[MEDIUM] new option "independant-streams" to stop updating read timeout on writes

By default, when data is sent over a socket, both the write timeout and the
read timeout for that socket are refreshed, because we consider that there is
activity on that socket, and we have no other means of guessing if we should
receive data or not.

While this default behaviour is desirable for almost all applications, there
exists a situation where it is desirable to disable it, and only refresh the
read timeout if there are incoming data. This happens on sessions with large
timeouts and low amounts of exchanged data such as telnet session. If the
server suddenly disappears, the output data accumulates in the system's
socket buffers, both timeouts are correctly refreshed, and there is no way
to know the server does not receive them, so we don't timeout. However, when
the underlying protocol always echoes sent data, it would be enough by itself
to detect the issue using the read timeout. Note that this problem does not
happen with more verbose protocols because data won't accumulate long in the
socket buffers.

When this option is set on the frontend, it will disable read timeout updates
on data sent to the client. There probably is little use of this case. When
the option is set on the backend, it will disable read timeout updates on
data sent to the server. Doing so will typically break large HTTP posts from
slow lines, so use it with caution.

doc/configuration.txt

@@ -747,6 +747,8 @@ option httpchk              X          -         X         X
 [no] option httpclose       X          X         X         X
 option httplog              X          X         X         X
 [no] option http_proxy      X          X         X         X
+[no] option independant-
+            streams         X          X         X         X
 [no] option log-health-     X          -         X         X
             checks
 [no] option log-separate-
@@ -2477,6 +2479,39 @@ no option http_proxy
   See also : "option httpclose"
 
 
+option independant-streams
+no option independant-streams
+  Enable or disable independant timeout processing for both directions
+  May be used in sections :   defaults | frontend | listen | backend
+                                 yes   |    yes   |   yes  |  yes
+  Arguments : none
+
+  By default, when data is sent over a socket, both the write timeout and the
+  read timeout for that socket are refreshed, because we consider that there is
+  activity on that socket, and we have no other means of guessing if we should
+  receive data or not.
+
+  While this default behaviour is desirable for almost all applications, there
+  exists a situation where it is desirable to disable it, and only refresh the
+  read timeout if there are incoming data. This happens on sessions with large
+  timeouts and low amounts of exchanged data such as telnet session. If the
+  server suddenly disappears, the output data accumulates in the system's
+  socket buffers, both timeouts are correctly refreshed, and there is no way
+  to know the server does not receive them, so we don't timeout. However, when
+  the underlying protocol always echoes sent data, it would be enough by itself
+  to detect the issue using the read timeout. Note that this problem does not
+  happen with more verbose protocols because data won't accumulate long in the
+  socket buffers.
+
+  When this option is set on the frontend, it will disable read timeout updates
+  on data sent to the client. There probably is little use of this case. When
+  the option is set on the backend, it will disable read timeout updates on
+  data sent to the server. Doing so will typically break large HTTP posts from
+  slow lines, so use it with caution.
+
+  See also : "timeout client" and "timeout server"
+
+
 option log-health-checks
 no option log-health-checks
   Enable or disable logging of health checks

include/types/proxy.h

@@ -119,6 +119,7 @@
 #define PR_O2_RDPC_PRST	0x00000200      /* Actvate rdp cookie analyser */
 #define PR_O2_CLFLOG	0x00000400      /* log into clf format */
 #define PR_O2_LOGHCHKS	0x00000800	/* log health checks */
+#define PR_O2_INDEPSTR	0x00001000	/* independant streams, don't update rex on write */
 
 struct error_snapshot {
 	struct timeval when;		/* date of this event, (tv_sec == 0) means "never" */

include/types/stream_interface.h

@@ -69,6 +69,7 @@ enum {
 	SI_FL_WAIT_DATA  = 0x0008,  /* waiting for more data to send */
 	SI_FL_CAP_SPLTCP = 0x0010,  /* splicing possible from/to TCP */
 	SI_FL_DONT_WAKE  = 0x0020,  /* resync in progress, don't wake up */
+	SI_FL_INDEP_STR  = 0x0040,  /* independant streams = don't update rex on write */
 };
 
 #define SI_FL_CAP_SPLICE (SI_FL_CAP_SPLTCP)

src/cfgparse.c

@@ -141,6 +141,7 @@ static const struct cfg_opt cfg_opts2[] =
 	{ "log-health-checks",            PR_O2_LOGHCHKS,  PR_CAP_BE, 0 },
 	{ "tcp-smart-accept",             PR_O2_SMARTACC,  PR_CAP_FE, 0 },
 	{ "tcp-smart-connect",            PR_O2_SMARTCON,  PR_CAP_BE, 0 },
+	{ "independant-streams",          PR_O2_INDEPSTR,  PR_CAP_FE|PR_CAP_BE, 0 },
 	{ NULL, 0, 0, 0 }
 };
 

src/client.c

@@ -200,6 +200,8 @@ int event_accept(int fd) {
 		s->si[0].iohandler = NULL;
 		s->si[0].fd = cfd;
 		s->si[0].flags = SI_FL_NONE | SI_FL_CAP_SPLTCP; /* TCP splicing capable */
+		if (s->fe->options2 & PR_O2_INDEPSTR)
+			s->si[0].flags |= SI_FL_INDEP_STR;
 		s->si[0].exp = TICK_ETERNITY;
 
 		s->si[1].state = s->si[1].prev_state = SI_ST_INI;
@@ -216,6 +218,8 @@ int event_accept(int fd) {
 		s->si[1].exp = TICK_ETERNITY;
 		s->si[1].fd = -1; /* just to help with debugging */
 		s->si[1].flags = SI_FL_NONE;
+		if (s->be->options2 & PR_O2_INDEPSTR)
+			s->si[1].flags |= SI_FL_INDEP_STR;
 
 		s->srv = s->prev_srv = s->srv_conn = NULL;
 		s->pend_pos = NULL;

src/proto_uxst.c

@@ -455,6 +455,8 @@ int uxst_event_accept(int fd) {
 		s->si[0].iohandler = NULL;
 		s->si[0].fd = cfd;
 		s->si[0].flags = SI_FL_NONE;
+		if (s->fe->options2 & PR_O2_INDEPSTR)
+			s->si[0].flags |= SI_FL_INDEP_STR;
 		s->si[0].exp = TICK_ETERNITY;
 
 		s->si[1].state = s->si[1].prev_state = SI_ST_INI;
@@ -464,6 +466,9 @@ int uxst_event_accept(int fd) {
 		s->si[1].exp = TICK_ETERNITY;
 		s->si[1].fd = -1; /* just to help with debugging */
 		s->si[1].flags = SI_FL_NONE;
+		if (s->be->options2 & PR_O2_INDEPSTR)
+			s->si[1].flags |= SI_FL_INDEP_STR;
+
 		stream_int_register_handler(&s->si[1], stats_io_handler);
 		s->si[1].private = s;
 		s->si[1].st0 = s->si[1].st1 = 0;

src/proxy.c

@@ -653,6 +653,10 @@ int session_set_backend(struct session *s, struct proxy *be)
 	s->rep->rto = s->req->wto = be->timeout.server;
 	s->req->cto = be->timeout.connect;
 	s->conn_retries = be->conn_retries;
+	s->si[1].flags &= ~SI_FL_INDEP_STR;
+	if (be->options2 & PR_O2_INDEPSTR)
+		s->si[1].flags |= SI_FL_INDEP_STR;
+
 	if (be->options2 & PR_O2_RSPBUG_OK)
 		s->txn.rsp.err_pos = -1; /* let buggy responses pass */
 	s->flags |= SN_BE_ASSIGNED;

src/stream_interface.c

@@ -123,13 +123,17 @@ void stream_int_update_embedded(struct stream_interface *si)
 	if ((si->ib->flags & (BF_FULL|BF_SHUTR)) == BF_FULL)
 		si->flags |= SI_FL_WAIT_ROOM;
 
-	if (si->ob->flags & BF_WRITE_ACTIVITY || si->ib->flags & BF_READ_ACTIVITY) {
-		if (tick_isset(si->ib->rex))
-			si->ib->rex = tick_add_ifset(now_ms, si->ib->rto);
+	if (si->ob->flags & BF_WRITE_ACTIVITY) {
 		if (tick_isset(si->ob->wex))
 			si->ob->wex = tick_add_ifset(now_ms, si->ob->wto);
 	}
 
+	if (si->ib->flags & BF_READ_ACTIVITY ||
+	    (si->ob->flags & BF_WRITE_ACTIVITY && !(si->flags & SI_FL_INDEP_STR))) {
+		if (tick_isset(si->ib->rex))
+			si->ib->rex = tick_add_ifset(now_ms, si->ib->rto);
+	}
+
 	if (si->ob->flags & BF_WRITE_PARTIAL)
 		si->ob->prod->chk_rcv(si->ob->prod);
 

src/stream_sock.c

@@ -764,12 +764,14 @@ int stream_sock_write(int fd)
 			b->wex = tick_add_ifset(now_ms, b->wto);
 
 	out_wakeup:
-		if (tick_isset(si->ib->rex)) {
+		if (tick_isset(si->ib->rex) && !(si->flags & SI_FL_INDEP_STR)) {
 			/* Note: to prevent the client from expiring read timeouts
-			 * during writes, we refresh it. A better solution would be
-			 * to merge read+write timeouts into a unique one, although
-			 * that needs some study particularly on full-duplex TCP
-			 * connections.
+			 * during writes, we refresh it. We only do this if the
+			 * interface is not configured for "independant streams",
+			 * because for some applications it's better not to do this,
+			 * for instance when continuously exchanging small amounts
+			 * of data which can full the socket buffers long before a
+			 * write timeout is detected.
 			 */
 			si->ib->rex = tick_add_ifset(now_ms, si->ib->rto);
 		}
@@ -943,11 +945,12 @@ void stream_sock_data_finish(struct stream_interface *si)
 			EV_FD_COND_S(fd, DIR_WR);
 			if (!tick_isset(ob->wex) || ob->flags & BF_WRITE_ACTIVITY) {
 				ob->wex = tick_add_ifset(now_ms, ob->wto);
-				if (tick_isset(ib->rex)) {
+				if (tick_isset(ib->rex) && !(si->flags & SI_FL_INDEP_STR)) {
 					/* Note: depending on the protocol, we don't know if we're waiting
 					 * for incoming data or not. So in order to prevent the socket from
 					 * expiring read timeouts during writes, we refresh the read timeout,
-					 * except if it was already infinite.
+					 * except if it was already infinite or if we have explicitly setup
+					 * independant streams.
 					 */
 					ib->rex = tick_add_ifset(now_ms, ib->rto);
 				}
@@ -1069,12 +1072,14 @@ void stream_sock_chk_snd(struct stream_interface *si)
 		if ((ob->flags & (BF_OUT_EMPTY|BF_SHUTW|BF_WRITE_PARTIAL)) == BF_WRITE_PARTIAL)
 			ob->wex = tick_add_ifset(now_ms, ob->wto);
 
-		if (tick_isset(si->ib->rex)) {
+		if (tick_isset(si->ib->rex) && !(si->flags & SI_FL_INDEP_STR)) {
 			/* Note: to prevent the client from expiring read timeouts
-			 * during writes, we refresh it. A better solution would be
-			 * to merge read+write timeouts into a unique one, although
-			 * that needs some study particularly on full-duplex TCP
-			 * connections.
+			 * during writes, we refresh it. We only do this if the
+			 * interface is not configured for "independant streams",
+			 * because for some applications it's better not to do this,
+			 * for instance when continuously exchanging small amounts
+			 * of data which can full the socket buffers long before a
+			 * write timeout is detected.
 			 */
 			si->ib->rex = tick_add_ifset(now_ms, si->ib->rto);
 		}