libuv dead cycle in cmpxchgi #2462
Labels
Comments
|
Can you fill out the issue template? Kernel and libuv version are important to know. I infer that "dead cycle" means it busy-loops? Can you attach to the process with gdb and post the output of Is there an easy way for me to reproduce the busy loop? |
The libuv I complied is V1.x , debug version. Linux localhost.localdomain 3.10.0-862.14.4.el7.x86_64 #1 SMP Wed Sep 26 15:12:11 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux The "busy loop" run in main thread 1, what I do is I hope the socket wirting opration happend in main thread , so I use the uv_asyn_t object to notify loop thread , the uv_asyn_init callback to write data in loop thread . But I create so many uv_asyn_t object in task thread .for every write req , I create on uv_asyn_t and put them back to a queue in loop thread after its callback finished. long time , this happend. |
|
Okay, thanks. From your description it sounds like libuv runs on thread 1 but you call That won't work, only |
|
Thanks for your patiention .Yes，I use uv_async_init in other thread which uv_loop is not run in. And uv_async_init is not thread-safe，why process doesn't crash？for example，uv_write，uv_close both cause visual errors ，process coredump，if they don't run in uv_run thread.By the way，based on my experiment using the libuv，it seems that it just could work with single io thread mode，all socket stay in one loop. Can I create mutile threads and distribute socket to different loop of different thread？ |
Because it's undefined behavior. There's no guarantee about what, if anything, will happen.
Yes, with some effort. See I'll close this out because it's not a libuv bug. If you have follow-up question, then please post them to the libuv/help repo. |
|
okay，thanks very much.------------------ 原始邮件 ------------------
发件人: "Ben Noordhuis"<notifications@github.com>
发送时间: 2019年9月11日(星期三) 凌晨3:28
收件人: "libuv/libuv"<libuv@noreply.github.com>;
抄送: "wangshangxiu"<894616049@qq.com>;"Author"<author@noreply.github.com>;
主题: Re: [libuv/libuv] libuv dead cycle in cmpxchgi (#2462)
why process doesn't crash？
Because it's undefined behavior. There's no guarantee about what, if anything, will happen.
Can I create mutile threads and distribute socket to different loop of different thread？
Yes, with some effort. See test/benchmark-multi-accept.c for an example, the uv_write2() call in particular.
I'll close this out because it's not a libuv bug. If you have follow-up question, then please post them to the libuv/help repo.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or mute the thread.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Stack:
top:
Stack is the last stack of cm_server base on libuv , top is cm_server runing status, in fact ,it doesn't run , just in libuv.
The text was updated successfully, but these errors were encountered: