Problem GW with Underflow

[santiago_figueroa]

Hello I have a problem with GW I get this mesage:

Oct 18 17:24:29.389: %PQUICC_ETHER-3-UNDERFLO: Unit 0, underflow error

Oct 19 19:37:01.373: %TCP-2-BADREFCNT: Tty0: Bad refcnt for packet 0x83E5FF80 during retransmit, 10.100.2.253(46158) to 180.33.186.120(1720), state 4 -Traceback= 0x801BA7C0 0x804F94D4 0x804F996C 0x804FFB98 0x80500080 0x802FEACC 0x803022D0
Oct 19 19:37:01.373: %SYS-2-BADSHARE: Bad refcount in datagram_done, ptr=8388AAEC, count=0 -Traceback= 0x801BA7C0 0x8025B4A4 0x805371AC 0x80537D7C 0x80537DB0 0x8054A188 0x804F91B8 0x804F9DA4 0x804FFB98 0x80500080 0x802FEACC 0x803022D0

Oct 19 19:37:06.426: %TCP-2-BADREFCNT: Tty0: Bad refcnt for packet 0x83E5FF80 during retransmit, 10.102.1.253(46158) to 190.34.187.130(1720), state 4 -Traceback= 0x801BA7C0 0x804F94D4 0x804F996C 0x804FFB98 0x80500080 0x802FEACC 0x803022D0
Oct 19 19:37:19.766:
TCP0: extra packet reference for pak 0x83E5FF80 found:
Oct 19 19:37:19.766: %TCP-2-BADQUEUE: Multiple entry for packet 83E5FF80 -Process= "CCH323_CT", ipl= 0, pid= 167 -Traceback= 0x801BA7C0 0x804FE880 0x80502338 0x80504698 0x805E961C 0x805EFFB8 0x816EFB38 0x816DE3B4 0x816745D8 0x816748A4 0x8167D0B4 0x81681C88 0x816B2494 0x81687E4C 0x81655310 0x8166C44C
Oct 19 19:37:19.766:     pak: UNKNOWN:46158, UNKNOWN:1720, seq 1532416589 length 138

I have a flash:c1700-k9o3sv8y7-mz.124-25d.bin

anyone can helpme?

Thanks

[Calin C.]

Hello,

Here you have it:

1. %PQUICC_ETHER-3-UNDERFLO: Unit [dec], underflow error During transmission of a frame, the local buffer of the controller  chip received  insufficient data because the data could not be transferred to the chip  fast enough to keep pace with  its output rate. Normally, such a problem is temporary, depending on  transient peak loads within the  system.

Recommended Action: The system should recover. No action is required. If the problem  recurs, it  indicates a hardware error that might be related to data traffic  patterns. If this is the case, copy the  error message exactly as it appears on the console or in the system log,  contact your Cisco technical  support representative, and provide the representative with the gathered  information.

Related documents- No specific documents apply to this  error message.

Maybe you can check the load of your system?

Cheers,

Calin

[Timothy Stewart]

Hello,

As the other poster already noted, the UNDERFLO message indicates a condition where the ethernet controller's transmit FIFO went empty while it was trying to transmit the frame.  Generically speaking, most drivers / chipset will use the following process when transmitting:

- The packet is stored in a buffer and the pointer for this packet buffer is passed into the Tx ring of the egress interface.

- The chipset will see that a packet is ready for transmit so it DMA's the packet from the TX ring into it's FIFO queue.

- As soon as the part of the packet is in the FIFO the chipset begins the transmit (while the DMA is still going on)

- If the chipset runs into a condition where the transmit is happening faster than the DMA it will display the UNDERFLO message and that single packet will be aborted.

The condition can occur under a number of scenerio's.  On the 1700 platform, certain show commands or other conditions will lock the system bus which can interfer with the DMA process which in turn may cause this message to occur.  Since you have only seen it a few times, it is safe to ignore.

The other messages in your log are unrelated to the UNDERFLO message.  They are more serious since they are indicating a situtation where the IOS dropped a TCP packet because the internal IOS data structures were either corrupted or the packet had previously been sent but the data structures where not cleaned up properly.  In this case, the packet in the buffer was a TCP packet which was being resent due to a timeout and it looks like some other process may have already dealt with that packet (i.e. two processes tried to send the same packet).  I would suggest you open a TAC case to have this situation looked into.  The TAC will need a show tec and a show log from your router.

Thanks

Tim