Hi Christoph, On 2023-12-04T21:12:47-08:00 Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > On Tue, Dec 05, 2023 at 02:28:55AM +0000, SeongJae Park wrote: > > DAMON cannot be used from loadable modules since it is not exporting its > > symbols. > > And that is a good thing. We should absolutely not allow random modules > probing MM internals. I agree. > What you posted here is basically just adding hooks without even real > in-kernel users. I was thinking someone might be able to think even the sample module as real usage since there was actually some questions about it from real users. That said, those were more like questions than strong requests, so I still think this change is somewhat better to be made for at least some folks, but I also agree that this wouldn't be not really essential for everyone, and could be only long term maintenance threat. I personally don't have strong opinion, and thank you for raising your concern. I will hold this patchset unless someone request this change again with good rationale. Btw, I know there were many concerns about unnecessarily exporting symbols, but do we have a formal guideline or documentation about the requirements for exporting symbols in sepcific subsystem? I was hoping to have such one so that I could provide better answer to DAMON's loadable module support questions, but I was unable to find such one with my poor searching skill. This reply could be used for the purpose meanwhile, though, so appreciate again. :) Thanks, SJ
