Skip to: Abstract | Comments | Solution | Related Segments | Migration | Download
APAR NUMBER: PJ29617 PRODUCT: TPF4 FUNCTIONAL AREA: CRAS SUPPORT SHIPPED IN PUT: 19 ABSTRACT: Soft IPL while PRC is on an alternate CRAS set may result in incorrect data being stored in the CRAS table. APAR CONTENTS: FEATURE TYPE STATUS NAME Base Copy Changed source/cp/base/cpsf40.cpy Base Copy Changed source/cp/base/ct4140.cpy PREREQUISITE APARS: To be applied in order listed. Base Copy Changed source/cp/base/cpsf40.cpy PJ25497(09) PJ25654(09) PJ25760(09) PJ26311(10) PJ26320(11) PJ26686(11) PJ26707(11) PJ27387(13) PJ27333(13) PJ27393(13) PJ27469(13) PJ26826(14) PJ28097(15) PJ28649(17) PJ28902(17) PJ28908(17) PJ28622(17) Base Copy Changed source/cp/base/ct4140.cpy PJ26707(11) PJ26687(11) PJ27023(12) PJ27446(14) PJ27814(15) PJ28133(15) PJ28097(15) PJ27785(15) PJ28136(15) PJ28799(17) PJ28758(17) PJ28622(17) PJ29005(17) PJ29111(18) COMMENTS: Information about the console configurations of each processor in a TPF complex is stored in the system CRAS table. This table is contained in keypoint C, whose file copy is shared among all the processors in a complex, and whose core copy is identical and kept in sync among all the processors in a complex. Because of this design, when a fallback occurs for a particular processor's PRC or RO, whether it be a result of a manual or an automatic fallback, any network terminal that is defined in the same CRAS table as an alternate CRAS set is eligible to be used as the new PRC or RO. In fact, the alternate CRAS terminal may even be owned by another processor in the complex. As a result, after the fallback occurs, the console LNIATA and CPUID fields of the system are updated in the CRAS table. If the TPF operator then enters a ZRIPL command to re-IPL the TPF system while the PRC is still on the alternate CRAS set, when the system comes back up through CTIN in copy member CT41 or CCCTIN, TPF will restore the LNIATA of the PRC and RO their appropriate addresses (010000 and 000000 for 3270 console support, or 010000 and 010002 for 3215 console support). However, the CPUID associated with the PRC and RO fields in the CRAS table are not reinitialized and may still contain what is now the incorrect CPUID of the alternate CRAS device from before the soft IPL. As a result, as the system continues to IPL through restart and into 1052 state or above, each message that is sent to the system console may result in the message either being routed to another system console, or may result in an OPR-I00034C (message undeliverable) from real-time segment CVAA. In a related condition, during cycle-down processing after the ZRIPL command has been entered from the alternate CRAS set (which is now the PRC for the particular system), CCCPSF will hang in cycle-down processing and will prevent the TPF system re-IPL from continuing. This scenario can occur in a system defined with 3215 console support if an automatic fallback occurred from the primary 3215 console on a system that does not have an alternate 3215 PRC device defined or available. In this automatic fallback scenario, the Line Status Table (LSTB) for the 3215 PRC can be left in such a way that prevents CCCPSF from mounting the correct 3215 PRC device address. During system cycle-down, as part of re-IPL processing, CCCPSF makes sure that it is able to mount the 3215 PRC if it needs to send a message to the system PRC to indicate a critical error (such as being unable to mount the IPL volume). Because the automatic fallback of the PRC occurred and the alternate 3215 PRC device was unavailable (causing automatic fallback to continue to the alternate CRAS set), when CCCPSF examines the 3215 PRC LSTB entry, it tries to mount the incorrect 3215 device address. The mount failure of that 3215 PRC device causes CCCPSF to store a message into a special message area, at label CPSFM06A, indicating that the PRC cannot be mounted. CCCPSF then goes into a tight loop while waiting for the operator to take a stand-alone dump and then issues a hard system IPL. SOLUTION: Copy member CT41 of CP CSECT CCCTIN has been modified to reinitialize the owning CPUID fields of that system's RO and PRC console in the CRAS status table at the point that CT41 is also reinitializing the LNIATA of the consoles. This will prevent routing errors and OPR-I00034C dumps as the system continues through restart. Copy member CPSF of CP CSECT CCCPSF has been modified to do additional checking on the status of the PRC if it is unable to mount the PRC device address specified in the PRC LSTB entry during ZRIPL cycle-down processing. If CPSF is unable to mount the 3215 PRC and it sees that the PRC LSTB entry also indicates that the PRC is currently not valid, it will bypass the hard loop because the condition is most likely that the PRC is on a network device. Now, only when the PRC cannot be mounted and some other critical error has occurred for which CCCPSF needs to send a message to the console, will the system go into a hard loop and the message field at label CPSFM06A will contain the more relevant message of what the problem is (rather than "PRC DEVICE NOT ABLE TO BE MOUNTED"). Though this does not eliminate the condition where CCCPSF will get into a tight loop, it makes the condition much more unlikely. DEPENDENCIES SEGMENTS TO BE ASSEMBLED OR COMPILED: CCCTIN, CCCPSF SEGMENTS TO BE LINK EDITED: cps040.exe COREQS: None. MIGRATION CONSIDERATIONS: None. UPDATED INFORMATION UNITS: None. See your IBM representative if you need additional information. DOWNLOAD INSTRUCTIONS: http://www.ibm.com/software/htp/tpf/pages/maint.htm -- END APAR PJ29617
Download file(s): Login once to access server, leave window open, then click on link(s) below.Source