Mailing List Archive

# HG changeset patch
# User kfraser@localhost.localdomain
# Date 1186585757 -3600
# Node ID 25e5c1b9faad01e03bb3a35b709d79c00697bf30
# Parent 359707941ae84944d757faa1c4c2b76c59fd7333
hvm: Unify %cr0 handling.
Signed-off-by: Keir Fraser <keir@xensource.com>
---
xen/arch/x86/hvm/hvm.c | 103 ++++++++++++++++++++--
xen/arch/x86/hvm/svm/svm.c | 144 +++++++++----------------------
xen/arch/x86/hvm/svm/vmcb.c | 12 +-
xen/arch/x86/hvm/vmx/vmx.c | 176 +++++++-------------------------------
xen/include/asm-x86/hvm/hvm.h | 9 +
xen/include/asm-x86/hvm/support.h | 1
xen/include/asm-x86/hvm/vcpu.h | 6 -
7 files changed, 190 insertions(+), 261 deletions(-)

diff -r 359707941ae8 -r 25e5c1b9faad xen/arch/x86/hvm/hvm.c
--- a/xen/arch/x86/hvm/hvm.c Wed Aug 08 15:03:40 2007 +0100
+++ b/xen/arch/x86/hvm/hvm.c Wed Aug 08 16:09:17 2007 +0100
@@ -76,13 +76,6 @@ void hvm_enable(struct hvm_function_tabl
hvm_enabled = 1;
}

-void hvm_stts(struct vcpu *v)
-{
- /* FPU state already dirty? Then no need to setup_fpu() lazily. */
- if ( !v->fpu_dirtied )
- hvm_funcs.stts(v);
-}
-
void hvm_set_guest_time(struct vcpu *v, u64 gtime)
{
u64 host_tsc;
@@ -112,7 +105,8 @@ void hvm_do_resume(struct vcpu *v)
{
ioreq_t *p;

- hvm_stts(v);
+ if ( !v->fpu_dirtied )
+ hvm_funcs.stts(v);

pt_thaw_time(v);

@@ -518,6 +512,99 @@ void hvm_triple_fault(void)
gdprintk(XENLOG_INFO, "Triple fault on VCPU%d - "
"invoking HVM system reset.\n", v->vcpu_id);
domain_shutdown(v->domain, SHUTDOWN_reboot);
+}
+
+int hvm_set_cr0(unsigned long value)
+{
+ struct vcpu *v = current;
+ unsigned long mfn, old_base_mfn, old_value = v->arch.hvm_vcpu.guest_cr[0];
+
+ HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR0 value = %lx", value);
+
+ if ( (u32)value != value )
+ {
+ HVM_DBG_LOG(DBG_LEVEL_1,
+ "Guest attempts to set upper 32 bits in CR0: %lx",
+ value);
+ hvm_inject_exception(TRAP_gp_fault, 0, 0);
+ return 0;
+ }
+
+ value &= ~HVM_CR0_GUEST_RESERVED_BITS;
+
+ /* ET is reserved and should be always be 1. */
+ value |= X86_CR0_ET;
+
+ if ( (value & (X86_CR0_PE|X86_CR0_PG)) == X86_CR0_PG )
+ {
+ hvm_inject_exception(TRAP_gp_fault, 0, 0);
+ return 0;
+ }
+
+ if ( (value & X86_CR0_PG) && !(old_value & X86_CR0_PG) )
+ {
+ if ( v->arch.hvm_vcpu.guest_efer & EFER_LME )
+ {
+ if ( !(v->arch.hvm_vcpu.guest_cr[4] & X86_CR4_PAE) )
+ {
+ HVM_DBG_LOG(DBG_LEVEL_1, "Enable paging before PAE enable");
+ hvm_inject_exception(TRAP_gp_fault, 0, 0);
+ return 0;
+ }
+ HVM_DBG_LOG(DBG_LEVEL_1, "Enabling long mode");
+ v->arch.hvm_vcpu.guest_efer |= EFER_LMA;
+ hvm_update_guest_efer(v);
+ }
+
+ if ( !paging_mode_hap(v->domain) )
+ {
+ /* The guest CR3 must be pointing to the guest physical. */
+ mfn = get_mfn_from_gpfn(v->arch.hvm_vcpu.guest_cr[3]>>PAGE_SHIFT);
+ if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain))
+ {
+ gdprintk(XENLOG_ERR, "Invalid CR3 value = %lx (mfn=%lx)\n",
+ v->arch.hvm_vcpu.guest_cr[3], mfn);
+ domain_crash(v->domain);
+ return 0;
+ }
+
+ /* Now arch.guest_table points to machine physical. */
+ old_base_mfn = pagetable_get_pfn(v->arch.guest_table);
+ v->arch.guest_table = pagetable_from_pfn(mfn);
+ if ( old_base_mfn )
+ put_page(mfn_to_page(old_base_mfn));
+
+ HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx, mfn = %lx",
+ v->arch.hvm_vcpu.guest_cr[3], mfn);
+ }
+ }
+ else if ( !(value & X86_CR0_PG) && (old_value & X86_CR0_PG) )
+ {
+ /* When CR0.PG is cleared, LMA is cleared immediately. */
+ if ( hvm_long_mode_enabled(v) )
+ {
+ v->arch.hvm_vcpu.guest_efer &= ~EFER_LMA;
+ hvm_update_guest_efer(v);
+ }
+
+ if ( !paging_mode_hap(v->domain) && v->arch.hvm_vcpu.guest_cr[3] )
+ {
+ put_page(mfn_to_page(get_mfn_from_gpfn(
+ v->arch.hvm_vcpu.guest_cr[3] >> PAGE_SHIFT)));
+ v->arch.guest_table = pagetable_null();
+ }
+ }
+
+ v->arch.hvm_vcpu.guest_cr[0] = value;
+ v->arch.hvm_vcpu.hw_cr[0] = value;
+ if ( !paging_mode_hap(v->domain) )
+ v->arch.hvm_vcpu.hw_cr[0] |= X86_CR0_PG | X86_CR0_WP;
+ hvm_update_guest_cr(v, 0);
+
+ if ( (value ^ old_value) & X86_CR0_PG )
+ paging_update_paging_modes(v);
+
+ return 1;
}

int hvm_set_cr3(unsigned long value)
diff -r 359707941ae8 -r 25e5c1b9faad xen/arch/x86/hvm/svm/svm.c
--- a/xen/arch/x86/hvm/svm/svm.c Wed Aug 08 15:03:40 2007 +0100
+++ b/xen/arch/x86/hvm/svm/svm.c Wed Aug 08 16:09:17 2007 +0100
@@ -344,7 +344,8 @@ int svm_vmcb_restore(struct vcpu *v, str
vmcb->rflags = c->rflags;

v->arch.hvm_vcpu.guest_cr[0] = c->cr0;
- vmcb->cr0 = c->cr0 | X86_CR0_WP | X86_CR0_ET | X86_CR0_PG;
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] =
+ c->cr0 | X86_CR0_WP | X86_CR0_ET | X86_CR0_PG;

v->arch.hvm_vcpu.guest_cr[2] = c->cr2;

@@ -392,7 +393,7 @@ int svm_vmcb_restore(struct vcpu *v, str
}

skip_cr3:
- vmcb->cr4 = c->cr4 | HVM_CR4_HOST_MASK;
+ vmcb->cr4 = v->arch.hvm_vcpu.hw_cr[4] = c->cr4 | HVM_CR4_HOST_MASK;
v->arch.hvm_vcpu.guest_cr[4] = c->cr4;

vmcb->idtr.limit = c->idtr_limit;
@@ -448,10 +449,10 @@ int svm_vmcb_restore(struct vcpu *v, str

if ( paging_mode_hap(v->domain) )
{
- vmcb->cr0 = v->arch.hvm_vcpu.guest_cr[0];
- vmcb->cr4 = (v->arch.hvm_vcpu.guest_cr[4] |
- (HVM_CR4_HOST_MASK & ~X86_CR4_PAE));
- vmcb->cr3 = c->cr3;
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] = v->arch.hvm_vcpu.guest_cr[0];
+ vmcb->cr4 = v->arch.hvm_vcpu.hw_cr[4] =
+ v->arch.hvm_vcpu.guest_cr[4] | (HVM_CR4_HOST_MASK & ~X86_CR4_PAE);
+ vmcb->cr3 = v->arch.hvm_vcpu.hw_cr[3] = c->cr3;
vmcb->np_enable = 1;
vmcb->g_pat = 0x0007040600070406ULL; /* guest PAT */
vmcb->h_cr3 = pagetable_get_paddr(v->domain->arch.phys_table);
@@ -580,18 +581,38 @@ static void svm_update_host_cr3(struct v

static void svm_update_guest_cr(struct vcpu *v, unsigned int cr)
{
+ struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
+
switch ( cr )
{
+ case 0:
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0];
+ break;
+ case 2:
+ vmcb->cr2 = v->arch.hvm_vcpu.hw_cr[2];
+ break;
case 3:
- v->arch.hvm_svm.vmcb->cr3 = v->arch.hvm_vcpu.hw_cr[3];
+ vmcb->cr3 = v->arch.hvm_vcpu.hw_cr[3];
svm_asid_inv_asid(v);
break;
case 4:
- v->arch.hvm_svm.vmcb->cr4 = v->arch.hvm_vcpu.hw_cr[4];
+ vmcb->cr4 = v->arch.hvm_vcpu.hw_cr[4];
break;
default:
BUG();
}
+}
+
+static void svm_update_guest_efer(struct vcpu *v)
+{
+#ifdef __x86_64__
+ struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
+
+ if ( v->arch.hvm_vcpu.guest_efer & EFER_LMA )
+ vmcb->efer |= EFER_LME | EFER_LMA;
+ else
+ vmcb->efer &= ~(EFER_LME | EFER_LMA);
+#endif
}

static void svm_flush_guest_tlbs(void)
@@ -703,7 +724,7 @@ static void svm_stts(struct vcpu *v)
if ( !(v->arch.hvm_vcpu.guest_cr[0] & X86_CR0_TS) )
{
v->arch.hvm_svm.vmcb->exception_intercepts |= 1U << TRAP_no_device;
- vmcb->cr0 |= X86_CR0_TS;
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] |= X86_CR0_TS;
}
}

@@ -928,6 +949,7 @@ static struct hvm_function_table svm_fun
.get_segment_register = svm_get_segment_register,
.update_host_cr3 = svm_update_host_cr3,
.update_guest_cr = svm_update_guest_cr,
+ .update_guest_efer = svm_update_guest_efer,
.flush_guest_tlbs = svm_flush_guest_tlbs,
.update_vtpr = svm_update_vtpr,
.stts = svm_stts,
@@ -1023,7 +1045,7 @@ static void svm_do_no_device_fault(struc
vmcb->exception_intercepts &= ~(1U << TRAP_no_device);

if ( !(v->arch.hvm_vcpu.guest_cr[0] & X86_CR0_TS) )
- vmcb->cr0 &= ~X86_CR0_TS;
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] &= ~X86_CR0_TS;
}

/* Reserved bits ECX: [31:14], [12:4], [2:1]*/
@@ -1597,31 +1619,11 @@ static int svm_set_cr0(unsigned long val
static int svm_set_cr0(unsigned long value)
{
struct vcpu *v = current;
- unsigned long mfn, old_value = v->arch.hvm_vcpu.guest_cr[0];
- struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
- unsigned long old_base_mfn;
-
- HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR0 value = %lx", value);
-
- if ( (u32)value != value )
- {
- HVM_DBG_LOG(DBG_LEVEL_1,
- "Guest attempts to set upper 32 bits in CR0: %lx",
- value);
- svm_inject_exception(v, TRAP_gp_fault, 1, 0);
+ struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
+ int rc = hvm_set_cr0(value);
+
+ if ( rc == 0 )
return 0;
- }
-
- value &= ~HVM_CR0_GUEST_RESERVED_BITS;
-
- /* ET is reserved and should be always be 1. */
- value |= X86_CR0_ET;
-
- if ( (value & (X86_CR0_PE|X86_CR0_PG)) == X86_CR0_PG )
- {
- svm_inject_exception(v, TRAP_gp_fault, 1, 0);
- return 0;
- }

/* TS cleared? Then initialise FPU now. */
if ( !(value & X86_CR0_TS) )
@@ -1629,67 +1631,6 @@ static int svm_set_cr0(unsigned long val
setup_fpu(v);
vmcb->exception_intercepts &= ~(1U << TRAP_no_device);
}
-
- if ( (value & X86_CR0_PG) && !(old_value & X86_CR0_PG) )
- {
- if ( svm_lme_is_set(v) )
- {
- if ( !(v->arch.hvm_vcpu.guest_cr[4] & X86_CR4_PAE) )
- {
- HVM_DBG_LOG(DBG_LEVEL_1, "Enable paging before PAE enable");
- svm_inject_exception(v, TRAP_gp_fault, 1, 0);
- return 0;
- }
- HVM_DBG_LOG(DBG_LEVEL_1, "Enable the Long mode");
- v->arch.hvm_vcpu.guest_efer |= EFER_LMA;
- vmcb->efer |= EFER_LMA | EFER_LME;
- }
-
- if ( !paging_mode_hap(v->domain) )
- {
- /* The guest CR3 must be pointing to the guest physical. */
- mfn = get_mfn_from_gpfn(v->arch.hvm_vcpu.guest_cr[3] >> PAGE_SHIFT);
- if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain))
- {
- gdprintk(XENLOG_ERR, "Invalid CR3 value = %lx (mfn=%lx)\n",
- v->arch.hvm_vcpu.guest_cr[3], mfn);
- domain_crash(v->domain);
- return 0;
- }
-
- /* Now arch.guest_table points to machine physical. */
- old_base_mfn = pagetable_get_pfn(v->arch.guest_table);
- v->arch.guest_table = pagetable_from_pfn(mfn);
- if ( old_base_mfn )
- put_page(mfn_to_page(old_base_mfn));
-
- HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx, mfn = %lx",
- v->arch.hvm_vcpu.guest_cr[3], mfn);
- }
- }
- else if ( !(value & X86_CR0_PG) && (old_value & X86_CR0_PG) )
- {
- /* When CR0.PG is cleared, LMA is cleared immediately. */
- if ( hvm_long_mode_enabled(v) )
- {
- vmcb->efer &= ~(EFER_LME | EFER_LMA);
- v->arch.hvm_vcpu.guest_efer &= ~EFER_LMA;
- }
-
- if ( !paging_mode_hap(v->domain) && v->arch.hvm_vcpu.guest_cr[3] )
- {
- put_page(mfn_to_page(get_mfn_from_gpfn(
- v->arch.hvm_vcpu.guest_cr[3] >> PAGE_SHIFT)));
- v->arch.guest_table = pagetable_null();
- }
- }
-
- vmcb->cr0 = v->arch.hvm_vcpu.guest_cr[0] = value;
- if ( !paging_mode_hap(v->domain) )
- vmcb->cr0 |= X86_CR0_PG | X86_CR0_WP;
-
- if ( (value ^ old_value) & X86_CR0_PG )
- paging_update_paging_modes(v);

return 1;
}
@@ -1833,7 +1774,7 @@ static void svm_cr_access(
/* TS being cleared means that it's time to restore fpu state. */
setup_fpu(current);
vmcb->exception_intercepts &= ~(1U << TRAP_no_device);
- vmcb->cr0 &= ~X86_CR0_TS; /* clear TS */
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] &= ~X86_CR0_TS; /* clear TS */
v->arch.hvm_vcpu.guest_cr[0] &= ~X86_CR0_TS; /* clear TS */
break;

@@ -2144,20 +2085,21 @@ static int svm_reset_to_realmode(struct

memset(regs, 0, sizeof(struct cpu_user_regs));

- vmcb->cr0 = X86_CR0_ET | X86_CR0_PG | X86_CR0_WP;
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] =
+ X86_CR0_ET | X86_CR0_PG | X86_CR0_WP;
v->arch.hvm_vcpu.guest_cr[0] = X86_CR0_ET;

vmcb->cr2 = 0;
vmcb->efer = EFER_SVME;

- vmcb->cr4 = HVM_CR4_HOST_MASK;
+ vmcb->cr4 = v->arch.hvm_vcpu.hw_cr[4] = HVM_CR4_HOST_MASK;
v->arch.hvm_vcpu.guest_cr[4] = 0;

if ( paging_mode_hap(v->domain) )
{
- vmcb->cr0 = v->arch.hvm_vcpu.guest_cr[0];
- vmcb->cr4 = (v->arch.hvm_vcpu.guest_cr[4] |
- (HVM_CR4_HOST_MASK & ~X86_CR4_PAE));
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] = v->arch.hvm_vcpu.guest_cr[0];
+ vmcb->cr4 = v->arch.hvm_vcpu.hw_cr[4] =
+ v->arch.hvm_vcpu.guest_cr[4] | (HVM_CR4_HOST_MASK & ~X86_CR4_PAE);
}

/* This will jump to ROMBIOS */
diff -r 359707941ae8 -r 25e5c1b9faad xen/arch/x86/hvm/svm/vmcb.c
--- a/xen/arch/x86/hvm/svm/vmcb.c Wed Aug 08 15:03:40 2007 +0100
+++ b/xen/arch/x86/hvm/svm/vmcb.c Wed Aug 08 16:09:17 2007 +0100
@@ -217,24 +217,26 @@ static int construct_vmcb(struct vcpu *v
vmcb->tr.limit = 0xff;

/* Guest CR0. */
- vmcb->cr0 = read_cr0();
- v->arch.hvm_vcpu.guest_cr[0] = vmcb->cr0 & ~(X86_CR0_PG | X86_CR0_TS);
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] = read_cr0();
+ v->arch.hvm_vcpu.guest_cr[0] =
+ v->arch.hvm_vcpu.hw_cr[0] & ~(X86_CR0_PG | X86_CR0_TS);

/* Guest CR4. */
v->arch.hvm_vcpu.guest_cr[4] =
read_cr4() & ~(X86_CR4_PGE | X86_CR4_PSE | X86_CR4_PAE);
- vmcb->cr4 = v->arch.hvm_vcpu.guest_cr[4] | HVM_CR4_HOST_MASK;
+ vmcb->cr4 = v->arch.hvm_vcpu.hw_cr[4] =
+ v->arch.hvm_vcpu.guest_cr[4] | HVM_CR4_HOST_MASK;

paging_update_paging_modes(v);
vmcb->cr3 = v->arch.hvm_vcpu.hw_cr[3];

if ( paging_mode_hap(v->domain) )
{
- vmcb->cr0 = v->arch.hvm_vcpu.guest_cr[0];
+ vmcb->cr0 = v->arch.hvm_vcpu.hw_cr[0] = v->arch.hvm_vcpu.guest_cr[0];
vmcb->np_enable = 1; /* enable nested paging */
vmcb->g_pat = 0x0007040600070406ULL; /* guest PAT */
vmcb->h_cr3 = pagetable_get_paddr(v->domain->arch.phys_table);
- vmcb->cr4 = v->arch.hvm_vcpu.guest_cr[4] =
+ vmcb->cr4 = v->arch.hvm_vcpu.hw_cr[4] = v->arch.hvm_vcpu.guest_cr[4] =
HVM_CR4_HOST_MASK & ~X86_CR4_PAE;
vmcb->exception_intercepts = HVM_TRAP_MASK;

diff -r 359707941ae8 -r 25e5c1b9faad xen/arch/x86/hvm/vmx/vmx.c
--- a/xen/arch/x86/hvm/vmx/vmx.c Wed Aug 08 15:03:40 2007 +0100
+++ b/xen/arch/x86/hvm/vmx/vmx.c Wed Aug 08 16:09:17 2007 +0100
@@ -61,6 +61,7 @@ static int vmx_alloc_vlapic_mapping(str
static int vmx_alloc_vlapic_mapping(struct domain *d);
static void vmx_free_vlapic_mapping(struct domain *d);
static void vmx_install_vlapic_mapping(struct vcpu *v);
+static void vmx_update_guest_efer(struct vcpu *v);

static int vmx_domain_initialise(struct domain *d)
{
@@ -101,33 +102,6 @@ static void vmx_vcpu_destroy(struct vcpu
}

#ifdef __x86_64__
-
-static int vmx_lme_is_set(struct vcpu *v)
-{
- return v->arch.hvm_vcpu.guest_efer & EFER_LME;
-}
-
-static void vmx_enable_long_mode(struct vcpu *v)
-{
- unsigned long vm_entry_value;
-
- vm_entry_value = __vmread(VM_ENTRY_CONTROLS);
- vm_entry_value |= VM_ENTRY_IA32E_MODE;
- __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value);
-
- v->arch.hvm_vcpu.guest_efer |= EFER_LMA;
-}
-
-static void vmx_disable_long_mode(struct vcpu *v)
-{
- unsigned long vm_entry_value;
-
- vm_entry_value = __vmread(VM_ENTRY_CONTROLS);
- vm_entry_value &= ~VM_ENTRY_IA32E_MODE;
- __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value);
-
- v->arch.hvm_vcpu.guest_efer &= ~EFER_LMA;
-}

static DEFINE_PER_CPU(struct vmx_msr_state, host_msr_state);

@@ -378,13 +352,6 @@ static void vmx_restore_guest_msrs(struc

#else /* __i386__ */

-static int vmx_lme_is_set(struct vcpu *v)
-{ return 0; }
-static void vmx_enable_long_mode(struct vcpu *v)
-{ BUG(); }
-static void vmx_disable_long_mode(struct vcpu *v)
-{ BUG(); }
-
#define vmx_save_host_msrs() ((void)0)

static void vmx_restore_host_msrs(void)
@@ -644,8 +611,7 @@ int vmx_vmcs_restore(struct vcpu *v, str
skip_cr3:
v->arch.hvm_vcpu.guest_cr[3] = c->cr3;

- if ( hvm_long_mode_enabled(v) )
- vmx_enable_long_mode(v);
+ vmx_update_guest_efer(v);

__vmwrite(GUEST_CR4, (c->cr4 | HVM_CR4_HOST_MASK));
v->arch.hvm_vcpu.guest_cr[4] = c->cr4;
@@ -1095,6 +1061,14 @@ static void vmx_update_guest_cr(struct v

switch ( cr )
{
+ case 0:
+ v->arch.hvm_vcpu.hw_cr[0] |= X86_CR0_PE | X86_CR0_NE;
+ __vmwrite(GUEST_CR0, v->arch.hvm_vcpu.hw_cr[0]);
+ __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vcpu.guest_cr[0]);
+ break;
+ case 2:
+ /* CR2 is updated in exit stub. */
+ break;
case 3:
__vmwrite(GUEST_CR3, v->arch.hvm_vcpu.hw_cr[3]);
break;
@@ -1107,6 +1081,26 @@ static void vmx_update_guest_cr(struct v
}

vmx_vmcs_exit(v);
+}
+
+static void vmx_update_guest_efer(struct vcpu *v)
+{
+#ifdef __x86_64__
+ unsigned long vm_entry_value;
+
+ ASSERT((v == current) || !vcpu_runnable(v));
+
+ vmx_vmcs_enter(v);
+
+ vm_entry_value = __vmread(VM_ENTRY_CONTROLS);
+ if ( v->arch.hvm_vcpu.guest_efer & EFER_LMA )
+ vm_entry_value |= VM_ENTRY_IA32E_MODE;
+ else
+ vm_entry_value &= ~VM_ENTRY_IA32E_MODE;
+ __vmwrite(VM_ENTRY_CONTROLS, vm_entry_value);
+
+ vmx_vmcs_exit(v);
+#endif
}

static void vmx_flush_guest_tlbs(void)
@@ -1172,6 +1166,7 @@ static struct hvm_function_table vmx_fun
.get_segment_register = vmx_get_segment_register,
.update_host_cr3 = vmx_update_host_cr3,
.update_guest_cr = vmx_update_guest_cr,
+ .update_guest_efer = vmx_update_guest_efer,
.flush_guest_tlbs = vmx_flush_guest_tlbs,
.update_vtpr = vmx_update_vtpr,
.stts = vmx_stts,
@@ -2110,108 +2105,17 @@ static int vmx_set_cr0(unsigned long val
static int vmx_set_cr0(unsigned long value)
{
struct vcpu *v = current;
- unsigned long mfn;
unsigned long eip;
- int paging_enabled;
- unsigned long old_cr0;
- unsigned long old_base_mfn;
-
- HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR0 value = %lx", value);
-
- if ( (u32)value != value )
- {
- HVM_DBG_LOG(DBG_LEVEL_1,
- "Guest attempts to set upper 32 bits in CR0: %lx",
- value);
- vmx_inject_hw_exception(v, TRAP_gp_fault, 0);
+ int rc = hvm_set_cr0(value);
+
+ if ( rc == 0 )
return 0;
- }
-
- value &= ~HVM_CR0_GUEST_RESERVED_BITS;
-
- /* ET is reserved and should be always be 1. */
- value |= X86_CR0_ET;
-
- if ( (value & (X86_CR0_PE | X86_CR0_PG)) == X86_CR0_PG )
- {
- vmx_inject_hw_exception(v, TRAP_gp_fault, 0);
- return 0;
- }

/* TS cleared? Then initialise FPU now. */
if ( !(value & X86_CR0_TS) )
{
setup_fpu(v);
__vm_clear_bit(EXCEPTION_BITMAP, TRAP_no_device);
- }
-
- old_cr0 = v->arch.hvm_vcpu.guest_cr[0];
- paging_enabled = old_cr0 & X86_CR0_PG;
-
- v->arch.hvm_vcpu.hw_cr[0] = (value | X86_CR0_PE | X86_CR0_PG |
- X86_CR0_NE | X86_CR0_WP);
- __vmwrite(GUEST_CR0, v->arch.hvm_vcpu.hw_cr[0]);
-
- v->arch.hvm_vcpu.guest_cr[0] = value;
- __vmwrite(CR0_READ_SHADOW, v->arch.hvm_vcpu.guest_cr[0]);
-
- /* Trying to enable paging. */
- if ( (value & X86_CR0_PE) && (value & X86_CR0_PG) && !paging_enabled )
- {
- if ( vmx_lme_is_set(v) && !hvm_long_mode_enabled(v) )
- {
- if ( !(v->arch.hvm_vcpu.guest_cr[4] & X86_CR4_PAE) )
- {
- HVM_DBG_LOG(DBG_LEVEL_1, "Guest enabled paging "
- "with EFER.LME set but not CR4.PAE");
- vmx_inject_hw_exception(v, TRAP_gp_fault, 0);
- return 0;
- }
-
- HVM_DBG_LOG(DBG_LEVEL_1, "Enabling long mode");
- vmx_enable_long_mode(v);
- }
-
- /*
- * The guest CR3 must be pointing to the guest physical.
- */
- mfn = get_mfn_from_gpfn(v->arch.hvm_vcpu.guest_cr[3] >> PAGE_SHIFT);
- if ( !mfn_valid(mfn) || !get_page(mfn_to_page(mfn), v->domain) )
- {
- gdprintk(XENLOG_ERR, "Invalid CR3 value = %lx (mfn=%lx)\n",
- v->arch.hvm_vcpu.guest_cr[3], mfn);
- domain_crash(v->domain);
- return 0;
- }
-
- /*
- * Now arch.guest_table points to machine physical.
- */
- old_base_mfn = pagetable_get_pfn(v->arch.guest_table);
- v->arch.guest_table = pagetable_from_pfn(mfn);
- if ( old_base_mfn )
- put_page(mfn_to_page(old_base_mfn));
-
- HVM_DBG_LOG(DBG_LEVEL_VMMU, "Update CR3 value = %lx, mfn = %lx",
- v->arch.hvm_vcpu.guest_cr[3], mfn);
-
- paging_update_paging_modes(v);
- }
-
- /* Trying to disable paging. */
- if ( ((value & (X86_CR0_PE | X86_CR0_PG)) != (X86_CR0_PE | X86_CR0_PG)) &&
- paging_enabled )
- {
- /* When CR0.PG is cleared, LMA is cleared immediately. */
- if ( hvm_long_mode_enabled(v) )
- vmx_disable_long_mode(v);
-
- if ( v->arch.hvm_vcpu.guest_cr[3] )
- {
- put_page(mfn_to_page(get_mfn_from_gpfn(
- v->arch.hvm_vcpu.guest_cr[3] >> PAGE_SHIFT)));
- v->arch.guest_table = pagetable_null();
- }
}

/*
@@ -2219,14 +2123,8 @@ static int vmx_set_cr0(unsigned long val
* real-mode by performing a world switch to VMXAssist whenever
* a partition disables the CR0.PE bit.
*/
- if ( (value & X86_CR0_PE) == 0 )
- {
- if ( value & X86_CR0_PG )
- {
- vmx_inject_hw_exception(v, TRAP_gp_fault, 0);
- return 0;
- }
-
+ if ( !(value & X86_CR0_PE) )
+ {
if ( vmx_assist(v, VMX_ASSIST_INVOKE) )
{
eip = __vmread(GUEST_RIP);
@@ -2247,8 +2145,6 @@ static int vmx_set_cr0(unsigned long val
return 0; /* do not update eip! */
}
}
- else if ( (value & (X86_CR0_PE | X86_CR0_PG)) == X86_CR0_PE )
- paging_update_paging_modes(v);

return 1;
}
diff -r 359707941ae8 -r 25e5c1b9faad xen/include/asm-x86/hvm/hvm.h
--- a/xen/include/asm-x86/hvm/hvm.h Wed Aug 08 15:03:40 2007 +0100
+++ b/xen/include/asm-x86/hvm/hvm.h Wed Aug 08 16:09:17 2007 +0100
@@ -112,9 +112,10 @@ struct hvm_function_table {
void (*update_host_cr3)(struct vcpu *v);

/*
- * Called to inform HVM layer that a guest control register has changed.
+ * Called to inform HVM layer that a guest CRn or EFER has changed.
*/
void (*update_guest_cr)(struct vcpu *v, unsigned int cr);
+ void (*update_guest_efer)(struct vcpu *v);

/*
* Called to ensure than all guest-specific mappings in a tagged TLB
@@ -225,6 +226,11 @@ static inline void hvm_update_guest_cr(s
hvm_funcs.update_guest_cr(v, cr);
}

+static inline void hvm_update_guest_efer(struct vcpu *v)
+{
+ hvm_funcs.update_guest_efer(v);
+}
+
static inline void
hvm_flush_guest_tlbs(void)
{
@@ -250,7 +256,6 @@ hvm_get_segment_register(struct vcpu *v,

void hvm_cpuid(unsigned int input, unsigned int *eax, unsigned int *ebx,
unsigned int *ecx, unsigned int *edx);
-void hvm_stts(struct vcpu *v);
void hvm_migrate_timers(struct vcpu *v);
void hvm_do_resume(struct vcpu *v);

diff -r 359707941ae8 -r 25e5c1b9faad xen/include/asm-x86/hvm/support.h
--- a/xen/include/asm-x86/hvm/support.h Wed Aug 08 15:03:40 2007 +0100
+++ b/xen/include/asm-x86/hvm/support.h Wed Aug 08 16:09:17 2007 +0100
@@ -234,6 +234,7 @@ void hvm_hlt(unsigned long rflags);
void hvm_hlt(unsigned long rflags);
void hvm_triple_fault(void);

+int hvm_set_cr0(unsigned long value);
int hvm_set_cr3(unsigned long value);
int hvm_set_cr4(unsigned long value);

diff -r 359707941ae8 -r 25e5c1b9faad xen/include/asm-x86/hvm/vcpu.h
--- a/xen/include/asm-x86/hvm/vcpu.h Wed Aug 08 15:03:40 2007 +0100
+++ b/xen/include/asm-x86/hvm/vcpu.h Wed Aug 08 16:09:17 2007 +0100
@@ -33,11 +33,7 @@ struct hvm_vcpu {
unsigned long guest_cr[5];
unsigned long guest_efer;

- /*
- * Processor-visible CR0-4 while guest executes.
- * Only CR3 is guaranteed to be valid: all other array entries are private
- * to the specific HVM implementation (e.g., VMX, SVM).
- */
+ /* Processor-visible control-register values, while guest executes. */
unsigned long hw_cr[5];

struct hvm_io_op io_op;

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