Skip to contentSkip to conversation

vmx: Fix EPT accessed and dirty flag test

PATCH Superseded Cathy Avery 2 participants 2 messages Jan 21, 2022 KVM

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 vmx: Fix EPT accessed and dirty flag test

If ept_ad is not supported by the processor or has been turned off via kvm module param, test_ept_eptp() will incorrectly leave EPTP_AD_FLAG set in variable eptp causing the following failures of subsequent test_vmx_valid_controls calls:

FAIL: Enable-EPT enabled; reserved bits [11:7] 0: vmlaunch succeeds
FAIL: Enable-EPT enabled; reserved bits [63:N] 0: vmlaunch succeeds

---
 x86/vmx_tests.c | 1 +
 1 file changed, 1 insertion(+)
x86/vmx_tests.c +1 -0
index 3d57ed6..54f2aaa 100644
@@ -4783,6 +4783,7 @@ static void test_ept_eptp(void)
 
 		eptp |= EPTP_AD_FLAG;
 		test_eptp_ad_bit(eptp, false);
+		eptp &= ~EPTP_AD_FLAG;
 	}
 
 	/*
Signature · 1 line
2.31.1
#1 lore raw
Quoted 19 lines · Cathy Avery

If ept_ad is not supported by the processor or has been turned off via kvm module param, test_ept_eptp() will incorrectly leave EPTP_AD_FLAG set in variable eptp causing the following failures of subsequent test_vmx_valid_controls calls:

FAIL: Enable-EPT enabled; reserved bits [11:7] 0: vmlaunch succeeds
FAIL: Enable-EPT enabled; reserved bits [63:N] 0: vmlaunch succeeds

---
 x86/vmx_tests.c | 1 +
 1 file changed, 1 insertion(+)
x86/vmx_tests.c +0 -0
index 3d57ed6..54f2aaa 100644

x86/vmx_tests.c static void test_ept_eptp(void) · line 4787

Quoted 5 lines
x86/vmx_tests.c +1 -0
@@ -4783,6 +4783,7 @@ static void test_ept_eptp(void)
 
 		eptp |= EPTP_AD_FLAG;
 		test_eptp_ad_bit(eptp, false);
+		eptp &= ~EPTP_AD_FLAG;
 	}

Heh, or we could get cute and do:

x86/vmx_tests.c +3 -3
index e67eaea..9a8f7c2 100644
@@ -4785,11 +4785,11 @@ static void test_ept_eptp(void)
                test_eptp_ad_bit(eptp, true);
        } else {
                report_info("Processor does not supports accessed and dirty flag");
-               eptp &= ~EPTP_AD_FLAG;
-               test_eptp_ad_bit(eptp, true);
-
                eptp |= EPTP_AD_FLAG;
                test_eptp_ad_bit(eptp, false);
+
+               eptp &= ~EPTP_AD_FLAG;
+               test_eptp_ad_bit(eptp, true);
        }
 
        /*

More seriously, I would much prefer we use eptp_saved to restore the known good eptp instead of manually clearing the bits that were set. Does this work?

---
 x86/vmx_tests.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)
x86/vmx_tests.c +5 -7
index e67eaea..8116b0c 100644
@@ -4749,8 +4749,7 @@ static void test_ept_eptp(void)
 			test_vmx_invalid_controls();
 		report_prefix_pop();
 	}
-
-	eptp = (eptp & ~EPT_MEM_TYPE_MASK) | 6ul;
+	eptp = eptp_saved;
 
 	/*
 	 * Page walk length (bits 5:3).  Note, the value in VMCS.EPTP "is 1
@@ -4769,9 +4768,7 @@ static void test_ept_eptp(void)
 			test_vmx_invalid_controls();
 		report_prefix_pop();
 	}
-
-	eptp = (eptp & ~EPTP_PG_WALK_LEN_MASK) |
-	    3ul << EPTP_PG_WALK_LEN_SHIFT;
+	eptp = eptp_saved;
 
 	/*
 	 * Accessed and dirty flag (bit 6)
@@ -4791,6 +4788,7 @@ static void test_ept_eptp(void)
 		eptp |= EPTP_AD_FLAG;
 		test_eptp_ad_bit(eptp, false);
 	}
+	eptp = eptp_saved;
 
 	/*
 	 * Reserved bits [11:7] and [63:N]
@@ -4809,8 +4807,7 @@ static void test_ept_eptp(void)
 			test_vmx_invalid_controls();
 		report_prefix_pop();
 	}
-
-	eptp = (eptp & ~(EPTP_RESERV_BITS_MASK << EPTP_RESERV_BITS_SHIFT));
+	eptp = eptp_saved;
 
 	maxphysaddr = cpuid_maxphyaddr();
 	for (i = 0; i < (63 - maxphysaddr + 1); i++) {
@@ -4829,6 +4826,7 @@ static void test_ept_eptp(void)
 			test_vmx_invalid_controls();
 		report_prefix_pop();
 	}
+	eptp = eptp_saved;
 
 	secondary &= ~(CPU_EPT | CPU_URG);
 	vmcs_write(CPU_EXEC_CTRL1, secondary);

--