lkml.org 
[lkml]   [2021]   [Jul]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH 04/16] KVM: arm64: Add MMIO checking infrastructure
On Tuesday 20 Jul 2021 at 14:15:56 (+0100), Marc Zyngier wrote:
> On Tue, 20 Jul 2021 12:13:20 +0100,
> Quentin Perret <qperret@google.com> wrote:
> >
> > On Thursday 15 Jul 2021 at 17:31:47 (+0100), Marc Zyngier wrote:
> > > +struct s2_walk_data {
> > > + kvm_pte_t pteval;
> > > + u32 level;
> > > +};
> > > +
> > > +static int s2_walker(u64 addr, u64 end, u32 level, kvm_pte_t *ptep,
> > > + enum kvm_pgtable_walk_flags flag, void * const arg)
> > > +{
> > > + struct s2_walk_data *data = arg;
> > > +
> > > + data->level = level;
> > > + data->pteval = *ptep;
> > > + return 0;
> > > +}
> > > +
> > > +/* Assumes mmu_lock taken */
> > > +static bool __check_ioguard_page(struct kvm_vcpu *vcpu, gpa_t ipa)
> > > +{
> > > + struct s2_walk_data data;
> > > + struct kvm_pgtable_walker walker = {
> > > + .cb = s2_walker,
> > > + .flags = KVM_PGTABLE_WALK_LEAF,
> > > + .arg = &data,
> > > + };
> > > +
> > > + kvm_pgtable_walk(vcpu->arch.hw_mmu->pgt, ALIGN_DOWN(ipa, PAGE_SIZE),
> > > + PAGE_SIZE, &walker);
> > > +
> > > + /* Must be a PAGE_SIZE mapping with our annotation */
> > > + return (BIT(ARM64_HW_PGTABLE_LEVEL_SHIFT(data.level)) == PAGE_SIZE &&
> > > + data.pteval == MMIO_NOTE);
> >
> > Nit: you could do this check in the walker directly and check the return
> > value of kvm_pgtable_walk() instead. That would allow to get rid of
> > struct s2_walk_data.
> >
> > Also, though the compiler might be able to optimize, maybe simplify the
> > level check to level == (KVM_PGTABLE_MAX_LEVELS - 1)?
>
> Yup, all good points. I guess I could do the same in my other series
> that parses the userspace PT to extract the level.

Well, actually, let me take that back. I think something like you have
would be useful, but in pgtable.c directly and re-usable for stage-1 and
stage-2 walks. Maybe something like the below (totally untested)?

I could use such a walker in several places as well in the memory
ownership series:

- following the idea of [1], I could remove the
kvm_pgtable_stage2_find_range() function entirely;

- [2] defines 2 custom walkers that do nothing but walk host stage-2
and hyp stage-1 page-tables to check permissions and such -- they
could be removed/re-implemented easily as well.

And you seem to need something similar here, so clearly there is a need.
WDYT?

Thanks,
Quentin

[1] https://lore.kernel.org/kvmarm/20210719104735.3681732-3-qperret@google.com/
[2] https://lore.kernel.org/kvmarm/20210719104735.3681732-14-qperret@google.com/

diff --git a/arch/arm64/kvm/hyp/pgtable.c b/arch/arm64/kvm/hyp/pgtable.c
index e0ae57dca827..bd6d26f27e1a 100644
--- a/arch/arm64/kvm/hyp/pgtable.c
+++ b/arch/arm64/kvm/hyp/pgtable.c
@@ -357,6 +357,38 @@ int kvm_pgtable_walk(struct kvm_pgtable *pgt, u64 addr, u64 size,
return _kvm_pgtable_walk(&walk_data);
}

+struct get_leaf_data {
+ kvm_pte_t *ptep;
+ u32 *level;
+};
+
+static int get_leaf_walker(u64 addr, u64 end, u32 level, kvm_pte_t *ptep,
+ enum kvm_pgtable_walk_flags flag, void * const arg)
+{
+ struct get_leaf_data *data = arg;
+
+ *(data->ptep) = *ptep;
+ *(data->level) = level;
+
+ return 0;
+}
+
+int kvm_pgtable_get_leaf(struct kvm_pgtable *pgt, u64 addr, kvm_pte_t *ptep,
+ u32 *level)
+{
+ struct get_leaf_data data = {
+ .ptep = ptep,
+ .level = level,
+ };
+ struct kvm_pgtable_walker __get_leaf_walker = {
+ .cb = get_leaf_walker,
+ .flags = KVM_PGTABLE_WALK_LEAF,
+ .arg = &data,
+ };
+
+ return kvm_pgtable_walk(pgt, addr, PAGE_SIZE, &__get_leaf_walker);
+}
+
struct hyp_map_data {
u64 phys;
kvm_pte_t attr;
\
 
 \ /
  Last update: 2021-07-20 17:56    [W:0.133 / U:0.552 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site