journal.c

bind-3.2.

	dns_journal_destroy(&j);

	return (result);
}

/**************************************************************************/
/*
 * Miscellaneous accessors.
 */
isc_uint32_t dns_journal_first_serial(dns_journal_t *j) {
	return (j->header.begin.serial);
}

isc_uint32_t dns_journal_last_serial(dns_journal_t *j) {
	return (j->header.end.serial);
}

/**************************************************************************/
/*
 * Iteration support.
 *
 * When serving an outgoing IXFR, we transmit a part the journal starting
 * at the serial number in the IXFR request and ending at the serial
 * number that is current when the IXFR request arrives.  The ending
 * serial number is not necessarily at the end of the journal:
 * the journal may grow while the IXFR is in progress, but we stop
 * when we reach the serial number that was current when the IXFR started.
 */

static isc_result_t read_one_rr(dns_journal_t *j);

/*
 * Make sure the buffer 'b' is has at least 'size' bytes
 * allocated, and clear it.
 *
 * Requires:
 *	Either b->base is NULL, or it points to b->length bytes of memory
 *	previously allocated by isc_mem_get().
 */

static isc_result_t
size_buffer(isc_mem_t *mctx, isc_buffer_t *b, unsigned size) {
	if (b->length < size) {
		void *mem = isc_mem_get(mctx, size);
		if (mem == NULL)
			return (ISC_R_NOMEMORY);
		if (b->base != NULL)
			isc_mem_put(mctx, b->base, b->length);
		b->base = mem;
		b->length = size;
	}
	isc_buffer_clear(b);
	return (ISC_R_SUCCESS);
}

isc_result_t
dns_journal_iter_init(dns_journal_t *j,
		      isc_uint32_t begin_serial, isc_uint32_t end_serial)
{
	isc_result_t result;

	CHECK(journal_find(j, begin_serial, &j->it.bpos));
	INSIST(j->it.bpos.serial == begin_serial);

	CHECK(journal_find(j, end_serial, &j->it.epos));
	INSIST(j->it.epos.serial == end_serial);

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


isc_result_t
dns_journal_first_rr(dns_journal_t *j) {
	isc_result_t result;

	/*
	 * Seek to the beginning of the first transaction we are
	 * interested in.
	 */
	CHECK(journal_seek(j, j->it.bpos.offset));
	j->it.current_serial = j->it.bpos.serial;

	j->it.xsize = 0;  /* We have no transaction data yet... */
	j->it.xpos = 0;	  /* ...and haven't used any of it. */

	return (read_one_rr(j));

 failure:
	return (result);
}

static isc_result_t
read_one_rr(dns_journal_t *j) {
	isc_result_t result;

	dns_rdatatype_t rdtype;
	dns_rdataclass_t rdclass;
	unsigned int rdlen;
	isc_uint32_t ttl;
	journal_xhdr_t xhdr;
	journal_rrhdr_t rrhdr;

	INSIST(j->offset <= j->it.epos.offset);
	if (j->offset == j->it.epos.offset)
		return (ISC_R_NOMORE);
	if (j->it.xpos == j->it.xsize) {
		/*
		 * We are at a transaction boundary.
		 * Read another transaction header.
		 */
		CHECK(journal_read_xhdr(j, &xhdr));
		if (xhdr.size == 0) {
			isc_log_write(JOURNAL_COMMON_LOGARGS, ISC_LOG_ERROR,
				      "journal corrupt: empty transaction");
			FAIL(ISC_R_UNEXPECTED);
		}
		if (xhdr.serial0 != j->it.current_serial) {
			isc_log_write(JOURNAL_COMMON_LOGARGS, ISC_LOG_ERROR,
					 "%s: journal file corrupt: "
					 "expected serial %u, got %u",
					 j->filename,
					 j->it.current_serial, xhdr.serial0);
			FAIL(ISC_R_UNEXPECTED);
		}
		j->it.xsize = xhdr.size;
		j->it.xpos = 0;
	}
	/*
	 * Read an RR.
	 */
	result = journal_read_rrhdr(j, &rrhdr);
	/*
	 * Perform a sanity check on the journal RR size.
	 * The smallest possible RR has a 1-byte owner name
	 * and a 10-byte header.  The largest possible
	 * RR has 65535 bytes of data, a header, and a maximum-
	 * size owner name, well below 70 k total.
	 */
	if (rrhdr.size < 1+10 || rrhdr.size > 70000) {
		isc_log_write(JOURNAL_COMMON_LOGARGS, ISC_LOG_ERROR,
				 "%s: journal corrupt: impossible RR size "
				 "(%d bytes)", j->filename, rrhdr.size);
		FAIL(ISC_R_UNEXPECTED);
	}

	CHECK(size_buffer(j->mctx, &j->it.source, rrhdr.size));
	CHECK(journal_read(j, j->it.source.base, rrhdr.size));
	isc_buffer_add(&j->it.source, rrhdr.size);

	/*
	 * The target buffer is made the same size
	 * as the source buffer, with the assumption that when
	 * no compression in present, the output of dns_*_fromwire()
	 * is no larger than the input.
	 */
	CHECK(size_buffer(j->mctx, &j->it.target, rrhdr.size));

	/*
	 * Parse the owner name.  We don't know where it
	 * ends yet, so we make the entire "remaining"
	 * part of the buffer "active".
	 */
	isc_buffer_setactive(&j->it.source,
			     j->it.source.used - j->it.source.current);
	CHECK(dns_name_fromwire(&j->it.name, &j->it.source,
				&j->it.dctx, ISC_FALSE, &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,
				 ISC_FALSE, &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]);

	CHECK(dns_journal_open(mctx, journal_filename, ISC_TRUE, &journal));

	CHECK(dns_db_createiterator(db[0], ISC_FALSE, &dbit[0]));
	CHECK(dns_db_createiterator(db[1], ISC_FALSE, &dbit[1]));

	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));
	dns_diff_clear(&resultdiff);

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