accel/tcg: split CpusAccel into three TCG variants

split up the CpusAccel tcg_cpus into three TCG variants:

tcg_cpus_rr (single threaded, round robin cpus)
tcg_cpus_icount (same as rr, but with instruction counting enabled)
tcg_cpus_mttcg (multi-threaded cpus)

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20201015143217.29337-2-cfontana@suse.de>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

1 files changed, 27 insertions, 457 deletions

diff --git a/accel/tcg/tcg-cpus.c b/accel/tcg/tcg-cpus.c
index da1c63d8f6..f2b9bbf99e 100644
--- a/accel/tcg/tcg-cpus.c
+++ b/accel/tcg/tcg-cpus.c
@@ -1,5 +1,7 @@
 /*
- * QEMU System Emulator
+ * QEMU TCG vCPU common functionality
+ *
+ * Functionality common to all TCG vCPU variants: mttcg, rr and icount.
  *
  * Copyright (c) 2003-2008 Fabrice Bellard
  * Copyright (c) 2014 Red Hat Inc.
@@ -33,436 +35,12 @@
 #include "hw/boards.h"
 
 #include "tcg-cpus.h"
+#include "tcg-cpus-mttcg.h"
+#include "tcg-cpus-rr.h"
 
-/* Kick all RR vCPUs */
-static void qemu_cpu_kick_rr_cpus(void)
-{
-    CPUState *cpu;
-
-    CPU_FOREACH(cpu) {
-        cpu_exit(cpu);
-    };
-}
-
-static void tcg_kick_vcpu_thread(CPUState *cpu)
-{
-    if (qemu_tcg_mttcg_enabled()) {
-        cpu_exit(cpu);
-    } else {
-        qemu_cpu_kick_rr_cpus();
-    }
-}
-
-/*
- * TCG vCPU kick timer
- *
- * The kick timer is responsible for moving single threaded vCPU
- * emulation on to the next vCPU. If more than one vCPU is running a
- * timer event with force a cpu->exit so the next vCPU can get
- * scheduled.
- *
- * The timer is removed if all vCPUs are idle and restarted again once
- * idleness is complete.
- */
-
-static QEMUTimer *tcg_kick_vcpu_timer;
-static CPUState *tcg_current_rr_cpu;
-
-#define TCG_KICK_PERIOD (NANOSECONDS_PER_SECOND / 10)
-
-static inline int64_t qemu_tcg_next_kick(void)
-{
-    return qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + TCG_KICK_PERIOD;
-}
-
-/* Kick the currently round-robin scheduled vCPU to next */
-static void qemu_cpu_kick_rr_next_cpu(void)
-{
-    CPUState *cpu;
-    do {
-        cpu = qatomic_mb_read(&tcg_current_rr_cpu);
-        if (cpu) {
-            cpu_exit(cpu);
-        }
-    } while (cpu != qatomic_mb_read(&tcg_current_rr_cpu));
-}
-
-static void kick_tcg_thread(void *opaque)
-{
-    timer_mod(tcg_kick_vcpu_timer, qemu_tcg_next_kick());
-    qemu_cpu_kick_rr_next_cpu();
-}
-
-static void start_tcg_kick_timer(void)
-{
-    assert(!mttcg_enabled);
-    if (!tcg_kick_vcpu_timer && CPU_NEXT(first_cpu)) {
-        tcg_kick_vcpu_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
-                                           kick_tcg_thread, NULL);
-    }
-    if (tcg_kick_vcpu_timer && !timer_pending(tcg_kick_vcpu_timer)) {
-        timer_mod(tcg_kick_vcpu_timer, qemu_tcg_next_kick());
-    }
-}
-
-static void stop_tcg_kick_timer(void)
-{
-    assert(!mttcg_enabled);
-    if (tcg_kick_vcpu_timer && timer_pending(tcg_kick_vcpu_timer)) {
-        timer_del(tcg_kick_vcpu_timer);
-    }
-}
-
-static void qemu_tcg_destroy_vcpu(CPUState *cpu)
-{
-}
-
-static void qemu_tcg_rr_wait_io_event(void)
-{
-    CPUState *cpu;
-
-    while (all_cpu_threads_idle()) {
-        stop_tcg_kick_timer();
-        qemu_cond_wait_iothread(first_cpu->halt_cond);
-    }
+/* common functionality among all TCG variants */
 
-    start_tcg_kick_timer();
-
-    CPU_FOREACH(cpu) {
-        qemu_wait_io_event_common(cpu);
-    }
-}
-
-static int64_t tcg_get_icount_limit(void)
-{
-    int64_t deadline;
-
-    if (replay_mode != REPLAY_MODE_PLAY) {
-        /*
-         * Include all the timers, because they may need an attention.
-         * Too long CPU execution may create unnecessary delay in UI.
-         */
-        deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
-                                              QEMU_TIMER_ATTR_ALL);
-        /* Check realtime timers, because they help with input processing */
-        deadline = qemu_soonest_timeout(deadline,
-                qemu_clock_deadline_ns_all(QEMU_CLOCK_REALTIME,
-                                           QEMU_TIMER_ATTR_ALL));
-
-        /*
-         * Maintain prior (possibly buggy) behaviour where if no deadline
-         * was set (as there is no QEMU_CLOCK_VIRTUAL timer) or it is more than
-         * INT32_MAX nanoseconds ahead, we still use INT32_MAX
-         * nanoseconds.
-         */
-        if ((deadline < 0) || (deadline > INT32_MAX)) {
-            deadline = INT32_MAX;
-        }
-
-        return icount_round(deadline);
-    } else {
-        return replay_get_instructions();
-    }
-}
-
-static void notify_aio_contexts(void)
-{
-    /* Wake up other AioContexts.  */
-    qemu_clock_notify(QEMU_CLOCK_VIRTUAL);
-    qemu_clock_run_timers(QEMU_CLOCK_VIRTUAL);
-}
-
-static void handle_icount_deadline(void)
-{
-    assert(qemu_in_vcpu_thread());
-    if (icount_enabled()) {
-        int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
-                                                      QEMU_TIMER_ATTR_ALL);
-
-        if (deadline == 0) {
-            notify_aio_contexts();
-        }
-    }
-}
-
-static void prepare_icount_for_run(CPUState *cpu)
-{
-    if (icount_enabled()) {
-        int insns_left;
-
-        /*
-         * These should always be cleared by process_icount_data after
-         * each vCPU execution. However u16.high can be raised
-         * asynchronously by cpu_exit/cpu_interrupt/tcg_handle_interrupt
-         */
-        g_assert(cpu_neg(cpu)->icount_decr.u16.low == 0);
-        g_assert(cpu->icount_extra == 0);
-
-        cpu->icount_budget = tcg_get_icount_limit();
-        insns_left = MIN(0xffff, cpu->icount_budget);
-        cpu_neg(cpu)->icount_decr.u16.low = insns_left;
-        cpu->icount_extra = cpu->icount_budget - insns_left;
-
-        replay_mutex_lock();
-
-        if (cpu->icount_budget == 0 && replay_has_checkpoint()) {
-            notify_aio_contexts();
-        }
-    }
-}
-
-static void process_icount_data(CPUState *cpu)
-{
-    if (icount_enabled()) {
-        /* Account for executed instructions */
-        icount_update(cpu);
-
-        /* Reset the counters */
-        cpu_neg(cpu)->icount_decr.u16.low = 0;
-        cpu->icount_extra = 0;
-        cpu->icount_budget = 0;
-
-        replay_account_executed_instructions();
-
-        replay_mutex_unlock();
-    }
-}
-
-static int tcg_cpu_exec(CPUState *cpu)
-{
-    int ret;
-#ifdef CONFIG_PROFILER
-    int64_t ti;
-#endif
-
-    assert(tcg_enabled());
-#ifdef CONFIG_PROFILER
-    ti = profile_getclock();
-#endif
-    cpu_exec_start(cpu);
-    ret = cpu_exec(cpu);
-    cpu_exec_end(cpu);
-#ifdef CONFIG_PROFILER
-    qatomic_set(&tcg_ctx->prof.cpu_exec_time,
-                tcg_ctx->prof.cpu_exec_time + profile_getclock() - ti);
-#endif
-    return ret;
-}
-
-/*
- * Destroy any remaining vCPUs which have been unplugged and have
- * finished running
- */
-static void deal_with_unplugged_cpus(void)
-{
-    CPUState *cpu;
-
-    CPU_FOREACH(cpu) {
-        if (cpu->unplug && !cpu_can_run(cpu)) {
-            qemu_tcg_destroy_vcpu(cpu);
-            cpu_thread_signal_destroyed(cpu);
-            break;
-        }
-    }
-}
-
-/*
- * Single-threaded TCG
- *
- * In the single-threaded case each vCPU is simulated in turn. If
- * there is more than a single vCPU we create a simple timer to kick
- * the vCPU and ensure we don't get stuck in a tight loop in one vCPU.
- * This is done explicitly rather than relying on side-effects
- * elsewhere.
- */
-
-static void *tcg_rr_cpu_thread_fn(void *arg)
-{
-    CPUState *cpu = arg;
-
-    assert(tcg_enabled());
-    rcu_register_thread();
-    tcg_register_thread();
-
-    qemu_mutex_lock_iothread();
-    qemu_thread_get_self(cpu->thread);
-
-    cpu->thread_id = qemu_get_thread_id();
-    cpu->can_do_io = 1;
-    cpu_thread_signal_created(cpu);
-    qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
-    /* wait for initial kick-off after machine start */
-    while (first_cpu->stopped) {
-        qemu_cond_wait_iothread(first_cpu->halt_cond);
-
-        /* process any pending work */
-        CPU_FOREACH(cpu) {
-            current_cpu = cpu;
-            qemu_wait_io_event_common(cpu);
-        }
-    }
-
-    start_tcg_kick_timer();
-
-    cpu = first_cpu;
-
-    /* process any pending work */
-    cpu->exit_request = 1;
-
-    while (1) {
-        qemu_mutex_unlock_iothread();
-        replay_mutex_lock();
-        qemu_mutex_lock_iothread();
-        /* Account partial waits to QEMU_CLOCK_VIRTUAL.  */
-        icount_account_warp_timer();
-
-        /*
-         * Run the timers here.  This is much more efficient than
-         * waking up the I/O thread and waiting for completion.
-         */
-        handle_icount_deadline();
-
-        replay_mutex_unlock();
-
-        if (!cpu) {
-            cpu = first_cpu;
-        }
-
-        while (cpu && cpu_work_list_empty(cpu) && !cpu->exit_request) {
-
-            qatomic_mb_set(&tcg_current_rr_cpu, cpu);
-            current_cpu = cpu;
-
-            qemu_clock_enable(QEMU_CLOCK_VIRTUAL,
-                              (cpu->singlestep_enabled & SSTEP_NOTIMER) == 0);
-
-            if (cpu_can_run(cpu)) {
-                int r;
-
-                qemu_mutex_unlock_iothread();
-                prepare_icount_for_run(cpu);
-
-                r = tcg_cpu_exec(cpu);
-
-                process_icount_data(cpu);
-                qemu_mutex_lock_iothread();
-
-                if (r == EXCP_DEBUG) {
-                    cpu_handle_guest_debug(cpu);
-                    break;
-                } else if (r == EXCP_ATOMIC) {
-                    qemu_mutex_unlock_iothread();
-                    cpu_exec_step_atomic(cpu);
-                    qemu_mutex_lock_iothread();
-                    break;
-                }
-            } else if (cpu->stop) {
-                if (cpu->unplug) {
-                    cpu = CPU_NEXT(cpu);
-                }
-                break;
-            }
-
-            cpu = CPU_NEXT(cpu);
-        } /* while (cpu && !cpu->exit_request).. */
-
-        /* Does not need qatomic_mb_set because a spurious wakeup is okay.  */
-        qatomic_set(&tcg_current_rr_cpu, NULL);
-
-        if (cpu && cpu->exit_request) {
-            qatomic_mb_set(&cpu->exit_request, 0);
-        }
-
-        if (icount_enabled() && all_cpu_threads_idle()) {
-            /*
-             * When all cpus are sleeping (e.g in WFI), to avoid a deadlock
-             * in the main_loop, wake it up in order to start the warp timer.
-             */
-            qemu_notify_event();
-        }
-
-        qemu_tcg_rr_wait_io_event();
-        deal_with_unplugged_cpus();
-    }
-
-    rcu_unregister_thread();
-    return NULL;
-}
-
-/*
- * Multi-threaded TCG
- *
- * In the multi-threaded case each vCPU has its own thread. The TLS
- * variable current_cpu can be used deep in the code to find the
- * current CPUState for a given thread.
- */
-
-static void *tcg_cpu_thread_fn(void *arg)
-{
-    CPUState *cpu = arg;
-
-    assert(tcg_enabled());
-    g_assert(!icount_enabled());
-
-    rcu_register_thread();
-    tcg_register_thread();
-
-    qemu_mutex_lock_iothread();
-    qemu_thread_get_self(cpu->thread);
-
-    cpu->thread_id = qemu_get_thread_id();
-    cpu->can_do_io = 1;
-    current_cpu = cpu;
-    cpu_thread_signal_created(cpu);
-    qemu_guest_random_seed_thread_part2(cpu->random_seed);
-
-    /* process any pending work */
-    cpu->exit_request = 1;
-
-    do {
-        if (cpu_can_run(cpu)) {
-            int r;
-            qemu_mutex_unlock_iothread();
-            r = tcg_cpu_exec(cpu);
-            qemu_mutex_lock_iothread();
-            switch (r) {
-            case EXCP_DEBUG:
-                cpu_handle_guest_debug(cpu);
-                break;
-            case EXCP_HALTED:
-                /*
-                 * during start-up the vCPU is reset and the thread is
-                 * kicked several times. If we don't ensure we go back
-                 * to sleep in the halted state we won't cleanly
-                 * start-up when the vCPU is enabled.
-                 *
-                 * cpu->halted should ensure we sleep in wait_io_event
-                 */
-                g_assert(cpu->halted);
-                break;
-            case EXCP_ATOMIC:
-                qemu_mutex_unlock_iothread();
-                cpu_exec_step_atomic(cpu);
-                qemu_mutex_lock_iothread();
-            default:
-                /* Ignore everything else? */
-                break;
-            }
-        }
-
-        qatomic_mb_set(&cpu->exit_request, 0);
-        qemu_wait_io_event(cpu);
-    } while (!cpu->unplug || cpu_can_run(cpu));
-
-    qemu_tcg_destroy_vcpu(cpu);
-    cpu_thread_signal_destroyed(cpu);
-    qemu_mutex_unlock_iothread();
-    rcu_unregister_thread();
-    return NULL;
-}
-
-static void tcg_start_vcpu_thread(CPUState *cpu)
+void tcg_start_vcpu_thread(CPUState *cpu)
 {
     char thread_name[VCPU_THREAD_NAME_SIZE];
     static QemuCond *single_tcg_halt_cond;
@@ -518,29 +96,36 @@ static void tcg_start_vcpu_thread(CPUState *cpu)
     }
 }
 
-static int64_t tcg_get_virtual_clock(void)
+void qemu_tcg_destroy_vcpu(CPUState *cpu)
 {
-    if (icount_enabled()) {
-        return icount_get();
-    }
-    return cpu_get_clock();
+    cpu_thread_signal_destroyed(cpu);
 }
 
-static int64_t tcg_get_elapsed_ticks(void)
+int tcg_cpu_exec(CPUState *cpu)
 {
-    if (icount_enabled()) {
-        return icount_get();
-    }
-    return cpu_get_ticks();
+    int ret;
+#ifdef CONFIG_PROFILER
+    int64_t ti;
+#endif
+    assert(tcg_enabled());
+#ifdef CONFIG_PROFILER
+    ti = profile_getclock();
+#endif
+    cpu_exec_start(cpu);
+    ret = cpu_exec(cpu);
+    cpu_exec_end(cpu);
+#ifdef CONFIG_PROFILER
+    qatomic_set(&tcg_ctx->prof.cpu_exec_time,
+                tcg_ctx->prof.cpu_exec_time + profile_getclock() - ti);
+#endif
+    return ret;
 }
 
 /* mask must never be zero, except for A20 change call */
-static void tcg_handle_interrupt(CPUState *cpu, int mask)
+void tcg_handle_interrupt(CPUState *cpu, int mask)
 {
-    int old_mask;
     g_assert(qemu_mutex_iothread_locked());
 
-    old_mask = cpu->interrupt_request;
     cpu->interrupt_request |= mask;
 
     /*
@@ -551,20 +136,5 @@ static void tcg_handle_interrupt(CPUState *cpu, int mask)
         qemu_cpu_kick(cpu);
     } else {
         qatomic_set(&cpu_neg(cpu)->icount_decr.u16.high, -1);
-        if (icount_enabled() &&
-            !cpu->can_do_io
-            && (mask & ~old_mask) != 0) {
-            cpu_abort(cpu, "Raised interrupt while not in I/O function");
-        }
     }
 }
-
-const CpusAccel tcg_cpus = {
-    .create_vcpu_thread = tcg_start_vcpu_thread,
-    .kick_vcpu_thread = tcg_kick_vcpu_thread,
-
-    .handle_interrupt = tcg_handle_interrupt,
-
-    .get_virtual_clock = tcg_get_virtual_clock,
-    .get_elapsed_ticks = tcg_get_elapsed_ticks,
-};