vdi: Handle failure for potentially large allocations

[qemu.git] / thread-pool.c

diff --git a/thread-pool.c b/thread-pool.c
index 204f70b7b55f639f3e117c98795536b6b9595cf6..23888dcfc4d74d9c9343462271dd15cb6eaa926d 100644 (file)
--- a/thread-pool.c
+++ b/thread-pool.c
@@ -15,16 +15,16 @@
  * GNU GPL, version 2 or (at your option) any later version.
  */
 #include "qemu-common.h"
  * GNU GPL, version 2 or (at your option) any later version.
  */
 #include "qemu-common.h"

-#include "qemu-queue.h"
-#include "qemu-thread.h"
-#include "osdep.h"
-#include "qemu-coroutine.h"
+#include "qemu/queue.h"
+#include "qemu/thread.h"
+#include "qemu/osdep.h"
+#include "block/coroutine.h"

 #include "trace.h"
 #include "trace.h"

-#include "block_int.h"
-#include "event_notifier.h"
-#include "thread-pool.h"
+#include "block/block_int.h"
+#include "block/thread-pool.h"
+#include "qemu/main-loop.h"

 
 

-static void do_spawn_thread(void);
+static void do_spawn_thread(ThreadPool *pool);

 
 typedef struct ThreadPoolElement ThreadPoolElement;
 
 
 typedef struct ThreadPoolElement ThreadPoolElement;
 


@@ -37,6 +37,7 @@ enum ThreadState {
 
 struct ThreadPoolElement {
     BlockDriverAIOCB common;
 
 struct ThreadPoolElement {
     BlockDriverAIOCB common;

+    ThreadPool *pool;

     ThreadPoolFunc *func;
     void *arg;
 
     ThreadPoolFunc *func;
     void *arg;
 


@@ -54,49 +55,56 @@ struct ThreadPoolElement {
     QLIST_ENTRY(ThreadPoolElement) all;
 };
 
     QLIST_ENTRY(ThreadPoolElement) all;
 };
 

-static EventNotifier notifier;
-static QemuMutex lock;
-static QemuCond check_cancel;
-static QemuSemaphore sem;
-static int max_threads = 64;
-static QEMUBH *new_thread_bh;
-
-/* The following variables are protected by the global mutex.  */
-static QLIST_HEAD(, ThreadPoolElement) head;
-
-/* The following variables are protected by lock.  */
-static QTAILQ_HEAD(, ThreadPoolElement) request_list;
-static int cur_threads;
-static int idle_threads;
-static int new_threads;     /* backlog of threads we need to create */
-static int pending_threads; /* threads created but not running yet */
-static int pending_cancellations; /* whether we need a cond_broadcast */
-
-static void *worker_thread(void *unused)
+struct ThreadPool {
+    AioContext *ctx;
+    QEMUBH *completion_bh;
+    QemuMutex lock;
+    QemuCond check_cancel;
+    QemuCond worker_stopped;
+    QemuSemaphore sem;
+    int max_threads;
+    QEMUBH *new_thread_bh;
+
+    /* The following variables are only accessed from one AioContext. */
+    QLIST_HEAD(, ThreadPoolElement) head;
+
+    /* The following variables are protected by lock.  */
+    QTAILQ_HEAD(, ThreadPoolElement) request_list;
+    int cur_threads;
+    int idle_threads;
+    int new_threads;     /* backlog of threads we need to create */
+    int pending_threads; /* threads created but not running yet */
+    int pending_cancellations; /* whether we need a cond_broadcast */
+    bool stopping;
+};
+
+static void *worker_thread(void *opaque)

 {
 {

-    qemu_mutex_lock(&lock);
-    pending_threads--;
-    do_spawn_thread();
+    ThreadPool *pool = opaque;

 
 

-    while (1) {
+    qemu_mutex_lock(&pool->lock);
+    pool->pending_threads--;
+    do_spawn_thread(pool);
+
+    while (!pool->stopping) {

         ThreadPoolElement *req;
         int ret;
 
         do {
         ThreadPoolElement *req;
         int ret;
 
         do {

-            idle_threads++;
-            qemu_mutex_unlock(&lock);
-            ret = qemu_sem_timedwait(&sem, 10000);
-            qemu_mutex_lock(&lock);
-            idle_threads--;
-        } while (ret == -1 && !QTAILQ_EMPTY(&request_list));
-        if (ret == -1) {
+            pool->idle_threads++;
+            qemu_mutex_unlock(&pool->lock);
+            ret = qemu_sem_timedwait(&pool->sem, 10000);
+            qemu_mutex_lock(&pool->lock);
+            pool->idle_threads--;
+        } while (ret == -1 && !QTAILQ_EMPTY(&pool->request_list));
+        if (ret == -1 || pool->stopping) {

             break;
         }
 
             break;
         }
 

-        req = QTAILQ_FIRST(&request_list);
-        QTAILQ_REMOVE(&request_list, req, reqs);
+        req = QTAILQ_FIRST(&pool->request_list);
+        QTAILQ_REMOVE(&pool->request_list, req, reqs);

         req->state = THREAD_ACTIVE;
         req->state = THREAD_ACTIVE;

-        qemu_mutex_unlock(&lock);
+        qemu_mutex_unlock(&pool->lock);

 
         ret = req->func(req->arg);
 
 
         ret = req->func(req->arg);
 


@@ -105,45 +113,48 @@ static void *worker_thread(void *unused)
         smp_wmb();
         req->state = THREAD_DONE;
 
         smp_wmb();
         req->state = THREAD_DONE;
 

-        qemu_mutex_lock(&lock);
-        if (pending_cancellations) {
-            qemu_cond_broadcast(&check_cancel);
+        qemu_mutex_lock(&pool->lock);
+        if (pool->pending_cancellations) {
+            qemu_cond_broadcast(&pool->check_cancel);

         }
 
         }
 

-        event_notifier_set(&notifier);
+        qemu_bh_schedule(pool->completion_bh);

     }
 
     }
 

-    cur_threads--;
-    qemu_mutex_unlock(&lock);
+    pool->cur_threads--;
+    qemu_cond_signal(&pool->worker_stopped);
+    qemu_mutex_unlock(&pool->lock);

     return NULL;
 }
 
     return NULL;
 }
 

-static void do_spawn_thread(void)
+static void do_spawn_thread(ThreadPool *pool)

 {
     QemuThread t;
 
     /* Runs with lock taken.  */
 {
     QemuThread t;
 
     /* Runs with lock taken.  */

-    if (!new_threads) {
+    if (!pool->new_threads) {

         return;
     }
 
         return;
     }
 

-    new_threads--;
-    pending_threads++;
+    pool->new_threads--;
+    pool->pending_threads++;

 
 

-    qemu_thread_create(&t, worker_thread, NULL, QEMU_THREAD_DETACHED);
+    qemu_thread_create(&t, "worker", worker_thread, pool, QEMU_THREAD_DETACHED);

 }
 
 static void spawn_thread_bh_fn(void *opaque)
 {
 }
 
 static void spawn_thread_bh_fn(void *opaque)
 {

-    qemu_mutex_lock(&lock);
-    do_spawn_thread();
-    qemu_mutex_unlock(&lock);
+    ThreadPool *pool = opaque;
+
+    qemu_mutex_lock(&pool->lock);
+    do_spawn_thread(pool);
+    qemu_mutex_unlock(&pool->lock);

 }
 
 }
 

-static void spawn_thread(void)
+static void spawn_thread(ThreadPool *pool)

 {
 {

-    cur_threads++;
-    new_threads++;
+    pool->cur_threads++;
+    pool->new_threads++;

     /* If there are threads being created, they will spawn new workers, so
      * we don't spend time creating many threads in a loop holding a mutex or
      * starving the current vcpu.
     /* If there are threads being created, they will spawn new workers, so
      * we don't spend time creating many threads in a loop holding a mutex or
      * starving the current vcpu.


@@ -151,28 +162,35 @@ static void spawn_thread(void)
      * If there are no idle threads, ask the main thread to create one, so we
      * inherit the correct affinity instead of the vcpu affinity.
      */
      * If there are no idle threads, ask the main thread to create one, so we
      * inherit the correct affinity instead of the vcpu affinity.
      */

-    if (!pending_threads) {
-        qemu_bh_schedule(new_thread_bh);
+    if (!pool->pending_threads) {
+        qemu_bh_schedule(pool->new_thread_bh);

     }
 }
 
     }
 }
 

-static void event_notifier_ready(EventNotifier *notifier)
+static void thread_pool_completion_bh(void *opaque)

 {
 {

+    ThreadPool *pool = opaque;

     ThreadPoolElement *elem, *next;
 
     ThreadPoolElement *elem, *next;
 

-    event_notifier_test_and_clear(notifier);

 restart:
 restart:

-    QLIST_FOREACH_SAFE(elem, &head, all, next) {
+    QLIST_FOREACH_SAFE(elem, &pool->head, all, next) {

         if (elem->state != THREAD_CANCELED && elem->state != THREAD_DONE) {
             continue;
         }
         if (elem->state == THREAD_DONE) {
         if (elem->state != THREAD_CANCELED && elem->state != THREAD_DONE) {
             continue;
         }
         if (elem->state == THREAD_DONE) {

-            trace_thread_pool_complete(elem, elem->common.opaque, elem->ret);
+            trace_thread_pool_complete(pool, elem, elem->common.opaque,
+                                       elem->ret);

         }
         if (elem->state == THREAD_DONE && elem->common.cb) {
             QLIST_REMOVE(elem, all);
             /* Read state before ret.  */
             smp_rmb();
         }
         if (elem->state == THREAD_DONE && elem->common.cb) {
             QLIST_REMOVE(elem, all);
             /* Read state before ret.  */
             smp_rmb();

+
+            /* Schedule ourselves in case elem->common.cb() calls aio_poll() to
+             * wait for another request that completed at the same time.
+             */
+            qemu_bh_schedule(pool->completion_bh);
+

             elem->common.cb(elem->common.opaque, elem->ret);
             qemu_aio_release(elem);
             goto restart;
             elem->common.cb(elem->common.opaque, elem->ret);
             qemu_aio_release(elem);
             goto restart;


@@ -184,36 +202,33 @@ restart:
     }
 }
 
     }
 }
 

-static int thread_pool_active(EventNotifier *notifier)
-{
-    return !QLIST_EMPTY(&head);
-}
-

 static void thread_pool_cancel(BlockDriverAIOCB *acb)
 {
     ThreadPoolElement *elem = (ThreadPoolElement *)acb;
 static void thread_pool_cancel(BlockDriverAIOCB *acb)
 {
     ThreadPoolElement *elem = (ThreadPoolElement *)acb;

+    ThreadPool *pool = elem->pool;

 
     trace_thread_pool_cancel(elem, elem->common.opaque);
 
 
     trace_thread_pool_cancel(elem, elem->common.opaque);
 

-    qemu_mutex_lock(&lock);
+    qemu_mutex_lock(&pool->lock);

     if (elem->state == THREAD_QUEUED &&
         /* No thread has yet started working on elem. we can try to "steal"
          * the item from the worker if we can get a signal from the
          * semaphore.  Because this is non-blocking, we can do it with
          * the lock taken and ensure that elem will remain THREAD_QUEUED.
          */
     if (elem->state == THREAD_QUEUED &&
         /* No thread has yet started working on elem. we can try to "steal"
          * the item from the worker if we can get a signal from the
          * semaphore.  Because this is non-blocking, we can do it with
          * the lock taken and ensure that elem will remain THREAD_QUEUED.
          */

-        qemu_sem_timedwait(&sem, 0) == 0) {
-        QTAILQ_REMOVE(&request_list, elem, reqs);
+        qemu_sem_timedwait(&pool->sem, 0) == 0) {
+        QTAILQ_REMOVE(&pool->request_list, elem, reqs);

         elem->state = THREAD_CANCELED;
         elem->state = THREAD_CANCELED;

-        event_notifier_set(&notifier);
+        qemu_bh_schedule(pool->completion_bh);

     } else {
     } else {

-        pending_cancellations++;
+        pool->pending_cancellations++;

         while (elem->state != THREAD_CANCELED && elem->state != THREAD_DONE) {
         while (elem->state != THREAD_CANCELED && elem->state != THREAD_DONE) {

-            qemu_cond_wait(&check_cancel, &lock);
+            qemu_cond_wait(&pool->check_cancel, &pool->lock);

         }
         }

-        pending_cancellations--;
+        pool->pending_cancellations--;

     }
     }

-    qemu_mutex_unlock(&lock);
+    qemu_mutex_unlock(&pool->lock);
+    thread_pool_completion_bh(pool);

 }
 
 static const AIOCBInfo thread_pool_aiocb_info = {
 }
 
 static const AIOCBInfo thread_pool_aiocb_info = {


@@ -221,7 +236,8 @@ static const AIOCBInfo thread_pool_aiocb_info = {
     .cancel             = thread_pool_cancel,
 };
 
     .cancel             = thread_pool_cancel,
 };
 

-BlockDriverAIOCB *thread_pool_submit_aio(ThreadPoolFunc *func, void *arg,
+BlockDriverAIOCB *thread_pool_submit_aio(ThreadPool *pool,
+        ThreadPoolFunc *func, void *arg,

         BlockDriverCompletionFunc *cb, void *opaque)
 {
     ThreadPoolElement *req;
         BlockDriverCompletionFunc *cb, void *opaque)
 {
     ThreadPoolElement *req;


@@ -230,18 +246,19 @@ BlockDriverAIOCB *thread_pool_submit_aio(ThreadPoolFunc *func, void *arg,
     req->func = func;
     req->arg = arg;
     req->state = THREAD_QUEUED;
     req->func = func;
     req->arg = arg;
     req->state = THREAD_QUEUED;

+    req->pool = pool;

 
 

-    QLIST_INSERT_HEAD(&head, req, all);
+    QLIST_INSERT_HEAD(&pool->head, req, all);

 
 

-    trace_thread_pool_submit(req, arg);
+    trace_thread_pool_submit(pool, req, arg);

 
 

-    qemu_mutex_lock(&lock);
-    if (idle_threads == 0 && cur_threads < max_threads) {
-        spawn_thread();
+    qemu_mutex_lock(&pool->lock);
+    if (pool->idle_threads == 0 && pool->cur_threads < pool->max_threads) {
+        spawn_thread(pool);

     }
     }

-    QTAILQ_INSERT_TAIL(&request_list, req, reqs);
-    qemu_mutex_unlock(&lock);
-    qemu_sem_post(&sem);
+    QTAILQ_INSERT_TAIL(&pool->request_list, req, reqs);
+    qemu_mutex_unlock(&pool->lock);
+    qemu_sem_post(&pool->sem);

     return &req->common;
 }
 
     return &req->common;
 }
 


@@ -258,32 +275,76 @@ static void thread_pool_co_cb(void *opaque, int ret)
     qemu_coroutine_enter(co->co, NULL);
 }
 
     qemu_coroutine_enter(co->co, NULL);
 }
 

-int coroutine_fn thread_pool_submit_co(ThreadPoolFunc *func, void *arg)
+int coroutine_fn thread_pool_submit_co(ThreadPool *pool, ThreadPoolFunc *func,
+                                       void *arg)

 {
     ThreadPoolCo tpc = { .co = qemu_coroutine_self(), .ret = -EINPROGRESS };
     assert(qemu_in_coroutine());
 {
     ThreadPoolCo tpc = { .co = qemu_coroutine_self(), .ret = -EINPROGRESS };
     assert(qemu_in_coroutine());

-    thread_pool_submit_aio(func, arg, thread_pool_co_cb, &tpc);
+    thread_pool_submit_aio(pool, func, arg, thread_pool_co_cb, &tpc);

     qemu_coroutine_yield();
     return tpc.ret;
 }
 
     qemu_coroutine_yield();
     return tpc.ret;
 }
 

-void thread_pool_submit(ThreadPoolFunc *func, void *arg)
+void thread_pool_submit(ThreadPool *pool, ThreadPoolFunc *func, void *arg)

 {
 {

-    thread_pool_submit_aio(func, arg, NULL, NULL);
+    thread_pool_submit_aio(pool, func, arg, NULL, NULL);

 }
 
 }
 

-static void thread_pool_init(void)
+static void thread_pool_init_one(ThreadPool *pool, AioContext *ctx)

 {
 {

-    QLIST_INIT(&head);
-    event_notifier_init(&notifier, false);
-    qemu_mutex_init(&lock);
-    qemu_cond_init(&check_cancel);
-    qemu_sem_init(&sem, 0);
-    qemu_aio_set_event_notifier(&notifier, event_notifier_ready,
-                                thread_pool_active);
-
-    QTAILQ_INIT(&request_list);
-    new_thread_bh = qemu_bh_new(spawn_thread_bh_fn, NULL);
+    if (!ctx) {
+        ctx = qemu_get_aio_context();
+    }
+
+    memset(pool, 0, sizeof(*pool));
+    pool->ctx = ctx;
+    pool->completion_bh = aio_bh_new(ctx, thread_pool_completion_bh, pool);
+    qemu_mutex_init(&pool->lock);
+    qemu_cond_init(&pool->check_cancel);
+    qemu_cond_init(&pool->worker_stopped);
+    qemu_sem_init(&pool->sem, 0);
+    pool->max_threads = 64;
+    pool->new_thread_bh = aio_bh_new(ctx, spawn_thread_bh_fn, pool);
+
+    QLIST_INIT(&pool->head);
+    QTAILQ_INIT(&pool->request_list);

 }
 
 }
 

-block_init(thread_pool_init)
+ThreadPool *thread_pool_new(AioContext *ctx)
+{
+    ThreadPool *pool = g_new(ThreadPool, 1);
+    thread_pool_init_one(pool, ctx);
+    return pool;
+}
+
+void thread_pool_free(ThreadPool *pool)
+{
+    if (!pool) {
+        return;
+    }
+
+    assert(QLIST_EMPTY(&pool->head));
+
+    qemu_mutex_lock(&pool->lock);
+
+    /* Stop new threads from spawning */
+    qemu_bh_delete(pool->new_thread_bh);
+    pool->cur_threads -= pool->new_threads;
+    pool->new_threads = 0;
+
+    /* Wait for worker threads to terminate */
+    pool->stopping = true;
+    while (pool->cur_threads > 0) {
+        qemu_sem_post(&pool->sem);
+        qemu_cond_wait(&pool->worker_stopped, &pool->lock);
+    }
+
+    qemu_mutex_unlock(&pool->lock);
+
+    qemu_bh_delete(pool->completion_bh);
+    qemu_sem_destroy(&pool->sem);
+    qemu_cond_destroy(&pool->check_cancel);
+    qemu_cond_destroy(&pool->worker_stopped);
+    qemu_mutex_destroy(&pool->lock);
+    g_free(pool);
+}