Re: [EFM] flow control in 802.3 - HPNA 1.0
HPNA 1.0 did this in the past. The period of the TX_CLK was changed to
accomodate the additional physical layer time required. It is described
(albeit in somewhat sparse detail) in section 3 of the 1M8 HPNA
specification.
charlie raasch
conexant systems.
"Vladimir Oksman"
<oksman@xxxxxxxxxxxx> To: "Hugh Barrass" <hbarrass@xxxxxxxxx>
Sent by: cc: "MARC KIMPE" <marc.kimpe@adtran.com>, stds-802-3-efm@ieee.org
owner-stds-802-3-efm@majordom Subject: Re: [EFM] flow control in 802.3
o.ieee.org
11/01/01 07:27 AM
Hugh,
I also have the same feeling - an artificial collision signal is
something we can use. It doesn't seems good from the first glance to add
new signals.
Vladimir.
Hugh Barrass wrote:
> Marc,
>
> What a good question!
>
> This is a problem that we (almost certainly) have to solve in the copper
track. We are looking for suggestions, the ones that immediately come to
mind are:
>
> MII clock adjustment
> IPG stretching
> .3x pause frames
> half-duplex back pressure (force collision)
> new signal for pause/continue
>
> The matter was discussed at length during .3ae - take a look at 10G
website...
>
> http://grouper.ieee.org/groups/802/3/ae/public/jul00/muller_1_0700.pdf
>
> There are various other discussions of this on 10G. It's not clear that
the arguments and conclusions are valid for .3ah in the same way as .3ae -
but it's always worth taking a look.
>
> Hugh.
>
> PS - Any of the 10G'ers care to comment?
>
> MARC KIMPE wrote:
>
> > All,
> >
> > Is there a way to transmit flow control information across an MII
interface from the PHY direction to the MAC ? If yes could someone point me
to the pertaining clause of 802.3 ?
> >
> > In other words, let's assume that a PHY for some reason or other can
only transmit at 3Mbps while information to that PHY flows at 10Mbps. How
can the PHY tell the MAC to stop sending ? I saw some presentations related
to 10 Gigabit Ethernet that proposed various mechanisms but I do not know
if anything was adopted.
> >
> > Any help would be appreciated.
> >
> > Thanks,
> > Marc Kimpe
> > Adtran