[bug] 偶现，在脚本中使用大量go(function () {})处理任务偶尔会卡死在epoll_wait

[omigafu]

Please answer these questions before submitting your issue.
不是在http_server中运行，是在脚本模式下

1. What did you do? If possible, provide a simple script for reproducing the error.

Swoole\Coroutine\run(function ()  {
    ...
    foreach ($postData as $k => $v) {
        go(function () use ($chan, $k, $v) {
             $client = new Swoole\Coroutine\Client(SWOOLE_SOCK_TCP);
            $client->connect($v['ip'], $v['port'], 2)
            $client->send($in);
            $rs = $client->recv(1);
            $client->close();
            $chan->push(['id' => $k, 'body' => $rs], 4);
        });
    }
    for ($i = 0; $i < $realCount; $i++) {
        $result = $chan->pop(4);
        $returns[$result['id']] = $result['body'];
    }
});

3. What did you expect to see?
   期望能正常执行完任务

4. What did you see instead?
   进程停在了epoll_wait

#0  0x00007f7f416d6e87 in epoll_wait () from /lib64/libc.so.6
#1  0x00007f7f3e9eff9d in swoole::ReactorEpoll::wait (this=0x36fffd0, timeo=<optimized out>) at /usr/local/services/swoole-src-4.8.13/include/swoole_reactor.h:269
#2  0x00007f7f3e93c52b in swoole::Reactor::wait (this=<optimized out>, timeout=0x0) at /usr/local/services/swoole-src-4.8.13/include/swoole_reactor.h:165
#3  php_swoole_event_wait () at /usr/local/services/swoole-src-4.8.13/ext-src/swoole_event.cc:278
#4  php_swoole_event_wait () at /usr/local/services/swoole-src-4.8.13/ext-src/swoole_event.cc:262
#5  0x00007f7f3e93801f in zim_swoole_coroutine_scheduler_start (execute_data=<optimized out>, return_value=0x7f7f3f216550) at /usr/local/services/swoole-src-4.8.13/ext-src/swoole_coroutine_scheduler.cc:319
#6  0x00000000009e31b6 in ZEND_DO_FCALL_SPEC_RETVAL_USED_HANDLER () at /usr/local/services/php-7.4.12/Zend/zend_vm_execute.h:1730
#7  execute_ex (ex=0xf) at /usr/local/services/php-7.4.12/Zend/zend_vm_execute.h:53865
#8  0x0000000000954246 in zend_call_function (fci=fci@entry=0x7ffcb4dd1300, fci_cache=<optimized out>, fci_cache@entry=0x7ffcb4dd1370) at /usr/local/services/php-7.4.12/Zend/zend_execute_API.c:820
#9  0x00007f7f3e90b298 in sw_zend_call_function_ex (retval=0x0, params=0x7f7f3f213bd0, param_count=1, fci_cache=0x7ffcb4dd1370, function_name=0x0) at /usr/local/services/swoole-src-4.8.13/ext-src/php_swoole_private.h:974
#10 zend::function::call (fci_cache=fci_cache@entry=0x7ffcb4dd1370, argc=argc@entry=1, argv=argv@entry=0x7f7f3f213bd0, retval=retval@entry=0x0, enable_coroutine=<optimized out>) at /usr/local/services/swoole-src-4.8.13/ext-src/php_swoole_cxx.cc:100
#11 0x00007f7f3e962098 in php_swoole_process_start (process=0x7f7f2e97f0a0, zobject=0x7f7f3f213bd0) at /usr/local/services/swoole-src-4.8.13/ext-src/swoole_process.cc:739
#12 0x00007f7f3e962117 in zim_swoole_process_start (execute_data=0x7f7f3f213bb0, return_value=0x7ffcb4dd1430) at /usr/local/services/swoole-src-4.8.13/ext-src/swoole_process.cc:771
#13 0x00000000009e3548 in ZEND_DO_FCALL_SPEC_RETVAL_UNUSED_HANDLER () at /usr/local/services/php-7.4.12/Zend/zend_vm_execute.h:1618
#14 execute_ex (ex=0xf) at /usr/local/services/php-7.4.12/Zend/zend_vm_execute.h:53861
#15 0x0000000000954246 in zend_call_function (fci=fci@entry=0x7ffcb4dd1560, fci_cache=<optimized out>, fci_cache@entry=0x7ffcb4dd15d0) at /usr/local/services/php-7.4.12/Zend/zend_execute_API.c:820
#16 0x00007f7f3e90b298 in sw_zend_call_function_ex (retval=0x0, params=0x7f7f3f213350, param_count=1, fci_cache=0x7ffcb4dd15d0, function_name=0x0) at /usr/local/services/swoole-src-4.8.13/ext-src/php_swoole_private.h:974
#17 zend::function::call (fci_cache=fci_cache@entry=0x7ffcb4dd15d0, argc=argc@entry=1, argv=argv@entry=0x7f7f3f213350, retval=retval@entry=0x0, enable_coroutine=<optimized out>) at /usr/local/services/swoole-src-4.8.13/ext-src/php_swoole_cxx.cc:100
#18 0x00007f7f3e962098 in php_swoole_process_start (process=0x7f7f3f39b1e0, zobject=0x7f7f3f213350) at /usr/local/services/swoole-src-4.8.13/ext-src/swoole_process.cc:739
#19 0x00007f7f3e962117 in zim_swoole_process_start (execute_data=0x7f7f3f213330, return_value=0x7ffcb4dd1690) at /usr/local/services/swoole-src-4.8.13/ext-src/swoole_process.cc:771
#20 0x00000000009e3548 in ZEND_DO_FCALL_SPEC_RETVAL_UNUSED_HANDLER () at /usr/local/services/php-7.4.12/Zend/zend_vm_execute.h:1618
#21 execute_ex (ex=0xf) at /usr/local/services/php-7.4.12/Zend/zend_vm_execute.h:53861
#22 0x00000000009e40a3 in zend_execute (op_array=0x7f7f3f27bee0, return_value=<optimized out>) at /usr/local/services/php-7.4.12/Zend/zend_vm_execute.h:57957
#23 0x0000000000962044 in zend_execute_scripts (type=type@entry=8, retval=0x7f7f3f213020, retval@entry=0x0, file_count=file_count@entry=3) at /usr/local/services/php-7.4.12/Zend/zend.c:1677
#24 0x00000000009043e0 in php_execute_script (primary_file=primary_file@entry=0x7ffcb4dd3c30) at /usr/local/services/php-7.4.12/main/main.c:2621
#25 0x00000000009e6133 in do_cli (argc=5, argv=0x340f030) at /usr/local/services/php-7.4.12/sapi/cli/php_cli.c:964
#26 0x0000000000642c88 in main (argc=5, argv=0x340f030) at /usr/local/services/php-7.4.12/sapi/cli/php_cli.c:1359

5. What version of Swoole are you using (show your php --ri swoole)?
   swoole

Swoole => enabled
Author => Swoole Team [email protected]
Version => 4.8.13
Built => Nov 13 2023 15:55:26
coroutine => enabled with boost asm context
epoll => enabled
eventfd => enabled
signalfd => enabled
cpu_affinity => enabled
spinlock => enabled
rwlock => enabled
sockets => enabled
openssl => OpenSSL 1.1.1k FIPS 25 Mar 2021
dtls => enabled
zlib => 1.2.11
mutex_timedlock => enabled
pthread_barrier => enabled
futex => enabled
async_redis => enabled

Directive => Local Value => Master Value
swoole.enable_coroutine => On => On
swoole.enable_library => On => On
swoole.enable_preemptive_scheduler => Off => Off
swoole.display_errors => On => On
swoole.use_shortname => On => On
swoole.unixsock_buffer_size => 8388608 => 8388608

swoole

Swoole => enabled
Author => Swoole Team [email protected]
Version => 5.1.2
Built => Aug 13 2024 17:02:50
coroutine => enabled with boost asm context
epoll => enabled
eventfd => enabled
signalfd => enabled
cpu_affinity => enabled
spinlock => enabled
rwlock => enabled
sockets => enabled
openssl => OpenSSL 3.0.12 24 Oct 2023
dtls => enabled
http2 => enabled
json => enabled
zlib => 1.2.13
brotli => E16781312/D16781312
mutex_timedlock => enabled
pthread_barrier => enabled
futex => enabled
async_redis => enabled

Directive => Local Value => Master Value
swoole.enable_coroutine => On => On
swoole.enable_library => On => On
swoole.enable_fiber_mock => Off => Off
swoole.enable_preemptive_scheduler => Off => Off
swoole.display_errors => On => On
swoole.use_shortname => On => On
swoole.unixsock_buffer_size => 8388608 => 8388608
6. What is your machine environment used (show your uname -a & php -v & gcc -v) ?
PHP 7.4.12

PHP 8.3.7

2个php版本和swoole版本都存在这个问题，不过是偶现，没有任何报错，进程就卡死在epoll_wait上，请问有人遇到过类似的问题吗？

[NathanFreeman]

多协程并行要用Swoole\Coroutine\WaitGroup保证主协程不退出。
也需要检查一下是不是send过程中数据没过去

[omigafu]

多协程并行要用Swoole\Coroutine\WaitGroup保证主协程不退出。 也需要检查一下是不是send过程中数据没过去

1. WaitGroup也有使用过，也是一样的情况
2. 如果是send失败的话，recv有设置超时时间，正常应该也不会出现这个问题
3. 复现难度还是比较大，要执行大规模离线任务才可能复现

[matyhtf]

这一般都是你的 php 代码层面存在逻辑错误，主线程在 epoll_wait 处轮询是符合预期的，没有任何事件时将进入 epoll 等待。

你可以记录协程 id，然后通过 Coroutine::getBackTrace() 获取对应协程挂起的位置。

[omigafu]

这一般都是你的 php 代码层面存在逻辑错误，主线程在 epoll_wait 处轮询是符合预期的，没有任何事件时将进入 epoll 等待。

你可以记录协程 id，然后通过 Coroutine::getBackTrace() 获取对应协程挂起的位置。

感谢，我通过这个方式再尝试下定位。
我们有办法给Coroutine设置一个超时时间吗？

[LIngMax]

我这边也遇到类似的问题 不停创建协程进入阻塞时会遇到 该问题

协程1 调用$lock->lock() ;sleep(1);$lock->unlock() ;
协程2调用 $lock->lock();sleep(1); $lock->unlock() ;

协程3调用 $lock->lock();sleep(1); $lock->unlock() ;
同时启动协程会造成协程3彻底卡死 再也唤不起

通常代替方案 浪费点性能 解决这个问题
while(!$lock->trylock())usleep(1000*10);#10ms#等待锁#协程挂起#让出cpu

至于chan没有检查是否阻塞 那就没办法了

[LIngMax]

   push写入前要判断下是通道   否满了     Swoole\Coroutine\Channel->isFull()

[NathanFreeman]

协程中不能用互斥锁和sleep，容易导致死锁

[LIngMax]

while($chan->isFull())usleep(1000*10);#10ms#等待 通道有空位置#协程挂起#让出cpu
这种只是个笨办法

希望官方能换个阻塞处理方式吧
希望给事件循环去唤醒 而不是阻塞卡死全部协程

[matyhtf]

Channel 本身就具有协程同步的能力，不需要使用锁。相反如果使用了锁，可能会导致进程死锁。

目前 Swoole\Lock 是进程间锁，Swoole\Thread\Lock 是线程间锁，均无法用于协程环境。