On 10/11/18 12:33, Alan Tull wrote: > On Thu, Oct 11, 2018 at 12:39 AM Frank Rowand <frowand.list@xxxxxxxxx> wrote: > > [resend of my messed up rejected email of a minute ago, sorry] > >> >> On 10/10/18 14:03, Frank Rowand wrote: < snip > > I understand you're quite busy with all this, but I'm wondering > whether it might be worth it go ahead and make the properties be > kernel objects also at this point. That would be an improvement for > the case of overlay properties added to non-overlay nodes, so the > lifespan of the overlay property memory can be coupled with the > properties kobj's instead of the node kobj's. > > Alan > That is one of the approaches that I am thinking about to handle the potential memory leaks from those properties. I'd like to make these changes in a step wise fashion, to let each major change get some exposure and use before moving on to the next step. Making properties into kernel objects would impact a lot of code. So not in this series. But thanks for thinking about it. -Frank