[PATCH] Fix for module initialization failure

[PATCH] Fix for module initialization failure

[Rusty Russell]

Need a module_arch_cleanup() after a successful module_finalize()
call: it was missing in the case where module->init() fails.  Since
module_arch_cleanup() is a noop on x86, I didn't spot it earlier.

Thanks to Paul Mackerras for prodding me about this again...

Rusty.

Name: Always call module_arch_cleanup
Author: Rusty Russell
Status: Booted on 2.6.0-test5-bk1

D: Bug reported by Paul Mackerras: if a module parameter fails, we didn't
D: call module_arch_cleanup().

diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal linux-2.6.0-test5-bk1/kernel/module.c working-2.6.0-test5-bk1-module-greg-orig/kernel/module.c
--- linux-2.6.0-test5-bk1/kernel/module.c	2003-09-09 10:35:05.000000000 +1000
+++ working-2.6.0-test5-bk1-module-greg-orig/kernel/module.c	2003-09-11 15:14:32.000000000 +1000
@@ -1653,7 +1660,7 @@ static struct module *load_module(void _
 				 NULL);
 	}
 	if (err < 0)
-		goto cleanup;
+		goto arch_cleanup;
 
 	/* Get rid of temporary copy */
 	vfree(hdr);
@@ -1661,6 +1668,8 @@ static struct module *load_module(void _
 	/* Done! */
 	return mod;
 
+ arch_cleanup:
+	module_arch_cleanup(mod);
  cleanup:
 	module_unload_free(mod);
 	module_free(mod, mod->module_init);

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

Re: [PATCH] Fix for module initialization failure

[Paul Mackerras]

Rusty Russell writes:

> Need a module_arch_cleanup() after a successful module_finalize()
> call: it was missing in the case where module->init() fails.  Since
> module_arch_cleanup() is a noop on x86, I didn't spot it earlier.
> 
> Thanks to Paul Mackerras for prodding me about this again...

Linus: this fixes a real bug which causes an oops on PPC in the case
where you try to load one module, but the init routine fails, and then
you try to load another module.  What happens is that the
module_finalize() call on the first module adds the module to a list
used in BUG() handling.  If the init routine fails, the code currently
doesn't call module_arch_cleanup(), which is where we remove the
module from the list used for BUG handling.  When you try to load the
next module, the BUG list head is pointing to a module which is no
longer there and you get an oops.

Please apply.

Thanks,
Paul.