On 30/10/2025 17:18, Brendan Jackman wrote:
On Wed Sep 24, 2025 at 3:22 PM UTC, Patrick Roy wrote:
Add a selftest that loads itself into guest_memfd (via
GUEST_MEMFD_FLAG_MMAP) and triggers an MMIO exit when executed. This
exercises x86 MMIO emulation code inside KVM for guest_memfd-backed
memslots where the guest_memfd folios are direct map removed.
Particularly, it validates that x86 MMIO emulation code (guest page
table walks + instruction fetch) correctly accesses gmem through the VMA
that's been reflected into the memslot's userspace_addr field (instead
of trying to do direct map accesses).
Signed-off-by: Patrick Roy <[email protected]>
---
.../selftests/kvm/set_memory_region_test.c | 50 +++++++++++++++++--
1 file changed, 46 insertions(+), 4 deletions(-)
diff --git a/tools/testing/selftests/kvm/set_memory_region_test.c
b/tools/testing/selftests/kvm/set_memory_region_test.c
index ce3ac0fd6dfb..cb3bc642d376 100644
--- a/tools/testing/selftests/kvm/set_memory_region_test.c
+++ b/tools/testing/selftests/kvm/set_memory_region_test.c
@@ -603,6 +603,41 @@ static void test_mmio_during_vectoring(void)
kvm_vm_free(vm);
}
+
+static void guest_code_trigger_mmio(void)
+{
+ /*
+ * Read some GPA that is not backed by a memslot. KVM consider this
+ * as MMIO and tell userspace to emulate the read.
+ */
+ READ_ONCE(*((uint64_t *)MEM_REGION_GPA));
+
+ GUEST_DONE();
+}
+
+static void test_guest_memfd_mmio(void)
+{
+ struct kvm_vm *vm;
+ struct kvm_vcpu *vcpu;
+ struct vm_shape shape = {
+ .mode = VM_MODE_DEFAULT,
+ .src_type = VM_MEM_SRC_GUEST_MEMFD_NO_DIRECT_MAP,
+ };
+ pthread_t vcpu_thread;
+
+ pr_info("Testing MMIO emulation for instructions in gmem\n");
+
+ vm = __vm_create_shape_with_one_vcpu(shape, &vcpu, 0,
guest_code_trigger_mmio);
When I run this test on my minimal config in a nested VM I get:
[root@testvm:~]#
/nix/store/xlxd60n7v1qfr6s5zxda410zrzdd0xc2-kselftests/bin/run_kselftest.sh -t
kvm:set_memory_region_test
TAP version 13
1..1
# timeout set to 120
# selftests: kvm: set_memory_region_test
# Random seed: 0x6b8b4567
# Testing KVM_RUN with zero added memory regions
# Testing MMIO during vectoring error handling
# Allowed number of memory slots: 32764
# Adding slots 0..32763, each memory region with 2048K size
# Testing MMIO emulation for instructions in gmem
# ==== Test Assertion Failure ====
# lib/kvm_util.c:1118: region->mmap_start != MAP_FAILED
# pid=614 tid=614 errno=19 - No such device
# 1 0x0000000000407b02: vm_mem_add at ??:?
# 2 0x000000000040a924: __vm_create at ??:?
# 3 0x000000000040ab16: __vm_create_shape_with_one_vcpu at ??:?
# 4 0x00000000004042cf: main at ??:?
# 5 0x00007faa6b08a47d: ?? ??:0
# 6 0x00007faa6b08a538: ?? ??:0
# 7 0x0000000000404384: _start at ??:?
# mmap() failed, rc: -1 errno: 19 (No such device)
Here's the kconfig I'm using (basically defconfig+KVM):
https://gist.githubusercontent.com/bjackman/4ea941ef5072606769211f3b000c8ed7/raw/73808882ddae6ff2ae8a0be85ac977b2980f7292/kconfig.txt
Sorry I'm too ignorant about KVM to know what I'm missing here but I
guess it's a missing TEST_REQUIRE()?
Hi Brendan,
I ran the set_memory_region_test on the v8 [1] with your config with
minor modifications (I had to enable XFS to be able to mount my rootfs
and THP as the test appeared to depend on it), and was not able to
reproduce the failure. Please let me know whether the v8 still fails
for you if you have time to try it.
[1] https://lore.kernel.org/kvm/[email protected]
Nikita
