linux-kernel - Re: [PATCH] KVM: x86/mmu: Register MMU shrinker only when necessary

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20240819214014.GA2313467.vipinsh@google.com>
Date: Mon, 19 Aug 2024 14:40:14 -0700
From: Vipin Sharma <vipinsh@...gle.com>
To: Sean Christopherson <seanjc@...gle.com>
Cc: Liang Chen <liangchen.linux@...il.com>, pbonzini@...hat.com,
	kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
	David Matlack <dmatlack@...gle.com>
Subject: Re: [PATCH] KVM: x86/mmu: Register MMU shrinker only when necessary

On 2024-08-14 07:41:19, Sean Christopherson wrote:
> +Vipin and David
> 
> On Wed, Aug 14, 2024, Liang Chen wrote:
> > The shrinker is allocated with TDP MMU, which is meaningless except for
> > nested VMs, and 'count_objects' is also called each time the reclaim
> > path tries to shrink slab caches. Let's allocate the shrinker only when
> > necessary.
> 
> This is definitely not worth the complexity.  In its current form, KVM's shrinker
> is quite useless[1], and there were plans to repurpose the shrinker to free pages
> from the so called "mmu caches"[2], i.e. free pages that are guaranteed to not be
> in use.
> 
> Vipin/David, what happened to that series?  Are we still working on it?
> 
> [1] https://lore.kernel.org/lkml/Y45dldZnI6OIf+a5@google.com
> [2] https://lore.kernel.org/all/20221222023457.1764-2-vipinsh@google.com

NUMA aware page table series got deprioritized, so, MMU shrinker changes
also moved to back burner.

I will extract the patch [2] above and send it as a separate series which
just changes the Shrinker behavior as it is independent of NUMA aware
page table effort.