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

Pull kvm fixes from Paolo Bonzini:
 "x86:

   - fix lock initialization race in gfn-to-pfn cache (+selftests)

   - fix two refcounting errors

   - emulator fixes

   - mask off reserved bits in CPUID

   - fix bug with disabling SGX

  RISC-V:

   - update MAINTAINERS"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86/xen: Fix eventfd error handling in kvm_xen_eventfd_assign()
  KVM: x86: smm: number of GPRs in the SMRAM image depends on the image format
  KVM: x86: emulator: update the emulation mode after CR0 write
  KVM: x86: emulator: update the emulation mode after rsm
  KVM: x86: emulator: introduce emulator_recalc_and_set_mode
  KVM: x86: emulator: em_sysexit should update ctxt->mode
  KVM: selftests: Mark "guest_saw_irq" as volatile in xen_shinfo_test
  KVM: selftests: Add tests in xen_shinfo_test to detect lock races
  KVM: Reject attempts to consume or refresh inactive gfn_to_pfn_cache
  KVM: Initialize gfn_to_pfn_cache locks in dedicated helper
  KVM: VMX: fully disable SGX if SECONDARY_EXEC_ENCLS_EXITING unavailable
  KVM: x86: Exempt pending triple fault from event injection sanity check
  MAINTAINERS: git://github -> https://github.com for kvm-riscv
  KVM: debugfs: Return retval of simple_attr_open() if it fails
  KVM: x86: Reduce refcount if single_open() fails in kvm_mmu_rmaps_stat_open()
  KVM: x86: Mask off reserved bits in CPUID.8000001FH
  KVM: x86: Mask off reserved bits in CPUID.8000001AH
  KVM: x86: Mask off reserved bits in CPUID.80000008H
  KVM: x86: Mask off reserved bits in CPUID.80000006H
  KVM: x86: Mask off reserved bits in CPUID.80000001H

1 files changed, 141 insertions, 1 deletions

diff --git a/tools/testing/selftests/kvm/x86_64/xen_shinfo_test.c b/tools/testing/selftests/kvm/x86_64/xen_shinfo_test.c
index 8a5cb800f50e..2a5727188c8d 100644
--- a/tools/testing/selftests/kvm/x86_64/xen_shinfo_test.c
+++ b/tools/testing/selftests/kvm/x86_64/xen_shinfo_test.c
@@ -15,9 +15,13 @@
 #include <time.h>
 #include <sched.h>
 #include <signal.h>
+#include <pthread.h>
 
 #include <sys/eventfd.h>
 
+/* Defined in include/linux/kvm_types.h */
+#define GPA_INVALID		(~(ulong)0)
+
 #define SHINFO_REGION_GVA	0xc0000000ULL
 #define SHINFO_REGION_GPA	0xc0000000ULL
 #define SHINFO_REGION_SLOT	10
@@ -44,6 +48,8 @@
 
 #define MIN_STEAL_TIME		50000
 
+#define SHINFO_RACE_TIMEOUT	2	/* seconds */
+
 #define __HYPERVISOR_set_timer_op	15
 #define __HYPERVISOR_sched_op		29
 #define __HYPERVISOR_event_channel_op	32
@@ -126,7 +132,7 @@ struct {
 	struct kvm_irq_routing_entry entries[2];
 } irq_routes;
 
-bool guest_saw_irq;
+static volatile bool guest_saw_irq;
 
 static void evtchn_handler(struct ex_regs *regs)
 {
@@ -148,6 +154,7 @@ static void guest_wait_for_irq(void)
 static void guest_code(void)
 {
 	struct vcpu_runstate_info *rs = (void *)RUNSTATE_VADDR;
+	int i;
 
 	__asm__ __volatile__(
 		"sti\n"
@@ -325,6 +332,49 @@ static void guest_code(void)
 	guest_wait_for_irq();
 
 	GUEST_SYNC(21);
+	/* Racing host ioctls */
+
+	guest_wait_for_irq();
+
+	GUEST_SYNC(22);
+	/* Racing vmcall against host ioctl */
+
+	ports[0] = 0;
+
+	p = (struct sched_poll) {
+		.ports = ports,
+		.nr_ports = 1,
+		.timeout = 0
+	};
+
+wait_for_timer:
+	/*
+	 * Poll for a timer wake event while the worker thread is mucking with
+	 * the shared info.  KVM XEN drops timer IRQs if the shared info is
+	 * invalid when the timer expires.  Arbitrarily poll 100 times before
+	 * giving up and asking the VMM to re-arm the timer.  100 polls should
+	 * consume enough time to beat on KVM without taking too long if the
+	 * timer IRQ is dropped due to an invalid event channel.
+	 */
+	for (i = 0; i < 100 && !guest_saw_irq; i++)
+		asm volatile("vmcall"
+			     : "=a" (rax)
+			     : "a" (__HYPERVISOR_sched_op),
+			       "D" (SCHEDOP_poll),
+			       "S" (&p)
+			     : "memory");
+
+	/*
+	 * Re-send the timer IRQ if it was (likely) dropped due to the timer
+	 * expiring while the event channel was invalid.
+	 */
+	if (!guest_saw_irq) {
+		GUEST_SYNC(23);
+		goto wait_for_timer;
+	}
+	guest_saw_irq = false;
+
+	GUEST_SYNC(24);
 }
 
 static int cmp_timespec(struct timespec *a, struct timespec *b)
@@ -352,11 +402,36 @@ static void handle_alrm(int sig)
 	TEST_FAIL("IRQ delivery timed out");
 }
 
+static void *juggle_shinfo_state(void *arg)
+{
+	struct kvm_vm *vm = (struct kvm_vm *)arg;
+
+	struct kvm_xen_hvm_attr cache_init = {
+		.type = KVM_XEN_ATTR_TYPE_SHARED_INFO,
+		.u.shared_info.gfn = SHINFO_REGION_GPA / PAGE_SIZE
+	};
+
+	struct kvm_xen_hvm_attr cache_destroy = {
+		.type = KVM_XEN_ATTR_TYPE_SHARED_INFO,
+		.u.shared_info.gfn = GPA_INVALID
+	};
+
+	for (;;) {
+		__vm_ioctl(vm, KVM_XEN_HVM_SET_ATTR, &cache_init);
+		__vm_ioctl(vm, KVM_XEN_HVM_SET_ATTR, &cache_destroy);
+		pthread_testcancel();
+	};
+
+	return NULL;
+}
+
 int main(int argc, char *argv[])
 {
 	struct timespec min_ts, max_ts, vm_ts;
 	struct kvm_vm *vm;
+	pthread_t thread;
 	bool verbose;
+	int ret;
 
 	verbose = argc > 1 && (!strncmp(argv[1], "-v", 3) ||
 			       !strncmp(argv[1], "--verbose", 10));
@@ -785,6 +860,71 @@ int main(int argc, char *argv[])
 			case 21:
 				TEST_ASSERT(!evtchn_irq_expected,
 					    "Expected event channel IRQ but it didn't happen");
+				alarm(0);
+
+				if (verbose)
+					printf("Testing shinfo lock corruption (KVM_XEN_HVM_EVTCHN_SEND)\n");
+
+				ret = pthread_create(&thread, NULL, &juggle_shinfo_state, (void *)vm);
+				TEST_ASSERT(ret == 0, "pthread_create() failed: %s", strerror(ret));
+
+				struct kvm_irq_routing_xen_evtchn uxe = {
+					.port = 1,
+					.vcpu = vcpu->id,
+					.priority = KVM_IRQ_ROUTING_XEN_EVTCHN_PRIO_2LEVEL
+				};
+
+				evtchn_irq_expected = true;
+				for (time_t t = time(NULL) + SHINFO_RACE_TIMEOUT; time(NULL) < t;)
+					__vm_ioctl(vm, KVM_XEN_HVM_EVTCHN_SEND, &uxe);
+				break;
+
+			case 22:
+				TEST_ASSERT(!evtchn_irq_expected,
+					    "Expected event channel IRQ but it didn't happen");
+
+				if (verbose)
+					printf("Testing shinfo lock corruption (SCHEDOP_poll)\n");
+
+				shinfo->evtchn_pending[0] = 1;
+
+				evtchn_irq_expected = true;
+				tmr.u.timer.expires_ns = rs->state_entry_time +
+							 SHINFO_RACE_TIMEOUT * 1000000000ULL;
+				vcpu_ioctl(vcpu, KVM_XEN_VCPU_SET_ATTR, &tmr);
+				break;
+
+			case 23:
+				/*
+				 * Optional and possibly repeated sync point.
+				 * Injecting the timer IRQ may fail if the
+				 * shinfo is invalid when the timer expires.
+				 * If the timer has expired but the IRQ hasn't
+				 * been delivered, rearm the timer and retry.
+				 */
+				vcpu_ioctl(vcpu, KVM_XEN_VCPU_GET_ATTR, &tmr);
+
+				/* Resume the guest if the timer is still pending. */
+				if (tmr.u.timer.expires_ns)
+					break;
+
+				/* All done if the IRQ was delivered. */
+				if (!evtchn_irq_expected)
+					break;
+
+				tmr.u.timer.expires_ns = rs->state_entry_time +
+							 SHINFO_RACE_TIMEOUT * 1000000000ULL;
+				vcpu_ioctl(vcpu, KVM_XEN_VCPU_SET_ATTR, &tmr);
+				break;
+			case 24:
+				TEST_ASSERT(!evtchn_irq_expected,
+					    "Expected event channel IRQ but it didn't happen");
+
+				ret = pthread_cancel(thread);
+				TEST_ASSERT(ret == 0, "pthread_cancel() failed: %s", strerror(ret));
+
+				ret = pthread_join(thread, 0);
+				TEST_ASSERT(ret == 0, "pthread_join() failed: %s", strerror(ret));
 				goto done;
 
 			case 0x20: