优先级反转那些事儿

开发
通过观察发现,出问题的线程都有QOS:BACKGROUND标记。整体看起来持有锁的子线程仍然在执行,只是留给主线程的时间不够了。为什么这些子线程在持有锁的情况下,需要执行这么久,直到主线程的 8s 卡死?一种情况就是真的如此耗时,另一种则是出现了优先级反转。

1. 从一个线上问题说起

最近在线上遇到了一些​​[HMDConfigManager remoteConfigWithAppID:]​​卡死

1.1 初步分析

观察了下主线程堆栈,用到的锁是读写锁:

图片

随后又去翻了下持有着锁的子线程,有各种各样的情况,且基本都处于正常的执行状态,例如有的处于打开文件状态,有的处于​​read​​状态,有的正在执行​​NSUserDefaults​​的方法···

图片

图片

通过观察发现,出问题的线程都有​​QOS:BACKGROUND​​标记。整体看起来持有锁的子线程仍然在执行,只是留给主线程的时间不够了。为什么这些子线程在持有锁的情况下,需要执行这么久,直到主线程的 8s 卡死?一种情况就是真的如此耗时,另一种则是出现了优先级反转。

1.2 解决办法

这个案例里,持有读写锁且优先级低的线程迟迟得不到调度(又或者得到调度的时候又被抢占了,或者得到调度的时候时间已然不够了)而具有高优先级的线程由于拿不到读写锁,一直被阻塞,所以互相死锁。​​iOS8​​之后引入了​​QualityOfService​​的概念,类似于线程的优先级,设置不同的​​QualityOfService​​的值后系统会分配不同的​​CPU​​时间、网络资源和硬盘资源等,因此我们可以通过这个设置队列的优先级。

1.2.1方案一:去除对 NSOperationQueue 的优先级设置

在 Threading Programming Guide 文档中,苹果给出了提示:

Important: It is generally a good idea to leave the priorities of your threads at their default values. Increasing the priorities of some threads also increases the likelihood of starvation among lower-priority threads. If your application contains high-priority and low-priority threads that must interact with each other, the starvation of lower-priority threads may block other threads and create performance bottlenecks.

苹果的建议是不要随意修改线程的优先级,尤其是这些高低优先级线程之间存在临界资源竞争的情况。所以删除相关优先级设置代码即可解决问题。

1.2.2 方案二:临时修改线程优先级

在 pthread_rwlock_rdlock(3pthread) 发现了如下提示:

Realtime applications may encounter priority inversion when using read-write locks. The problem occurs when a high priority thread "locks" a read-write lock that is about to be "unlocked" by a low priority thread, but the low priority thread is preempted by a medium priority thread. This scenario leads to priority inversion; a high priority thread is blocked by lower priority threads for an unlimited period of time. During system design, realtime programmers must take into account the possibility of this kind of priority inversion. They can deal with it in a number of ways, such as by having critical sections that are guarded by read-write locks execute at a high priority, so that a thread cannot be preempted while executing in its critical section.

尽管针对的是实时系统,但是还是有一些启示和帮助。按照提示,对有问题的代码进行了修改:在线程通过 ​​pthread_rwlock_wrlock​​ 拿到 ​​_rwlock​​ 的时候,临时提升其优先级,在释放 ​​_rwlock​​ 之后,恢复其原先的优先级。

- (id)remoteConfigWithAppID:(NSString *)appID
{
.......
pthread_rwlock_rdlock(&_rwlock);
HMDHeimdallrConfig *result = ....... // get existing config
pthread_rwlock_unlock(&_rwlock);

if(result == nil) {
result = [[HMDHeimdallrConfig alloc] init]; // make a new config
pthread_rwlock_wrlock(&_rwlock);

qos_class_t oldQos = qos_class_self();
BOOL needRecover = NO;

// 临时提升线程优先级
if (_enablePriorityInversionProtection && oldQos < QOS_CLASS_USER_INTERACTIVE) {
int ret = pthread_set_qos_class_self_np(QOS_CLASS_USER_INTERACTIVE, 0);
needRecover = (ret == 0);
}

......

pthread_rwlock_unlock(&_rwlock);

// 恢复线程优先级
if (_enablePriorityInversionProtection && needRecover) {
pthread_set_qos_class_self_np(oldQos, 0);
}
}

return result;
}

值得注意的是,这里只能使用​​pthread​​的​​api​​,​​NSThread​​提供的​​API​​是不可行的

1.3  Demo 验证

为了验证上述的手动调整线程优先级是否有一定的效果,这里通过​​demo​​进行本地实验:定义了​​2000​​个​​operation​​(目的是为了​​CPU​​繁忙),优先级设置​​NSQualityOfServiceUserInitiated​​,且对其中可以被​​100​​整除的​​operation​​的优先级调整为​​NSQualityOfServiceBackground​​,在每个​​operation​​执行相同的耗时任务,然后对这被选中的​​10​​个​​operation​​进行耗时统计。

for (int j = 0; j < 2000; ++j) {
NSOperationQueue *operation = [[NSOperationQueue alloc] init];
operation.maxConcurrentOperationCount = 1;
operation.qualityOfService = NSQualityOfServiceUserInitiated;

// 模块1
// if (j % 100 == 0) {
// operation.qualityOfService = NSQualityOfServiceBackground;
// }
// 模块1

[operation addOperationWithBlock:^{
// 模块2
// qos_class_t oldQos = qos_class_self();
// pthread_set_qos_class_self_np(QOS_CLASS_USER_INITIATED, 0);
// 模块2

NSTimeInterval start = CFAbsoluteTimeGetCurrent();
double sum = 0;
for (int i = 0; i < 100000; ++i) {
sum += sin(i) + cos(i) + sin(i*2) + cos(i*2);
}
start = CFAbsoluteTimeGetCurrent() - start;
if (j % 100 == 0) {
printf("%.8f\n", start * 1000);
}

// 模块2
// pthread_set_qos_class_self_np(oldQos, 0);
// 模块2
}];
}

统计信息如下表所示:

图片

可以看到:

  1. 正常情况下,每个任务的平均耗时为:11.8190561;
  2. operation被设置为低优先级时,其耗时大幅度提升为:94.70210189;
  3. operation被设置为低优先级时,又在Block中手动恢复其原有的优先级,其耗时已经大幅度降低:15.04005137(耗时比正常情况高,大家可以思考下为什么)

通过​​Demo​​可以发现,通过手动调整其优先级,低优先级任务的整体耗时得到大幅度的降低,这样在持有锁的情况下,可以减少对主线程的阻塞时间。

1.4 上线效果

图片

该问题的验证过程分为​​2​​个阶段:

  1. 第一个阶段如第1个红框所示,从36号开始在版本19.7上有较大幅度的下降,主要原因:堆栈中被等待的队列信息由QOS:BACKGROUND变为了com.apple.root.default-qos,队列的优先级从QOS_CLASS_BACKGROUND提升为QOS_CLASS_DEFAULT,相当于实施了方案一,使用了默认优先级。
  2. 第二个阶段如第2个红框所示,从424号在版本20.3上开始验证。目前看起来效果暂时不明显,推测一个主要原因是:demo中是把优先级从QOS_CLASS_BACKGROUND提升为QOS_CLASS_USER_INITIATED,而线上相当于把队列的优先级从默认的优先级QOS_CLASS_DEFAULT提升为QOS_CLASS_USER_INITIATED
  1. ​QOS_CLASS_BACKGROUND​​的Mach层级优先级数是4;
  2. ​QOS_CLASS_DEFAULT​​的Mach层级优先级数是31;
  3. ​QOS_CLASS_USER_INITIATED​​的Mach层级优先级数是37。

所以相对来说,线上的提升相对有限。

2.  深刻理解优先级反转

那么是否所有锁都需要像上文一样,手动提升持有锁的线程优先级?系统是否会自动调整线程的优先级?如果有这样的机制,是否可以覆盖所有的锁?要理解这些问题,需要深刻认识优先级反转。

2.1 什么是优先级反转?

优先级反转,是指某同步资源被较低优先级的进程/线程所拥有,较高优先级的进程/线程竞争该同步资源未获得该资源,而使得较高优先级进程/线程反而推迟被调度执行的现象。根据阻塞类型的不同,优先级反转又被分为​​Bounded priority inversion​​和​​Unbounded priority inversion​​。

这里借助 Introduction to RTOS - Solution to Part 11 的图进行示意。

2.1.1 Bounded priority inversion

如图所示,高优先级任务(​​Task H​​)被持有锁的低优先级任务(​​Task L​​)阻塞,由于阻塞的时间取决于低优先级任务在临界区的时间(持有锁的时间),所以被称为​​bounded priority inversion​​。只要​​Task L​​一直持有锁,​​Task H​​就会一直被阻塞,低优先级的任务运行在高优先级任务的前面,优先级被反转。

这里的任务也可以理解为线程

图片

2.1.2 Unbounded priority inversion

在​​Task L​​持有锁的情况下,如果有一个中间优先级的任务(​​Task M​​)打断了​​Task L​​,前面的​​bounded​​就会变为​​unbounded​​,因为​​Task M​​只要抢占了​​Task L​​的​​CPU​​,就可能会阻塞​​Task H​​任意多的时间(​​Task M​​可能不止​​1​​个)。

图片

2.2  优先级反转常规解决思路

目前解决​​Unbounded priority inversion​​有​​2​​种方法:一种被称作优先权极限(​​priority ceiling protocol​​),另一种被称作优先级继承(​​priority inheritance​​)。

2.2.1 Priority ceiling protocol

在优先权极限方案中,系统把每一个临界资源与 1 个极限优先权相关联。当1个任务进入临界区时,系统便把这个极限优先权传递给这个任务,使得这个任务的优先权最高;当这个任务退出临界区后,系统立即把它的优先权恢复正常,从而保证系统不会出现优先权反转的情况。该极限优先权的值是由所有需要该临界资源的任务的最大优先级来决定的。

如图所示,锁的极限优先权是 3。当​​Task L​​持有锁的时候,它的优先级将会被提升到3,和​​Task H​​一样的优先级。这样就可以阻止​​Task M​​(优先级是2)的运行,直到​​Task L​​和​​Task H​​不再需要该锁。

图片

2.2.2 Priority inheritance

在优先级继承方案中,大致原理是:高优先级任务在尝试获取锁的时候,如果该锁正好被低优先级任务持有,此时会临时把高优先级线程的优先级转移给拥有锁的低优先级线程,使低优先级线程能更快的执行并释放同步资源,释放同步资源后再恢复其原来的优先级。

图片

​priority ceiling protocol​​和​​priority inheritance​​都会在释放锁的时候,恢复低优先级任务的优先级。同时要注意,以上​​2​​种方法只能阻止​​Unbounded priority inversion​​,而无法阻止​​Bounded priority inversion​​(​​Task H​​必须等待​​Task L​​执行完毕才能执行,这个反转是无法避免的)。

可以通过以下几种发生来避免或者转移​​Bounded priority inversion​​:

  1. 减少临界区的执行时间,减少Bounded priority inversion的反转耗时;
  2. 避免使用会阻塞高优先级任务的临界区资源;
  3. 专门使用一个队列来管理资源,避免使用锁。

优先级继承必须是可传递的。举个栗子:当​​T1​​阻塞在被​​T2​​持有的资源上,而​​T2​​又阻塞在​​T3​​持有的一个资源上。如果​​T1​​的优先级高于​​T2​​和​​T3​​的优先级,​​T3​​必须通过​​T2​​继承​​T1​​的优先级。否则,如果另外一个优先级高于​​T2​​和​​T3​​,小于​​T1​​的线程​​T4​​,将抢占​​T3​​,引发相对于​​T1​​的优先级反转。因此,线程所继承的优先级必须是直接或者间接阻塞的线程的最高优先级。

3. 如何避免优先级反转?

3.1 QoS 传递

iOS 系统主要使用以下两种机制来在不同线程(或 ​​queue​​)间传递 ​​QoS​​:

  • 机制 1:dispatch_async
  • ​dispatch_async()​​ automatically propagates the QoS from the calling thread, though it will translate User Interactive to User Initiated to avoid assigning that priority to non-main threads.
  • Capturedattimeofblocksubmission,translateuserinteractivetouserinitiated.UsedifdestinationqueuedoesnothaveaQoSanddoesnotlowertheQoS(exdispatch_asyncbacktothemainthread).
  • 机制 2:基于 XPC 的进程间通信(IPC

系统的 QoS 传递规则比较复杂,主要参考以下信息:

  • 当前线程的QoS
  • 如果是使用dispatch_block_create() 方法生成的 dispatch_block,则考虑生成 block时所调用的参数
  • ​dispatch_async​​ 或 IPC 的目标 queue 或线程的 QoS

调度程序会根据这些信息决定 ​​block​​ 以什么优先级运行。

  1. 如果没有其他线程同步地等待此block,则 block就按上面所说的优先级来运行。
  2. ​如果出现了线程间同步等待的情况,则调度程序会根据情况调整线程的运行优先级。​

3.2 如何触发优先级反转避免机制?

如果当前线程因等待某线程(线程 1)上正在进行的操作(如 ​​block1​​)而受阻,而系统知道 ​​block1​​ 所在的目标线程(​​owner​​),系统会通过提高相关线程的优先级来解决优先级反转的问题。反之如果系统不知道 ​​block1​​ 所在目标线程,则无法知道应该提高谁的优先级,也就无法解决反转问题;

记录了持有者信息(​​owner​​)的系统 API 如下:

  1. ​pthread mutex​​、os_unfair_lock、以及基于这二者实现的上层API
    a.dispatch_once 的实现是基于 os_unfair_lock
    b.NSLockNSRecursiveLock@synchronized 等的实现是基于 pthreadmutex
  2. ​dispatch_sync​​、dispatch_wait
  3. ​xpc_connection_send_with_message_sync​

使用以上这些 ​​API​​ 能够在发生优先级反转时使系统启用优先级反转避免机制。

3.3  基础 API 验证

接下来对前文提到的各种「基础系统​​API​​」进行验证

测试验证环境:模拟器 iOS15.2

3.3.1  pthread mutex

​pthread mutex​​的数据结构​​pthread_mutex_s​​其中有一个​​m_tid​​字段,专门来记录持有该锁的线程​​Id​​。

// types_internal.h
struct pthread_mutex_s {
long sig;
_pthread_lock lock;
union {
uint32_t value;
struct pthread_mutex_options_s options;
} mtxopts;
int16_t prioceiling;
int16_t priority;
#if defined(__LP64__)
uint32_t _pad;
#endif
union {
struct {
uint32_t m_tid[2]; // thread id of thread that has mutex locked
uint32_t m_seq[2]; // mutex sequence id
uint32_t m_mis[2]; // for misaligned locks m_tid/m_seq will span into here
} psynch;
struct _pthread_mutex_ulock_s ulock;
};
#if defined(__LP64__)
uint32_t _reserved[4];
#else
uint32_t _reserved[1];
#endif
};

代码来验证一下:线程优先级是否会被提升?

// printThreadPriority用来打印线程的优先级信息
void printThreadPriority() {
thread_t cur_thread = mach_thread_self();
mach_port_deallocate(mach_task_self(), cur_thread);
mach_msg_type_number_t thread_info_count = THREAD_INFO_MAX;
thread_info_data_t thinfo;
kern_return_t kr = thread_info(cur_thread, THREAD_EXTENDED_INFO, (thread_info_t)thinfo, &thread_info_count);
if (kr != KERN_SUCCESS) {
return;
}
thread_extended_info_t extend_info = (thread_extended_info_t)thinfo;
printf("pth_priority: %d, pth_curpri: %d, pth_maxpriority: %d\n", extend_info->pth_priority, extend_info->pth_curpri, extend_info->pth_maxpriority);
}

先在子线程上锁并休眠,然后主线程请求该锁。

dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_BACKGROUND, 0), ^{
printf("begin : \n");
printThreadPriority();
printf("queue before lock \n");
pthread_mutex_lock(&_lock); //确保 backgroundQueue 先得到锁
printf("queue lock \n");
printThreadPriority();
dispatch_async(dispatch_get_main_queue(), ^{
printf("before main lock\n");
pthread_mutex_lock(&_lock);
printf("in main lock\n");
pthread_mutex_unlock(&_lock);
printf("after main unlock\n");
});
sleep(10);
printThreadPriority();
printf("queue unlock\n");
pthread_mutex_unlock(&_lock);
printf("queue after unlock\n");
});
begin : 
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63
queue before lock
queue lock
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63
before main lock
pth_priority: 47, pth_curpri: 47, pth_maxpriority: 63
queue unlock
in main lock
after main unlock
queue after unlock

可以看到,低优先级子线程先持有锁,当时的优先级为​​4​​,而该锁被主线程请求的时候,子线程的优先级被提升为​​47​

3.3.2  os_unfair_lock

​os_unfair_lock​​用来替换​​OSSpinLock​​,解决优先级反转问题。等待​​os_unfair_lock​​锁的线程会处于休眠状态,从用户态切换到内核态,而并非忙等。​​os_unfair_lock​​将线程​​ID​​保存到了锁的内部,锁的等待者会把自己的优先级让出来,从而避免优先级反转。验证一下:

  dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_BACKGROUND, 0), ^{
printf("begin : \n");
printThreadPriority();
printf("queue before lock \n");
os_unfair_lock_lock(&_unfair_lock); //确保 backgroundQueue 先得到锁
printf("queue lock \n");
printThreadPriority();
dispatch_async(dispatch_get_main_queue(), ^{
printf("before main lock\n");
os_unfair_lock_lock(&_unfair_lock);
printf("in main lock\n");
os_unfair_lock_unlock(&_unfair_lock);
printf("after main unlock\n");
});
sleep(10);
printThreadPriority();
printf("queue unlock\n");
os_unfair_lock_unlock(&_unfair_lock);
printf("queue after unlock\n");
});
begin : 
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63
queue before lock
queue lock
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63
before main lock
pth_priority: 47, pth_curpri: 47, pth_maxpriority: 63
queue unlock
in main lock
after main unlock
queue after unlock

结果和​​pthread mutex​​一致。

3.3.3  pthread_rwlock_t

在 pthread_rwlock_init 有如下提示:

Caveats: Beware of priority inversion when using read-write locks. A high-priority thread may be blocked waiting on a read-write lock locked by a low-priority thread. The microkernel has no knowledge of read-write locks, and therefore can't boost the low-priority thread to prevent the priority inversion.

大意是内核不感知读写锁,无法提升低优先级线程的优先级,从而无法避免优先级反转。通过查询定义发现:​​pthread_rwlock_s​​包含了字段​​rw_tid​​,专门来记录持有写锁的线程,这不由令人好奇:为什么​​pthread_rwlock_s​​有​​owner​​信息却仍然无法避免优先级反转?

struct pthread_rwlock_s {
long sig;
_pthread_lock lock;
uint32_t
unused:29,
misalign:1,
pshared:2;
uint32_t rw_flags;
#if defined(__LP64__)
uint32_t _pad;
#endif
uint32_t rw_tid[2]; // thread id of thread that has exclusive (write) lock
uint32_t rw_seq[4]; // rw sequence id (at 128-bit aligned boundary)
uint32_t rw_mis[4]; // for misaligned locks rw_seq will span into here
#if defined(__LP64__)
uint32_t _reserved[34];
#else
uint32_t _reserved[18];
#endif
};

https://news.ycombinator.com/item?id=21751269 链接中提到:

xnu supports priority inheritance through "turnstiles", a kernel-internal mechanism which is used by default by a number of locking primitives (list at [1]), including normal pthread mutexes (though not read-write locks [2]), as well as the os_unfair_lock API (via the ulock syscalls). With pthread mutexes, you can actually explicitly request priority inheritance by calling pthread_mutexattr_setprotocol [3] with PTHREAD_PRIO_INHERIT; the Apple implementation supports it, but currently ignores the protocol setting and just gives all mutexes priority inheritance.

大意是:​​XNU​使用 ​​​turnstiles​​​ 内核机制进行优先级继承,这种机制被应用在 ​​pthread mutex​​  ​​os_unfair_lock​​ 上。

顺藤摸瓜,在​​ksyn_wait​​方法中找到了​​_kwq_use_turnstile​​的调用,其中的注释对读写锁解释的比较委婉,添加了 ​​at least sometimes​

pthread mutexes and rwlocks both (at least sometimes) know their owner and can use turnstiles. Otherwise, we pass NULL as the tstore to the shims so they wait on the global waitq.

// libpthread/kern/kern_synch.c
int
ksyn_wait(ksyn_wait_queue_t kwq, kwq_queue_type_t kqi, uint32_t lockseq,
int fit, uint64_t abstime, uint16_t kwe_flags,
thread_continue_t continuation, block_hint_t block_hint)
{
thread_t th = current_thread();
uthread_t uth = pthread_kern->get_bsdthread_info(th);
struct turnstile **tstore = NULL;
int res;

assert(continuation != THREAD_CONTINUE_NULL);

ksyn_waitq_element_t kwe = pthread_kern->uthread_get_uukwe(uth);
bzero(kwe, sizeof(*kwe));
kwe->kwe_count = 1;
kwe->kwe_lockseq = lockseq & PTHRW_COUNT_MASK;
kwe->kwe_state = KWE_THREAD_INWAIT;
kwe->kwe_uth = uth;
kwe->kwe_thread = th;
kwe->kwe_flags = kwe_flags;

res = ksyn_queue_insert(kwq, kqi, kwe, lockseq, fit);
if (res != 0) {
//panic("psynch_rw_wrlock: failed to enqueue\n"); // XXX ksyn_wqunlock(kwq);
return res;
}

PTHREAD_TRACE(psynch_mutex_kwqwait, kwq->kw_addr, kwq->kw_inqueue,
kwq->kw_prepost.count, kwq->kw_intr.count);

if (_kwq_use_turnstile(kwq)) {
// pthread mutexes and rwlocks both (at least sometimes) know their
// owner and can use turnstiles. Otherwise, we pass NULL as the
// tstore to the shims so they wait on the global waitq.
tstore = &kwq->kw_turnstile;
}
......
}

再去查看​​_kwq_use_turnstile​​的定义,代码还是很诚实的,只有在​​KSYN_WQTYPE_MTX​​才会启用​​turnstile​​进行优先级反转保护,而读写锁的类型为​​KSYN_WQTYPE_RWLOCK​​,这说明读写锁不会使用​​_kwq_use_turnstile​​,所以无法避免优先级反转。

#define KSYN_WQTYPE_MTX         0x01
#define KSYN_WQTYPE_CVAR 0x02
#define KSYN_WQTYPE_RWLOCK 0x04
#define KSYN_WQTYPE_SEMA 0x08

static inline bool
_kwq_use_turnstile(ksyn_wait_queue_t kwq)
{
// If we had writer-owner information from the
// rwlock then we could use the turnstile to push on it. For now, only
// plain mutexes use it.
return (_kwq_type(kwq) == KSYN_WQTYPE_MTX);
}

另外在​​_pthread_find_owner​​也可以看到,读写锁的​​owner​​是​​0​

void
_pthread_find_owner(thread_t thread,
struct stackshot_thread_waitinfo * waitinfo)
{
ksyn_wait_queue_t kwq = _pthread_get_thread_kwq(thread);
switch (waitinfo->wait_type) {
case kThreadWaitPThreadMutex:
assert((kwq->kw_type & KSYN_WQTYPE_MASK) == KSYN_WQTYPE_MTX);
waitinfo->owner = thread_tid(kwq->kw_owner);
waitinfo->context = kwq->kw_addr;
break;
/* Owner of rwlock not stored in kernel space due to races. Punt
* and hope that the userspace address is helpful enough. */
case kThreadWaitPThreadRWLockRead:
case kThreadWaitPThreadRWLockWrite:
assert((kwq->kw_type & KSYN_WQTYPE_MASK) == KSYN_WQTYPE_RWLOCK);
waitinfo->owner = 0;
waitinfo->context = kwq->kw_addr;
break;
/* Condvars don't have owners, so just give the userspace address. */
case kThreadWaitPThreadCondVar:
assert((kwq->kw_type & KSYN_WQTYPE_MASK) == KSYN_WQTYPE_CVAR);
waitinfo->owner = 0;
waitinfo->context = kwq->kw_addr;
break;
case kThreadWaitNone:
default:
waitinfo->owner = 0;
waitinfo->context = 0;
break;
}
}

把锁更换为读写锁,验证一下前面的理论是否正确:

pthread_rwlock_init(&_rwlock, NULL);
dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_BACKGROUND, 0), ^{
printf("begin : \n");
printThreadPriority();
printf("queue before lock \n");
pthread_rwlock_rdlock(&_rwlock); //确保 backgroundQueue 先得到锁
printf("queue lock \n");
printThreadPriority();
dispatch_async(dispatch_get_main_queue(), ^{
printf("before main lock\n");
pthread_rwlock_wrlock(&_rwlock);
printf("in main lock\n");
pthread_rwlock_unlock(&_rwlock);
printf("after main unlock\n");
});
sleep(10);
printThreadPriority();
printf("queue unlock\n");
pthread_rwlock_unlock(&_rwlock);
printf("queue after unlock\n");
});
begin : 
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63
queue before lock
queue lock
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63
before main lock
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63
queue unlock
queue after unlock
in main lock
after main unlock

可以看到读写锁不会发生优先级提升。

3.3.4  dispatch_sync

这个​​API​​都比较熟悉了,这里直接验证:

// 当前线程为主线程
dispatch_queue_attr_t qosAttribute = dispatch_queue_attr_make_with_qos_class(DISPATCH_QUEUE_SERIAL, QOS_CLASS_BACKGROUND, 0);
_queue = dispatch_queue_create("com.demo.test", qosAttribute);
printThreadPriority();
dispatch_async(_queue, ^{
printf("dispatch_async before dispatch_sync : \n");
printThreadPriority();
});
dispatch_sync(_queue, ^{
printf("dispatch_sync: \n");
printThreadPriority();
});
dispatch_async(_queue, ^{
printf("dispatch_async after dispatch_sync: \n");
printThreadPriority();
});
pth_priority: 47, pth_curpri: 47, pth_maxpriority: 63 
dispatch_async before dispatch_sync :
pth_priority: 47, pth_curpri: 47, pth_maxpriority: 63
dispatch_sync:
pth_priority: 47, pth_curpri: 47, pth_maxpriority: 63
dispatch_async after dispatch_sync:
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63

​_queue​​是一个低优先级队列(​​QOS_CLASS_BACKGROUND​​),可以看到​​dispatch_sync​​调用压入队列的任务,以及在这之前​​dispatch_async​​压入的任务,都被提升到较高的优先级​​47​​(和主线程一致),而最后一个​​dispatch_async​​的任务则以优先级​​4​​来执行。

3.3.5  dispatch_wait

// 当前线程为主线程
dispatch_queue_attr_t qosAttribute = dispatch_queue_attr_make_with_qos_class(DISPATCH_QUEUE_SERIAL, QOS_CLASS_BACKGROUND, 0);
_queue = dispatch_queue_create("com.demo.test", qosAttribute);
printf("main thread\n");
printThreadPriority();
dispatch_block_t block = dispatch_block_create(DISPATCH_BLOCK_INHERIT_QOS_CLASS, ^{
printf("sub thread\n");
sleep(2);
printThreadPriority();
});
dispatch_async(_queue, block);
dispatch_wait(block, DISPATCH_TIME_FOREVER);

​_queue​​是一个低优先级队列(​​QOS_CLASS_BACKGROUND​​),当在当前主线程使用​​dispatch_wait​​进行等待时,输出如下,低优先级的任务被提升到优先级​​47​

main thread
pth_priority: 47, pth_curpri: 47, pth_maxpriority: 63
sub thread
pth_priority: 47, pth_curpri: 47, pth_maxpriority: 63

而如果将​​dispatch_wait(block, DISPATCH_TIME_FOREVER)​​注释掉之后,输出如下:

main thread
pth_priority: 47, pth_curpri: 47, pth_maxpriority: 63
sub thread
pth_priority: 4, pth_curpri: 4, pth_maxpriority: 63

值得注意的是,​​dispatch_wait​​是一个宏(​​C11​​的泛型),或者是一个入口函数,它可以接受​​dispatch_block_t​​,​​dispatch_group_t​​,​​dispatch_semaphore_t​​ ​​3​​种类型的参数,但是这里的具体含义应该是指​​dispatch_block_wait​​,只有​​dispatch_block_wait​​会调整优先级,避免优先级反转。

intptr_t
dispatch_wait(void *object, dispatch_time_t timeout);
#if __has_extension(c_generic_selections)
#define dispatch_wait(object, timeout) \
_Generic((object), \
dispatch_block_t:dispatch_block_wait, \
dispatch_group_t:dispatch_group_wait, \
dispatch_semaphore_t:dispatch_semaphore_wait \
)((object),(timeout))
#endif

3.4  神秘的信号量

3.4.1 dispatch_semaphore

之前对​​dispatch_semaphore​​的认知非常浅薄,经常把二值信号量和互斥锁划等号。但是通过调研后发现:​​dispatch_semaphore​​ 没有 ​​QoS​​ 的概念,没有记录当前持有信号量的线程(​​owner​​),所以有高优先级的线程在等待锁时,内核无法知道该提高哪个线程的调试优先级(​​QoS​​)。如果锁持有者优先级比其他线程低,高优先级的等待线程将一直等待。Mutexvs Semaphore: What’s the Difference? 一文详细比对了​​Mutex​​和​​Semaphore​​之间的区别。

Semaphores are for signaling (sames a condition variables, events) while mutexes are for mutual exclusion. Technically, you can also use semaphores for mutual exclusion (a mutex can be thought as a binary semaphore) but you really shouldn't.

Right, but libdispatch doesn't have a mutex. It has semaphores and queues. So if you're trying to use libdispatch and you don't want the closure-based aspect of queues, you might be tempted to use a semaphore instead. Don't do that, use os_unfair_lock or pthread_mutex (or a higher-level construct like NSLock) instead.

这些是一些警示,可以看到​​dispatch_semaphore​​十分危险,使用需要特别小心。

这里通过苹果官方提供的demo进行解释:

__block NSString *taskName = nil;
dispatch_semaphore_t sema = dispatch_semaphore_create(0);
[self.connection.remoteObjectProxy requestCurrentTaskName:^(NSString *task) {
taskName = task;
dispatch_semaphore_signal(sema);
}];
dispatch_semaphore_wait(sema, DISPATCH_TIME_FOREVER);
return taskName;
  1. 假设在主线程执行这段代码,那么当前线程的优先级是QOS_CLASS_USER_INTERACTIVE
  2. 由于从主线程进行了异步,异步任务队列的QoS将会被提升为QOS_CLASS_USER_INITIATED
  3. 主线程被信号量sema阻塞,而负责释放该信号量的异步任务的优先级QOS_CLASS_USER_INITIATED低于主线程的优先级QOS_CLASS_USER_INTERACTIVE,因此可能会发生优先级反转。

值得一提的是,​​Clang​​专门针对这种情况进行了静态检测:https://github.com/llvm-mirror/clang/blob/master/lib/StaticAnalyzer/Checkers/GCDAntipatternChecker.cpp

static auto findGCDAntiPatternWithSemaphore() -> decltype(compoundStmt()) {

const char *SemaphoreBinding = "semaphore_name";
auto SemaphoreCreateM = callExpr(allOf(
callsName("dispatch_semaphore_create"),
hasArgument(0, ignoringParenCasts(integerLiteral(equals(0))))));

auto SemaphoreBindingM = anyOf(
forEachDescendant(
varDecl(hasDescendant(SemaphoreCreateM)).bind(SemaphoreBinding)),
forEachDescendant(binaryOperator(bindAssignmentToDecl(SemaphoreBinding),
hasRHS(SemaphoreCreateM))));

auto HasBlockArgumentM = hasAnyArgument(hasType(
hasCanonicalType(blockPointerType())
));

auto ArgCallsSignalM = hasAnyArgument(stmt(hasDescendant(callExpr(
allOf(
callsName("dispatch_semaphore_signal"),
equalsBoundArgDecl(0, SemaphoreBinding)
)))));

auto HasBlockAndCallsSignalM = allOf(HasBlockArgumentM, ArgCallsSignalM);

auto HasBlockCallingSignalM =
forEachDescendant(
stmt(anyOf(
callExpr(HasBlockAndCallsSignalM),
objcMessageExpr(HasBlockAndCallsSignalM)
)));

auto SemaphoreWaitM = forEachDescendant(
callExpr(
allOf(
callsName("dispatch_semaphore_wait"),
equalsBoundArgDecl(0, SemaphoreBinding)
)
).bind(WarnAtNode));

return compoundStmt(
SemaphoreBindingM, HasBlockCallingSignalM, SemaphoreWaitM);
}

如果想使用该功能,只需要打开​​xcode​​设置即可:

另外,​​dispatch_group​​ 跟 ​​semaphore​​ 类似,在调用 ​​enter()​​ 方法时,无法预知谁会调用 ​​leave()​​,所以系统也无法知道其 ​​owner​​是谁,所以同样不会有优先级提升的问题。

3.4.2  信号量卡死现身说法

​dispatch_semaphore​​给笔者的印象非常深刻,之前写过一段这样的代码:使用信号量在主线程同步等待相机授权结果。

__block BOOL auth = NO;
dispatch_semaphore_t semaphore = dispatch_semaphore_create(0);
[KTAuthorizeService requestAuthorizationWithType:KTPermissionsTypeCamera completionHandler:^(BOOL allow) {
auth = allow;
dispatch_semaphore_signal(semaphore);
}];
dispatch_semaphore_wait(semaphore, DISPATCH_TIME_FOREVER);

上线后长期占据卡死​​top1​​,当时百思不得其解,在深入了解到信号量无法避免优先级反转后,终于豁然开朗,一扫之前心中的阴霾。

图片

这类问题一般通过​​2​​种方式来解决:

  1. 使用同步API
BOOL auth = [KTAuthorizeService authorizationWithType:KTPermissionsTypeCamera];
// do something next
  1. 异步回调,不要在当前线程等待
[KTAuthorizeService requestAuthorizationWithType:KTPermissionsTypeCamera completionHandler:^(BOOL allow) {
BOOL auth = allow;
// do something next via callback
}];

4.  几个概念

4.1  turnstile

前文提到​​XNU​​使用​​turnstile​​进行优先级继承,这里对​​turnstile​​机制进行简单的描述和理解。在​​XNU​​内核中,存在着大量的同步对象(例如​​lck_mtx_t​​),为了解决优先级反转的问题,每个同步对象都必须对应一个分离的数据结构来维护大量的信息,例如阻塞在这个同步对象上的线程队列。可以想象一下,如果每个同步对象都要分配一个这样的数据结构,将造成极大的内存浪费。

为了解决这个问题,​​XNU​​采用了​​turnstile​​机制,一种空间利用率很高的解决方案。该方案的提出依据是同一个线程在同一时刻不能同时阻塞于多个同步对象上。这一事实允许所有同步对象只需要保留一个指向​​turnstile​​的指针,且在需要的时候去分配一个​​turnstile​​即可,而​​turnstile​​则包含了操作一个同步对象需要的所有信息,例如阻塞线程的队列、拥有这个同步对象的线程指针。​​turnstile​​是从池中动态分配的,这个池的大小会随着系统中已分配的线程数目增加而增加,所以​​turnstile​​总数将始终低于或等于线程数,这也决定了​​turnstile​​的数目是可控的。​​turnstile​​由阻塞在该同步对象上的第一个线程负责分配,当没有更多线程阻塞在该同步对象上,​​turnstile​​会被释放,回收到池中。

​turnstile​​的数据结构如下:

struct turnstile {
struct waitq ts_waitq; /* waitq embedded in turnstile */
turnstile_inheritor_t ts_inheritor; /* thread/turnstile inheriting the priority (IL, WL) */
union {
struct turnstile_list ts_free_turnstiles; /* turnstile free list (IL) */
SLIST_ENTRY(turnstile) ts_free_elm; /* turnstile free list element (IL) */
};
struct priority_queue_sched_max ts_inheritor_queue; /* Queue of turnstile with us as an inheritor (WL) */
union {
struct priority_queue_entry_sched ts_inheritor_links; /* Inheritor queue links */
struct mpsc_queue_chain ts_deallocate_link; /* thread deallocate link */
};
SLIST_ENTRY(turnstile) ts_htable_link; /* linkage for turnstile in global hash table */
uintptr_t ts_proprietor; /* hash key lookup turnstile (IL) */
os_refcnt_t ts_refcount; /* reference count for turnstiles */
_Atomic uint32_t ts_type_gencount; /* gen count used for priority chaining (IL), type of turnstile (IL) */
uint32_t ts_port_ref; /* number of explicit refs from ports on send turnstile */
turnstile_update_flags_t ts_inheritor_flags; /* flags for turnstile inheritor (IL, WL) */
uint8_t ts_priority; /* priority of turnstile (WL) */

#if DEVELOPMENT || DEBUG
uint8_t ts_state; /* current state of turnstile (IL) */
queue_chain_t ts_global_elm; /* global turnstile chain */
thread_t ts_thread; /* thread the turnstile is attached to */
thread_t ts_prev_thread; /* thread the turnstile was attached before donation */
#endif
};

4.2  优先级数值

在验证环节有一些优先级数值,这里借助「Mac OS® X and iOS Internals」解释一下:实验中涉及到的优先级数值都是相对于​​Mach​​层而言的,且都是用户线程数值。

  1. 用户线程的优先级是0~63;
    a.NSQualityOfServiceBackgroundMach层级优先级数是4;
    b.NSQualityOfServiceUtility的Mach层级优先级数是20;
    c.NSQualityOfServiceDefaultMach层级优先级数是31;
    d.NSQualityOfServiceUserInitiated的Mach层级优先级数是37;
    e.NSQualityOfServiceUserInteractive的Mach层级优先级是47。
  2. 内核线程的优先级是80~95;
  3. 实时系统线程的优先级是96~127;
  4. 64~79被保留给系统使用。

图片

5. 总结

本文主要阐述了优先级反转的一些概念和解决思路,并结合​​iOS​​平台的几种锁进行了详细的调研。通过深入的理解,可以去规避一些不必要的优先级反转,从而进一步避免卡死异常。字节跳动 ​​APM​​团队也针对线程的优先级做了监控处理,进而达到发现和预防优先级反转的目的。

6. 参考文档

责任编辑:庞桂玉 来源: 字节跳动技术团队
相关推荐

2012-08-14 09:38:29

WAN优化

2010-09-13 17:30:07

CSS优先级

2010-09-01 14:10:36

CSS优先级

2010-08-31 11:04:48

CSS优先级

2009-08-28 17:10:59

C#线程优先级

2010-03-18 14:09:20

Java线程同步

2020-09-30 09:07:37

DevOps

2021-04-06 10:45:18

React前端优先级

2015-06-02 11:26:29

产品团队

2009-08-12 13:07:13

C#运算符优先级

2010-09-06 14:53:27

CSS优先级

2012-06-27 10:59:05

WAN优化应用交付

2023-11-03 08:22:09

Android系统算法

2023-01-05 08:48:57

技术管理排优先级

2011-03-14 10:17:40

JavaScript

2020-06-04 08:36:55

Linux内核线程

2023-08-08 08:28:03

消息消费端Spring

2024-04-23 14:13:38

开发配置文件

2011-02-25 14:35:00

2021-06-02 08:33:31

TPCTPC-H系统
点赞
收藏

51CTO技术栈公众号