| Message ID | 151632015123.21271.9060883997507739532.stgit@dwillia2-desk3.amr.corp.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 19/01/2018 01:02, Dan Williams wrote: > Commit 75f139aaf896 "KVM: x86: Add memory barrier on vmcs field lookup" > added a raw 'asm("lfence");' to prevent a bounds check bypass of > 'vmcs_field_to_offset_table'. This does not work for some AMD cpus, see > the 'ifence' helper, The code never runs on AMD cpus (it's for Intel virtualization extensions), so it'd be nice if you could fix up the commit message. Apart from this, obviously Acked-by: Paolo Bonzini <pbonzini@redhat.com> Thanks! Paolo > and it otherwise does not use the common > 'array_ptr' helper designed for these types of fixes. Convert this to > use 'array_ptr'. > > Cc: Andrew Honig <ahonig@google.com> > Cc: Jim Mattson <jmattson@google.com> > Cc: Paolo Bonzini <pbonzini@redhat.com> > Cc: <stable@vger.kernel.org> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> > --- > arch/x86/kvm/vmx.c | 19 +++++++------------ > 1 file changed, 7 insertions(+), 12 deletions(-) > > diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c > index c829d89e2e63..20b9b0b5e336 100644 > --- a/arch/x86/kvm/vmx.c > +++ b/arch/x86/kvm/vmx.c > @@ -34,6 +34,7 @@ > #include <linux/tboot.h> > #include <linux/hrtimer.h> > #include <linux/frame.h> > +#include <linux/nospec.h> > #include "kvm_cache_regs.h" > #include "x86.h" > > @@ -898,21 +899,15 @@ static const unsigned short vmcs_field_to_offset_table[] = { > > static inline short vmcs_field_to_offset(unsigned long field) > { > - BUILD_BUG_ON(ARRAY_SIZE(vmcs_field_to_offset_table) > SHRT_MAX); > - > - if (field >= ARRAY_SIZE(vmcs_field_to_offset_table)) > - return -ENOENT; > + const unsigned short *offset; > > - /* > - * FIXME: Mitigation for CVE-2017-5753. To be replaced with a > - * generic mechanism. > - */ > - asm("lfence"); > + BUILD_BUG_ON(ARRAY_SIZE(vmcs_field_to_offset_table) > SHRT_MAX); > > - if (vmcs_field_to_offset_table[field] == 0) > + offset = array_ptr(vmcs_field_to_offset_table, field, > + ARRAY_SIZE(vmcs_field_to_offset_table)); > + if (!offset || *offset == 0) > return -ENOENT; > - > - return vmcs_field_to_offset_table[field]; > + return *offset; > } > > static inline struct vmcs12 *get_vmcs12(struct kvm_vcpu *vcpu) >
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index c829d89e2e63..20b9b0b5e336 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -34,6 +34,7 @@ #include <linux/tboot.h> #include <linux/hrtimer.h> #include <linux/frame.h> +#include <linux/nospec.h> #include "kvm_cache_regs.h" #include "x86.h" @@ -898,21 +899,15 @@ static const unsigned short vmcs_field_to_offset_table[] = { static inline short vmcs_field_to_offset(unsigned long field) { - BUILD_BUG_ON(ARRAY_SIZE(vmcs_field_to_offset_table) > SHRT_MAX); - - if (field >= ARRAY_SIZE(vmcs_field_to_offset_table)) - return -ENOENT; + const unsigned short *offset; - /* - * FIXME: Mitigation for CVE-2017-5753. To be replaced with a - * generic mechanism. - */ - asm("lfence"); + BUILD_BUG_ON(ARRAY_SIZE(vmcs_field_to_offset_table) > SHRT_MAX); - if (vmcs_field_to_offset_table[field] == 0) + offset = array_ptr(vmcs_field_to_offset_table, field, + ARRAY_SIZE(vmcs_field_to_offset_table)); + if (!offset || *offset == 0) return -ENOENT; - - return vmcs_field_to_offset_table[field]; + return *offset; } static inline struct vmcs12 *get_vmcs12(struct kvm_vcpu *vcpu)
Commit 75f139aaf896 "KVM: x86: Add memory barrier on vmcs field lookup" added a raw 'asm("lfence");' to prevent a bounds check bypass of 'vmcs_field_to_offset_table'. This does not work for some AMD cpus, see the 'ifence' helper, and it otherwise does not use the common 'array_ptr' helper designed for these types of fixes. Convert this to use 'array_ptr'. Cc: Andrew Honig <ahonig@google.com> Cc: Jim Mattson <jmattson@google.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: <stable@vger.kernel.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com> --- arch/x86/kvm/vmx.c | 19 +++++++------------ 1 file changed, 7 insertions(+), 12 deletions(-)