qemu/util/qemu-thread-posix.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

687 lines
17 KiB
C
Raw Normal View History

/*
* Wrappers around mutex/cond/thread functions
*
* Copyright Red Hat, Inc. 2009
*
* Author:
* Marcelo Tosatti <mtosatti@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/osdep.h"
#include "qemu/thread.h"
#include "qemu/atomic.h"
#include "qemu/notify.h"
#include "qemu-thread-common.h"
#include "qemu/tsan.h"
#include "qemu/bitmap.h"
#ifdef CONFIG_PTHREAD_SET_NAME_NP
#include <pthread_np.h>
#endif
static bool name_threads;
void qemu_thread_naming(bool enable)
{
name_threads = enable;
#if !defined CONFIG_PTHREAD_SETNAME_NP_W_TID && \
!defined CONFIG_PTHREAD_SETNAME_NP_WO_TID && \
!defined CONFIG_PTHREAD_SET_NAME_NP
/* This is a debugging option, not fatal */
if (enable) {
fprintf(stderr, "qemu: thread naming not supported on this host\n");
}
#endif
}
static void error_exit(int err, const char *msg)
{
fprintf(stderr, "qemu: %s: %s\n", msg, strerror(err));
abort();
}
static inline clockid_t qemu_timedwait_clockid(void)
{
#ifdef CONFIG_PTHREAD_CONDATTR_SETCLOCK
return CLOCK_MONOTONIC;
#else
return CLOCK_REALTIME;
#endif
}
static void compute_abs_deadline(struct timespec *ts, int ms)
{
clock_gettime(qemu_timedwait_clockid(), ts);
ts->tv_nsec += (ms % 1000) * 1000000;
ts->tv_sec += ms / 1000;
if (ts->tv_nsec >= 1000000000) {
ts->tv_sec++;
ts->tv_nsec -= 1000000000;
}
}
void qemu_mutex_init(QemuMutex *mutex)
{
int err;
err = pthread_mutex_init(&mutex->lock, NULL);
if (err)
error_exit(err, __func__);
qemu_mutex_post_init(mutex);
}
void qemu_mutex_destroy(QemuMutex *mutex)
{
int err;
assert(mutex->initialized);
mutex->initialized = false;
err = pthread_mutex_destroy(&mutex->lock);
if (err)
error_exit(err, __func__);
}
void qemu_mutex_lock_impl(QemuMutex *mutex, const char *file, const int line)
{
int err;
assert(mutex->initialized);
qemu_mutex_pre_lock(mutex, file, line);
err = pthread_mutex_lock(&mutex->lock);
if (err)
error_exit(err, __func__);
qemu_mutex_post_lock(mutex, file, line);
}
int qemu_mutex_trylock_impl(QemuMutex *mutex, const char *file, const int line)
{
int err;
assert(mutex->initialized);
err = pthread_mutex_trylock(&mutex->lock);
if (err == 0) {
qemu_mutex_post_lock(mutex, file, line);
return 0;
}
if (err != EBUSY) {
error_exit(err, __func__);
}
return -EBUSY;
}
void qemu_mutex_unlock_impl(QemuMutex *mutex, const char *file, const int line)
{
int err;
assert(mutex->initialized);
qemu_mutex_pre_unlock(mutex, file, line);
err = pthread_mutex_unlock(&mutex->lock);
if (err)
error_exit(err, __func__);
}
void qemu_rec_mutex_init(QemuRecMutex *mutex)
{
int err;
pthread_mutexattr_t attr;
pthread_mutexattr_init(&attr);
pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE);
err = pthread_mutex_init(&mutex->m.lock, &attr);
pthread_mutexattr_destroy(&attr);
if (err) {
error_exit(err, __func__);
}
mutex->m.initialized = true;
}
void qemu_rec_mutex_destroy(QemuRecMutex *mutex)
{
qemu_mutex_destroy(&mutex->m);
}
void qemu_rec_mutex_lock_impl(QemuRecMutex *mutex, const char *file, int line)
{
qemu_mutex_lock_impl(&mutex->m, file, line);
}
int qemu_rec_mutex_trylock_impl(QemuRecMutex *mutex, const char *file, int line)
{
return qemu_mutex_trylock_impl(&mutex->m, file, line);
}
void qemu_rec_mutex_unlock_impl(QemuRecMutex *mutex, const char *file, int line)
{
qemu_mutex_unlock_impl(&mutex->m, file, line);
}
void qemu_cond_init(QemuCond *cond)
{
pthread_condattr_t attr;
int err;
err = pthread_condattr_init(&attr);
if (err) {
error_exit(err, __func__);
}
#ifdef CONFIG_PTHREAD_CONDATTR_SETCLOCK
err = pthread_condattr_setclock(&attr, qemu_timedwait_clockid());
if (err) {
error_exit(err, __func__);
}
#endif
err = pthread_cond_init(&cond->cond, &attr);
if (err) {
error_exit(err, __func__);
}
err = pthread_condattr_destroy(&attr);
if (err) {
error_exit(err, __func__);
}
cond->initialized = true;
}
void qemu_cond_destroy(QemuCond *cond)
{
int err;
assert(cond->initialized);
cond->initialized = false;
err = pthread_cond_destroy(&cond->cond);
if (err)
error_exit(err, __func__);
}
void qemu_cond_signal(QemuCond *cond)
{
int err;
assert(cond->initialized);
err = pthread_cond_signal(&cond->cond);
if (err)
error_exit(err, __func__);
}
void qemu_cond_broadcast(QemuCond *cond)
{
int err;
assert(cond->initialized);
err = pthread_cond_broadcast(&cond->cond);
if (err)
error_exit(err, __func__);
}
void qemu_cond_wait_impl(QemuCond *cond, QemuMutex *mutex, const char *file, const int line)
{
int err;
assert(cond->initialized);
qemu_mutex_pre_unlock(mutex, file, line);
err = pthread_cond_wait(&cond->cond, &mutex->lock);
qemu_mutex_post_lock(mutex, file, line);
if (err)
error_exit(err, __func__);
}
static bool TSA_NO_TSA
qemu_cond_timedwait_ts(QemuCond *cond, QemuMutex *mutex, struct timespec *ts,
const char *file, const int line)
{
int err;
assert(cond->initialized);
trace_qemu_mutex_unlock(mutex, file, line);
err = pthread_cond_timedwait(&cond->cond, &mutex->lock, ts);
trace_qemu_mutex_locked(mutex, file, line);
if (err && err != ETIMEDOUT) {
error_exit(err, __func__);
}
return err != ETIMEDOUT;
}
bool qemu_cond_timedwait_impl(QemuCond *cond, QemuMutex *mutex, int ms,
const char *file, const int line)
{
struct timespec ts;
compute_abs_deadline(&ts, ms);
return qemu_cond_timedwait_ts(cond, mutex, &ts, file, line);
}
void qemu_sem_init(QemuSemaphore *sem, int init)
{
qemu_mutex_init(&sem->mutex);
qemu_cond_init(&sem->cond);
if (init < 0) {
error_exit(EINVAL, __func__);
}
sem->count = init;
}
void qemu_sem_destroy(QemuSemaphore *sem)
{
qemu_cond_destroy(&sem->cond);
qemu_mutex_destroy(&sem->mutex);
}
void qemu_sem_post(QemuSemaphore *sem)
{
qemu_mutex_lock(&sem->mutex);
if (sem->count == UINT_MAX) {
error_exit(EINVAL, __func__);
} else {
sem->count++;
qemu_cond_signal(&sem->cond);
}
qemu_mutex_unlock(&sem->mutex);
}
int qemu_sem_timedwait(QemuSemaphore *sem, int ms)
{
bool rc = true;
struct timespec ts;
compute_abs_deadline(&ts, ms);
qemu_mutex_lock(&sem->mutex);
while (sem->count == 0) {
if (ms == 0) {
rc = false;
} else {
rc = qemu_cond_timedwait_ts(&sem->cond, &sem->mutex, &ts,
__FILE__, __LINE__);
}
if (!rc) { /* timeout */
break;
}
}
if (rc) {
--sem->count;
}
qemu_mutex_unlock(&sem->mutex);
return (rc ? 0 : -1);
}
void qemu_sem_wait(QemuSemaphore *sem)
{
qemu_mutex_lock(&sem->mutex);
while (sem->count == 0) {
qemu_cond_wait(&sem->cond, &sem->mutex);
}
--sem->count;
qemu_mutex_unlock(&sem->mutex);
}
#ifdef __linux__
#include "qemu/futex.h"
#else
static inline void qemu_futex_wake(QemuEvent *ev, int n)
{
assert(ev->initialized);
pthread_mutex_lock(&ev->lock);
if (n == 1) {
pthread_cond_signal(&ev->cond);
} else {
pthread_cond_broadcast(&ev->cond);
}
pthread_mutex_unlock(&ev->lock);
}
static inline void qemu_futex_wait(QemuEvent *ev, unsigned val)
{
assert(ev->initialized);
pthread_mutex_lock(&ev->lock);
if (ev->value == val) {
pthread_cond_wait(&ev->cond, &ev->lock);
}
pthread_mutex_unlock(&ev->lock);
}
#endif
/* Valid transitions:
* - free->set, when setting the event
* - busy->set, when setting the event, followed by qemu_futex_wake
* - set->free, when resetting the event
* - free->busy, when waiting
*
* set->busy does not happen (it can be observed from the outside but
* it really is set->free->busy).
*
* busy->free provably cannot happen; to enforce it, the set->free transition
* is done with an OR, which becomes a no-op if the event has concurrently
* transitioned to free or busy.
*/
#define EV_SET 0
#define EV_FREE 1
#define EV_BUSY -1
void qemu_event_init(QemuEvent *ev, bool init)
{
#ifndef __linux__
pthread_mutex_init(&ev->lock, NULL);
pthread_cond_init(&ev->cond, NULL);
#endif
ev->value = (init ? EV_SET : EV_FREE);
ev->initialized = true;
}
void qemu_event_destroy(QemuEvent *ev)
{
assert(ev->initialized);
ev->initialized = false;
#ifndef __linux__
pthread_mutex_destroy(&ev->lock);
pthread_cond_destroy(&ev->cond);
#endif
}
void qemu_event_set(QemuEvent *ev)
{
assert(ev->initialized);
/*
* Pairs with both qemu_event_reset() and qemu_event_wait().
*
* qemu_event_set has release semantics, but because it *loads*
* ev->value we need a full memory barrier here.
*/
smp_mb();
if (qatomic_read(&ev->value) != EV_SET) {
int old = qatomic_xchg(&ev->value, EV_SET);
/* Pairs with memory barrier in kernel futex_wait system call. */
smp_mb__after_rmw();
if (old == EV_BUSY) {
/* There were waiters, wake them up. */
qemu_futex_wake(ev, INT_MAX);
}
}
}
void qemu_event_reset(QemuEvent *ev)
{
assert(ev->initialized);
/*
* If there was a concurrent reset (or even reset+wait),
* do nothing. Otherwise change EV_SET->EV_FREE.
*/
qatomic_or(&ev->value, EV_FREE);
/*
* Order reset before checking the condition in the caller.
* Pairs with the first memory barrier in qemu_event_set().
*/
smp_mb__after_rmw();
}
void qemu_event_wait(QemuEvent *ev)
{
unsigned value;
assert(ev->initialized);
/*
* qemu_event_wait must synchronize with qemu_event_set even if it does
* not go down the slow path, so this load-acquire is needed that
* synchronizes with the first memory barrier in qemu_event_set().
*
* If we do go down the slow path, there is no requirement at all: we
* might miss a qemu_event_set() here but ultimately the memory barrier in
* qemu_futex_wait() will ensure the check is done correctly.
*/
value = qatomic_load_acquire(&ev->value);
if (value != EV_SET) {
if (value == EV_FREE) {
/*
* Leave the event reset and tell qemu_event_set that there are
* waiters. No need to retry, because there cannot be a concurrent
* busy->free transition. After the CAS, the event will be either
* set or busy.
*
* This cmpxchg doesn't have particular ordering requirements if it
* succeeds (moving the store earlier can only cause qemu_event_set()
* to issue _more_ wakeups), the failing case needs acquire semantics
* like the load above.
*/
if (qatomic_cmpxchg(&ev->value, EV_FREE, EV_BUSY) == EV_SET) {
return;
}
}
/*
* This is the final check for a concurrent set, so it does need
* a smp_mb() pairing with the second barrier of qemu_event_set().
* The barrier is inside the FUTEX_WAIT system call.
*/
qemu_futex_wait(ev, EV_BUSY);
}
}
util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX Our current implementation of qemu_thread_atexit* is broken on OSX. This is because it works by cerating a piece of thread-specific data with pthread_key_create() and using the destructor function for that data to run the notifier function passed to it by the caller of qemu_thread_atexit_add(). The expected use case is that the caller uses a __thread variable as the notifier, and uses the callback to clean up information that it is keeping per-thread in __thread variables. Unfortunately, on OSX this does not work, because on OSX a __thread variable may be destroyed (freed) before the pthread_key_create() destructor runs. (POSIX imposes no ordering constraint here; the OSX implementation happens to implement __thread variables in terms of pthread_key_create((), whereas Linux uses different mechanisms that mean the __thread variables will still be present when the pthread_key_create() destructor is run.) Fix this by switching to a scheme similar to the one qemu-thread-win32 uses for qemu_thread_atexit: keep the thread's notifiers on a __thread variable, and run the notifiers on calls to qemu_thread_exit() and on return from the start routine passed to qemu_thread_start(). We do this with the pthread_cleanup_push() API. We take advantage of the qemu_thread_atexit_add() API permission not to run thread notifiers on process exit to avoid having to special case the main thread. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-11-05 16:55:38 +03:00
static __thread NotifierList thread_exit;
util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX Our current implementation of qemu_thread_atexit* is broken on OSX. This is because it works by cerating a piece of thread-specific data with pthread_key_create() and using the destructor function for that data to run the notifier function passed to it by the caller of qemu_thread_atexit_add(). The expected use case is that the caller uses a __thread variable as the notifier, and uses the callback to clean up information that it is keeping per-thread in __thread variables. Unfortunately, on OSX this does not work, because on OSX a __thread variable may be destroyed (freed) before the pthread_key_create() destructor runs. (POSIX imposes no ordering constraint here; the OSX implementation happens to implement __thread variables in terms of pthread_key_create((), whereas Linux uses different mechanisms that mean the __thread variables will still be present when the pthread_key_create() destructor is run.) Fix this by switching to a scheme similar to the one qemu-thread-win32 uses for qemu_thread_atexit: keep the thread's notifiers on a __thread variable, and run the notifiers on calls to qemu_thread_exit() and on return from the start routine passed to qemu_thread_start(). We do this with the pthread_cleanup_push() API. We take advantage of the qemu_thread_atexit_add() API permission not to run thread notifiers on process exit to avoid having to special case the main thread. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-11-05 16:55:38 +03:00
/*
* Note that in this implementation you can register a thread-exit
* notifier for the main thread, but it will never be called.
* This is OK because main thread exit can only happen when the
* entire process is exiting, and the API allows notifiers to not
* be called on process exit.
*/
void qemu_thread_atexit_add(Notifier *notifier)
{
util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX Our current implementation of qemu_thread_atexit* is broken on OSX. This is because it works by cerating a piece of thread-specific data with pthread_key_create() and using the destructor function for that data to run the notifier function passed to it by the caller of qemu_thread_atexit_add(). The expected use case is that the caller uses a __thread variable as the notifier, and uses the callback to clean up information that it is keeping per-thread in __thread variables. Unfortunately, on OSX this does not work, because on OSX a __thread variable may be destroyed (freed) before the pthread_key_create() destructor runs. (POSIX imposes no ordering constraint here; the OSX implementation happens to implement __thread variables in terms of pthread_key_create((), whereas Linux uses different mechanisms that mean the __thread variables will still be present when the pthread_key_create() destructor is run.) Fix this by switching to a scheme similar to the one qemu-thread-win32 uses for qemu_thread_atexit: keep the thread's notifiers on a __thread variable, and run the notifiers on calls to qemu_thread_exit() and on return from the start routine passed to qemu_thread_start(). We do this with the pthread_cleanup_push() API. We take advantage of the qemu_thread_atexit_add() API permission not to run thread notifiers on process exit to avoid having to special case the main thread. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-11-05 16:55:38 +03:00
notifier_list_add(&thread_exit, notifier);
}
void qemu_thread_atexit_remove(Notifier *notifier)
{
notifier_remove(notifier);
}
util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX Our current implementation of qemu_thread_atexit* is broken on OSX. This is because it works by cerating a piece of thread-specific data with pthread_key_create() and using the destructor function for that data to run the notifier function passed to it by the caller of qemu_thread_atexit_add(). The expected use case is that the caller uses a __thread variable as the notifier, and uses the callback to clean up information that it is keeping per-thread in __thread variables. Unfortunately, on OSX this does not work, because on OSX a __thread variable may be destroyed (freed) before the pthread_key_create() destructor runs. (POSIX imposes no ordering constraint here; the OSX implementation happens to implement __thread variables in terms of pthread_key_create((), whereas Linux uses different mechanisms that mean the __thread variables will still be present when the pthread_key_create() destructor is run.) Fix this by switching to a scheme similar to the one qemu-thread-win32 uses for qemu_thread_atexit: keep the thread's notifiers on a __thread variable, and run the notifiers on calls to qemu_thread_exit() and on return from the start routine passed to qemu_thread_start(). We do this with the pthread_cleanup_push() API. We take advantage of the qemu_thread_atexit_add() API permission not to run thread notifiers on process exit to avoid having to special case the main thread. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-11-05 16:55:38 +03:00
static void qemu_thread_atexit_notify(void *arg)
{
util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX Our current implementation of qemu_thread_atexit* is broken on OSX. This is because it works by cerating a piece of thread-specific data with pthread_key_create() and using the destructor function for that data to run the notifier function passed to it by the caller of qemu_thread_atexit_add(). The expected use case is that the caller uses a __thread variable as the notifier, and uses the callback to clean up information that it is keeping per-thread in __thread variables. Unfortunately, on OSX this does not work, because on OSX a __thread variable may be destroyed (freed) before the pthread_key_create() destructor runs. (POSIX imposes no ordering constraint here; the OSX implementation happens to implement __thread variables in terms of pthread_key_create((), whereas Linux uses different mechanisms that mean the __thread variables will still be present when the pthread_key_create() destructor is run.) Fix this by switching to a scheme similar to the one qemu-thread-win32 uses for qemu_thread_atexit: keep the thread's notifiers on a __thread variable, and run the notifiers on calls to qemu_thread_exit() and on return from the start routine passed to qemu_thread_start(). We do this with the pthread_cleanup_push() API. We take advantage of the qemu_thread_atexit_add() API permission not to run thread notifiers on process exit to avoid having to special case the main thread. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-11-05 16:55:38 +03:00
/*
* Called when non-main thread exits (via qemu_thread_exit()
* or by returning from its start routine.)
*/
notifier_list_notify(&thread_exit, NULL);
}
qemu-thread: fix races on threads that exit very quickly If we create a thread with QEMU_THREAD_DETACHED mode, QEMU may get a segfault with low probability. The backtrace is: #0 0x00007f46c60291d7 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56 #1 0x00007f46c602a8c8 in __GI_abort () at abort.c:90 #2 0x00000000008543c9 in PAT_abort () #3 0x000000000085140d in patchIllInsHandler () #4 <signal handler called> #5 pthread_detach (th=139933037614848) at pthread_detach.c:50 #6 0x0000000000829759 in qemu_thread_create (thread=thread@entry=0x7ffdaa8205e0, name=name@entry=0x94d94a "io-task-worker", start_routine=start_routine@entry=0x7eb9a0 <qio_task_thread_worker>, arg=arg@entry=0x3f5cf70, mode=mode@entry=1) at util/qemu_thread_posix.c:512 #7 0x00000000007ebc96 in qio_task_run_in_thread (task=0x31db2c0, worker=worker@entry=0x7e7e40 <qio_channel_socket_connect_worker>, opaque=0xcd23380, destroy=0x7f1180 <qapi_free_SocketAddress>) at io/task.c:141 #8 0x00000000007e7f33 in qio_channel_socket_connect_async (ioc=ioc@entry=0x626c0b0, addr=<optimized out>, callback=callback@entry=0x55e080 <qemu_chr_socket_connected>, opaque=opaque@entry=0x42862c0, destroy=destroy@entry=0x0) at io/channel_socket.c:194 #9 0x000000000055bdd1 in socket_reconnect_timeout (opaque=0x42862c0) at qemu_char.c:4744 #10 0x00007f46c72483b3 in g_timeout_dispatch () from /usr/lib64/libglib-2.0.so.0 #11 0x00007f46c724799a in g_main_context_dispatch () from /usr/lib64/libglib-2.0.so.0 #12 0x000000000076c646 in glib_pollfds_poll () at main_loop.c:228 #13 0x000000000076c6eb in os_host_main_loop_wait (timeout=348000000) at main_loop.c:273 #14 0x000000000076c815 in main_loop_wait (nonblocking=nonblocking@entry=0) at main_loop.c:521 #15 0x000000000056a511 in main_loop () at vl.c:2076 #16 0x0000000000420705 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4940 The cause of this problem is a glibc bug; for more information, see https://sourceware.org/bugzilla/show_bug.cgi?id=19951. The solution for this bug is to use pthread_attr_setdetachstate. There is a similar issue with pthread_setname_np, which is moved from creating thread to created thread. Signed-off-by: linzhecheng <linzhecheng@huawei.com> Message-Id: <20171128044656.10592-1-linzhecheng@huawei.com> Reviewed-by: Fam Zheng <famz@redhat.com> [Simplify the code by removing qemu_thread_set_name, and free the arguments before invoking the start routine. - Paolo] Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-11-28 07:46:56 +03:00
typedef struct {
void *(*start_routine)(void *);
void *arg;
char *name;
} QemuThreadArgs;
static void *qemu_thread_start(void *args)
{
QemuThreadArgs *qemu_thread_args = args;
void *(*start_routine)(void *) = qemu_thread_args->start_routine;
void *arg = qemu_thread_args->arg;
util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX Our current implementation of qemu_thread_atexit* is broken on OSX. This is because it works by cerating a piece of thread-specific data with pthread_key_create() and using the destructor function for that data to run the notifier function passed to it by the caller of qemu_thread_atexit_add(). The expected use case is that the caller uses a __thread variable as the notifier, and uses the callback to clean up information that it is keeping per-thread in __thread variables. Unfortunately, on OSX this does not work, because on OSX a __thread variable may be destroyed (freed) before the pthread_key_create() destructor runs. (POSIX imposes no ordering constraint here; the OSX implementation happens to implement __thread variables in terms of pthread_key_create((), whereas Linux uses different mechanisms that mean the __thread variables will still be present when the pthread_key_create() destructor is run.) Fix this by switching to a scheme similar to the one qemu-thread-win32 uses for qemu_thread_atexit: keep the thread's notifiers on a __thread variable, and run the notifiers on calls to qemu_thread_exit() and on return from the start routine passed to qemu_thread_start(). We do this with the pthread_cleanup_push() API. We take advantage of the qemu_thread_atexit_add() API permission not to run thread notifiers on process exit to avoid having to special case the main thread. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-11-05 16:55:38 +03:00
void *r;
qemu-thread: fix races on threads that exit very quickly If we create a thread with QEMU_THREAD_DETACHED mode, QEMU may get a segfault with low probability. The backtrace is: #0 0x00007f46c60291d7 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56 #1 0x00007f46c602a8c8 in __GI_abort () at abort.c:90 #2 0x00000000008543c9 in PAT_abort () #3 0x000000000085140d in patchIllInsHandler () #4 <signal handler called> #5 pthread_detach (th=139933037614848) at pthread_detach.c:50 #6 0x0000000000829759 in qemu_thread_create (thread=thread@entry=0x7ffdaa8205e0, name=name@entry=0x94d94a "io-task-worker", start_routine=start_routine@entry=0x7eb9a0 <qio_task_thread_worker>, arg=arg@entry=0x3f5cf70, mode=mode@entry=1) at util/qemu_thread_posix.c:512 #7 0x00000000007ebc96 in qio_task_run_in_thread (task=0x31db2c0, worker=worker@entry=0x7e7e40 <qio_channel_socket_connect_worker>, opaque=0xcd23380, destroy=0x7f1180 <qapi_free_SocketAddress>) at io/task.c:141 #8 0x00000000007e7f33 in qio_channel_socket_connect_async (ioc=ioc@entry=0x626c0b0, addr=<optimized out>, callback=callback@entry=0x55e080 <qemu_chr_socket_connected>, opaque=opaque@entry=0x42862c0, destroy=destroy@entry=0x0) at io/channel_socket.c:194 #9 0x000000000055bdd1 in socket_reconnect_timeout (opaque=0x42862c0) at qemu_char.c:4744 #10 0x00007f46c72483b3 in g_timeout_dispatch () from /usr/lib64/libglib-2.0.so.0 #11 0x00007f46c724799a in g_main_context_dispatch () from /usr/lib64/libglib-2.0.so.0 #12 0x000000000076c646 in glib_pollfds_poll () at main_loop.c:228 #13 0x000000000076c6eb in os_host_main_loop_wait (timeout=348000000) at main_loop.c:273 #14 0x000000000076c815 in main_loop_wait (nonblocking=nonblocking@entry=0) at main_loop.c:521 #15 0x000000000056a511 in main_loop () at vl.c:2076 #16 0x0000000000420705 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4940 The cause of this problem is a glibc bug; for more information, see https://sourceware.org/bugzilla/show_bug.cgi?id=19951. The solution for this bug is to use pthread_attr_setdetachstate. There is a similar issue with pthread_setname_np, which is moved from creating thread to created thread. Signed-off-by: linzhecheng <linzhecheng@huawei.com> Message-Id: <20171128044656.10592-1-linzhecheng@huawei.com> Reviewed-by: Fam Zheng <famz@redhat.com> [Simplify the code by removing qemu_thread_set_name, and free the arguments before invoking the start routine. - Paolo] Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-11-28 07:46:56 +03:00
/* Attempt to set the threads name; note that this is for debug, so
* we're not going to fail if we can't set it.
*/
if (name_threads && qemu_thread_args->name) {
# if defined(CONFIG_PTHREAD_SETNAME_NP_W_TID)
pthread_setname_np(pthread_self(), qemu_thread_args->name);
# elif defined(CONFIG_PTHREAD_SETNAME_NP_WO_TID)
pthread_setname_np(qemu_thread_args->name);
# elif defined(CONFIG_PTHREAD_SET_NAME_NP)
pthread_set_name_np(pthread_self(), qemu_thread_args->name);
# endif
}
QEMU_TSAN_ANNOTATE_THREAD_NAME(qemu_thread_args->name);
qemu-thread: fix races on threads that exit very quickly If we create a thread with QEMU_THREAD_DETACHED mode, QEMU may get a segfault with low probability. The backtrace is: #0 0x00007f46c60291d7 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56 #1 0x00007f46c602a8c8 in __GI_abort () at abort.c:90 #2 0x00000000008543c9 in PAT_abort () #3 0x000000000085140d in patchIllInsHandler () #4 <signal handler called> #5 pthread_detach (th=139933037614848) at pthread_detach.c:50 #6 0x0000000000829759 in qemu_thread_create (thread=thread@entry=0x7ffdaa8205e0, name=name@entry=0x94d94a "io-task-worker", start_routine=start_routine@entry=0x7eb9a0 <qio_task_thread_worker>, arg=arg@entry=0x3f5cf70, mode=mode@entry=1) at util/qemu_thread_posix.c:512 #7 0x00000000007ebc96 in qio_task_run_in_thread (task=0x31db2c0, worker=worker@entry=0x7e7e40 <qio_channel_socket_connect_worker>, opaque=0xcd23380, destroy=0x7f1180 <qapi_free_SocketAddress>) at io/task.c:141 #8 0x00000000007e7f33 in qio_channel_socket_connect_async (ioc=ioc@entry=0x626c0b0, addr=<optimized out>, callback=callback@entry=0x55e080 <qemu_chr_socket_connected>, opaque=opaque@entry=0x42862c0, destroy=destroy@entry=0x0) at io/channel_socket.c:194 #9 0x000000000055bdd1 in socket_reconnect_timeout (opaque=0x42862c0) at qemu_char.c:4744 #10 0x00007f46c72483b3 in g_timeout_dispatch () from /usr/lib64/libglib-2.0.so.0 #11 0x00007f46c724799a in g_main_context_dispatch () from /usr/lib64/libglib-2.0.so.0 #12 0x000000000076c646 in glib_pollfds_poll () at main_loop.c:228 #13 0x000000000076c6eb in os_host_main_loop_wait (timeout=348000000) at main_loop.c:273 #14 0x000000000076c815 in main_loop_wait (nonblocking=nonblocking@entry=0) at main_loop.c:521 #15 0x000000000056a511 in main_loop () at vl.c:2076 #16 0x0000000000420705 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4940 The cause of this problem is a glibc bug; for more information, see https://sourceware.org/bugzilla/show_bug.cgi?id=19951. The solution for this bug is to use pthread_attr_setdetachstate. There is a similar issue with pthread_setname_np, which is moved from creating thread to created thread. Signed-off-by: linzhecheng <linzhecheng@huawei.com> Message-Id: <20171128044656.10592-1-linzhecheng@huawei.com> Reviewed-by: Fam Zheng <famz@redhat.com> [Simplify the code by removing qemu_thread_set_name, and free the arguments before invoking the start routine. - Paolo] Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-11-28 07:46:56 +03:00
g_free(qemu_thread_args->name);
g_free(qemu_thread_args);
/*
* GCC 11 with glibc 2.17 on PowerPC reports
*
* qemu-thread-posix.c:540:5: error: __sigsetjmp accessing 656 bytes
* in a region of size 528 [-Werror=stringop-overflow=]
* 540 | pthread_cleanup_push(qemu_thread_atexit_notify, NULL);
* | ^~~~~~~~~~~~~~~~~~~~
*
* which is clearly nonsense.
*/
#pragma GCC diagnostic push
#ifndef __clang__
#pragma GCC diagnostic ignored "-Wstringop-overflow"
#endif
util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX Our current implementation of qemu_thread_atexit* is broken on OSX. This is because it works by cerating a piece of thread-specific data with pthread_key_create() and using the destructor function for that data to run the notifier function passed to it by the caller of qemu_thread_atexit_add(). The expected use case is that the caller uses a __thread variable as the notifier, and uses the callback to clean up information that it is keeping per-thread in __thread variables. Unfortunately, on OSX this does not work, because on OSX a __thread variable may be destroyed (freed) before the pthread_key_create() destructor runs. (POSIX imposes no ordering constraint here; the OSX implementation happens to implement __thread variables in terms of pthread_key_create((), whereas Linux uses different mechanisms that mean the __thread variables will still be present when the pthread_key_create() destructor is run.) Fix this by switching to a scheme similar to the one qemu-thread-win32 uses for qemu_thread_atexit: keep the thread's notifiers on a __thread variable, and run the notifiers on calls to qemu_thread_exit() and on return from the start routine passed to qemu_thread_start(). We do this with the pthread_cleanup_push() API. We take advantage of the qemu_thread_atexit_add() API permission not to run thread notifiers on process exit to avoid having to special case the main thread. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-11-05 16:55:38 +03:00
pthread_cleanup_push(qemu_thread_atexit_notify, NULL);
r = start_routine(arg);
pthread_cleanup_pop(1);
#pragma GCC diagnostic pop
util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX Our current implementation of qemu_thread_atexit* is broken on OSX. This is because it works by cerating a piece of thread-specific data with pthread_key_create() and using the destructor function for that data to run the notifier function passed to it by the caller of qemu_thread_atexit_add(). The expected use case is that the caller uses a __thread variable as the notifier, and uses the callback to clean up information that it is keeping per-thread in __thread variables. Unfortunately, on OSX this does not work, because on OSX a __thread variable may be destroyed (freed) before the pthread_key_create() destructor runs. (POSIX imposes no ordering constraint here; the OSX implementation happens to implement __thread variables in terms of pthread_key_create((), whereas Linux uses different mechanisms that mean the __thread variables will still be present when the pthread_key_create() destructor is run.) Fix this by switching to a scheme similar to the one qemu-thread-win32 uses for qemu_thread_atexit: keep the thread's notifiers on a __thread variable, and run the notifiers on calls to qemu_thread_exit() and on return from the start routine passed to qemu_thread_start(). We do this with the pthread_cleanup_push() API. We take advantage of the qemu_thread_atexit_add() API permission not to run thread notifiers on process exit to avoid having to special case the main thread. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-11-05 16:55:38 +03:00
return r;
}
void qemu_thread_create(QemuThread *thread, const char *name,
void *(*start_routine)(void*),
void *arg, int mode)
{
sigset_t set, oldset;
int err;
pthread_attr_t attr;
QemuThreadArgs *qemu_thread_args;
err = pthread_attr_init(&attr);
if (err) {
error_exit(err, __func__);
}
qemu-thread: fix races on threads that exit very quickly If we create a thread with QEMU_THREAD_DETACHED mode, QEMU may get a segfault with low probability. The backtrace is: #0 0x00007f46c60291d7 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56 #1 0x00007f46c602a8c8 in __GI_abort () at abort.c:90 #2 0x00000000008543c9 in PAT_abort () #3 0x000000000085140d in patchIllInsHandler () #4 <signal handler called> #5 pthread_detach (th=139933037614848) at pthread_detach.c:50 #6 0x0000000000829759 in qemu_thread_create (thread=thread@entry=0x7ffdaa8205e0, name=name@entry=0x94d94a "io-task-worker", start_routine=start_routine@entry=0x7eb9a0 <qio_task_thread_worker>, arg=arg@entry=0x3f5cf70, mode=mode@entry=1) at util/qemu_thread_posix.c:512 #7 0x00000000007ebc96 in qio_task_run_in_thread (task=0x31db2c0, worker=worker@entry=0x7e7e40 <qio_channel_socket_connect_worker>, opaque=0xcd23380, destroy=0x7f1180 <qapi_free_SocketAddress>) at io/task.c:141 #8 0x00000000007e7f33 in qio_channel_socket_connect_async (ioc=ioc@entry=0x626c0b0, addr=<optimized out>, callback=callback@entry=0x55e080 <qemu_chr_socket_connected>, opaque=opaque@entry=0x42862c0, destroy=destroy@entry=0x0) at io/channel_socket.c:194 #9 0x000000000055bdd1 in socket_reconnect_timeout (opaque=0x42862c0) at qemu_char.c:4744 #10 0x00007f46c72483b3 in g_timeout_dispatch () from /usr/lib64/libglib-2.0.so.0 #11 0x00007f46c724799a in g_main_context_dispatch () from /usr/lib64/libglib-2.0.so.0 #12 0x000000000076c646 in glib_pollfds_poll () at main_loop.c:228 #13 0x000000000076c6eb in os_host_main_loop_wait (timeout=348000000) at main_loop.c:273 #14 0x000000000076c815 in main_loop_wait (nonblocking=nonblocking@entry=0) at main_loop.c:521 #15 0x000000000056a511 in main_loop () at vl.c:2076 #16 0x0000000000420705 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4940 The cause of this problem is a glibc bug; for more information, see https://sourceware.org/bugzilla/show_bug.cgi?id=19951. The solution for this bug is to use pthread_attr_setdetachstate. There is a similar issue with pthread_setname_np, which is moved from creating thread to created thread. Signed-off-by: linzhecheng <linzhecheng@huawei.com> Message-Id: <20171128044656.10592-1-linzhecheng@huawei.com> Reviewed-by: Fam Zheng <famz@redhat.com> [Simplify the code by removing qemu_thread_set_name, and free the arguments before invoking the start routine. - Paolo] Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-11-28 07:46:56 +03:00
if (mode == QEMU_THREAD_DETACHED) {
pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
}
/* Leave signal handling to the iothread. */
sigfillset(&set);
/* Blocking the signals can result in undefined behaviour. */
sigdelset(&set, SIGSEGV);
sigdelset(&set, SIGFPE);
sigdelset(&set, SIGILL);
/* TODO avoid SIGBUS loss on macOS */
pthread_sigmask(SIG_SETMASK, &set, &oldset);
qemu_thread_args = g_new0(QemuThreadArgs, 1);
qemu_thread_args->name = g_strdup(name);
qemu_thread_args->start_routine = start_routine;
qemu_thread_args->arg = arg;
err = pthread_create(&thread->thread, &attr,
qemu_thread_start, qemu_thread_args);
qemu-thread: fix races on threads that exit very quickly If we create a thread with QEMU_THREAD_DETACHED mode, QEMU may get a segfault with low probability. The backtrace is: #0 0x00007f46c60291d7 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56 #1 0x00007f46c602a8c8 in __GI_abort () at abort.c:90 #2 0x00000000008543c9 in PAT_abort () #3 0x000000000085140d in patchIllInsHandler () #4 <signal handler called> #5 pthread_detach (th=139933037614848) at pthread_detach.c:50 #6 0x0000000000829759 in qemu_thread_create (thread=thread@entry=0x7ffdaa8205e0, name=name@entry=0x94d94a "io-task-worker", start_routine=start_routine@entry=0x7eb9a0 <qio_task_thread_worker>, arg=arg@entry=0x3f5cf70, mode=mode@entry=1) at util/qemu_thread_posix.c:512 #7 0x00000000007ebc96 in qio_task_run_in_thread (task=0x31db2c0, worker=worker@entry=0x7e7e40 <qio_channel_socket_connect_worker>, opaque=0xcd23380, destroy=0x7f1180 <qapi_free_SocketAddress>) at io/task.c:141 #8 0x00000000007e7f33 in qio_channel_socket_connect_async (ioc=ioc@entry=0x626c0b0, addr=<optimized out>, callback=callback@entry=0x55e080 <qemu_chr_socket_connected>, opaque=opaque@entry=0x42862c0, destroy=destroy@entry=0x0) at io/channel_socket.c:194 #9 0x000000000055bdd1 in socket_reconnect_timeout (opaque=0x42862c0) at qemu_char.c:4744 #10 0x00007f46c72483b3 in g_timeout_dispatch () from /usr/lib64/libglib-2.0.so.0 #11 0x00007f46c724799a in g_main_context_dispatch () from /usr/lib64/libglib-2.0.so.0 #12 0x000000000076c646 in glib_pollfds_poll () at main_loop.c:228 #13 0x000000000076c6eb in os_host_main_loop_wait (timeout=348000000) at main_loop.c:273 #14 0x000000000076c815 in main_loop_wait (nonblocking=nonblocking@entry=0) at main_loop.c:521 #15 0x000000000056a511 in main_loop () at vl.c:2076 #16 0x0000000000420705 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4940 The cause of this problem is a glibc bug; for more information, see https://sourceware.org/bugzilla/show_bug.cgi?id=19951. The solution for this bug is to use pthread_attr_setdetachstate. There is a similar issue with pthread_setname_np, which is moved from creating thread to created thread. Signed-off-by: linzhecheng <linzhecheng@huawei.com> Message-Id: <20171128044656.10592-1-linzhecheng@huawei.com> Reviewed-by: Fam Zheng <famz@redhat.com> [Simplify the code by removing qemu_thread_set_name, and free the arguments before invoking the start routine. - Paolo] Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-11-28 07:46:56 +03:00
if (err)
error_exit(err, __func__);
pthread_sigmask(SIG_SETMASK, &oldset, NULL);
pthread_attr_destroy(&attr);
}
int qemu_thread_set_affinity(QemuThread *thread, unsigned long *host_cpus,
unsigned long nbits)
{
#if defined(CONFIG_PTHREAD_AFFINITY_NP)
const size_t setsize = CPU_ALLOC_SIZE(nbits);
unsigned long value;
cpu_set_t *cpuset;
int err;
cpuset = CPU_ALLOC(nbits);
g_assert(cpuset);
CPU_ZERO_S(setsize, cpuset);
value = find_first_bit(host_cpus, nbits);
while (value < nbits) {
CPU_SET_S(value, setsize, cpuset);
value = find_next_bit(host_cpus, nbits, value + 1);
}
err = pthread_setaffinity_np(thread->thread, setsize, cpuset);
CPU_FREE(cpuset);
return err;
#else
return -ENOSYS;
#endif
}
int qemu_thread_get_affinity(QemuThread *thread, unsigned long **host_cpus,
unsigned long *nbits)
{
#if defined(CONFIG_PTHREAD_AFFINITY_NP)
unsigned long tmpbits;
cpu_set_t *cpuset;
size_t setsize;
int i, err;
tmpbits = CPU_SETSIZE;
while (true) {
setsize = CPU_ALLOC_SIZE(tmpbits);
cpuset = CPU_ALLOC(tmpbits);
g_assert(cpuset);
err = pthread_getaffinity_np(thread->thread, setsize, cpuset);
if (err) {
CPU_FREE(cpuset);
if (err != -EINVAL) {
return err;
}
tmpbits *= 2;
} else {
break;
}
}
/* Convert the result into a proper bitmap. */
*nbits = tmpbits;
*host_cpus = bitmap_new(tmpbits);
for (i = 0; i < tmpbits; i++) {
if (CPU_ISSET(i, cpuset)) {
set_bit(i, *host_cpus);
}
}
CPU_FREE(cpuset);
return 0;
#else
return -ENOSYS;
#endif
}
void qemu_thread_get_self(QemuThread *thread)
{
thread->thread = pthread_self();
}
bool qemu_thread_is_self(QemuThread *thread)
{
return pthread_equal(pthread_self(), thread->thread);
}
void qemu_thread_exit(void *retval)
{
pthread_exit(retval);
}
void *qemu_thread_join(QemuThread *thread)
{
int err;
void *ret;
err = pthread_join(thread->thread, &ret);
if (err) {
error_exit(err, __func__);
}
return ret;
}