Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 9 Dec 2022 15:14:32 -0800
From:   David Matlack <dmatlack@google.com>
To:     Ben Gardon <bgardon@google.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        Peter Xu <peterx@redhat.com>,
        Sean Christopherson <seanjc@google.com>,
        Vipin Sharma <vipinsh@google.com>
Subject: Re: [PATCH 0/7] KVM: x86/MMU: Factor rmap operations out of mmu.c
Message-ID: <Y5PBWLEXXMjcyKTS@google.com>
References: <20221206173601.549281-1-bgardon@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20221206173601.549281-1-bgardon@google.com>
Precedence: bulk

On Tue, Dec 06, 2022 at 05:35:54PM +0000, Ben Gardon wrote:
> Move the basic operations for manipulating an rmap out of mmu.c, into
> their own files.

memslot_rmap_alloc() and memslot_rmap_free() from x86.c look like good
candidates to move to rmap.c as well as they are pure setup/teardown.

> 
> This is the first step in a series of series to split the Shadow MMU out of
> mmu.c. Splitting the shadow MMU out results in about a 50% reduction in line
> count of mmu.c, from ~7K to ~3.5K. The rmap operations only represent about
> 10% of the Shadow MMU but are split out first becase the are relatively
> self-contained.
> 
> This split may also help facilitate the addition of kUnit tests for rmap
> manipulation, especially the fiddly bit flag which diferentiates a direct
> pointer from a pte_list_desc.
> 
> This effort is complimentary to David Matlack's proposal to refactor
> the TDP MMU into an arch-neutral implementation because it further
> clarifies the distinction between the Shadow MMU and TDP MMU within x86.
> 
> This series contains no functional changes. It's just a direct movement
> of code from one file to another.
> 
> Whether this rmap code should stay in its own file or get merged with
> the rest of the shadow MMU code as it is factored out is open for
> debate.
> 
> Ben Gardon (7):
>   KVM: x86/MMU: Move pte_list operations to rmap.c
>   KVM: x86/MMU: Move rmap_iterator to rmap.h
>   KVM: x86/MMU: Move gfn_to_rmap() to rmap.c
>   KVM: x86/MMU: Move rmap_can_add() and rmap_remove() to rmap.c
>   KVM: x86/MMU: Move the rmap walk iterator out of mmu.c
>   KVM: x86/MMU: Move rmap zap operations to rmap.c
>   KVM: x86/MMU: Move rmap_add() to rmap.c
> 
>  arch/x86/kvm/Makefile           |   2 +-
>  arch/x86/kvm/debugfs.c          |   1 +
>  arch/x86/kvm/mmu/mmu.c          | 438 +-------------------------------
>  arch/x86/kvm/mmu/mmu_internal.h |   9 +-
>  arch/x86/kvm/mmu/rmap.c         | 368 +++++++++++++++++++++++++++
>  arch/x86/kvm/mmu/rmap.h         | 106 ++++++++
>  6 files changed, 492 insertions(+), 432 deletions(-)
>  create mode 100644 arch/x86/kvm/mmu/rmap.c
>  create mode 100644 arch/x86/kvm/mmu/rmap.h
> 
> -- 
> 2.39.0.rc0.267.gcb52ba06e7-goog
>