The vulnerability is actually unproven, (No POC) and it's been six (6) years since it has been listed to the CVE database. I bet it has been fixed since then up to the recent version of OpenSER. The overview description will get you going:
"Buffer overflow in the fetchsms function in the SMS handling module
(libsms_getsms.c) in OpenSER 1.2.0 and earlier allows remote attackers
to cause a denial of service (crash) via a crafted SMS message,
triggering memory corruption when the "beginning" buffer is copied to
the third (pdu) argument."
Since there is no proof of concept or even a sample snippet of an actual exploit code, I can't assume the validity of this claim.
http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2006-6876
http://xforce.iss.net/xforce/xfdb/31137
http://www.securityfocus.com/bid/21800/exploit
For a POC, you might want to check exploit-db for that.
There are many types of buffer overflow, it can be heap-based or stack-based etc. The very concept of an overflow is that the attacker may send more data in a buffer (the temporary storage in memory) than it can hold. This is usually the account of programming error and not setting bound checking which can lead to the running application to crash.