Skip to main content

How to debug a deadlock in multithreaded c/c++ program

Debugging a deadlock in multithreaded program can be little tricky .
So the rescuer GDB is there for the help.I am just giving here the head start one needs to debug a multithreaded program for deadlocks.
Run your program in GDB, wait until the deadlock happens and then use the following commands.
1.info threads
which will list all of the threads that are running and a little * on the one which is currently running.
Sample output 
5 Thread 0x7ffff6be3700 (LWP 20175) 0x00000032cf6de8b3 in select () from /lib64/libc.so.6
3 Thread 0x7ffff75e4700 (LWP 20173) 0x00000032cf6ab15d in nanosleep () from /lib64/libc.so.6
2 Thread 0x7ffff7fe5700 (LWP 20172) 0x00000032cf6cd177 in sched_yield () from /lib64/libc.so.6
* 1 Thread 0x7ffff7fe7720 (LWP 20168) 0x00000032cf6de8b3 in select () from /lib64/libc.so.6

 So one can see where all the threads are waiting at this point, so if happens to be a dead lock, threads must be stuck on lock or unlock.

 to see whats the thread name and all one can use the following commands.
2. thr 1
which will take us to thread 1 code. You should go to thread which is waiting on some lock or which you think is causing the deadlock.
Sample output
[Switching to thread 1 (Thread 0x7ffff7fe7720 (LWP 20168))]#0 0x00000032cf6de8b3 in select () from /lib64/libc.so.6

 3. bt
back trace, tada you should now come to know about ur thread name and all the other details.
Sample output
#0 0x00000032cf6de8b3 in select () from /lib64/libc.so.6
#1 0x000000000040c1fd in blm_rule_recv_worker_start () at some_program.cpp:549
#2 0x0000000000404977 in main (argc=2, argv=0x7fffffffe118) at main_some_prog.cpp:356

 Best of luck :D

Comments

Post a Comment

Popular posts from this blog

Enable stats GUI on haproxy.

Add bottom snippet to the haproxy.conf below the defaults section. listen  stats         bind 19.41.259.10:1234         mode            http         log             global         maxconn 10         clitimeout      100s         srvtimeout      100s         contimeout      100s         timeout queue   100s         stats enable         stats hide-version         stats refresh 30s         stats show-node         stats auth admin:password         stats uri  /haproxy?stats Make sure you are updating the IP address on the bind to your VIP and if you want, you can change th...
Post a Comment
Read more

Sending a SIGHUP signal to some external process from Python script

Code : import psutil import os import signal pids = psutil.get_pid_list() for pid in pids: if psutil.Process(pid).name == "process_name": os.kill(pid,signal.SIGHUP) break Steps to follow. 1.Get the PID of the process, in this case  "process_name"   to which you want to send out a SIGHUP signal. 2.Use os.kill(pid,sig) command to send out the SIGHUP signal to that process. 1.Get the PID of the process to which you want to send out a SIGHUP signal. One has to install a package called psutil by the following command. easy_install psutil Check out the following links for more details https://code.google.com/p/psutil/ https://pypi.python.org/pypi/psutil use psutil.get_pid_list() to get all of the PIDs. psutil.get_pid_list() works in the following manner.  pids = [ int ( x ) for x in os . listdir ( '/proc' ) if x . isdigit ()] return pids once you get all the PIDs get the PID you are i...
Post a Comment
Read more

How to enable Openstack Octavia, LBaaS V2 with devstack.

Little Intro to Octavia. Octavia is a service manager for Openstack Load balancer as a service. Neutron LBaaS V2 plugin talks to Octavia through Octavia driver. Octavia driver talks to Octavia api(o-api) and that in turn talks to Octavia control worker(o-cw). Neutron LBaaS V2 Plugin ----> Octavia plugin driver -----> o-api ----> o-cw Other than o-api and o-cw Octavia has 2 more components, housekeeping(o-hk) and health manager(o-hm). o-api Octavia api server which receives all the request from octavia LBaaS driver and passes it to o-cw. o-cw It's the main workhorse, it creates the load balancer VMs( amphorae ), configures them. o-hk keeps track of spare amphorae. o-hm Manages the health of the amphorae through heartbeats, collects the stats and takes care of fail over by creating a new amphora when it fails to get the heartbeat. How to enable Openstack Octavia, LBaaS V2 with Devstack. Since octavia uses a VM for loadbalancer its needs good a...
Post a Comment
Read more