KVM: selftests: Automatically do init_ucall() for non-barebones VMs

Do init_ucall() automatically during VM creation to kill two (three?)
birds with one stone.

First, initializing ucall immediately after VM creations allows forcing
aarch64's MMIO ucall address to immediately follow memslot0. This is
still somewhat fragile as tests could clobber the MMIO address with a
new memslot, but it's safe-ish since tests have to be conversative when
accounting for memslot0. And this can be hardened in the future by
creating a read-only memslot for the MMIO page (KVM ARM exits with MMIO
if the guest writes to a read-only memslot). Add a TODO to document that
selftests can and should use a memslot for the ucall MMIO (doing so
requires yet more rework because tests assumes thay can use all memslots
except memslot0).

Second, initializing ucall for all VMs prepares for making ucall
initialization meaningful on all architectures. aarch64 is currently the
only arch that needs to do any setup, but that will change in the future
by switching to a pool-based implementation (instead of the current
stack-based approach).

Lastly, defining the ucall MMIO address from common code will simplify
switching all architectures (except s390) to a common MMIO-based ucall
implementation (if there's ever sufficient motivation to do so).

Cc: Oliver Upton <[e-mail]>
Reviewed-by: Andrew Jones <[e-mail]>
Tested-by: Peter Gonda <[e-mail]>
Signed-off-by: Sean Christopherson <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]

tools/testing/selftests/kvm/aarch64/aarch32_id_regs.c

···
 
 	TEST_REQUIRE(vcpu_aarch64_only(vcpu));
 
-	ucall_init(vm, NULL);
-
 	test_user_raz_wi(vcpu);
 	test_user_raz_invariant(vcpu);
 	test_guest_raz(vcpu);

tools/testing/selftests/kvm/aarch64/arch_timer.c

···
 	for (i = 0; i < nr_vcpus; i++)
 		vcpu_init_descriptor_tables(vcpus[i]);
 
-	ucall_init(vm, NULL);
 	test_init_timer_irq(vm);
 	gic_fd = vgic_v3_setup(vm, nr_vcpus, 64, GICD_BASE_GPA, GICR_BASE_GPA);
 	__TEST_REQUIRE(gic_fd >= 0, "Failed to create vgic-v3");

tools/testing/selftests/kvm/aarch64/debug-exceptions.c

···
 	int stage;
 
 	vm = vm_create_with_one_vcpu(&vcpu, guest_code);
-	ucall_init(vm, NULL);
 
 	vm_init_descriptor_tables(vm);
 	vcpu_init_descriptor_tables(vcpu);
···
 	struct kvm_guest_debug debug = {};
 
 	vm = vm_create_with_one_vcpu(&vcpu, guest_code_ss);
-	ucall_init(vm, NULL);
 	run = vcpu->run;
 	vcpu_args_set(vcpu, 1, test_cnt);
 

tools/testing/selftests/kvm/aarch64/hypercalls.c

···
 
 	vm = vm_create_with_one_vcpu(vcpu, guest_code);
 
-	ucall_init(vm, NULL);
 	steal_time_init(*vcpu);
 
 	return vm;

tools/testing/selftests/kvm/aarch64/psci_test.c

···
 	struct kvm_vm *vm;
 
 	vm = vm_create(2);
-	ucall_init(vm, NULL);
 
 	vm_ioctl(vm, KVM_ARM_PREFERRED_TARGET, &init);
 	init.features[0] |= (1 << KVM_ARM_VCPU_PSCI_0_2);

tools/testing/selftests/kvm/aarch64/vgic_init.c

···
 /* we don't want to assert on run execution, hence that helper */
 static int run_vcpu(struct kvm_vcpu *vcpu)
 {
-	ucall_init(vcpu->vm, NULL);
-
 	return __vcpu_run(vcpu) ? -errno : 0;
 }
 

tools/testing/selftests/kvm/aarch64/vgic_irq.c

···
 	print_args(&args);
 
 	vm = vm_create_with_one_vcpu(&vcpu, guest_code);
-	ucall_init(vm, NULL);
 
 	vm_init_descriptor_tables(vm);
 	vcpu_init_descriptor_tables(vcpu);

tools/testing/selftests/kvm/dirty_log_test.c

···
 	/* Cache the HVA pointer of the region */
 	host_test_mem = addr_gpa2hva(vm, (vm_paddr_t)guest_test_phys_mem);
 
-	ucall_init(vm, NULL);
-
 	/* Export the shared variables to the guest */
 	sync_global_to_guest(vm, host_page_size);
 	sync_global_to_guest(vm, guest_page_size);

tools/testing/selftests/kvm/include/ucall_common.h

···
 	uint64_t args[UCALL_MAX_ARGS];
 };
 
-void ucall_arch_init(struct kvm_vm *vm, void *arg);
+void ucall_arch_init(struct kvm_vm *vm, vm_paddr_t mmio_gpa);
 void ucall_arch_uninit(struct kvm_vm *vm);
 void ucall_arch_do_ucall(vm_vaddr_t uc);
 void *ucall_arch_get_ucall(struct kvm_vcpu *vcpu);
···
 void ucall(uint64_t cmd, int nargs, ...);
 uint64_t get_ucall(struct kvm_vcpu *vcpu, struct ucall *uc);
 
-static inline void ucall_init(struct kvm_vm *vm, void *arg)
+static inline void ucall_init(struct kvm_vm *vm, vm_paddr_t mmio_gpa)
 {
-	ucall_arch_init(vm, arg);
+	ucall_arch_init(vm, mmio_gpa);
 }
 
 static inline void ucall_uninit(struct kvm_vm *vm)

tools/testing/selftests/kvm/kvm_page_table_test.c

···
 	host_test_mem = addr_gpa2hva(vm, (vm_paddr_t)guest_test_phys_mem);
 
 	/* Export shared structure test_args to guest */
-	ucall_init(vm, NULL);
 	sync_global_to_guest(vm, test_args);
 
 	ret = sem_init(&test_stage_updated, 0, 0);

tools/testing/selftests/kvm/lib/aarch64/ucall.c

···
 
 static vm_vaddr_t *ucall_exit_mmio_addr;
 
-static bool ucall_mmio_init(struct kvm_vm *vm, vm_paddr_t gpa)
+void ucall_arch_init(struct kvm_vm *vm, vm_paddr_t mmio_gpa)
 {
-	if (kvm_userspace_memory_region_find(vm, gpa, gpa + 1))
-		return false;
+	virt_pg_map(vm, mmio_gpa, mmio_gpa);
 
-	virt_pg_map(vm, gpa, gpa);
-
-	ucall_exit_mmio_addr = (vm_vaddr_t *)gpa;
+	ucall_exit_mmio_addr = (vm_vaddr_t *)mmio_gpa;
 	sync_global_to_guest(vm, ucall_exit_mmio_addr);
-
-	return true;
-}
-
-void ucall_arch_init(struct kvm_vm *vm, void *arg)
-{
-	vm_paddr_t gpa, start, end, step, offset;
-	unsigned int bits;
-	bool ret;
-
-	if (arg) {
-		gpa = (vm_paddr_t)arg;
-		ret = ucall_mmio_init(vm, gpa);
-		TEST_ASSERT(ret, "Can't set ucall mmio address to %lx", gpa);
-		return;
-	}
-
-	/*
-	 * Find an address within the allowed physical and virtual address
-	 * spaces, that does _not_ have a KVM memory region associated with
-	 * it. Identity mapping an address like this allows the guest to
-	 * access it, but as KVM doesn't know what to do with it, it
-	 * will assume it's something userspace handles and exit with
-	 * KVM_EXIT_MMIO. Well, at least that's how it works for AArch64.
-	 * Here we start with a guess that the addresses around 5/8th
-	 * of the allowed space are unmapped and then work both down and
-	 * up from there in 1/16th allowed space sized steps.
-	 *
-	 * Note, we need to use VA-bits - 1 when calculating the allowed
-	 * virtual address space for an identity mapping because the upper
-	 * half of the virtual address space is the two's complement of the
-	 * lower and won't match physical addresses.
-	 */
-	bits = vm->va_bits - 1;
-	bits = min(vm->pa_bits, bits);
-	end = 1ul << bits;
-	start = end * 5 / 8;
-	step = end / 16;
-	for (offset = 0; offset < end - start; offset += step) {
-		if (ucall_mmio_init(vm, start - offset))
-			return;
-		if (ucall_mmio_init(vm, start + offset))
-			return;
-	}
-	TEST_FAIL("Can't find a ucall mmio address");
 }
 
 void ucall_arch_uninit(struct kvm_vm *vm)

tools/testing/selftests/kvm/lib/kvm_util.c

···
 {
 	uint64_t nr_pages = vm_nr_pages_required(mode, nr_runnable_vcpus,
 						 nr_extra_pages);
+	struct userspace_mem_region *slot0;
 	struct kvm_vm *vm;
 
 	vm = ____vm_create(mode, nr_pages);
 
 	kvm_vm_elf_load(vm, program_invocation_name);
 
+	/*
+	 * TODO: Add proper defines to protect the library's memslots, and then
+	 * carve out memslot1 for the ucall MMIO address.  KVM treats writes to
+	 * read-only memslots as MMIO, and creating a read-only memslot for the
+	 * MMIO region would prevent silently clobbering the MMIO region.
+	 */
+	slot0 = memslot2region(vm, 0);
+	ucall_init(vm, slot0->region.guest_phys_addr + slot0->region.memory_size);
+
 #ifdef __x86_64__
 	vm_create_irqchip(vm);
 #endif
+
 	return vm;
 }
 

tools/testing/selftests/kvm/lib/memstress.c

···
 		memstress_setup_nested(vm, nr_vcpus, vcpus);
 	}
 
-	ucall_init(vm, NULL);
-
 	/* Export the shared variables to the guest. */
 	sync_global_to_guest(vm, memstress_args);
 

tools/testing/selftests/kvm/lib/riscv/ucall.c

···
 #include "kvm_util.h"
 #include "processor.h"
 
-void ucall_arch_init(struct kvm_vm *vm, void *arg)
+void ucall_arch_init(struct kvm_vm *vm, vm_paddr_t mmio_gpa)
 {
 }
 

tools/testing/selftests/kvm/lib/s390x/ucall.c

···
  */
 #include "kvm_util.h"
 
-void ucall_arch_init(struct kvm_vm *vm, void *arg)
+void ucall_arch_init(struct kvm_vm *vm, vm_paddr_t mmio_gpa)
 {
 }
 

tools/testing/selftests/kvm/lib/x86_64/ucall.c

···
 
 #define UCALL_PIO_PORT ((uint16_t)0x1000)
 
-void ucall_arch_init(struct kvm_vm *vm, void *arg)
+void ucall_arch_init(struct kvm_vm *vm, vm_paddr_t mmio_gpa)
 {
 }
 

tools/testing/selftests/kvm/memslot_perf_test.c

···
 	TEST_ASSERT(data->hva_slots, "malloc() fail");
 
 	data->vm = __vm_create_with_one_vcpu(&data->vcpu, mempages, guest_code);
-	ucall_init(data->vm, NULL);
 
 	pr_info_v("Adding slots 1..%i, each slot with %"PRIu64" pages + %"PRIu64" extra pages last\n",
 		max_mem_slots - 1, data->pages_per_slot, rempages);

tools/testing/selftests/kvm/rseq_test.c

···
 	 * CPU affinity.
 	 */
 	vm = vm_create_with_one_vcpu(&vcpu, guest_code);
-	ucall_init(vm, NULL);
 
 	pthread_create(&migration_thread, NULL, migration_worker,
 		       (void *)(unsigned long)syscall(SYS_gettid));

tools/testing/selftests/kvm/steal_time.c

···
 	gpages = vm_calc_num_guest_pages(VM_MODE_DEFAULT, STEAL_TIME_SIZE * NR_VCPUS);
 	vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS, ST_GPA_BASE, 1, gpages, 0);
 	virt_map(vm, ST_GPA_BASE, ST_GPA_BASE, gpages);
-	ucall_init(vm, NULL);
 
 	TEST_REQUIRE(is_steal_time_supported(vcpus[0]));
 

tools/testing/selftests/kvm/system_counter_offset_test.c

···
 
 	vm = vm_create_with_one_vcpu(&vcpu, guest_main);
 	check_preconditions(vcpu);
-	ucall_init(vm, NULL);
 
 	enter_guest(vcpu);
 	kvm_vm_free(vm);