Re: [PATCH 2/3] module: Add text_mutex lockdep assertions for page attribute changes

[Josh Poimboeuf <jpoimboe@redhat.com>]

On Fri, Jun 14, 2019 at 04:04:57PM +0200, Petr Mladek wrote:
> On Thu 2019-06-13 20:07:23, Josh Poimboeuf wrote:
> > External callers of the module page attribute change functions now need
> > to have the text_mutex.  Enforce that with lockdep assertions.
> > 
> > diff --git a/kernel/module.c b/kernel/module.c
> > index 6e6712b3aaf5..e43a90ee2d23 100644
> > --- a/kernel/module.c
> > +++ b/kernel/module.c
> > @@ -3519,7 +3534,7 @@ static noinline int do_init_module(struct module *mod)
> >  	/* Switch to core kallsyms now init is done: kallsyms may be walking! */
> >  	rcu_assign_pointer(mod->kallsyms, &mod->core_kallsyms);
> >  #endif
> > -	module_enable_ro(mod, true);
> > +	__module_enable_ro(mod, true);
> 
> This one must be called under text_mutex. Otherwise it might get
> called when ftrace is in the middle of modifying the functions.
> 
> It should be enough to take text_mutex right around this call.
> It will prevent making the code ro when ftrace is doing
> the modification. It safe also the other way.
> set_all_modules_text_ro() does not call frob_ro_after_init().
> Therefore ftrace could not make the after_init section RO prematurely.
> 
> >  	mod_tree_remove_init(mod);
> >  	module_arch_freeing_init(mod);
> >  	mod->init_layout.base = NULL;
> > @@ -3626,8 +3641,8 @@ static int complete_formation(struct module *mod, struct load_info *info)
> >  	/* This relies on module_mutex for list integrity. */
> >  	module_bug_finalize(info->hdr, info->sechdrs, mod);
> >  
> > -	module_enable_ro(mod, false);
> > -	module_enable_nx(mod);
> > +	__module_enable_ro(mod, false);
> > +	__module_enable_nx(mod);
> 
> This one is OK. It is called when the module is in
> MODULE_STATE_UNFORMED. Therefore it is ignored by ftrace.
> The module state is manipulated and checked under module_mutex.

Yes, good catch.  Thanks.

-- 
Josh