Re: [ns] memory leak problem

[John Heidemann <johnh@ISI.EDU>]

>What TCP Agents are you using?  I'm also fighting a leak when using
>FullTcpNewreno.  The number of packets allocated (with "new") keeps
>growing unboundedly, indicating that somewhere some packets are not being
>freed.  When I find a fix for this, I'll drop you a note.
>
>George
>
>					
>			-George F. Riley (riley@cc.gatech.edu)
>
>On Tue, 14 Nov 2000, Milind Dilip Kopikare wrote:
>
>> Hi.
>>   I am trying to simulate 1000 TCP flows sending packets over a 100Mb link
>> to 1000 TCP receivers at the other end. The topology is as shown.
>> 
>>     s1---------|		  |--------r1
>>     s2---------|		  |--------r2
>>    :----------|int1----------int2|--------:
>>    :----------|	100Mb	  |--------:
>>     s1000------|		  |--------r1000
>> 
>> Each of the sources transmit at 100Mb so the link between int1 and int2 is
>> always full. However I am running into memory leak problems. I tries using
>> Eric Weigle's fix of "delete [] he_old" in heap_insert() in scheduler.cc.
>> But this hasn't solved the problem. Would anyone be able to help as to why
>> this memory leak is taking place?
>> 

A good way to attack memory leaks is to run dmalloc and then
dmalloc_summarize... the places that are leaking should bubble to the
top.  This should also differentiate a leak from just using a lot of
memory.

   -John Heidemann