x86 part: was: Re: [PATCH v6] livepatch: Clear relocation targets on a module removal

[Petr Mladek]

On Mon 2022-11-28 17:57:06, Song Liu wrote:
> On Fri, Nov 18, 2022 at 8:24 AM Petr Mladek <pmladek at suse.com> wrote:
> --- a/arch/x86/kernel/module.c
> +++ b/arch/x86/kernel/module.c
> > This duplicates a lot of code. Please, rename apply_relocate_add() the
> > same way as __apply_clear_relocate_add() and add the "apply" parameter.
> > Then add the wrappers for this:
> >
> > int write_relocate_add(Elf64_Shdr *sechdrs,
> >                        const char *strtab,
> >                        unsigned int symindex,
> >                        unsigned int relsec,
> >                        struct module *me,
> >                        bool apply)
> > {
> >         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);
> >         }
> 
> How about we move the "early" logic into __write_relocate_add()?

If I get it correctly then __write_relocate_add() has three different
return paths. I am not sure if this could be moved there a reasonable
way.

Anyway, I do not resist on the above proposal. Feel free to find
another solution that reduces the duplicated code and looks
reasonable. I am sure that there are more possibilities.

Best Regards,
Petr