Hello, main issue being reported is "Memory Leaks" in net-snmp v5.7.2 and v5.7.1. Support Key: 44ae-0f1e-cd36-1abb
First off I am totally new to net-snmp and the support forum mechanism usage, so I hope I am posting to the right forum. I just want to get this memory leaking issues reported and fixed. I have been developing a snmp trap forwarder using net-snmp v5.7.1 package on a Linux RHEL5 server. I had made a lot of modifications to the snmptrapd so that it can filter and/or manipulate traps based on trap OID and then forwards the desired traps on to 3 destinations. Thus, I have modified the snmptrapd.c and snmptrapd_handlers.c programs. I have added logic to the snmptrapd.c program to open 3 destination ports (sessions) and added reading in my own custom configuration file to specify the trap filtering parameters. The 3 opened sessions remain opened for the duration of running the snmptrapd program. In the snmptrapd_handlers.c program, I've added logic in the snmp_input callback section to examine the received traps and set filters to either forward the trap or to ignore and not forward the trap to the 3 forwarding destination session ports. I clone the original pdu several times to send to the various destination ports. (I free all pdu's after sending or blocking the trap). I also added logic to the snmptrapd_handlers.c (forward_handler callback routine) modify certain traps adding and appending the agent ipaddress as an extra varbind to some trap pdu's before forwarding it. I also have a special case where I have to clone just the pdu header and then re-populate the trap's varbinds with translated to text varbind values. In all the new logic that I've added, I've checked over and over to make sure I am free'ing up any allocated memory. valgrind utility reports no memory lost and no errors now. But when I run the snmptrapd program, it slowly creeps up in memory usage until it consumes all of the systems memory! I have in the snmptrapd.conf file these handlers turned on: snmpTrapdAddr 192.168.1.16:162 authCommunity log,net public forward default udp:192.168.1.15:162 format1 %t %.4y-%.2m-%.2l %.2h:%.2j:%.2k Received event %s %A (via [%b]; %V;%v\n format2 %t %.4y-%.2m-%.2l %.2h:%.2j:%.2k Received event %s %A (via [%b]; %V;%v\n I have done as thorough a check as I can up to this point to ensure none of my own added code is leaking and locking up the system memory. I have tried disabling the forward_handler and run it with the logging handler only to isolate the trap forwarder logic out of the equation and it still leaks memory. I have tried disabling the log handler and run only the forward handler too and it still leaks memory. The memory usage gradually increases as trap activity is received no matter which handler is enabled or disabled. This leads me to suspect that somewhere in the net-snmp handler dispatching or callback mechanism or elsewhere is continually allocating bits of memory to variables that it keeps around forever. Thus eating up memory as time runs on. Also, I should mention that I do have configured snmp.conf file to load in a bunch of MIBS along with the default standard MIBS that come with the net-snmp package. When I installed and built the package I used the following configure command: ./configure --enable-embedded-perl --enable-shared I ran make I ran make test I ran as root, make install using the default install directories of /usr/local I am at my wits end. Where can there be memory allocations happening in the net-snmp v5.7.1 and v5.7.2 package that eats up memory like this? Is it in the logging format checking routines? Is it in the MIB parsing routines? Is it in the handler dispatching routines? Where in net-snmp code can it constantly be adding to a memory variable increasing it in size with time? I had developed this code for a critical project which was about to be deployed but then during User Acceptance Testing they discovered that snmptrapd was eating up all the system memory and now the project is delayed and on hold. Desperately need help in locating the memory allocations in net-snmp causing this. Also if it helps, this happens on both a Fedora OS as well as Linux RHEL5 OS server. Also, I'm pretty sure even if I compile the original snmptrapd programs that come with the package and run it, it also will show memory usage creeping up. So it is inherent in the package. Any ideas or patches to fix this? Thank You! Bill Yee ------------------------------------------------------------------------------ LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial Remotely access PCs and mobile devices and provide instant support Improve your efficiency, and focus on delivering more value-add services Discover what IT Professionals Know. Rescue delivers http://p.sf.net/sfu/logmein_12329d2d _______________________________________________ Net-snmp-coders mailing list Net-snmp-coders@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
