[PATCH 0/4][kvm-unit-test nVMX]: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry control on vmentry of nested guests

[PATCH 0/4][kvm-unit-test nVMX]: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry control on vmentry of nested guests

[Krish Sadhukhan]

This set contains the unit test, and related changes, for the "load
IA32_PERF_GLOBAL_CONTROL" VM-entry control that was enabled in my previous
patchset titled:

	[KVM nVMX]: Check "load IA32_PERF_GLOBAL_CTRL" on vmentry of nested guests


[PATCH 1/4][kvm-unit-test nVMX]: Rename guest_pat_main to guest_state_test_main
[PATCH 2/4][kvm-unit-test nVMX]: Rename report_guest_pat_test to
[PATCH 3/4][kvm-unit-test nVMX]: Add #define for "load IA32_PERF_GLOBAL_CONTROL" bit
[PATCH 4/4][kvm-unit-test nVMX]: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry

 x86/vmx.h       |   1 +
 x86/vmx_tests.c | 108 +++++++++++++++++++++++++++++++++++++++++++-----------
 2 files changed, 87 insertions(+), 22 deletions(-)

Krish Sadhukhan (4):
      nVMX: Rename guest_pat_main to guest_state_test_main
      nVMX: Rename report_guest_pat_test to report_guest_state_test
      nVMX: Add #define for "load IA32_PERF_GLOBAL_CONTROL" bit
      nVMX: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry control on vmentry of nested guests

[PATCH 1/4][kvm-unit-test nVMX]: Rename guest_pat_main to guest_state_test_main

[Krish Sadhukhan]

  ...so that it can be re-used by other tests.

Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Reviewed-by: Karl Heubaum <karl.heubaum@oracle.com>
---
 x86/vmx_tests.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/x86/vmx_tests.c b/x86/vmx_tests.c
index c01fa9d..ee96596 100644
--- a/x86/vmx_tests.c
+++ b/x86/vmx_tests.c
@@ -4995,7 +4995,7 @@ static void test_sysenter_field(u32 field, const char *name)
 	vmcs_write(field, addr_saved);
 }
 
-static void guest_pat_main(void)
+static void guest_state_test_main(void)
 {
 	while (1) {
 		if (vmx_get_test_stage() != 2)
@@ -5097,7 +5097,7 @@ static void test_pat(u32 fld, const char * fld_name, u32 ctrl_fld, u64 ctrl_bit)
 	vmcs_clear_bits(ctrl_fld, ctrl_bit);
 	if (fld == GUEST_PAT) {
 		vmx_set_test_stage(1);
-		test_set_guest(guest_pat_main);
+		test_set_guest(guest_state_test_main);
 	}
 
 	for (i = 0; i < 256; i = (i < PAT_VAL_LIMIT) ? i + 1 : i * 2) {
-- 
2.20.1

[PATCH 2/4][kvm-unit-test nVMX]: Rename report_guest_pat_test to report_guest_state_test

[Krish Sadhukhan]

  ...so that it can be re-used by other tests.

Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Reviewed-by: Karl Heubaum <karl.heubaum@oracle.com>
---
 x86/vmx_tests.c | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/x86/vmx_tests.c b/x86/vmx_tests.c
index ee96596..a339bb3 100644
--- a/x86/vmx_tests.c
+++ b/x86/vmx_tests.c
@@ -5007,13 +5007,14 @@ static void guest_state_test_main(void)
 	asm volatile("fnop");
 }
 
-static void report_guest_pat_test(const char *test, u32 xreason, u64 guest_pat)
+static void report_guest_state_test(const char *test, u32 xreason,
+				    u64 field, const char * field_name)
 {
 	u32 reason = vmcs_read(EXI_REASON);
 	u64 guest_rip;
 	u32 insn_len;
 
-	report("%s, GUEST_PAT %lx", reason == xreason, test, guest_pat);
+	report("%s, %s %lx", reason == xreason, test, field_name, field);
 
 	guest_rip = vmcs_read(GUEST_RIP);
 	insn_len = vmcs_read(EXI_INST_LEN);
@@ -5112,8 +5113,9 @@ static void test_pat(u32 fld, const char * fld_name, u32 ctrl_fld, u64 ctrl_bit)
 
 			} else {	// GUEST_PAT
 				__enter_guest(ABORT_ON_EARLY_VMENTRY_FAIL);
-				report_guest_pat_test("ENT_LOAD_PAT enabled",
-						       VMX_VMCALL, val);
+				report_guest_state_test("ENT_LOAD_PAT enabled",
+							VMX_VMCALL, val,
+							"GUEST_PAT");
 			}
 		}
 	}
@@ -5139,17 +5141,19 @@ static void test_pat(u32 fld, const char * fld_name, u32 ctrl_fld, u64 ctrl_bit)
 			} else {	// GUEST_PAT
 				if (i == 0x2 || i == 0x3 || i == 0x8) {
 					__enter_guest(ABORT_ON_EARLY_VMENTRY_FAIL);
-					report_guest_pat_test("ENT_LOAD_PAT "
-								"enabled",
+					report_guest_state_test("ENT_LOAD_PAT "
+							     "enabled",
 							     VMX_FAIL_STATE |
 							     VMX_ENTRY_FAILURE,
-							     val);
+							     val,
+							     "GUEST_PAT");
 				} else {
 					__enter_guest(ABORT_ON_EARLY_VMENTRY_FAIL);
-					report_guest_pat_test("ENT_LOAD_PAT "
+					report_guest_state_test("ENT_LOAD_PAT "
 							      "enabled",
 							      VMX_VMCALL,
-							      val);
+							      val,
+							      "GUEST_PAT");
 				}
 			}
 
-- 
2.20.1

[PATCH 3/4][kvm-unit-test nVMX]: Add #define for "load IA32_PERF_GLOBAL_CONTROL" bit

[Krish Sadhukhan]

Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Reviewed-by: Karl Heubaum <karl.heubaum@oracle.com>
---
 x86/vmx.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/x86/vmx.h b/x86/vmx.h
index eefd5dc..a4ceac3 100644
--- a/x86/vmx.h
+++ b/x86/vmx.h
@@ -370,6 +370,7 @@ enum Ctrl_exi {
 enum Ctrl_ent {
 	ENT_LOAD_DBGCTLS	= 1UL << 2,
 	ENT_GUEST_64		= 1UL << 9,
+	ENT_LOAD_PERF		= 1UL << 13,
 	ENT_LOAD_PAT		= 1UL << 14,
 	ENT_LOAD_EFER		= 1UL << 15,
 };
-- 
2.20.1

[PATCH 4/4][kvm-unit-test nVMX]: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry control on vmentry of nested guests

[Krish Sadhukhan]

 ..to verify KVM performs the appropriate consistency checks for loading
   IA32_PERF_GLOBAL_CONTROL as part of running a nested guest.

According to section "Checking and Loading Guest State" in Intel SDM
vol 3C, the following check is performed on vmentry:

   If the "load IA32_PERF_GLOBAL_CTRL" VM-exit control is 1, bits reserved
   in the IA32_PERF_GLOBAL_CTRL MSR must be 0 in the field for that register.

Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Reviewed-by: Karl Heubaum <karl.heubaum@oracle.com>
---
 x86/vmx_tests.c | 82 ++++++++++++++++++++++++++++++++++++++++++-------
 1 file changed, 71 insertions(+), 11 deletions(-)

diff --git a/x86/vmx_tests.c b/x86/vmx_tests.c
index a339bb3..65772ce 100644
--- a/x86/vmx_tests.c
+++ b/x86/vmx_tests.c
@@ -5031,29 +5031,69 @@ static void test_perf_global_ctl(u32 field, const char * field_name,
 	u64 perf_global_saved = vmcs_read(field);
 	u64 i, val;
 
+	if (field == GUEST_PERF_GLOBAL_CTRL) {
+		vmx_set_test_stage(1);
+		test_set_guest(guest_state_test_main);
+	}
+
 	vmcs_write(ctrl_field, ctrl_saved & ~ctrl_bit);
 	for (i = 0; i < 64; i++) {
 		val = 1ull << i;
 		vmcs_write(field, val);
-		report_prefix_pushf("\"load IA32_PERF_GLOBAL_CTRL\" "
-		    "VM-exit control is off, HOST_PERF_GLOBAL_CTRL %lx", val);
-		test_vmx_vmlaunch(0, false);
-		report_prefix_pop();
+		if (field == HOST_PERF_GLOBAL_CTRL) {
+			report_prefix_pushf("\"load IA32_PERF_GLOBAL_CTRL\" "
+					    "VM-exit control is off, "
+					    "HOST_PERF_GLOBAL_CTRL %lx", val);
+			test_vmx_vmlaunch(0, false);
+			report_prefix_pop();
+		} else {        // GUEST_PERF_GLOBAL_CTRL
+			enter_guest();
+			report_guest_state_test("ENT_LOAD_PERF disabled",
+						 VMX_VMCALL, val,
+						"GUEST_PERF_GLOBAL_CTRL");
+		}
 	}
 
 	vmcs_write(ctrl_field, ctrl_saved | ctrl_bit);
 	for (i = 0; i < 64; i++) {
 		val = 1ull << i;
 		vmcs_write(field, val);
-		report_prefix_pushf("\"load IA32_PERF_GLOBAL_CTRL\" "
-		    "VM-exit control is on, HOST_PERF_GLOBAL_CTRL %lx", val);
-		if (PERF_GLOBAL_CTRL_VALID_BITS & (1ull << i)) {
-			test_vmx_vmlaunch(0, false);
-		} else {
-			test_vmx_vmlaunch(VMXERR_ENTRY_INVALID_HOST_STATE_FIELD,
-					  false);
+		if (field == HOST_PERF_GLOBAL_CTRL) {
+			report_prefix_pushf("\"load IA32_PERF_GLOBAL_CTRL\" "
+			"VM-exit control is on, HOST_PERF_GLOBAL_CTRL %lx",
+			val);
+			if (PERF_GLOBAL_CTRL_VALID_BITS & (1ull << i)) {
+				test_vmx_vmlaunch(0, false);
+			} else {
+				test_vmx_vmlaunch(
+					VMXERR_ENTRY_INVALID_HOST_STATE_FIELD,
+					false);
 		}
 		report_prefix_pop();
+		} else {        // GUEST_PERF_GLOBAL_CTRL
+			enter_guest();
+			if (PERF_GLOBAL_CTRL_VALID_BITS & (1ull << i)) {
+				report_guest_state_test("ENT_LOAD_PERF enabled",
+						VMX_FAIL_STATE |
+						VMX_ENTRY_FAILURE,
+						val,
+						"GUEST_PERF_GLOBAL_CTRL");
+			} else {
+				report_guest_state_test("ENT_LOAD_PERF enabled",
+						VMX_VMCALL,
+						val,
+						"GUEST_PERF_GLOBAL_CTRL");
+			}
+		}
+	}
+
+	if (field == GUEST_PERF_GLOBAL_CTRL) {
+		/*
+		 * Let the guest finish execution
+		 */
+		vmx_set_test_stage(2);
+		vmcs_write(field, perf_global_saved);
+		enter_guest();
 	}
 
 	vmcs_write(ctrl_field, ctrl_saved);
@@ -5079,6 +5119,25 @@ static void test_host_perf_global_ctl(void)
 			     EXI_CONTROLS, EXI_LOAD_PERF);
 }
 
+/*
+ * If the "load IA32_PERF_GLOBAL_CTRL" VM-entry control is 1, bits reserved
+ * in the IA32_PERF_GLOBAL_CTRL MSR must be 0 in the field for that
+ * register.
+ *
+ *  [Intel SDM]
+ */
+static void test_guest_perf_global_ctl(void)
+{
+	if (!(ctrl_exit_rev.clr & ENT_LOAD_PERF)) {
+		printf("\"load IA32_PERF_GLOBAL_CTRL\" VM-entry control not "
+			"supported\n");
+		return;
+	}
+
+	test_perf_global_ctl(GUEST_PERF_GLOBAL_CTRL, "GUEST_PERF_GLOBAL_CTRL",
+			     ENT_CONTROLS, ENT_LOAD_PERF);
+}
+
 /*
  * PAT values higher than 8 are uninteresting since they're likely lumped
  * in with "8". We only test values above 8 one bit at a time,
@@ -5243,6 +5302,7 @@ static void test_load_guest_pat(void)
  */
 static void vmx_guest_state_area_test(void)
 {
+	test_guest_perf_global_ctl();
 	test_load_guest_pat();
 }
 
-- 
2.20.1

Re: [PATCH 0/4][kvm-unit-test nVMX]: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry control on vmentry of nested guests

[Paolo Bonzini]

On 09/05/19 23:20, Krish Sadhukhan wrote:
> This set contains the unit test, and related changes, for the "load
> IA32_PERF_GLOBAL_CONTROL" VM-entry control that was enabled in my previous
> patchset titled:
> 
> 	[KVM nVMX]: Check "load IA32_PERF_GLOBAL_CTRL" on vmentry of nested guests
> 
> 
> [PATCH 1/4][kvm-unit-test nVMX]: Rename guest_pat_main to guest_state_test_main
> [PATCH 2/4][kvm-unit-test nVMX]: Rename report_guest_pat_test to
> [PATCH 3/4][kvm-unit-test nVMX]: Add #define for "load IA32_PERF_GLOBAL_CONTROL" bit
> [PATCH 4/4][kvm-unit-test nVMX]: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry
> 
>  x86/vmx.h       |   1 +
>  x86/vmx_tests.c | 108 +++++++++++++++++++++++++++++++++++++++++++-----------
>  2 files changed, 87 insertions(+), 22 deletions(-)
> 
> Krish Sadhukhan (4):
>       nVMX: Rename guest_pat_main to guest_state_test_main
>       nVMX: Rename report_guest_pat_test to report_guest_state_test
>       nVMX: Add #define for "load IA32_PERF_GLOBAL_CONTROL" bit
>       nVMX: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry control on vmentry of nested guests
> 

Queued 1-3, but patch 4 does not apply.  It seems like you have another
patch that this sits on top of?

Paolo

Re: [PATCH 0/4][kvm-unit-test nVMX]: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry control on vmentry of nested guests

[Krish Sadhukhan]

On 06/06/2019 05:39 AM, Paolo Bonzini wrote:
> On 09/05/19 23:20, Krish Sadhukhan wrote:
>> This set contains the unit test, and related changes, for the "load
>> IA32_PERF_GLOBAL_CONTROL" VM-entry control that was enabled in my previous
>> patchset titled:
>>
>> 	[KVM nVMX]: Check "load IA32_PERF_GLOBAL_CTRL" on vmentry of nested guests
>>
>>
>> [PATCH 1/4][kvm-unit-test nVMX]: Rename guest_pat_main to guest_state_test_main
>> [PATCH 2/4][kvm-unit-test nVMX]: Rename report_guest_pat_test to
>> [PATCH 3/4][kvm-unit-test nVMX]: Add #define for "load IA32_PERF_GLOBAL_CONTROL" bit
>> [PATCH 4/4][kvm-unit-test nVMX]: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry
>>
>>   x86/vmx.h       |   1 +
>>   x86/vmx_tests.c | 108 +++++++++++++++++++++++++++++++++++++++++++-----------
>>   2 files changed, 87 insertions(+), 22 deletions(-)
>>
>> Krish Sadhukhan (4):
>>        nVMX: Rename guest_pat_main to guest_state_test_main
>>        nVMX: Rename report_guest_pat_test to report_guest_state_test
>>        nVMX: Add #define for "load IA32_PERF_GLOBAL_CONTROL" bit
>>        nVMX: Test "load IA32_PERF_GLOBAL_CONTROL" VM-entry control on vmentry of nested guests
>>
> Queued 1-3, but patch 4 does not apply.  It seems like you have another
> patch that this sits on top of?

Yes, the following patchset:

             [KVM nVMX]: Check "load IA32_PERF_GLOBAL_CTRL" on vmentry 
of nested guests

I need to send out v2 of that set and patch# 4 of this set builds on top 
of that.

You don't need to pull 1-3 out because they are independent of the 
pending patchset. I will attach patch# 4 of this set to the v2 of the 
pending set.

>
> Paolo