On Fri, Jun 8, 2012 at 11:43 AM, Minchan Kim minchan@kernel.org wrote:
So, the solution would be then two-fold:
- Use your memory pressure notifications. They must be quite fast when
we starting to feel the high pressure. (I see the you use zone_page_state() and friends, which is vm_stat, and it is updated
VM has other information like nr_reclaimed, nr_scanned, nr_congested, recent_scanned, recent_rotated, too. I hope we can make math by them and improve as we improve VM reclaimer.
very infrequently, but to get accurate notification we have to update it much more frequently, but this is very expensive. So KOSAKI and Christoph will complain. :-)
Reclaimer already have used that and if we need accuracy, we handled it like zone_watermark_ok_safe. If it's very inaccurate, VM should be fixed, too.
Exactly. I don't know why people think pushing vmevents to userspace is going to fix any of the hard problems.
Anton, Lenoid, do you see any fundamental issues from userspace point of view with going forward what Minchan is proposing?