WARNING - OLD ARCHIVES

This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-devel

[Xen-devel] [PATCH] x86: fix EFER handling

To: <xen-devel@xxxxxxxxxxxxxxxxxxx>
Subject: [Xen-devel] [PATCH] x86: fix EFER handling
From: "Jan Beulich" <jbeulich@xxxxxxxxxx>
Date: Mon, 14 May 2007 17:36:07 +0200
Delivery-date: Mon, 14 May 2007 08:33:43 -0700
Envelope-to: www-data@xxxxxxxxxxxxxxxxxx
List-help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-id: Xen developer discussion <xen-devel.lists.xensource.com>
List-post: <mailto:xen-devel@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
Sender: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
Introduce a per-CPU shadow of what is currently in EFER, as context switch
code must re-write this MSR so that all guests run with appropriate EFER.SCE
and EFER.NX settings. Remove EFER from the set of MSRs that VMX deals with
in a generic fashion.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxxxx>

Index: 2007-05-14/xen/arch/x86/domain.c
===================================================================
--- 2007-05-14.orig/xen/arch/x86/domain.c       2007-05-14 14:27:23.000000000 
+0200
+++ 2007-05-14/xen/arch/x86/domain.c    2007-05-14 14:31:12.000000000 +0200
@@ -47,6 +47,7 @@
 #endif
 
 DEFINE_PER_CPU(struct vcpu *, curr_vcpu);
+DEFINE_PER_CPU(__u64, efer);
 
 static void paravirt_ctxt_switch_from(struct vcpu *v);
 static void paravirt_ctxt_switch_to(struct vcpu *v);
@@ -1138,21 +1139,18 @@ void context_switch(struct vcpu *prev, s
         __context_switch();
 
 #ifdef CONFIG_COMPAT
-        if ( is_idle_vcpu(prev) ||
-             (is_pv_32on64_domain(prev->domain) !=
-              is_pv_32on64_domain(next->domain)) )
+        if ( !is_hvm_vcpu(next) &&
+             (is_idle_vcpu(prev) ||
+              is_hvm_vcpu(prev) ||
+              is_pv_32on64_vcpu(prev) != is_pv_32on64_vcpu(next)) )
         {
-            uint32_t efer_lo, efer_hi;
+            uint64_t efer = read_efer();
 
             local_flush_tlb_one(GDT_VIRT_START(next) +
                                 FIRST_RESERVED_GDT_BYTE);
 
-            rdmsr(MSR_EFER, efer_lo, efer_hi);
-            if ( !is_pv_32on64_domain(next->domain) == !(efer_lo & EFER_SCE) )
-            {
-                efer_lo ^= EFER_SCE;
-                wrmsr(MSR_EFER, efer_lo, efer_hi);
-            }
+            if ( !is_pv_32on64_vcpu(next) == !(efer & EFER_SCE) )
+                write_efer(efer ^ EFER_SCE);
         }
 #endif
 
Index: 2007-05-14/xen/arch/x86/hvm/svm/svm.c
===================================================================
--- 2007-05-14.orig/xen/arch/x86/hvm/svm/svm.c  2007-05-14 13:47:25.000000000 
+0200
+++ 2007-05-14/xen/arch/x86/hvm/svm/svm.c       2007-05-14 14:28:19.000000000 
+0200
@@ -96,11 +96,8 @@ static inline void svm_inject_exception(
 
 static void stop_svm(void)
 {
-    u32 eax, edx;    
     /* We turn off the EFER_SVME bit. */
-    rdmsr(MSR_EFER, eax, edx);
-    eax &= ~EFER_SVME;
-    wrmsr(MSR_EFER, eax, edx);
+    write_efer(read_efer() & ~EFER_SVME);
 }
 
 static void svm_store_cpu_guest_regs(
@@ -145,7 +142,13 @@ static inline int long_mode_do_msr_write
     {
     case MSR_EFER:
         /* Offending reserved bit will cause #GP. */
-        if ( msr_content & ~(EFER_LME | EFER_LMA | EFER_NX | EFER_SCE) )
+#ifdef __x86_64__
+        if ( (msr_content & ~(EFER_LME | EFER_LMA | EFER_NX | EFER_SCE)) ||
+#else
+        if ( (msr_content & ~(EFER_NX | EFER_SCE)) ||
+#endif
+             (!cpu_has_nx && (msr_content & EFER_NX)) ||
+             (!cpu_has_syscall && (msr_content & EFER_SCE)) )
         {
             gdprintk(XENLOG_WARNING, "Trying to set reserved bit in "
                      "EFER: %"PRIx64"\n", msr_content);
@@ -502,7 +505,7 @@ int svm_vmcb_restore(struct vcpu *v, str
 }
 
         
-void svm_save_cpu_state(struct vcpu *v, struct hvm_hw_cpu *data)
+static void svm_save_cpu_state(struct vcpu *v, struct hvm_hw_cpu *data)
 {
     struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
 
@@ -518,7 +521,7 @@ void svm_save_cpu_state(struct vcpu *v, 
 }
 
 
-void svm_load_cpu_state(struct vcpu *v, struct hvm_hw_cpu *data)
+static void svm_load_cpu_state(struct vcpu *v, struct hvm_hw_cpu *data)
 {
     struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
 
@@ -537,13 +540,13 @@ void svm_load_cpu_state(struct vcpu *v, 
     hvm_set_guest_time(v, data->tsc);
 }
 
-void svm_save_vmcb_ctxt(struct vcpu *v, struct hvm_hw_cpu *ctxt)
+static void svm_save_vmcb_ctxt(struct vcpu *v, struct hvm_hw_cpu *ctxt)
 {
     svm_save_cpu_state(v, ctxt);
     svm_vmcb_save(v, ctxt);
 }
 
-int svm_load_vmcb_ctxt(struct vcpu *v, struct hvm_hw_cpu *ctxt)
+static int svm_load_vmcb_ctxt(struct vcpu *v, struct hvm_hw_cpu *ctxt)
 {
     svm_load_cpu_state(v, ctxt);
     if (svm_vmcb_restore(v, ctxt)) {
@@ -871,6 +874,7 @@ static struct hvm_function_table svm_fun
     .paging_enabled       = svm_paging_enabled,
     .long_mode_enabled    = svm_long_mode_enabled,
     .pae_enabled          = svm_pae_enabled,
+    .nx_enabled           = svm_nx_enabled,
     .interrupts_enabled   = svm_interrupts_enabled,
     .guest_x86_mode       = svm_guest_x86_mode,
     .get_guest_ctrl_reg   = svm_get_ctrl_reg,
@@ -927,9 +931,7 @@ int start_svm(void)
          ((root_vmcb[cpu] = alloc_vmcb()) == NULL) )
         return 0;
 
-    rdmsr(MSR_EFER, eax, edx);
-    eax |= EFER_SVME;
-    wrmsr(MSR_EFER, eax, edx);
+    write_efer(read_efer() | EFER_SVME);
 
     svm_npt_detect();
 
Index: 2007-05-14/xen/arch/x86/hvm/vmx/vmcs.c
===================================================================
--- 2007-05-14.orig/xen/arch/x86/hvm/vmx/vmcs.c 2007-05-03 09:45:09.000000000 
+0200
+++ 2007-05-14/xen/arch/x86/hvm/vmx/vmcs.c      2007-05-14 14:28:19.000000000 
+0200
@@ -285,11 +285,6 @@ static void construct_vmcs(struct vcpu *
 
     vmx_vmcs_enter(v);
 
-    v->arch.hvm_vmx.cpu_cr2 = 0;
-    v->arch.hvm_vmx.cpu_cr3 = 0;
-    memset(&v->arch.hvm_vmx.msr_state, 0, sizeof(v->arch.hvm_vmx.msr_state));
-    v->arch.hvm_vmx.vmxassist_enabled = 0;
-
     /* VMCS controls. */
     __vmwrite(PIN_BASED_VM_EXEC_CONTROL, vmx_pin_based_exec_control);
     __vmwrite(VM_EXIT_CONTROLS, vmx_vmexit_control);
Index: 2007-05-14/xen/arch/x86/hvm/vmx/vmx.c
===================================================================
--- 2007-05-14.orig/xen/arch/x86/hvm/vmx/vmx.c  2007-05-14 13:47:25.000000000 
+0200
+++ 2007-05-14/xen/arch/x86/hvm/vmx/vmx.c       2007-05-14 14:28:19.000000000 
+0200
@@ -89,7 +89,7 @@ static DEFINE_PER_CPU(struct vmx_msr_sta
 static u32 msr_index[VMX_MSR_COUNT] =
 {
     MSR_LSTAR, MSR_STAR, MSR_CSTAR,
-    MSR_SYSCALL_MASK, MSR_EFER,
+    MSR_SYSCALL_MASK
 };
 
 static void vmx_save_host_msrs(void)
@@ -117,8 +117,7 @@ static inline int long_mode_do_msr_read(
 
     switch ( (u32)regs->ecx ) {
     case MSR_EFER:
-        HVM_DBG_LOG(DBG_LEVEL_2, "EFER msr_content 0x%"PRIx64, msr_content);
-        msr_content = guest_msr_state->msrs[VMX_INDEX_MSR_EFER];
+        msr_content = v->arch.hvm_vmx.efer;
         break;
 
     case MSR_FS_BASE:
@@ -130,7 +129,7 @@ static inline int long_mode_do_msr_read(
         goto check_long_mode;
 
     case MSR_SHADOW_GS_BASE:
-        msr_content = guest_msr_state->shadow_gs;
+        msr_content = v->arch.hvm_vmx.shadow_gs;
     check_long_mode:
         if ( !(vmx_long_mode_enabled(v)) )
         {
@@ -182,7 +181,9 @@ static inline int long_mode_do_msr_write
     {
     case MSR_EFER:
         /* offending reserved bit will cause #GP */
-        if ( msr_content & ~(EFER_LME | EFER_LMA | EFER_NX | EFER_SCE) )
+        if ( (msr_content & ~(EFER_LME | EFER_LMA | EFER_NX | EFER_SCE)) ||
+             (!cpu_has_nx && (msr_content & EFER_NX)) ||
+             (!cpu_has_syscall && (msr_content & EFER_SCE)) )
         {
             gdprintk(XENLOG_WARNING, "Trying to set reserved bit in "
                      "EFER: %"PRIx64"\n", msr_content);
@@ -190,7 +191,7 @@ static inline int long_mode_do_msr_write
         }
 
         if ( (msr_content & EFER_LME)
-             &&  !(guest_msr_state->msrs[VMX_INDEX_MSR_EFER] & EFER_LME) )
+             &&  !(v->arch.hvm_vmx.efer & EFER_LME) )
         {
             if ( unlikely(vmx_paging_enabled(v)) )
             {
@@ -200,7 +201,7 @@ static inline int long_mode_do_msr_write
             }
         }
         else if ( !(msr_content & EFER_LME)
-                  && (guest_msr_state->msrs[VMX_INDEX_MSR_EFER] & EFER_LME) )
+                  && (v->arch.hvm_vmx.efer & EFER_LME) )
         {
             if ( unlikely(vmx_paging_enabled(v)) )
             {
@@ -210,7 +211,11 @@ static inline int long_mode_do_msr_write
             }
         }
 
-        guest_msr_state->msrs[VMX_INDEX_MSR_EFER] = msr_content;
+        if ( (msr_content ^ v->arch.hvm_vmx.efer) & (EFER_NX|EFER_SCE) )
+            write_efer((read_efer() & ~(EFER_NX|EFER_SCE)) |
+                       (msr_content & (EFER_NX|EFER_SCE)));
+
+        v->arch.hvm_vmx.efer = msr_content;
         break;
 
     case MSR_FS_BASE:
@@ -228,7 +233,7 @@ static inline int long_mode_do_msr_write
             __vmwrite(GUEST_GS_BASE, msr_content);
         else
         {
-            v->arch.hvm_vmx.msr_state.shadow_gs = msr_content;
+            v->arch.hvm_vmx.shadow_gs = msr_content;
             wrmsrl(MSR_SHADOW_GS_BASE, msr_content);
         }
 
@@ -280,12 +285,14 @@ static void vmx_restore_host_msrs(void)
         wrmsrl(msr_index[i], host_msr_state->msrs[i]);
         clear_bit(i, &host_msr_state->flags);
     }
+    if ( !(read_efer() & EFER_NX) )
+        write_efer(read_efer() | EFER_NX);
 }
 
 static void vmx_save_guest_msrs(struct vcpu *v)
 {
     /* MSR_SHADOW_GS_BASE may have been changed by swapgs instruction. */
-    rdmsrl(MSR_SHADOW_GS_BASE, v->arch.hvm_vmx.msr_state.shadow_gs);
+    rdmsrl(MSR_SHADOW_GS_BASE, v->arch.hvm_vmx.shadow_gs);
 }
 
 static void vmx_restore_guest_msrs(struct vcpu *v)
@@ -297,11 +304,9 @@ static void vmx_restore_guest_msrs(struc
     guest_msr_state = &v->arch.hvm_vmx.msr_state;
     host_msr_state = &this_cpu(host_msr_state);
 
-    wrmsrl(MSR_SHADOW_GS_BASE, guest_msr_state->shadow_gs);
+    wrmsrl(MSR_SHADOW_GS_BASE, v->arch.hvm_vmx.shadow_gs);
 
     guest_flags = guest_msr_state->flags;
-    if ( !guest_flags )
-        return;
 
     while ( guest_flags ) {
         i = find_first_set_bit(guest_flags);
@@ -313,23 +318,90 @@ static void vmx_restore_guest_msrs(struc
         wrmsrl(msr_index[i], guest_msr_state->msrs[i]);
         clear_bit(i, &guest_flags);
     }
+
+    if ( (v->arch.hvm_vmx.efer ^ read_efer()) & (EFER_NX|EFER_SCE) )
+    {
+        HVM_DBG_LOG(DBG_LEVEL_2,
+                    "restore guest's EFER with value %lx",
+                    v->arch.hvm_vmx.efer);
+        write_efer((read_efer() & ~(EFER_NX|EFER_SCE)) |
+                   (v->arch.hvm_vmx.efer & (EFER_NX|EFER_SCE)));
+    }
 }
 
 #else  /* __i386__ */
 
 #define vmx_save_host_msrs()        ((void)0)
-#define vmx_restore_host_msrs()     ((void)0)
+
+static void vmx_restore_host_msrs(void)
+{
+    if ( !(read_efer() & EFER_NX) )
+        write_efer(read_efer() | EFER_NX);
+}
+
 #define vmx_save_guest_msrs(v)      ((void)0)
-#define vmx_restore_guest_msrs(v)   ((void)0)
+
+static void vmx_restore_guest_msrs(struct vcpu *v)
+{
+    if ( (v->arch.hvm_vmx.efer ^ read_efer()) & EFER_NX )
+    {
+        HVM_DBG_LOG(DBG_LEVEL_2,
+                    "restore guest's EFER with value %lx",
+                    v->arch.hvm_vmx.efer);
+        write_efer((read_efer() & ~EFER_NX) |
+                   (v->arch.hvm_vmx.efer & EFER_NX));
+    }
+}
 
 static inline int long_mode_do_msr_read(struct cpu_user_regs *regs)
 {
-    return 0;
+    u64 msr_content = 0;
+    struct vcpu *v = current;
+
+    switch ( regs->ecx ) {
+    case MSR_EFER:
+        msr_content = v->arch.hvm_vmx.efer;
+        break;
+
+    default:
+        return 0;
+    }
+
+    regs->eax = msr_content >>  0;
+    regs->edx = msr_content >> 32;
+
+    return 1;
 }
 
 static inline int long_mode_do_msr_write(struct cpu_user_regs *regs)
 {
-    return 0;
+    u64 msr_content = regs->eax | ((u64)regs->edx << 32);
+    struct vcpu *v = current;
+
+    switch ( regs->ecx )
+    {
+    case MSR_EFER:
+        /* offending reserved bit will cause #GP */
+        if ( (msr_content & ~EFER_NX) ||
+             (!cpu_has_nx && (msr_content & EFER_NX)) )
+        {
+            gdprintk(XENLOG_WARNING, "Trying to set reserved bit in "
+                     "EFER: %"PRIx64"\n", msr_content);
+            vmx_inject_hw_exception(v, TRAP_gp_fault, 0);
+            return 0;
+        }
+
+        if ( (msr_content ^ v->arch.hvm_vmx.efer) & EFER_NX )
+            write_efer((read_efer() & ~EFER_NX) | (msr_content & EFER_NX));
+
+        v->arch.hvm_vmx.efer = msr_content;
+        break;
+
+    default:
+        return 0;
+    }
+
+    return 1;
 }
 
 #endif /* __i386__ */
@@ -637,7 +709,7 @@ int vmx_vmcs_restore(struct vcpu *v, str
     return -EINVAL;
 }
 
-#ifdef HVM_DEBUG_SUSPEND
+#if defined(__x86_64__) && defined(HVM_DEBUG_SUSPEND)
 static void dump_msr_state(struct vmx_msr_state *m)
 {
     int i = 0;
@@ -648,17 +720,16 @@ static void dump_msr_state(struct vmx_ms
     printk("\n");
 }
 #else
-static void dump_msr_state(struct vmx_msr_state *m)
-{
-}
+#define dump_msr_state(m) ((void)0)
 #endif
         
-void vmx_save_cpu_state(struct vcpu *v, struct hvm_hw_cpu *data)
+static void vmx_save_cpu_state(struct vcpu *v, struct hvm_hw_cpu *data)
 {
+#ifdef __x86_64__
     struct vmx_msr_state *guest_state = &v->arch.hvm_vmx.msr_state;
     unsigned long guest_flags = guest_state->flags;
 
-    data->shadow_gs = guest_state->shadow_gs;
+    data->shadow_gs = v->arch.hvm_vmx.shadow_gs;
 
     /* save msrs */
     data->msr_flags        = guest_flags;
@@ -666,15 +737,18 @@ void vmx_save_cpu_state(struct vcpu *v, 
     data->msr_star         = guest_state->msrs[VMX_INDEX_MSR_STAR];
     data->msr_cstar        = guest_state->msrs[VMX_INDEX_MSR_CSTAR];
     data->msr_syscall_mask = guest_state->msrs[VMX_INDEX_MSR_SYSCALL_MASK];
-    data->msr_efer         = guest_state->msrs[VMX_INDEX_MSR_EFER];
+#endif
+
+    data->msr_efer = v->arch.hvm_vmx.efer;
 
     data->tsc = hvm_get_guest_time(v);
     
     dump_msr_state(guest_state);
 }
 
-void vmx_load_cpu_state(struct vcpu *v, struct hvm_hw_cpu *data)
+static void vmx_load_cpu_state(struct vcpu *v, struct hvm_hw_cpu *data)
 {
+#ifdef __x86_64__
     struct vmx_msr_state *guest_state = &v->arch.hvm_vmx.msr_state;
 
     /* restore msrs */
@@ -683,9 +757,11 @@ void vmx_load_cpu_state(struct vcpu *v, 
     guest_state->msrs[VMX_INDEX_MSR_STAR]         = data->msr_star;
     guest_state->msrs[VMX_INDEX_MSR_CSTAR]        = data->msr_cstar;
     guest_state->msrs[VMX_INDEX_MSR_SYSCALL_MASK] = data->msr_syscall_mask;
-    guest_state->msrs[VMX_INDEX_MSR_EFER]         = data->msr_efer;
 
-    guest_state->shadow_gs = data->shadow_gs;
+    v->arch.hvm_vmx.shadow_gs = data->shadow_gs;
+#endif
+
+    v->arch.hvm_vmx.efer = data->msr_efer;
 
     v->arch.hvm_vmx.vmxassist_enabled = !(data->cr0 & X86_CR0_PE);
 
@@ -695,7 +771,7 @@ void vmx_load_cpu_state(struct vcpu *v, 
 }
 
 
-void vmx_save_vmcs_ctxt(struct vcpu *v, struct hvm_hw_cpu *ctxt)
+static void vmx_save_vmcs_ctxt(struct vcpu *v, struct hvm_hw_cpu *ctxt)
 {
     vmx_save_cpu_state(v, ctxt);
     vmx_vmcs_enter(v);
@@ -703,7 +779,7 @@ void vmx_save_vmcs_ctxt(struct vcpu *v, 
     vmx_vmcs_exit(v);
 }
 
-int vmx_load_vmcs_ctxt(struct vcpu *v, struct hvm_hw_cpu *ctxt)
+static int vmx_load_vmcs_ctxt(struct vcpu *v, struct hvm_hw_cpu *ctxt)
 {
     vmx_load_cpu_state(v, ctxt);
     if (vmx_vmcs_restore(v, ctxt)) {
@@ -1017,6 +1093,11 @@ static int vmx_pae_enabled(struct vcpu *
     return (vmx_paging_enabled(v) && (cr4 & X86_CR4_PAE));
 }
 
+static int vmx_nx_enabled(struct vcpu *v)
+{
+    return v->arch.hvm_vmx.efer & EFER_NX;
+}
+
 static int vmx_interrupts_enabled(struct vcpu *v) 
 {
     unsigned long eflags = __vmread(GUEST_RFLAGS); 
@@ -1097,6 +1178,7 @@ static struct hvm_function_table vmx_fun
     .paging_enabled       = vmx_paging_enabled,
     .long_mode_enabled    = vmx_long_mode_enabled,
     .pae_enabled          = vmx_pae_enabled,
+    .nx_enabled           = vmx_nx_enabled,
     .interrupts_enabled   = vmx_interrupts_enabled,
     .guest_x86_mode       = vmx_guest_x86_mode,
     .get_guest_ctrl_reg   = vmx_get_ctrl_reg,
@@ -1997,8 +2079,7 @@ static int vmx_set_cr0(unsigned long val
             else
             {
                 HVM_DBG_LOG(DBG_LEVEL_1, "Enabling long mode\n");
-                v->arch.hvm_vmx.msr_state.msrs[VMX_INDEX_MSR_EFER]
-                    |= EFER_LMA;
+                v->arch.hvm_vmx.efer |= EFER_LMA;
                 vm_entry_value = __vmread(VM_ENTRY_CONTROLS);
                 vm_entry_value |= VM_ENTRY_IA32E_MODE;
                 __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value);
@@ -2047,8 +2128,7 @@ static int vmx_set_cr0(unsigned long val
              */
             if ( vmx_long_mode_enabled(v) )
             {
-                v->arch.hvm_vmx.msr_state.msrs[VMX_INDEX_MSR_EFER]
-                    &= ~EFER_LMA;
+                v->arch.hvm_vmx.efer &= ~EFER_LMA;
                 vm_entry_value = __vmread(VM_ENTRY_CONTROLS);
                 vm_entry_value &= ~VM_ENTRY_IA32E_MODE;
                 __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value);
@@ -2080,7 +2160,7 @@ static int vmx_set_cr0(unsigned long val
     {
         if ( vmx_long_mode_enabled(v) )
         {
-            v->arch.hvm_vmx.msr_state.msrs[VMX_INDEX_MSR_EFER] &= ~EFER_LMA;
+            v->arch.hvm_vmx.efer &= ~EFER_LMA;
             vm_entry_value = __vmread(VM_ENTRY_CONTROLS);
             vm_entry_value &= ~VM_ENTRY_IA32E_MODE;
             __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value);
Index: 2007-05-14/xen/arch/x86/mm/shadow/multi.c
===================================================================
--- 2007-05-14.orig/xen/arch/x86/mm/shadow/multi.c      2007-05-03 
09:45:09.000000000 +0200
+++ 2007-05-14/xen/arch/x86/mm/shadow/multi.c   2007-05-14 14:28:19.000000000 
+0200
@@ -181,11 +181,11 @@ guest_supports_superpages(struct vcpu *v
 static inline int
 guest_supports_nx(struct vcpu *v)
 {
+    if ( GUEST_PAGING_LEVELS == 2 || !cpu_has_nx )
+        return 0;
     if ( !is_hvm_vcpu(v) )
-        return cpu_has_nx;
-
-    // XXX - fix this!
-    return 1;
+        return 1;
+    return hvm_nx_enabled(v);
 }
 
 
Index: 2007-05-14/xen/include/asm-x86/cpufeature.h
===================================================================
--- 2007-05-14.orig/xen/include/asm-x86/cpufeature.h    2007-04-23 
10:01:46.000000000 +0200
+++ 2007-05-14/xen/include/asm-x86/cpufeature.h 2007-05-14 14:28:19.000000000 
+0200
@@ -114,6 +114,7 @@
 #define cpu_has_xmm2           boot_cpu_has(X86_FEATURE_XMM2)
 #define cpu_has_xmm3           boot_cpu_has(X86_FEATURE_XMM3)
 #define cpu_has_ht             boot_cpu_has(X86_FEATURE_HT)
+#define cpu_has_syscall                boot_cpu_has(X86_FEATURE_SYSCALL)
 #define cpu_has_mp             boot_cpu_has(X86_FEATURE_MP)
 #define cpu_has_nx             boot_cpu_has(X86_FEATURE_NX)
 #define cpu_has_k6_mtrr                boot_cpu_has(X86_FEATURE_K6_MTRR)
@@ -136,6 +137,7 @@
 #define cpu_has_xmm2           1
 #define cpu_has_xmm3           boot_cpu_has(X86_FEATURE_XMM3)
 #define cpu_has_ht             boot_cpu_has(X86_FEATURE_HT)
+#define cpu_has_syscall                1
 #define cpu_has_mp             1
 #define cpu_has_nx             boot_cpu_has(X86_FEATURE_NX)
 #define cpu_has_k6_mtrr                0
Index: 2007-05-14/xen/include/asm-x86/hvm/hvm.h
===================================================================
--- 2007-05-14.orig/xen/include/asm-x86/hvm/hvm.h       2007-04-23 
10:01:46.000000000 +0200
+++ 2007-05-14/xen/include/asm-x86/hvm/hvm.h    2007-05-14 14:28:19.000000000 
+0200
@@ -93,14 +93,17 @@ struct hvm_function_table {
      * 1) determine whether paging is enabled,
      * 2) determine whether long mode is enabled,
      * 3) determine whether PAE paging is enabled,
-     * 4) determine whether interrupts are enabled or not,
-     * 5) determine the mode the guest is running in,
-     * 6) return the current guest control-register value
-     * 7) return the current guest segment descriptor base
+     * 4) determine whether NX is enabled,
+     * 5) determine whether interrupts are enabled or not,
+     * 6) determine the mode the guest is running in,
+     * 7) return the current guest control-register value
+     * 8) return the current guest segment descriptor base
+     * 9) return the current guest segment descriptor
      */
     int (*paging_enabled)(struct vcpu *v);
     int (*long_mode_enabled)(struct vcpu *v);
     int (*pae_enabled)(struct vcpu *v);
+    int (*nx_enabled)(struct vcpu *v);
     int (*interrupts_enabled)(struct vcpu *v);
     int (*guest_x86_mode)(struct vcpu *v);
     unsigned long (*get_guest_ctrl_reg)(struct vcpu *v, unsigned int num);
@@ -199,6 +202,12 @@ hvm_interrupts_enabled(struct vcpu *v)
 }
 
 static inline int
+hvm_nx_enabled(struct vcpu *v)
+{
+    return hvm_funcs.nx_enabled(v);
+}
+
+static inline int
 hvm_guest_x86_mode(struct vcpu *v)
 {
     return hvm_funcs.guest_x86_mode(v);
Index: 2007-05-14/xen/include/asm-x86/hvm/svm/svm.h
===================================================================
--- 2007-05-14.orig/xen/include/asm-x86/hvm/svm/svm.h   2007-04-23 
10:01:46.000000000 +0200
+++ 2007-05-14/xen/include/asm-x86/hvm/svm/svm.h        2007-05-14 
14:28:19.000000000 +0200
@@ -62,6 +62,11 @@ static inline int svm_pae_enabled(struct
     return svm_paging_enabled(v) && (guest_cr4 & X86_CR4_PAE);
 }
 
+static inline int svm_nx_enabled(struct vcpu *v)
+{
+    return v->arch.hvm_svm.cpu_shadow_efer & EFER_NX;
+}
+
 static inline int svm_pgbit_test(struct vcpu *v)
 {
     return v->arch.hvm_svm.cpu_shadow_cr0 & X86_CR0_PG;
Index: 2007-05-14/xen/include/asm-x86/hvm/vmx/vmcs.h
===================================================================
--- 2007-05-14.orig/xen/include/asm-x86/hvm/vmx/vmcs.h  2007-05-03 
09:45:09.000000000 +0200
+++ 2007-05-14/xen/include/asm-x86/hvm/vmx/vmcs.h       2007-05-14 
14:28:19.000000000 +0200
@@ -39,7 +39,6 @@ enum {
     VMX_INDEX_MSR_STAR,
     VMX_INDEX_MSR_CSTAR,
     VMX_INDEX_MSR_SYSCALL_MASK,
-    VMX_INDEX_MSR_EFER,
 
     VMX_MSR_COUNT
 };
@@ -47,7 +46,6 @@ enum {
 struct vmx_msr_state {
     unsigned long flags;
     unsigned long msrs[VMX_MSR_COUNT];
-    unsigned long shadow_gs;
 };
 
 struct arch_vmx_struct {
@@ -76,7 +74,11 @@ struct arch_vmx_struct {
     unsigned long        cpu_shadow_cr4; /* copy of guest read shadow CR4 */
     unsigned long        cpu_cr2; /* save CR2 */
     unsigned long        cpu_cr3;
+#ifdef __x86_64__
     struct vmx_msr_state msr_state;
+    unsigned long        shadow_gs;
+#endif
+    unsigned long        efer;
     unsigned long        vmxassist_enabled:1;
 };
 
Index: 2007-05-14/xen/include/asm-x86/hvm/vmx/vmx.h
===================================================================
--- 2007-05-14.orig/xen/include/asm-x86/hvm/vmx/vmx.h   2007-04-23 
10:01:46.000000000 +0200
+++ 2007-05-14/xen/include/asm-x86/hvm/vmx/vmx.h        2007-05-14 
14:28:19.000000000 +0200
@@ -261,14 +261,12 @@ static inline int vmx_paging_enabled(str
 
 static inline int vmx_long_mode_enabled(struct vcpu *v)
 {
-    u64 efer = v->arch.hvm_vmx.msr_state.msrs[VMX_INDEX_MSR_EFER];
-    return efer & EFER_LMA;
+    return v->arch.hvm_vmx.efer & EFER_LMA;
 }
 
 static inline int vmx_lme_is_set(struct vcpu *v)
 {
-    u64 efer = v->arch.hvm_vmx.msr_state.msrs[VMX_INDEX_MSR_EFER];
-    return efer & EFER_LME;
+    return v->arch.hvm_vmx.efer & EFER_LME;
 }
 
 static inline int vmx_pgbit_test(struct vcpu *v)
Index: 2007-05-14/xen/include/asm-x86/msr.h
===================================================================
--- 2007-05-14.orig/xen/include/asm-x86/msr.h   2007-04-23 10:01:46.000000000 
+0200
+++ 2007-05-14/xen/include/asm-x86/msr.h        2007-05-14 14:28:19.000000000 
+0200
@@ -3,6 +3,9 @@
 
 #ifndef __ASSEMBLY__
 
+#include <xen/smp.h>
+#include <xen/percpu.h>
+
 #define rdmsr(msr,val1,val2) \
      __asm__ __volatile__("rdmsr" \
                          : "=a" (val1), "=d" (val2) \
@@ -142,6 +145,25 @@ static inline void wrmsrl(unsigned int m
 #define EFER_NX (1<<_EFER_NX)
 #define EFER_SVME (1<<_EFER_SVME)
 
+#ifndef __ASSEMBLY__
+
+DECLARE_PER_CPU(__u64, efer);
+
+static inline __u64 read_efer(void)
+{
+    if (!this_cpu(efer))
+        rdmsrl(MSR_EFER, this_cpu(efer));
+    return this_cpu(efer);
+}
+
+static inline void write_efer(__u64 val)
+{
+    this_cpu(efer) = val;
+    wrmsrl(MSR_EFER, val);
+}
+
+#endif
+
 /* Intel MSRs. Some also available on other CPUs */
 #define MSR_IA32_PLATFORM_ID   0x17
 



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

<Prev in Thread] Current Thread [Next in Thread>
  • [Xen-devel] [PATCH] x86: fix EFER handling, Jan Beulich <=