* [PATCH 0/2 v2] KVM: nVMX: Check GUEST_DR7 on vmentry of nested guests @ 2020-01-15 1:25 Krish Sadhukhan 2020-01-15 1:25 ` [PATCH 1/2 " Krish Sadhukhan 2020-01-15 1:25 ` [PATCH 2/2 v2] kvm-unit-test: nVMX: Test " Krish Sadhukhan 0 siblings, 2 replies; 4+ messages in thread From: Krish Sadhukhan @ 2020-01-15 1:25 UTC (permalink / raw) To: kvm; +Cc: pbonzini, jmattson I sent out v1 a few months back. Here are the changes: v1 -> v2: 1. Patch# 1 has been dropped as we do not want to check GUEST_DEBUGCTL software. 2. Patch# 3 has been dropped. 3. Patch# 4 has been modified to include only the tests for GUEST_DR7 and to be run directly on hardware as well. [PATCH 1/2 v2] KVM: nVMX: Check GUEST_DR7 on vmentry of nested guests [PATCH 2/2 v2] kvm-unit-test: nVMX: Test GUEST_DR7 on vmentry of nested guests arch/x86/kvm/vmx/nested.c | 6 ++++++ arch/x86/kvm/x86.c | 2 +- arch/x86/kvm/x86.h | 6 ++++++ 3 files changed, 13 insertions(+), 1 deletion(-) Krish Sadhukhan (1): nVMX: Check GUEST_DR7 on vmentry of nested guests x86/vmx_tests.c | 44 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 44 insertions(+) Krish Sadhukhan (1): nVMX: Test GUEST_DR7 on vmentry of nested guests ^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH 1/2 v2] KVM: nVMX: Check GUEST_DR7 on vmentry of nested guests 2020-01-15 1:25 [PATCH 0/2 v2] KVM: nVMX: Check GUEST_DR7 on vmentry of nested guests Krish Sadhukhan @ 2020-01-15 1:25 ` Krish Sadhukhan 2020-01-15 17:13 ` Sean Christopherson 2020-01-15 1:25 ` [PATCH 2/2 v2] kvm-unit-test: nVMX: Test " Krish Sadhukhan 1 sibling, 1 reply; 4+ messages in thread From: Krish Sadhukhan @ 2020-01-15 1:25 UTC (permalink / raw) To: kvm; +Cc: pbonzini, jmattson According to section "Checks on Guest Control Registers, Debug Registers, and and MSRs" in Intel SDM vol 3C, the following checks are performed on vmentry of nested guests: If the "load debug controls" VM-entry control is 1, bits 63:32 in the DR7 field must be 0. Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com> Reviewed-by: Karl Heubaum <karl.heubaum@oracle.com> --- arch/x86/kvm/vmx/nested.c | 6 ++++++ arch/x86/kvm/x86.c | 2 +- arch/x86/kvm/x86.h | 6 ++++++ 3 files changed, 13 insertions(+), 1 deletion(-) diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c index 4aea7d304beb..acde8a2f13e2 100644 --- a/arch/x86/kvm/vmx/nested.c +++ b/arch/x86/kvm/vmx/nested.c @@ -2899,6 +2899,12 @@ static int nested_vmx_check_guest_state(struct kvm_vcpu *vcpu, CC(!nested_guest_cr4_valid(vcpu, vmcs12->guest_cr4))) return -EINVAL; +#ifdef CONFIG_X86_64 + if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS) && + !kvm_dr7_valid(vmcs12->guest_dr7)) + return -EINVAL; +#endif + if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PAT) && CC(!kvm_pat_valid(vmcs12->guest_ia32_pat))) return -EINVAL; diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index cf917139de6b..220f20a2f9c3 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -1064,7 +1064,7 @@ static int __kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val) case 5: /* fall through */ default: /* 7 */ - if (val & 0xffffffff00000000ULL) + if (!kvm_dr7_valid(val)) return -1; /* #GP */ vcpu->arch.dr7 = (val & DR7_VOLATILE) | DR7_FIXED_1; kvm_update_dr7(vcpu); diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h index 29391af8871d..76cd389ecf60 100644 --- a/arch/x86/kvm/x86.h +++ b/arch/x86/kvm/x86.h @@ -369,6 +369,12 @@ static inline bool kvm_pat_valid(u64 data) return (data | ((data & 0x0202020202020202ull) << 1)) == data; } +static inline bool kvm_dr7_valid(u64 data) +{ + /* Bits [63:32] are reserved */ + return ((data & 0xFFFFFFFF00000000ull) ? false : true); +} + void kvm_load_guest_xsave_state(struct kvm_vcpu *vcpu); void kvm_load_host_xsave_state(struct kvm_vcpu *vcpu); -- 2.20.1 ^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH 1/2 v2] KVM: nVMX: Check GUEST_DR7 on vmentry of nested guests 2020-01-15 1:25 ` [PATCH 1/2 " Krish Sadhukhan @ 2020-01-15 17:13 ` Sean Christopherson 0 siblings, 0 replies; 4+ messages in thread From: Sean Christopherson @ 2020-01-15 17:13 UTC (permalink / raw) To: Krish Sadhukhan; +Cc: kvm, pbonzini, jmattson On Tue, Jan 14, 2020 at 08:25:40PM -0500, Krish Sadhukhan wrote: > According to section "Checks on Guest Control Registers, Debug Registers, and > and MSRs" in Intel SDM vol 3C, the following checks are performed on vmentry > of nested guests: > > If the "load debug controls" VM-entry control is 1, bits 63:32 in the DR7 > field must be 0. Please explain *why* the check is being added to KVM. Quoting the SDM is very helpful in proving the correctness of the code, but it doesn't provide any insight into why a guest field is being checked in software. A tweaked version of Jim's anaylsis from v1[*] would be perfect. https://lkml.kernel.org/r/CALMp9eR2GQ_aerH-arOEpa08k8ZdtYCA5ftxHfDCo5fS1r3VtA@mail.gmail.com > Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com> > Reviewed-by: Karl Heubaum <karl.heubaum@oracle.com> > --- > arch/x86/kvm/vmx/nested.c | 6 ++++++ > arch/x86/kvm/x86.c | 2 +- > arch/x86/kvm/x86.h | 6 ++++++ > 3 files changed, 13 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c > index 4aea7d304beb..acde8a2f13e2 100644 > --- a/arch/x86/kvm/vmx/nested.c > +++ b/arch/x86/kvm/vmx/nested.c > @@ -2899,6 +2899,12 @@ static int nested_vmx_check_guest_state(struct kvm_vcpu *vcpu, > CC(!nested_guest_cr4_valid(vcpu, vmcs12->guest_cr4))) > return -EINVAL; > > +#ifdef CONFIG_X86_64 Hmm, I'd prefer not to wrap this with CONFIG_X86_64. From an architectural perspective, the consistency check is performed if the CPU *supports* long mode, irrespective of whether the CPU is actually in long mode. KVM could technically do something like static_cpu_has(X86_FEATURE_LM), but that's a waste of code for everyone except the 0.00000000000001% of the population running on Yonah, and nested 32-bit on 64-bit already fudges things with respect to 64-bit CPU behavior. Functionally, it'll be the same end result (and possibly a waste of cycles on 32-bit KVM if the compiler doesn't optimize out kvm_dr7_valid()) as having the CONFIG_X86_64 since kvm_dr7_valid() will always return true on 32-bit KVM (assuming @data is changed to an unsigned long). Architecturally, 32-bit KVM on 64-bit harware is already in a grey area, e.g. hardware VM-Entry still performs checks like GUEST_DR7[63:32]!=0, they just can't fail on 32-bit KVM because KVM's VMWRITE to propgate vmcs12->guest_dr7 to vmcs02.GUEST_DR7 will drop bits 63:32. In other words, it's not an issue of functionality, I'd just prefer to keep keep the constency checks themselves aligned with the SDM. > + if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS) && > + !kvm_dr7_valid(vmcs12->guest_dr7)) Wrap !kvm_dr7_valid() with CC() so that it's traced. > + return -EINVAL; > +#endif > + > if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PAT) && > CC(!kvm_pat_valid(vmcs12->guest_ia32_pat))) > return -EINVAL; > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > index cf917139de6b..220f20a2f9c3 100644 > --- a/arch/x86/kvm/x86.c > +++ b/arch/x86/kvm/x86.c > @@ -1064,7 +1064,7 @@ static int __kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val) > case 5: > /* fall through */ > default: /* 7 */ > - if (val & 0xffffffff00000000ULL) > + if (!kvm_dr7_valid(val)) > return -1; /* #GP */ > vcpu->arch.dr7 = (val & DR7_VOLATILE) | DR7_FIXED_1; > kvm_update_dr7(vcpu); > diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h > index 29391af8871d..76cd389ecf60 100644 > --- a/arch/x86/kvm/x86.h > +++ b/arch/x86/kvm/x86.h > @@ -369,6 +369,12 @@ static inline bool kvm_pat_valid(u64 data) > return (data | ((data & 0x0202020202020202ull) << 1)) == data; > } > > +static inline bool kvm_dr7_valid(u64 data) Per Jim's feedback on v1, @data should be "unsigned long". > +{ > + /* Bits [63:32] are reserved */ > + return ((data & 0xFFFFFFFF00000000ull) ? false : true); Per Jim's feedback in v1, the ternary operator and second set of parantheses are unnecessary. return !(data & 0xFFFFFFFF00000000ull); or return data == (u32)data; or return !(data >> 32); I prefer the last one because (IMO) it's easier to visually parse than the "& 0xFF..." variant and more explicit in what it's doing than the casting variant. > +} > + > void kvm_load_guest_xsave_state(struct kvm_vcpu *vcpu); > void kvm_load_host_xsave_state(struct kvm_vcpu *vcpu); > > -- > 2.20.1 > ^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH 2/2 v2] kvm-unit-test: nVMX: Test GUEST_DR7 on vmentry of nested guests 2020-01-15 1:25 [PATCH 0/2 v2] KVM: nVMX: Check GUEST_DR7 on vmentry of nested guests Krish Sadhukhan 2020-01-15 1:25 ` [PATCH 1/2 " Krish Sadhukhan @ 2020-01-15 1:25 ` Krish Sadhukhan 1 sibling, 0 replies; 4+ messages in thread From: Krish Sadhukhan @ 2020-01-15 1:25 UTC (permalink / raw) To: kvm; +Cc: pbonzini, jmattson According to section "Checks on Guest Control Registers, Debug Registers, and and MSRs" in Intel SDM vol 3C, the following checks are performed on vmentry of nested guests: If the "load debug controls" VM-entry control is 1, - bits 63:32 in the DR7 field must be 0. Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com> Reviewed-by: Karl Heubaum <karl.heubaum@oracle.com> Co-developed-by: Jim Mattson <jmattson@google.com> Signed-off-by: Jim Mattson <jmattson@google.com> --- x86/vmx_tests.c | 44 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 44 insertions(+) diff --git a/x86/vmx_tests.c b/x86/vmx_tests.c index fce773c..b773872 100644 --- a/x86/vmx_tests.c +++ b/x86/vmx_tests.c @@ -7442,6 +7442,49 @@ static void vmx_host_state_area_test(void) test_load_host_perf_global_ctrl(); } +/* + * If the "load debug controls" VM-entry control is 1, bits 63:32 in + * the DR7 field must be 0. + * + * [Intel SDM] + */ +static void test_guest_dr7(void) +{ + u32 ent_saved = vmcs_read(ENT_CONTROLS); + u64 dr7_saved = vmcs_read(GUEST_DR7); + u64 val; + int i; + + if (ctrl_enter_rev.set & ENT_LOAD_DBGCTLS) { + vmcs_clear_bits(ENT_CONTROLS, ENT_LOAD_DBGCTLS); + for (i = 0; i < 64; i++) { + val = 1ull << i; + vmcs_write(GUEST_DR7, val); + enter_guest(); + report_guest_state_test("ENT_LOAD_DBGCTLS disabled", + VMX_VMCALL, val, "GUEST_DR7"); + } + } + if (ctrl_enter_rev.clr & ENT_LOAD_DBGCTLS) { + vmcs_set_bits(ENT_CONTROLS, ENT_LOAD_DBGCTLS); + for (i = 0; i < 64; i++) { + val = 1ull << i; + vmcs_write(GUEST_DR7, val); + if (i < 32) + enter_guest(); + else + enter_guest_with_invalid_guest_state(); + report_guest_state_test("ENT_LOAD_DBGCTLS enabled", + i < 32 ? VMX_VMCALL : + VMX_ENTRY_FAILURE | + VMX_FAIL_STATE, + val, "GUEST_DR7"); + } + } + vmcs_write(GUEST_DR7, dr7_saved); + vmcs_write(ENT_CONTROLS, ent_saved); +} + /* * If the "load IA32_PAT" VM-entry control is 1, the value of the field * for the IA32_PAT MSR must be one that could be written by WRMSR @@ -7480,6 +7523,7 @@ static void vmx_guest_state_area_test(void) test_canonical(GUEST_SYSENTER_ESP, "GUEST_SYSENTER_ESP", false); test_canonical(GUEST_SYSENTER_EIP, "GUEST_SYSENTER_EIP", false); + test_guest_dr7(); test_load_guest_pat(); test_guest_efer(); test_load_guest_perf_global_ctrl(); -- 2.20.1 ^ permalink raw reply related [flat|nested] 4+ messages in thread
end of thread, other threads:[~2020-01-15 17:13 UTC | newest] Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2020-01-15 1:25 [PATCH 0/2 v2] KVM: nVMX: Check GUEST_DR7 on vmentry of nested guests Krish Sadhukhan 2020-01-15 1:25 ` [PATCH 1/2 " Krish Sadhukhan 2020-01-15 17:13 ` Sean Christopherson 2020-01-15 1:25 ` [PATCH 2/2 v2] kvm-unit-test: nVMX: Test " Krish Sadhukhan
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).