Re: [PATCH v2 7/9] x86/module: Use text_poke() for late relocations

[Josh Poimboeuf]

On Fri, Apr 17, 2020 at 04:29:44PM +0200, Peter Zijlstra wrote:
> On Fri, Apr 17, 2020 at 09:04:32AM -0500, Josh Poimboeuf wrote:
> > +int apply_relocate_add(Elf64_Shdr *sechdrs,
> > +		   const char *strtab,
> > +		   unsigned int symindex,
> > +		   unsigned int relsec,
> > +		   struct module *me)
> > +{
> > +	int ret;
> > +	bool early = me->state == MODULE_STATE_UNFORMED;
> > +	void *(*write)(void *, const void *, size_t) = memcpy;
> > +
> > +	if (!early) {
> > +		write = text_poke;
> > +		mutex_lock(&text_mutex);
> > +	}
> > +
> > +	ret = __apply_relocate_add(sechdrs, strtab, symindex, relsec, me,
> > +				   write);
> > +
> > +	if (!early) {
> > +		mutex_unlock(&text_mutex);
> > +		text_poke_sync();
> 
> I'm thinking text_poke_sync() wants to be inside text_mutex. Although
> given that nothing should be running that text, it really doesn't
> matter.

Yeah, makes sense.

diff --git a/arch/x86/kernel/module.c b/arch/x86/kernel/module.c
index 2a997afa04c6..23c95a53d20e 100644
--- a/arch/x86/kernel/module.c
+++ b/arch/x86/kernel/module.c
@@ -237,8 +237,8 @@ int apply_relocate_add(Elf64_Shdr *sechdrs,
 				   write);
 
 	if (!early) {
-		mutex_unlock(&text_mutex);
 		text_poke_sync();
+		mutex_unlock(&text_mutex);
 	}
 
 	return ret;