From a56de89a4cd467c5cca7a11032cbf6bd5288b95b Mon Sep 17 00:00:00 2001
From: Samuel Thibault <samuel.thibault@ens-lyon.org>
Date: Sun, 30 Mar 2025 12:01:16 +0200
Subject: [PATCH] kernel accel: if switching takes a long time, back off

That can happen even with not many sessions, e.g. on a loaded machine with a
lot of routes
---
 l2tpns.c | 26 +++++++++++++++++++-------
 1 file changed, 19 insertions(+), 7 deletions(-)

diff --git a/l2tpns.c b/l2tpns.c
index 2e0d89b..febcab0 100644
--- a/l2tpns.c
+++ b/l2tpns.c
@@ -140,6 +140,7 @@ char main_quit = 0;			// True if we're in the process of exiting.
 static char main_reload = 0;		// Re-load pending
 #define MAX_KERNEL_SWITCHES 20		// Maximum number of kernel switches per 1/10th second
 static int kernel_switches = 0;		// How many kernel switches we performed since last cleanup
+static double last_poll_time = 0;	// Last time we polled fds
 linked_list *loaded_plugins;
 linked_list *plugins[MAX_PLUGIN_TYPES];
 
@@ -1387,7 +1388,7 @@ static int delete_kernel_accel(sessionidt s)
 // Enable (set=1) or disable (set=0) kernel PPP acceleration
 // This basically calls create/delete_kernel_accel, but also updates routes
 // If now is 0, we may delay this if we have already made a lot of switches since last cleanup
-static void set_kernel_accel(sessionidt s, int set, int now)
+static void set_kernel_accel(sessionidt s, int set, int nodelay)
 {
 	if (set && !can_kernel_accel(s))
 		/* Still cannot enable it */
@@ -1418,10 +1419,14 @@ static void set_kernel_accel(sessionidt s, int set, int now)
 		}
 	}
 
-	if (set && !now && kernel_switches >= MAX_KERNEL_SWITCHES)
+	double cur_switch;
+	now(&cur_switch);
+	if (set && !nodelay &&
+	    (kernel_switches >= MAX_KERNEL_SWITCHES
+	     || cur_switch - last_poll_time > 0.02))
 	{
-		// We already performed many switches, throttle a bit by just
-		// marking as pending
+		// We have already been switching a lot or working at for 20ms since last poll,
+		// throttle a bit by just marking as pending
 		sess_local[s].needs_switch = 1;
 		return;
 	}
@@ -5999,7 +6004,7 @@ static void mainloop(void)
 		n = epoll_wait(epollfd, events, maxevent, 100); // timeout 100ms (1/10th sec)
 		STAT(select_called);
 
-		TIME = now(NULL);
+		TIME = now(&last_poll_time);
 		if (n < 0)
 		{
 			if (errno == EINTR ||
@@ -6387,7 +6392,7 @@ static void mainloop(void)
 		// Handle trying to enable kernel accel
 		{
 			static double last_switch = 0;
-			double this_switch;
+			double this_switch, cur_switch;
 			double diff;
 
 			TIME = now(&this_switch);
@@ -6399,14 +6404,21 @@ static void mainloop(void)
 			{
 				kernel_switches = 0;
 
+				cur_switch = this_switch;
 				for (i = 1; i <= config->cluster_highest_sessionid; i++)
 				{
 					// Delayed kernel switch
 					if (session[i].ppp.lcp == Opened && sess_local[i].needs_switch)
+					{
 						set_kernel_accel(i, can_kernel_accel(i), 0);
+						TIME = now(&cur_switch);
+						if (cur_switch - this_switch > 0.02)
+							// We have already taken 20ms sec doing switches, backoff
+							break;
+					}
 				}
 
-				last_switch = this_switch;
+				last_switch = cur_switch;
 			}
 		}