On Wed, Aug 26, 2020 at 8:15 AM Jann Horn <jannh@xxxxxxxxxx> wrote: > > A downside of this approach is that we now need a bigger amount of kernel > memory per userspace VMA in the normal ELF case, and that we need O(n) > kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't > be terribly bad. So this looks much simpler now. But it also makes it more obvious how that dump-size callback is kind of pointless. Why does elf_fdpic have different heuristics than regular elf? And not in meaningful ways - the heuristics look basically identical, just with different logging and probably random other differences that have mostly just grown over time. So rather than the callback function pointer, I think you should just copy the ELF version of the dump_size() logic, and get rid of a very odd and strange callback. But even in this form, at least this patch doesn't make the code look _worse_ than it used to, so while I would like to see a further cleanup I no longer dislike it. Linus