On Wed, 29 Mar 2000, Ben Greear wrote:
> I think I've figured out most of the fields, but these are baffling me. Anyone
> have any insight to share?
>
> Is this like an interrupt miss or something?
> unsigned long rx_missed_errors; /* receiver missed packet */
It's the count of packets missed by the card.
Typically this is because of a lack of buffers or interrupt attention. But
any known source of packet lossage is counted.
Most chips have missed packet counters. A few only have a "I missed at
least one" flags.
> There are no comments on these...and at best I could only guess.
> /* detailed tx_errors */
> unsigned long tx_aborted_errors;
> unsigned long tx_carrier_errors;
> unsigned long tx_fifo_errors;
> unsigned long tx_heartbeat_errors;
> unsigned long tx_window_errors;
1) Aborted due to various reasons.
2) A carrier error was reported.
3) A FIFO error (design specific) was reported.
4) A transceiver heartbeat error
5) Out-of-window collision.
All of these are well defined and widely understood for Ethernet. The only
one that is Ethernet specific is (4), which means "the self-test indicates
my transceiver might be broken". (5) means that the network topology or
configuration is invalid, which has an analogy in many network types.
> Offhand, how well are these counters supported in the tulip driver?
You shouldn't need to ask this question: The source and databooks are
readily available.
The Tulip does a good job of reporting most statistics.
> Or does it also depend on the ether-card?
The major variation is the ability of the adapter to report the collision
count. Some adapters can only report that at least one collision occurred,
rather than a count. For those adapters the collision count field is
incremented rather than added to. Other adapters have bits for 0, 1, and >1
collisions, which we treat as just a 0 or 1.
Donald Becker
Scyld Computing Corporation, [EMAIL PROTECTED]
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
