2023-07-22 01:30:17

by Sean Christopherson

Subject: [PATCH 3/5] KVM: x86/mmu: Harden TDP MMU iteration against root w/o shadow page

Explicitly check that tdp_iter_start() is handed a valid shadow page
to harden KVM against bugs where

Opportunistically stop the TDP MMU iteration instead of continuing on
with garbage if the incoming root is bogus. Attempting to walk a garbage
root is more likely to caused major problems than doing nothing.

Signed-off-by: Sean Christopherson <[email protected]>
---
arch/x86/kvm/mmu/tdp_iter.c | 11 ++++++-----
1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
index d2eb0d4f8710..bd30ebfb2f2c 100644
--- a/arch/x86/kvm/mmu/tdp_iter.c
+++ b/arch/x86/kvm/mmu/tdp_iter.c
@@ -39,13 +39,14 @@ void tdp_iter_restart(struct tdp_iter *iter)
void tdp_iter_start(struct tdp_iter *iter, struct kvm_mmu_page *root,
int min_level, gfn_t next_last_level_gfn)
{
- int root_level = root->role.level;
-
- WARN_ON(root_level < 1);
- WARN_ON(root_level > PT64_ROOT_MAX_LEVEL);
+ if (WARN_ON_ONCE(!root || (root->role.level < 1) ||
+ (root->role.level > PT64_ROOT_MAX_LEVEL))) {
+ iter->valid = false;
+ return;
+ }

iter->next_last_level_gfn = next_last_level_gfn;
- iter->root_level = root_level;
+ iter->root_level = root->role.level;
iter->min_level = min_level;
iter->pt_path[iter->root_level - 1] = (tdp_ptep_t)root->spt;
iter->as_id = kvm_mmu_page_as_id(root);
--
2.41.0.487.g6d72f3e995-goog

2023-07-25 10:53:12

by Yu Zhang

Subject: Re: [PATCH 3/5] KVM: x86/mmu: Harden TDP MMU iteration against root w/o shadow page

On Fri, Jul 21, 2023 at 06:23:48PM -0700, Sean Christopherson wrote:
> Explicitly check that tdp_iter_start() is handed a valid shadow page
> to harden KVM against bugs where

Sorry, where?

It's not about guest using an invisible GFN, it's about a KVM bug, right?

>
> Opportunistically stop the TDP MMU iteration instead of continuing on
> with garbage if the incoming root is bogus. Attempting to walk a garbage
> root is more likely to caused major problems than doing nothing.
>
> Signed-off-by: Sean Christopherson <[email protected]>
> ---
> arch/x86/kvm/mmu/tdp_iter.c | 11 ++++++-----
> 1 file changed, 6 insertions(+), 5 deletions(-)
>
> diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
> index d2eb0d4f8710..bd30ebfb2f2c 100644
> --- a/arch/x86/kvm/mmu/tdp_iter.c
> +++ b/arch/x86/kvm/mmu/tdp_iter.c
> @@ -39,13 +39,14 @@ void tdp_iter_restart(struct tdp_iter *iter)
> void tdp_iter_start(struct tdp_iter *iter, struct kvm_mmu_page *root,
> int min_level, gfn_t next_last_level_gfn)
> {
> - int root_level = root->role.level;
> -
> - WARN_ON(root_level < 1);
> - WARN_ON(root_level > PT64_ROOT_MAX_LEVEL);
> + if (WARN_ON_ONCE(!root || (root->role.level < 1) ||
> + (root->role.level > PT64_ROOT_MAX_LEVEL))) {
> + iter->valid = false;
> + return;
> + }
>

I saw many usages of WARN_ON_ONCE() and WARN_ON() in KVM. And just wonder,
is there any criteria for KVM when to use which?

B.R.
Yu

2023-07-25 16:20:50

by Sean Christopherson

Subject: Re: [PATCH 3/5] KVM: x86/mmu: Harden TDP MMU iteration against root w/o shadow page

On Tue, Jul 25, 2023, Yu Zhang wrote:
> On Fri, Jul 21, 2023 at 06:23:48PM -0700, Sean Christopherson wrote:
> > Explicitly check that tdp_iter_start() is handed a valid shadow page
> > to harden KVM against bugs where
>
> Sorry, where?

Gah, I must have seen something shiny when writing the changelog.

> It's not about guest using an invisible GFN, it's about a KVM bug, right?

Yes, the intent is to guard against a KVM bug, e.g. if KVM managed to get into
the TDP MMU with an invalid root, or a root belonging to a shadow MMU. I'll fix
the changelog in v2.

> > Opportunistically stop the TDP MMU iteration instead of continuing on
> > with garbage if the incoming root is bogus. Attempting to walk a garbage
> > root is more likely to caused major problems than doing nothing.
> >
> > Signed-off-by: Sean Christopherson <[email protected]>
> > ---
> > arch/x86/kvm/mmu/tdp_iter.c | 11 ++++++-----
> > 1 file changed, 6 insertions(+), 5 deletions(-)
> >
> > diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
> > index d2eb0d4f8710..bd30ebfb2f2c 100644
> > --- a/arch/x86/kvm/mmu/tdp_iter.c
> > +++ b/arch/x86/kvm/mmu/tdp_iter.c
> > @@ -39,13 +39,14 @@ void tdp_iter_restart(struct tdp_iter *iter)
> > void tdp_iter_start(struct tdp_iter *iter, struct kvm_mmu_page *root,
> > int min_level, gfn_t next_last_level_gfn)
> > {
> > - int root_level = root->role.level;
> > -
> > - WARN_ON(root_level < 1);
> > - WARN_ON(root_level > PT64_ROOT_MAX_LEVEL);
> > + if (WARN_ON_ONCE(!root || (root->role.level < 1) ||
> > + (root->role.level > PT64_ROOT_MAX_LEVEL))) {
> > + iter->valid = false;
> > + return;
> > + }
> >
>
> I saw many usages of WARN_ON_ONCE() and WARN_ON() in KVM. And just wonder,
> is there any criteria for KVM when to use which?

Heh, already a step ahead of you :-)

https://lore.kernel.org/all/[email protected]

2023-07-26 04:09:04

by Yu Zhang

Subject: Re: [PATCH 3/5] KVM: x86/mmu: Harden TDP MMU iteration against root w/o shadow page

On Tue, Jul 25, 2023 at 08:56:32AM -0700, Sean Christopherson wrote:
> On Tue, Jul 25, 2023, Yu Zhang wrote:
> > On Fri, Jul 21, 2023 at 06:23:48PM -0700, Sean Christopherson wrote:
> > > Explicitly check that tdp_iter_start() is handed a valid shadow page
> > > to harden KVM against bugs where
> >
> > Sorry, where?
>
> Gah, I must have seen something shiny when writing the changelog.
>
> > It's not about guest using an invisible GFN, it's about a KVM bug, right?
>
> Yes, the intent is to guard against a KVM bug, e.g. if KVM managed to get into
> the TDP MMU with an invalid root, or a root belonging to a shadow MMU. I'll fix
> the changelog in v2.
>
> > > Opportunistically stop the TDP MMU iteration instead of continuing on
> > > with garbage if the incoming root is bogus. Attempting to walk a garbage
> > > root is more likely to caused major problems than doing nothing.
> > >
> > > Signed-off-by: Sean Christopherson <[email protected]>
> > > ---
> > > arch/x86/kvm/mmu/tdp_iter.c | 11 ++++++-----
> > > 1 file changed, 6 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
> > > index d2eb0d4f8710..bd30ebfb2f2c 100644
> > > --- a/arch/x86/kvm/mmu/tdp_iter.c
> > > +++ b/arch/x86/kvm/mmu/tdp_iter.c
> > > @@ -39,13 +39,14 @@ void tdp_iter_restart(struct tdp_iter *iter)
> > > void tdp_iter_start(struct tdp_iter *iter, struct kvm_mmu_page *root,
> > > int min_level, gfn_t next_last_level_gfn)
> > > {
> > > - int root_level = root->role.level;
> > > -
> > > - WARN_ON(root_level < 1);
> > > - WARN_ON(root_level > PT64_ROOT_MAX_LEVEL);
> > > + if (WARN_ON_ONCE(!root || (root->role.level < 1) ||
> > > + (root->role.level > PT64_ROOT_MAX_LEVEL))) {
> > > + iter->valid = false;
> > > + return;
> > > + }
> > >
> >
> > I saw many usages of WARN_ON_ONCE() and WARN_ON() in KVM. And just wonder,
> > is there any criteria for KVM when to use which?
>
> Heh, already a step ahead of you :-)
>
> https://lore.kernel.org/all/[email protected]

Haha! That patch lies just above this series, and the explanation is very
convincing. :) Thanks!

B.R.
Yu