Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:

 - Fix for compilation of selftests on non-x86 architectures

 - Fix for kvm_run->if_flag on SEV-ES

 - Fix for page table use-after-free if yielding during exit_mm()

 - Improve behavior when userspace starts a nested guest with invalid
   state

 - Fix missed wakeup with assigned devices but no VT-d posted interrupts

 - Do not tell userspace to save/restore an unsupported PMU MSR

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: VMX: Wake vCPU when delivering posted IRQ even if vCPU == this vCPU
  KVM: selftests: Add test to verify TRIPLE_FAULT on invalid L2 guest state
  KVM: VMX: Fix stale docs for kvm-intel.emulate_invalid_guest_state
  KVM: nVMX: Synthesize TRIPLE_FAULT for L2 if emulation is required
  KVM: VMX: Always clear vmx->fail on emulation_required
  selftests: KVM: Fix non-x86 compiling
  KVM: x86: Always set kvm_run->if_flag
  KVM: x86/mmu: Don't advance iterator after restart due to yielding
  KVM: x86: remove PMU FIXED_CTR3 from msrs_to_save_all

5 files changed, 113 insertions, 9 deletions

diff --git a/tools/testing/selftests/kvm/.gitignore b/tools/testing/selftests/kvm/.gitignore
index 00814c0f87a6..3cb5ac5da087 100644
--- a/tools/testing/selftests/kvm/.gitignore
+++ b/tools/testing/selftests/kvm/.gitignore
@@ -35,6 +35,7 @@
 /x86_64/vmx_apic_access_test
 /x86_64/vmx_close_while_nested_test
 /x86_64/vmx_dirty_log_test
+/x86_64/vmx_invalid_nested_guest_state
 /x86_64/vmx_preemption_timer_test
 /x86_64/vmx_set_nested_state_test
 /x86_64/vmx_tsc_adjust_test
diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
index f307c9f61981..17342b575e85 100644
--- a/tools/testing/selftests/kvm/Makefile
+++ b/tools/testing/selftests/kvm/Makefile
@@ -64,6 +64,7 @@ TEST_GEN_PROGS_x86_64 += x86_64/userspace_msr_exit_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_apic_access_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_close_while_nested_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_dirty_log_test
+TEST_GEN_PROGS_x86_64 += x86_64/vmx_invalid_nested_guest_state
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_set_nested_state_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_tsc_adjust_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_nested_tsc_scaling_test
diff --git a/tools/testing/selftests/kvm/include/kvm_util.h b/tools/testing/selftests/kvm/include/kvm_util.h
index da2b702da71a..2d62edc49d67 100644
--- a/tools/testing/selftests/kvm/include/kvm_util.h
+++ b/tools/testing/selftests/kvm/include/kvm_util.h
@@ -71,15 +71,6 @@ enum vm_guest_mode {
 
 #endif
 
-#if defined(__x86_64__)
-unsigned long vm_compute_max_gfn(struct kvm_vm *vm);
-#else
-static inline unsigned long vm_compute_max_gfn(struct kvm_vm *vm)
-{
-	return ((1ULL << vm->pa_bits) >> vm->page_shift) - 1;
-}
-#endif
-
 #define MIN_PAGE_SIZE		(1U << MIN_PAGE_SHIFT)
 #define PTES_PER_MIN_PAGE	ptes_per_page(MIN_PAGE_SIZE)
 
@@ -330,6 +321,7 @@ bool vm_is_unrestricted_guest(struct kvm_vm *vm);
 
 unsigned int vm_get_page_size(struct kvm_vm *vm);
 unsigned int vm_get_page_shift(struct kvm_vm *vm);
+unsigned long vm_compute_max_gfn(struct kvm_vm *vm);
 uint64_t vm_get_max_gfn(struct kvm_vm *vm);
 int vm_get_fd(struct kvm_vm *vm);
 
diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index daf6fdb217a7..53d2b5d04b82 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -2328,6 +2328,11 @@ unsigned int vm_get_page_shift(struct kvm_vm *vm)
 	return vm->page_shift;
 }
 
+unsigned long __attribute__((weak)) vm_compute_max_gfn(struct kvm_vm *vm)
+{
+	return ((1ULL << vm->pa_bits) >> vm->page_shift) - 1;
+}
+
 uint64_t vm_get_max_gfn(struct kvm_vm *vm)
 {
 	return vm->max_gfn;
diff --git a/tools/testing/selftests/kvm/x86_64/vmx_invalid_nested_guest_state.c b/tools/testing/selftests/kvm/x86_64/vmx_invalid_nested_guest_state.c
new file mode 100644
index 000000000000..489fbed4ca6f
--- /dev/null
+++ b/tools/testing/selftests/kvm/x86_64/vmx_invalid_nested_guest_state.c
@@ -0,0 +1,105 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include "test_util.h"
+#include "kvm_util.h"
+#include "processor.h"
+#include "vmx.h"
+
+#include <string.h>
+#include <sys/ioctl.h>
+
+#include "kselftest.h"
+
+#define VCPU_ID	0
+#define ARBITRARY_IO_PORT 0x2000
+
+static struct kvm_vm *vm;
+
+static void l2_guest_code(void)
+{
+	/*
+	 * Generate an exit to L0 userspace, i.e. main(), via I/O to an
+	 * arbitrary port.
+	 */
+	asm volatile("inb %%dx, %%al"
+		     : : [port] "d" (ARBITRARY_IO_PORT) : "rax");
+}
+
+static void l1_guest_code(struct vmx_pages *vmx_pages)
+{
+#define L2_GUEST_STACK_SIZE 64
+	unsigned long l2_guest_stack[L2_GUEST_STACK_SIZE];
+
+	GUEST_ASSERT(prepare_for_vmx_operation(vmx_pages));
+	GUEST_ASSERT(load_vmcs(vmx_pages));
+
+	/* Prepare the VMCS for L2 execution. */
+	prepare_vmcs(vmx_pages, l2_guest_code,
+		     &l2_guest_stack[L2_GUEST_STACK_SIZE]);
+
+	/*
+	 * L2 must be run without unrestricted guest, verify that the selftests
+	 * library hasn't enabled it.  Because KVM selftests jump directly to
+	 * 64-bit mode, unrestricted guest support isn't required.
+	 */
+	GUEST_ASSERT(!(vmreadz(CPU_BASED_VM_EXEC_CONTROL) & CPU_BASED_ACTIVATE_SECONDARY_CONTROLS) ||
+		     !(vmreadz(SECONDARY_VM_EXEC_CONTROL) & SECONDARY_EXEC_UNRESTRICTED_GUEST));
+
+	GUEST_ASSERT(!vmlaunch());
+
+	/* L2 should triple fault after main() stuffs invalid guest state. */
+	GUEST_ASSERT(vmreadz(VM_EXIT_REASON) == EXIT_REASON_TRIPLE_FAULT);
+	GUEST_DONE();
+}
+
+int main(int argc, char *argv[])
+{
+	vm_vaddr_t vmx_pages_gva;
+	struct kvm_sregs sregs;
+	struct kvm_run *run;
+	struct ucall uc;
+
+	nested_vmx_check_supported();
+
+	vm = vm_create_default(VCPU_ID, 0, (void *) l1_guest_code);
+
+	/* Allocate VMX pages and shared descriptors (vmx_pages). */
+	vcpu_alloc_vmx(vm, &vmx_pages_gva);
+	vcpu_args_set(vm, VCPU_ID, 1, vmx_pages_gva);
+
+	vcpu_run(vm, VCPU_ID);
+
+	run = vcpu_state(vm, VCPU_ID);
+
+	/*
+	 * The first exit to L0 userspace should be an I/O access from L2.
+	 * Running L1 should launch L2 without triggering an exit to userspace.
+	 */
+	TEST_ASSERT(run->exit_reason == KVM_EXIT_IO,
+		    "Expected KVM_EXIT_IO, got: %u (%s)\n",
+		    run->exit_reason, exit_reason_str(run->exit_reason));
+
+	TEST_ASSERT(run->io.port == ARBITRARY_IO_PORT,
+		    "Expected IN from port %d from L2, got port %d",
+		    ARBITRARY_IO_PORT, run->io.port);
+
+	/*
+	 * Stuff invalid guest state for L2 by making TR unusuable.  The next
+	 * KVM_RUN should induce a TRIPLE_FAULT in L2 as KVM doesn't support
+	 * emulating invalid guest state for L2.
+	 */
+	memset(&sregs, 0, sizeof(sregs));
+	vcpu_sregs_get(vm, VCPU_ID, &sregs);
+	sregs.tr.unusable = 1;
+	vcpu_sregs_set(vm, VCPU_ID, &sregs);
+
+	vcpu_run(vm, VCPU_ID);
+
+	switch (get_ucall(vm, VCPU_ID, &uc)) {
+	case UCALL_DONE:
+		break;
+	case UCALL_ABORT:
+		TEST_FAIL("%s", (const char *)uc.args[0]);
+	default:
+		TEST_FAIL("Unexpected ucall: %lu", uc.cmd);
+	}
+}