Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Flow control

0 views
Skip to first unread message

Christian Hennecke

unread,
Apr 15, 2002, 3:24:32 PM4/15/02
to
Hi all!

Can somebody please explain the following?

My NICs' configuration program displays data regarding a feature called
"flow control" and allows for four different settings:

TX disabled, RX disabled
TX enabled, RX disabled
TX disabled, RX enabled
TX enabled, RX enabled

So the question is: What does it do and which setting is the best?
--
Christian Hennecke
The OS/2 Files - Die OS/2 Akten: http://www.os2world.com/os2files/

William L. Hartzell

unread,
Apr 15, 2002, 7:11:29 PM4/15/02
to
Sir: Christian Hennecke wrote: > Hi all! > Can somebody please explain the following? > My NICs' configuration program displays data regarding a feature called > "flow control" and allows for four different settings: > TX disabled, RX disabled > TX enabled, RX disabled > TX disabled, RX enabled > TX enabled, RX enabled > So the question is: What does it do and which setting is the best? This sounds like the flow control (aka hardware handshaking) between a serial port and a modem. If it is, then it should be enabled upon both Tx and Rx. If this is between two NICs, it should be disabled upon both TX and Rx to avoid problems. I don't have enough information to tell you what to do. Bill <Be Kind to the Taxman on His Day>

Veit Kannegieser

unread,
Apr 16, 2002, 4:12:12 AM4/16/02
to
Christian Hennecke wrote:

> TX disabled, RX disabled
> TX enabled, RX disabled
> TX disabled, RX enabled
> TX enabled, RX enabled
>
> So the question is: What does it do and which setting is the best?

It is used in full duplex modes only.
Chip documentation says

"9.7 Flow Control
The RTL8139C(L) supports IEEE802.3X flow control to improve
performance in full︸uplex mode.
It detects PAUSE packet to achieve flow control task.
9.7.1. Control Frame Transmission
When RTL8139C(L) detects its free receive buffer less than 3K bytes,
it sends a PAUSE
packet with pause_time(=FFFFh) to inform the source station to stop
transmission for the
specified period of time. After the driver has processed the packets
in the receive buffer and
updated the boundary pointer, the RTL8139C(L) sends the other PAUSE
packet with
pause_time(=0000h) to wake up the source station to restart
transmission.
9.7.2. Control Frame Reception
RTL8139C(L) enters backoff state for the specified period of time when
it receives a valid
PAUSE packet with pause_time(=n). If the PAUSE packet is received
while RTL8139C(L)
is transmitting, RTL8139C(L) starts to backoff after current
transmission completes.
RTL8139C(L) frees to transmit next packets again when it receives a
valid PAUSE packet
with pause_time(=0000h) or the backoff timer(=n*512 bit time) elapses.

Note: The PAUSE operation cannot be used to inhibit transmission of
MAC Control frames (e.g.
PAUSE packet). The N﹁ay flow control capability can be disabled,
please refer to Section 6.
EEPROM (93C46 or 93C56) Contents for detailed description. "

Veit Kannegieser

William L. Hartzell

unread,
Apr 16, 2002, 5:42:10 AM4/16/02
to
Sir:

So the question really becomes, does the NIC driver on both/all ends
support this protocol? If no, then disable, else enable. This is
uncommon on the NIC drivers that I know (this being the first). Thanks
for the heads up.
--
Bill
<Here comes May Day>

Christian Hennecke

unread,
Apr 18, 2002, 8:19:16 PM4/18/02
to
On Tue, 16 Apr 2002 09:42:10 UTC, "William L. Hartzell"
<wlhar...@attbi.com> wrote:

> So the question really becomes, does the NIC driver on both/all ends
> support this protocol? If no, then disable, else enable. This is
> uncommon on the NIC drivers that I know (this being the first). Thanks
> for the heads up.

Judging from my experiences over the last days I can only say "disable
it". I tried all combinations and unless both TX and RX were disabled,
my second machine "lost the connection" for TCP/IP after a while and
NetBIOS suffered from a severe performance penalty. This is true at
least for using the RealTek NIC with a 3COM NIC. Maybe it would work if
both NICs had RealTek chips.

0 new messages