On Thu, Feb 23, 2017 at 04:31:07PM +0100, Michal Hocko wrote:

On Thu 23-02-17 14:36:39, Laurent Dufour wrote:

quoted

Until a soft limit is set to a cgroup, the soft limit data are useless
so delay this allocation when a limit is set.

Hmm, I am still undecided whether this is actually worth it. On one hand
distribution kernels tend to have quite large NUMA_SHIFT (e.g. SLES has
NUMA_SHIFT=10 and then we will save 8kB+12kB which is not hell of a lot
but always good if we can save that, especially for a rarely used
feature. The code grown on the other hand (it was in __init section
previously) which is a minus, on the other hand.

What do you think Johannes?

Hohumm, saving 5 pages on a NUMA machine vs. the additional complexity
and the increased risk of memory problems when somebody sets up a soft
limit after some uptime... I don't think I can give a strong yes or no
on this one, so inertia wins for me; I'd just leave it alone.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>