[ovs-discuss] 答复: ????: memory leak in openvswitch-2.0.0

Yinpeijun yinpeijun at huawei.com
Wed Apr 9 07:36:08 UTC 2014


>On Wed, Apr 09, 2014 at 02:23:13AM +0000, Yinpeijun wrote:
>> >On Tue, Apr 08, 2014 at 03:51:33AM +0000, Yinpeijun wrote:
>> >> Hello everyone, Recently, I use openvswitch-2.0.0 to test send 
>> >> packets between the vms, duration the test I use valgrind to
>> >> 
>> >> catch if there is memory leak in ovs-vswitched process , after test I see the valgrind log as follow :
>...
>> >> So , I should annotation or delete the code above? Any advice on 
>> >> this would be greatly appreciated.
>> >
>> >This is certainly a bug that should be fixed, but I'm having trouble 
>> >matching up the line numbers in the valgrind output above with the code that you quoted.  What are the line numbers for the code you quoted above?  Are you using exactly the openvswitch 2.0.0 release, or a different release, or a modified version of some release?
>> 
>> Thank you for your reply Ben. 
>> sorry for my mistake ,I exactly use a openvswitch 2.0.0 release but I have added some our own code, and the valgrind output above match the line numbers as follow:
>
>Can you explain how to trigger the issue?  I spent a little time running

I use two servers and create 20 vms in each , which are divided into three classes: no vlan ,vlan 20 and vxlan 65536(our own scheme ), then vms to vms send TCP or UDP packets with netperf .

>2.0.0 today under valgrind and didn't manage to trigger it.  (Also, since you've made significant changes to OVS, it is possible that one of those changes introduced the problem.)

To ensure whether our changes introduced the problem,I will use openvswitch-2.0.0 release without modification to test in the same environment above. 

Thank you for you reply.


More information about the discuss mailing list