static void qemu_tcg_wait_io_event(void)
{
    CPUArchState *env;

    while (all_cpu_threads_idle()) {
#ifdef CONFIG_S2E_DEBUG
        s2e_debug_print("CPU: qemu_tcg_wait_io_event: waiting for awaken cpu %p\n", tcg_halt_cond);
#endif
        /* Start accounting real time to the virtual clock if the CPUs
          are idle.  */
        qemu_clock_warp(vm_clock);
        qemu_cond_wait(tcg_halt_cond, &qemu_global_mutex);
    }

    while (iothread_requesting_mutex) {
#ifdef CONFIG_S2E_DEBUG
        s2e_debug_print("CPU: qemu_tcg_wait_io_event iothread_requesting_mutex\n");
#endif
        qemu_cond_wait(&qemu_io_proceeded_cond, &qemu_global_mutex);
    }

    for (env = first_cpu; env != NULL; env = env->next_cpu) {
        qemu_wait_io_event_common(env);
    }
}

void pause_all_vcpus(void)
{
    CPUState *cpu = first_cpu;

    qemu_clock_enable(QEMU_CLOCK_VIRTUAL, false);
    while (cpu) {
        cpu->stop = true;
        qemu_cpu_kick(cpu);
        cpu = cpu->next_cpu;
    }

    if (qemu_in_vcpu_thread()) {
        cpu_stop_current();
        if (!kvm_enabled()) {
            cpu = first_cpu;
            while (cpu) {
                cpu->stop = false;
                cpu->stopped = true;
                cpu = cpu->next_cpu;
            }
            return;
        }
    }

    while (!all_vcpus_paused()) {
        qemu_cond_wait(&qemu_pause_cond, &qemu_global_mutex);
        cpu = first_cpu;
        while (cpu) {
            qemu_cpu_kick(cpu);
            cpu = cpu->next_cpu;
        }
    }
}

static void qemu_tcg_wait_io_event(void)
{
    CPUState *env;

    while (all_cpu_threads_idle()) {
       /* Start accounting real time to the virtual clock if the CPUs
          are idle.  */
        qemu_clock_warp(vm_clock);
        qemu_cond_wait(tcg_halt_cond, &qemu_global_mutex);
    }

    qemu_mutex_unlock(&qemu_global_mutex);

    /*
     * Users of qemu_global_mutex can be starved, having no chance
     * to acquire it since this path will get to it first.
     * So use another lock to provide fairness.
     */
    qemu_mutex_lock(&qemu_fair_mutex);
    qemu_mutex_unlock(&qemu_fair_mutex);

    qemu_mutex_lock(&qemu_global_mutex);

    for (env = first_cpu; env != NULL; env = env->next_cpu) {
        qemu_wait_io_event_common(env);
    }
}

static void qemu_tcg_init_vcpu(void *_env)
{
    CPUArchState *env = _env;

    /* share a single thread for all cpus with TCG */
    if (!tcg_cpu_thread) {
#ifdef CONFIG_S2E
        /* Forks inherit parent's memory, therefore we do not want
           to allocate new memory regions, just overwrite them. */
        if (!s2e_is_forking()) {
            env->thread = g_malloc0(sizeof(QemuThread));
            env->halt_cond = g_malloc0(sizeof(QemuCond));
        }
#else
        env->thread = g_malloc0(sizeof(QemuThread));
        env->halt_cond = g_malloc0(sizeof(QemuCond));
#endif
        qemu_cond_init(env->halt_cond);
        tcg_halt_cond = env->halt_cond;
        qemu_thread_create(env->thread, qemu_tcg_cpu_thread_fn, env,
                           QEMU_THREAD_JOINABLE);
#ifdef _WIN32
        env->hThread = qemu_thread_get_handle(env->thread);
#endif
        while (env->created == 0) {
            qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex);
        }
        tcg_cpu_thread = env->thread;
    } else {
        env->thread = tcg_cpu_thread;
        env->halt_cond = tcg_halt_cond;
    }
}

static void qemu_archipelago_close(BlockDriverState *bs)
{
    int r, targetlen;
    char *target;
    struct xseg_request *req;
    BDRVArchipelagoState *s = bs->opaque;

    s->stopping = true;

    qemu_mutex_lock(&s->request_mutex);
    while (!s->th_is_signaled) {
        qemu_cond_wait(&s->request_cond,
                       &s->request_mutex);
    }
    qemu_mutex_unlock(&s->request_mutex);
    qemu_thread_join(&s->request_th);
    qemu_cond_destroy(&s->request_cond);
    qemu_mutex_destroy(&s->request_mutex);

    qemu_cond_destroy(&s->archip_cond);
    qemu_mutex_destroy(&s->archip_mutex);

    targetlen = strlen(s->volname);
    req = xseg_get_request(s->xseg, s->srcport, s->vportno, X_ALLOC);
    if (!req) {
        archipelagolog("Cannot get XSEG request\n");
        goto err_exit;
    }
    r = xseg_prep_request(s->xseg, req, targetlen, 0);
    if (r < 0) {
        xseg_put_request(s->xseg, req, s->srcport);
        archipelagolog("Cannot prepare XSEG close request\n");
        goto err_exit;
    }

    target = xseg_get_target(s->xseg, req);
    memcpy(target, s->volname, targetlen);
    req->size = req->datalen;
    req->offset = 0;
    req->op = X_CLOSE;

    xport p = xseg_submit(s->xseg, req, s->srcport, X_ALLOC);
    if (p == NoPort) {
        xseg_put_request(s->xseg, req, s->srcport);
        archipelagolog("Cannot submit XSEG close request\n");
        goto err_exit;
    }

    xseg_signal(s->xseg, p);
    wait_reply(s->xseg, s->srcport, s->port, req);

    xseg_put_request(s->xseg, req, s->srcport);

err_exit:
    g_free(s->volname);
    g_free(s->segment_name);
    xseg_quit_local_signal(s->xseg, s->srcport);
    xseg_leave_dynport(s->xseg, s->port);
    xseg_leave(s->xseg);
}

void run_on_cpu(CPUState *cpu, void (*func)(void *data), void *data)
{
    struct qemu_work_item wi;

    if (qemu_cpu_is_self(cpu)) {
        func(data);
        return;
    }

    wi.func = func;
    wi.data = data;
    if (cpu->queued_work_first == NULL) {
        cpu->queued_work_first = &wi;
    } else {
        cpu->queued_work_last->next = &wi;
    }
    cpu->queued_work_last = &wi;
    wi.next = NULL;
    wi.done = false;

    qemu_cpu_kick(cpu);
    while (!wi.done) {
        CPUArchState *self_env = cpu_single_env;

        qemu_cond_wait(&qemu_work_cond, &qemu_global_mutex);
        cpu_single_env = self_env;
    }
}

/* Start an exclusive operation.
   Must only be called from outside cpu_exec.  */
void start_exclusive(void)
{
    CPUState *other_cpu;
    int running_cpus;

    qemu_mutex_lock(&qemu_cpu_list_lock);
    exclusive_idle();

    /* Make all other cpus stop executing.  */
    atomic_set(&pending_cpus, 1);

    /* Write pending_cpus before reading other_cpu->running.  */
    smp_mb();
    running_cpus = 0;
    CPU_FOREACH(other_cpu) {
        if (atomic_read(&other_cpu->running)) {
            other_cpu->has_waiter = true;
            running_cpus++;
            qemu_cpu_kick(other_cpu);
        }
    }

    atomic_set(&pending_cpus, running_cpus + 1);
    while (pending_cpus > 1) {
        qemu_cond_wait(&exclusive_cond, &qemu_cpu_list_lock);
    }

    /* Can release mutex, no one will enter another exclusive
     * section until end_exclusive resets pending_cpus to 0.
     */
    qemu_mutex_unlock(&qemu_cpu_list_lock);
}

static void *qemu_tcg_cpu_thread_fn(void *arg)
{
    CPUState *env = arg;

    qemu_tcg_init_cpu_signals();
    qemu_thread_get_self(env->thread);

    /* signal CPU creation */
    qemu_mutex_lock(&qemu_global_mutex);
    for (env = first_cpu; env != NULL; env = env->next_cpu) {
        env->thread_id = qemu_get_thread_id();
        env->created = 1;
    }
    qemu_cond_signal(&qemu_cpu_cond);

    /* wait for initial kick-off after machine start */
    while (first_cpu->stopped) {
        qemu_cond_wait(tcg_halt_cond, &qemu_global_mutex);
    }

    while (1) {
        cpu_exec_all();
        if (use_icount && qemu_next_icount_deadline() <= 0) {
            qemu_notify_event();
        }
        qemu_tcg_wait_io_event();
    }

    return NULL;
}

static void iothread_complete(UserCreatable *obj, Error **errp)
{
    Error *local_error = NULL;
    IOThread *iothread = IOTHREAD(obj);
    char *name, *thread_name;

    iothread->stopping = false;
    iothread->thread_id = -1;
    iothread->ctx = aio_context_new(&local_error);
    if (!iothread->ctx) {
        error_propagate(errp, local_error);
        return;
    }

    qemu_mutex_init(&iothread->init_done_lock);
    qemu_cond_init(&iothread->init_done_cond);

    /* This assumes we are called from a thread with useful CPU affinity for us
     * to inherit.
     */
    name = object_get_canonical_path_component(OBJECT(obj));
    thread_name = g_strdup_printf("IO %s", name);
    qemu_thread_create(&iothread->thread, thread_name, iothread_run,
                       iothread, QEMU_THREAD_JOINABLE);
    g_free(thread_name);
    g_free(name);

    /* Wait for initialization to complete */
    qemu_mutex_lock(&iothread->init_done_lock);
    while (iothread->thread_id == -1) {
        qemu_cond_wait(&iothread->init_done_cond,
                       &iothread->init_done_lock);
    }
    qemu_mutex_unlock(&iothread->init_done_lock);
}

void run_on_cpu(CPUState *env, void (*func)(void *data), void *data)
{
    struct qemu_work_item wi;

    if (qemu_cpu_self(env)) {
        func(data);
        return;
    }

    wi.func = func;
    wi.data = data;
    if (!env->queued_work_first)
        env->queued_work_first = &wi;
    else
        env->queued_work_last->next = &wi;
    env->queued_work_last = &wi;
    wi.next = NULL;
    wi.done = false;

    qemu_cpu_kick(env);
    while (!wi.done) {
        CPUState *self_env = cpu_single_env;

        qemu_cond_wait(&qemu_work_cond, &qemu_global_mutex);
        cpu_single_env = self_env;
    }
}

void pause_all_vcpus(void)
{
    CPUArchState *penv = first_cpu;

    qemu_clock_enable(vm_clock, false);
    while (penv) {
        CPUState *pcpu = ENV_GET_CPU(penv);
        pcpu->stop = true;
        qemu_cpu_kick(pcpu);
        penv = penv->next_cpu;
    }

    if (qemu_in_vcpu_thread()) {
        cpu_stop_current();
        if (!kvm_enabled()) {
            while (penv) {
                CPUState *pcpu = ENV_GET_CPU(penv);
                pcpu->stop = 0;
                pcpu->stopped = true;
                penv = penv->next_cpu;
            }
            return;
        }
    }

    while (!all_vcpus_paused()) {
        qemu_cond_wait(&qemu_pause_cond, &qemu_global_mutex);
        penv = first_cpu;
        while (penv) {
            qemu_cpu_kick(ENV_GET_CPU(penv));
            penv = penv->next_cpu;
        }
    }
}

void pause_all_vcpus(void)
{
    CPUArchState *penv = first_cpu;

    qemu_clock_enable(vm_clock, false);
    while (penv) {
        penv->stop = 1;
        qemu_cpu_kick(penv);
        penv = penv->next_cpu;
    }

    if (!qemu_thread_is_self(&io_thread)) {
        cpu_stop_current();
        if (!kvm_enabled()) {
            while (penv) {
                penv->stop = 0;
                penv->stopped = 1;
                penv = penv->next_cpu;
            }
            return;
        }
    }

    while (!all_vcpus_paused()) {
        qemu_cond_wait(&qemu_pause_cond, &qemu_global_mutex);
        penv = first_cpu;
        while (penv) {
            qemu_cpu_kick(penv);
            penv = penv->next_cpu;
        }
    }
}

static void iothread_complete(UserCreatable *obj, Error **errp)
{
    Error *local_error = NULL;
    IOThread *iothread = IOTHREAD(obj);

    iothread->stopping = false;
    iothread->thread_id = -1;
    iothread->ctx = aio_context_new(&local_error);
    if (!iothread->ctx) {
        error_propagate(errp, local_error);
        return;
    }

    qemu_mutex_init(&iothread->init_done_lock);
    qemu_cond_init(&iothread->init_done_cond);

    /* This assumes we are called from a thread with useful CPU affinity for us
     * to inherit.
     */
    qemu_thread_create(&iothread->thread, "iothread", iothread_run,
                       iothread, QEMU_THREAD_JOINABLE);

    /* Wait for initialization to complete */
    qemu_mutex_lock(&iothread->init_done_lock);
    while (iothread->thread_id == -1) {
        qemu_cond_wait(&iothread->init_done_cond,
                       &iothread->init_done_lock);
    }
    qemu_mutex_unlock(&iothread->init_done_lock);
}

static void qemu_kvm_wait_io_event(CPUState *env)
{
    while (cpu_thread_is_idle(env)) {
        qemu_cond_wait(env->halt_cond, &qemu_global_mutex);
    }

    qemu_kvm_eat_signals(env);
    qemu_wait_io_event_common(env);
}

static void qemu_tcg_wait_io_event(void)
{
    CPUArchState *env;

    while (all_cpu_threads_idle()) {
       /* Start accounting real time to the virtual clock if the CPUs
          are idle.  */
        qemu_clock_warp(vm_clock);
        qemu_cond_wait(tcg_halt_cond, &qemu_global_mutex);
    }

    while (iothread_requesting_mutex) {
        qemu_cond_wait(&qemu_io_proceeded_cond, &qemu_global_mutex);
    }

    for (env = first_cpu; env != NULL; env = env->next_cpu) {
        qemu_wait_io_event_common(ENV_GET_CPU(env));
    }
}

static void qemu_tcg_wait_io_event(void)
{
    CPUState *cpu;

    while (all_cpu_threads_idle()) {
       /* Start accounting real time to the virtual clock if the CPUs
          are idle.  */
        qemu_clock_warp(QEMU_CLOCK_VIRTUAL);
        qemu_cond_wait(tcg_halt_cond, &qemu_global_mutex);
    }

    while (iothread_requesting_mutex) {
        qemu_cond_wait(&qemu_io_proceeded_cond, &qemu_global_mutex);
    }

    for (cpu = first_cpu; cpu != NULL; cpu = cpu->next_cpu) {
        qemu_wait_io_event_common(cpu);
    }
}

static void qemu_kvm_start_vcpu(CPUState *env)
{
    env->thread = g_malloc0(sizeof(QemuThread));
    env->halt_cond = g_malloc0(sizeof(QemuCond));
    qemu_cond_init(env->halt_cond);
    qemu_thread_create(env->thread, qemu_kvm_cpu_thread_fn, env);
    while (env->created == 0) {
        qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex);
    }
}

static void qemu_kvm_wait_io_event(CPUArchState *env)
{
    CPUState *cpu = ENV_GET_CPU(env);

    while (cpu_thread_is_idle(env)) {
        qemu_cond_wait(cpu->halt_cond, &qemu_global_mutex);
    }

    qemu_kvm_eat_signals(env);
    qemu_wait_io_event_common(cpu);
}

static void qemu_dummy_start_vcpu(CPUArchState *env)
{
    env->thread = g_malloc0(sizeof(QemuThread));
    env->halt_cond = g_malloc0(sizeof(QemuCond));
    qemu_cond_init(env->halt_cond);
    qemu_thread_create(env->thread, qemu_dummy_cpu_thread_fn, env,
                       QEMU_THREAD_JOINABLE);
    while (env->created == 0) {
        qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex);
    }
}

static void qemu_dummy_start_vcpu(CPUState *cpu)
{
    cpu->thread = g_malloc0(sizeof(QemuThread));
    cpu->halt_cond = g_malloc0(sizeof(QemuCond));
    qemu_cond_init(cpu->halt_cond);
    qemu_thread_create(cpu->thread, qemu_dummy_cpu_thread_fn, cpu,
                       QEMU_THREAD_JOINABLE);
    while (!cpu->created) {
        qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex);
    }
}