On Wed 2023-01-25 15:22:48, Josh Poimboeuf wrote: > On Tue, Jan 17, 2023 at 04:01:57PM +0100, Petr Mladek wrote: > > IMHO, this is the reason why we should make it per-object. > > > > If the shadow variable was used by a livepatched module and we remove > > this module then the shadow variables would get unmaintained. It would > > results in the problem described in this paragraph. > > Yes, that makes sense. Ok, I'm convinced. Thanks! > BTW, this is yet another unfortunate consequence of our decision many > years ago to break the module dependency between a livepatch module and > the modules it patches. We already have a lot of technical debt as a > result of that decision and it continues to pile up. > > In that vein see also Song's and my recent patches to fix module > re-patching. Yeah. Just for record, I have played with splitting the livepatch module some years ago. It was quite tricky. The main problem was loading all the needed livepatch modules and synchronizing their load with the livepatched modules. Few more details were mentioned in https://lore.kernel.org/r/Ytp+u2mGPk5+7Tvf@alley Best Regards, Petr
