[Freeipmi-devel] RMCP ping/pongs figured out, maybe a BMC bug??

[Albert Chu]

Hey AB, Bill ...

I figured the RMCP problem.  It seems the RMCP ASF Message Tag field (
IPMI spec section 12.2.3) *MUST* be changed between sending consecutive
RMCP ping messages.

According to the ASF Specification (http://www.dmtf.org/standards/asf):

"When a duplicate message is received, i.e. one with the same Message
Tag, the consumer of the message determines whether the message is
accepted or rejected."

It seems the tiger 4 BMCs choose to reject the message.  As far as I can
tell, this isn't specified in the IPMI spec.  It seems quite important.
Perhaps it should be added as an errata Bill??

As for "maybe a BMC bug", the ASF specification also says:

"A value of 255 (FFh) indicates that the associated message is not a
request-response type message."

I interpret this statement to mean that a client should always respond
to a message with a tag of 255, irregardless if it is a duplicate.  When
I set consecutive RMCP ping messages to have a message tag of 255, later
packets with a tag of 255 seem to be ignored.  To me, this sounds like
improper behavior.  Am I interpreting this statement incorrectly??

Thanks,

Al

--
Albert Chu
address@hidden
Lawrence Livermore National Laboratory