[xen master] x86: fix GS-base-dirty determination

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

[xen master] x86: fix GS-base-dirty determination

commit 91f85280b9b80852352fcad73d94ed29fafb88da
Author:     Jan Beulich <[hidden email]>
AuthorDate: Tue Oct 24 18:12:31 2017 +0200
Commit:     Jan Beulich <[hidden email]>
CommitDate: Tue Oct 24 18:12:31 2017 +0200

    x86: fix GS-base-dirty determination
    load_segments() writes the two MSRs in their "canonical" positions
    (GS_BASE for the user base, SHADOW_GS_BASE for the kernel one) and uses
    SWAPGS to switch them around if the incoming vCPU is in kernel mode. In
    order to not leave a stale kernel address in GS_BASE when the incoming
    guest is in user mode, the check on the outgoing vCPU needs to be
    dependent upon the mode it is currently in, rather than blindly looking
    at the user base.
    Signed-off-by: Jan Beulich <[hidden email]>
    Reviewed-by: Andrew Cooper <[hidden email]>
    Release-acked-by: Julien Grall <[hidden email]>
 xen/arch/x86/domain.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
index c28ac38..735f45c 100644
--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -1479,7 +1479,8 @@ static void save_segments(struct vcpu *v)
         if ( regs->gs & ~3 )
             v->arch.pv_vcpu.gs_base_user = 0;
-    if ( v->arch.pv_vcpu.gs_base_user )
+    if ( v->arch.flags & TF_kernel_mode ? v->arch.pv_vcpu.gs_base_kernel
+                                        : v->arch.pv_vcpu.gs_base_user )
         dirty_segment_mask |= DIRTY_GS_BASE_USER;
     this_cpu(dirty_segment_mask) = dirty_segment_mask;
generated by git-patchbot for /home/xen/git/xen.git#master

Xen-changelog mailing list
[hidden email]