Update to nsd 4.1.14

OK sthen@

9 files changed, 105 insertions, 48 deletions

diff --git a/usr.sbin/nsd/configlexer.lex b/usr.sbin/nsd/configlexer.lex
index e6b26f947e1..42dbd152f16 100644
--- a/usr.sbin/nsd/configlexer.lex
+++ b/usr.sbin/nsd/configlexer.lex
@@ -66,15 +66,6 @@ static void config_start_include(const char* filename)
 		c_error_msg("include %s: malloc failure", filename);
 		return;
 	}
-	if (cfg_parser->chroot) {
-		int l = strlen(cfg_parser->chroot); /* chroot has trailing slash */
-		if (strncmp(cfg_parser->chroot, filename, l) != 0) {
-			c_error_msg("include file '%s' is not relative to chroot '%s'",
-				filename, cfg_parser->chroot);
-			return;
-		}
-		filename += l - 1; /* strip chroot without trailing slash */
-	}
 	nm = strdup(filename);
 	if(!nm) {
 		c_error_msg("include %s: strdup failure", filename);
@@ -103,12 +94,23 @@ static void config_start_include(const char* filename)
 
 static void config_start_include_glob(const char* filename)
 {
-	 /* check for wildcards */
+	/* check for wildcards */
 #ifdef HAVE_GLOB
-	 glob_t g;
-	 size_t i;
-	 int r, flags;
-	 if(!(!strchr(filename, '*') && !strchr(filename, '?') &&
+	glob_t g;
+	size_t i;
+	int r, flags;
+#endif /* HAVE_GLOB */
+	if (cfg_parser->chroot) {
+		int l = strlen(cfg_parser->chroot); /* chroot has trailing slash */
+		if (strncmp(cfg_parser->chroot, filename, l) != 0) {
+			c_error_msg("include file '%s' is not relative to chroot '%s'",
+				filename, cfg_parser->chroot);
+			return;
+		}
+		filename += l - 1; /* strip chroot without trailing slash */
+	}
+#ifdef HAVE_GLOB
+	if(!(!strchr(filename, '*') && !strchr(filename, '?') &&
 		 !strchr(filename, '[') && !strchr(filename, '{') &&
 		 !strchr(filename, '~'))) {
 		 flags = 0
@@ -141,9 +143,9 @@ static void config_start_include_glob(const char* filename)
 		}
 		globfree(&g);
 		return;
-	 }
+	}
 #endif /* HAVE_GLOB */
-	 config_start_include(filename);
+	config_start_include(filename);
 }
 
 static void config_end_include(void)
diff --git a/usr.sbin/nsd/configure.ac b/usr.sbin/nsd/configure.ac
index e2aca4d531e..029193574f0 100644
--- a/usr.sbin/nsd/configure.ac
+++ b/usr.sbin/nsd/configure.ac
@@ -4,7 +4,7 @@ dnl
 
 sinclude(acx_nlnetlabs.m4)
 
-AC_INIT(NSD,4.1.13,nsd-bugs@nlnetlabs.nl)
+AC_INIT(NSD,4.1.14,nsd-bugs@nlnetlabs.nl)
 AC_CONFIG_HEADER([config.h])
 
 CFLAGS="$CFLAGS"
diff --git a/usr.sbin/nsd/radtree.c b/usr.sbin/nsd/radtree.c
index 8ca4a5bd5ba..fd003795a4a 100644
--- a/usr.sbin/nsd/radtree.c
+++ b/usr.sbin/nsd/radtree.c
@@ -636,11 +636,13 @@ radnode_cleanup_onechild(struct region* region, struct radnode* n,
 		return 0;
 	}
 	/* we know that .str and join are malloced, thus aligned */
-	memcpy(join, par->array[pidx].str, par->array[pidx].len);
+	if(par->array[pidx].str)
+	    memcpy(join, par->array[pidx].str, par->array[pidx].len);
 	/* the array lookup is gone, put its character in the lookup string*/
 	join[par->array[pidx].len] = child->pidx + n->offset;
 	/* but join+len may not be aligned */
-	memmove(join+par->array[pidx].len+1, n->array[0].str, n->array[0].len);
+	if(n->array[0].str)
+	    memmove(join+par->array[pidx].len+1, n->array[0].str, n->array[0].len);
 	region_recycle(region, par->array[pidx].str, par->array[pidx].len);
 	par->array[pidx].str = join;
 	par->array[pidx].len = joinlen;
diff --git a/usr.sbin/nsd/remote.c b/usr.sbin/nsd/remote.c
index d961dbf4146..baa9985389a 100644
--- a/usr.sbin/nsd/remote.c
+++ b/usr.sbin/nsd/remote.c
@@ -1469,7 +1469,7 @@ repat_interrupt_zones(xfrd_state_t* xfrd, nsd_options_t* newopt)
 			xz->master = 0;
 			xz->master_num = 0;
 			xz->next_master = -1;
-			xz->round_num = 0; /* fresh set of retries */
+			xz->round_num = -1; /* fresh set of retries */
 		}
 	}
 	/* if notify list changed:
diff --git a/usr.sbin/nsd/xfrd-disk.c b/usr.sbin/nsd/xfrd-disk.c
index c724a2c9cc6..a86240d4a57 100644
--- a/usr.sbin/nsd/xfrd-disk.c
+++ b/usr.sbin/nsd/xfrd-disk.c
@@ -165,8 +165,15 @@ xfrd_read_state(struct xfrd_state* xfrd)
 		region_destroy(tempregion);
 		return;
 	}
-	if(!xfrd_read_check_str(in, XFRD_FILE_MAGIC) ||
-	   !xfrd_read_check_str(in, "filetime:") ||
+	if(!xfrd_read_check_str(in, XFRD_FILE_MAGIC)) {
+		/* older file version; reset everything */
+		DEBUG(DEBUG_XFRD,1, (LOG_INFO, "xfrd: file %s is old version. refreshing all zones.",
+			statefile));
+		fclose(in);
+		region_destroy(tempregion);
+		return;
+	}
+	if(!xfrd_read_check_str(in, "filetime:") ||
 	   !xfrd_read_i32(in, &filetime) ||
 	   (time_t)filetime > xfrd_time()+15 ||
 	   !xfrd_read_check_str(in, "numzones:") ||
@@ -183,7 +190,7 @@ xfrd_read_state(struct xfrd_state* xfrd)
 		char *p;
 		xfrd_zone_t* zone;
 		const dname_type* dname;
-		uint32_t state, masnum, nextmas, round_num, timeout;
+		uint32_t state, masnum, nextmas, round_num, timeout, backoff;
 		xfrd_soa_t soa_nsd_read, soa_disk_read, soa_notified_read;
 		time_t soa_nsd_acquired_read,
 			soa_disk_acquired_read, soa_notified_acquired_read;
@@ -214,6 +221,8 @@ xfrd_read_state(struct xfrd_state* xfrd)
 		   !xfrd_read_i32(in, &round_num) ||
 		   !xfrd_read_check_str(in, "next_timeout:") ||
 		   !xfrd_read_i32(in, &timeout) ||
+		   !xfrd_read_check_str(in, "backoff:") ||
+		   !xfrd_read_i32(in, &backoff) ||
 		   !xfrd_read_state_soa(in, "soa_nsd_acquired:", "soa_nsd:",
 			&soa_nsd_read, &soa_nsd_acquired_read) ||
 		   !xfrd_read_state_soa(in, "soa_disk_acquired:", "soa_disk:",
@@ -249,6 +258,7 @@ xfrd_read_state(struct xfrd_state* xfrd)
 		zone->round_num = round_num;
 		zone->timeout.tv_sec = timeout;
 		zone->timeout.tv_usec = 0;
+		zone->fresh_xfr_timeout = backoff*XFRD_TRANSFER_TIMEOUT_START;
 
 		/* read the zone OK, now set the master properly */
 		zone->master = acl_find_num(zone->zone_options->pattern->
@@ -296,10 +306,14 @@ xfrd_read_state(struct xfrd_state* xfrd)
 			zone->state = state;
 			xfrd_set_timer(zone, timeout);
 		}	
-		if(zone->soa_nsd_acquired == 0 && soa_nsd_acquired_read == 0 &&
-			soa_disk_acquired_read == 0) {
-			/* continue expon backoff where we were + check now */
-			zone->fresh_xfr_timeout = timeout;
+		if((zone->soa_nsd_acquired == 0 && soa_nsd_acquired_read == 0 &&
+			soa_disk_acquired_read == 0) ||
+			(zone->state != xfrd_zone_ok && timeout != 0)) {
+			/* but don't check now, because that would mean a
+			 * storm of attempts on some master servers */
+			xfrd_deactivate_zone(zone);
+			zone->state = state;
+			xfrd_set_timer(zone, timeout);
 		}
 
 		/* handle as an incoming SOA. */
@@ -318,7 +332,8 @@ xfrd_read_state(struct xfrd_state* xfrd)
 		{
 			xfrd_send_expire_notification(zone);
 		}
-		xfrd_handle_incoming_soa(zone, &incoming_soa, incoming_acquired);
+		if(incoming_acquired != 0)
+			xfrd_handle_incoming_soa(zone, &incoming_soa, incoming_acquired);
 	}
 
 	if(!xfrd_read_check_str(in, XFRD_FILE_MAGIC)) {
@@ -477,6 +492,7 @@ xfrd_write_state(struct xfrd_state* xfrd)
 			neato_timeout(out, "\t# =", zone->timeout.tv_sec);
 		}
 		fprintf(out, "\n");
+		fprintf(out, "\tbackoff: %d\n", zone->fresh_xfr_timeout/XFRD_TRANSFER_TIMEOUT_START);
 		xfrd_write_state_soa(out, "soa_nsd", &zone->soa_nsd,
 			zone->soa_nsd_acquired, zone->apex);
 		xfrd_write_state_soa(out, "soa_disk", &zone->soa_disk,
diff --git a/usr.sbin/nsd/xfrd-disk.h b/usr.sbin/nsd/xfrd-disk.h
index b7e2d10b237..acd4e4c8333 100644
--- a/usr.sbin/nsd/xfrd-disk.h
+++ b/usr.sbin/nsd/xfrd-disk.h
@@ -14,7 +14,7 @@ struct xfrd_state;
 struct nsd;
 
 /* magic string to identify xfrd state file */
-#define XFRD_FILE_MAGIC "NSDXFRD1"
+#define XFRD_FILE_MAGIC "NSDXFRD2"
 
 /* read from state file as many zones as possible (until error/eof).*/
 void xfrd_read_state(struct xfrd_state* xfrd);
diff --git a/usr.sbin/nsd/xfrd-tcp.c b/usr.sbin/nsd/xfrd-tcp.c
index a4a7a23fe29..506b2520589 100644
--- a/usr.sbin/nsd/xfrd-tcp.c
+++ b/usr.sbin/nsd/xfrd-tcp.c
@@ -490,6 +490,14 @@ xfrd_tcp_open(xfrd_tcp_set_t* set, struct xfrd_tcp_pipeline* tp,
 	int fd, family, conn;
 	struct timeval tv;
 	assert(zone->tcp_conn != -1);
+
+	/* if there is no next master, fallback to use the first one */
+	/* but there really should be a master set */
+	if(!zone->master) {
+		zone->master = zone->zone_options->pattern->request_xfr;
+		zone->master_num = 0;
+	}
+
 	DEBUG(DEBUG_XFRD,1, (LOG_INFO, "xfrd: zone %s open tcp conn to %s",
 		zone->apex_str, zone->master->ip_address_spec));
 	tp->tcp_r->is_reading = 1;
@@ -879,7 +887,7 @@ xfrd_tcp_read(struct xfrd_tcp_pipeline* tp)
 			assert(zone->round_num == -1);
 			break;
 		case xfrd_packet_notimpl:
-			zone->master->ixfr_disabled = time(NULL);
+			xfrd_disable_ixfr(zone);
 			xfrd_tcp_release(xfrd->tcp_set, zone);
 			/* query next server */
 			xfrd_make_request(zone);
diff --git a/usr.sbin/nsd/xfrd.c b/usr.sbin/nsd/xfrd.c
index a0b34446ae5..415602f014b 100644
--- a/usr.sbin/nsd/xfrd.c
+++ b/usr.sbin/nsd/xfrd.c
@@ -30,13 +30,11 @@
 #include "ipc.h"
 #include "remote.h"
 
-#define XFRD_TRANSFER_TIMEOUT_START 10 /* empty zone timeout is between x and 2*x seconds */
-#define XFRD_TRANSFER_TIMEOUT_MAX 86400 /* empty zone timeout max expbackoff */
 #define XFRD_UDP_TIMEOUT 10 /* seconds, before a udp request times out */
 #define XFRD_NO_IXFR_CACHE 172800 /* 48h before retrying ixfr's after notimpl */
 #define XFRD_LOWERBOUND_REFRESH 1 /* seconds, smallest refresh timeout */
 #define XFRD_LOWERBOUND_RETRY 1 /* seconds, smallest retry timeout */
-#define XFRD_MAX_ROUNDS 3 /* max number of rounds along the masters */
+#define XFRD_MAX_ROUNDS 1 /* max number of rounds along the masters */
 #define XFRD_TSIG_MAX_UNSIGNED 103 /* max number of packets without tsig in a tcp stream. */
 			/* rfc recommends 100, +3 for offbyone errors/interoperability. */
 #define XFRD_CHILD_REAP_TIMEOUT 60 /* seconds to wakeup and reap lost children */
@@ -412,7 +410,7 @@ xfrd_init_slave_zone(xfrd_state_t* xfrd, zone_options_t* zone_opt)
 	xzone->state = xfrd_zone_refreshing;
 	xzone->zone_options = zone_opt;
 	/* first retry will use first master */
-	xzone->master = 0;
+	xzone->master = xzone->zone_options->pattern->request_xfr;
 	xzone->master_num = 0;
 	xzone->next_master = 0;
 	xzone->fresh_xfr_timeout = XFRD_TRANSFER_TIMEOUT_START;
@@ -727,11 +725,24 @@ static void
 xfrd_set_timer_retry(xfrd_zone_t* zone)
 {
 	time_t set_retry;
+	int mult;
+	/* perform exponential backoff in all the cases */
+	if(zone->fresh_xfr_timeout == 0)
+		zone->fresh_xfr_timeout = XFRD_TRANSFER_TIMEOUT_START;
+	else {
+		/* exponential backoff - some master data in zones is paid-for
+		   but non-working, and will not get fixed. */
+		zone->fresh_xfr_timeout *= 2;
+		if(zone->fresh_xfr_timeout > XFRD_TRANSFER_TIMEOUT_MAX)
+			zone->fresh_xfr_timeout = XFRD_TRANSFER_TIMEOUT_MAX;
+	}
+	/* exponential backoff multiplier, starts at 1, backs off */
+	mult = zone->fresh_xfr_timeout / XFRD_TRANSFER_TIMEOUT_START;
+	if(mult == 0) mult = 1;
+
 	/* set timer for next retry or expire timeout if earlier. */
 	if(zone->soa_disk_acquired == 0) {
 		/* if no information, use reasonable timeout */
-		if(zone->fresh_xfr_timeout == 0)
-			zone->fresh_xfr_timeout = XFRD_TRANSFER_TIMEOUT_START;
 #ifdef HAVE_ARC4RANDOM_UNIFORM
 		xfrd_set_timer(zone, zone->fresh_xfr_timeout
 			+ arc4random_uniform(zone->fresh_xfr_timeout));
@@ -742,16 +753,12 @@ xfrd_set_timer_retry(xfrd_zone_t* zone)
 		xfrd_set_timer(zone, zone->fresh_xfr_timeout
 			+ random()%zone->fresh_xfr_timeout);
 #endif
-		/* exponential backoff - some master data in zones is paid-for
-		   but non-working, and will not get fixed. */
-		zone->fresh_xfr_timeout *= 2;
-		if(zone->fresh_xfr_timeout > XFRD_TRANSFER_TIMEOUT_MAX)
-			zone->fresh_xfr_timeout = XFRD_TRANSFER_TIMEOUT_MAX;
 	} else if(zone->state == xfrd_zone_expired ||
-		xfrd_time() + (time_t)ntohl(zone->soa_disk.retry) <
+		xfrd_time() + (time_t)ntohl(zone->soa_disk.retry)*mult <
 		zone->soa_disk_acquired + (time_t)ntohl(zone->soa_disk.expire))
 	{
 		set_retry = ntohl(zone->soa_disk.retry);
+		set_retry *= mult;
 		if(set_retry > (time_t)zone->zone_options->pattern->max_retry_time)
 			set_retry = zone->zone_options->pattern->max_retry_time;
 		else if(set_retry < (time_t)zone->zone_options->pattern->min_retry_time)
@@ -760,13 +767,14 @@ xfrd_set_timer_retry(xfrd_zone_t* zone)
 			set_retry =  XFRD_LOWERBOUND_RETRY;
 		xfrd_set_timer(zone, set_retry);
 	} else {
-		if(ntohl(zone->soa_disk.expire) < XFRD_LOWERBOUND_RETRY)
+		set_retry = ntohl(zone->soa_disk.expire);
+		if(set_retry < XFRD_LOWERBOUND_RETRY)
 			xfrd_set_timer(zone, XFRD_LOWERBOUND_RETRY);
 		else {
-			if(zone->soa_disk_acquired + (time_t)ntohl(zone->soa_disk.expire) < xfrd_time())
+			if(zone->soa_disk_acquired + set_retry < xfrd_time())
 				xfrd_set_timer(zone, XFRD_LOWERBOUND_RETRY);
 			else xfrd_set_timer(zone, zone->soa_disk_acquired +
-				ntohl(zone->soa_disk.expire) - xfrd_time());
+				set_retry - xfrd_time());
 		}
 	}
 }
@@ -1130,6 +1138,8 @@ xfrd_handle_incoming_soa(xfrd_zone_t* zone,
 		zone->soa_nsd = zone->soa_disk;
 		zone->soa_nsd_acquired = zone->soa_disk_acquired;
 		xfrd->write_zonefile_needed = 1;
+		/* reset exponential backoff, we got a normal timer now */
+		zone->fresh_xfr_timeout = 0;
 		if(xfrd_time() - zone->soa_disk_acquired
 			< (time_t)ntohl(zone->soa_disk.refresh))
 		{
@@ -1269,6 +1279,19 @@ xfrd_udp_release(xfrd_zone_t* zone)
 		xfrd->udp_use_num--;
 }
 
+/** disable ixfr for master */
+void
+xfrd_disable_ixfr(xfrd_zone_t* zone)
+{
+	if(!(zone->master->ixfr_disabled &&
+		(zone->master->ixfr_disabled + XFRD_NO_IXFR_CACHE) <= time(NULL))) {
+		/* start new round, with IXFR disabled */
+		zone->round_num = 0;
+		zone->next_master = zone->master_num;
+	}
+	zone->master->ixfr_disabled = time(NULL);
+}
+
 static void
 xfrd_udp_read(xfrd_zone_t* zone)
 {
@@ -1276,7 +1299,7 @@ xfrd_udp_read(xfrd_zone_t* zone)
 	if(!xfrd_udp_read_packet(xfrd->packet, zone->zone_handler.ev_fd)) {
 		zone->master->bad_xfr_count++;
 		if (zone->master->bad_xfr_count > 2) {
-			zone->master->ixfr_disabled = time(NULL);
+			xfrd_disable_ixfr(zone);
 			zone->master->bad_xfr_count = 0;
 		}
 		/* drop packet */
@@ -1303,7 +1326,7 @@ xfrd_udp_read(xfrd_zone_t* zone)
 			xfrd_udp_release(zone);
 			break;
 		case xfrd_packet_notimpl:
-			zone->master->ixfr_disabled = time(NULL);
+			xfrd_disable_ixfr(zone);
 			/* drop packet */
 			xfrd_udp_release(zone);
 			/* query next server */
@@ -1320,7 +1343,7 @@ xfrd_udp_read(xfrd_zone_t* zone)
 		default:
 			zone->master->bad_xfr_count++;
 			if (zone->master->bad_xfr_count > 2) {
-				zone->master->ixfr_disabled = time(NULL);
+				xfrd_disable_ixfr(zone);
 				zone->master->bad_xfr_count = 0;
 			}
 			/* drop packet */
diff --git a/usr.sbin/nsd/xfrd.h b/usr.sbin/nsd/xfrd.h
index 24541dbfc44..b601c2aae3e 100644
--- a/usr.sbin/nsd/xfrd.h
+++ b/usr.sbin/nsd/xfrd.h
@@ -244,6 +244,9 @@ enum xfrd_packet_result {
 #define XFRD_MAX_UDP 64 /* max number of UDP sockets at a time for IXFR */
 #define XFRD_MAX_UDP_NOTIFY 64 /* max concurrent UDP sockets for NOTIFY */
 
+#define XFRD_TRANSFER_TIMEOUT_START 10 /* empty zone timeout is between x and 2*x seconds */
+#define XFRD_TRANSFER_TIMEOUT_MAX 86400 /* empty zone timeout max expbackoff */
+
 extern xfrd_state_t* xfrd;
 
 /* start xfrd, new start. Pass socket to server_main. */
@@ -256,6 +259,9 @@ void xfrd_init_slave_zone(xfrd_state_t* xfrd, zone_options_t* zone_opt);
 /* delete slave zone */
 void xfrd_del_slave_zone(xfrd_state_t* xfrd, const dname_type* dname);
 
+/* disable ixfr for a while for zone->master */
+void xfrd_disable_ixfr(xfrd_zone_t* zone);
+
 /* get the current time epoch. Cached for speed. */
 time_t xfrd_time(void);