zcfh opened a new pull request, #2918:
URL: https://github.com/apache/brpc/pull/2918
### What problem does this PR solve?
Issue Number:
Problem Summary:
`bthread_usleep`/`bthread_yield` 包含对 tls_task_group 的访问，在LTO模式下，会触发 #2156
提到的问题。在example下会出现 `bthread=4294967808 sched_to itself!`的错误，实际服务中会导致程序出core。
#### 复现方式
1. 使用编译器clang-17.0.6 （测试过 clang-11不会）
2. 增加编译选项 -flto=thin,
3. 新增链接选项 -fuse-ld=lld
4. 编译 parallel_echo_c++
5. 执行 `./parallel_echo_client --use_bthread=true` 后 `Ctrl` + `C`
退出，server可以不需要启动。
#### 汇编对比
clang-11 + thinlto。会内联bthread_usleep，但不会做tls变量访问的优化，此时不会出现问题。
```as
0000000000781ba0 <brpc::SocketMap::WatchConnections()>:
; brpc::SocketMap::WatchConnections():
; /home/wuminghui03/workspace/svn/brpc/src/brpc/socket_map.cpp:354
......
; /home/wuminghui03/workspace/svn/brpc/src/brpc/socket_map.cpp:359
781c10: bf 40 42 0f 00 movl $0xf4240, %edi
# imm = 0xF4240
781c15: e8 46 98 fc ff callq 0x74b460
<bthread_usleep>
......
781ca0: 0f 84 6a ff ff ff je 0x781c10
<brpc::SocketMap::WatchConnections()+0x70>
......
000000000074b460 <bthread_usleep>:
; bthread_usleep():
......
; /home/wuminghui03/workspace/svn/brpc/src/bthread/bthread.cpp:525
74b46b: 64 48 8b 04 25 38 fe ff ff movq %fs:-0x1c8, %rax
74b474: 48 89 45 f8 movq %rax, -0x8(%rbp)
......
```
clang-17 + thinlto, 在循环外缓存了tls变量的地址，此时会出现问题
```as
0000000000781bf0 <brpc::SocketMap::WatchConnections()>:
; brpc::SocketMap::WatchConnections():
; /home/wuminghui03/workspace/svn/brpc/src/brpc/socket_map.cpp:354
......
781c4a: 64 48 8b 04 25 00 00 00 00 movq %fs:0x0, %rax
781c53: 48 8d 80 38 fe ff ff leaq -0x1c8(%rax), %rax
781c5a: 48 89 85 48 ff ff ff movq %rax, -0xb8(%rbp)
781c61: 31 c0 xorl %eax, %eax
781c63: 48 89 45 a0 movq %rax, -0x60(%rbp)
781c67: 4c 8d 6d c8 leaq -0x38(%rbp), %r13
781c6b: 48 89 7d a8 movq %rdi, -0x58(%rbp)
781c6f: 90 nop
781c70: 48 8b 85 48 ff ff ff movq -0xb8(%rbp), %rax
; /home/wuminghui03/workspace/svn/brpc/src/bthread/bthread.cpp:525
781c77: 48 8b 00 movq (%rax), %rax
781c7a: 48 89 45 d0 movq %rax, -0x30(%rbp)
......
781d2f: 0f 84 3b ff ff ff je 0x781c70
<brpc::SocketMap::WatchConnections()+0x80>
......
```
### What is changed and the side effects?
#### Changed:
bthread_usleep 和 bthread_yield 禁用inline优化。
#### 替代方案
使用 BAIDU_GET_VOLATILE_THREAD_LOCAL 宏的方式，访问
tls_task_group，但只要避免bthread_usleep，被inline，就不会做后续的优化。看各位有什么建议。
Side effects:
- Performance effects(性能影响):
对于非LTO模式下，无影响；
对于LTO模式下，会让bthread_usleep 和 bthread_yield 不再被内联。
- Breaking backward compatibility(向后兼容性):
---
### Check List:
- Please make sure your changes are compilable(请确保你的更改可以通过编译).
- When providing us with a new feature, it is best to add related
tests(如果你向我们增加一个新的功能, 请添加相关测试).
- Please follow [Contributor Covenant Code of
Conduct](https://github.com/apache/brpc/blob/master/CODE_OF_CONDUCT.md).(请遵循贡献者准则).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: dev-unsubscr...@brpc.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@brpc.apache.org
For additional commands, e-mail: dev-h...@brpc.apache.org
