1 files changed, 269 insertions, 2 deletions

diff --git a/target/i386/kvm/xen-emu.c b/target/i386/kvm/xen-emu.c
index 3b46cab..c210ff9 100644
--- a/target/i386/kvm/xen-emu.c
+++ b/target/i386/kvm/xen-emu.c
@@ -38,6 +38,9 @@
 
 #include "xen-compat.h"
 
+static void xen_vcpu_singleshot_timer_event(void *opaque);
+static void xen_vcpu_periodic_timer_event(void *opaque);
+
 #ifdef TARGET_X86_64
 #define hypercall_compat32(longmode) (!(longmode))
 #else
@@ -201,6 +204,23 @@ int kvm_xen_init_vcpu(CPUState *cs)
     env->xen_vcpu_time_info_gpa = INVALID_GPA;
     env->xen_vcpu_runstate_gpa = INVALID_GPA;
 
+    qemu_mutex_init(&env->xen_timers_lock);
+    env->xen_singleshot_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
+                                             xen_vcpu_singleshot_timer_event,
+                                             cpu);
+    if (!env->xen_singleshot_timer) {
+        return -ENOMEM;
+    }
+    env->xen_singleshot_timer->opaque = cs;
+
+    env->xen_periodic_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
+                                           xen_vcpu_periodic_timer_event,
+                                           cpu);
+    if (!env->xen_periodic_timer) {
+        return -ENOMEM;
+    }
+    env->xen_periodic_timer->opaque = cs;
+
     return 0;
 }
 
@@ -232,7 +252,8 @@ static bool kvm_xen_hcall_xen_version(struct kvm_xen_exit *exit, X86CPU *cpu,
                          1 << XENFEAT_writable_descriptor_tables |
                          1 << XENFEAT_auto_translated_physmap |
                          1 << XENFEAT_supervisor_mode_kernel |
-                         1 << XENFEAT_hvm_callback_vector;
+                         1 << XENFEAT_hvm_callback_vector |
+                         1 << XENFEAT_hvm_safe_pvclock;
         }
 
         err = kvm_copy_to_gva(CPU(cpu), arg, &fi, sizeof(fi));
@@ -878,13 +899,208 @@ static int vcpuop_register_runstate_info(CPUState *cs, CPUState *target,
     return 0;
 }
 
+static uint64_t kvm_get_current_ns(void)
+{
+    struct kvm_clock_data data;
+    int ret;
+
+    ret = kvm_vm_ioctl(kvm_state, KVM_GET_CLOCK, &data);
+    if (ret < 0) {
+        fprintf(stderr, "KVM_GET_CLOCK failed: %s\n", strerror(ret));
+                abort();
+    }
+
+    return data.clock;
+}
+
+static void xen_vcpu_singleshot_timer_event(void *opaque)
+{
+    CPUState *cpu = opaque;
+    CPUX86State *env = &X86_CPU(cpu)->env;
+    uint16_t port = env->xen_virq[VIRQ_TIMER];
+
+    if (likely(port)) {
+        xen_evtchn_set_port(port);
+    }
+
+    qemu_mutex_lock(&env->xen_timers_lock);
+    env->xen_singleshot_timer_ns = 0;
+    qemu_mutex_unlock(&env->xen_timers_lock);
+}
+
+static void xen_vcpu_periodic_timer_event(void *opaque)
+{
+    CPUState *cpu = opaque;
+    CPUX86State *env = &X86_CPU(cpu)->env;
+    uint16_t port = env->xen_virq[VIRQ_TIMER];
+    int64_t qemu_now;
+
+    if (likely(port)) {
+        xen_evtchn_set_port(port);
+    }
+
+    qemu_mutex_lock(&env->xen_timers_lock);
+
+    qemu_now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+    timer_mod_ns(env->xen_periodic_timer,
+                 qemu_now + env->xen_periodic_timer_period);
+
+    qemu_mutex_unlock(&env->xen_timers_lock);
+}
+
+static int do_set_periodic_timer(CPUState *target, uint64_t period_ns)
+{
+    CPUX86State *tenv = &X86_CPU(target)->env;
+    int64_t qemu_now;
+
+    timer_del(tenv->xen_periodic_timer);
+
+    qemu_mutex_lock(&tenv->xen_timers_lock);
+
+    qemu_now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+    timer_mod_ns(tenv->xen_periodic_timer, qemu_now + period_ns);
+    tenv->xen_periodic_timer_period = period_ns;
+
+    qemu_mutex_unlock(&tenv->xen_timers_lock);
+    return 0;
+}
+
+#define MILLISECS(_ms)  ((int64_t)((_ms) * 1000000ULL))
+#define MICROSECS(_us)  ((int64_t)((_us) * 1000ULL))
+#define STIME_MAX ((time_t)((int64_t)~0ull >> 1))
+/* Chosen so (NOW() + delta) wont overflow without an uptime of 200 years */
+#define STIME_DELTA_MAX ((int64_t)((uint64_t)~0ull >> 2))
+
+static int vcpuop_set_periodic_timer(CPUState *cs, CPUState *target,
+                                     uint64_t arg)
+{
+    struct vcpu_set_periodic_timer spt;
+
+    qemu_build_assert(sizeof(spt) == 8);
+    if (kvm_copy_from_gva(cs, arg, &spt, sizeof(spt))) {
+        return -EFAULT;
+    }
+
+    if (spt.period_ns < MILLISECS(1) || spt.period_ns > STIME_DELTA_MAX) {
+        return -EINVAL;
+    }
+
+    return do_set_periodic_timer(target, spt.period_ns);
+}
+
+static int vcpuop_stop_periodic_timer(CPUState *target)
+{
+    CPUX86State *tenv = &X86_CPU(target)->env;
+
+    qemu_mutex_lock(&tenv->xen_timers_lock);
+
+    timer_del(tenv->xen_periodic_timer);
+    tenv->xen_periodic_timer_period = 0;
+
+    qemu_mutex_unlock(&tenv->xen_timers_lock);
+    return 0;
+}
+
+static int do_set_singleshot_timer(CPUState *cs, uint64_t timeout_abs,
+                                   bool future, bool linux_wa)
+{
+    CPUX86State *env = &X86_CPU(cs)->env;
+    int64_t now = kvm_get_current_ns();
+    int64_t qemu_now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+    int64_t delta = timeout_abs - now;
+
+    if (future && timeout_abs < now) {
+        return -ETIME;
+    }
+
+    if (linux_wa && unlikely((int64_t)timeout_abs < 0 ||
+                             (delta > 0 && (uint32_t)(delta >> 50) != 0))) {
+        /*
+         * Xen has a 'Linux workaround' in do_set_timer_op() which checks
+         * for negative absolute timeout values (caused by integer
+         * overflow), and for values about 13 days in the future (2^50ns)
+         * which would be caused by jiffies overflow. For those cases, it
+         * sets the timeout 100ms in the future (not *too* soon, since if
+         * a guest really did set a long timeout on purpose we don't want
+         * to keep churning CPU time by waking it up).
+         */
+        delta = (100 * SCALE_MS);
+        timeout_abs = now + delta;
+    }
+
+    qemu_mutex_lock(&env->xen_timers_lock);
+
+    timer_mod_ns(env->xen_singleshot_timer, qemu_now + delta);
+    env->xen_singleshot_timer_ns = now + delta;
+
+    qemu_mutex_unlock(&env->xen_timers_lock);
+    return 0;
+}
+
+static int vcpuop_set_singleshot_timer(CPUState *cs, uint64_t arg)
+{
+    struct vcpu_set_singleshot_timer sst = { 0 };
+
+    /*
+     * The struct is a uint64_t followed by a uint32_t. On 32-bit that
+     * makes it 12 bytes. On 64-bit it gets padded to 16. The parts
+     * that get used are identical, and there's four bytes of padding
+     * unused at the end. For true Xen compatibility we should attempt
+     * to copy the full 16 bytes from 64-bit guests, and return -EFAULT
+     * if we can't get the padding too. But that's daft. Just copy what
+     * we need.
+     */
+    qemu_build_assert(offsetof(struct vcpu_set_singleshot_timer, flags) == 8);
+    qemu_build_assert(sizeof(sst) >= 12);
+
+    if (kvm_copy_from_gva(cs, arg, &sst, 12)) {
+        return -EFAULT;
+    }
+
+    return do_set_singleshot_timer(cs, sst.timeout_abs_ns,
+                                   !!(sst.flags & VCPU_SSHOTTMR_future),
+                                   false);
+}
+
+static int vcpuop_stop_singleshot_timer(CPUState *cs)
+{
+    CPUX86State *env = &X86_CPU(cs)->env;
+
+    qemu_mutex_lock(&env->xen_timers_lock);
+
+    timer_del(env->xen_singleshot_timer);
+    env->xen_singleshot_timer_ns = 0;
+
+    qemu_mutex_unlock(&env->xen_timers_lock);
+    return 0;
+}
+
+static bool kvm_xen_hcall_set_timer_op(struct kvm_xen_exit *exit, X86CPU *cpu,
+                                       uint64_t timeout)
+{
+    int err;
+
+    if (unlikely(timeout == 0)) {
+        err = vcpuop_stop_singleshot_timer(CPU(cpu));
+    } else {
+        err = do_set_singleshot_timer(CPU(cpu), timeout, false, true);
+    }
+    exit->u.hcall.result = err;
+    return true;
+}
+
 static bool kvm_xen_hcall_vcpu_op(struct kvm_xen_exit *exit, X86CPU *cpu,
                                   int cmd, int vcpu_id, uint64_t arg)
 {
-    CPUState *dest = qemu_get_cpu(vcpu_id);
     CPUState *cs = CPU(cpu);
+    CPUState *dest = cs->cpu_index == vcpu_id ? cs : qemu_get_cpu(vcpu_id);
     int err;
 
+    if (!dest) {
+        err = -ENOENT;
+        goto out;
+    }
+
     switch (cmd) {
     case VCPUOP_register_runstate_memory_area:
         err = vcpuop_register_runstate_info(cs, dest, arg);
@@ -895,11 +1111,34 @@ static bool kvm_xen_hcall_vcpu_op(struct kvm_xen_exit *exit, X86CPU *cpu,
     case VCPUOP_register_vcpu_info:
         err = vcpuop_register_vcpu_info(cs, dest, arg);
         break;
+    case VCPUOP_set_singleshot_timer: {
+        if (cs->cpu_index == vcpu_id) {
+            err = vcpuop_set_singleshot_timer(dest, arg);
+        } else {
+            err = -EINVAL;
+        }
+        break;
+    }
+    case VCPUOP_stop_singleshot_timer:
+        if (cs->cpu_index == vcpu_id) {
+            err = vcpuop_stop_singleshot_timer(dest);
+        } else {
+            err = -EINVAL;
+        }
+        break;
+    case VCPUOP_set_periodic_timer: {
+        err = vcpuop_set_periodic_timer(cs, dest, arg);
+        break;
+    }
+    case VCPUOP_stop_periodic_timer:
+        err = vcpuop_stop_periodic_timer(dest);
+        break;
 
     default:
         return false;
     }
 
+ out:
     exit->u.hcall.result = err;
     return true;
 }
@@ -1249,6 +1488,16 @@ static bool do_kvm_xen_handle_exit(X86CPU *cpu, struct kvm_xen_exit *exit)
     }
 
     switch (code) {
+    case __HYPERVISOR_set_timer_op:
+        if (exit->u.hcall.longmode) {
+            return kvm_xen_hcall_set_timer_op(exit, cpu,
+                                              exit->u.hcall.params[0]);
+        } else {
+            /* In 32-bit mode, the 64-bit timer value is in two args. */
+            uint64_t val = ((uint64_t)exit->u.hcall.params[1]) << 32 |
+                (uint32_t)exit->u.hcall.params[0];
+            return kvm_xen_hcall_set_timer_op(exit, cpu, val);
+        }
     case __HYPERVISOR_grant_table_op:
         return kvm_xen_hcall_gnttab_op(exit, cpu, exit->u.hcall.params[0],
                                        exit->u.hcall.params[1],
@@ -1358,7 +1607,25 @@ int kvm_put_xen_state(CPUState *cs)
         }
     }
 
+    if (env->xen_periodic_timer_period) {
+        ret = do_set_periodic_timer(cs, env->xen_periodic_timer_period);
+        if (ret < 0) {
+            return ret;
+        }
+    }
+
     if (!kvm_xen_has_cap(EVTCHN_SEND)) {
+        /*
+         * If the kernel has EVTCHN_SEND support then it handles timers too,
+         * so the timer will be restored by kvm_xen_set_vcpu_timer() below.
+         */
+        if (env->xen_singleshot_timer_ns) {
+            ret = do_set_singleshot_timer(cs, env->xen_singleshot_timer_ns,
+                                    false, false);
+            if (ret < 0) {
+                return ret;
+            }
+        }
         return 0;
     }