Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: XGMII New ElectricalI Characteristics




The way I originally wrote the service primitives, the DATA_COMPLETE was
sent when any control character was received (after receiving a Start).  As
you know, this can cause a runt because of the Error code.  Treating XAUI as
an error-free interface (i.e., negligible error rate), we won't worry if
creation of some other control character causes a runt.  We took care of the
main thing which is the errors occuring on the cable.  The RS need only be
resilient to other errors.  I proposed a number of alternatives (e.g., wait
for T or I) but the committee was comfortable with the "control character
but not Error".

BTW, /E/, /S/ etc. are not used in the RS, Error, Start, etc. control
character is used as PCS independent.  Also, the RS operates on a byte
stream, with no reference to a column (only lanes).

--Bob


-----Original Message-----
From: Louis Lin [mailto:louislin@lsil.com]
Sent: Monday, November 13, 2000 3:01 PM
To: Grow, Bob
Cc: HSSG
Subject: Re: XGMII New ElectricalI Characteristics



Hi Bob,

One of the decision on the clause is to end receiving frame when there is a
non-error control data on RXD between /S/ and /T/. Can you give more
details on the background of this decision? Of course this is rare, and
something
is very wrong when it happens. Or it could be timing problems on the RXC
inputs. As long the MAC can recover to receive next packet, it should be
fine.

But ending the receiving process in this condition creates runt packets and
management won't have the right info. of the failure. If we can treat it
like
we
seeing an /E/, of course it will a different error condition. And keeping
receiving
the frame until seeing /E/ or /I/ column. Host can always drop this frame
becasue
of the error condition, and it will have more info about what is wrong.

I think that other conditions that can end a receiving frame is /T/ and /I/
column.

Thanks,

Louis Lin