On Wed, Nov 07, 2012 at 02:53:49AM -0800, Anton Vorontsov wrote:
Hi all,
This is the third RFC. As suggested by Minchan Kim, the API is much simplified now (comparing to vmevent_fd):
- As well as Minchan, KOSAKI Motohiro didn't like the timers, so the timers are gone now;
- Pekka Enberg didn't like the complex attributes matching code, and so it is no longer there;
- Nobody liked the raw vmstat attributes, and so they were eliminated too.
But, conceptually, it is the exactly the same approach as in v2: three discrete levels of the pressure -- low, medium and oom. The levels are based on the reclaimer inefficiency index as proposed by Mel Gorman, but userland does not see the raw index values. The description why I moved away from reporting the raw 'reclaimer inefficiency index' can be found in v2: http://lkml.org/lkml/2012/10/22/177
While the new API is very simple, it is still extensible (i.e. versioned).
Sorry, I didn't follow previous discussion on this, but could you explain what's wrong with memory notifications from memcg? As I can see you can get pretty similar functionality using memory thresholds on the root cgroup. What's the point?