On Thu, 16 Jul 2015, Miroslav Benes wrote: > And I made a mistake too. Of course we have access to all these > information as long as we are hooked somewhere in load_module. So... > > 1. We can switch from COMING notifier to hook in a load_module(). This is > certainly possible (we do it in kGraft that way). We would have all the > information need to correctly process dynrelas in > klp_write_module_reloc(), because arch specific things are freed at the > end of load_module(). > > It would allow us to reject loading a patch module if the patching fails > somehow. > > BUT this does not work for the to-be-patched modules loaded afterwards. It > is only a variant of your approach (which offloads the work to kernel > module loader) but we still do it ourselves in klp. > > EDIT: The hook in load_module() is not even necessary. COMING notifiers > are called before arch specific info is freed (if I am not wrong) so we > have all the info in a notifier even now (I did not test it though). And the info should be there even when module init function is called, so we wouldn't need to process relocations for patch module in the notifier (which would be ugly :)). Miroslav -- To unsubscribe from this list: send the line "unsubscribe live-patching" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
