All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks
@ 2016-10-14 10:06 Andrew Cooper
  2016-10-14 10:06 ` [PATCH 2/2] x86/hvm: Clobber %cs.L when LME becomes set Andrew Cooper
                   ` (2 more replies)
  0 siblings, 3 replies; 6+ messages in thread
From: Andrew Cooper @ 2016-10-14 10:06 UTC (permalink / raw)
  To: Xen-devel; +Cc: Andrew Cooper, Jan Beulich

Contrary to the description in the software manuals, in Long Mode, attempts to
load %cs check that D is not set in combination with L before the present flag
is checked.

This can be observed because the L/D check fails with #GP before the presence
check failes with #NP

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
---
CC: Jan Beulich <JBeulich@suse.com>
---
 xen/arch/x86/x86_emulate/x86_emulate.c | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/xen/arch/x86/x86_emulate/x86_emulate.c b/xen/arch/x86/x86_emulate/x86_emulate.c
index 793ce30..b23cd99 100644
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -1405,6 +1405,14 @@ protmode_load_seg(
                /* Non-conforming segment: check RPL and DPL against CPL. */
                : rpl > cpl || dpl != cpl )
             goto raise_exn;
+        /*
+         * 64-bit code segments (L bit set) must have D bit clear.
+         * Experimentally in long mode, the L and D bits are checked before
+         * the Present bit.
+         */
+        if ( in_longmode(ctxt, ops) &&
+             (desc.b & (1 << 21)) && (desc.b & (1 << 22)) )
+            goto raise_exn;
         sel = (sel ^ rpl) | cpl;
         break;
     case x86_seg_ss:
@@ -1444,11 +1452,6 @@ protmode_load_seg(
         goto raise_exn;
     }
 
-    /* 64-bit code segments (L bit set) must have D bit clear. */
-    if ( seg == x86_seg_cs && in_longmode(ctxt, ops) &&
-         (desc.b & (1 << 21)) && (desc.b & (1 << 22)) )
-        goto raise_exn;
-
     /* Ensure Accessed flag is set. */
     if ( a_flag && !(desc.b & a_flag) )
     {
-- 
2.1.4


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

^ permalink raw reply related	[flat|nested] 6+ messages in thread
* [PATCH 2/2] x86/hvm: Clobber %cs.L when LME becomes set
  2016-10-14 10:06 [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks Andrew Cooper
@ 2016-10-14 10:06 ` Andrew Cooper
  2016-10-14 10:29   ` Jan Beulich
  2016-10-14 10:28 ` [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks Jan Beulich
  2016-10-14 14:10 ` Konrad Rzeszutek Wilk
  2 siblings, 1 reply; 6+ messages in thread
From: Andrew Cooper @ 2016-10-14 10:06 UTC (permalink / raw)
  To: Xen-devel; +Cc: Andrew Cooper, Jan Beulich

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
---
CC: Jan Beulich <JBeulich@suse.com>
---
 xen/arch/x86/hvm/hvm.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index ceb89c7..3c90ecd 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -2037,6 +2037,30 @@ int hvm_set_efer(uint64_t value)
         return X86EMUL_EXCEPTION;
     }
 
+    if ( (value & EFER_LME) && !(v->arch.hvm_vcpu.guest_efer & EFER_LME) )
+    {
+        struct segment_register cs;
+
+        hvm_get_segment_register(v, x86_seg_cs, &cs);
+
+        /*
+         * %cs may be loaded with both .D and .L set in legacy mode, and both
+         * are captured in the VMCS/VMCB.
+         *
+         * If a guest does this and then tries to transition into long mode,
+         * the vmentry from setting LME fails due to invalid guest state,
+         * because %cr0.PG is still clear.
+         *
+         * When LME becomes set, clobber %cs.L to keep the guest firmly in
+         * compatibility mode until it reloads %cs itself.
+         */
+        if ( cs.attr.fields.l )
+        {
+            cs.attr.fields.l = 0;
+            hvm_set_segment_register(v, x86_seg_cs, &cs);
+        }
+    }
+
     if ( nestedhvm_enabled(v->domain) && cpu_has_svm &&
        ((value & EFER_SVME) == 0 ) &&
        ((value ^ v->arch.hvm_vcpu.guest_efer) & EFER_SVME) )
-- 
2.1.4


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

^ permalink raw reply related	[flat|nested] 6+ messages in thread
* Re: [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks
  2016-10-14 10:06 [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks Andrew Cooper
  2016-10-14 10:06 ` [PATCH 2/2] x86/hvm: Clobber %cs.L when LME becomes set Andrew Cooper
@ 2016-10-14 10:28 ` Jan Beulich
  2016-10-14 10:32   ` Andrew Cooper
  2016-10-14 14:10 ` Konrad Rzeszutek Wilk
  2 siblings, 1 reply; 6+ messages in thread
From: Jan Beulich @ 2016-10-14 10:28 UTC (permalink / raw)
  To: Andrew Cooper; +Cc: Xen-devel

>>> On 14.10.16 at 12:06, <andrew.cooper3@citrix.com> wrote:
> Contrary to the description in the software manuals, in Long Mode, attempts to
> load %cs check that D is not set in combination with L before the present flag
> is checked.
> 
> This can be observed because the L/D check fails with #GP before the presence
> check failes with #NP
> 
> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>

Reviewed-by: Jan Beulich <jbeulich@suse.com>

But I think it would be worthwhile mentioning that this restores
things to v1 of what became commit 78ff18c90.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: [PATCH 2/2] x86/hvm: Clobber %cs.L when LME becomes set
  2016-10-14 10:06 ` [PATCH 2/2] x86/hvm: Clobber %cs.L when LME becomes set Andrew Cooper
@ 2016-10-14 10:29   ` Jan Beulich
  0 siblings, 0 replies; 6+ messages in thread
From: Jan Beulich @ 2016-10-14 10:29 UTC (permalink / raw)
  To: Andrew Cooper; +Cc: Xen-devel

>>> On 14.10.16 at 12:06, <andrew.cooper3@citrix.com> wrote:
> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>

Reviewed-by: Jan Beulich <jbeulich@suse.com>


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks
  2016-10-14 10:28 ` [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks Jan Beulich
@ 2016-10-14 10:32   ` Andrew Cooper
  0 siblings, 0 replies; 6+ messages in thread
From: Andrew Cooper @ 2016-10-14 10:32 UTC (permalink / raw)
  To: Jan Beulich; +Cc: Xen-devel

On 14/10/16 11:28, Jan Beulich wrote:
>>>> On 14.10.16 at 12:06, <andrew.cooper3@citrix.com> wrote:
>> Contrary to the description in the software manuals, in Long Mode, attempts to
>> load %cs check that D is not set in combination with L before the present flag
>> is checked.
>>
>> This can be observed because the L/D check fails with #GP before the presence
>> check failes with #NP
>>
>> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
> Reviewed-by: Jan Beulich <jbeulich@suse.com>
>
> But I think it would be worthwhile mentioning that this restores
> things to v1 of what became commit 78ff18c90.

I will update the comment to include this.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

^ permalink raw reply	[flat|nested] 6+ messages in thread
* Re: [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks
  2016-10-14 10:06 [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks Andrew Cooper
  2016-10-14 10:06 ` [PATCH 2/2] x86/hvm: Clobber %cs.L when LME becomes set Andrew Cooper
  2016-10-14 10:28 ` [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks Jan Beulich
@ 2016-10-14 14:10 ` Konrad Rzeszutek Wilk
  2 siblings, 0 replies; 6+ messages in thread
From: Konrad Rzeszutek Wilk @ 2016-10-14 14:10 UTC (permalink / raw)
  To: Andrew Cooper, paul.c.lai, sherry.hurwitz; +Cc: Jan Beulich, Xen-devel

On Fri, Oct 14, 2016 at 11:06:55AM +0100, Andrew Cooper wrote:
> Contrary to the description in the software manuals, in Long Mode, attempts to
> load %cs check that D is not set in combination with L before the present flag
> is checked.
> 
> This can be observed because the L/D check fails with #GP before the presence
> check failes with #NP

CC-ing Paul and Sherry.

Perhaps the SDMs should mention this as well?

> 
> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
> ---
> CC: Jan Beulich <JBeulich@suse.com>
> ---
>  xen/arch/x86/x86_emulate/x86_emulate.c | 13 ++++++++-----
>  1 file changed, 8 insertions(+), 5 deletions(-)
> 
> diff --git a/xen/arch/x86/x86_emulate/x86_emulate.c b/xen/arch/x86/x86_emulate/x86_emulate.c
> index 793ce30..b23cd99 100644
> --- a/xen/arch/x86/x86_emulate/x86_emulate.c
> +++ b/xen/arch/x86/x86_emulate/x86_emulate.c
> @@ -1405,6 +1405,14 @@ protmode_load_seg(
>                 /* Non-conforming segment: check RPL and DPL against CPL. */
>                 : rpl > cpl || dpl != cpl )
>              goto raise_exn;
> +        /*
> +         * 64-bit code segments (L bit set) must have D bit clear.
> +         * Experimentally in long mode, the L and D bits are checked before
> +         * the Present bit.
> +         */
> +        if ( in_longmode(ctxt, ops) &&
> +             (desc.b & (1 << 21)) && (desc.b & (1 << 22)) )
> +            goto raise_exn;
>          sel = (sel ^ rpl) | cpl;
>          break;
>      case x86_seg_ss:
> @@ -1444,11 +1452,6 @@ protmode_load_seg(
>          goto raise_exn;
>      }
>  
> -    /* 64-bit code segments (L bit set) must have D bit clear. */
> -    if ( seg == x86_seg_cs && in_longmode(ctxt, ops) &&
> -         (desc.b & (1 << 21)) && (desc.b & (1 << 22)) )
> -        goto raise_exn;
> -
>      /* Ensure Accessed flag is set. */
>      if ( a_flag && !(desc.b & a_flag) )
>      {
> -- 
> 2.1.4
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> https://lists.xen.org/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

^ permalink raw reply	[flat|nested] 6+ messages in thread
end of thread, other threads:[~2016-10-14 14:10 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-10-14 10:06 [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks Andrew Cooper
2016-10-14 10:06 ` [PATCH 2/2] x86/hvm: Clobber %cs.L when LME becomes set Andrew Cooper
2016-10-14 10:29   ` Jan Beulich
2016-10-14 10:28 ` [PATCH 1/2] x86/hvm: Correct the position of the %cs L/D checks Jan Beulich
2016-10-14 10:32   ` Andrew Cooper
2016-10-14 14:10 ` Konrad Rzeszutek Wilk

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.