Catalyst 2960G optical port autonegotiation

matija.pecotic · ‎04-26-2013

Hello,

I started to work with catalyst 2960G and I'm observing strange issue that 1000BASE-X autonegotiation is turned on at the port, partner has autonegotiation turned off, and link is still up.

I have played a bit with

show controllers ethernet-controller g0/46 phy detail

and here is what I belive is happening:

- you are using marvell 88E1112 on this port

- Marvell 88E1112 has "Serial Interface Auto-negotitation Bypass Mode" option which is enabled in register 26, bit 15, page 1. This option is enabled on this port:

Page 1 Registers:

.

001A: 8083 PHY Specific Control Register2 : 1000 0000 1000 0011

- If the link came up due to autonegotiation bypassed can be monitored in register 17, page 1, bit 5

- When I set the partner in autonegotiation enabled mode, link comes up and this bit is cleared.

0011: 3CC8 PHY Specific Status Register : 0011 1100 1100 1000

- When I set partner to autonegotation off, this bit is active and and link is up (as mentioned before)

0011: 24E8 PHY Specific Status Register : 0010 0100 1110 1000

My questions is, whether this analysis holds, and, how to disable this, is there some configuration option, etc. I dont mind writing to phy directly. In the end of this post you'll find info on the device used. If you need some more input, please say.

Additional info - I'm always plugging in and out cable, it isnt the case link survives turning off autonegotiation on the fact it was once negotiated with.

Partner has autonegotiation turned off in this case all the time.

Thanks and regards,

Matija

Interface configuration:

Switch#show interfaces GigabitEthernet0/46

GigabitEthernet0/46 is up, line protocol is up (connected)

Hardware is Gigabit Ethernet, address is 0019.e7fb.d3ae (bia 0019.e7fb.d3ae)

MTU 1500 bytes, BW 1000000 Kbit, DLY 10 usec,

reliability 255/255, txload 1/255, rxload 1/255

Encapsulation ARPA, loopback not set

Keepalive not set

Full-duplex, 1000Mb/s, link type is auto, media type is unsupported

input flow-control is off, output flow-control is unsupported

ARP type: ARPA, ARP Timeout 04:00:00

Last input never, output 00:00:00, output hang never

Last clearing of "show interface" counters never

Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0

Queueing strategy: fifo

Output queue: 0/40 (size/max)

5 minute input rate 0 bits/sec, 0 packets/sec

5 minute output rate 1000 bits/sec, 2 packets/sec

31 packets input, 2510 bytes, 0 no buffer

Received 19 broadcasts (0 multicast)

0 runts, 0 giants, 0 throttles

0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored

0 watchdog, 6 multicast, 0 pause input

0 input packets with dribble condition detected

5372 packets output, 485419 bytes, 0 underruns

0 output errors, 0 collisions, 1 interface resets

0 babbles, 0 late collision, 0 deferred

0 lost carrier, 0 no carrier, 0 PAUSE output

0 output buffer failures, 0 output buffers swapped out

Switch#show version

Cisco IOS Software, C2960 Software (C2960-LANBASE-M), Version 12.2(25)SEE2, RELEASE SOFTWARE (fc1)

Compiled Fri 28-Jul-06 04:33 by yenanh

Image text-base: 0x00003000, data-base: 0x00AA2F34

ROM: Bootstrap program is C2960 boot loader

BOOTLDR: C2960 Boot Loader (C2960-HBOOT-M) Version 12.2(25r)SEE1, RELEASE SOFTWARE (fc1)

Switch uptime is 1 hour, 3 minutes

System returned to ROM by power-on

System image file is "flash:c2960-lanbase-mz.122-25.SEE2/c2960-lanbase-mz.122-25.SEE2.bin"

cisco WS-C2960G-48TC-L (PowerPC405) processor (revision A0) with 61440K/4088K bytes of memory.

Processor board ID FOC1043X0E2

Last reset from power-on

1 Virtual Ethernet interface

48 Gigabit Ethernet interfaces

The password-recovery mechanism is enabled.

64K bytes of flash-simulated non-volatile configuration memory.

Base ethernet MAC Address : 00:19:E7:FB:D3:80

Motherboard assembly number : 73-10300-06

Power supply part number : 341-0098-02

Motherboard serial number : FOC10421SEW

Power supply serial number : DCA104191Y3

Model revision number : A0

Motherboard revision number : B0

Model number : WS-C2960G-48TC-L

System serial number : FOC1043X0E2

Top Assembly Part Number : 800-27071-01

Top Assembly Revision Number : C0

Version ID : V01

CLEI Code Number : COM4A10BRA

Hardware Board Revision Number : 0x01

Switch Ports Model SW Version SW Image

------ ----- ----- ---------- ----------

* 1 48 WS-C2960G-48TC-L 12.2(25)SEE2 C2960-LANBASE-M

Configuration register is 0xF

Message was edited by: Matija Glavinic Pecotic Reason: Additional clarifications

Gregory Snipes · ‎04-26-2013

Is the partner switch also a WS-C2960G-48TC-L? Could you post up the interface configuration from each side?

matija.pecotic · ‎04-26-2013

Hi Gregory,

partner is embedded system I'm working on. Is output of ethtool sufficient? I can give phy register dumps.. But anyway, this piece of equipment works fine in 1000BASE-X mode, I would say Cisco bends the standard here. I will post 2690 configuration on Monday.

Regards,

Matija

Gregory Snipes · ‎04-26-2013

This issue is complicated by the fact the the WS-C2960G-48TC-L's ports have a dual personality. They can be set to use SFP or RJ45. In RJ-45 mode they can be hard coded or auto negotiate, in SFP mode they need to be hard set. It is possible that the controller flag is referring to a mode of operation that you are not using.

matija.pecotic · ‎04-26-2013

I dont think they need to be hard set in SFP mode. SFP mode refers to 1000BASE-X mode of operation, this one can be set either to fixed speed or autonegotiation. The only important thing is that both partners agree. In 1000BASE-T however(lets say we are only interested in gigabit), we have to enable autonegotation, it is mandated by the standard. BTW, I set interface media-type to sfp, just to be on the safe side, I dont like autodetections.

Regarding the bit I'm referring to, this bit is specifically aimed to bypass SGMII or 1000BASE-X autonegotiation. Also, this register resides on page 1 where fiber configuration takes place.

Story is like this: if partner has autonegotiation enabled, it expects /C/ (configuration) symbols on the line which hold the remote partner capabilities ( in fiber, this is duplex, pause and remote fault usually). But if the remote partner has aneg disabled, he will send /I/ ( idle) symbols and therefore aneg will never complete. But, this phy has this option which after certain timeout will notice that partner is sending only /I/dles and will link up. And this option is enabled after phy reset. IMHO, this option isnt good, it just brings confusion why stuff works and it shouldnt.

Gregory Snipes · ‎04-26-2013

I understand what you are saying, and while the SFPs do negotiate, they only do so for flow-control parameters and remote fault information. The speed and duplex are fixed by the sfp type. If you wish to fully disable negotiation on an SFP fiber link you can issue the "speed nonegotiate" to stop all negotiation on the interface or conversly "no speed nonegotiate" enable it.

matija.pecotic · ‎04-26-2013

just a little addendum, standard defines both half and full duplex, it is vendor's decision not to implement certain option.

Regarding the disablement of the autonegotiation, thanks for the hint, but that is not the behaviour I want. What I would like to see is a standard defined behavior, and that is link down if partner has autonegotiation turned off.

This is also stated on cisco web pages, here is the compatibility table and source below

Table 6—Gigabit Autonegotiation Configuration Table

Autonegotiation Setting	NIC Gigabit Autonegotiation Setting	Switch Port Gigabit Link	Alternate Switch Link/NIC Link
Enabled	Enabled	Up	Up
Disabled	Disabled	Up	Up
Enabled	Disabled	Down	Up
Disabled	Enabled	Up	Down

http://www.cisco.com/en/US/products/hw/switches/ps708/products_tech_note09186a00800a7af0.shtml

Gregory Snipes · ‎04-26-2013

I am not fully understanding what you are looking for. You should either have the negotiation off on both sides on on on both sides. I have always observed Cisco switches behave exactly as describe in the table above and am not sure of the issue.

matija.pecotic · ‎04-26-2013

that is the issue. Switch not behaving according to the table given above. I have case 3, aneg on on the switch, and aneg off at the partner. And switch declares link is up and traffic passes through. Note I'm referring to switch, not the partner, partner also states link is up, but that is fine in this case. Switch is not.

And ofcourse I should have same settings on both sides, if I was asked, I would ban manual settings for good. But also equipment should behave accordingly regardless of my 'missconfiguration'.

I hope I have mentioned I was just testing something when I came up to this, I amnt trying to deploy anything.

matija.pecotic · ‎04-29-2013

I've added switch configuration to the first post.

Do you think I should also post this question to optical networking on these discussion boards? Or maybe some other list? I'm perfeclty aware people here do not concern usually with such low level stuff

matija.pecotic · ‎05-01-2013

Discussion moved to optical networking as it may seem to be a better place for it. Hope this is fine!