accel/tcg: split tcg_start_vcpu_thread

after the initial split into 3 tcg variants, we proceed to also
split tcg_start_vcpu_thread.

We actually split it in 2 this time, since the icount variant
just uses the round robin function.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Message-Id: <20201015143217.29337-3-cfontana@suse.de>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

accel/tcg/tcg-all.c

@@ -105,6 +105,11 @@ static int tcg_init(MachineState *ms)
     tcg_exec_init(s->tb_size * 1024 * 1024);
     mttcg_enabled = s->mttcg_enabled;
 
+    /*
+     * Initialize TCG regions
+     */
+    tcg_region_init();
+
     if (mttcg_enabled) {
         cpus_register_accel(&tcg_cpus_mttcg);
     } else if (icount_enabled()) {

accel/tcg/tcg-cpus-icount.c

@@ -138,7 +138,7 @@ static void icount_handle_interrupt(CPUState *cpu, int mask)
 }
 
 const CpusAccel tcg_cpus_icount = {
-    .create_vcpu_thread = tcg_start_vcpu_thread,
+    .create_vcpu_thread = rr_start_vcpu_thread,
     .kick_vcpu_thread = qemu_cpu_kick_rr_cpus,
 
     .handle_interrupt = icount_handle_interrupt,

accel/tcg/tcg-cpus-mttcg.c

@@ -33,7 +33,6 @@
 #include "hw/boards.h"
 
 #include "tcg-cpus.h"
-#include "tcg-cpus-mttcg.h"
 
 /*
  * In the multi-threaded case each vCPU has its own thread. The TLS
@@ -41,7 +40,7 @@
  * current CPUState for a given thread.
  */
 
-void *tcg_cpu_thread_fn(void *arg)
+static void *tcg_cpu_thread_fn(void *arg)
 {
     CPUState *cpu = arg;
 
@@ -109,8 +108,32 @@ static void mttcg_kick_vcpu_thread(CPUState *cpu)
     cpu_exit(cpu);
 }
 
+static void mttcg_start_vcpu_thread(CPUState *cpu)
+{
+    char thread_name[VCPU_THREAD_NAME_SIZE];
+
+    g_assert(tcg_enabled());
+
+    parallel_cpus = (current_machine->smp.max_cpus > 1);
+
+    cpu->thread = g_malloc0(sizeof(QemuThread));
+    cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+    qemu_cond_init(cpu->halt_cond);
+
+    /* create a thread per vCPU with TCG (MTTCG) */
+    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/TCG",
+             cpu->cpu_index);
+
+    qemu_thread_create(cpu->thread, thread_name, tcg_cpu_thread_fn,
+                       cpu, QEMU_THREAD_JOINABLE);
+
+#ifdef _WIN32
+    cpu->hThread = qemu_thread_get_handle(cpu->thread);
+#endif
+}
+
 const CpusAccel tcg_cpus_mttcg = {
-    .create_vcpu_thread = tcg_start_vcpu_thread,
+    .create_vcpu_thread = mttcg_start_vcpu_thread,
     .kick_vcpu_thread = mttcg_kick_vcpu_thread,
 
     .handle_interrupt = tcg_handle_interrupt,

accel/tcg/tcg-cpus-mttcg.h

@@ -1,21 +0,0 @@
-/*
- * QEMU TCG Multi Threaded vCPUs implementation
- *
- * Copyright 2020 SUSE LLC
- *
- * This work is licensed under the terms of the GNU GPL, version 2 or later.
- * See the COPYING file in the top-level directory.
- */
-
-#ifndef TCG_CPUS_MTTCG_H
-#define TCG_CPUS_MTTCG_H
-
-/*
- * 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.
- */
-
-void *tcg_cpu_thread_fn(void *arg);
-
-#endif /* TCG_CPUS_MTTCG_H */

accel/tcg/tcg-cpus-rr.c

@@ -144,7 +144,7 @@ static void deal_with_unplugged_cpus(void)
  * elsewhere.
  */
 
-void *tcg_rr_cpu_thread_fn(void *arg)
+static void *tcg_rr_cpu_thread_fn(void *arg)
 {
     CPUState *cpu = arg;
 
@@ -262,8 +262,43 @@ void *tcg_rr_cpu_thread_fn(void *arg)
     return NULL;
 }
 
+void rr_start_vcpu_thread(CPUState *cpu)
+{
+    char thread_name[VCPU_THREAD_NAME_SIZE];
+    static QemuCond *single_tcg_halt_cond;
+    static QemuThread *single_tcg_cpu_thread;
+
+    g_assert(tcg_enabled());
+    parallel_cpus = false;
+
+    if (!single_tcg_cpu_thread) {
+        cpu->thread = g_malloc0(sizeof(QemuThread));
+        cpu->halt_cond = g_malloc0(sizeof(QemuCond));
+        qemu_cond_init(cpu->halt_cond);
+
+        /* share a single thread for all cpus with TCG */
+        snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "ALL CPUs/TCG");
+        qemu_thread_create(cpu->thread, thread_name,
+                           tcg_rr_cpu_thread_fn,
+                           cpu, QEMU_THREAD_JOINABLE);
+
+        single_tcg_halt_cond = cpu->halt_cond;
+        single_tcg_cpu_thread = cpu->thread;
+#ifdef _WIN32
+        cpu->hThread = qemu_thread_get_handle(cpu->thread);
+#endif
+    } else {
+        /* we share the thread */
+        cpu->thread = single_tcg_cpu_thread;
+        cpu->halt_cond = single_tcg_halt_cond;
+        cpu->thread_id = first_cpu->thread_id;
+        cpu->can_do_io = 1;
+        cpu->created = true;
+    }
+}
+
 const CpusAccel tcg_cpus_rr = {
-    .create_vcpu_thread = tcg_start_vcpu_thread,
+    .create_vcpu_thread = rr_start_vcpu_thread,
     .kick_vcpu_thread = qemu_cpu_kick_rr_cpus,
 
     .handle_interrupt = tcg_handle_interrupt,

accel/tcg/tcg-cpus-rr.h

@@ -15,6 +15,7 @@
 /* Kick all RR vCPUs. */
 void qemu_cpu_kick_rr_cpus(CPUState *unused);
 
-void *tcg_rr_cpu_thread_fn(void *arg);
+/* start the round robin vcpu thread */
+void rr_start_vcpu_thread(CPUState *cpu);
 
 #endif /* TCG_CPUS_RR_H */

accel/tcg/tcg-cpus.c

@@ -35,67 +35,9 @@
 #include "hw/boards.h"
 
 #include "tcg-cpus.h"
-#include "tcg-cpus-mttcg.h"
-#include "tcg-cpus-rr.h"
 
 /* common functionality among all TCG variants */
 
-void tcg_start_vcpu_thread(CPUState *cpu)
-{
-    char thread_name[VCPU_THREAD_NAME_SIZE];
-    static QemuCond *single_tcg_halt_cond;
-    static QemuThread *single_tcg_cpu_thread;
-    static int tcg_region_inited;
-
-    assert(tcg_enabled());
-    /*
-     * Initialize TCG regions--once. Now is a good time, because:
-     * (1) TCG's init context, prologue and target globals have been set up.
-     * (2) qemu_tcg_mttcg_enabled() works now (TCG init code runs before the
-     *     -accel flag is processed, so the check doesn't work then).
-     */
-    if (!tcg_region_inited) {
-        tcg_region_inited = 1;
-        tcg_region_init();
-        parallel_cpus = qemu_tcg_mttcg_enabled() && current_machine->smp.max_cpus > 1;
-    }
-
-    if (qemu_tcg_mttcg_enabled() || !single_tcg_cpu_thread) {
-        cpu->thread = g_malloc0(sizeof(QemuThread));
-        cpu->halt_cond = g_malloc0(sizeof(QemuCond));
-        qemu_cond_init(cpu->halt_cond);
-
-        if (qemu_tcg_mttcg_enabled()) {
-            /* create a thread per vCPU with TCG (MTTCG) */
-            snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/TCG",
-                 cpu->cpu_index);
-
-            qemu_thread_create(cpu->thread, thread_name, tcg_cpu_thread_fn,
-                               cpu, QEMU_THREAD_JOINABLE);
-
-        } else {
-            /* share a single thread for all cpus with TCG */
-            snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "ALL CPUs/TCG");
-            qemu_thread_create(cpu->thread, thread_name,
-                               tcg_rr_cpu_thread_fn,
-                               cpu, QEMU_THREAD_JOINABLE);
-
-            single_tcg_halt_cond = cpu->halt_cond;
-            single_tcg_cpu_thread = cpu->thread;
-        }
-#ifdef _WIN32
-        cpu->hThread = qemu_thread_get_handle(cpu->thread);
-#endif
-    } else {
-        /* For non-MTTCG cases we share the thread */
-        cpu->thread = single_tcg_cpu_thread;
-        cpu->halt_cond = single_tcg_halt_cond;
-        cpu->thread_id = first_cpu->thread_id;
-        cpu->can_do_io = 1;
-        cpu->created = true;
-    }
-}
-
 void qemu_tcg_destroy_vcpu(CPUState *cpu)
 {
     cpu_thread_signal_destroyed(cpu);

accel/tcg/tcg-cpus.h

@@ -18,7 +18,6 @@ extern const CpusAccel tcg_cpus_mttcg;
 extern const CpusAccel tcg_cpus_icount;
 extern const CpusAccel tcg_cpus_rr;
 
-void tcg_start_vcpu_thread(CPUState *cpu);
 void qemu_tcg_destroy_vcpu(CPUState *cpu);
 int tcg_cpu_exec(CPUState *cpu);
 void tcg_handle_interrupt(CPUState *cpu, int mask);