RE: [slurm-dev] Gres-plugin leaks memory?

[Jette, Moe]

The attached patch should fix the memory leak. This occurs only with
select/cons_res, sched/backfill, and generic resources configured.
The change will be in SLURM version 2.2.5. Please verify that it fixes 
your memory leak.

Index: src/plugins/select/cons_res/select_cons_res.c
===================================================================
--- src/plugins/select/cons_res/select_cons_res.c       (revision 23019)
+++ src/plugins/select/cons_res/select_cons_res.c       (working copy)
@@ -493,10 +493,15 @@
 static void _destroy_node_data(struct node_use_record *node_usage,
                               struct node_res_record *node_data)
 {
+       int i;
+
        xfree(node_data);
        if (node_usage) {
-               if (node_usage->gres_list)
-                       list_destroy(node_usage->gres_list);
+               for (i = 0; i < select_node_cnt; i++) {
+                       if (node_usage[i].gres_list) {
+                               list_destroy(node_usage[i].gres_list);
+                       }
+               }
                xfree(node_usage);
        }
 }

________________________________________
From: owner-slurm-...@lists.llnl.gov [owner-slurm-...@lists.llnl.gov] On Behalf 
Of Tommi T [tommi_...@yahoo.com]
Sent: Tuesday, April 05, 2011 12:33 AM
To: slurm-dev@lists.llnl.gov
Subject: [slurm-dev] Gres-plugin leaks memory?

Hello

After configuring gres-plugin, slurmctld started to leak memory a lot:

==25614== 29,259,064 (12,594,792 direct, 16,664,272 indirect) bytes in 1,229 
blocks are definitely lost in loss record 369 of 369
==25614==    at 0x4C20E1C: malloc (vg_replace_malloc.c:195)
==25614==    by 0x478ADC: slurm_xmalloc (xmalloc.c:94)
==25614==    by 0x47C25E: list_alloc_aux (list.c:964)
==25614==    by 0x47C951: list_create (list.c:905)
==25614==    by 0x4FBC3C: gres_plugin_node_state_dup (gres.c:1709)
==25614==    by 0x65B950A: _dup_node_usage (select_cons_res.c:401)

Here are config files and full valgrind log

http://www.puuppa.org/~teve/slurm-current.conf
http://www.puuppa.org/~teve/val2.out.bz2
http://www.puuppa.org/~teve/gres.conf

TIA,

Tommi




Index: src/plugins/select/cons_res/select_cons_res.c
===================================================================
--- src/plugins/select/cons_res/select_cons_res.c	(revision 23019)
+++ src/plugins/select/cons_res/select_cons_res.c	(working copy)
@@ -493,10 +493,15 @@
 static void _destroy_node_data(struct node_use_record *node_usage,
 			       struct node_res_record *node_data)
 {
+	int i;
+
 	xfree(node_data);
 	if (node_usage) {
-		if (node_usage->gres_list)
-			list_destroy(node_usage->gres_list);
+		for (i = 0; i < select_node_cnt; i++) {
+			if (node_usage[i].gres_list) {
+				list_destroy(node_usage[i].gres_list);
+			}
+		}
 		xfree(node_usage);
 	}
 }