如何定位死循环或高CPU使用率(linux) -

来源:互联网 发布:小仙有毒 知乎 编辑:程序博客网 时间:2024/04/19 18:33

确定是CPU过高

使用top观察是否存在CPU使用率过高现象

找出线程

对CPU使用率过高的进程的所有线程进行排序

ps H -e -o pid,tid,pcpu,cmd --sort=pcpu |grep xxx
得到如下结果,其中线程2909使用了7.8%的CPU. 
2907 2913 0.0 ./xxx 
2907 2909 7.8 ./xxx
也可以通过查看/proc中的信息来确定高CPU线程. 打印了4列,线程ID,线程名,用户时间和内核时间(排名未分先后) 
awk '{print $1,$2,$14,$15}' /proc/2907/task/*/stat

找出调用栈

使用gdb attach nmsagent所在的进程,在gdb中使用 info threads显示所有线程

gdbgdb>attach 2907gdb>info threads

得到如下结果,可以发现2909线程的编号是12

  13 Thread 0xad5f2b70 (LWP 2908)  0x004ef0d7 in mq_timedreceive () from /lib/tls/i686/cmov/librt.so.1  12 Thread 0xad58eb70 (LWP 2909)  0x006e0422 in __kernel_vsyscall ()  11 Thread 0xad52ab70 (LWP 2910)  0x006e0422 in __kernel_vsyscall ()  10 Thread 0xad4f8b70 (LWP 2911)  0x006e0422 in __kernel_vsyscall ()  9 Thread 0xad4c6b70 (LWP 2912)  0x006e0422 in __kernel_vsyscall ()  8 Thread 0xad3feb70 (LWP 2913)  0x004ef0d7 in mq_timedreceive () from /lib/tls/i686/cmov/librt.so.1  7 Thread 0xace08b70 (LWP 2914)  0x004ef0d7 in mq_timedreceive () from /lib/tls/i686/cmov/librt.so.1  6 Thread 0xac607b70 (LWP 2915)  0x006e0422 in __kernel_vsyscall ()  5 Thread 0xac5e6b70 (LWP 2916)  0x006e0422 in __kernel_vsyscall ()  4 Thread 0xac361b70 (LWP 2917)  0x006e0422 in __kernel_vsyscall ()  3 Thread 0xac2fdb70 (LWP 2918)  0x006e0422 in __kernel_vsyscall ()  2 Thread 0xac1fcb70 (LWP 2919)  0x004ef0d7 in mq_timedreceive () from /lib/tls/i686/cmov/librt.so.1* 1 Thread 0xb78496d0 (LWP 2907)  0x006e0422 in __kernel_vsyscall ()

使用thread 切换线程,使用bt显示线程栈

gdb>thread 12gdb>bt

得到如下线程栈

#0  0x006e0422 in __kernel_vsyscall ()#1  0x001cca26 in nanosleep () from /lib/tls/i686/cmov/libc.so.6#2  0x001fc2dc in usleep () from /lib/tls/i686/cmov/libc.so.6#3  0x0806b510 in OspTaskDelay ()#4  0x0805c710 in CDispatchTask::NodeMsgSendToSock() ()#5  0x0805cc74 in DispatchTaskEntry ()#6  0x0806a8e9 in OspTaskTemplateFunc(void*) ()#7  0x00d4780e in start_thread () from /lib/tls/i686/cmov/libpthread.so.0 #8  0x002027ee in clone () from /lib/tls/i686/cmov/libc.so.6

 ps + strace

得到进程ID 21465

ps -e |grep cmu 4996 ?        00:00:25 cmu_fjga_sp321465 pts/5    00:08:10 cmu

得到线程时间, 其中最占CPU的是 EpollRecvTask 21581

ps -eL |grep 21465 
21465 21579 pts/5 00:00:00 CamApp 
21465 21580 pts/5 00:00:00 TimerMan Task 
21465 21581 pts/5 00:09:02 EpollRecvTask 
21465 21582 pts/5 00:00:00 

使用 strace -p 21581 得到线程栈

原创粉丝点击