Re: [PATCH] RISC-V: KVM: Teardown riscv specific bits after kvm_exit

From: Sean Christopherson <seanjc@google.com>
To: Atish Kumar Patra <atishp@rivosinc.com>
Cc: Anup Patel <anup@brainfault.org>,
	Atish Patra <atishp@atishpatra.org>,
	 Paul Walmsley <paul.walmsley@sifive.com>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	 Alexandre Ghiti <alex@ghiti.fr>,
	Andrew Jones <ajones@ventanamicro.com>,
	 Anup Patel <apatel@ventanamicro.com>,
	kvm@vger.kernel.org,  kvm-riscv@lists.infradead.org,
	linux-riscv@lists.infradead.org,  linux-kernel@vger.kernel.org
Subject: Re: [PATCH] RISC-V: KVM: Teardown riscv specific bits after kvm_exit
Date: Wed, 26 Mar 2025 18:55:55 -0700	[thread overview]
Message-ID: <Z-SwK1xO_S4phG2o@google.com> (raw)
In-Reply-To: <CAHBxVyFLeZFwEnJYa-oUbAKVimdVsr=Ct76Jf=TyWeoAkHe8yQ@mail.gmail.com>

On Thu, Mar 20, 2025, Atish Kumar Patra wrote:
> On Mon, Mar 17, 2025 at 9:08 AM Sean Christopherson <seanjc@google.com> wrote:
> > On Mon, Mar 17, 2025, Atish Patra wrote:
> > >  arch/riscv/kvm/main.c | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/arch/riscv/kvm/main.c b/arch/riscv/kvm/main.c
> > > index 1fa8be5ee509..4b24705dc63a 100644
> > > --- a/arch/riscv/kvm/main.c
> > > +++ b/arch/riscv/kvm/main.c
> > > @@ -172,8 +172,8 @@ module_init(riscv_kvm_init);
> > >
> > >  static void __exit riscv_kvm_exit(void)
> > >  {
> > > -     kvm_riscv_teardown();
> > > -
> > >       kvm_exit();
> > > +
> > > +     kvm_riscv_teardown();
> >
> > I wonder if there's a way we can guard against kvm_init()/kvm_exit() being called
> > too early/late.  x86 had similar bugs for a very long time, e.g. see commit
> > e32b120071ea ("KVM: VMX: Do _all_ initialization before exposing /dev/kvm to userspace").
> >
> > E.g. maybe we do something like create+destroy a VM at the end of kvm_init() and
> > the beginning of kvm_exit()?  Not sure if that would work for kvm_exit(), but it
> > should definitely be fine for kvm_init().
> >
> Yes. That would be super useful. I am not sure about the exact
> mechanism to achieve that though.

Me either :-)

> Do you just test code guarded within a new config that just
> creates/destroys a dummy VM ?

That's only idea I could come up with too, but I don't particulary like it.

> May be kunit test for KVM fits here in some way ?

I don't think a kunit test would be a good fit, there are likely too many
dependencies, and I'm pretty sure we'd still need to hack KVM.

-- 
kvm-riscv mailing list
kvm-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kvm-riscv