Document Audience: | INTERNAL |
Document ID: | I1138-1 |
Title: | The QFE network interface is reporting excessive input packet errors when running back to back stress tests. |
Copyright Notice: | Copyright © 2005 Sun Microsystems, Inc. All Rights Reserved |
Update Date: | 2004-11-05 |
------------------------------------------------------------
- Sun Proprietary/Confidential: Internal Use Only -
------------------------------------------------------------------------
*** Sun Confidential: Internal Use and Authorized VARs Only ***
________________________________________________________________________
This message including any attachments is confidential information
of Sun Microsystems, Inc. Disclosure, copying or distribution is
prohibited without permission of Sun. If you are not the intended
recipient, please reply to the sender and then delete this message.
________________________________________________________________________
FIELD INFORMATION NOTICE
(For Authorized Distribution by Sun Service)
FIN #: I1138-1
Synopsis: The QFE network interface is reporting excessive input packet errors when running back to back stress tests.Create Date: Nov/05/04
SunAlert: No
Top FIN/FCO Report: No
Products Reference: Sun Fire /4800/4810/6800E4900/E6900 Servers
Product Category: Server / Diag-Doc-Service
Product Affected:
Systems Affected:
-----------------
Mkt_ID Platform Model Description Serial Number
------ -------- ----- ----------- -------------
- E4900 ALL Sun Fire 4900 -
- E6900 ALL Sun Fire 6900 -
- S12 ALL Sun Fire 4800 -
- S12i ALL Sun Fire 4810 -
- S24 ALL Sun Fire 6800 -
X-Options Affected:
-------------------
Mkt_ID Platform Model Description Serial Number
------ -------- ----- ----------- -------------
X1034A - - Quad Fast Ethernet (QFE) PCI Adapter -
Parts Affected:
----------------------
Part Number Description Model
----------- ----------- -----
501-5406-01 Sun Quad FastEthernet (QFE/P) -
501-5406-06 Sun Quad FastEthernet (QFE/P) -
501-5406-07 Sun Quad FastEthernet (QFE/P) -
References:
BugId: 5041625 - QFE is showing too many ierror pkts running back to
back stress.
Issue Description:
While running back to back configuration stress tests using TTCP on the
PCI+ IO Board, the QFE interface reports excessive input packet errors
"ierrors". This has been determined to be a QFE issue, which possibly
accentuates in operations utilizing the XMITS ASIC since XMITS has higher
throughput.
This issue needs to be documented for PCI+ release, since customers who
upgrade their PCI IO assemblies to PCI+ IO assemblies, may still have
QFE NICs in their existing configuration.
Running back to back tests using the TTCP TCP/UDP benchmarking tool the
user will observe that "netstat -i" will report many more Ierrs for the
QFE interfaces than any of the other configured network interfaces being
tested. Testing only the QFE interface and not the other interfaces
yielded similar results.
# netstat -i
Name Mtu Net/Dest Address Ipkts Ierrs Opkts Oerrs Collis Queue
ba0 9180 atmb-0 atmb-0 502593 0 472294 2 0 0
lo0 8232 localhost localhost 299 0 299 0 0 0
ce0 1500 ceb-0 ceb-0 1194862 0 1062732 0 0 0
ce1 1500 ceb-1 ceb-1 1062132 0 983099 0 0 0
ce2 1500 kuhb-2 kuhb-2 305675 0 582944 0 0 0
ce3 1500 qgeb-3 qgeb-3 1154890 0 1001651 0 0 0
ce5 1500 qgeb-5 qgeb-5 1215951 0 1057197 0 0 0
ce6 1500 qgeb-6 qgeb-6 1238271 0 1042729 0 0 0
ge0 1500 gigb-0 gigb-0 448942 0 452411 0 0 0
hme1 1500 qads4-b qads4-b 3704 0 1588 0 0 0
qfe0 1500 qfeb-0 qfeb-0 1034566 2390 1104153 0 0 0
qfe1 1500 qfeb-1 qfeb-1 934923 2397 1061761 0 0 0
qfe2 1500 qfeb-2 qfeb-2 1200896 1814 1154268 0 0 0
qfe3 1500 qfeb-3 qfeb-3 679806 2383 900015 0 0 0
Affected parts can be examined through either a recent explorer output
as in the following example:
# /explorer../sysconfig/prtdiag-v.out
Or run prtdiag on the system: (abbreviated output)
# prtdiag -v
System Configuration: Sun Microsystems sun4u Sun Fire 6800
System clock frequency: 150 MHz
Memory size: 24576 Megabytes
.....
========================= IO Cards =========================
Bus Max
IO Port Bus Freq Bus Dev,
FRU Name Type ID Side Slot MHz Freq Func State Name Model
---------- ---- ---- ---- ---- ---- ---- ---- ----- ---------------- ---------
/N0/IB9/P0 PCI 30 B 0 33 33 1,0 ok pci-pci1011,24.3/ pci-bridge
pci108e,1000
/N0/IB9/P0 PCI 30 B 1 33 33 2,0 ok pci-pci1011,25.4/ pci-bridge
pci108e,1000
/N0/IB9/P0 PCI 30 B 1 33 33 0,0 ok pci108e,1000-pci108e,1000.1
/N0/IB9/P0 PCI 30 B 1 33 33 0,1 ok SUNW,qfe-pci108e,1001
SUNW,pci-qfe
/N0/IB9/P0 PCI 30 B 1 33 33 1,0 ok pci108e,1000-pci108e,1000.1
/N0/IB9/P0 PCI 30 B 1 33 33 1,1 ok SUNW,qfe-pci108e,1001
SUNW,pci-qfe
/N0/IB9/P0 PCI 30 B 1 33 33 2,0 ok pci108e,1000-pci108e,1000.1
/N0/IB9/P0 PCI 30 B 1 33 33 2,1 ok UNW,qfe-pci108e,1001
SUNW,pci-qfe
/N0/IB9/P0 PCI 30 B 1 33 33 3,0 ok pci108e,1000-pci108e,1000.1
/N0/IB9/P0 PCI 30 B 1 33 33 3,1 ok SUNW,qfe-pci108e,1001
SUNW,pci-qfe
========================= Active Boards for Domain ===========================
Board Receptacle Occupant
FRU Name Type Status Status Condition Info
--------- ----------- ---------- ---------- --------- ------------------
/N0/SB3 CPU_V3 connected configured ok powered-on, assigned
/N0/SB4 CPU_V3 connected configured ok powered-on, assigned
/N0/SB5 CPU_V3 connected configured ok powered-on, assigned
/N0/IB7 PCI+_I/O_Bo connected configured ok powered-on, assigned
/N0/IB9 PCI+_I/O_Bo connected configured ok powered-on, assigned
========================= Available Boards/Slots for Domain ===================
......
========================= HW Revisions =======================================
System PROM revisions:
----------------------
OBP 5.17.0 02/14/04 13:09
IO ASIC revisions:
------------------
Port
FRU Name Model ID Status Version
----------- --------------- ---- ------ -------
/N0/IB7/P0 SUNW,xmits 26 ok 3
/N0/IB7/P1 SUNW,xmits 27 ok 3
/N0/IB9/P0 SUNW,xmits 30 ok 3
/N0/IB9/P1 SUNW,xmits 31 ok 3
/N0/IB7/P0 SUNW,sgsbbc 26 ok 2
/N0/IB9/P0 SUNW,sgsbbc 30 ok 2
"ierrors" can happen for various reasons (parity error, alignment
error, fifo overflow, etc.). The reason for ierrors in this case is
because the QFE hardware card has a 2 K byte receive buffer and the
receive (Rx) DMA engine could not get free Rx descriptors. As a result
the Rx DMA engine drops the packet. In normal TCP network operations
the packets will be retransmitted and NO data will be lost.
This is not the issue in the QFE driver - this is a limitation of the
QFE hardware. Starting with Solaris 9 Update 5 and continuing with
Solaris 10, kernel/stream performance enhancements now allow the
network stress tests (TTCP, etc.) to generate higher packets rates than
Solaris 8 and earlier versions of Solaris 9.
The QFE/P network interface card is a EOL'ed NIC in E4900, E6900,
F4810, F4800 and F6800 midrange servers. There are no plans to modify
the existing hardware. If this issue is seen at a customer site after
upgrading to PCI+, the only option for the customer is to replace the
QFE IO card with one of the NICs currently available for these systems.
Implementation:
---
| | MANDATORY (Fully Proactive)
---
---
| | CONTROLLED PROACTIVE (per Sun Geo Plan)
---
---
| X | REACTIVE (As Required)
---
Corrective Action:
The following recommendation is provided as a guideline for authorized
Sun Services Field Representatives who may encounter the above
mentioned issue.
This is a hardware limitation of the QFE/P interface card. Potential
customers should be aware of this issue and of the possibility that by
continuing to use a EOL'ed NIC that they are at risk of this issue
occurring because of the higher throughput possible with the new PCI+
IO assembly.
Customers should take this opportunity to upgrade the QFE interfaces to
one of the currently available and fully supported network interface
cards.
Comments:
If a customer is considering upgrading their existing Sun Fire 6800 and
4800/4810 servers to the new PCI+ IO assembly they will be aware of this
issue. If there is a chance that the customer may run into this
problem the field should suggest that the customer take this
opportunity to upgrade to a supported NIC.
============================================================================
NOTE: FIN Tracking Instructions for Radiance/SPWeb:
--------------------------------------------------
If a Radiance case involves the application of a FIN to solve a customer
issue, please complete the following steps in Radiance/SPWeb prior to
closing the case:
o Select "Field Information Notice" in the REFERENCE TYPE field.
o Enter FIN ID number in the REFERENCE ID field.
For example; I1111-1.
If possible, include additional details in the REFERENCE SUMMARY field
(ie. implementation complete, customer declined, etc.)
--------------------------------------------------------------------------
Implementation Notes:
--------------------
In case of "Mandatory" FINs, Sun Services will attempt to contact
all known customers to recommend proactive implementation.
For "Controlled Proactive" FINs, Sun Services mission critical
support teams will initiate proactive implementation efforts for
their respective accounts as required.
For "Reactive" FINs, Sun Services and partners will implement
the necessary corrective actions as the need arises.
Billing Information:
-------------------
Warranty: On-Site Labor Rates are based on specified Warranty deliverables
for the affected product.
Contract: On-Site Labor Rates are based on the type of service contract.
Non Contract: On-Site implementation by Sun is available based on On-Site
Labor Rates defined in the Price List.
--------------------------------------------------------------------------
All FIN documents are accessible via Internal SunSolve. Type "sunsolve"
in a browser and follow the prompts to Search Collections.
For questions on this document, please email:
[email protected]
The FIN and FCO homepage is available at:
http://sdpsweb.central/FIN_FCO/index.html
For more information on how to submit a FIN, go to:
http://pronto.central/fin.html
To access the Service Partner Exchange, use:
https://spe.sun.com
--------------------------------------------------------------------------