MEDIUM: config: don't enforce a low frontend maxconn value anymore

Historically the default frontend's maxconn used to be quite low (2000),
which was sufficient two decades ago but often proved to be a problem
when users had purposely set the global maxconn value but forgot to set
the frontend's.

There is no point in keeping this arbitrary limit for frontends : when
the global maxconn is lower, it's already too high and when the global
maxconn is much higher, it becomes a limiting factor which causes trouble
in production.

This commit allows the value to be set to zero, which becomes the new
default value, to mean it's not directly limited, or in fact it's set
to the global maxconn. Since this operation used to be performed before
computing a possibly automatic global maxconn based on memory limits,
the calculation of the maxconn value and its propagation to the backends'
fullconn has now moved to a dedicated function, proxy_adjust_all_maxconn(),
which is called once the global maxconn is stabilized.

This comes with two benefits :
  1) a configuration missing "maxconn" in the defaults section will not
     limit itself to a magically hardcoded value but will scale up to the
     global maxconn ;

  2) when the global maxconn is not set and memory limits are used instead,
     the frontends' maxconn automatically adapts, and the backends' fullconn
     as well.
diff --git a/src/cfgparse.c b/src/cfgparse.c
index e29ab97..1a03d97 100644
--- a/src/cfgparse.c
+++ b/src/cfgparse.c
@@ -125,7 +125,7 @@
 
 char *cursection = NULL;
 struct proxy defproxy = { };		/* fake proxy used to assign default values on all instances */
-int cfg_maxpconn = DEFAULT_MAXCONN;	/* # of simultaneous connections per proxy (-N) */
+int cfg_maxpconn = 0;                   /* # of simultaneous connections per proxy (-N) */
 int cfg_maxconn = 0;			/* # of simultaneous connections, (-n) */
 char *cfg_scope = NULL;                 /* the current scope during the configuration parsing */
 
@@ -2524,8 +2524,6 @@
 			} else {
 				free(curproxy->defbe.name);
 				curproxy->defbe.be = target;
-				/* Update tot_fe_maxconn for a further fullconn's computation */
-				target->tot_fe_maxconn += curproxy->maxconn;
 				/* Emit a warning if this proxy also has some servers */
 				if (curproxy->srv) {
 					ha_warning("In proxy '%s', the 'default_backend' rule always has precedence over the servers, which will never be used.\n",
@@ -2535,14 +2533,6 @@
 			}
 		}
 
-		if (!curproxy->defbe.be && (curproxy->cap & PR_CAP_LISTEN) == PR_CAP_LISTEN) {
-			/* Case of listen without default backend
-			 * The curproxy will be its own default backend
-			 * so we update tot_fe_maxconn for a further
-			 * fullconn's computation */
-			curproxy->tot_fe_maxconn += curproxy->maxconn;
-		}
-
 		/* find the target proxy for 'use_backend' rules */
 		list_for_each_entry(rule, &curproxy->switching_rules, list) {
 			struct proxy *target;
@@ -2611,23 +2601,6 @@
 			} else {
 				free((void *)rule->be.name);
 				rule->be.backend = target;
-				/* For each target of switching rules, we update
-				 * their tot_fe_maxconn, except if a previous rule point
-				 * on the same backend or on the default backend */
-				if (rule->be.backend != curproxy->defbe.be) {
-					struct switching_rule *swrule;
-
-					list_for_each_entry(swrule, &curproxy->switching_rules, list) {
-						if (rule == swrule) {
-							target->tot_fe_maxconn += curproxy->maxconn;
-							break;
-						}
-						else if (!swrule->dynamic && swrule->be.backend == rule->be.backend) {
-							/* there is multiple ref of this backend */
-							break;
-						}
-					}
-				}
 			}
 		}
 
@@ -3817,24 +3790,6 @@
 		}
 	}
 
-	/* automatically compute fullconn if not set. We must not do it in the
-	 * loop above because cross-references are not yet fully resolved.
-	 */
-	for (curproxy = proxies_list; curproxy; curproxy = curproxy->next) {
-		/* If <fullconn> is not set, let's set it to 10% of the sum of
-		 * the possible incoming frontend's maxconns.
-		 */
-		if (!curproxy->fullconn && (curproxy->cap & PR_CAP_BE)) {
-			/* we have the sum of the maxconns in <total>. We only
-			 * keep 10% of that sum to set the default fullconn, with
-			 * a hard minimum of 1 (to avoid a divide by zero).
-			 */
-			curproxy->fullconn = (curproxy->tot_fe_maxconn + 9) / 10;
-			if (!curproxy->fullconn)
-				curproxy->fullconn = 1;
-		}
-	}
-
 	/*
 	 * Recount currently required checks.
 	 */
diff --git a/src/haproxy.c b/src/haproxy.c
index 942c075..59da1fd 100644
--- a/src/haproxy.c
+++ b/src/haproxy.c
@@ -2069,6 +2069,8 @@
 				global.maxsock += p->peers_fe->maxconn;
 	}
 
+	proxy_adjust_all_maxconn();
+
 	if (global.tune.maxpollevents <= 0)
 		global.tune.maxpollevents = MAX_POLL_EVENTS;
 
diff --git a/src/proxy.c b/src/proxy.c
index 7f442b9..e60b34a 100644
--- a/src/proxy.c
+++ b/src/proxy.c
@@ -1488,6 +1488,75 @@
 	HA_SPIN_UNLOCK(PROXY_LOCK, &proxy->lock);
 }
 
+/* Configure all proxies which lack a maxconn setting to use the global one by
+ * default. This avoids the common mistake consisting in setting maxconn only
+ * in the global section and discovering the hard way that it doesn't propagate
+ * through the frontends. These values are also propagated through the various
+ * targetted backends, whose fullconn is finally calculated if not yet set.
+ */
+void proxy_adjust_all_maxconn()
+{
+	struct proxy *curproxy;
+	struct switching_rule *swrule1, *swrule2;
+
+	for (curproxy = proxies_list; curproxy; curproxy = curproxy->next) {
+		if (curproxy->state == PR_STSTOPPED)
+			continue;
+
+		if (!(curproxy->cap & PR_CAP_FE))
+			continue;
+
+		if (!curproxy->maxconn)
+			curproxy->maxconn = global.maxconn;
+
+		/* update the target backend's fullconn count : default_backend */
+		if (curproxy->defbe.be)
+			curproxy->defbe.be->tot_fe_maxconn += curproxy->maxconn;
+		else if ((curproxy->cap & PR_CAP_LISTEN) == PR_CAP_LISTEN)
+			curproxy->tot_fe_maxconn += curproxy->maxconn;
+
+		list_for_each_entry(swrule1, &curproxy->switching_rules, list) {
+			/* For each target of switching rules, we update their
+			 * tot_fe_maxconn, except if a previous rule points to
+			 * the same backend or to the default backend.
+			 */
+			if (swrule1->be.backend != curproxy->defbe.be) {
+				list_for_each_entry(swrule2, &curproxy->switching_rules, list) {
+					if (swrule2 == swrule1) {
+						swrule1->be.backend->tot_fe_maxconn += curproxy->maxconn;
+						break;
+					}
+					else if (!swrule2->dynamic && swrule2->be.backend == swrule1->be.backend) {
+						/* there are multiple refs of this backend */
+						break;
+					}
+				}
+			}
+		}
+	}
+
+	/* automatically compute fullconn if not set. We must not do it in the
+	 * loop above because cross-references are not yet fully resolved.
+	 */
+	for (curproxy = proxies_list; curproxy; curproxy = curproxy->next) {
+		if (curproxy->state == PR_STSTOPPED)
+			continue;
+
+		/* If <fullconn> is not set, let's set it to 10% of the sum of
+		 * the possible incoming frontend's maxconns.
+		 */
+		if (!curproxy->fullconn && (curproxy->cap & PR_CAP_BE)) {
+			/* we have the sum of the maxconns in <total>. We only
+			 * keep 10% of that sum to set the default fullconn, with
+			 * a hard minimum of 1 (to avoid a divide by zero).
+			 */
+			curproxy->fullconn = (curproxy->tot_fe_maxconn + 9) / 10;
+			if (!curproxy->fullconn)
+				curproxy->fullconn = 1;
+		}
+	}
+}
+
 /* Config keywords below */
 
 static struct cfg_kw_list cfg_kws = {ILH, {