[v3,07/13] nEPT: Fix wrong test in kvm_set_cr3

Commit Message

Nakajima, Jun May 19, 2013, 4:52 a.m. UTC

From: Nadav Har'El <[e-mail]>

kvm_set_cr3() attempts to check if the new cr3 is a valid guest physical
address. The problem is that with nested EPT, cr3 is an *L2* physical
address, not an L1 physical address as this test expects.

As the comment above this test explains, it isn't necessary, and doesn't
correspond to anything a real processor would do. So this patch removes it.

Note that this wrong test could have also theoretically caused problems
in nested NPT, not just in nested EPT. However, in practice, the problem
was avoided: nested_svm_vmexit()/vmrun() do not call kvm_set_cr3 in the
nested NPT case, and instead set the vmcb (and arch.cr3) directly, thus
circumventing the problem. Additional potential calls to the buggy function
are avoided in that we don't trap cr3 modifications when nested NPT is
enabled. However, because in nested VMX we did want to use kvm_set_cr3()
(as requested in Avi Kivity's review of the original nested VMX patches),
we can't avoid this problem and need to fix it.

Signed-off-by: Nadav Har'El <[e-mail]>
Signed-off-by: Jun Nakajima <[e-mail]>
Signed-off-by: Xinhao Xu <[e-mail]>
---
 arch/x86/kvm/x86.c | 11 -----------
 1 file changed, 11 deletions(-)

Comments

Paolo Bonzini May 20, 2013, 1:17 p.m. UTC | #1

Il 19/05/2013 06:52, Jun Nakajima ha scritto:
> From: Nadav Har'El <[e-mail]>
> 
> kvm_set_cr3() attempts to check if the new cr3 is a valid guest physical
> address. The problem is that with nested EPT, cr3 is an *L2* physical
> address, not an L1 physical address as this test expects.
> 
> As the comment above this test explains, it isn't necessary, and doesn't
> correspond to anything a real processor would do. So this patch removes it.
> 
> Note that this wrong test could have also theoretically caused problems
> in nested NPT, not just in nested EPT. However, in practice, the problem
> was avoided: nested_svm_vmexit()/vmrun() do not call kvm_set_cr3 in the
> nested NPT case, and instead set the vmcb (and arch.cr3) directly, thus
> circumventing the problem. Additional potential calls to the buggy function
> are avoided in that we don't trap cr3 modifications when nested NPT is
> enabled. However, because in nested VMX we did want to use kvm_set_cr3()
> (as requested in Avi Kivity's review of the original nested VMX patches),
> we can't avoid this problem and need to fix it

Makes sense, but did you test what happens (without nesting, and both
with/without EPT) if L1 points CR3 at an invalid physical address?  Does
a basic level of sanity remain?

Paolo
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 094b5d9..7b36ec6 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -683,17 +683,6 @@  int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
 		 */
 	}
 
-	/*
-	 * Does the new cr3 value map to physical memory? (Note, we
-	 * catch an invalid cr3 even in real-mode, because it would
-	 * cause trouble later on when we turn on paging anyway.)
-	 *
-	 * A real CPU would silently accept an invalid cr3 and would
-	 * attempt to use it - with largely undefined (and often hard
-	 * to debug) behavior on the guest side.
-	 */
-	if (unlikely(!gfn_to_memslot(vcpu->kvm, cr3 >> PAGE_SHIFT)))
-		return 1;
 	vcpu->arch.cr3 = cr3;
 	__set_bit(VCPU_EXREG_CR3, (ulong *)&vcpu->arch.regs_avail);
 	vcpu->arch.mmu.new_cr3(vcpu);