[patch 1/4] openmosix/kcomd_thread-verbose-functionning.patch Verbose messages for kcomd_thread

Subject: [patch @num@/@total@] @name@ Verbose messages for kcomd_thread

This adds verbose debug messages for kcomd_thread (this is quite useful to debug it)

Messages are control using the debugfs system (OMDEBUG_KCOMD) as usual

Index: linux/hpc/kcomd.c
===================================================================
--- linux.orig/hpc/kcomd.c	2006-11-13 16:02:03.000000000 +0100
+++ linux/hpc/kcomd.c	2006-11-13 16:02:53.000000000 +0100
@@ -480,6 +480,7 @@
 	struct kcom_task *task;
 	/* for each nodes (set fds.in && fds.out) and reset the waiting fds */
 
+	OMDEBUG_KCOMD( 3, "Resetting select fdsets (%d fdsets)\n", n);
 	zero_fd_set(n, sockets_fds.in);
 	zero_fd_set(n, sockets_fds.out);
 	zero_fd_set(n, sockets_fds.ex);
@@ -488,11 +489,13 @@
 	list_for_each_entry(node, &kcom_nodes, list) {
 
 		if (node->fd == -1 || node->fd > maxfds) {
-			read_unlock(&kcom_nodes_lock);
+			OMDEBUG_KCOMD( 3, "KCOMD found invalid fd ?\n");
+			schedule_timeout_interruptible(HZ);
 			continue;
 		}
 
 		/* if a file descriptor is open, we want select to pay attention.*/
+		OMDEBUG_KCOMD( 3, "Setting fd %d\n", node->fd );
 		set_bit(node->fd, sockets_fds.in);
 		set_bit(node->fd, sockets_fds.ex);
 
@@ -500,6 +503,7 @@
 		read_lock(&node->tasks_lock);
 		list_for_each_entry(task, &node->tasks, list)
 			if (!list_empty(&task->out_packs)) {
+				OMDEBUG_KCOMD(3, "Setting out flag for fd %d\n", node->fd );
 				set_bit(node->fd, sockets_fds.out);
 			}
 		read_unlock(&node->tasks_lock);
@@ -572,7 +576,6 @@
 	int ret;
 	struct kcom_node *node, *node_next;
 	s64 timeout = -1;
-	int n = -1;
 	siginfo_t info; /* matt*/
 	int sig;
 	int err;
@@ -591,6 +594,8 @@
 	kcomd_task=current;
 
 retry_listen:
+	OMDEBUG_KCOMD( 3, "KCOMD: Creating (retrying) sockets\n" );
+
 	fd4 = socket_listen_ip4(DAEMON_IP4_PORT, &lsock4);
 	fd6 = socket_listen_ip6(DAEMON_IP6_PORT, &lsock6);
 
@@ -600,13 +605,18 @@
 	}
 
 
-	n=alloc_fd_bitmap(max(fd4, fd6));
+	alloc_fd_bitmap(max(fd4, fd6));
 
 	while (kcomd_done==0)
 	{
-		n = maxfds;
+		OMDEBUG_KCOMD( 3, "KCOMD: Entering (restarting) loop\n" );
+
+ 		/* Check if we must leave ... */
+ 		if (kcomd_module_exit) {
+ 			if (0 == kcomd_leave_cluster()) kcomd_done = 1;
+ 		}
 
-		kcomd_thread_set_fds(n);
+		kcomd_thread_set_fds( maxfds );
 
 		/* add listening sockets to the set */
 		if (fd4!=-1)
@@ -619,12 +629,19 @@
 
 		allow_signal (SIGHUP);
 		timeout = -1;
-		ret = do_select(n + 1, &sockets_fds, &timeout);
-		disallow_signal (SIGHUP);
-		/* SIGHUP is sent by kcom_send to wake up kcomd so it can send the packet*/
-		while (signal_pending (current)) {
-			sig = dequeue_signal (current, &current->blocked, &info);
+		OMDEBUG_KCOMD( 3, "KCOMD: Waiting for events ...\n");
+		ret = do_select(maxfds + 1, &sockets_fds, &timeout);
+		if (0 == ret && !signal_pending(current)) {
+		    OMBUG( "KCOMD: do_select returned 0 but no signal pending !\n" );
+		    continue;
+		} else {
+		    /* SIGHUP is sent by kcom_send to wake up kcomd so it can send the packet*/
+		    sig = dequeue_signal (current, &current->blocked, &info);
+		    OMDEBUG_KCOMD( 3, "KCOMD: got signal %d\n", sig);
 		}
+
+		disallow_signal (SIGHUP);
+
 		if (ret == 0) { /* -1=error; 0=signal*/
 			/* New kernel security, prohibits sharing file descriptors between kernel threads.*/
 			/* We have to allocate them here.*/
@@ -633,6 +650,7 @@
 			read_lock(&kcom_nodes_lock);
 			list_for_each_entry(node, &kcom_nodes, list)
 				if (node->fd==0) { /* unmapped*/
+					OMDEBUG_KCOMD( 3, "Found unmapped fd %d\n", node->fd );
 					node->fd = sock_map_fd(node->sock);
 					alloc_fd_bitmap(node->fd);
 				}
@@ -641,7 +659,7 @@
 			continue;
 
 		} else if (ret < 0) {/* -1=error; 0=signal*/
-			OMBUG("do_select returned an error. \n");
+			OMBUG("do_select returned an error (%d). \n", ret);
 			schedule_timeout_interruptible(HZ);
 		}
 
@@ -662,9 +680,10 @@
 		list_for_each_entry_safe(node, node_next, &kcom_nodes, list) {
 
 			if (test_bit(node->fd, sockets_fds.res_in)) {
+				OMDEBUG_KCOMD( 3, "KCOMD: receiving on fd %d\n", node->fd );
 				err=pkt_read(node);
 				if (err!=0) {
-					OMBUG("ERROR receiving data => ignoring packet.\n");
+					printk(KERN_ERR "ERROR receiving data => ignoring packet.\n");
 					continue;
 				}
 			}
@@ -674,23 +693,16 @@
 		}
 		write_unlock(&kcom_nodes_lock);
 
- 		/* Check if we must leave ... */
- 		if (kcomd_module_exit) {
- 			if (0 == kcomd_leave_cluster()) kcomd_done = 1;
- 		}
 	}
 
 /* DONE*/
  	printk(KERN_INFO "kcomd: openMosix kernel communication daemon cleaning up.\n");
-	if (fd4!=-1) {
- 		if (lsock4) sock_release(lsock4);
+
+	if (fd4!=-1)
 		sys_close(fd4);
-	}
 
-	if (fd6!=-1) {
- 		if (lsock6) sock_release(lsock6);
+	if (fd6!=-1)
 		sys_close(fd6);
-	}
 
 
 	kcomd_thread_free_ressources();
Index: linux/hpc/kcom.c
===================================================================
--- linux.orig/hpc/kcom.c	2006-11-13 16:01:58.000000000 +0100
+++ linux/hpc/kcom.c	2006-11-13 16:02:53.000000000 +0100
@@ -615,6 +615,8 @@
 {
 	struct kcom_node *node;
 
+	OMDEBUG_KCOMD( 3, "Adding new socket node\n");
+
 	node=kmem_cache_alloc(kcom_node_cachep, SLAB_KERNEL);
 	if (!node) {
 		OMBUG("Unable to allocate node space.\n");