On Mon, Aug 06, 2018 at 12:05:30PM +0200, Stefan Sperling wrote: > On Mon, Aug 06, 2018 at 11:42:51AM +0200, Florian Obser wrote: > > I don't understand why we have a limit on the join list. > > The ioctl interface is easier to deal with if the max size is known in > advance. The recalloc dance in benno's diff isn't actually necessary > (nevermind that it's still enforcing some other arbitrary limit of ten > times the node cache size). > > The current limit seems to be too low for some people. I'd suggest > we just keep bumping it up until people stop complaining.
Well, I disagree, but no matter, I said my piece and will not argue this further. Anyway, we have bigger fish to fry. For example telling userland that we hit the limit and not just silently ignore it... > > > It's the users decision to drive their kernel out of memory. > > But it's also not a great idea to deliberately allow userspace to do that. > I don't see why this list should be allowed to run the kernel out of memory. > > Having some fixed size limit also helps if we decide to add another > persistent storage solution later on. AFAIK this is still being debated. > -- I'm not entirely sure you are real.
