add win32 qemu-thread implementation

For now, qemu_cond_timedwait and qemu_mutex_timedlock are left as
POSIX-only functions.  They can be removed later, once the patches
that remove their uses are in.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Makefile.objs

@@ -142,8 +142,8 @@ endif
 common-obj-y += $(addprefix ui/, $(ui-obj-y))
 
 common-obj-y += iov.o acl.o
-common-obj-$(CONFIG_THREAD) += qemu-thread.o
-common-obj-$(CONFIG_POSIX) += compatfd.o
+common-obj-$(CONFIG_POSIX) += qemu-thread-posix.o compatfd.o
+common-obj-$(CONFIG_WIN32) += qemu-thread-win32.o
 common-obj-y += notify.o event_notifier.o
 common-obj-y += qemu-timer.o qemu-timer-common.o
 

qemu-thread-posix.h

@@ -0,0 +1,18 @@
+#ifndef __QEMU_THREAD_POSIX_H
+#define __QEMU_THREAD_POSIX_H 1
+#include "pthread.h"
+
+struct QemuMutex {
+    pthread_mutex_t lock;
+};
+
+struct QemuCond {
+    pthread_cond_t cond;
+};
+
+struct QemuThread {
+    pthread_t thread;
+};
+
+void qemu_thread_signal(QemuThread *thread, int sig);
+#endif

qemu-thread-win32.c

@@ -0,0 +1,260 @@
+/*
+ * Win32 implementation for mutex/cond/thread functions
+ *
+ * Copyright Red Hat, Inc. 2010
+ *
+ * Author:
+ *  Paolo Bonzini <pbonzini@redhat.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+#include "qemu-common.h"
+#include "qemu-thread.h"
+#include <process.h>
+#include <assert.h>
+#include <limits.h>
+
+static void error_exit(int err, const char *msg)
+{
+    char *pstr;
+
+    FormatMessage(FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_ALLOCATE_BUFFER,
+                  NULL, err, 0, (LPTSTR)&pstr, 2, NULL);
+    fprintf(stderr, "qemu: %s: %s\n", msg, pstr);
+    LocalFree(pstr);
+    exit(1);
+}
+
+void qemu_mutex_init(QemuMutex *mutex)
+{
+    mutex->owner = 0;
+    InitializeCriticalSection(&mutex->lock);
+}
+
+void qemu_mutex_lock(QemuMutex *mutex)
+{
+    EnterCriticalSection(&mutex->lock);
+
+    /* Win32 CRITICAL_SECTIONs are recursive.  Assert that we're not
+     * using them as such.
+     */
+    assert(mutex->owner == 0);
+    mutex->owner = GetCurrentThreadId();
+}
+
+int qemu_mutex_trylock(QemuMutex *mutex)
+{
+    int owned;
+
+    owned = TryEnterCriticalSection(&mutex->lock);
+    if (owned) {
+        assert(mutex->owner == 0);
+        mutex->owner = GetCurrentThreadId();
+    }
+    return !owned;
+}
+
+void qemu_mutex_unlock(QemuMutex *mutex)
+{
+    assert(mutex->owner == GetCurrentThreadId());
+    mutex->owner = 0;
+    LeaveCriticalSection(&mutex->lock);
+}
+
+void qemu_cond_init(QemuCond *cond)
+{
+    memset(cond, 0, sizeof(*cond));
+
+    cond->sema = CreateSemaphore(NULL, 0, LONG_MAX, NULL);
+    if (!cond->sema) {
+        error_exit(GetLastError(), __func__);
+    }
+    cond->continue_event = CreateEvent(NULL,    /* security */
+                                       FALSE,   /* auto-reset */
+                                       FALSE,   /* not signaled */
+                                       NULL);   /* name */
+    if (!cond->continue_event) {
+        error_exit(GetLastError(), __func__);
+    }
+}
+
+void qemu_cond_signal(QemuCond *cond)
+{
+    DWORD result;
+
+    /*
+     * Signal only when there are waiters.  cond->waiters is
+     * incremented by pthread_cond_wait under the external lock,
+     * so we are safe about that.
+     */
+    if (cond->waiters == 0) {
+        return;
+    }
+
+    /*
+     * Waiting threads decrement it outside the external lock, but
+     * only if another thread is executing pthread_cond_broadcast and
+     * has the mutex.  So, it also cannot be decremented concurrently
+     * with this particular access.
+     */
+    cond->target = cond->waiters - 1;
+    result = SignalObjectAndWait(cond->sema, cond->continue_event,
+                                 INFINITE, FALSE);
+    if (result == WAIT_ABANDONED || result == WAIT_FAILED) {
+        error_exit(GetLastError(), __func__);
+    }
+}
+
+void qemu_cond_broadcast(QemuCond *cond)
+{
+    BOOLEAN result;
+    /*
+     * As in pthread_cond_signal, access to cond->waiters and
+     * cond->target is locked via the external mutex.
+     */
+    if (cond->waiters == 0) {
+        return;
+    }
+
+    cond->target = 0;
+    result = ReleaseSemaphore(cond->sema, cond->waiters, NULL);
+    if (!result) {
+        error_exit(GetLastError(), __func__);
+    }
+
+    /*
+     * At this point all waiters continue. Each one takes its
+     * slice of the semaphore. Now it's our turn to wait: Since
+     * the external mutex is held, no thread can leave cond_wait,
+     * yet. For this reason, we can be sure that no thread gets
+     * a chance to eat *more* than one slice. OTOH, it means
+     * that the last waiter must send us a wake-up.
+     */
+    WaitForSingleObject(cond->continue_event, INFINITE);
+}
+
+void qemu_cond_wait(QemuCond *cond, QemuMutex *mutex)
+{
+    /*
+     * This access is protected under the mutex.
+     */
+    cond->waiters++;
+
+    /*
+     * Unlock external mutex and wait for signal.
+     * NOTE: we've held mutex locked long enough to increment
+     * waiters count above, so there's no problem with
+     * leaving mutex unlocked before we wait on semaphore.
+     */
+    qemu_mutex_unlock(mutex);
+    WaitForSingleObject(cond->sema, INFINITE);
+
+    /* Now waiters must rendez-vous with the signaling thread and
+     * let it continue.  For cond_broadcast this has heavy contention
+     * and triggers thundering herd.  So goes life.
+     *
+     * Decrease waiters count.  The mutex is not taken, so we have
+     * to do this atomically.
+     *
+     * All waiters contend for the mutex at the end of this function
+     * until the signaling thread relinquishes it.  To ensure
+     * each waiter consumes exactly one slice of the semaphore,
+     * the signaling thread stops until it is told by the last
+     * waiter that it can go on.
+     */
+    if (InterlockedDecrement(&cond->waiters) == cond->target) {
+        SetEvent(cond->continue_event);
+    }
+
+    qemu_mutex_lock(mutex);
+}
+
+struct QemuThreadData {
+    QemuThread *thread;
+    void *(*start_routine)(void *);
+    void *arg;
+};
+
+static int qemu_thread_tls_index = TLS_OUT_OF_INDEXES;
+
+static unsigned __stdcall win32_start_routine(void *arg)
+{
+    struct QemuThreadData data = *(struct QemuThreadData *) arg;
+    QemuThread *thread = data.thread;
+
+    free(arg);
+    TlsSetValue(qemu_thread_tls_index, thread);
+
+    /*
+     * Use DuplicateHandle instead of assigning thread->thread in the
+     * creating thread to avoid races.  It's simpler this way than with
+     * synchronization.
+     */
+    DuplicateHandle(GetCurrentProcess(), GetCurrentThread(),
+                    GetCurrentProcess(), &thread->thread,
+                    0, FALSE, DUPLICATE_SAME_ACCESS);
+
+    qemu_thread_exit(data.start_routine(data.arg));
+    abort();
+}
+
+void qemu_thread_exit(void *arg)
+{
+    QemuThread *thread = TlsGetValue(qemu_thread_tls_index);
+    thread->ret = arg;
+    CloseHandle(thread->thread);
+    thread->thread = NULL;
+    ExitThread(0);
+}
+
+static inline void qemu_thread_init(void)
+{
+    if (qemu_thread_tls_index == TLS_OUT_OF_INDEXES) {
+        qemu_thread_tls_index = TlsAlloc();
+        if (qemu_thread_tls_index == TLS_OUT_OF_INDEXES) {
+            error_exit(ERROR_NO_SYSTEM_RESOURCES, __func__);
+        }
+    }
+}
+
+
+void qemu_thread_create(QemuThread *thread,
+                       void *(*start_routine)(void *),
+                       void *arg)
+{
+    HANDLE hThread;
+
+    struct QemuThreadData *data;
+    qemu_thread_init();
+    data = qemu_malloc(sizeof *data);
+    data->thread = thread;
+    data->start_routine = start_routine;
+    data->arg = arg;
+
+    hThread = (HANDLE) _beginthreadex(NULL, 0, win32_start_routine,
+                                      data, 0, NULL);
+    if (!hThread) {
+        error_exit(GetLastError(), __func__);
+    }
+    CloseHandle(hThread);
+}
+
+void qemu_thread_get_self(QemuThread *thread)
+{
+    if (!thread->thread) {
+        /* In the main thread of the process.  Initialize the QemuThread
+           pointer in TLS, and use the dummy GetCurrentThread handle as
+           the identifier for qemu_thread_is_self.  */
+        qemu_thread_init();
+        TlsSetValue(qemu_thread_tls_index, thread);
+        thread->thread = GetCurrentThread();
+    }
+}
+
+int qemu_thread_is_self(QemuThread *thread)
+{
+    QemuThread *this_thread = TlsGetValue(qemu_thread_tls_index);
+    return this_thread->thread == thread->thread;
+}

qemu-thread-win32.h

@@ -0,0 +1,21 @@
+#ifndef __QEMU_THREAD_WIN32_H
+#define __QEMU_THREAD_WIN32_H 1
+#include "windows.h"
+
+struct QemuMutex {
+    CRITICAL_SECTION lock;
+    LONG owner;
+};
+
+struct QemuCond {
+    LONG waiters, target;
+    HANDLE sema;
+    HANDLE continue_event;
+};
+
+struct QemuThread {
+    HANDLE thread;
+    void *ret;
+};
+
+#endif

qemu-thread.h

@@ -1,24 +1,16 @@
 #ifndef __QEMU_THREAD_H
 #define __QEMU_THREAD_H 1
-#include "semaphore.h"
-#include "pthread.h"
-
-struct QemuMutex {
-    pthread_mutex_t lock;
-};
-
-struct QemuCond {
-    pthread_cond_t cond;
-};
-
-struct QemuThread {
-    pthread_t thread;
-};
 
 typedef struct QemuMutex QemuMutex;
 typedef struct QemuCond QemuCond;
 typedef struct QemuThread QemuThread;
 
+#ifdef _WIN32
+#include "qemu-thread-win32.h"
+#else
+#include "qemu-thread-posix.h"
+#endif
+
 void qemu_mutex_init(QemuMutex *mutex);
 void qemu_mutex_destroy(QemuMutex *mutex);
 void qemu_mutex_lock(QemuMutex *mutex);
@@ -28,6 +20,12 @@ void qemu_mutex_unlock(QemuMutex *mutex);
 
 void qemu_cond_init(QemuCond *cond);
 void qemu_cond_destroy(QemuCond *cond);
+
+/*
+ * IMPORTANT: The implementation does not guarantee that pthread_cond_signal
+ * and pthread_cond_broadcast can be called except while the same mutex is
+ * held as in the corresponding pthread_cond_wait calls!
+ */
 void qemu_cond_signal(QemuCond *cond);
 void qemu_cond_broadcast(QemuCond *cond);
 void qemu_cond_wait(QemuCond *cond, QemuMutex *mutex);
@@ -36,7 +34,6 @@ int qemu_cond_timedwait(QemuCond *cond, QemuMutex *mutex, uint64_t msecs);
 void qemu_thread_create(QemuThread *thread,
                        void *(*start_routine)(void*),
                        void *arg);
-void qemu_thread_signal(QemuThread *thread, int sig);
 void qemu_thread_get_self(QemuThread *thread);
 int qemu_thread_is_self(QemuThread *thread);
 void qemu_thread_exit(void *retval);