Hello All , I am having a dickens of a time with auto negotiation
between a netgear fsm726s & a TrendNet TEG-PCISX+ . Fyi , Which
I bought just for this card .
The switch reports negotiating ...
Port Name Link On/Off State Rate/Duplex Flow Ctrl
26GB Not Defined Up On Forwarding (10 Full) (Disabled)
ns83820 reports , eth0: link now 1000F mbps, full duplex and up.
The thruput to a 10/100 device is dismal at best . Which
lead me to the idea that something was wrong with negotiation .
See (*) below .
I have attempted to set the switch port speed/duplex/flow manually
then I am unable to ping thru the ns83820 card .
I am quite aware that this could well be a difficulty in the
switch still . So I am looking for pointers on where to look ?
I already tried the netgear suport site ;-} . That is why I am
running the lastest code for the switch (1.0.4) .
Tia , JimL
# uname -a
Linux filesrv1 2.4.21-pre3 #1 SMP Sat Mar 8 13:44:59 EST 2003 i686 unknown
# dmesg | grep eth0
ns83820.c: National Semiconductor DP83820 10/100/1000 driver.
eth0: ns83820.c: 0x22c: f022100b, subsystem: 100b:f022
eth0: detected 64 bit PCI data bus.
eth0: enabling optical transceiver
eth0: ns83820 v0.20: DP83820 v1.3: 00:40:f4:66:df:ed io=0xfeafe000 irq=20 f=sg
...
eth0: link now 1000F mbps, full duplex and up.
eth0: no IPv6 routers present
cut-n-paste from serial console of ...
FSM726S Managed Stackable Switch
Unit 1 Set-up > Port Configuration
Port Name Link On/Off State Rate/Duplex Flow Ctrl
...
2 Not Defined Up On Forwarding (100 Full) (Enabled )
...
26GB Not Defined Up On Forwarding (10 Full) (Disabled)
(*)
# tftp filesrv1
tftp> mode binary
tftp> get getme
Received 495321088 bytes in 334.0 seconds
tftp> quit
# bc
This is free software with ABSOLUTELY NO WARRANTY.
For details type `warranty'.
scale=10
495321088/334
1482997.2694610778 < Bytes/sec
last*8
11863978.1556886224 < ~ bits/sec
quit
#
--
+------------------------------------------------------------------+
| James W. Laferriere | System Techniques | Give me VMS |
| Network Engineer | P.O. Box 854 | Give me Linux |
| [email protected] | Coudersport PA 16915 | only on AXP |
+------------------------------------------------------------------+
On Wed, Mar 12, 2003 at 09:11:02AM -0500, Mr. James W. Laferriere wrote:
> The switch reports negotiating ...
> Port Name Link On/Off State Rate/Duplex Flow Ctrl
> 26GB Not Defined Up On Forwarding (10 Full) (Disabled)
>
> ns83820 reports , eth0: link now 1000F mbps, full duplex and up.
...
> I am quite aware that this could well be a difficulty in the
> switch still . So I am looking for pointers on where to look ?
> I already tried the netgear suport site ;-} . That is why I am
> running the lastest code for the switch (1.0.4) .
It's entirely possible the card has a different polarity for the phy bits
as compared to the fibre card (Netgear) that the driver is already tested
on. Also, I've only managed to test on a cisco switch -- is there any
other hardware you can test against (ie using the fibre cable for cross
over) to narrow things down? Enabling debug and dumping the status bits
might hint as to what has to be changed. Thankfully Trendnet seems to
have programmed the subsystem id, so I'll be able to include the change
automatically.
-ben
Hello Ben ,
On Wed, 12 Mar 2003, Benjamin LaHaise wrote:
> On Wed, Mar 12, 2003 at 09:11:02AM -0500, Mr. James W. Laferriere wrote:
> > The switch reports negotiating ...
> > Port Name Link On/Off State Rate/Duplex Flow Ctrl
> > 26GB Not Defined Up On Forwarding (10 Full) (Disabled)
> >
> > ns83820 reports , eth0: link now 1000F mbps, full duplex and up.
> ...
> > I am quite aware that this could well be a difficulty in the
> > switch still . So I am looking for pointers on where to look ?
> > I already tried the netgear suport site ;-} . That is why I am
> > running the lastest code for the switch (1.0.4) .
> It's entirely possible the card has a different polarity for the phy bits
> as compared to the fibre card (Netgear) that the driver is already tested
> on. Also, I've only managed to test on a cisco switch -- is there any
> other hardware you can test against (ie using the fibre cable for cross
> over) to narrow things down? Enabling debug and dumping the status bits
> might hint as to what has to be changed. Thankfully Trendnet seems to
> have programmed the subsystem id, so I'll be able to include the change
> automatically.
I have access to another one of those cards at work it reports
the same items in the dmesg , I just looked at the switch it is
attached to & it shows fine(*) . That switch manufacturer is an
old Cabletron SSR4000 . But the thruput there is even worse (**).
THO the ssr4k network isn't idle . And I am unable to aportion a
time to make it idle . The previous emails test network was idle .
After I get back from a road trip next week I'll try a back to
back test . At least I hope I can when I get back . Twyl , JimL
(*)
# port show port-status gi.4.2
Flags: M - Mirroring enabled S - SmartTRUNK port
Link Admin
Port Port Type Duplex Speed Negotiation State State
Flags
---- --------- ------ ----- ----------- ----- ----- -----
gi.4.2 1 Gigabit Ethernet Full 1 Gbits Auto Up Up
(**)
$ tftp hostname
tftp> mode binary
tftp> get getme
Received 675315712 bytes in 1169.8 seconds
tftp> quit
$ bc
scale=10
675315712/1169.8
577291.5985638570 << bytes/sec
last*8
4618332.7885108560 << bits/sec
quit
--
+------------------------------------------------------------------+
| James W. Laferriere | System Techniques | Give me VMS |
| Network Engineer | P.O. Box 854 | Give me Linux |
| [email protected] | Coudersport PA 16915 | only on AXP |
+------------------------------------------------------------------+