From: Sean Christopherson <seanjc@google.com>
To: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Cc: kvm@vger.kernel.org, pbonzini@redhat.com, jmattson@google.com
Subject: Re: [PATCH 1/3 v2] nSVM: Check addresses of MSR and IO bitmap
Date: Tue, 19 Jan 2021 16:45:36 -0800 [thread overview]
Message-ID: <YAd9MBkpDjC1MY6E@google.com> (raw)
In-Reply-To: <20210116022039.7316-2-krish.sadhukhan@oracle.com>
On Sat, Jan 16, 2021, Krish Sadhukhan wrote:
> According to section "Canonicalization and Consistency Checks" in APM vol 2,
> the following guest state is illegal:
>
> "The MSR or IOIO intercept tables extend to a physical address that
> is greater than or equal to the maximum supported physical address."
>
> Also check that these addresses are aligned on page boundary.
>
> Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
> ---
> arch/x86/kvm/svm/nested.c | 18 ++++++++++++------
> 1 file changed, 12 insertions(+), 6 deletions(-)
>
> diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
> index cb4c6ee10029..2419f392a13d 100644
> --- a/arch/x86/kvm/svm/nested.c
> +++ b/arch/x86/kvm/svm/nested.c
> @@ -211,7 +211,8 @@ static bool svm_get_nested_state_pages(struct kvm_vcpu *vcpu)
> return true;
> }
>
> -static bool nested_vmcb_check_controls(struct vmcb_control_area *control)
> +static bool nested_vmcb_check_controls(struct kvm_vcpu *vcpu,
> + struct vmcb_control_area *control)
> {
> if ((vmcb_is_intercept(control, INTERCEPT_VMRUN)) == 0)
> return false;
> @@ -223,10 +224,15 @@ static bool nested_vmcb_check_controls(struct vmcb_control_area *control)
> !npt_enabled)
> return false;
>
> + if (!page_address_valid(vcpu, control->msrpm_base_pa))
> + return false;
> + if (!page_address_valid(vcpu, control->iopm_base_pa))
These checks are wrong. The MSRPM is 8kb in size, and the IOPM is 12kb, and the
APM explicitly states that the last byte is checked:
if the address of the last byte in the table is greater than or equal to the
maximum supported physical address, this is treated as illegal VMCB state and
causes a #VMEXIT(VMEXIT_INVALID).
KVM can't check just the last byte, as that would fail to detect a wrap of the
64-bit boundary. Might be worth adding yet another helper? I think this will
work, though I'm sure Paolo has a much more clever solution :-)
static inline bool page_range_valid(struct kvm_vcpu *vcpu, gpa_t gpa, int size)
{
gpa_t last_page = gpa + size - PAGE_SIZE;
if (last_page < gpa)
return false;
return page_address_valid(last_page);
}
Note, the IOPM is 12kb in size, but KVM allocates and initializes 16kb, i.e.
using IOPM_ALLOC_ORDER for the check would be wrong. Maybe define the actual
size for both bitmaps and use get_order() instead of hardcoding the order? That
would make it easy to "fix" svm_hardware_setup() so that it doesn't initialize
unused memory.
> + return false;
> +
> return true;
> }
>
> -static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12)
> +static bool nested_vmcb_checks(struct kvm_vcpu *vcpu, struct vmcb *vmcb12)
> {
> bool vmcb12_lma;
>
> @@ -255,10 +261,10 @@ static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12)
> (vmcb12->save.cr3 & MSR_CR3_LONG_MBZ_MASK))
> return false;
> }
> - if (!kvm_is_valid_cr4(&svm->vcpu, vmcb12->save.cr4))
> + if (!kvm_is_valid_cr4(vcpu, vmcb12->save.cr4))
> return false;
>
> - return nested_vmcb_check_controls(&vmcb12->control);
> + return nested_vmcb_check_controls(vcpu, &vmcb12->control);
> }
>
> static void load_nested_vmcb_control(struct vcpu_svm *svm,
> @@ -485,7 +491,7 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
> if (WARN_ON_ONCE(!svm->nested.initialized))
> return -EINVAL;
>
> - if (!nested_vmcb_checks(svm, vmcb12)) {
> + if (!nested_vmcb_checks(&svm->vcpu, vmcb12)) {
> vmcb12->control.exit_code = SVM_EXIT_ERR;
> vmcb12->control.exit_code_hi = 0;
> vmcb12->control.exit_info_1 = 0;
> @@ -1173,7 +1179,7 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,
> goto out_free;
>
> ret = -EINVAL;
> - if (!nested_vmcb_check_controls(ctl))
> + if (!nested_vmcb_check_controls(vcpu, ctl))
> goto out_free;
>
> /*
> --
> 2.27.0
>
next prev parent reply other threads:[~2021-01-20 0:46 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-16 2:20 [PATCH 0/3 v2] KVM: nSVM: Check addresses of MSR bitmap and IO bitmap tables on vmrun of nested guests Krish Sadhukhan
2021-01-16 2:20 ` [PATCH 1/3 v2] nSVM: Check addresses of MSR and IO bitmap Krish Sadhukhan
2021-01-20 0:45 ` Sean Christopherson [this message]
2021-01-21 0:36 ` Krish Sadhukhan
2021-01-16 2:20 ` [PATCH 2/3 v2] Test: nSVM: Test MSR and IO bitmap address Krish Sadhukhan
2021-01-16 2:20 ` [PATCH 3/3 v2] Test: SVM: Use ALIGN macro when aligning 'io_bitmap_area' Krish Sadhukhan
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=YAd9MBkpDjC1MY6E@google.com \
--to=seanjc@google.com \
--cc=jmattson@google.com \
--cc=krish.sadhukhan@oracle.com \
--cc=kvm@vger.kernel.org \
--cc=pbonzini@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).