*ttl = j->it.ttl;
	*rdata = &j->it.rdata;
}

/**************************************************************************/
/*
 * Generating diffs from databases
 */

/*
 * Construct a diff containing all the RRs at the current name of the
 * database iterator 'dbit' in database 'db', version 'ver'.
 * Set '*name' to the current name, and append the diff to 'diff'.
 * All new tuples will have the operation 'op'.
 *
 * Requires: 'name' must have buffer large enough to hold the name.
 * Typically, a dns_fixedname_t would be used.
 */
static isc_result_t
get_name_diff(dns_db_t *db, dns_dbversion_t *ver, isc_stdtime_t now,
	      dns_dbiterator_t *dbit, dns_name_t *name, dns_diffop_t op,
	      dns_diff_t *diff)
{
	isc_result_t result;
	dns_dbnode_t *node = NULL;
	dns_rdatasetiter_t *rdsiter = NULL;
	dns_difftuple_t *tuple = NULL;

	result = dns_dbiterator_current(dbit, &node, name);
	if (result != ISC_R_SUCCESS)
		return (result);

	result = dns_db_allrdatasets(db, node, ver, now, &rdsiter);
	if (result != ISC_R_SUCCESS)
		goto cleanup_node;

	for (result = dns_rdatasetiter_first(rdsiter);
	     result == ISC_R_SUCCESS;
	     result = dns_rdatasetiter_next(rdsiter))
	{
		dns_rdataset_t rdataset;

		dns_rdataset_init(&rdataset);
		dns_rdatasetiter_current(rdsiter, &rdataset);

		for (result = dns_rdataset_first(&rdataset);
		     result == ISC_R_SUCCESS;
		     result = dns_rdataset_next(&rdataset))
		{
			dns_rdata_t rdata = DNS_RDATA_INIT;
			dns_rdataset_current(&rdataset, &rdata);
			result = dns_difftuple_create(diff->mctx, op, name,
						      rdataset.ttl, &rdata,
						      &tuple);
			if (result != ISC_R_SUCCESS) {
				dns_rdataset_disassociate(&rdataset);
				goto cleanup_iterator;
			}
			dns_diff_append(diff, &tuple);
		}
		dns_rdataset_disassociate(&rdataset);
		if (result != ISC_R_NOMORE)
			goto cleanup_iterator;
	}
	if (result != ISC_R_NOMORE)
		goto cleanup_iterator;

	result = ISC_R_SUCCESS;

 cleanup_iterator:
	dns_rdatasetiter_destroy(&rdsiter);

 cleanup_node:
	dns_db_detachnode(db, &node);

	return (result);
}

/*
 * Comparison function for use by dns_diff_subtract when sorting
 * the diffs to be subtracted.  The sort keys are the rdata type
 * and the rdata itself.  The owner name is ignored, because
 * it is known to be the same for all tuples.
 */
static int
rdata_order(const void *av, const void *bv) {
	dns_difftuple_t const * const *ap = av;
	dns_difftuple_t const * const *bp = bv;
	dns_difftuple_t const *a = *ap;
	dns_difftuple_t const *b = *bp;
	int r;
	r = (b->rdata.type - a->rdata.type);
	if (r != 0)
		return (r);
	r = dns_rdata_compare(&a->rdata, &b->rdata);
	return (r);
}

static isc_result_t
dns_diff_subtract(dns_diff_t diff[2], dns_diff_t *r) {
	isc_result_t result;
	dns_difftuple_t *p[2];
	int i, t;
	CHECK(dns_diff_sort(&diff[0], rdata_order));
	CHECK(dns_diff_sort(&diff[1], rdata_order));

	for (;;) {
		p[0] = ISC_LIST_HEAD(diff[0].tuples);
		p[1] = ISC_LIST_HEAD(diff[1].tuples);
		if (p[0] == NULL && p[1] == NULL)
			break;

		for (i = 0; i < 2; i++)
			if (p[!i] == NULL) {
				ISC_LIST_UNLINK(diff[i].tuples, p[i], link);
				ISC_LIST_APPEND(r->tuples, p[i], link);
				goto next;
			}
		t = rdata_order(&p[0], &p[1]);
		if (t < 0) {
			ISC_LIST_UNLINK(diff[0].tuples, p[0], link);
			ISC_LIST_APPEND(r->tuples, p[0], link);
			goto next;
		}
		if (t > 0) {
			ISC_LIST_UNLINK(diff[1].tuples, p[1], link);
			ISC_LIST_APPEND(r->tuples, p[1], link);
			goto next;
		}
		INSIST(t == 0);
		/*
		 * Identical RRs in both databases; skip them both.
		 */
		for (i = 0; i < 2; i++) {
			ISC_LIST_UNLINK(diff[i].tuples, p[i], link);
			dns_difftuple_free(&p[i]);
		}
	next: ;
	}
	result = ISC_R_SUCCESS;
 failure:
	return (result);
}

/*
 * Compare the databases 'dba' and 'dbb' and generate a journal
 * entry containing the changes to make 'dba' from 'dbb' (note
 * the order).  This journal entry will consist of a single,
 * possibly very large transaction.
 */

isc_result_t
dns_db_diff(isc_mem_t *mctx,
	    dns_db_t *dba, dns_dbversion_t *dbvera,
	    dns_db_t *dbb, dns_dbversion_t *dbverb,
	    const char *journal_filename)
{
	dns_db_t *db[2];
	dns_dbversion_t *ver[2];
	dns_dbiterator_t *dbit[2] = { NULL, NULL };
	isc_boolean_t have[2] = { ISC_FALSE, ISC_FALSE };
	dns_fixedname_t fixname[2];
	isc_result_t result, itresult[2];
	dns_diff_t diff[2], resultdiff;
	int i, t;
	dns_journal_t *journal = NULL;

	db[0] = dba, db[1] = dbb;
	ver[0] = dbvera, ver[1] = dbverb;

	dns_diff_init(mctx, &diff[0]);
	dns_diff_init(mctx, &diff[1]);
	dns_diff_init(mctx, &resultdiff);

	dns_fixedname_init(&fixname[0]);
	dns_fixedname_init(&fixname[1]);

	result = dns_journal_open(mctx, journal_filename, ISC_TRUE, &journal);
	if (result != ISC_R_SUCCESS)
		return (result);

	result = dns_db_createiterator(db[0], ISC_FALSE, &dbit[0]);
	if (result != ISC_R_SUCCESS)
		goto cleanup_journal;
	result = dns_db_createiterator(db[1], ISC_FALSE, &dbit[1]);
	if (result != ISC_R_SUCCESS)
		goto cleanup_interator0;

	itresult[0] = dns_dbiterator_first(dbit[0]);
	itresult[1] = dns_dbiterator_first(dbit[1]);

	for (;;) {
		for (i = 0; i < 2; i++) {
			if (! have[i] && itresult[i] == ISC_R_SUCCESS) {
				CHECK(get_name_diff(db[i], ver[i], 0, dbit[i],
					    dns_fixedname_name(&fixname[i]),
					    i == 0 ?
					    DNS_DIFFOP_ADD :
					    DNS_DIFFOP_DEL,
					    &diff[i]));
				itresult[i] = dns_dbiterator_next(dbit[i]);
				have[i] = ISC_TRUE;
			}
		}

		if (! have[0] && ! have[1]) {
			INSIST(ISC_LIST_EMPTY(diff[0].tuples));
			INSIST(ISC_LIST_EMPTY(diff[1].tuples));
			break;
		}

		for (i = 0; i < 2; i++) {
			if (! have[!i]) {
				ISC_LIST_APPENDLIST(resultdiff.tuples,
						    diff[i].tuples, link);
				INSIST(ISC_LIST_EMPTY(diff[i].tuples));
				have[i] = ISC_FALSE;
				goto next;
			}
		}

		t = dns_name_compare(dns_fixedname_name(&fixname[0]),
				     dns_fixedname_name(&fixname[1]));
		if (t < 0) {
			ISC_LIST_APPENDLIST(resultdiff.tuples,
					    diff[0].tuples, link);
			INSIST(ISC_LIST_EMPTY(diff[0].tuples));
			have[0] = ISC_FALSE;
			continue;
		}
		if (t > 0) {
			ISC_LIST_APPENDLIST(resultdiff.tuples,
					    diff[1].tuples, link);
			INSIST(ISC_LIST_EMPTY(diff[1].tuples));
			have[1] = ISC_FALSE;
			continue;
		}
		INSIST(t == 0);
		CHECK(dns_diff_subtract(diff, &resultdiff));
		INSIST(ISC_LIST_EMPTY(diff[0].tuples));
		INSIST(ISC_LIST_EMPTY(diff[1].tuples));
		have[0] = have[1] = ISC_FALSE;
	next: ;
	}
	if (itresult[0] != ISC_R_NOMORE)
		FAIL(itresult[0]);
	if (itresult[1] != ISC_R_NOMORE)
		FAIL(itresult[1]);

	if (ISC_LIST_EMPTY(resultdiff.tuples)) {
		isc_log_write(JOURNAL_DEBUG_LOGARGS(3), "no changes");
	} else {
		CHECK(dns_journal_write_transaction(journal, &resultdiff));
	}
	INSIST(ISC_LIST_EMPTY(diff[0].tuples));
	INSIST(ISC_LIST_EMPTY(diff[1].tuples));

 failure:
	dns_diff_clear(&resultdiff);
	dns_dbiterator_destroy(&dbit[1]);
 cleanup_interator0:
	dns_dbiterator_destroy(&dbit[0]);
 cleanup_journal:
	dns_journal_destroy(&journal);
	return (result);
}

isc_result_t
dns_journal_compact(isc_mem_t *mctx, char *filename, isc_uint32_t serial,
		    isc_uint32_t target_size)
{
	unsigned int i;
	journal_pos_t best_guess;
	journal_pos_t current_pos;
	dns_journal_t *j = NULL;
	journal_rawheader_t rawheader;
	unsigned int copy_length;
	unsigned int len;
	char *buf = NULL;
	unsigned int size = 0;
	isc_result_t result;
	unsigned int indexend;

	CHECK(journal_open(mctx, filename, ISC_TRUE, ISC_FALSE, &j));

	if (JOURNAL_EMPTY(&j->header)) {
		dns_journal_destroy(&j);
		return (ISC_R_SUCCESS);
	}
		
	if (DNS_SERIAL_GT(j->header.begin.serial, serial) ||
	    DNS_SERIAL_GT(serial, j->header.end.serial)) {
		dns_journal_destroy(&j);
		return (ISC_R_RANGE);
	}

	/*
	 * Cope with very small target sizes.
	 */
	indexend = sizeof(journal_rawheader_t) +
		   j->header.index_size * sizeof(journal_rawpos_t);
	if (target_size < indexend * 2)
		target_size = target_size/2 + indexend;

	/*
	 * See if there is any work to do.
	 */
	if ((isc_uint32_t) j->header.end.offset < target_size) {
		dns_journal_destroy(&j);
		return (ISC_R_SUCCESS);
	}
	
	/*
	 * Remove overhead so space test below can succeed.
	 */
	if (target_size >= indexend)
		target_size -= indexend;

	/*
	 * Find if we can create enough free space.
	 */
	best_guess = j->header.begin;
	for (i = 0; i < j->header.index_size; i++) {
		if (POS_VALID(j->index[i]) &&
		    DNS_SERIAL_GE(serial, j->index[i].serial) &&
		    ((isc_uint32_t)(j->header.end.offset - j->index[i].offset)
		     >= target_size / 2) &&
		    j->index[i].offset > best_guess.offset)
			best_guess = j->index[i];
	}

	current_pos = best_guess;
	while (current_pos.serial != serial) {
		CHECK(journal_next(j, &current_pos));
		if (current_pos.serial == j->header.end.serial)
			break;

		if (DNS_SERIAL_GE(serial, current_pos.serial) &&
		   ((isc_uint32_t)(j->header.end.offset - current_pos.offset)
		     >= (target_size / 2)) &&
		    current_pos.offset > best_guess.offset)
			best_guess = current_pos;
		else
			break;
	}

	INSIST(best_guess.serial != j->header.end.serial);
	if (best_guess.serial != serial)
		CHECK(journal_next(j, &best_guess));

	/*
	 * Enough space to proceed?
	 */
	if ((isc_uint32_t) (j->header.end.offset - best_guess.offset) >
	     (isc_uint32_t) (best_guess.offset - indexend)) {
		dns_journal_destroy(&j);
		return (ISC_R_NOSPACE);
	}

	copy_length = j->header.end.offset - best_guess.offset;

	/*
	 * Invalidate entire index, will be rebuilt at end.
	 */
	for (i = 0; i < j->header.index_size; i++) {
		if (POS_VALID(j->index[i]))
			POS_INVALIDATE(j->index[i]);
	}

	/*
	 * Convert the index into on-disk format and write
	 * it to disk.
	 */
	CHECK(index_to_disk(j));
	CHECK(journal_fsync(j));

	/*
	 * Update the journal header.
	 */
	if (copy_length == 0) {
		j->header.begin.serial = 0;
		j->header.end.serial = 0;
		j->header.begin.offset = 0;
		j->header.end.offset = 0;
	} else {
		j->header.begin = best_guess;
	}
	journal_header_encode(&j->header, &rawheader);
	CHECK(journal_seek(j, 0));
	CHECK(journal_write(j, &rawheader, sizeof(rawheader)));
	CHECK(journal_fsync(j));

	if (copy_length != 0) {
		/*
		 * Copy best_guess to end into space just freed.
		 */
		size = 64*1024;
		if (copy_length < size)
			size = copy_length;
		buf = isc_mem_get(mctx, size);
		if (buf == NULL) {
			result = ISC_R_NOMEMORY;
			goto failure;
		}
	
		for (i = 0; i < copy_length; i += size) {
			len = (copy_length - i) > size ? size :
							 (copy_length - i);
			CHECK(journal_seek(j, best_guess.offset + i));
			CHECK(journal_read(j, buf, len));
			CHECK(journal_seek(j, indexend + i));
			CHECK(journal_write(j, buf, len));
		}

		CHECK(journal_fsync(j));

		/*
		 * Compute new header.
		 */
		j->header.begin.offset = indexend;
		j->header.end.offset = indexend + copy_length;
		/*
		 * Update the journal header.
		 */
		journal_header_encode(&j->header, &rawheader);
		CHECK(journal_seek(j, 0));
		CHECK(journal_write(j, &rawheader, sizeof(rawheader)));
		CHECK(journal_fsync(j));

		/*
		 * Build new index.
		 */
		current_pos = j->header.begin;
		while (current_pos.serial != j->header.end.serial) {
			index_add(j, &current_pos);
			CHECK(journal_next(j, &current_pos));
		}

		/*
		 * Write index.
		 */
		CHECK(index_to_disk(j));
		CHECK(journal_fsync(j));

		indexend = j->header.end.offset;
	}
	dns_journal_destroy(&j);
	(void)isc_file_truncate(filename, (isc_offset_t)indexend);
	result = ISC_R_SUCCESS;

 failure:
	if (buf != NULL)
		isc_mem_put(mctx, buf, size);
	if (j != NULL)
		dns_journal_destroy(&j);
	return (result);
}

static isc_result_t
index_to_disk(dns_journal_t *j) {
	isc_result_t result = ISC_R_SUCCESS;

	if (j->header.index_size != 0) {
		unsigned int i;
		unsigned char *p;
		unsigned int rawbytes;

		rawbytes = j->header.index_size * sizeof(journal_rawpos_t);

		p = j->rawindex;
		for (i = 0; i < j->header.index_size; i++) {
			encode_uint32(j->index[i].serial, p);
			p += 4;
			encode_uint32(j->index[i].offset, p);
			p += 4;
		}
		INSIST(p == j->rawindex + rawbytes);

		CHECK(journal_seek(j, sizeof(journal_rawheader_t)));
		CHECK(journal_write(j, j->rawindex, rawbytes));
	}
failure:
	return (result);
}