On Wed, 14 Nov 2012, Andrew Morton wrote: > For this reason and for general ease-of-testing: can and should we add > a knob which will enable users to disable the feature at runtime? That > way if it causes problems or if we suspect it's causing problems, we > can easily verify the theory and offer users a temporary fix. > I think it would be best to add a tunable under /sys/kernel/mm/transparent_hugepage and enable it by default whenever /sys/kernel/mm/transparent_hugepage/enabled is "always" or "madvise" and allocate the huge zero page under such circumstances. Then we can free it if disabled (or if enabled is set to "never") and avoid all the refcounting and lazy allocation that causes a regression on Kirill's benchmark. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>