Skip to contentSkip to conversation

VMX Unrestricted mode support

PATCH In review Nitin A Kamble 5 participants 13 messages May 28, 2009 – Jun 4, 2009

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 VMX Unrestricted mode support

Avi,

A new VMX feature "Unrestricted Guest" feature is added in the VMX specification. You can look at the latest Intel processor manual for details of the feature here:

http://www.intel.com/products/processor/manuals

It allows kvm guests to run real mode and unpaged mode
code natively in the VMX mode when EPT is turned on. With the unrestricted guest there is no need to emulate the guest real mode code in the vm86 container or in the emulator. Also the guest big real mode code works like native.

The attached patch enhances KVM to use the unrestricted guest feature
if available on the processor. It also adds a new kernel/module parameter to disable the unrestricted guest feature at the boot time.

Thanks & Regards,
Nitin

(quoted diff) +0 -0

b/arch/x86/include/asm/kvm_host.h
index d2b082d..7832599 100644

arch/x86/include/asm/kvm_host.h +6 -2
@@ -40,9 +40,13 @@
 #define KVM_GUEST_CR0_MASK				   \
 	(X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
 	 | X86_CR0_NW | X86_CR0_CD)
+#define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST				\
+	(X86_CR0_WP | X86_CR0_NE | X86_CR0_TS | X86_CR0_MP)
+#define KVM_VM_CR0_ALWAYS_ON_RESTRICTED_GUEST				\
+	(KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST | X86_CR0_PG | X86_CR0_PE)
 #define KVM_VM_CR0_ALWAYS_ON						\
-	(X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
-	 | X86_CR0_MP)
+	(enable_unrestricted_guest ? KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST \
+	: KVM_VM_CR0_ALWAYS_ON_RESTRICTED_GUEST)
 #define KVM_GUEST_CR4_MASK						\
 	(X86_CR4_VME | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_PGE | X86_CR4_VMXE)
 #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE)
arch/x86/include/asm/vmx.h +1 -0
index 498f944..c73da02 100644
@@ -55,6 +55,7 @@
 #define SECONDARY_EXEC_ENABLE_EPT               0x00000002
 #define SECONDARY_EXEC_ENABLE_VPID              0x00000020
 #define SECONDARY_EXEC_WBINVD_EXITING		0x00000040
+#define SECONDARY_EXEC_UNRESTRICTED_GUEST	0x00000080
 
 
 #define PIN_BASED_EXT_INTR_MASK                 0x00000001
arch/x86/kvm/vmx.c +0 -0
index 25f1239..703d2c4 100644
@@ -50,6 +50,10 @@ module_param_named(flexpriority,

flexpriority_enabled, bool, S_IRUGO);
static int __read_mostly enable_ept = 1; module_param_named(ept, enable_ept, bool, S_IRUGO);

+static int __read_mostly enable_unrestricted_guest = 1; +module_param_named(unrestricted_guest,
+ enable_unrestricted_guest, bool, S_IRUGO);
+
static int __read_mostly emulate_invalid_guest_state = 0; module_param(emulate_invalid_guest_state, bool, S_IRUGO);

@@ -268,6 +272,12 @@ static inline int cpu_has_vmx_ept(void)
SECONDARY_EXEC_ENABLE_EPT;
}

+static inline int cpu_has_vmx_unrestricted_guest(void)
+{
+ return vmcs_config.cpu_based_2nd_exec_ctrl &
+ SECONDARY_EXEC_UNRESTRICTED_GUEST;
+}
+
static inline int vm_need_virtualize_apic_accesses(struct kvm *kvm) {
return flexpriority_enabled &&
@@ -731,7 +741,7 @@ static unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu)

static void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) {
- if (vcpu->arch.rmode.active)
+ if (vcpu->arch.rmode.active && !enable_unrestricted_guest)
rflags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM;
vmcs_writel(GUEST_RFLAGS, rflags);
}
@@ -1195,7 +1205,8 @@ static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf)
opt2 = SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES |
SECONDARY_EXEC_WBINVD_EXITING | SECONDARY_EXEC_ENABLE_VPID |
- SECONDARY_EXEC_ENABLE_EPT;
+ SECONDARY_EXEC_ENABLE_EPT |
+ SECONDARY_EXEC_UNRESTRICTED_GUEST;
if (adjust_vmx_controls(min2, opt2,
MSR_IA32_VMX_PROCBASED_CTLS2, &_cpu_based_2nd_exec_control) < 0)
@@ -1325,8 +1336,13 @@ static __init int hardware_setup(void)
if (!cpu_has_vmx_vpid())
enable_vpid = 0;

- if (!cpu_has_vmx_ept())
+ if (!cpu_has_vmx_ept()) {
enable_ept = 0;
+ enable_unrestricted_guest = 0;
+ }
+
+ if (!cpu_has_vmx_unrestricted_guest())
+ enable_unrestricted_guest = 0;

if (!cpu_has_vmx_flexpriority())
flexpriority_enabled = 0;
@@ -1363,9 +1379,17 @@ static void enter_pmode(struct kvm_vcpu *vcpu)
unsigned long flags;
struct vcpu_vmx *vmx = to_vmx(vcpu);

- vmx->emulation_required = 1;
vcpu->arch.rmode.active = 0;

+ update_exception_bitmap(vcpu);
+
+ if (enable_unrestricted_guest) {
+ vmx->emulation_required = 0;
+ return;
+ }
+
+ vmx->emulation_required = 1;
+
vmcs_writel(GUEST_TR_BASE, vcpu->arch.rmode.tr.base); vmcs_write32(GUEST_TR_LIMIT, vcpu->arch.rmode.tr.limit); vmcs_write32(GUEST_TR_AR_BYTES, vcpu->arch.rmode.tr.ar);
@@ -1378,8 +1402,6 @@ static void enter_pmode(struct kvm_vcpu *vcpu)
vmcs_writel(GUEST_CR4, (vmcs_readl(GUEST_CR4) & ~X86_CR4_VME) |
(vmcs_readl(CR4_READ_SHADOW) & X86_CR4_VME));

- update_exception_bitmap(vcpu);
-
if (emulate_invalid_guest_state)
return;

@@ -1425,8 +1447,15 @@ static void enter_rmode(struct kvm_vcpu *vcpu)
unsigned long flags;
struct vcpu_vmx *vmx = to_vmx(vcpu);

- vmx->emulation_required = 1;
vcpu->arch.rmode.active = 1;
+ update_exception_bitmap(vcpu);
+
+ if (enable_unrestricted_guest) {
+ vmx->emulation_required = 0;
+ goto continue_rmode;
+ }
+
+ vmx->emulation_required = 1;

vcpu->arch.rmode.tr.base = vmcs_readl(GUEST_TR_BASE); vmcs_writel(GUEST_TR_BASE, rmode_tss_base(vcpu->kvm));
@@ -1445,7 +1474,6 @@ static void enter_rmode(struct kvm_vcpu *vcpu)

vmcs_writel(GUEST_RFLAGS, flags); vmcs_writel(GUEST_CR4, vmcs_readl(GUEST_CR4) | X86_CR4_VME);
- update_exception_bitmap(vcpu);

if (emulate_invalid_guest_state)
goto continue_rmode;
@@ -1563,7 +1591,6 @@ static void ept_update_paging_mode_cr0(unsigned long *hw_cr0,
CPU_BASED_CR3_STORE_EXITING));
vcpu->arch.cr0 = cr0;
vmx_set_cr4(vcpu, vcpu->arch.cr4);
- *hw_cr0 |= X86_CR0_PE | X86_CR0_PG;
*hw_cr0 &= ~X86_CR0_WP;
} else if (!is_paging(vcpu)) {
/* From nonpaging to paging */
@@ -1733,6 +1760,23 @@ static u32 vmx_segment_access_rights(struct kvm_segment *var)
return ar;
}

+static inline u32 get_segment_ar(int seg)
+{
+ if (!enable_unrestricted_guest)
+ return 0xf3;
+
+ switch (seg) {
+ case VCPU_SREG_CS:
+ return 0x9b;
+ case VCPU_SREG_TR:
+ return 0x8b;
+ case VCPU_SREG_LDTR:
+ return 0x82;
+ default:
+ return 0x93;
+ }
+}
+
static void vmx_set_segment(struct kvm_vcpu *vcpu,
struct kvm_segment *var, int seg)
{
@@ -1755,7 +1799,7 @@ static void vmx_set_segment(struct kvm_vcpu *vcpu,
*/
if (var->base == 0xffff0000 && var->selector == 0xf000)
vmcs_writel(sf->base, 0xf0000);
- ar = 0xf3;
+ ar = get_segment_ar(seg);
} else
ar = vmx_segment_access_rights(var);
vmcs_write32(sf->ar_bytes, ar);
@@ -2058,7 +2102,7 @@ static void seg_setup(int seg)
vmcs_write16(sf->selector, 0);
vmcs_writel(sf->base, 0);
vmcs_write32(sf->limit, 0xffff);
- vmcs_write32(sf->ar_bytes, 0xf3);
+ vmcs_write32(sf->ar_bytes, get_segment_ar(seg));
}

static int alloc_apic_access_page(struct kvm *kvm)
@@ -2201,6 +2245,8 @@ static int vmx_vcpu_setup(struct vcpu_vmx *vmx)
exec_control &= ~SECONDARY_EXEC_ENABLE_VPID;
if (!enable_ept)
exec_control &= ~SECONDARY_EXEC_ENABLE_EPT;
+ if (!enable_unrestricted_guest)
+ exec_control &= ~SECONDARY_EXEC_UNRESTRICTED_GUEST;
vmcs_write32(SECONDARY_VM_EXEC_CONTROL, exec_control);
}

arch/x86/include/asm/kvm_host.h +6 -2
index d2b082d..7832599 100644
@@ -40,9 +40,13 @@
 #define KVM_GUEST_CR0_MASK				   \
 	(X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
 	 | X86_CR0_NW | X86_CR0_CD)
+#define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST				\
+	(X86_CR0_WP | X86_CR0_NE | X86_CR0_TS | X86_CR0_MP)
+#define KVM_VM_CR0_ALWAYS_ON_RESTRICTED_GUEST				\
+	(KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST | X86_CR0_PG | X86_CR0_PE)
 #define KVM_VM_CR0_ALWAYS_ON						\
-	(X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
-	 | X86_CR0_MP)
+	(enable_unrestricted_guest ? KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST \
+	: KVM_VM_CR0_ALWAYS_ON_RESTRICTED_GUEST)
 #define KVM_GUEST_CR4_MASK						\
 	(X86_CR4_VME | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_PGE | X86_CR4_VMXE)
 #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE)
arch/x86/include/asm/vmx.h +1 -0
index 498f944..c73da02 100644
@@ -55,6 +55,7 @@
 #define SECONDARY_EXEC_ENABLE_EPT               0x00000002
 #define SECONDARY_EXEC_ENABLE_VPID              0x00000020
 #define SECONDARY_EXEC_WBINVD_EXITING		0x00000040
+#define SECONDARY_EXEC_UNRESTRICTED_GUEST	0x00000080
 
 
 #define PIN_BASED_EXT_INTR_MASK                 0x00000001
arch/x86/kvm/vmx.c +57 -11

156 lines not on this page · raw message

#1 lore raw

VMX Unrestricted mode support -- looks like very interesting (and useful !) feature.

Which CPUs support it ?

Core i7 900-series (Nehalem) ?

Signature · 1 line
-Alexey Eromenko "Technologov"
#2 lore raw

Re: [patch] VMX Unrestricted guest mode support

Unrestricted guest features is introduced in Westmere processor. Westmere is next to nehalem. And all the following processors will support it. Westmere would be the 1st processor being built on the 32nm process.

Thanks,
Nitin

Quoted 7 lines · Alexey Eremenko

VMX Unrestricted mode support -- looks like very interesting (and useful !) feature.

Which CPUs support it ?

Core i7 900-series (Nehalem) ?

#3 lore raw
In reply to Nitin A Kamble
Quoted 14 lines · Nitin A Kamble

Avi,

A new VMX feature "Unrestricted Guest" feature is added in the VMX specification. You can look at the latest Intel processor manual for details of the feature here:

http://www.intel.com/products/processor/manuals

It allows kvm guests to run real mode and unpaged mode
code natively in the VMX mode when EPT is turned on. With the unrestricted guest there is no need to emulate the guest real mode code in the vm86 container or in the emulator. Also the guest big real mode code works like native.

Quoted 4 lines

The attached patch enhances KVM to use the unrestricted guest feature
if available on the processor. It also adds a new kernel/module parameter to disable the unrestricted guest feature at the boot time.

Thanks, this is much needed. Review comments below.

arch/x86/include/asm/kvm_host.h line 49

Quoted 13 lines

#define KVM_GUEST_CR0_MASK \

arch/x86/include/asm/kvm_host.h +6 -2
 	(X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
 	 | X86_CR0_NW | X86_CR0_CD)
+#define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST				\
+	(X86_CR0_WP | X86_CR0_NE | X86_CR0_TS | X86_CR0_MP)
+#define KVM_VM_CR0_ALWAYS_ON_RESTRICTED_GUEST				\
+	(KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST | X86_CR0_PG | X86_CR0_PE)
 #define KVM_VM_CR0_ALWAYS_ON						\
-	(X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
-	 | X86_CR0_MP)
+	(enable_unrestricted_guest ? KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST \
+	: KVM_VM_CR0_ALWAYS_ON_RESTRICTED_GUEST)

Please avoid hiding computations in macros. Just change the call sites.

Quoted 12 lines

static inline int vm_need_virtualize_apic_accesses(struct kvm *kvm) {

arch/x86/kvm/vmx.c +1 -1
 	return flexpriority_enabled &&
@@ -731,7 +741,7 @@ static unsigned long vmx_get_rflags(struct kvm_vcpu
 *vcpu)
 
 static void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags)
 {
-	if (vcpu->arch.rmode.active)
+	if (vcpu->arch.rmode.active && !enable_unrestricted_guest)
 		rflags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM;
 	vmcs_writel(GUEST_RFLAGS, rflags);
Quoted 2 lines

}

Instead of changing all the checks like this, you can make rmode.active be false when unrestricted guest is enabled. We can interpret rmode.active as "emulating real mode via vm86", not as "guest is in real mode".

You can just have enter_rmode() exit immediately when called.

Quoted 22 lines
arch/x86/kvm/vmx.c +17 -0
+static inline u32 get_segment_ar(int seg)
+{
+	if (!enable_unrestricted_guest)
+		return 0xf3;
+
+	switch (seg) {
+	case VCPU_SREG_CS:
+		return 0x9b;
+	case VCPU_SREG_TR:
+		return 0x8b;
+	case VCPU_SREG_LDTR:
+		return 0x82;
+	default:
+		return 0x93;
+	}
+}
+
 static void vmx_set_segment(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg)
 {

arch/x86/kvm/vmx.c static void vmx_set_segment(struct kvm_vcpu *vcpu, · line 1803

Quoted 6 lines
arch/x86/kvm/vmx.c +1 -1
@@ -1755,7 +1799,7 @@ static void vmx_set_segment(struct kvm_vcpu *vcpu,
 		 */
 		if (var->base == 0xffff0000 && var->selector == 0xf000)
 			vmcs_writel(sf->base, 0xf0000);
-		ar = 0xf3;
+		ar = get_segment_ar(seg);
 

I think this can go away if rmode.active == 0.

Signature · 1 line
error compiling committee.c: too many arguments to function
#4 lore raw
Quoted 5 lines · Avi Kivity

Thanks, this is much needed. Review comments below.

Please avoid hiding computations in macros. Just change the call sites.

Quoted 9 lines

Instead of changing all the checks like this, you can make rmode.active be false when unrestricted guest is enabled. We can interpret rmode.active as "emulating real mode via vm86", not as "guest is in real mode".

You can just have enter_rmode() exit immediately when called.

I think this can go away if rmode.active == 0.

Avi,
Thanks for your comments. I will cook another patch for this feature
as per your comments.

Thanks & Regards,
Nitin

#5 lore raw
Quoted 6 lines · Nitin A Kamble
Quoted 4 lines · Avi Kivity

Instead of changing all the checks like this, you can make rmode.active be false when unrestricted guest is enabled. We can interpret rmode.active as "emulating real mode via vm86", not as "guest is in real mode".

Avi,
How about renaming rmode.active to rmode.vm86_active ?

Thanks & Regards,
Nitin

#6 lore raw
Quoted 3 lines · Nitin A Kamble

Avi,
How about renaming rmode.active to rmode.vm86_active ?

Sure. But if you do that, then do it in a separate patch please.

Signature · 1 line
Do not meddle in the internals of kernels, for they are subtle and quick to panic.
#7 lore raw

Hi Avi,
I find that the qemu processor reset state is not per the IA32
processor specifications. (Sections 8.1.1 of
http://www.intel.com/Assets/PDF/manual/253668.pdf)

In qemu-kvm.git in file target-i386/helper.c in function cpu_reset the segment registers are initialized as follows:

cpu_x86_load_seg_cache(env, R_CS, 0xf000, 0xffff0000, 0xffff,
DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK |
DESC_R_MASK);
cpu_x86_load_seg_cache(env, R_DS, 0, 0, 0xffff,

                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_ES, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_SS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_FS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_GS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);

While the IA32 cpu reset state specification says that Segment Accessed bit is also 1 at the time of cpu reset. so the above code should look like this:

cpu_x86_load_seg_cache(env, R_CS, 0xf000, 0xffff0000, 0xffff,
DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK | DESC_R_MASK | DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_DS, 0, 0, 0xffff,

                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK | DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_ES, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK| DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_SS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK |DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_FS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_GS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);

This discrepancy is adding the need of the following function in the unrestricted guest patch.

+static inline u32 get_segment_ar(int seg)
+{
+ if (!enable_unrestricted_guest)
+ return 0xf3;
+
+ switch (seg) {
+ case VCPU_SREG_CS:
+ return 0x9b;
+ case VCPU_SREG_TR:
+ return 0x8b;
+ case VCPU_SREG_LDTR:
+ return 0x82;
+ default:
+ return 0x93;
+ }
+}
+

For the unrestricted guest support either we can fix this discrepancy in the qemu code, or have a functionality like get_segment_ar() in the kvm vmx code.
what do you suggest ?

Thanks & Regards,
Nitin

#8 lore raw

Avi,
I also tested the qemu change is not affecting vm86 (no unrestricted
guest) code path.

Thanks & Regards,
Nitin

Quoted 74 lines · Nitin A Kamble

Hi Avi,
I find that the qemu processor reset state is not per the IA32
processor specifications. (Sections 8.1.1 of
http://www.intel.com/Assets/PDF/manual/253668.pdf)

In qemu-kvm.git in file target-i386/helper.c in function cpu_reset the segment registers are initialized as follows:

cpu_x86_load_seg_cache(env, R_CS, 0xf000, 0xffff0000, 0xffff,
DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK |
DESC_R_MASK);
cpu_x86_load_seg_cache(env, R_DS, 0, 0, 0xffff,

                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_ES, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_SS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_FS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_GS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);

While the IA32 cpu reset state specification says that Segment Accessed bit is also 1 at the time of cpu reset. so the above code should look like this:

cpu_x86_load_seg_cache(env, R_CS, 0xf000, 0xffff0000, 0xffff,
DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK | DESC_R_MASK | DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_DS, 0, 0, 0xffff,

                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK | DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_ES, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK| DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_SS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK |DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_FS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_GS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);

This discrepancy is adding the need of the following function in the unrestricted guest patch.

arch/x86/kvm/vmx.c +17 -0
+static inline u32 get_segment_ar(int seg)
+{
+       if (!enable_unrestricted_guest)
+               return 0xf3;
+
+       switch (seg) {
+       case VCPU_SREG_CS:
+               return 0x9b;
+       case VCPU_SREG_TR:
+               return 0x8b;
+       case VCPU_SREG_LDTR:
+               return 0x82;
+       default:
+               return 0x93;
+       }
+}
+

For the unrestricted guest support either we can fix this discrepancy in the qemu code, or have a functionality like get_segment_ar() in the kvm vmx code.
what do you suggest ?

Thanks & Regards,
Nitin

--
To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#9 lore raw
In reply to Nitin A Kamble
Quoted 61 lines · Nitin A Kamble

Hi Avi,
I find that the qemu processor reset state is not per the IA32
processor specifications. (Sections 8.1.1 of
http://www.intel.com/Assets/PDF/manual/253668.pdf)

In qemu-kvm.git in file target-i386/helper.c in function cpu_reset the segment registers are initialized as follows:

cpu_x86_load_seg_cache(env, R_CS, 0xf000, 0xffff0000, 0xffff,
DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK |
DESC_R_MASK);
cpu_x86_load_seg_cache(env, R_DS, 0, 0, 0xffff,

                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_ES, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_SS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_FS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_GS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);

While the IA32 cpu reset state specification says that Segment Accessed bit is also 1 at the time of cpu reset. so the above code should look like this:

cpu_x86_load_seg_cache(env, R_CS, 0xf000, 0xffff0000, 0xffff,
DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK | DESC_R_MASK | DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_DS, 0, 0, 0xffff,

                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK | DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_ES, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK| DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_SS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK |DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_FS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_GS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);

This discrepancy is adding the need of the following function in the unrestricted guest patch.

arch/x86/kvm/vmx.c +17 -0
+static inline u32 get_segment_ar(int seg)
+{
+       if (!enable_unrestricted_guest)
+               return 0xf3;
+
+       switch (seg) {
+       case VCPU_SREG_CS:
+               return 0x9b;
+       case VCPU_SREG_TR:
+               return 0x8b;
+       case VCPU_SREG_LDTR:
+               return 0x82;
+       default:
+               return 0x93;
+       }
+}
+
Quoted 6 lines

For the unrestricted guest support either we can fix this discrepancy in the qemu code, or have a functionality like get_segment_ar() in the kvm vmx code.
what do you suggest ?

Qemu should be fixed of course, but we want kvm to keep working with older qemu. So please also have KVM_SET_SREGS set the A bit on segments which are not unusable.

Signature · 2 lines
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.
#10 lore raw
In reply to Nitin A Kamble
Quoted 40 lines · Nitin A Kamble

Hi Avi,
I find that the qemu processor reset state is not per the IA32
processor specifications. (Sections 8.1.1 of
http://www.intel.com/Assets/PDF/manual/253668.pdf)

In qemu-kvm.git in file target-i386/helper.c in function cpu_reset the segment registers are initialized as follows:

cpu_x86_load_seg_cache(env, R_CS, 0xf000, 0xffff0000, 0xffff,
DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK |
DESC_R_MASK);
cpu_x86_load_seg_cache(env, R_DS, 0, 0, 0xffff,

                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_ES, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_SS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_FS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_GS, 0, 0, 0xffff,
                           DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);

While the IA32 cpu reset state specification says that Segment Accessed bit is also 1 at the time of cpu reset. so the above code should look like this:

cpu_x86_load_seg_cache(env, R_CS, 0xf000, 0xffff0000, 0xffff,
DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK | DESC_R_MASK | DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_DS, 0, 0, 0xffff,

                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK | DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_ES, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK| DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_SS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK |DESC_A_MASK);
cpu_x86_load_seg_cache(env, R_FS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
cpu_x86_load_seg_cache(env, R_GS, 0, 0, 0xffff,
                 DESC_P_MASK | DESC_S_MASK | DESC_W_MASK);
Quoted 2 lines

This discrepancy is adding the need of the following function in the unrestricted guest patch.

As Avi already indicated: Independent of the kvm workaround for older qemu versions, please post (to qemu-devel) a patch against upstream's git to fix the discrepancy.

Jan

Signature · 2 lines
Siemens AG, Corporate Technology, CT SE 2
Corporate Competence Center Embedded Linux
#11 lore raw
In reply to Nitin A Kamble
Quoted 14 lines · Nitin A Kamble

Avi,

A new VMX feature "Unrestricted Guest" feature is added in the VMX specification. You can look at the latest Intel processor manual for details of the feature here:

http://www.intel.com/products/processor/manuals

It allows kvm guests to run real mode and unpaged mode
code natively in the VMX mode when EPT is turned on. With the unrestricted guest there is no need to emulate the guest real mode code in the vm86 container or in the emulator. Also the guest big real mode code works like native.

Quoted 9 lines

The attached patch enhances KVM to use the unrestricted guest feature
if available on the processor. It also adds a new kernel/module parameter to disable the unrestricted guest feature at the boot time.

Thanks & Regards,
Nitin

It may make sense to expose this as a CAP down to userspace.

We could possibly make better error messages in userspace when we get an unknown vmexit failure on pre-wsm hardware.

Regards,

Anthony Liguori

#12 lore raw
Quoted 27 lines · Anthony Liguori
Quoted 14 lines · Nitin A Kamble

Avi,

A new VMX feature "Unrestricted Guest" feature is added in the VMX specification. You can look at the latest Intel processor manual for details of the feature here:

http://www.intel.com/products/processor/manuals

It allows kvm guests to run real mode and unpaged mode
code natively in the VMX mode when EPT is turned on. With the unrestricted guest there is no need to emulate the guest real mode code in the vm86 container or in the emulator. Also the guest big real mode code works like native.

Quoted 10 lines

The attached patch enhances KVM to use the unrestricted guest feature
if available on the processor. It also adds a new kernel/module parameter to disable the unrestricted guest feature at the boot time.

Thanks & Regards,
Nitin

Quoted 5 lines

It may make sense to expose this as a CAP down to userspace.

We could possibly make better error messages in userspace when we get an unknown vmexit failure on pre-wsm hardware.

If we see the dreaded 0x80000021 and CR0.PE=0, then we're likely seeing a real mode emulation problem.

Signature · 1 line
error compiling committee.c: too many arguments to function