Send notification and wedge the connection on too many timeouts.

svn path=/trunk/; revision=2190

5 files changed, 50 insertions, 15 deletions

diff --git a/NEWS b/NEWS
index a58d8988..e1b1d79a 100644
--- a/NEWS
+++ b/NEWS
@@ -4,6 +4,7 @@ fetchmail-4.6.7 ():
 * Gerald Britton's patches to enable Hesiod support.
 * Postmaster option fix from Bill Metzenthen <billm@melbpc.org.au>.
 * Socks library support forom Guiseppe Guerini.
+* Notification mail is now sent to the user on 20 consecutive timeouts.
 
 There are 248 people on fetchmail-friends and 306 on fetchmail-announce.
 
diff --git a/driver.c b/driver.c
index f13f7b01..11ea4d49 100644
--- a/driver.c
+++ b/driver.c
@@ -65,7 +65,8 @@
 #endif /* KERBEROS_V5 */
 
 #include "socket.h"
-#include  "fetchmail.h"
+#include "fetchmail.h"
+#include "tunable.h"
 
 #if INET6
 #define	SMTP_PORT	"smtp"	/* standard SMTP service port */
@@ -92,6 +93,7 @@ static int tagnum;
 
 static char shroud[PASSWORDLEN];	/* string to shroud in debug output */
 static int mytimeout;			/* value of nonreponse timeout */
+static int timeoutcount;		/* count consecutive timeouts */
 static int msglen;			/* actual message length */
 
 void set_timeout(int timeleft)
@@ -100,6 +102,9 @@ void set_timeout(int timeleft)
 #ifndef __EMX__
     struct itimerval ntimeout;
 
+    if (timeleft == 0)
+	timeoutcount = 0;
+
     ntimeout.it_interval.tv_sec = ntimeout.it_interval.tv_usec = 0;
     ntimeout.it_value.tv_sec  = timeleft;
     ntimeout.it_value.tv_usec = 0;
@@ -110,6 +115,7 @@ void set_timeout(int timeleft)
 static void timeout_handler (int signal)
 /* handle server-timeout SIGALRM signal */
 {
+    timeoutcount++;
     longjmp(restart, 1);
 }
 
@@ -1364,11 +1370,37 @@ const struct method *proto;	/* protocol method table */
 	    close(ctl->smtp_socket);
 	if (sock != -1)
 	    SockClose(sock);
+
+	/*
+	 * If we've exceeded our threshold for consecutive timeouts, 
+	 * try to notify the user, then mark the connection wedged.
+	 */
+	if (timeoutcount > MAX_TIMEOUTS && !open_warning_by_mail(ctl))
+	{
+	    stuff_warning(ctl,
+			  "Subject: fetchmail sees repeated timeouts\r\n");
+	    stuff_warning(ctl,
+			  "Fetchmail saw more than %d timouts while attempting to get mail from %s@%s.", 
+			  MAX_TIMEOUTS,
+			  ctl->remotename,
+			  ctl->server.truename);
+	    stuff_warning(ctl, 
+			  "This could mean that your mailserver is stuck, or that your SMTP listener");
+	    stuff_warning(ctl, 
+			  "is wedged, or that your mailbox file on the server has been corrupted by");
+	    stuff_warning(ctl, 
+			  "a server error.  You can run `fetchmail -v -v' to diagnose the problem.");
+	    stuff_warning(ctl,
+			  "Fetchmail won't poll this mailbox again until you restart it.");
+	    close_warning_by_mail(ctl);
+	    ctl->wedged = TRUE;
+	}
+
 	ok = PS_ERROR;
     }
     else
     {
-	char buf [POPBUFSIZE+1], *realhost;
+	char buf[POPBUFSIZE+1], *realhost;
 	int len, num, count, new, bytes, deletions = 0, *msgsizes = NULL;
 #if INET6
 	int fetches, dispatches;
@@ -1504,7 +1536,7 @@ const struct method *proto;	/* protocol method table */
 		     * failure the first time it happens.
 		     */
 		    if (run.poll_interval
-			&& !ctl->authfailcount && !open_warning_by_mail(ctl))
+			&& !ctl->wedged && !open_warning_by_mail(ctl))
 		    {
 			stuff_warning(ctl,
 			       "Subject: fetchmail authentication failed\r\n");
@@ -1517,7 +1549,7 @@ const struct method *proto;	/* protocol method table */
 			stuff_warning(ctl, 
 			       "This probably means your password is invalid.");
 			close_warning_by_mail(ctl);
-			ctl->authfailcount++;
+			ctl->wedged = TRUE;
 		    }
 		}
 		goto cleanUp;
diff --git a/fetchmail.c b/fetchmail.c
index 95848491..b36c2d83 100644
--- a/fetchmail.c
+++ b/fetchmail.c
@@ -377,8 +377,6 @@ int main (int argc, char **argv)
     /* pick up interactively any passwords we need but don't have */ 
     for (ctl = querylist; ctl; ctl = ctl->next)
     {
-	ctl->authfailcount = 0;
-
 	if (ctl->active && !(implicitmode && ctl->server.skip)&&!ctl->password)
 	{
 	    if (ctl->server.preauthenticate == A_KERBEROS_V4 ||
@@ -505,10 +503,10 @@ int main (int argc, char **argv)
 	batchcount = 0;
 	for (ctl = querylist; ctl; ctl = ctl->next)
 	{
-	    if (ctl->authfailcount)
+	    if (ctl->wedged)
 	    {
 		error(0, -1, 
-		      "poll of %s skipped until authentication is unwedged",
+		      "poll of %s skipped (failed authentication or too many timeouts)",
 		      ctl->server.pollname);
 		continue;
 	    }
@@ -603,14 +601,14 @@ int main (int argc, char **argv)
 	     * should softly and silently vanish away, rather than
 	     * spinning uselessly.
 	     */
-	    int auth_ok = 0;
+	    int unwedged = 0;
 
 	    for (ctl = querylist; ctl; ctl = ctl->next)
-		if (!ctl->authfailcount)
-		    auth_ok++;
-	    if (!auth_ok)
+		if (!ctl->wedged)
+		    unwedged++;
+	    if (!unwedged)
 	    {
-		error(0, -1, "All authentications have failed.  Exiting.");
+		error(0, -1, "All connections are wedged.  Exiting.");
 		exit(PS_AUTHFAIL);
 	    }
 
@@ -896,6 +894,8 @@ static int load_params(int argc, char **argv, int optind)
     /* merge in wired defaults, do sanity checks and prepare internal fields */
     for (ctl = querylist; ctl; ctl = ctl->next)
     {
+	ctl->wedged = FALSE;
+
 	if (configdump || (ctl->active && !(implicitmode && ctl->server.skip)))
 	{
 	    /* merge in defaults */
diff --git a/fetchmail.h b/fetchmail.h
index 9b4f03b5..d40f25da 100644
--- a/fetchmail.h
+++ b/fetchmail.h
@@ -255,7 +255,7 @@ struct query
     flag active;		/* should we actually poll this server? */
     const char *destaddr;	/* destination host for this query */
     int errcount;		/* count transient errors in last pass */
-    int authfailcount;		/* count authentication failures this run */
+    int wedged;			/* wedged by auth failures or timeouts? */
     char *smtphost;		/* actual SMTP host we connected to */
     int smtp_socket;		/* socket descriptor for SMTP connection */
     unsigned int uid;		/* UID of user to deliver to */
diff --git a/tunable.h b/tunable.h
index 9ee1e6c9..8046afc1 100644
--- a/tunable.h
+++ b/tunable.h
@@ -10,6 +10,8 @@
 /* default timeout period in seconds if server connection dies */
 #define	CLIENT_TIMEOUT		300
 
+/* maximum consecutive timeouts to accept */
+#define MAX_TIMEOUTS		20
+
 /* default skipped message warning interval in seconds */
 #define WARNING_INTERVAL	3600
-