OpenVMS ALPF11X06_071 Alpha V7.1 - V7.1-1H2 F11BXQP ECO Summary
TITLE: OpenVMS ALPF11X06_071 Alpha V7.1 - V7.1-1H2 F11BXQP ECO Summary
Modification Date: 07-OCT-1999
Modification Type: Updated Kit: Supersedes ALPF11X05_071
NOTE: An OpenVMS saveset or PCSI installation file is stored
on the Internet in a self-expanding compressed file.
The name of the compressed file will be kit_name-dcx_vaxexe
for OpenVMS VAX or kit_name-dcx_axpexe for OpenVMS Alpha.
Once the file is copied to your system, it can be expanded
by typing RUN compressed_file. The resultant file will
be the OpenVMS saveset or PCSI installation file which
can be used to install the ECO.
Copyright (c) Compaq Computer Corporation 1998, 1999. All rights reserved.
OP/SYS: OpenVMS Alpha
COMPONENTS: F11BXQP
FILESERV
VMS$CONFIG-050_CACHE_SERVER.COM
SOURCE: Compaq Computer Corporation
ECO INFORMATION:
ECO Kit Name: ALPF11X06_071
ECO Kits Superseded by This ECO Kit: ALPF11X05_071
ALPF11X04_071
ALPF11X03_071
ECO Kit Approximate Size: 990 Blocks
Kit Applies To: OpenVMS Alpha V7.1, V7.1-1H1, V7.1-1H2
System/Cluster Reboot Necessary: Yes
Rolling Re-boot Supported: Yes
Installation Rating: INSTALL_1
1 - To be installed on all systems running
the listed version(s) of OpenVMS.
Kit Dependencies:
The following remedial kit(s) must be installed BEFORE
installation of this kit:
ALPBASE02_071 or later
In order to receive all the corrections listed in this
kit, the following remedial kits should also be installed:
ALPSYSA03_071 or later
ECO KIT SUMMARY:
An ECO kit exists for F11BXQP on OpenVMS Alpha V7.1 through V7.1-1H2.
This kit addresses the following problems:
Problems addressed in ALPF11X06_071:
o The ALPF11X05_071 installation checks to see if both the
ALPY2K01_071 and ALPBASE02_071 kits are installed. If not,
the installation fails. The ALPBASE02_071 remedial kit
includes the ALPY2K01_071 functionality. Therefore, the
check for the ALPY2K01_071 kit is redundant and unnecessary.
This kit corrects this problem by only checking for the
ALPBASE02_071 kit. There are no functional changes in this
kit. If you have installed the ALPF11X05_071 kit, you do
not need to install this kit.
Problems addressed in APF11X05_071:
o When two processes access a file via the MOVEFILE and
READATTR/FID_TO_SPEC mechanism, both processes try to delete
the 'primary_fcb' used to get the information in question.
This causes a NOTFCBFCB bugcheck to occur.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o If a process attempts to mount a bound volume set (BVS)
and all the members of the BVS are not present, an attempt
to lock the volume for REBUILDing the meta-data on the volume
will fail. However, the blocking lock (F11B$b) is left
with the process.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o An XQPERR bugcheck in LOCKERS can occur when the retry
limit on the F11B$x lock is reached.
This problem can occur when:
1. The owner of the $x lock is running at a high process
priority; and
2. A number of processes are in a clustered system are also
trying to validate this lock, but at a lower process
priority.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o After releasing the current process' IPL/Fork lock, a system
can crash with a SPLACQERR bugcheck
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o A directory file becomes "corrupt" and DUMP/DIRECTORY
displays a block similar to the following:
Virtual block number 3574 (00000DF6), 512 (0200) bytes
0000 Directory Entry:
0000 Size: 508
0002 Version limit: 32767
0004 Type: 0 (FID)
0005 Name count: 24
0006 Name: COSLR1201_01_JUPICC2.LIS
001E Version: 7859 FID: (40993,5,0)
0026 Version: 7858 FID: (40990,1,0)
002E Version: 7857 FID: (40988,3,0)
...
01E6 Version: 7802 FID: (40455,1,0)
01EE Version: 7801 FID: (40454,1,0)
01F6 Version: 32767 FID: (16744447,65535,0)
01FE End of records (-1)
The directory shuffle code creates the above erroneous
directory entry for a couple of reasons:
1. So that a new directory buffer will have a valid structure
(this allows VALIDATE_DIRBLK to write the block to disk)
2. So that the entry will be spotted as bogus (via VERIFY) if
we crash in the middle of this shuffle.
After the directory block (with the eroneous directory entry)
is written to disk, the bad entry is removed. A subsequent
call to READ_BLOCK assumes that the block comes from the
buffer cache and not from disk. Under heavy load, this
assumption may not be true as the directory block may have
been kicked out of the cache.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o Update the conversion of the ODS-1 date field to comply
with the RSX-11 version of date handling. Note that this
change is for the obsolete string format dates of DDMMMYY
of RSX and do not affect the 64 bit binary date stored on disk.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o Under the following circumstances:
1. A directory with multiple headers (from a large ACL for
example) is deleted on one node (A) in a cluster and
2. the directory had been previously accessed on another node
(B) in the cluster
The files created with the previously deleted headers in step
1 would show up on node B with the error:
%SYSTEM-F-NOSUCHFILE, no such file
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
Problems addressed in APF11X04_071:
o A reserved operand fault bugcheck occurs on a $QIO exit.
During the mounting of the system disk, the error message that
the disk is mounted with a reduced cache is suppressed. Hence,
the System Manager may be unaware that the performance of the
system disk and all others attached to the same cache block is
questionable.
Image Affected: [SYSEXE]FILESERV.EXE
o When deleting a large file (such as a system dump file), a
UNXSIGNAL Bugcheck may occur. This particular bugcheck occurs
because a variable in the code causes a reference to memory
data that the file system does not own and an internal access
violation occurs (ACCVIO).
Image Affected: [SYS$LDR]F11BXQP.EXE
o On some systems with higher rates of system paging or with
WSDEC set to a non-standard value, a system can crash with a
PGFLIPLHI (Page Fault IPL too High) bugcheck. This problem
happened when the system returns from a SMP$ACQUIRE call.
Image Affected: [SYS$LDR]F11BXQP.EXE
o An XQPERR can occur in the RDBLOK module during disk cleanup
using DCL.
Image Affected: [SYS$LDR]F11BXQP.EXE
o When storing the value of a directory index buffer, the system
may crash with a PGFLIPLHI, SSRVEXCPTN error.
Image Affected: [SYS$LDR]F11BXQP.EXE
o A dismount on a shadowded device results in an unnecessary
copy.
Image Affected: [SYS$LDR]F11BXQP.EXE
o An XQPERR bugcheck (crash) in the RETURN_CREDITS module can
occur during DISMOUNT.
Image Affected: [SYS$LDR]F11BXQP.EXE
o An XQPERR Bugcheck (crash) in XQP can occur during a SET ACL
(SET FILE/ACL) operation.
Image Affected: [SYS$LDR]F11BXQP.EXE
o When two processes are competing to dismount a volume, one
process may be just a bit faster than the other and delete the
VCB and other structures before the second process has time to
finish up its processing. The result is a UNXSIGNAL/ACCVIO
crash.
Image Affected: [SYS$LDR]F11BXQP.EXE
o A device reporting a read error (SS$_PARITY) during read/write
processing in the XQP will attempt to record the bad blocks
and FID in the BADLOG.SYS file. When the internal close
operation occurs (on BADLOG), the system XQPERR bugchecks
when it finds the process's dirty buffers have not been
written out.
Image Affected: [SYS$LDR]F11BXQP.EXE
o A UNXSIGNAL/ACCVIO error can occur at module F11BXQP. This
problem occurs during mount, when the primary volume is not
yet mounted.
Image Affected: [SYS$LDR]F11BXQP.EXE
o Processes can hang (deadlock) when dismounting a device.
Image Affected: [SYS$LDR]F11BXQP.EXE
o A 'no such file' error can occur on directory extension FCBs.
This problem can occur in at least two ways:
1. A file appears normal on one node but has an 'no such
file' error from another node.
2. BACKUP or DUMP /HEADER encounters a read attributes
error of NOSUCHFILE. This error occurs when an attempt
is made to read a file header, for which the FCB for the
old header is still in memory.
Image Affected: [SYS$LDR]F11BXQP.EXE
o Occasional false end-of-file (EOF) errors can occur on a read
operation.
Image Affected: [SYS$LDR]F11BXQP.EXE
o The XQP fails after an IO$_DEACCESS call with an SS$_BADPARAM
error. One cannot determine whether a file is still open or
not due to the failed IO$_DEACCESS call.
Image Affected: [SYS$LDR]F11BXQP.EXE
o Non-privileged users can change the revision date (and count)
of a file for which they should have only READ access. For
example, if a non-privileged user with READ-only file access
tries to set the file protection, a failure occurs with an
SS$_NOPRIV error as expected. However, the revision date (and
count) are modified.
Image Affected: [SYS$LDR]F11BXQP.EXE
Problems addressed in APF11X03_071:
o An XQPERR bugcheck error, "all the index buffers are active",
occurs when running with a reduced cache or during a backup.
o An XQPERR bugcheck error, "wrong lock basis with FCB present",
occurs when files with ACLs are created on a full volume.
o It is possible to serialize on the wrong volume in a volume set.
o If more than one process queues for a volume's activity
blocking lock, the XQP can deadlock.
o Various XQPERR bugchecks occur in directory scanning/shuffling.
o A crash may occur during creation of a file with a version
limit of 1 if SYSTEM_CHECK is on or bit 5 or 6 of ACP_DATACHECK
is 1. An XQPERR bugcheck occurs due to a corrupted directory.
o The NOTIFY_USER for final status of an XQP request occurs too
early in XQP request completion to report any errors produced
in cleanup or auditing. This normally does not cause a problem
since USER_STATUS is not expected to change during cleanup.
However, if it does change, then the output of SET WATCH FILE
is misleading.
o XQP requests to create a new file with version limits set can
fail with an SS$_NOSUCHFILE error.
o Deleting a stale alias on a directory with extension headers
can bugcheck with an XQPERR "lock index has shifted" error.
This problem occurs when:
1. An alias to an existing file is created;
2. The original file is deleted; and
3. The original file's file header is reused as an
extension header of the alias's parent directory
(when ACEs are added to the directory).
o Prevent access to file headers beyond index file highwater-marking
(HWM).
It is possible to ACCESS by FID a file header beyond the
current end of the index file on a freshly-initialized volume.
Creating a new file accessed in this way can bugcheck the
system with the XQPERR error.
o Fix a reserved operand fault bugcheck on $QIO exit. The
$QIO failed on return because the IPL was set to zero but
was entered at IPL 2.
o $GET_SECURITY was reading the ORB on a file without any
synchronization with the file system. In the best case,
this problem can lead to bad information being returned.
In the worst case, if the filesystem was rebuilding the
ORB's ACL chain at the time, a kernel mode ACCVIO can occur.
*** Note: In order to get this fix, ALPSYSA01_071 must also
be installed on the system.
INSTALLATION NOTES:
The images in this kit will not take effect until the system is
rebooted. If there are other nodes in the VMScluster, they must
also be rebooted in order to make use of the new image(s).
If it is not possible or convenient to reboot the entire cluster at
this time, a rolling re-boot may be performed.
All trademarks are the property of their respective owners.
Copyright (c) Compaq Computer Corporation 1998, 1999. All rights reserved.
Modification Date: 07-OCT-1999
Modification Type: ARCHIVED
OP/SYS: OpenVMS Alpha
COMPONENTS: F11BXQP
FILESERV
VMS$CONFIG-050_CACHE_SERVER.COM
SOURCE: Compaq Computer Corporation
ECO INFORMATION:
ECO Kit Name: ALPF11X05_071
ECO Kits Superseded by This ECO Kit: ALPF11X04_071
ALPF11X03_071
ECO Kit Approximate Size: 990 Blocks
Kit Applies To: OpenVMS Alpha V7.1, V7.1-1H1, V7.1-1H2
System/Cluster Reboot Necessary: Yes
Rolling Re-boot Supported: Yes
Installation Rating: INSTALL_1
1 - To be installed on all systems running
the listed version(s) of OpenVMS.
Kit Dependencies:
The following remedial kit(s) must be installed BEFORE
installation of this kit:
ALPBASE02_071
ALPY2K01_071
In order to receive all the corrections listed in this
kit, the following remedial kits should also be installed:
ALPSYSA03_071
ECO KIT SUMMARY:
An ECO kit exists for F11BXQP on OpenVMS Alpha V7.1 through V7.1-1H2.
This kit addresses the following problems:
Problems addressed in APF11X05_071:
o When two processes access a file via the MOVEFILE and
READATTR/FID_TO_SPEC mechanism, both processes try to delete
the 'primary_fcb' used to get the information in question.
This causes a NOTFCBFCB bugcheck to occur.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o If a process attempts to mount a bound volume set (BVS)
and all the members of the BVS are not present, an attempt
to lock the volume for REBUILDing the meta-data on the volume
will fail. However, the blocking lock (F11B$b) is left
with the process.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o An XQPERR bugcheck in LOCKERS can occur when the retry
limit on the F11B$x lock is reached.
This problem can occur when:
1. The owner of the $x lock is running at a high process
priority; and
2. A number of processes are in a clustered system are also
trying to validate this lock, but at a lower process
priority.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o After releasing the current process' IPL/Fork lock, a system
can crash with a SPLACQERR bugcheck
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o A directory file becomes "corrupt" and DUMP/DIRECTORY
displays a block similar to the following:
Virtual block number 3574 (00000DF6), 512 (0200) bytes
0000 Directory Entry:
0000 Size: 508
0002 Version limit: 32767
0004 Type: 0 (FID)
0005 Name count: 24
0006 Name: COSLR1201_01_JUPICC2.LIS
001E Version: 7859 FID: (40993,5,0)
0026 Version: 7858 FID: (40990,1,0)
002E Version: 7857 FID: (40988,3,0)
...
01E6 Version: 7802 FID: (40455,1,0)
01EE Version: 7801 FID: (40454,1,0)
01F6 Version: 32767 FID: (16744447,65535,0)
01FE End of records (-1)
The directory shuffle code creates the above erroneous
directory entry for a couple of reasons:
1. So that a new directory buffer will have a valid structure
(this allows VALIDATE_DIRBLK to write the block to disk)
2. So that the entry will be spotted as bogus (via VERIFY) if
we crash in the middle of this shuffle.
After the directory block (with the eroneous directory entry)
is written to disk, the bad entry is removed. A subsequent
call to READ_BLOCK assumes that the block comes from the
buffer cache and not from disk. Under heavy load, this
assumption may not be true as the directory block may have
been kicked out of the cache.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o Update the conversion of the ODS-1 date field to comply
with the RSX-11 version of date handling. Note that this
change is for the obsolete string format dates of DDMMMYY
of RSX and do not affect the 64 bit binary date stored on disk.
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
o Under the following circumstances:
1. A directory with multiple headers (from a large ACL for
example) is deleted on one node (A) in a cluster and
2. the directory had been previously accessed on another node
(B) in the cluster
The files created with the previously deleted headers in step
1 would show up on node B with the error:
%SYSTEM-F-NOSUCHFILE, no such file
Image(s) Affected: [SYS$LDR]F11BXQP.EXE
Problems addressed in APF11X04_071:
o A reserved operand fault bugcheck occurs on a $QIO exit.
During the mounting of the system disk, the error message that
the disk is mounted with a reduced cache is suppressed. Hence,
the System Manager may be unaware that the performance of the
system disk and all others attached to the same cache block is
questionable.
Image Affected: [SYSEXE]FILESERV.EXE
o When deleting a large file (such as a system dump file), a
UNXSIGNAL Bugcheck may occur. This particular bugcheck occurs
because a variable in the code causes a reference to memory
data that the file system does not own and an internal access
violation occurs (ACCVIO).
Image Affected: [SYS$LDR]F11BXQP.EXE
o On some systems with higher rates of system paging or with
WSDEC set to a non-standard value, a system can crash with a
PGFLIPLHI (Page Fault IPL too High) bugcheck. This problem
happened when the system returns from a SMP$ACQUIRE call.
Image Affected: [SYS$LDR]F11BXQP.EXE
o An XQPERR can occur in the RDBLOK module during disk cleanup
using DCL.
Image Affected: [SYS$LDR]F11BXQP.EXE
o When storing the value of a directory index buffer, the system
may crash with a PGFLIPLHI, SSRVEXCPTN error.
Image Affected: [SYS$LDR]F11BXQP.EXE
o A dismount on a shadowded device results in an unnecessary
copy.
Image Affected: [SYS$LDR]F11BXQP.EXE
o An XQPERR bugcheck (crash) in the RETURN_CREDITS module can
occur during DISMOUNT.
Image Affected: [SYS$LDR]F11BXQP.EXE
o An XQPERR Bugcheck (crash) in XQP can occur during a SET ACL
(SET FILE/ACL) operation.
Image Affected: [SYS$LDR]F11BXQP.EXE
o When two processes are competing to dismount a volume, one
process may be just a bit faster than the other and delete the
VCB and other structures before the second process has time to
finish up its processing. The result is a UNXSIGNAL/ACCVIO
crash.
Image Affected: [SYS$LDR]F11BXQP.EXE
o A device reporting a read error (SS$_PARITY) during read/write
processing in the XQP will attempt to record the bad blocks
and FID in the BADLOG.SYS file. When the internal close
operation occurs (on BADLOG), the system XQPERR bugchecks
when it finds the process's dirty buffers have not been
written out.
Image Affected: [SYS$LDR]F11BXQP.EXE
o A UNXSIGNAL/ACCVIO error can occur at module F11BXQP. This
problem occurs during mount, when the primary volume is not
yet mounted.
Image Affected: [SYS$LDR]F11BXQP.EXE
o Processes can hang (deadlock) when dismounting a device.
Image Affected: [SYS$LDR]F11BXQP.EXE
o A 'no such file' error can occur on directory extension FCBs.
This problem can occur in at least two ways:
1. A file appears normal on one node but has an 'no such
file' error from another node.
2. BACKUP or DUMP /HEADER encounters a read attributes
error of NOSUCHFILE. This error occurs when an attempt
is made to read a file header, for which the FCB for the
old header is still in memory.
Image Affected: [SYS$LDR]F11BXQP.EXE
o Occasional false end-of-file (EOF) errors can occur on a read
operation.
Image Affected: [SYS$LDR]F11BXQP.EXE
o The XQP fails after an IO$_DEACCESS call with an SS$_BADPARAM
error. One cannot determine whether a file is still open or
not due to the failed IO$_DEACCESS call.
Image Affected: [SYS$LDR]F11BXQP.EXE
o Non-privileged users can change the revision date (and count)
of a file for which they should have only READ access. For
example, if a non-privileged user with READ-only file access
tries to set the file protection, a failure occurs with an
SS$_NOPRIV error as expected. However, the revision date (and
count) are modified.
Image Affected: [SYS$LDR]F11BXQP.EXE
Problems addressed in APF11X03_071:
o An XQPERR bugcheck error, "all the index buffers are active",
occurs when running with a reduced cache or during a backup.
o An XQPERR bugcheck error, "wrong lock basis with FCB present",
occurs when files with ACLs are created on a full volume.
o It is possible to serialize on the wrong volume in a volume set.
o If more than one process queues for a volume's activity
blocking lock, the XQP can deadlock.
o Various XQPERR bugchecks occur in directory scanning/shuffling.
o A crash may occur during creation of a file with a version
limit of 1 if SYSTEM_CHECK is on or bit 5 or 6 of ACP_DATACHECK
is 1. An XQPERR bugcheck occurs due to a corrupted directory.
o The NOTIFY_USER for final status of an XQP request occurs too
early in XQP request completion to report any errors produced
in cleanup or auditing. This normally does not cause a problem
since USER_STATUS is not expected to change during cleanup.
However, if it does change, then the output of SET WATCH FILE
is misleading.
o XQP requests to create a new file with version limits set can
fail with an SS$_NOSUCHFILE error.
o Deleting a stale alias on a directory with extension headers
can bugcheck with an XQPERR "lock index has shifted" error.
This problem occurs when:
1. An alias to an existing file is created;
2. The original file is deleted; and
3. The original file's file header is reused as an
extension header of the alias's parent directory
(when ACEs are added to the directory).
o Prevent access to file headers beyond index file highwater-marking
(HWM).
It is possible to ACCESS by FID a file header beyond the
current end of the index file on a freshly-initialized volume.
Creating a new file accessed in this way can bugcheck the
system with the XQPERR error.
o Fix a reserved operand fault bugcheck on $QIO exit. The
$QIO failed on return because the IPL was set to zero but
was entered at IPL 2.
o $GET_SECURITY was reading the ORB on a file without any
synchronization with the file system. In the best case,
this problem can lead to bad information being returned.
In the worst case, if the filesystem was rebuilding the
ORB's ACL chain at the time, a kernel mode ACCVIO can occur.
*** Note: In order to get this fix, ALPSYSA01_071 must also
be installed on the system.
INSTALLATION NOTES:
The images in this kit will not take effect until the system is
rebooted. If there are other nodes in the VMScluster, they must
also be rebooted in order to make use of the new image(s).
If it is not possible or convenient to reboot the entire cluster at
this time, a rolling re-boot may be performed.
All trademarks are the property of their respective owners.
This patch can be found at any of these sites:
Colorado Site
Georgia Site
Files on this server are as follows:
alpf11x06_071.README
alpf11x06_071.CHKSUM
alpf11x06_071.CVRLET_TXT
alpf11x06_071.a-dcx_axpexe
alpf11x06_071.CVRLET_TXT
|
