Skip to contentSkip to conversation

Modules race fix

PATCH New Rusty Russell 1 participant 1 message Feb 26, 2003

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Modules race fix

Linus, please apply.

Name: Fix two module races
Author: Bob Miller, Rusty Russell
Status: Trivial

D: Bob Miller points out that the try_module_get in use_module() can,
D: of course, fail. Secondly, there is a race between setting the module
D: live, and a simultaneous removal of it.

kernel/module.c +6 -1
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal linux-2.5.62-bk6/kernel/module.c working-2.5.62-bk6-modraces/kernel/module.c
@@ -173,16 +173,19 @@ static int use_module(struct module *a, 
 	struct module_use *use;
 	if (b == NULL || already_uses(a, b)) return 1;
 
+	if (!strong_try_module_get(b))
+		return 0;
+
 	DEBUGP("Allocating new usage for %s.\n", a->name);
 	use = kmalloc(sizeof(*use), GFP_ATOMIC);
 	if (!use) {
 		printk("%s: out of memory loading\n", a->name);
+		module_put(b);
 		return 0;
 	}
 
 	use->module_which_uses = a;
 	list_add(&use->list, &b->modules_which_use_me);
-	try_module_get(b); /* Can't fail */
 	return 1;
 }
 
@@ -1456,10 +1459,12 @@ sys_init_module(void *umod,
 	}
 
 	/* Now it's a first class citizen! */
+	down(&module_mutex);
 	mod->state = MODULE_STATE_LIVE;
 	module_free(mod, mod->module_init);
 	mod->module_init = NULL;
 	mod->init_size = 0;
+	up(&module_mutex);
 
 	return 0;
 }

--
Anyone who quotes me in their sig is an idiot. -- Rusty Russell.