[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20231002162133.GC5054@cmpxchg.org>
Date: Mon, 2 Oct 2023 12:21:33 -0400
From: Johannes Weiner <hannes@...xchg.org>
To: Michal Hocko <mhocko@...e.com>
Cc: Nhat Pham <nphamcs@...il.com>, akpm@...ux-foundation.org,
riel@...riel.com, roman.gushchin@...ux.dev, shakeelb@...gle.com,
muchun.song@...ux.dev, tj@...nel.org, lizefan.x@...edance.com,
shuah@...nel.org, mike.kravetz@...cle.com, yosryahmed@...gle.com,
linux-mm@...ck.org, kernel-team@...a.com,
linux-kernel@...r.kernel.org, cgroups@...r.kernel.org
Subject: Re: [PATCH v2 1/2] hugetlb: memcg: account hugetlb-backed memory in
memory controller
On Mon, Oct 02, 2023 at 03:43:19PM +0200, Michal Hocko wrote:
> We should also consider the global control for this functionality. I am
> especially worried about setups where a mixed bag of workloads
> (containers) is executed. While some of them will be ready for the new
> accounting mode many will leave in their own world without ever being
> modified. How do we deal with that situation?
It's possible to add more localized control on top of the global flag
should this come up. But this seems like a new and honestly pretty
hypothetical usecase, given the host-level coordination already
involved in real-world hugetlb setups.
The same could be said about other mount options, such as nsdelegate,
memory_localevents, and memory_recursiveprot. Those you'd expect to
have a much broader audience, and nobody has asked for mixed use.
Let's cross this bridge not when but if we have to.
Powered by blists - more mailing lists