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:   Tue, 9 Aug 2022 23:51:41 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     David Matlack <dmatlack@google.com>
Cc:     Vipin Sharma <vipinsh@google.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        kvm list <kvm@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] KVM: x86/mmu: Make page tables for eager page splitting
 NUMA aware
Message-ID: <YvLzDUjpJHmZtn0i@google.com>
References: <20220801151928.270380-1-vipinsh@google.com>
 <YuhPT2drgqL+osLl@google.com>
 <YuhoJUoPBOu5eMz8@google.com>
 <YulRZ+uXFOE1y2dj@google.com>
 <YuldSf4T2j4rIrIo@google.com>
 <4ccbafb5-9157-ec73-c751-ec71164f8688@redhat.com>
 <Yul3A4CmaAHMui2Z@google.com>
 <cedcced0-b92c-07bd-ef2b-272ae58fdf40@redhat.com>
 <CAHVum0c=s8DH=p8zJcGzYDsfLY_qHEmvD1uF58h5WoUk6ZF8rQ@mail.gmail.com>
 <CALzav=ccxkAWk7ddqbJ_qPL2-=bXVZUEpWgwKpJ1oCtc_8w7WQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CALzav=ccxkAWk7ddqbJ_qPL2-=bXVZUEpWgwKpJ1oCtc_8w7WQ@mail.gmail.com>
Precedence: bulk

On Tue, Aug 09, 2022, David Matlack wrote:
> On Fri, Aug 5, 2022 at 4:30 PM Vipin Sharma <vipinsh@google.com> wrote:
> > Approach B:
> > Ask page from the specific node on fault path with option to fallback
> > to the original cache and default task policy.
> >
> > This is what Sean's rough patch looks like.
> 
> This would definitely be a simpler approach but could increase the
> amount of time a vCPU thread holds the MMU lock when handling a fault,
> since KVM would start performing GFP_NOWAIT allocations under the
> lock. So my preference would be to try the cache approach first and
> see how complex it turns out to be.

Ya, as discussed off-list, I don't like my idea either :-)

The pfn and thus node information is available before mmu_lock is acquired, so I
don't see any reason to defer the allocation other than to reduce the memory
footprint, and that's a solvable problem one way or another.