gosnmp sets SnmpPDU.Value to nil when type is Bitstring about snmp_exporter HOT 4 CLOSED

Yes, I think this may need to be fixed upstream first.

from snmp_exporter.

I've done some more learning and think this may be more of a bug with Switch B. BITS "types" (not the ANS.1 Bitstring type, but rather an SNMP "construct", https://datatracker.ietf.org/doc/html/rfc1906#section-8 ) are meant to be encoded and reported as the ASN.1 Octet String type. Switch B seems to report the encoding type as 0x03, Bitstring, but then proceed to encode it as an Octet String. Somehow snmpget figures this out though, somewhere in here probably https://github.com/net-snmp/net-snmp/blob/master/snmplib/mib.c#L1809 . It is probably desirable for gosnmp to behave similarly, handling Bitstring as Octet String.

from snmp_exporter.

I'm not great with C, but it looks like this is where ASN Bitstring is parsed: https://github.com/net-snmp/net-snmp/blob/90a70ebd82ae1b545e7a58c1b3eb10981cb85adc/snmplib/asn1.c#L1824 . Doing some tracing on Switch B's output through this makes it seem like that Bitstring parsing code isn't that careful and essentially treats Bitstrings as Octet Strings.

Tracing that function also explains why Switch B reports Bitstring: 03 in the debug output when the actual BITS data is 3E. It is (unintentionally) just reporting 0x03, the Bitstring type code.

from snmp_exporter.

It seems the ASN.1 Bitstring type has no place, officially, in SNMP. Encoding a BITS construct as anything other than an Octet String is incorrect. It just happens to be that netsnmp (kind of) handles it. I'll close this and the other issue as I'm reasonably confident this is simply incorrect behavior from the switch.

from snmp_exporter.