On Tue Mar 26, 2024 at 1:50 AM EET, Masami Hiramatsu (Google) wrote:
> On Tue, 26 Mar 2024 00:09:42 +0200
> "Jarkko Sakkinen" <jarkko@xxxxxxxxxx> wrote:
>
> > On Mon Mar 25, 2024 at 11:55 PM EET, Jarkko Sakkinen wrote:
> > > +#ifdef CONFIG_MODULES
> > > if (register_module_notifier(&trace_kprobe_module_nb))
> > > return -EINVAL;
> > > +#endif /* CONFIG_MODULES */
> >
> > register_module_notifier() does have "dummy" version but what
> > would I pass to it. It makes more mess than it cleans to declare
> > also a "dummy" version of trace_kprobe_module_nb.
>
> That is better than having #ifdef in the function.
>
> >
> > The callback itself has too tight module subsystem bindings so
> > that they could be simply flagged with IS_DEFINED() (or correct
> > if I'm mistaken, this the conclusion I've ended up with).
>
> Please try this.
>
> -----
> diff --git a/kernel/kprobes.c b/kernel/kprobes.c
> index 70dc6179086e..bc98db14927f 100644
> --- a/kernel/kprobes.c
> +++ b/kernel/kprobes.c
> @@ -2625,6 +2625,7 @@ static void remove_module_kprobe_blacklist(struct module *mod)
> }
> }
>
> +#ifdef CONFIG_MODULES
> /* Module notifier call back, checking kprobes on the module */
> static int kprobes_module_callback(struct notifier_block *nb,
> unsigned long val, void *data)
> @@ -2675,6 +2676,9 @@ static int kprobes_module_callback(struct notifier_block *nb,
> mutex_unlock(&kprobe_mutex);
> return NOTIFY_DONE;
> }
> +#else
> +#define kprobes_module_callback (NULL)
> +#endif
>
> static struct notifier_block kprobe_module_nb = {
> .notifier_call = kprobes_module_callback,
> @@ -2739,7 +2743,7 @@ static int __init init_kprobes(void)
> err = arch_init_kprobes();
> if (!err)
> err = register_die_notifier(&kprobe_exceptions_nb);
> - if (!err)
> + if (!err && IS_ENABLED(CONFIG_MODULES))
> err = register_module_notifier(&kprobe_module_nb);
>
> kprobes_initialized = (err == 0);
OK, thanks for the suggestion WFM.
I'll give this also a spin with VisionFive2 RISC-V SBC before sending
v6.
BR, Jarkko
