Re: [PATCH 08/19] KVM: arm64: vgic-v3: Push user access into vgic_v3_cpu_sysregs_uaccess()

From: Reiji Watanabe <reijiw@google.com>
To: Marc Zyngier <maz@kernel.org>
Cc: kvmarm@lists.cs.columbia.edu, kvm@vger.kernel.org,
	Linux ARM <linux-arm-kernel@lists.infradead.org>,
	Schspa Shi <schspa@gmail.com>,
	kernel-team@android.com, Oliver Upton <oliver.upton@linux.dev>
Subject: Re: [PATCH 08/19] KVM: arm64: vgic-v3: Push user access into vgic_v3_cpu_sysregs_uaccess()
Date: Tue, 12 Jul 2022 20:26:25 -0700	[thread overview]
Message-ID: <CAAeT=FwQtyfKPWfSz4jzY4nAssqxoFJ2HkWsEcKedvgBGckqQw@mail.gmail.com> (raw)
In-Reply-To: <87fsj6lt4s.wl-maz@kernel.org>

Hi Marc,

On Mon, Jul 11, 2022 at 11:53 PM Marc Zyngier <maz@kernel.org> wrote:
>
> Hi Reiji,
>
> On Tue, 12 Jul 2022 07:11:39 +0100,
> Reiji Watanabe <reijiw@google.com> wrote:
> >
> > Hi Marc,
> >
> > On Wed, Jul 6, 2022 at 9:43 AM Marc Zyngier <maz@kernel.org> wrote:
> > >
> > > In order to start making the vgic sysreg access from userspace
> > > similar to all the other sysregs, push the userspace memory
> > > access one level down into vgic_v3_cpu_sysregs_uaccess().
> > >
> > > The next step will be to rely on the sysreg infrastructure
> > > to perform this task.
> > >
> > > Signed-off-by: Marc Zyngier <maz@kernel.org>
> > > ---
> > >  arch/arm64/kvm/vgic-sys-reg-v3.c      | 22 +++++++++++++------
> > >  arch/arm64/kvm/vgic/vgic-kvm-device.c | 31 ++++++---------------------
> > >  arch/arm64/kvm/vgic/vgic.h            |  4 ++--
> > >  3 files changed, 23 insertions(+), 34 deletions(-)
> > >
> > > diff --git a/arch/arm64/kvm/vgic-sys-reg-v3.c b/arch/arm64/kvm/vgic-sys-reg-v3.c
> > > index 85a5e1d15e9f..8c56e285fde9 100644
> > > --- a/arch/arm64/kvm/vgic-sys-reg-v3.c
> > > +++ b/arch/arm64/kvm/vgic-sys-reg-v3.c
> > > @@ -278,15 +278,21 @@ int vgic_v3_has_cpu_sysregs_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *
> > >         return -ENXIO;
> > >  }
> > >
> > > -int vgic_v3_cpu_sysregs_uaccess(struct kvm_vcpu *vcpu, bool is_write, u64 id,
> > > -                               u64 *reg)
> > > +int vgic_v3_cpu_sysregs_uaccess(struct kvm_vcpu *vcpu,
> > > +                               struct kvm_device_attr *attr,
> > > +                               bool is_write)
> > >  {
> > > +       u64 __user *uaddr = (u64 __user *)(long)attr->addr;
> > >         struct sys_reg_params params;
> > >         const struct sys_reg_desc *r;
> > > -       u64 sysreg = (id & KVM_DEV_ARM_VGIC_SYSREG_MASK) | KVM_REG_SIZE_U64;
> > > +       u64 sysreg;
> > >
> > > -       if (is_write)
> > > -               params.regval = *reg;
> > > +       sysreg = (attr->attr & KVM_DEV_ARM_VGIC_SYSREG_MASK) | KVM_REG_SIZE_U64;
> >
> > Why don't you use attr_to_id() here ?
>
> This actually happens in the following patch. Happy to move the change
> here though.

IMHO it makes more sense to have the change in this patch
because vgic_v3_cpu_sysregs_uaccess in this patch already
takes "attr" as the arguments.
Having said that, I'm fine with keeping the current code:)

> >
> >
> > > +
> > > +       if (is_write) {
> > > +               if (get_user(params.regval, uaddr))
> > > +                       return -EFAULT;
> > > +       }
> > >         params.is_write = is_write;
> > >
> > >         r = find_reg_by_id(sysreg, &params, gic_v3_icc_reg_descs,
> > > @@ -297,8 +303,10 @@ int vgic_v3_cpu_sysregs_uaccess(struct kvm_vcpu *vcpu, bool is_write, u64 id,
> > >         if (!r->access(vcpu, &params, r))
> > >                 return -EINVAL;
> > >
> > > -       if (!is_write)
> > > -               *reg = params.regval;
> > > +       if (!is_write) {
> > > +               if (put_user(params.regval, uaddr))
> > > +                       return -EFAULT;
> > > +       }
> > >
> > >         return 0;
> > >  }
> > > diff --git a/arch/arm64/kvm/vgic/vgic-kvm-device.c b/arch/arm64/kvm/vgic/vgic-kvm-device.c
> > > index c6d52a1fd9c8..d8269300632d 100644
> > > --- a/arch/arm64/kvm/vgic/vgic-kvm-device.c
> > > +++ b/arch/arm64/kvm/vgic/vgic-kvm-device.c
> > > @@ -561,14 +561,9 @@ static int vgic_v3_attr_regs_access(struct kvm_device *dev,
> > >                 if (!is_write)
> > >                         *reg = tmp32;
> > >                 break;
> > > -       case KVM_DEV_ARM_VGIC_GRP_CPU_SYSREGS: {
> > > -               u64 regid;
> > > -
> > > -               regid = (attr->attr & KVM_DEV_ARM_VGIC_SYSREG_INSTR_MASK);
> > > -               ret = vgic_v3_cpu_sysregs_uaccess(vcpu, is_write,
> > > -                                                 regid, reg);
> > > +       case KVM_DEV_ARM_VGIC_GRP_CPU_SYSREGS:
> > > +               ret = vgic_v3_cpu_sysregs_uaccess(vcpu, attr, is_write);
> >
> > Nit: Since @reg that is passed to vgic_v3_attr_regs_access() will be NULL
> > for KVM_DEV_ARM_VGIC_GRP_CPU_SYSREGS, I think it would be more clear
> > if you could update the comment for vgic_v3_attr_regs_access accordingly.
> >
> > ----
> > /*
> >  * vgic_v3_attr_regs_access - allows user space to access VGIC v3 state
> >  *
> >  * @dev:      kvm device handle
> >  * @attr:     kvm device attribute
> >  * @reg:      address the value is read or written
> >    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >  * @is_write: true if userspace is writing a register
> >  */
> > static int vgic_v3_attr_regs_access(struct kvm_device *dev,
> >                                     struct kvm_device_attr *attr,
> >                                     u64 *reg, bool is_write)
>
> @reg disappears completely in patch #12. Do you see value in rewriting
> this comment even if I end-up removing it 4 patches down the line?

Oh, I haven't seen PATCH#12 yet. Then let me withdraw my comment.

Thank you,
Reiji