cpu: Reclaim vCPU objects

In order to deal well with the kvm vcpus (which can not be removed without any
protection), we do not close KVM vcpu fd, just record and mark it as stopped
into a list, so that we can reuse it for the appending cpu hot-add request if
possible. It is also the approach that kvm guys suggested:
https://www.mail-archive.com/kvm@vger.kernel.org/msg102839.html

Signed-off-by: Chen Fan <chen.fan.fnst@cn.fujitsu.com>
Signed-off-by: Gu Zheng <guz.fnst@cn.fujitsu.com>
Signed-off-by: Zhu Guihua <zhugh.fnst@cn.fujitsu.com>
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
               [- Explicit CPU_REMOVE() from qemu_kvm/tcg_destroy_vcpu()
                  isn't needed as it is done from cpu_exec_exit()
                - Use iothread mutex instead of global mutex during
                  destroy
                - Don't cleanup vCPU object from vCPU thread context
                  but leave it to the callers (device_add/device_del)]
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

cpus.c

@@ -972,6 +972,18 @@ void async_run_on_cpu(CPUState *cpu, void (*func)(void *data), void *data)
     qemu_cpu_kick(cpu);
 }
 
+static void qemu_kvm_destroy_vcpu(CPUState *cpu)
+{
+    if (kvm_destroy_vcpu(cpu) < 0) {
+        error_report("kvm_destroy_vcpu failed");
+        exit(EXIT_FAILURE);
+    }
+}
+
+static void qemu_tcg_destroy_vcpu(CPUState *cpu)
+{
+}
+
 static void flush_queued_work(CPUState *cpu)
 {
     struct qemu_work_item *wi;
@@ -1061,7 +1073,7 @@ static void *qemu_kvm_cpu_thread_fn(void *arg)
     cpu->created = true;
     qemu_cond_signal(&qemu_cpu_cond);
 
-    while (1) {
+    do {
         if (cpu_can_run(cpu)) {
             r = kvm_cpu_exec(cpu);
             if (r == EXCP_DEBUG) {
@@ -1069,8 +1081,10 @@ static void *qemu_kvm_cpu_thread_fn(void *arg)
             }
         }
         qemu_kvm_wait_io_event(cpu);
-    }
+    } while (!cpu->unplug || cpu_can_run(cpu));
 
+    qemu_kvm_destroy_vcpu(cpu);
+    qemu_mutex_unlock_iothread();
     return NULL;
 }
 
@@ -1124,6 +1138,7 @@ static void tcg_exec_all(void);
 static void *qemu_tcg_cpu_thread_fn(void *arg)
 {
     CPUState *cpu = arg;
+    CPUState *remove_cpu = NULL;
 
     rcu_register_thread();
 
@@ -1161,6 +1176,16 @@ static void *qemu_tcg_cpu_thread_fn(void *arg)
             }
         }
         qemu_tcg_wait_io_event(QTAILQ_FIRST(&cpus));
+        CPU_FOREACH(cpu) {
+            if (cpu->unplug && !cpu_can_run(cpu)) {
+                remove_cpu = cpu;
+                break;
+            }
+        }
+        if (remove_cpu) {
+            qemu_tcg_destroy_vcpu(remove_cpu);
+            remove_cpu = NULL;
+        }
     }
 
     return NULL;
@@ -1317,6 +1342,13 @@ void resume_all_vcpus(void)
     }
 }
 
+void cpu_remove(CPUState *cpu)
+{
+    cpu->stop = true;
+    cpu->unplug = true;
+    qemu_cpu_kick(cpu);
+}
+
 /* For temporary buffers for forming a name */
 #define VCPU_THREAD_NAME_SIZE 16
 
@@ -1533,6 +1565,9 @@ static void tcg_exec_all(void)
                 break;
             }
         } else if (cpu->stop || cpu->stopped) {
+            if (cpu->unplug) {
+                next_cpu = CPU_NEXT(cpu);
+            }
             break;
         }
     }

include/qom/cpu.h

@@ -244,6 +244,7 @@ struct qemu_work_item {
  * @halted: Nonzero if the CPU is in suspended state.
  * @stop: Indicates a pending stop request.
  * @stopped: Indicates the CPU has been artificially stopped.
+ * @unplug: Indicates a pending CPU unplug request.
  * @crash_occurred: Indicates the OS reported a crash (panic) for this CPU
  * @tcg_exit_req: Set to force TCG to stop executing linked TBs for this
  *           CPU and return to its top level loop.
@@ -296,6 +297,7 @@ struct CPUState {
     bool created;
     bool stop;
     bool stopped;
+    bool unplug;
     bool crash_occurred;
     bool exit_request;
     bool tb_flushed;
@@ -762,6 +764,14 @@ void cpu_exit(CPUState *cpu);
  */
 void cpu_resume(CPUState *cpu);
 
+/**
+ * cpu_remove:
+ * @cpu: The CPU to remove.
+ *
+ * Requests the CPU to be removed.
+ */
+void cpu_remove(CPUState *cpu);
+
 /**
  * qemu_init_vcpu:
  * @cpu: The vCPU to initialize.

include/sysemu/kvm.h

@@ -216,6 +216,7 @@ int kvm_has_intx_set_mask(void);
 
 int kvm_init_vcpu(CPUState *cpu);
 int kvm_cpu_exec(CPUState *cpu);
+int kvm_destroy_vcpu(CPUState *cpu);
 
 #ifdef NEED_CPU_H
 #include "cpu.h"

kvm-all.c

@@ -61,6 +61,12 @@
 
 #define KVM_MSI_HASHTAB_SIZE    256
 
+struct KVMParkedVcpu {
+    unsigned long vcpu_id;
+    int kvm_fd;
+    QLIST_ENTRY(KVMParkedVcpu) node;
+};
+
 struct KVMState
 {
     AccelState parent_obj;
@@ -94,6 +100,7 @@ struct KVMState
     QTAILQ_HEAD(msi_hashtab, KVMMSIRoute) msi_hashtab[KVM_MSI_HASHTAB_SIZE];
 #endif
     KVMMemoryListener memory_listener;
+    QLIST_HEAD(, KVMParkedVcpu) kvm_parked_vcpus;
 };
 
 KVMState *kvm_state;
@@ -237,6 +244,53 @@ static int kvm_set_user_memory_region(KVMMemoryListener *kml, KVMSlot *slot)
     return kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, &mem);
 }
 
+int kvm_destroy_vcpu(CPUState *cpu)
+{
+    KVMState *s = kvm_state;
+    long mmap_size;
+    struct KVMParkedVcpu *vcpu = NULL;
+    int ret = 0;
+
+    DPRINTF("kvm_destroy_vcpu\n");
+
+    mmap_size = kvm_ioctl(s, KVM_GET_VCPU_MMAP_SIZE, 0);
+    if (mmap_size < 0) {
+        ret = mmap_size;
+        DPRINTF("KVM_GET_VCPU_MMAP_SIZE failed\n");
+        goto err;
+    }
+
+    ret = munmap(cpu->kvm_run, mmap_size);
+    if (ret < 0) {
+        goto err;
+    }
+
+    vcpu = g_malloc0(sizeof(*vcpu));
+    vcpu->vcpu_id = kvm_arch_vcpu_id(cpu);
+    vcpu->kvm_fd = cpu->kvm_fd;
+    QLIST_INSERT_HEAD(&kvm_state->kvm_parked_vcpus, vcpu, node);
+err:
+    return ret;
+}
+
+static int kvm_get_vcpu(KVMState *s, unsigned long vcpu_id)
+{
+    struct KVMParkedVcpu *cpu;
+
+    QLIST_FOREACH(cpu, &s->kvm_parked_vcpus, node) {
+        if (cpu->vcpu_id == vcpu_id) {
+            int kvm_fd;
+
+            QLIST_REMOVE(cpu, node);
+            kvm_fd = cpu->kvm_fd;
+            g_free(cpu);
+            return kvm_fd;
+        }
+    }
+
+    return kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)vcpu_id);
+}
+
 int kvm_init_vcpu(CPUState *cpu)
 {
     KVMState *s = kvm_state;
@@ -245,7 +299,7 @@ int kvm_init_vcpu(CPUState *cpu)
 
     DPRINTF("kvm_init_vcpu\n");
 
-    ret = kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)kvm_arch_vcpu_id(cpu));
+    ret = kvm_get_vcpu(s, kvm_arch_vcpu_id(cpu));
     if (ret < 0) {
         DPRINTF("kvm_create_vcpu failed\n");
         goto err;
@@ -1501,6 +1555,7 @@ static int kvm_init(MachineState *ms)
 #ifdef KVM_CAP_SET_GUEST_DEBUG
     QTAILQ_INIT(&s->kvm_sw_breakpoints);
 #endif
+    QLIST_INIT(&s->kvm_parked_vcpus);
     s->vmfd = -1;
     s->fd = qemu_open("/dev/kvm", O_RDWR);
     if (s->fd == -1) {

kvm-stub.c

@@ -32,6 +32,11 @@ bool kvm_allowed;
 bool kvm_readonly_mem_allowed;
 bool kvm_ioeventfd_any_length_allowed;
 
+int kvm_destroy_vcpu(CPUState *cpu)
+{
+    return -ENOSYS;
+}
+
 int kvm_init_vcpu(CPUState *cpu)
 {
     return -ENOSYS;