Re: [PATCH 08/14] KVM: arm64: Protect page table traversal with RCU

From: Oliver Upton <oliver.upton@linux.dev>
To: David Matlack <dmatlack@google.com>
Cc: Marc Zyngier <maz@kernel.org>, James Morse <james.morse@arm.com>,
	Alexandru Elisei <alexandru.elisei@arm.com>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will@kernel.org>,
	Quentin Perret <qperret@google.com>,
	Ricardo Koller <ricarkol@google.com>,
	Reiji Watanabe <reijiw@google.com>,
	Ben Gardon <bgardon@google.com>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Gavin Shan <gshan@redhat.com>, Peter Xu <peterx@redhat.com>,
	Sean Christopherson <seanjc@google.com>,
	linux-arm-kernel@lists.infradead.org,
	kvmarm@lists.cs.columbia.edu, kvm@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 08/14] KVM: arm64: Protect page table traversal with RCU
Date: Fri, 9 Sep 2022 10:55:59 +0100	[thread overview]
Message-ID: <YxsNr+79UUm5Go9x@google.com> (raw)
In-Reply-To: <YxkRXLsLuhjBNanT@google.com>

On Wed, Sep 07, 2022 at 02:47:08PM -0700, David Matlack wrote:
> On Tue, Aug 30, 2022 at 07:41:26PM +0000, Oliver Upton wrote:
> > The use of RCU is necessary to change the paging structures in parallel.
> > Acquire and release an RCU read lock when traversing the page tables.
> > 
> > Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
> > ---
> >  arch/arm64/include/asm/kvm_pgtable.h | 19 ++++++++++++++++++-
> >  arch/arm64/kvm/hyp/pgtable.c         |  7 ++++++-
> >  2 files changed, 24 insertions(+), 2 deletions(-)
> > 
> > diff --git a/arch/arm64/include/asm/kvm_pgtable.h b/arch/arm64/include/asm/kvm_pgtable.h
> > index 78fbb7be1af6..7d2de0a98ccb 100644
> > --- a/arch/arm64/include/asm/kvm_pgtable.h
> > +++ b/arch/arm64/include/asm/kvm_pgtable.h
> > @@ -578,9 +578,26 @@ enum kvm_pgtable_prot kvm_pgtable_stage2_pte_prot(kvm_pte_t pte);
> >   */
> >  enum kvm_pgtable_prot kvm_pgtable_hyp_pte_prot(kvm_pte_t pte);
> >  
> > +#if defined(__KVM_NVHE_HYPERVISOR___)
> > +
> 
> Future readers will wonder why NVHE stubs out RCU support and how that
> is even correct. Some comments here would be useful explain it.

Good point.

> > +static inline void kvm_pgtable_walk_begin(void) {}
> > +static inline void kvm_pgtable_walk_end(void) {}
> > +
> > +#define kvm_dereference_ptep rcu_dereference_raw
> 
> How does NVHE have access rcu_dereference_raw()?

rcu_dereference_raw() is inlined and simply recasts the pointer into the
kernel address space.

Perhaps it is less confusing to template this on kvm_pte_read() to avoid
polluting nVHE with an otherwise benign reference to RCU.

> > +
> > +#else	/* !defined(__KVM_NVHE_HYPERVISOR__) */
> > +
> > +#define kvm_pgtable_walk_begin	rcu_read_lock
> > +#define kvm_pgtable_walk_end	rcu_read_unlock
> > +#define kvm_dereference_ptep	rcu_dereference
> > +
> > +#endif	/* defined(__KVM_NVHE_HYPERVISOR__) */
> > +
> >  static inline kvm_pte_t kvm_pte_read(kvm_pte_t *ptep)
> >  {
> > -	return READ_ONCE(*ptep);
> > +	kvm_pte_t __rcu *p = (kvm_pte_t __rcu *)ptep;
> > +
> > +	return READ_ONCE(*kvm_dereference_ptep(p));
> 
> What about all the other places where page table memory is accessed?
> 
> If RCU is going to be used to protect page table memory, then all
> accesses have to go under an RCU critical section. This means that page
> table memory should only be accessed through __rcu annotated pointers
> and dereferenced with rcu_dereference().

Let me play around with this a bit, as the annoying part is trying to
sprinkle in RCU annotations w/o messing with nVHE. 

--
Thanks,
Oliver

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel