Re: [Pysnmp-dev] Potential oneliner bugs

[Mark M E. <mar...@ma...>]

Thanks for the quick fixes (and sorry for the slow response, I've  
been swamped).  I've verified the socket leak is fixed.  The changes  
for V1 GET NEXT looks good, but I haven't been able to test it  
against real hardware yet.

On Aug 15, 2007, at 6:31 AM, Ilya Etingof wrote:

>
> Both issues hopefully fixed in CVS (not with your patches). Please,  
> take a
> look.
>
> Thanks,
> ilya
>
> On Tue, 14 Aug 2007, Mark M Evans wrote:
>
>> I've run into a couple of issues that I think are bugs in the
>> oneliner implementation.  I'm still learning my way around PySNMP, so
>> apologies if my assumptions or conclusions are incorrect.
>>
>> Issue #1:  Oneliner commands "leaking" sockets.
>>
>> UdpTransportTarget.openClientMode() creates a transport via
>> UdpSocketTransport().  A side effect of using the default parameters
>> is that a new channel/socket is created and is registered in
>> asyncore.socket_map.  The socket is never removed from
>> asyncore.socket_map, it's never garbage collected and the socket is
>> never closed.  The result is that instantiating and destroying large
>> numbers of CommandGenerators will consume all of the systems socket
>> resources.
>>
>> I have a patch that works for me which is to pass a throw away
>> dictionary as the sockMap.  Example:
>>
>> pysnmp/entity/rfc3413/oneliner/cmdgen.py:
>>
>> class UdpTransportTarget:
>>      ...
>>      def openClientMode(self):
>>          self.transport = udp.UdpSocketTransport(sock=None,
>>                                                  sockMap=
>> {}).openClientMode()
>>          return self.transport
>>
>> Issue #2:  CommandGenerator().nextCmd() does not work with V1  
>> devices.
>>
>> V1 devices terminate the list list of OIDs by returning an
>> errorStatus of noSuchName(2), not by returning a value of endOfMib.
>> CommandGenerator().nextCmd() considers a non-false errorStatus as an
>> error (makes sense) and passes that failure to the caller.
>>
>> I've hacked my local version to explicitly check for an errorStatus
>> of 2, and clears that errorStatus/Index.  This seems to work, but
>> there may be a cleaner solution.
>>
>> pysnmp/entity/rfc3413/oneliner/cmdgen.py:
>>
>> class CommandGenerator(AsynCommandGenerator):
>>      ...
>>      def nextCmd(self, authData, transportTarget, *varNames):
>>          def __cbFun(
>>              sendRequestHandle, errorIndication, errorStatus,
>> errorIndex,
>>              varBindTable, (varBindHead, varBindTotalTable,  
>> appReturn)
>>              ):
>>              if errorStatus == 2:
>>                  # V1 terminates GETNEXT with an errorStatus of
>> noSuchName
>>                  # rather than a value of endOfMib (which v2c.PDUAPI
>> replaces
>>                  # with None).
>>                  errorStatus = errorStatus.clone(0)
>>                  errorIndex = errorIndex.clone(0)
>>                  pass
>>              if errorIndication or errorStatus:
>>              ...
>>
>> Hopefully this information is helpful.
>>
>> Regards,
>> Mark

--
This signature intentionally left blank.