Re: [RFC][PATCH] module: Propagate MODULE_STATE_COMING notifier errors

[Petr Mladek <pmladek@suse.com>]

On Wed 2019-06-19 13:12:12, Miroslav Benes wrote:
> On Mon, 17 Jun 2019, Peter Zijlstra wrote:
> > --- a/kernel/module.c
> > +++ b/kernel/module.c
> > @@ -3638,9 +3638,10 @@ static int prepare_coming_module(struct module *mod)
> >  	if (err)
> >  		return err;
> >  
> > -	blocking_notifier_call_chain(&module_notify_list,
> > -				     MODULE_STATE_COMING, mod);
> > -	return 0;
> > +	ret = blocking_notifier_call_chain(&module_notify_list,
> > +					   MODULE_STATE_COMING, mod);
> > +	ret = notifier_to_errno(ret);
> > +	return ret;
> >  }
> >  
> >  static int unknown_module_param_cb(char *param, char *val, const char *modname,
> > @@ -3780,7 +3781,7 @@ static int load_module(struct load_info *info, const char __user *uargs,
> >  
> >  	err = prepare_coming_module(mod);
> >  	if (err)
> > -		goto bug_cleanup;
> > +		goto coming_cleanup;
> 
> Not good. klp_module_going() is not prepared to be called without 
> klp_module_coming() succeeding. "Funny" things might happen.

We have discussed it with Miroslav. The best solution seems to
be to allow to call klp_module_going() even when
klp_module_comming() failed. Something like:

diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
index c4ce08f43bd6..42d72b62edb2 100644
--- a/kernel/livepatch/core.c
+++ b/kernel/livepatch/core.c
@@ -1188,6 +1188,10 @@ void klp_module_going(struct module *mod)
                return;
 
        mutex_lock(&klp_mutex);
+
+       /* Nope when klp_module_comming failed */
+       if (!mod->klp_alive)
+               goto out;
        /*
         * Each module has to know that klp_module_going()
         * has been called. We never know what module will
@@ -1196,7 +1200,7 @@ void klp_module_going(struct module *mod)
        mod->klp_alive = false;
 
        klp_cleanup_module_patches_limited(mod, NULL);
-
+out:
        mutex_unlock(&klp_mutex);
 }

Peter's patch actually allows to convert the klp hooks into
proper module notifiers. But we could do it later once this
patch is upstream or queued.

Best Regards,
Petr