Re: [PATCH 1/1] ksm: introduce ksm_max_page_sharing per page deduplication limit

[Rik van Riel <riel@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 840 bytes --]

On Tue, 2015-11-10 at 19:44 +0100, Andrea Arcangeli wrote:
> Without a max deduplication limit for each KSM page, the list of the
> rmap_items associated to each stable_node can grow infinitely
> large.
> 
> During the rmap walk each entry can take up to ~10usec to process
> because of IPIs for the TLB flushing (both for the primary MMU and
> the
> secondary MMUs with the MMU notifier). With only 16GB of address
> space
> shared in the same KSM page, that would amount to dozens of seconds
> of
> kernel runtime.

Silly question, but could we fix this problem
by building up a bitmask of all CPUs that have
a page-with-high-mapcount mapped, and simply
send out a global TLB flush to those CPUs once
we have changed the page tables, instead of
sending out IPIs at every page table change?

-- 
All rights reversed

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 473 bytes --]