From: Ondřej Surý <ondrej@isc.org>
Date: Wed, 13 Jul 2022 07:34:47 +0000 (+0200)
Subject: Handle the transient TCP connect() failures on FreeBSD
X-Git-Tag: v9.19.4~26^2~2
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=a280855f7ba743027440c274efdff4bb4c1e5668;p=thirdparty%2Fbind9.git

Handle the transient TCP connect() failures on FreeBSD

On FreeBSD (and perhaps other *BSD) systems, the TCP connect() call (via
uv_tcp_connect()) can fail with transient UV_EADDRINUSE error.  The UDP
code already handles this by trying three times (is a charm) before
giving up.  Add a code for the TCP, TCPDNS and TLSDNS layers to also try
three times before giving up by calling uv_tcp_connect() from the
callback two more time on UV_EADDRINUSE error.

Additionally, stop the timer only if we succeed or on hard error via
isc__nm_failed_connect_cb().
---

diff --git a/lib/isc/netmgr/netmgr-int.h b/lib/isc/netmgr/netmgr-int.h
index ebb57a33906..96f863fa4d0 100644
--- a/lib/isc/netmgr/netmgr-int.h
+++ b/lib/isc/netmgr/netmgr-int.h
@@ -377,6 +377,7 @@ struct isc__nm_uvreq {
 	isc__nm_cb_t cb;       /* callback */
 	void *cbarg;	       /* callback argument */
 	isc_nm_timer_t *timer; /* TCP write timer */
+	int connect_tries;     /* connect retries */
 
 	union {
 		uv_handle_t handle;
diff --git a/lib/isc/netmgr/netmgr.c b/lib/isc/netmgr/netmgr.c
index 4b416fae067..350117d6f7c 100644
--- a/lib/isc/netmgr/netmgr.c
+++ b/lib/isc/netmgr/netmgr.c
@@ -2465,7 +2465,10 @@ isc___nm_uvreq_get(isc_nm_t *mgr, isc_nmsocket_t *sock FLARG) {
 		req = isc_mem_get(mgr->mctx, sizeof(*req));
 	}
 
-	*req = (isc__nm_uvreq_t){ .magic = 0 };
+	*req = (isc__nm_uvreq_t){
+		.magic = 0,
+		.connect_tries = 3,
+	};
 	ISC_LINK_INIT(req, link);
 	req->uv_req.req.data = req;
 	isc___nmsocket_attach(sock, &req->sock FLARG_PASS);
diff --git a/lib/isc/netmgr/tcp.c b/lib/isc/netmgr/tcp.c
index f7af009081b..3d139b0dc19 100644
--- a/lib/isc/netmgr/tcp.c
+++ b/lib/isc/netmgr/tcp.c
@@ -228,9 +228,6 @@ tcp_connect_cb(uv_connect_t *uvreq, int status) {
 	REQUIRE(VALID_NMSOCK(sock));
 	REQUIRE(sock->tid == isc_nm_tid());
 
-	isc__nmsocket_timer_stop(sock);
-	uv_handle_set_data((uv_handle_t *)&sock->read_timer, sock);
-
 	req = uv_handle_get_data((uv_handle_t *)uvreq);
 
 	REQUIRE(VALID_UVREQ(req));
@@ -239,9 +236,7 @@ tcp_connect_cb(uv_connect_t *uvreq, int status) {
 	if (atomic_load(&sock->timedout)) {
 		result = ISC_R_TIMEDOUT;
 		goto error;
-	}
-
-	if (!atomic_load(&sock->connecting)) {
+	} else if (!atomic_load(&sock->connecting)) {
 		/*
 		 * The connect was cancelled from timeout; just clean up
 		 * the req.
@@ -260,11 +255,33 @@ tcp_connect_cb(uv_connect_t *uvreq, int status) {
 		/* Timeout status code here indicates hard error */
 		result = ISC_R_TIMEDOUT;
 		goto error;
+	} else if (status == UV_EADDRINUSE) {
+		/*
+		 * On FreeBSD the TCP connect() call sometimes results in a
+		 * spurious transient EADDRINUSE. Try a few more times before
+		 * giving up.
+		 */
+		if (--req->connect_tries > 0) {
+			r = uv_tcp_connect(&req->uv_req.connect,
+					   &sock->uv_handle.tcp,
+					   &req->peer.type.sa, tcp_connect_cb);
+			if (r != 0) {
+				isc__nm_incstats(sock, STATID_CONNECTFAIL);
+				result = isc_uverr2result(r);
+				goto error;
+			}
+			return;
+		}
+		result = isc_uverr2result(status);
+		goto error;
 	} else if (status != 0) {
 		result = isc_uverr2result(status);
 		goto error;
 	}
 
+	isc__nmsocket_timer_stop(sock);
+	uv_handle_set_data((uv_handle_t *)&sock->read_timer, sock);
+
 	isc__nm_incstats(sock, STATID_CONNECT);
 	r = uv_tcp_getpeername(&sock->uv_handle.tcp, (struct sockaddr *)&ss,
 			       &(int){ sizeof(ss) });
@@ -281,7 +298,6 @@ tcp_connect_cb(uv_connect_t *uvreq, int status) {
 	isc__nm_connectcb(sock, req, ISC_R_SUCCESS, false);
 
 	return;
-
 error:
 	isc__nm_failed_connect_cb(sock, req, result, false);
 }
diff --git a/lib/isc/netmgr/tcpdns.c b/lib/isc/netmgr/tcpdns.c
index e0da11b43f9..685268eed96 100644
--- a/lib/isc/netmgr/tcpdns.c
+++ b/lib/isc/netmgr/tcpdns.c
@@ -198,9 +198,6 @@ tcpdns_connect_cb(uv_connect_t *uvreq, int status) {
 	REQUIRE(VALID_NMSOCK(sock));
 	REQUIRE(sock->tid == isc_nm_tid());
 
-	isc__nmsocket_timer_stop(sock);
-	uv_handle_set_data((uv_handle_t *)&sock->read_timer, sock);
-
 	req = uv_handle_get_data((uv_handle_t *)uvreq);
 
 	REQUIRE(VALID_UVREQ(req));
@@ -209,9 +206,7 @@ tcpdns_connect_cb(uv_connect_t *uvreq, int status) {
 	if (atomic_load(&sock->timedout)) {
 		result = ISC_R_TIMEDOUT;
 		goto error;
-	}
-
-	if (isc__nm_closing(sock)) {
+	} else if (isc__nm_closing(sock)) {
 		/* Network manager shutting down */
 		result = ISC_R_SHUTTINGDOWN;
 		goto error;
@@ -223,11 +218,33 @@ tcpdns_connect_cb(uv_connect_t *uvreq, int status) {
 		/* Timeout status code here indicates hard error */
 		result = ISC_R_TIMEDOUT;
 		goto error;
+	} else if (status == UV_EADDRINUSE) {
+		/*
+		 * On FreeBSD the TCP connect() call sometimes results in a
+		 * spurious transient EADDRINUSE. Try a few more times before
+		 * giving up.
+		 */
+		if (--req->connect_tries > 0) {
+			r = uv_tcp_connect(
+				&req->uv_req.connect, &sock->uv_handle.tcp,
+				&req->peer.type.sa, tcpdns_connect_cb);
+			if (r != 0) {
+				isc__nm_incstats(sock, STATID_CONNECTFAIL);
+				result = isc_uverr2result(r);
+				goto error;
+			}
+			return;
+		}
+		result = isc_uverr2result(status);
+		goto error;
 	} else if (status != 0) {
 		result = isc_uverr2result(status);
 		goto error;
 	}
 
+	isc__nmsocket_timer_stop(sock);
+	uv_handle_set_data((uv_handle_t *)&sock->read_timer, sock);
+
 	isc__nm_incstats(sock, STATID_CONNECT);
 	r = uv_tcp_getpeername(&sock->uv_handle.tcp, (struct sockaddr *)&ss,
 			       &(int){ sizeof(ss) });
@@ -244,7 +261,6 @@ tcpdns_connect_cb(uv_connect_t *uvreq, int status) {
 	isc__nm_connectcb(sock, req, ISC_R_SUCCESS, false);
 
 	return;
-
 error:
 	isc__nm_failed_connect_cb(sock, req, result, false);
 }
diff --git a/lib/isc/netmgr/tlsdns.c b/lib/isc/netmgr/tlsdns.c
index 42f47cb4238..44c546836b5 100644
--- a/lib/isc/netmgr/tlsdns.c
+++ b/lib/isc/netmgr/tlsdns.c
@@ -246,9 +246,7 @@ tlsdns_connect_cb(uv_connect_t *uvreq, int status) {
 	if (atomic_load(&sock->timedout)) {
 		result = ISC_R_TIMEDOUT;
 		goto error;
-	}
-
-	if (isc__nm_closing(sock)) {
+	} else if (isc__nm_closing(sock)) {
 		/* Network manager shutting down */
 		result = ISC_R_SHUTTINGDOWN;
 		goto error;
@@ -260,6 +258,25 @@ tlsdns_connect_cb(uv_connect_t *uvreq, int status) {
 		/* Timeout status code here indicates hard error */
 		result = ISC_R_TIMEDOUT;
 		goto error;
+	} else if (status == UV_EADDRINUSE) {
+		/*
+		 * On FreeBSD the TCP connect() call sometimes results in a
+		 * spurious transient EADDRINUSE. Try a few more times before
+		 * giving up.
+		 */
+		if (--req->connect_tries > 0) {
+			r = uv_tcp_connect(
+				&req->uv_req.connect, &sock->uv_handle.tcp,
+				&req->peer.type.sa, tlsdns_connect_cb);
+			if (r != 0) {
+				isc__nm_incstats(sock, STATID_CONNECTFAIL);
+				result = isc_uverr2result(r);
+				goto error;
+			}
+			return;
+		}
+		result = isc_uverr2result(status);
+		goto error;
 	} else if (status != 0) {
 		result = isc_uverr2result(status);
 		goto error;
diff --git a/lib/isc/netmgr/udp.c b/lib/isc/netmgr/udp.c
index 37a3f0480a9..b4aa2ea7633 100644
--- a/lib/isc/netmgr/udp.c
+++ b/lib/isc/netmgr/udp.c
@@ -846,7 +846,6 @@ udp_connect_direct(isc_nmsocket_t *sock, isc__nm_uvreq_t *req) {
 	isc__networker_t *worker = NULL;
 	int uv_bind_flags = UV_UDP_REUSEADDR;
 	isc_result_t result = ISC_R_UNSET;
-	int tries = 3;
 	int r;
 
 	REQUIRE(isc__nm_in_netthread());
@@ -901,7 +900,7 @@ udp_connect_direct(isc_nmsocket_t *sock, isc__nm_uvreq_t *req) {
 	do {
 		r = isc_uv_udp_connect(&sock->uv_handle.udp,
 				       &req->peer.type.sa);
-	} while (r == UV_EADDRINUSE && --tries > 0);
+	} while (r == UV_EADDRINUSE && --req->connect_tries > 0);
 	if (r != 0) {
 		isc__nm_incstats(sock, STATID_CONNECTFAIL);
 		goto done;