Exception EasySNMPUnknownObjectIDError raised for an OID that exists

[nnathan]

I haven't investigated the code yet, but I found the following bug in easysnmp.

I have installed the stock SNMP agent with the default configuration.

I fetch a single OID using snmpget command:

$ snmpget -On -v 2c -c redacted localhost 1.3.6.1.4.1.2021.11.53.0
.1.3.6.1.4.1.2021.11.53.0 = Counter32: 444734235

Now, when I use easysnmp with the following script:

import easysnmp

oids = ['1.3.6.1.4.1.2021.11.53.0']

print easysnmp.easy.snmp_get(
        oids,
        hostname='localhost',
        version=2,
        community='redacted'
)

I get the following exception:

Traceback (most recent call last):
  File "test_easysnmp.py", line 9, in <module>
    community='redacted'
  File "/srv/virtualenvs/bpl-poll-consumer/local/lib/python2.7/site-packages/easysnmp/easy.py", line 20, in snmp_get
    return session.get(oids)
  File "/srv/virtualenvs/bpl-poll-consumer/local/lib/python2.7/site-packages/easysnmp/session.py", line 299, in get
    interface.get(self, varlist)
easysnmp.exceptions.EasySNMPUnknownObjectIDError: unknown object id (1.3.6.1.4.1.2021.11.53.0)

So the easysnmp.interface is raising the exception except the OID itself does exist as demonstrated by the above snmpget command.

[nnathan]

I have found the offending code and I'm working on a patch.

First, this line is a problem since py_netsnmp_attr_string(varbind, "oid_index", &iid, NULL) < 0 returns -1 since the attribute exists, is a None type, but not a string. This is because no index exists.

So I patched in the following function and fixed up the lines as follows:

/* returns 1 if None object, 0 if not None, and -1 if error */
static int py_netsnmp_attr_is_none_object(PyObject *obj, char *attr_name)
{
    int retval = -1;

    if (obj && attr_name)
    {
        PyObject *attr = PyObject_GetAttrString(obj, attr_name);

        if (attr)
        {
            /* 1 if Py_None or 0 otherwise */
            retval = (attr == Py_None);

            Py_DECREF(attr);
        }
    }

    return retval;
}
/* [...] */

            while (varlist_iter && (varbind = PyIter_Next(varlist_iter)))
            {
                if (py_netsnmp_attr_string(varbind, "oid", &tag, NULL) < 0 ||
                    (!py_netsnmp_attr_is_none_object(varbind, "oid_index") &&
                     py_netsnmp_attr_string(varbind, "oid_index", &iid, NULL) < 0)
                   )
/* [...] */

However now I'm getting a segfault here at this line:

    if (iid && *iid && oid_arr_len)
    {
        __concat_oid_str(oid_arr, oid_arr_len, iid);
    }
    return rtp;

Presumably, iid isn't set to NULL. I explicitly set it to NULL in the netsnmp_get function, and now I get the following output:

$ python test_easysnmp.py
([<SNMPVariable value=None (oid='1.3.6.1.4.1.2021.11.53.0', oid_index=None, snmp_type=None)>], True)
[<SNMPVariable value='444734235' (oid='iso.3.6.1.4.1.2021.11.53.0', oid_index='', snmp_type='COUNTER')>]

I'll work on a thorough patch that validates this. I'm surprised this isn't picked up by the unit tests.

[nnathan]

Actually, just looking at the successful output from my last response, it appears that the oid_index=''. Perhaps oid_index should always be set to an empty string by the easysnmp library. I'm not sure how that will affect the code.

Let me know your insight.

Thanks.

[bwks]

@nnathan Thought I would mention that if you add the leading '.' to the oid the exception is not raised.

oids = ['1.3.6.1.4.1.2021.11.53.0']
EasySNMPUnknownObjectIDError: unknown object id (1.3.6.1.4.1.2021.11.53.0)

oids = ['.1.3.6.1.4.1.2021.11.53.0']
EasySNMPTimeoutError: timed out while connecting to remote host (I dont have snmp open on localhost)