Excise the completely bogus testcase in test_host_addr_size() which
purports to verify that setting vmcs.HOST_RIP[63:32] to non-zero values is
allowed for 64-bit hosts. The testcase is mindbogglingly broken: setting
arbitrary, single bits above bit 46 creates a noncanonical address, and
setting arbitrary bits below bit 47 would send the test into the weeds as
a "successful" VMLAUNCH generates a VM-Exit, i.e. would load the garbage
RIP and immediately encounter a #PF.
The only reason the passes is because it does absolutely nothing useful:
vmlaunch() unconditionally writes HOST_RIP before VMLAUNCH, because not
jumping to a random RIP on a VM-Exit is mildly important.
Outright drop the testcase, trying to salvage anything from it would be a
waste of time as simply running any 64-bit guest will generate a huge
variety of RIPs with non-zero values in bits 63:32.
Signed-off-by: Sean Christopherson <seanjc@xxxxxxxxxx>
---
x86/vmx_tests.c | 10 ----------
1 file changed, 10 deletions(-)
diff --git a/x86/vmx_tests.c b/x86/vmx_tests.c
index 1a340242..9d0f2050 100644
--- a/x86/vmx_tests.c
+++ b/x86/vmx_tests.c
@@ -7616,8 +7616,6 @@ static void test_host_addr_size(void)
u64 cr4_saved = vmcs_read(HOST_CR4);
u64 rip_saved = vmcs_read(HOST_RIP);
u64 entry_ctrl_saved = vmcs_read(ENT_CONTROLS);
- int i;
- u64 tmp;
assert(vmcs_read(EXI_CONTROLS) & EXI_HOST_64);
assert(cr4_saved & X86_CR4_PAE);
@@ -7640,14 +7638,6 @@ static void test_host_addr_size(void)
vmcs_write(HOST_CR4, cr4_saved);
report_prefix_pop();
- for (i = 32; i <= 63; i = i + 4) {
- tmp = rip_saved | 1ull << i;
- vmcs_write(HOST_RIP, tmp);
- report_prefix_pushf("HOST_RIP %lx", tmp);
- test_vmx_vmlaunch(0);
- report_prefix_pop();
- }
-
vmcs_write(HOST_RIP, NONCANONICAL);
report_prefix_pushf("HOST_RIP %llx", NONCANONICAL);
test_vmx_vmlaunch(VMXERR_ENTRY_INVALID_HOST_STATE_FIELD);
--
2.42.0.283.g2d96d420d3-goog
