selftests: KVM: Add test to verify KVM doesn't explode on "bad" I/O
authorSean Christopherson <seanjc@google.com>
Mon, 25 Oct 2021 20:13:11 +0000 (13:13 -0700)
committerPaolo Bonzini <pbonzini@redhat.com>
Fri, 10 Dec 2021 14:38:02 +0000 (09:38 -0500)
Add an x86 selftest to verify that KVM doesn't WARN or otherwise explode
if userspace modifies RCX during a userspace exit to handle string I/O.
This is a regression test for a user-triggerable WARN introduced by
commit 3b27de271839 ("KVM: x86: split the two parts of emulator_pio_in").

Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20211025201311.1881846-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
tools/testing/selftests/kvm/.gitignore
tools/testing/selftests/kvm/Makefile
tools/testing/selftests/kvm/x86_64/userspace_io_test.c [new file with mode: 0644]

index 3763105..00814c0 100644 (file)
@@ -30,6 +30,7 @@
 /x86_64/svm_int_ctl_test
 /x86_64/sync_regs_test
 /x86_64/tsc_msrs_test
+/x86_64/userspace_io_test
 /x86_64/userspace_msr_exit_test
 /x86_64/vmx_apic_access_test
 /x86_64/vmx_close_while_nested_test
index c4e3471..f307c9f 100644 (file)
@@ -59,6 +59,7 @@ TEST_GEN_PROGS_x86_64 += x86_64/vmx_preemption_timer_test
 TEST_GEN_PROGS_x86_64 += x86_64/svm_vmcall_test
 TEST_GEN_PROGS_x86_64 += x86_64/svm_int_ctl_test
 TEST_GEN_PROGS_x86_64 += x86_64/sync_regs_test
+TEST_GEN_PROGS_x86_64 += x86_64/userspace_io_test
 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
diff --git a/tools/testing/selftests/kvm/x86_64/userspace_io_test.c b/tools/testing/selftests/kvm/x86_64/userspace_io_test.c
new file mode 100644 (file)
index 0000000..e4bef2e
--- /dev/null
@@ -0,0 +1,114 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ioctl.h>
+
+#include "test_util.h"
+
+#include "kvm_util.h"
+#include "processor.h"
+
+#define VCPU_ID                        1
+
+static void guest_ins_port80(uint8_t *buffer, unsigned int count)
+{
+       unsigned long end;
+
+       if (count == 2)
+               end = (unsigned long)buffer + 1;
+       else
+               end = (unsigned long)buffer + 8192;
+
+       asm volatile("cld; rep; insb" : "+D"(buffer), "+c"(count) : "d"(0x80) : "memory");
+       GUEST_ASSERT_1(count == 0, count);
+       GUEST_ASSERT_2((unsigned long)buffer == end, buffer, end);
+}
+
+static void guest_code(void)
+{
+       uint8_t buffer[8192];
+       int i;
+
+       /*
+        * Special case tests.  main() will adjust RCX 2 => 1 and 3 => 8192 to
+        * test that KVM doesn't explode when userspace modifies the "count" on
+        * a userspace I/O exit.  KVM isn't required to play nice with the I/O
+        * itself as KVM doesn't support manipulating the count, it just needs
+        * to not explode or overflow a buffer.
+        */
+       guest_ins_port80(buffer, 2);
+       guest_ins_port80(buffer, 3);
+
+       /* Verify KVM fills the buffer correctly when not stuffing RCX. */
+       memset(buffer, 0, sizeof(buffer));
+       guest_ins_port80(buffer, 8192);
+       for (i = 0; i < 8192; i++)
+               GUEST_ASSERT_2(buffer[i] == 0xaa, i, buffer[i]);
+
+       GUEST_DONE();
+}
+
+int main(int argc, char *argv[])
+{
+       struct kvm_regs regs;
+       struct kvm_run *run;
+       struct kvm_vm *vm;
+       struct ucall uc;
+       int rc;
+
+       /* Tell stdout not to buffer its content */
+       setbuf(stdout, NULL);
+
+       /* Create VM */
+       vm = vm_create_default(VCPU_ID, 0, guest_code);
+       run = vcpu_state(vm, VCPU_ID);
+
+       memset(&regs, 0, sizeof(regs));
+
+       while (1) {
+               rc = _vcpu_run(vm, VCPU_ID);
+
+               TEST_ASSERT(rc == 0, "vcpu_run failed: %d\n", rc);
+               TEST_ASSERT(run->exit_reason == KVM_EXIT_IO,
+                           "Unexpected exit reason: %u (%s),\n",
+                           run->exit_reason,
+                           exit_reason_str(run->exit_reason));
+
+               if (get_ucall(vm, VCPU_ID, &uc))
+                       break;
+
+               TEST_ASSERT(run->io.port == 0x80,
+                           "Expected I/O at port 0x80, got port 0x%x\n", run->io.port);
+
+               /*
+                * Modify the rep string count in RCX: 2 => 1 and 3 => 8192.
+                * Note, this abuses KVM's batching of rep string I/O to avoid
+                * getting stuck in an infinite loop.  That behavior isn't in
+                * scope from a testing perspective as it's not ABI in any way,
+                * i.e. it really is abusing internal KVM knowledge.
+                */
+               vcpu_regs_get(vm, VCPU_ID, &regs);
+               if (regs.rcx == 2)
+                       regs.rcx = 1;
+               if (regs.rcx == 3)
+                       regs.rcx = 8192;
+               memset((void *)run + run->io.data_offset, 0xaa, 4096);
+               vcpu_regs_set(vm, VCPU_ID, &regs);
+       }
+
+       switch (uc.cmd) {
+       case UCALL_DONE:
+               break;
+       case UCALL_ABORT:
+               TEST_FAIL("%s at %s:%ld : argN+1 = 0x%lx, argN+2 = 0x%lx",
+                         (const char *)uc.args[0], __FILE__, uc.args[1],
+                         uc.args[2], uc.args[3]);
+       default:
+               TEST_FAIL("Unknown ucall %lu", uc.cmd);
+       }
+
+       kvm_vm_free(vm);
+       return 0;
+}