| Message ID | 1504262665-28092-1-git-send-email-aisaila@bitdefender.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 01/09/17 11:44, Alexandru Isaila wrote: > diff --git a/xen/include/public/arch-x86/hvm/save.h b/xen/include/public/arch-x86/hvm/save.h > index fd7bf3f..e6e8e87 100644 > --- a/xen/include/public/arch-x86/hvm/save.h > +++ b/xen/include/public/arch-x86/hvm/save.h > @@ -134,6 +134,8 @@ struct hvm_hw_cpu { > /* msr for em64t */ > uint64_t shadow_gs; > > + uint64_t shadow_gs_base; > + You presumably haven't tried migrating across this boundary? (Things will explode rather impressively when you try to restore LSTAR into SMASK.) What's wrong with the shadow_gs in context here? ~Andrew > /* msr content saved/restored. */ > uint64_t msr_flags; /* Obsolete, ignored. */ > uint64_t msr_lstar;
>>> On 01.09.17 at 12:44, <aisaila@bitdefender.com> wrote: > This patch is adding an new param in the hvm_hw_cpu structure > so it can be exposed to user space. But you don't say why. > --- a/xen/include/public/arch-x86/hvm/save.h > +++ b/xen/include/public/arch-x86/hvm/save.h > @@ -134,6 +134,8 @@ struct hvm_hw_cpu { > /* msr for em64t */ > uint64_t shadow_gs; > > + uint64_t shadow_gs_base; You can't alter a public interface structure like this. The one here is used for migration, so migrating guests from an older hypervisor would break. But what's more funny - you're adding a field next to what looks like already has the value you're after. Jan
On 01.09.2017 13:55, Andrew Cooper wrote: > On 01/09/17 11:44, Alexandru Isaila wrote: >> diff --git a/xen/include/public/arch-x86/hvm/save.h b/xen/include/public/arch-x86/hvm/save.h >> index fd7bf3f..e6e8e87 100644 >> --- a/xen/include/public/arch-x86/hvm/save.h >> +++ b/xen/include/public/arch-x86/hvm/save.h >> @@ -134,6 +134,8 @@ struct hvm_hw_cpu { >> /* msr for em64t */ >> uint64_t shadow_gs; >> >> + uint64_t shadow_gs_base; >> + > > You presumably haven't tried migrating across this boundary? (Things > will explode rather impressively when you try to restore LSTAR into SMASK.) > > What's wrong with the shadow_gs in context here? Nothing, we've missed that, while grepping xen/ for MSR_SHADOW_GS_BASE and 0xc0000102. My fault, should have checked more thoroughly. Apologies. Thanks, Razvan
diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c index 6cb903d..519333c 100644 --- a/xen/arch/x86/hvm/hvm.c +++ b/xen/arch/x86/hvm/hvm.c @@ -834,6 +834,8 @@ static int hvm_save_cpu_ctxt(struct domain *d, hvm_domain_context_t *h) ctxt.gs_base = seg.base; ctxt.gs_arbytes = seg.attr; + rdmsrl(MSR_SHADOW_GS_BASE, ctxt.shadow_gs_base); + hvm_get_segment_register(v, x86_seg_tr, &seg); ctxt.tr_sel = seg.sel; ctxt.tr_limit = seg.limit; @@ -1090,6 +1092,8 @@ static int hvm_load_cpu_ctxt(struct domain *d, hvm_domain_context_t *h) seg.attr = ctxt.gs_arbytes; hvm_set_segment_register(v, x86_seg_gs, &seg); + wrmsrl(MSR_SHADOW_GS_BASE, ctxt.shadow_gs_base); + seg.sel = ctxt.tr_sel; seg.limit = ctxt.tr_limit; seg.base = ctxt.tr_base; diff --git a/xen/include/public/arch-x86/hvm/save.h b/xen/include/public/arch-x86/hvm/save.h index fd7bf3f..e6e8e87 100644 --- a/xen/include/public/arch-x86/hvm/save.h +++ b/xen/include/public/arch-x86/hvm/save.h @@ -134,6 +134,8 @@ struct hvm_hw_cpu { /* msr for em64t */ uint64_t shadow_gs; + uint64_t shadow_gs_base; + /* msr content saved/restored. */ uint64_t msr_flags; /* Obsolete, ignored. */ uint64_t msr_lstar;
This patch is adding an new param in the hvm_hw_cpu structure so it can be exposed to user space. Signed-off-by: Alexandru Isaila <aisaila@bitdefender.com> --- xen/arch/x86/hvm/hvm.c | 4 ++++ xen/include/public/arch-x86/hvm/save.h | 2 ++ 2 files changed, 6 insertions(+)