promiscuous mode, was Re: [HECnet] Multinet Tunnel Connections to SG1::

Paul_Koning at Dell.com Paul_Koning at Dell.com
Fri Jun 8 21:07:48 PDT 2012


On Jun 8, 2012, at 3:36 PM, Johnny Billquist wrote:

On 2012-06-08 21:28, Paul_Koning at Dell.com wrote:

On Jun 8, 2012, at 3:18 PM, Johnny Billquist wrote:
...
If you don't need 16 distinct values, you simply pad the table with extra copies of any of the meaningful values; that way the result is what you want.   The easiest way to do that is to make extra copies of the entry that specifies the MAC address, but that isn't necessary, it is mentioned only because it's easy to remember.

Right.
But there are some interesting passages in the manual.
"
More than one physical address may be specified, but in Normal mode, only the first is used for receiving datagrams, and as the source address for system ID messages generated by the DELQA. In DEQNA-lock mode the specifications of multiple physical Ethernet addesses will cause the DELQA to filter all such physical Ethernet addesses for packet reception.

NOTE
Enabling more than one physical address is not recommended under normal circumstances. This may have a substantial impact on performance.
"

What do you make of that?

	Johnny

Hm.   I wonder if the LQA uses a chip that doesn't natively support multiple MAC addresses (LANCE is one such, I think).   If so, it would mean that it would have to emulate this QNA feature by setting promiscuous mode on that chip and doing the 16-entry exact filtering in firmware.

Bummer.   That would imply that QNA actually has some benefits over LQA, which I didn't expect.

I wonder if "substantial" is actually true.   As I mentioned, the multiple individual address feature was introduced to allow LAT not to be broken by DECnet startup, so any OS that starts LAT before DECnet (VMS is one such, I think) would end up with multiple MAC addresses.   And since the main reason for LQA was to make Local Area VAXclusters work right, one would assume the performance impact from this mode is not that high, since clusters are rather picky about performance issues.

	paul



More information about the Hecnet-list mailing list