Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210427223635.2711774-1-bgardon@google.com> <20210427223635.2711774-7-bgardon@google.com>
 <1b598516-4478-4de2-4241-d4b517ec03fa@redhat.com>
In-Reply-To: <1b598516-4478-4de2-4241-d4b517ec03fa@redhat.com>
From:   Ben Gardon <bgardon@google.com>
Date:   Wed, 28 Apr 2021 09:45:56 -0700
Message-ID: <CANgfPd_o+UA4ry7Kpw3WbcPZYm32r+1o=hQmZdazsrZvO4aynA@mail.gmail.com>
Subject: Re: [PATCH 6/6] KVM: x86/mmu: Lazily allocate memslot rmaps
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     LKML <linux-kernel@vger.kernel.org>, kvm <kvm@vger.kernel.org>,
        Peter Xu <peterx@redhat.com>,
        Sean Christopherson <seanjc@google.com>,
        Peter Shier <pshier@google.com>,
        Junaid Shahid <junaids@google.com>,
        Jim Mattson <jmattson@google.com>,
        Yulei Zhang <yulei.kernel@gmail.com>,
        Wanpeng Li <kernellwp@gmail.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Xiao Guangrong <xiaoguangrong.eric@gmail.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Apr 28, 2021 at 3:04 AM Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> On 28/04/21 00:36, Ben Gardon wrote:
> > -static int kvm_alloc_memslot_metadata(struct kvm_memory_slot *slot,
> > +static int kvm_alloc_memslot_metadata(struct kvm *kvm,
> > +                                   struct kvm_memory_slot *slot,
> >                                     unsigned long npages)
> >   {
> >       int i;
> > @@ -10892,7 +10950,7 @@ static int kvm_alloc_memslot_metadata(struct kvm_memory_slot *slot,
> >        */
> >       memset(&slot->arch, 0, sizeof(slot->arch));
> >
> > -     r = alloc_memslot_rmap(slot, npages);
> > +     r = alloc_memslot_rmap(kvm, slot, npages);
> >       if (r)
> >               return r;
> >
>
> I wonder why you need alloc_memslot_rmap at all here in
> kvm_alloc_memslot_metadata, or alternatively why you need to do it in
> kvm_arch_assign_memslots.  It seems like only one of those would be
> necessary.

Oh, that's a good point. We need it in kvm_arch_assign_memslots
because of the race I identified in the thread for patch 5 of this
series, but we could remove it from kvm_alloc_memslot_metadata with
this patch.

Of course, it would be much nicer if we could just keep it in
kvm_alloc_memslot_metadata and find some other way to stop memslots
without rmaps from being inappropriately installed, if anyone has a
simpler way to accomplish that.

>
> Paolo
>