Re: [PATCH -tip 5/5] kprobes: Don't call BUG_ON if there is a kprobe in use on free list

From: Masami Hiramatsu
Date: Tue Sep 11 2018 - 02:36:38 EST


On Mon, 10 Sep 2018 14:23:56 +0200
Ingo Molnar <mingo@xxxxxxxxxx> wrote:

>
> * Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote:
>
> > Instead of calling BUG_ON, if we find a kprobe in use on free kprobe
> > list, just remove it from the list and keep it on kprobe hash list
> > as same as other in-use kprobes.
> >
> > Signed-off-by: Masami Hiramatsu <mhiramat@xxxxxxxxxx>
> > ---
> > kernel/kprobes.c | 8 +++++++-
> > 1 file changed, 7 insertions(+), 1 deletion(-)
> >
> > diff --git a/kernel/kprobes.c b/kernel/kprobes.c
> > index 63c342e5e6c3..e3420364b415 100644
> > --- a/kernel/kprobes.c
> > +++ b/kernel/kprobes.c
> > @@ -546,8 +546,14 @@ static void do_free_cleaned_kprobes(void)
> > struct optimized_kprobe *op, *tmp;
> >
> > list_for_each_entry_safe(op, tmp, &freeing_list, list) {
> > - BUG_ON(!kprobe_unused(&op->kp));
> > list_del_init(&op->list);
> > + if (!kprobe_unused(&op->kp)) {
> > + /*
> > + * This must not happen, but if there is a kprobe
> > + * still in use, keep it on kprobes hash list.
> > + */
> > + continue;
>
> If this is an 'impossible' code path then I think it would make sense to add a WARN_ON_ONCE()
> here.

I agree. This means something goes wrong. That is enough reason to warn user.

Thank you!

>
> Thanks,
>
> Ingo


--
Masami Hiramatsu <mhiramat@xxxxxxxxxx>