Processor Memory Parity Errors (PMPEs)

• Hard parity errors

  These errors occur when there is a chip or board failure that corrupts data. In this case, you need to re-seat or replace the affected component, which usually involves a memory chip swap or a board swap. There is a hard parity error when multiple parity errors occur at the same address. There are more complicated cases that are harder to identify. In general, if you see more than one parity error in a particular memory region in a relatively short period, you can consider it to be a hard parity error.

Studies have shown that soft parity errors are 10 to 100 times more frequent than hard parity errors. Therefore, Cisco highly recommends you to wait for a second parity error before you replace anything. This greatly reduces the impact on your network.

Isolate the Problem

A router has memory in different locations. Theoretically, any memory location can be affected by the parity error, but most memory problems occur in dynamic RAM (DRAM) or shared RAM (SRAM). Based on the platform, here is how you can find out what memory location has been affected, and, if it turns out to be a hard parity error, what part you must replace:

Cisco 4500 and 4700 Platforms

On the Cisco 4500 and 4700 platforms, the crashinfo file is not available in versions earlier than Cisco IOS® Software Release 12.2(10) and 12.2(10)T.

One way to find out where the error occurred is to look at the "restart reason" in the console logs, and in the output of the show version command:

• Parity Error in DRAM:

  If you did not manually reload the router after the crash, the show version output looks like this:

  System restarted by processor memory parity error at PC 0x601799C4,
   address 0x0 
  System image file is "flash:c4500-inr-mz.111-14.bin", booted via flash

  If a crashinfo file is available, or if console logs have been captured, you can also see something like this:

  *** Cache Error Exception *** 
   Cache Err Reg = 0xa0255c61 
   data reference, primary cache, data field error , error on SysAD Bus 
   PC = 0xbfc0edc0, Cause = 0xb800, Status Reg = 0x34408007

  Repeated occurrence of parity errors in DRAM indicates that either the DRAM or the chassis is defective. If you recently removed the chassis, or if you performed any hardware configuration changes, re-seat the DRAM chips to solve the problem. Otherwise, replace the DRAM as a first step. This must prevent the parity errors. If the router still crashes, replace the chassis.

• Parity Error in SRAM:

  If you did not manually reload the router after the crash, the show version command output looks like this:

  System  restarted by shared memory parity error at PC 0x60130F40 
  System image file is "flash:c4500-inr-mz.111-14.bin", booted via flash

  If a crashinfo file is available, or if console logs have been captured, you can also see something like this:

  *** Shared Memory Parity Error *** 
  shared memory control register= 0xffe3 
  error(s) reported for: CPU on byte(s): 0/1

  or

  %PAR-1-FATAL: Shared memory parity error 
  shared memory status register= 0xFFEF 
  error(s) reported for: CPU on byte(s): 0/1 2/3

  or

  *** Shared Memory Parity Error *** 
  shared memory control register= 0xffdf 
  error(s) reported for: NIM1 on byte(s): 0/1  2/3

Note:

• If the error is reported for the CPU, replace the SRAM.

• If the error is reported for NIM(x), replace the network module in slot (x). The SRAM allocated to slot (x) can also be affected. In this case, replace the SRAM.

  Repeated parity errors in SRAM indicate either defective SRAM chips, or a defective network module that has written bad parity in the SRAM. If you removed the chassis recently, or if you made any hardware configuration changes, re-seat the network modules and the SRAM chips to solve the problem. Otherwise, check where the error is reported in the console logs (see the output example above).

Route/Switch Processor (RSP), Network Processing Engine (NPE), and Route Processor (RP) Platforms

As with the Cisco 4000 series, the problem can be due to faulty DRAM or SRAM for these platforms. The problem can also be because of a defective processor card (RP, RSP or NPE). The Cisco 7000 and 7500 can also report parity errors generated by a faulty or badly seated Interface Processor (legacy xIP or VIP).

Check the crashinfo file and the console logs for one of these error messages:

Parity Error in DRAM or SRAM (MEMD)

For the RP, RSP and NPE, you usually see something like this:

Error: primary data cache, fields: data, (SysAD) 
virtual addr 0x6058A000, physical addr(21:3) 0x18A000, vAddr(14:12) 0x2000 
virtual address corresponds to main:data, cache word 0

or simply:

Error: primary data cache, fields: data, SysAD
phy21:3 0x201880, va14:12 0x1000, addr 63E01880

This indicates a problem on the RSP itself. If the problem only occurs once, it is most probably a transient issue.

Parity Error Pulled from SRAM

For the RSP, the message can look like this:

%RSP-3-ERROR: MD error 0000008000000200 
%RSP-3-ERROR: QA parity error (bytes 0:3) 02 
%RSP-3-ERROR: MEMD parity error condition 
%RSP-2-QAERROR: reused or zero link error, write at addr 0100 (QA) 
    log 22010000, data 00000000 00000000 
%RSP-3-RESTART: cbus complex

or

%RSP-3-ERROR: CyBus error 01 
%RSP-3-ERROR: read data parity 
%RSP-3-ERROR: read parity error (bytes 0:7) 20 
%RSP-3-ERROR: physical address (bits 20:15) 000000

If there is no indication of another interface processor that writes bad parity into the SRAM (for example, VIP2-1-MSG error messages), the most likely reason for the parity error is the SRAM itself. In this case, replace the RSP.

If other error messages indicate that an interface processor writes bad parity, it can be a faulty or badly-seated card.

Versatile Interface Processor

If you receive %VIP2-1-MSG: slot(x) messages in the logs or in the crashinfo file, refer to Troubleshooting VIP Crashes.

Recommended Actions

At the first occurrence of a parity error, it is not possible to distinguish between a soft or hard parity error. From experience, most parity occurrences are soft parity errors, and you can usually dismiss them. If you have recently changed some hardware or have moved the box, try to re-seat the affected part (DRAM, SRAM, NPE, RP, RSP, or VIP). Frequent multiple parity occurrences signify faulty hardware. Replace the affected part (DRAM, RSP, VIP, or motherboard) with the help of the instructions mentioned in this document.

Information to Collect if You Open a TAC Service Request

If you still need assistance after you follow the troubleshooting steps above and want to open a service request with the Cisco TAC, be sure to include this information:

Troubleshooting performed before you opened the service request. show technical-support command output (in enable mode if possible). show log command output or console captures if available. crashinfo file (if it is present, and not already included in the show technical-support command output. If multiple crashinfo files exist, include all of them). Number of reloads due to processor memory parity errors that you have seen and when they have occurred. Please attach the collected data to your case in non-zipped, plain text format (.txt). In order to attach information to your service request, upload it through the TAC Service Request Tool (registered customers only) . If you cannot access the Service Request Tool, attach the relevant information to your service request, and send it to attach@cisco.com with your service request number in the subject line of your message. Note: Do not manually reload or power-cycle the router before you collect the above information unless required to troubleshoot a processor memory parity error, because this can cause important information to be lost that is needed to determine the root cause of the problem.

Related Information

• Cisco 7200 Parity Error Fault Tree
• Troubleshooting Router Crashes
• Retrieving Information from the Crashinfo File
• Memory Errors, Detection, and Correction
• Technical Support - Cisco Systems