On Wed, Dec 18, 2019 at 12:39:43PM +0100, Christian Borntraeger wrote:

On 17.12.19 21:40, Sean Christopherson wrote:

quoted

Add a KVM selftest to test moving the base gfn of a userspace memory
region.  The test is primarily targeted at x86 to verify its memslot
metadata is correctly updated, but also provides basic functionality
coverage on other architectures.
+static void *vcpu_worker(void *data)
+{
+	struct kvm_vm *vm = data;
+	struct kvm_run *run;
+	struct ucall uc;
+	uint64_t cmd;
+
+	/*
+	 * Loop until the guest is done.  Re-enter the guest on all MMIO exits,
+	 * which will occur if the guest attempts to access a memslot while it
+	 * is being moved.
+	 */
+	run = vcpu_state(vm, VCPU_ID);
+	do {
+		vcpu_run(vm, VCPU_ID);
+	} while (run->exit_reason == KVM_EXIT_MMIO);
+
+	TEST_ASSERT(run->exit_reason == KVM_EXIT_IO,
+		    "Unexpected exit reason = %d", run->exit_reason);

This will also not work for s390. Maybe just make this test x86 specific for now?

Doh, that's obvious in hindsight.  I think the basic premise is also
broken on arm64 as it returns -EFAULT on is_error_noslot_pfn(pfn).  So
yeah, x86 only for now :-(