Petr Mladek <pmladek@xxxxxxxx> wrote on Thu [2021-Dec-16 15:14:38 +0100]:
> > Now it does. In the past, I think we did create some memory. I know
> > when we hook debugobjects up to kobjects (there's an external patch for
> > that floating around somewhere), that is one reason to keep the
> > kobject_put() rule, and there might have been other reasons in the past
> > 20+ years as well.
> >
> > So yes, while you are correct today, the "normal" reference counted
> > object model patern is "after the object is initialized, it MUST only be
> > freed by handling its reference count." So let's stick to that rule for
> > now.
>
> Good point.
Thanks for the discussion all. I think we've landed on the fact that this
is a refcounting bug that needs to be fixed, but isn't a leak in how the
kobject implementation exists today.
Petr - are you OK with me sending out a v3 of the patch with the following
changes:
- The patch description is updated to not claim that a leak is being
fixed, but rather that a kobject reference counting bug is being fixed.
- All of the NULL checking in klp_init_patch_early() is brought into
klp_enable_patch(), and klp_init_patch_early() is updated to be void,
per Josh's suggestion. This would address the refcounting issue and IMO
also simplifies and improves the code. I know you were onboard with
moving try_module_get() into klp_enable_patch(), but I don't think we
ever resolved the point about moving the rest of the NULL checking out
as well.
Regards,
David
