>Would this apply to z/vm 4.2?
>
>Would it fix the following problem?  SPXTAPE encountered a tape
>error, it hadn't yet cancelled itself and the SPXTAPE CANCEL command
>wasn't issued, but the userid was forced off.  This happened many
>weeks ago and the user is still hung.
>
>     - Tom.

Tom,

While the problem could occur on z/VM 4.2 (in the form of an SPO001
soft abend), it is not likely that this APAR is related to your problem.
The problem that the APAR fixes is caused by the status of DUMP
spool files that are selected to be written to tape by SPXTAPE DUMP.

HPC010 abends are new in z/VM 5.2.0, but we have seen several occurrences
of the SPO001 abend on earlier releases of z/VM. We used the
information from many of these occurrences to find the pattern that led
us to the cause of this problem.

There are PTFs for this APAR for z/VM 5.2.0, 5.1.0, and 4.4.0. The
description of the problem and solution are below. And yes, it WILL be
on an RSU.

John Franciscovich
z/VM Development
-------------------------------------------------------------------------
****************************************************************
* USERS AFFECTED: Users of SPXTAPE                             *
****************************************************************
* PROBLEM DESCRIPTION:                                         *
****************************************************************
* RECOMMENDATION: APPLY PTF                                    *
****************************************************************
ABENDHPC010 HPC010 ABENDSPO001 SPO001
These abends are caused by the interaction of APAR VM62853 and
SPXTAPE.  VM62853 changed HCPDMQ to turn off SPFOPEN and turn on
SPFERRPU when deleting an SPFBK that represents a pre-allocated
CP dump file.  This occurs when a CP dump file requires more
DASD space than is currently allocated.

HCPSPK adds files that are open for creation (including
pre-allocated CP dump files) to the list of files that meet the
specified selection criteria to be dumped (FLSPT).  Then HCPSPX
checks the SPFOPEN bit to decide if the file should actually be
dumped.  If SPFOPEN is ON, SPXTAPE will not dump the file.  If
it is OFF, SPXTAPE will dump it.

If HCPDMQ changes the SPFOPEN and SPFERRPU bits between the time
when HCPSPK selects the file and HCPSPX decides whether to
actually dump it, then HCPSPX sees it as closed (SPFOPEN = OFF)
and attempts to dump the file. Since the file has no data pages,
the abends occur during SPXTAPE dump processing and resource
clean up for the file.

This scenario is likely to happen since SPXTAPE uses so many
system virtual pages for SPDBK buffers, creating a greater need
for pre-allocated dump space.

CONCLUSION:
Code is added to HCPSPK and HCPSPX to disallow the selection
by SPXTAPE of open CP dump files and files with the SPFERRPU
bit on.

TEMPORARY FIX:
FOR RELEASE VM/ESA CP/ESA R440 :
PREREQ: NONE
CO-REQ: NONE
IF-REQ: NONE
FOR RELEASE VM/ESA CP/ESA R510 :
PREREQ: NONE
CO-REQ: NONE
IF-REQ: NONE
FOR RELEASE VM/ESA CP/ESA R520 :
PREREQ: NONE
CO-REQ: NONE
IF-REQ: NONE

Reply via email to