Re: multiple connections

[julian_satran@il.ibm.com]

My conceptual model was more like a chained list that directs the setup on
all NICS
and enforces sequence. Please note that several such chains can be
"executed" in parallel
as the sequence is important only whithin a list. Obviously the rgular
caution applies
as you have several "hands" on your NIC cards. And the chain excution means
only setup.

Julo

"Matt Wakeley" <matt_wakeley@agilent.com> on 07/09/2000 00:40:28

Please respond to Matt Wakeley <matt_wakeley@agilent.com>

To:   ips@ece.cmu.edu
cc:    (bcc: Julian Satran/Haifa/IBM)
Subject:  Re: multiple connections




julian_satran@il.ibm.com wrote:

> Command numbering won't be needed anymore - every command with an
> outstanding status is restarted.

I suppose that could be done, as long as the "retry" bit was still there,
so
there was no chance of a command being re-issued while the status is in
flight
back to the initiator...

> As for the 2 commands - synch can be achieved by setting data first and
then
> sending commands.

My point was that if there are two NICs using some kind of "mailbox"
interface
to the system (like I2O), the host will have to put the data command
describing
the scatter/gather list in the data NIC's mailbox, and wait for some kind
of
notification that the NIC has pulled the message from the mailbox (can you
say
extra unnecessary interrupt?) before it can post the command to the control
NIC's mailbox. If this synchronization was not done, there would be a small
possibility that the command could be sent to the target and the target
respond
with the data before the data NIC has set itself up to receive and properly
place the data.

>
>
> Julo

-Matt