Philipp Kern wrote: > Peter, > > am Mon, Jan 28, 2013 at 09:45:46AM +0100 hast du folgendes geschrieben: >> I would like to know if it would be a good idea to change the page >> allocator >> zeroing > policy from zero on allocate (for GFP_USER) to zero on free >> for the >> -virtual > kernel builds. > > wouldn't it be better to signal the host that the page in question can be > discarded? If the guest really needs to hand out the page and not the > mapping to the zeroed frame it keeps anyway, it seems like a disadvantage > to have a CoW breaking on the host immediately afterwards.
This would be the best solution anyway. But wouldn't this also add performance overhead? What I had in mind was using the zero-on-free page allocator only if inside a VM and if KSM is enabled on the host. In this case one would already have decided to trade in performance for memory. Peter > > Live migration wouldn't need to copy such pages and instead of applying > memory > scanning and merging on those pages they could simply be reassigned to > another > process by the host on memory pressure. > > The pages would probably need to get part of the balloon that's used to > dynamically change memory size, potentially with some additional > signalling > which pages are likely to still be allocated to the guest and a way to > leave > them in the free pool. > > Kind regards > Philipp Kern > -- > ubuntu-devel mailing list > [email protected] > Modify settings or unsubscribe at: > https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel > -- ubuntu-devel mailing list [email protected] Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
