On Thu, 2008-06-05 at 20:22 +0000, Chris Larson wrote: > I've run into a segmentation fault in net-snmp version 5.4 on our > product, when there's a cpuIdle > or < alert set up in snmpd.conf.
No mteTrigger_run experts that had a brilliant insight into this? :) > The segfault is in snmp_oid_compare, called on line 281 of > agent/mibgroup/disman/event/mteTrigger.c, in mteTrigger_run. vp2->name > is sometimes 0, sometimes other values. > > I've seen vp2 pointing out into the ether, not within the process's > memory space. entry->count is 2 at this point, vp2_prev->next_variable > is pointing out there. vp2_prev seems fine. > > I've also seen vp2 be fine, but vp2->name with an address like "0x131". > > I assume we have some sort of memory management bug here. Occasionally, > seemingly at random, I'll also get a glibc double free warning. This is > x86_64, in one of the montavista distributions. > > If anyone more familiar with this code than I happens to know what could > cause this, I'd appreciate it greatly :) I've been perusing > mteTrigger_run looking for problems surrounding the calls to > snmp_free_varbind, but haven't spotted anything just yet. Continuing to hammer away at it, but I'm probably missing something simple.. it usually is. Switching to gdb 6.8 helped, being able to nest commands blocks is a Good Thing(tm). Thanks again, -- Chris Larson Dedicated Engineer Montavista, Inc. Work: 480-517-5031 Home: 480-963-2636 Email: clarson AT mvista DOT com Email: clarson AT kergoth DOT com ------------------------------------------------------------------------- Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://sourceforge.net/services/buy/index.php _______________________________________________ Net-snmp-coders mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
