Dear Customer, Support Level Supplement (SLS) nsc1011g, the UnixWare 7.1.1 NonStop Clusters Release 7.1.1+IP Maintenance Pack, has bundled these UnixWare 7.1.1 patches for installation on the UnixWare 7 NonStop Clusters 7.1.1+IP product. Software Notes and Recommendations ---------------------------------- nsc1011g should only be installed on: UnixWare 7 NonStop Clusters Release 7.1.1+IP Installation Instructions ------------------------- 1. Download the nsc1011g.Z file to the /tmp directory on your machine. 2. Make sure all the nodes are up and have joined the cluster. Then as root, uncompress the file and add the package to your system using these commands: $ su Password: # uncompress /tmp/nsc1011g.Z # pkgadd -qd /tmp/nsc1011g all # rm /tmp/nsc1011g 3. Reboot the system after installing this package. The release notes displayed prior to installation can be found in: /var/sadm/pkg/nsc1011/install/nsc1011.txt Removal Instructions -------------------- 1. As root, remove the package using these commands: $ su Password: # pkgrm nsc1011 2. Reboot the system after removing this package. Software Notes and Recommendations ---------------------------------- If you have questions regarding this SLS, or the product on which it is installed, please contact your software supplier. SLS nsc1011g should only be installed on: UnixWare 7 NonStop Clusters Release 7.1.1+IP Problems Fixed in this Maintenance Pack --------------------------------------- The UnixWare 7.1.1 NonStop Clusters Release 7.1.1+IP Maintenance Pack addresses all of the issues listed below: 1. SLS ptf7080d: UnixWare 7.0.0, 7.0.1 and 7.1.1 ticots and ticotsor Supplement. A client/server program that uses socketpair(3) to write/read data and do I_SENDFD/I_RECVFD, can fail with EINVAL on read. 2. SLS ptf7410e: UnixWare 7.1.0 and 7.1.1 libthread supplement A. Forked processes of a multithreaded application, which frequently uses fork1() from different threads, may occasionally hang. libthread's _thr_forklock mutex needs to be cleared in the child process, in case it was locked by another thread of the parent at the instant of forking. Note that this problem only applied to UnixWare 7.1.0. B. An application using libthread's asyncio core dumps with SIGSEGV, because indexing into an array using a legal and valid, but high, fd number gives an array sized too small for this index. This number correlates to an address off the end of the heap segment, hence the segmentation violation(SIGSEGV). Note that this problem only applied to UnixWare 7.1.0. C. A race condition exists in libthread's signal handling. The handler sets the thread's signal mask to the state it should be in for running the application's signal handler and then invokes the handler. Before the handler is invoked, another signal may be processed. If that signal's handler does not return, the original signal will never be processed. This is dangerous for the internal thread library signal, SIGLWP, since missing this signal can mean missing a suspension, preemption or cancellation. D. The special _cond_wait_sig variant of cond_wait used by the Java virtual machine enables all signals, not just SIGLWP, before re-grabbing the mutex on a wake-up. When non-SIGLWP signal handlers exit with siglongjmp, the normal return from _cond_wait_sig is bypassed and ownership of the mutex is unpredictable. E. getitimer and setitimer incorrectly report the existing value of a thread's interval timer. If the thread is multiplexed, the two microseconds' fields of the itimer are always reported to be zero. F. A multiplexed thread will have the microseconds fields of its interval timer set to zero during a call to aio_memlock. ptf7410d only adds the ability to install in quiet mode. G. pthread mutex locks will be used instead of basic mutex locks. This handles the lock ownership correctly for the ERRORCHECK mutex locks. Additionally, proper support of pthread_join as being an implicit cancellation point was added. 3. SLS ptf7411b: UnixWare 7.1.0 passwd Display Patch This SLS addresses a problem where passwd -sa displays incorrect information about the status of passwords. For example, passwd -sa may list users as locked (LK) when they are unlocked. SLS ptf7411b also provides a new passwd binary to address a problem with ttymon when logging into accounts that are configured with "Force the user to change the password at the next login". A login to these accounts is not possible if the ttymon is configured with PPP support (-P option to ttyadm). For example, this message will repeatedly occur: Welcome to UnixWare 7.1.0 The system's name is BLUTACK. login: test UX:login: ERROR: You don't have a password. UX:login: TO FIX: Choose one. UX:passwd: ERR: Sorry 4. SLS ptf7426c: UnixWare 7.1.0 and 7.1.1 ksh Supplement A. Exiting from a shell function does not exit script. If a read executing in a shell script is interrupted and there is a trap handler for the signal (if exit is called in the trap handler) then the actual exit() does not take place. Instead, the execution restarts at the instruction following the interrupted read in the main body. B. The shell is not handling SIGHUP correctly, resulting in ksh not releasing login. 5. SLS ptf7601h: UnixWare 7.1.1 inet and socksys Driver Supplement A. With kma_paranoid running, the system can panic after finding corrupt free buffers. mtt_replace() can corrupt the links if both "new" and "old" currently have no minor number assigned. B. Applications that use t_optmgmt() may cause a system panic. C. If a connect(3) is attempted using AF_UNIX to a port without a listener, this correctly fails with ECONNREFUSED. However, if a listener is then started and the connect(3) attempted again without closing the socket, the connect(3) will succeed, but a subsequent write will fail with EPIPE. D. If listen(3) is called with a backlog of 0, UnixWare 7.1.1 does not accept any connections. E. In-Kernel Sockets causes an application built on SCO UnixWare 2.1.x to block without a SIGALARM signal. F. A special case in the tcp_uinput() routine sends an ACK immediately, even when tcp_delay_acks is set. (This is to cater to small data flow, such as telnet.) This may cause excessive ACKs to be sent on some workloads. G. Systems under heavy network load can panic at in_chek_mgmt_que+27. H. Systems under heavy network load can panic. The stack trace shows tcp_itmr_keep() in the stack trace. I. Systems under very heavy network load can appear to hang. J. A program that attempts a t_unbind(), after a t_connect() fails, can remain in a SYN_SENT state without timing out. K. Select returns invalid read status on udp sockets. L. If doing an asynchronous connect, calling getsockopt() to get the error reason for the connect failing will return a random error code. M. Earlier versions of ptf7601 caused various tcp panics, including a panic at tcp_io+18e when doing init 6 and at tcpclose+le0. N. System hangs and the osmlog display the message: NOTICE: Kernel virtual memory for buffer I/O temporarily exhausted; using reserved pool to allow paging to proceed. O. ip_mtt list corruption causes system panic. 6. SLS ptf7602c: UnixWare 7.1.1 vxfs Driver Supplement A. Some actions can consume the whole of memory, forcing the system to start swapping. B. VxFS causes a kma paranoid panic due to not checking for vnode softcounts before kmem_free()ing the vx_biginode. C. A problem was fixed in the vx_rename1() routine of the VxFS driver that caused a kernel panic. 7. SLS ptf7603g: UnixWare 7.1.1 fs Driver Supplement A. buf_breakup or bufbrkup_copy causes a stack overflow panic when used along with DDI8 HBAs that specify bcb_addrtypes BA_SCGTH. B. Some kernel overhead needed to be eliminated for keeping accurate waitio statistics. C. The performance of the kernel locking routines needed improvement. D. Various panics occurred, including a panic at pvn_memunresv_pp_l+2f. E. The cpio(1) tape device volume limit was increased from 2GB to 4GB. F. Machine hangs may occur if syslogd is enabled. G. Privileged processes could core dump. H. A system panic in logcons() at logcons+75 was fixed. I. Use of cmn_err() function with the '%s' format specifier caused spurious characters in /var/adm/log/osmlog file. J. System appears to run out of streams memory. K. The dump feature has been modified to lessen the chance of its failing. This improves double panic processing and reduces hangs during sysdump. L. Performance suffers when multiple controllers are accessed with multiple dd processes. Note: SLS ptf7603e was never released publicly due to a problem found with creating emergency recovery floppies. M. An inability to create emergency recovery floppies. N. buf_breakup or bufbrkup_copy causes a stack overflow panic when used along with DDI8 HBAs that specify bcb_addrtypes BA_SCGTH. O. Some kernel overhead needed to be eliminated for keeping accurate waitio statistics. P. The performance of the kernel locking routines needed improvement. Q. Various panics occurred, including a panic at pvn_memunresv_pp_l+2f. R. The cpio(1) tape device volume limit was increased from 2GB to 4GB. This fix was previously released in ptf7605c. S. Machine hangs may occur if syslogd is enabled. This fix was previously released in ptf7605c. T. Privileged processes could core dump. This fix was previously released in ptf7626. U. A system panic in logcons() at logcons+75 was fixed. V. Use of cmn_err() function with the '%s' format specifier caused W. System appears to run out of streams memory. X. An inability to create emergency recovery floppies. Y. Performance suffers when multiple controllers are accessed with multiple dd processes. Z. More Reliable System Dump provides the following:- An improved dump mechanism that increases the percentage of valid kernel dumps. Double-panic processing has been improved and the frequency of hangs during system dumps has been reduced. The dump I/O size has been increased from 8K to 64K. The overall speed of the dump process has also been improved. Following a system panic and before reboot begins, the stack backtrace and register values are now written to the /var/adm/syslog file. This acts as a ``fingerprint'' of what happened, in case the dump save was not successful. In addition, a system dump can now be obtained on demand, on a live system, with the new dumpmemory(1M) command. A dump, thus obtained, can be analyzed with crash(1M) just as any other dump. AA. Unattended Reboot provides the following: The unattended reboot feature improves the behavior of the PANICBOOT parameter and offers the potential to reduce downtime for you and your users. You can now save a memory dump after a panic, and reboot without needing to be present at the console to respond to prompts. To obtain the unattended reboot functionality on your system, you will need to change two parameters: PANICBOOT and TIME. It is good practice to use the System Defaults Manager in SCOadmin or the defadm(1M) command to change the values of system parameters. Do not edit system files by hand (unpredictable system behavior may result). Parameter Defined in file Default Value needed to obtain value unattended reboot functionality ___________________________________________________________________ PANICBOOT /etc/default/init NO YES TIME /etc/default/dump -- either 0 (zero) or a positive number When PANICBOOT is set to YES, a memory dump is created (in either the dump slice or the primary swap device) following a system panic, and the system automatically reboots. When TIME is set to zero, prompts about saving the dump are never asked at the console. When TIME is set to a positive number of seconds, the system asks whether to save the dump, waits TIME number of seconds, then continues to reboot. In both cases the dump is created on either a dump slice or the primary swap device. If TIME is not defined, or is a negative number, the system asks whether to create the dump and waits forever for a response. BB. erg501391 - fz511405 The dumpmemory command doesn't work for machines with over 2GB memory. erg501391 - fz511392 System hangs when a dump is taken during high disk activity. erg501391 - fz511396 The sysdump behavior is not in accordance to the tunables PANICBOOOT, and TIME. 9. SLS ptf7607e: UnixWare 7.1.1 libc Supplement The pt.c functions have been enhanced to support up to 99999 ptys. Previously, the Unix98 pty support routines could not allow more than 999 ptys to be allocated by a user process. A. When logging into accounts which are configured with "Force the user to change the password at the next login", a login to these accounts is not possible if the ttymon is configured with PPP support (-P option to ttyadm). B. ksh93 pattern expansion bug. C. Incoming baud rates of 115k do not seem to be handled correctly by ttymon. D. ecvt, ecvtl, fcvt and gcvt have rounding errors in libc. E. select(3) can return 0 on blocking call if poll(2) reports POLLHUP and rd/wr fdsets not set. SLS ptf7607e additionally addresses these issues: F. Installation of Oracle 8i (8.1.6) can fail (core dump) when set71101g, the SCO UnixWare Maintenance Set, or the stand alone SLS ptf7607d, (which is included in set71101g) is installed on the system. Also, in a post Oracle 8i (8.1.6) installation environment, Oracle processes can core dump. G. A UNIX 95 Test failure introduced by the (e,f,g)cvt() in ptf7607b existed. 10. SLS ptf7608c: UnixWare 7.1.1 sd01 Driver Supplement A. If an HBA fails to assign memory, it can return NULL. This can cause sdi_xtranslate() to return SDI_RET_RETRY. However, since there is no retry code in sd01strat1(), a retry can result in a system hang. B. There was excessive locking code in the sd01 driver. C. KMA abuse occurs in sd01getalts(). D. fz511368 erg711327 In an MPIO environment the system appears to hang when one path to a disk fails. 11. SLS ptf7609c: UnixWare 7.1.1 pkgadd Supplement The UnixWare 7.1.1 pkgadd Supplement addresses a problem where pkgadd(1M), used with the -p option, may fail if the tape contains a large package whose table of contents is written over more than one 512-byte block. The following is an example of a failing command and the error messages: #pkgadd -p -d ctape1 MyPackage Installation in progress. Do not remove the cartridge. UX:pkgadd: ERROR: attempt to process package from failed - bad format in datastream table-of-contents or: UX:cpio: HALT: Bad magic number/header. 1 error UX:pkgadd: ERROR: attempt to process package from failed The data is read in 512-byte blocks. If the data is not terminated with a null byte, it is possible for pkgadd to 'fall off the end' of the data 12. SLS ptf7610b: UnixWare st01 Driver Supplement tapecntl -c n /dev/rmt/ntape1 (where n is 0 to 3) does not control compression. SLS ptf7610b only adds the option to quietly install the ptf. 13. SLS ptf7611b: UnixWare pdimkdev Supplement sdighost -r [see sdighost(1M)] does not permanently remove the names of disks in the sdi database that do not correspond to any physical disks on the system ('phantom' disks). With SLS ptf7611a, sdighost -r removes unwanted "vtoc" entries from the resmgr database and updates the configuration regardless of whether there are any ghost disks or not. 14. SLS ptf7612c: UnixWare 7.1.1 ping Supplement This SLS addresses these problems found with the ping command on UnixWare 7.1.1: A. The ping command hangs and never times out. This occurs when alarm(1) becomes stuck in a loop of retries. B. ping -f fails. C. ping -s and ping -i will ping continuously. D. In an MPIO environment the system appears to hang when one path to a disk fails. 15. SLS ptf7613b: UnixWare 7.1.1 iasy Driver Unable to login from a serial terminal using an xon/xoff cable SLS ptf7613b only adds the option to quietly install the ptf. 16. SLS ptf7614b: UnixWare 7.1.1 scodb Driver The problem where scodb cannot be loaded. /car/asm/log/osmlog displays this message: Undefined symbol db_cp_active in loadable module /etc/conf/pack.d/scodb SLS ptf7614b only adds the option to quietly install the ptf. 17. SLS ptf7619a: UnixWare 7.1.1 psM_apic Driver On machines with an 8-way Intel Sabre motherboard, interrupts are not distributed to the eighth processor. Note: This SLS is only applicable to machines with an 8-way Intel Sabre motherboard. 18. SLS ptf7621b: UnixWare 7.1.1 segdev and segshm Driver Supplement An async IO read (via aio_read(2) or lio_listio(2)) could return incorrect data due to a bug in the regular Shared Memory driver (segdev). 19. SLS ptf7623b: UnixWare 7.1.1 mpio driver Supplement This feature is known as automatic path recovery. It enhances the failover strategy of MPIO. When an error is encountered on an active path andno more alternate paths are available, MPIO will attempt to repair all the failed paths. This improves the chances of recovery because a newly repaired path may be capable of serving the I/O. SLS ptf7623b only adds the option to quietly install the ptf. 20. SLS ptf7627b: UnixWare 7.1.1 osocket driver Supplement SLS ptf7627a addresses the following: OSR5 binaries that use rpcs (svc_register) will not run on UnixWare 7 Release 7.1.1. SLS ptf7627b only adds the option to quitely install the ptf. 21. SLS ptf7631c: UnixWare 7.1.1 libsocket Supplement Libsocket maintains a pointer to libresolv code; the pointer may become stale causing a core dump of Apache when used with PHP. 22. SLS ptf7633b: UnixWare 7.1.1 libnsl and timod Driver Supplement This SLS contains modifications to address networking-related errors in the t_sync library call. A. t_listen fails with t_error TBADQLEN. In a TLI/XTI application using t_sync, the qlen value for the specified transport endpoint can be incorrectly set to zero locally within the transport library. The negotiated qlen field of a transport endpoint is now correctly updated when t_sync is called. B. Some RPC routines, upon encountering an error condition, fail to clean up partially-built connections. In the case of TCP/IP, this can result in unused endpoints remaining in the CLOSE_WAIT state. A T_DISCON_REQ is now sent down the stream to ensure orderly cleanup. C. NIS name resolution can cause memory corruption for some applications, often resulting in random core dumps generated by a SIGSEGV. A memory allocation for an internal structure was increased to the correct size. 23. SLS ptf7634b: UnixWare 7.0.0 - 7.1.1 mtrr Driver Supplement This SLS addresses the problem of a memory leak in the mttr driver which may cause misprogrammed MTTRs or other problems. 24. SLS ptf7635b: UnixWare 7.0.0 - 7.1.1 char Driver Supplement An incorrect KMA buffer reference to km_type occurs after kernel memory has been freed. 25. SLS ptf7636a: UnixWare 7.1.0 and 7.1.1 vol Driver Supplement Panics can occur due to vxfs not checking for vnode softcounts before performing kmem_free() on the vx_bignode. 26. SLS ptf7637b: UnixWare 7.1.0 and 7.1.1 Processorfs Driver Supplement A. Incorrect KMA buffer references occur after kernel memory has been freed in processorfs. B. SLS ptf7637b additionally fixes:- - fz511891 erg501411 After installation of ptf7637b you get a PANIC 27. SLS ptf7638a: UnixWare 7.1.1 memfs Driver Supplement Incorrect KMA buffer references occur after kernel memory has been freed in memfs. 28. SLS ptf7639a: UnixWare 7.0.1, 7.1.0 and 7.1.1 vxvm-startup Supplement On boot-up, volumes in Vxvm disk groups other than rootdg may not start. These volumes will show up in vxprint as DISABLED. 29. SLS ptf7640a: UnixWare 7.0.1, 7.1.0 and 7.1.1 vxslicer Supplement The root disk cannot be encapsulated due to a gap between dump and swap. The gap is created during a previous attempt to encapsulate the root drive, and results from the use of swap as a temporary working area during encapsulation. If the swap is large enough, the end track lies beyond the 2GB limit from the start of the encapsulation area. Since vxslicer is not large file capable, this can result in transfer errors. 30. SLS ptf7644a: UnixWare 7.1.1 intmap, ldterm and ptem Supplement Screen corruption occurs when running an application over a telnet(1tcp) or rlogin(1tcp) session on multiprocessor systems. 31. SLS ptf7645b: UnixWare 7.1.1 ipc Driver Supplement ipcs -qa sets message queue to R but once msgrcv completed the flag is not reset.is not reset. 32. SLS ptf7649a: UnixWare 7.1.1 rtpm Supplement fz511609 erg711369 memory leak in rtpm 33. SLS ptf7652a: UnixWare 7.1.1 csh Supplement csh is supposed to support variable names as long as 20 characters, but the actual limit has been less. 34. SSE039: UnixWare 7.0.0 - 7.1.1 SU Security Patch Several security holes were found in the "su" program along with the iaf library. Without these patches, systems may be vulnerable to local security holes. 35. SSE053: UnixWare 7.0.0 - 7.1.1 System Security Enhancement Several security holes were found on UnixWare 7.0.0 through 7.1.1 in the packaging tools. 36. SSE059: UnixWare 7.1.1 SSE for rtpm Two security holes are closed with this new version of rtpm. Both could lead to local system compromises if left open. NonStop Clusters Problems Fixed ------------------------------- 1. vxdump and vxrestore try to use 64KB buffering, while the OS only supports 32KB. 2. A panic in ICSSIG_STOP() occurs. 3. swap -c needs to be clusterized. 4. sysinit.sh fails to reattach disk plexes. 5. prtvtoc fails trying to configure disk2. 6. A panic occurs in ip_q_matchfn(); userp field is NULL. 7. A previous fix causes a lock hierarchy violation in the debug kernel. 8. Improper locking in fbtok_clntinit() can lead to unkillable hung processes. 9. A screen corruption occurs in curses app over telnet. 10. NFS and RPC daemons do not always failover. 11. OSU - switchroot should not occur on shutdown to single-user (init s). 12. Booting node 2 with base (zero encoded) devices in /etc/swaptab can corrupt node 2 disk. 13. swapctl(SC_ADD,...) should add swap device at the disk server node (not give EREMOTE). 14. Route for the CVIP alias is not removed when the alias is removed. 15. init: can be killed by signals during failover: PANIC: Init died! (why = 3, what = 10). 16. The IP cluster panics in ics_mblk_bcopy() when the server returns ESRCH for the process. 17. There is an interrupt priority conflict between Regroup and CLMS. 18. When IP cluster uses cnet for ICS, the ICS link is lost if both nodes are not booted at exactly the same time. 19. Health and wellness /dev/entries are not created for all nodes. 20. Shared SCSI internal boot disk is not seen on dependent node. 21. If a bind() is done in a port in TIME_WAIT state, a bind to that port cannot be done even after it becomes free. 22. 2-node IP failover hangs; cnet-public; n100c with hub-private. 23. /dev/_tcp directory is deleted during dependent node installation. 24. A priority inversion causes a deadlock and the cluster hangs. 25. An incorrect error code is returned while doing a read from a filesystem that resides on a node that has gone down. 26. A second NIC on a dependent node cannot be configured. 27. A race condition in CFS causes a panic. 28. If snmpd dies while a client connection is open on its port 199, snmpd cannot bind to its port when it is restarted, because the the port is in FIN_WAIT state. 29. A new tuneable "in_cvipoutaddr" has been added and is configurable via the "inconfig" command. This tunable dictates whether outgoing connects will go out with a CVIP source address. By default this feature is disabled and the behavior is that a local interface address is selected for sockets that are not explicitly bound to a local address. 30. Executing "ls" on an NFS mounted file system that contains character special or block special files causes a panic. 31. Supplying Emergency Recovery commands for CNM, Dual-Hosted SCSI and FCAL systems. stand and root disk slices are restored with dd option. Instructions: Making Emergency Recovery Tape. ------------------------------- - Change run level to 1 shutdown -y -i1 -g0 OR from boot prompt [boot] INITSTATE=1 [boot] go - execute /sbin/emergency_rec -d ctape1 - select CNM, DHS or FCAL system - Remove the tape after its done. Making Emergency Recovery disk ----------------------------- The tape driver (st01) should be statically linked for emergency_disk to work. - to make the driver Static vi /etc/conf/sdevice.d/st01 insert line "$static" after line "$version" - rebuild kernel /etc/conf/bin/idbuild -B - reboot shutdown -y -i6 -g0 - Make emergency disk /sbin/emergency_disk -d / diskette1 You will need four diskettes. Label them 1/4 2/4 3/4 4/4. Restoring from Emergency Tape ---------------------------- Restoring FCAL system --------------------- Use the following procedure to restore node-1 of NonStop cluster with COMPAQ Fiber Channel Array (RA4x00) used as external storage shared between the root and the failover nodes. FCAL disks are not accessible if two array controllers are installed in redundant mode. Make sure there is only one controller is installed in the FCAL array. Pull out the second controller (connected to node-2) board from the RA4x00 box. If you have installed new FCAL disks then do the following - Boot from Compaq Smart-Start CD and configure FCAL for a logical disk volume - Boot from the emergency recovery diskette (4 of them) - Exit to unix shell - Type acu with a very large number to delete any disk configuration acu 999999999 - Next acu command will reconfigure the disks acu look for messages indicating disk is stamped etc. Ignore sidadd error message. - exit from shell. - Reboot with the emergency diskettes. - Insert tape - Select Restore Disk - Reboot the system when its done. If restoring to already configured FCAL disks then do the following - Boot with the emergency diskettes (4 of them). - Insert tape - Select Restore Disk - Reboot the system when its done. Restoring CNM or DHS systems. ------------------------------- Use the following procedure to restore node-1 of Cross Node Mirror or Dual Hosted SCSI type clusters. - Boot with the emergency diskettes (4 of them). - Insert tape - Select Restore Disk - Reboot the system when its done. If Online Software Upgrade is performed and the system is booted from the alternate root slice then emergency_rec will only save/restore the alternate root slice. The old root will not be restored by the emergency_rec procedure. 32. A problem exists with installing a dependent node with gigabit ethernet. 33. IP dependent node install doesn't handle hexadecimal netmask specification properly. 34. dlpid incorrectly re-enables n100c NIC interface after link is lost. 35. The "/sbin/dbms_guard" command is out-of-date. 36. NetBios daemon doesn't recognize CVIP. 37. "netstat -na" fails to show correct number of established connections. 38. CVIP failover takes over 50 seconds when the root node is shot. 39. keepalive doesn't honor probation period, always declares process "down" after mmaxerrors. 40. X Desktop does not fail over. 41. "swap -a" during nodeup fails if swap device is >2Gb. 42. Can't start dcu from scoadmin->hardware->dc. 43. TCP_NODELAY is not set for all connections in an IP ICS cluster. 44. TCP window sizes are set incorrectly for IP ICS ports. 45. NSC certification test 'runlevel' causes a node to panic when changing the init state. 46. /usr/lib/libc.so.1 is installed incorrectly. 47. There is no mouse driver in node2 resmgr when IP I/C is used (breaks X failover to node 2). If the mouse on node 2 doesn't work after this PTF is installed, run /etc/cluster/setup.d/cluster_dist_mouse_modules. 48. Fitz. 514367: NSC panic when running Oracle 8i. 49. Fitz. 513336 NSC set installation needs to avoid accepting leftover keys in input buffer as default option. 50. Fitz. 513421: Modem manager does not work correctly with NSC software. 51. cluster_dependent_setup script has a minor coding error. may cause SBA (split brain avoidance) config failure during Servernet Cluster dependent node install. 52. Fitz 513879: cron job to shutdown system is repeated on the cluster. 53. SCOadmin Shutdown manager on a single node issues "nodedown" not clusternode_shutdown. 54. Starting automount on an 711b cluster panics the node. 55. Marking of the primary external interface down result in inaccessablity of other IP interfaces from the cluster. 56. Panic in ICS NOTICE: ASSERTION CAUGHT: (flags & KM_NOSLEEP) || KS_HOLD0LOCKS(), file: kma.c, line: 1256 57. Default route is lost when all public interfaces are marked down. 58. Node failure during vxplex can cause auto recovery of mirrors to fail. 59. Deadlock VPROC: DEADLOCK: vpop_exit() & pvpop_reap(). 60. ttymon leaks file descriptors, causing the system to run out of pipes. 61. Failover node install problem with the new UW 7.1.1 CD's. 62. Fitz. 514885: Added a kernel tunable to run NSC in short PID mode. This feature will limit the Process ID (PID) to a 16 bit number. The 16 bit PID mode is required to run some old OSR5 binaries on NSC. A new tunable 'USE_SHORT_PID' is added to enable the short PID mode. The default value for this tunable is 0, in which case the long PID mode (32 bit PID) is used. In the long PID mode 16 MSBs are used for the node number and the remaining 16 LSBs are used for a unique PID within the node. If USE_SHORT_PID is set to 1 then 16-bit, or short PID, mode is used. In this mode the 3 MSBs are used for the node number and the remaining 13 LSBs are used for a unique process id within the node. Since only 3 bits are available for the node number, the short PID feature will limit the maximum number of NSC nodes to 7. Also, the maximum PID per node will be limited to 8192 (13 bits). To enable the short PID mode, set USE_SHORT_PID to 1 # /etc/conf/bin/idtune USE_SHORT_PID 1 build the kernel # /etc/conf/bin/idbuild -B and then reboot # shutdown -y -i6 -g0 63. Fitz. 517183: TCP connection with port numbers 3000 to 3012 fails. 64. Fitz. 518188: osr5 binary fails in short PID mode with F_O_GETLK error. 65. Fitz. 515550: Excessive system swapping when freemem > lotsfree. The swapping starts due to various assumptions in vmmeter and schedpaging. 66. Fitz. 516802: Excessive warning messages "WARNING: dnet_in_tcp_pcbbind: found conflicting pcb!" filling up the root file system. 67. "PANIC:sba_asy_send()-needs to check return from allocb()." 68. Fitz. 711480: TCP client can flood the server with lots data resulting in a large amount of 2K buffers being allocated and eventually exhausting the kernel VM and severely degrading system performance. Also, TCP flow control will "ack" a data byte probe even when its flow-control window is zero. 69. Fitz. 517679: Opening more than 28 file descriptors before running rsh causes a hang. 70. Added support for Compaq 71xx series GB ethernet NICS during IP cluster dependent node install. 71. Fitz. 515735: NSC license registration is not fully clusterized, disabled warning messages "NOTICE: Unregistered SCO software is installed on your system. 72. Fitz. 515680: A bad call to cluster_detailed transition which can block forever. 73. culd daemon fails to migrate after root node failover. 74. Fitz. 711680: In some rare situations nfs lockd demon may never exit grace period; this causes kernel lock manager requests not to be processed. 75. Fitz. 517029: nerd_bcache process consumes a large amount of CPU time. 76. Fitz. 517660: VxFS file system snapshot creates problem. 77. Fitz. 518243: Panic occurs at ics_mblk_copy+18. File Contents of this Maintenance Pack: --------------------------------------- /etc/cluster/inst.d/locale/C/menus/cluster_dependent_disksetup/funcs /etc/cluster/inst.d/locale/C/menus/cluster_dependent_disksetup/globals /etc/cluster/setup.d/cfglib /etc/cluster/setup.d/cluster_dependent_setup /etc/cluster/setup.d/cluster_dist_mouse_modules /etc/cluster/setup.d/cluster_failover_sb_config /etc/cluster/setup.d/cluster_get_mouse_modules /etc/cluster/setup.d/cluster_get_proots /etc/cluster/setup.d/cluster_try_resmgr_module /etc/cluster/setup.d/ipcluster_dependent_setup /etc/conf/autotune.d/kernel /etc/conf/autotune.d/nsc_cfs /etc/conf/bin/idcheck /etc/conf/bin/idinstall /etc/conf/bin/idmknod /etc/conf/bin/idmodreg /etc/conf/bin/idresadd /etc/conf/dtune.d/nsc /etc/conf/dtune.d/nsc_cfs /etc/conf/dtune.d/nsc_clms /etc/conf/dtune.d/nsc_cprocfs /etc/conf/dtune.d/nsc_dnet /etc/conf/dtune.d/nsc_dswap /etc/conf/dtune.d/nsc_glue /etc/conf/dtune.d/nsc_icstcp /etc/conf/dtune.d/nsc_nerd /etc/conf/dtune.d/svc /etc/conf/init.d/kernel /etc/conf/mdevice.d/nsc /etc/conf/mdevice.d/nsc_asyncio /etc/conf/mdevice.d/nsc_cfs /etc/conf/mdevice.d/nsc_clms /etc/conf/mdevice.d/nsc_clreg /etc/conf/mdevice.d/nsc_cprocfs /etc/conf/mdevice.d/nsc_cspecfs /etc/conf/mdevice.d/nsc_cts /etc/conf/mdevice.d/nsc_cvip /etc/conf/mdevice.d/nsc_dnet /etc/conf/mdevice.d/nsc_dswap /etc/conf/mdevice.d/nsc_gentty /etc/conf/mdevice.d/nsc_glue /etc/conf/mdevice.d/nsc_ics /etc/conf/mdevice.d/nsc_ipc /etc/conf/mdevice.d/nsc_nerd /etc/conf/mdevice.d/nsc_rcopy /etc/conf/mdevice.d/nsc_rdev /etc/conf/mdevice.d/nsc_rioc /etc/conf/mdevice.d/nsc_tserv /etc/conf/mdevice.d/nsc_tservloop /etc/conf/mdevice.d/procfs /etc/conf/mdevice.d/ts /etc/conf/mdevice.d/vproc /etc/conf/mtune.d/nsc /etc/conf/mtune.d/nsc_cfs /etc/conf/mtune.d/nsc_clms /etc/conf/mtune.d/nsc_cprocfs /etc/conf/mtune.d/nsc_dnet /etc/conf/mtune.d/nsc_dswap /etc/conf/mtune.d/nsc_glue /etc/conf/mtune.d/nsc_icstcp /etc/conf/mtune.d/nsc_nerd /etc/conf/mtune.d/svc /etc/conf/mtune.d/sysdump /etc/conf/node.d/nsc_cts /etc/conf/node.d/nsc_cvip /etc/conf/node.d/nsc_gentty /etc/conf/node.d/nsc_glue /etc/conf/node.d/sysmsg /etc/conf/pack.d/asyc/Driver_mp.o /etc/conf/pack.d/audit/Driver_mp.o /etc/conf/pack.d/audit/stubs.c /etc/conf/pack.d/bios/Driver_mp.o /etc/conf/pack.d/ca/Driver_mp.o /etc/conf/pack.d/char/Driver_mp.o /etc/conf/pack.d/fc/Driver_mp.o /etc/conf/pack.d/fdfs/Driver_mp.o /etc/conf/pack.d/fifofs/Driver_mp.o /etc/conf/pack.d/fp/Driver_mp.o /etc/conf/pack.d/fs/Driver_mp.o /etc/conf/pack.d/inet/Driver_mp.o /etc/conf/pack.d/inet/space.c /etc/conf/pack.d/intmap/Driver_mp.o /etc/conf/pack.d/io/Driver_mp.o /etc/conf/pack.d/ipc/Driver_mp.o /etc/conf/pack.d/kernel/Driver_mp.o /etc/conf/pack.d/krpc/Driver_mp.o /etc/conf/pack.d/ldterm/Driver_mp.o /etc/conf/pack.d/log/Driver_mp.o /etc/conf/pack.d/mem/Driver_mp.o /etc/conf/pack.d/memfs/Driver_mp.o /etc/conf/pack.d/mtrr/Driver.o /etc/conf/pack.d/mtrr/Driver_mp.o /etc/conf/pack.d/namefs/Driver_mp.o /etc/conf/pack.d/nfs/Driver_mp.o /etc/conf/pack.d/nfs/space.c /etc/conf/pack.d/nsc/Driver_mp.o /etc/conf/pack.d/nsc/space.c /etc/conf/pack.d/nsc_asyncio/Driver_mp.o /etc/conf/pack.d/nsc_cfs/Driver_mp.o /etc/conf/pack.d/nsc_cfs/space.c /etc/conf/pack.d/nsc_clms/Driver_mp.o /etc/conf/pack.d/nsc_clms/space.c /etc/conf/pack.d/nsc_clreg/Driver_mp.o /etc/conf/pack.d/nsc_cprocfs/Driver_mp.o /etc/conf/pack.d/nsc_cprocfs/space.c /etc/conf/pack.d/nsc_cspecfs/Driver_mp.o /etc/conf/pack.d/nsc_cts/Driver_mp.o /etc/conf/pack.d/nsc_cvip/Driver_mp.o /etc/conf/pack.d/nsc_dnet/Driver_mp.o /etc/conf/pack.d/nsc_dnet/space.c /etc/conf/pack.d/nsc_dswap/Driver_mp.o /etc/conf/pack.d/nsc_dswap/space.c /etc/conf/pack.d/nsc_gentty/Driver_mp.o /etc/conf/pack.d/nsc_glue/Driver_mp.o /etc/conf/pack.d/nsc_glue/space.c /etc/conf/pack.d/nsc_ics/Driver_mp.o /etc/conf/pack.d/nsc_icslti/Driver.o /etc/conf/pack.d/nsc_icslti/Driver_mp.o /etc/conf/pack.d/nsc_icstcp/Driver.o /etc/conf/pack.d/nsc_icstcp/Driver_mp.o /etc/conf/pack.d/nsc_icstcp/space.c /etc/conf/pack.d/nsc_icsutil/Driver_mp.o /etc/conf/pack.d/nsc_ipc/Driver_mp.o /etc/conf/pack.d/nsc_nerd/Driver_mp.o /etc/conf/pack.d/nsc_nerd/space.c /etc/conf/pack.d/nsc_rcopy/Driver_mp.o /etc/conf/pack.d/nsc_rdev/Driver_mp.o /etc/conf/pack.d/nsc_rgp/Driver_mp.o /etc/conf/pack.d/nsc_rioc/Driver_mp.o /etc/conf/pack.d/nsc_tserv/Driver_mp.o /etc/conf/pack.d/nsc_tservloop/Driver_mp.o /etc/conf/pack.d/postwait/Driver_mp.o /etc/conf/pack.d/proc/Driver_mp.o /etc/conf/pack.d/processorfs/Driver_mp.o /etc/conf/pack.d/procfs/Driver_mp.o /etc/conf/pack.d/psm_apic/Driver_mp.o /etc/conf/pack.d/ptem/Driver_mp.o /etc/conf/pack.d/resmgr/Driver_mp.o /etc/conf/pack.d/rlogin/Driver_mp.o /etc/conf/pack.d/sad/Driver_mp.o /etc/conf/pack.d/sd01/Driver.o /etc/conf/pack.d/sd01/Driver_mp.o /etc/conf/pack.d/segshm/Driver_mp.o /etc/conf/pack.d/segshm/space.c /etc/conf/pack.d/socksys/Driver_mp.o /etc/conf/pack.d/sp/Driver_mp.o /etc/conf/pack.d/specfs/Driver.o /etc/conf/pack.d/specfs/Driver_mp.o /etc/conf/pack.d/spew/Driver_mp.o /etc/conf/pack.d/st01/Driver.o /etc/conf/pack.d/st01/Driver_mp.o /etc/conf/pack.d/svc/Driver_mp.o /etc/conf/pack.d/svc/space.c /etc/conf/pack.d/sysdump/Driver_mp.o /etc/conf/pack.d/sysdump/space.c /etc/conf/pack.d/sysdump/stubs.c /etc/conf/pack.d/sysmsg/Driver_mp.o /etc/conf/pack.d/timod/Driver_mp.o /etc/conf/pack.d/tpath/Driver_mp.o /etc/conf/pack.d/ts/Driver_mp.o /etc/conf/pack.d/ttcompat/Driver_mp.o /etc/conf/pack.d/udev/Driver_mp.o /etc/conf/pack.d/uni/Driver_mp.o /etc/conf/pack.d/util/Driver_mp.o /etc/conf/pack.d/vol/Driver.o /etc/conf/pack.d/vol/Driver_mp.o /etc/conf/pack.d/voles/Driver.o /etc/conf/pack.d/voles/Driver_mp.o /etc/conf/pack.d/volspec/Driver.o /etc/conf/pack.d/volspec/Driver_mp.o /etc/conf/pack.d/vproc/Driver_mp.o /etc/conf/pack.d/vxfs/Driver_mp.o /etc/conf/pack.d/xpoll/Driver_mp.o /etc/conf/pack.d/xque/Driver_mp.o /etc/conf/sdevice.d/ipc /etc/conf/sdevice.d/nsc /etc/conf/sdevice.d/nsc_asyncio /etc/conf/sdevice.d/nsc_cfs /etc/conf/sdevice.d/nsc_clms /etc/conf/sdevice.d/nsc_clreg /etc/conf/sdevice.d/nsc_cprocfs /etc/conf/sdevice.d/nsc_cspecfs /etc/conf/sdevice.d/nsc_cts /etc/conf/sdevice.d/nsc_cvip /etc/conf/sdevice.d/nsc_dnet /etc/conf/sdevice.d/nsc_dswap /etc/conf/sdevice.d/nsc_gentty /etc/conf/sdevice.d/nsc_glue /etc/conf/sdevice.d/nsc_ics /etc/conf/sdevice.d/nsc_ipc /etc/conf/sdevice.d/nsc_nerd /etc/conf/sdevice.d/nsc_rcopy /etc/conf/sdevice.d/nsc_rdev /etc/conf/sdevice.d/nsc_rioc /etc/conf/sdevice.d/nsc_tserv /etc/conf/sdevice.d/nsc_tservloop /etc/conf/sdevice.d/ts /etc/conf/sdevice.d/vproc /etc/fs/vxfs/mount /etc/inet/inet.dfl /etc/init.d/MOUNTFSYS /etc/init.d/dtlogin /etc/init.d/mse /etc/init.d/nd /etc/init.d/vxvm-startup /etc/keepalive.d/dtlogin_restart /etc/keepalive.d/mse_restart /etc/rc2.d/S42ls /etc/scsi/dfocfg /etc/scsi/pdi_timeout /etc/spawndaemon.d/ka_dtlogin /etc/spawndaemon.d/ka_ifmond /etc/spawndaemon.d/ka_mse /etc/ttydefs /sbin/dbms_guard /sbin/dumpcheck /sbin/dumpsave /sbin/emergency_rec /sbin/fixroot /sbin/keepalive /sbin/onall /sbin/rc0 /sbin/shutdown /sbin/su /sbin/swap /sbin/wsinit /u95/bin/sh /usr/bin/csh /usr/bin/netstat /usr/bin/onnode /usr/bin/passwd /usr/bin/pkginfo /usr/bin/pkgparam /usr/bin/pkgtrans /usr/bin/su /usr/ccs/lib/libc.a /usr/ccs/lib/libc.so /usr/ccs/lib/libp/libc.a /usr/ccs/lib/libp/libc.so /usr/ccs/lib/libp/libc.so.1 /usr/include/sys/disp.h /usr/include/sys/file.h /usr/include/sys/timod.h /usr/lib/drf/acu /usr/lib/drf/changes.boot2diskette.tar /usr/lib/drf/conframdfs /usr/lib/drf/copyPart /usr/lib/drf/cut_flop /usr/lib/drf/drf_inst.gen /usr/lib/drf/drfram.proto /usr/lib/drf/emrec_cfglib /usr/lib/drf/get_emrec_disk /usr/lib/drf/netinst1.img.UW711 /usr/lib/drf/netinst2.img.UW711 /usr/lib/iaf/cr1/scheme /usr/lib/iaf/in.login/scheme /usr/lib/iaf/login/scheme /usr/lib/libc.so.1 /usr/lib/libiaf.so /usr/lib/libnsl_i.so /usr/lib/libsnmpio.a /usr/lib/libsnmpio.so /usr/lib/libthread.so /usr/lib/libthread.so.1 /usr/lib/libthreadT.so /usr/lib/libthreadT.so.1 /usr/lib/libvxvm.a /usr/lib/locale/C/LC_MESSAGES/NSC /usr/lib/nfs/lockd /usr/lib/netcfg/bin/ndcfg /usr/lib/saf/ttymon /usr/lib/scoadmin/common/nodelaunch /usr/lib/scoadmin/modem/modemGUI /usr/lib/scoadmin/netosa/serialOsa /usr/lib/scoadmin/netosa/uucpOsa /usr/lib/scoadmin/serial/serialGUI /usr/lib/scoadmin/shutdown/shutdownGUI /usr/lib/scoadmin/shutdown/shutdownUtil /usr/lib/vxvm/bin/vxslicer /usr/sbin/crash /usr/sbin/cron /usr/sbin/culd /usr/sbin/dlpid /usr/sbin/dumpmemory /usr/sbin/ifconfig /usr/sbin/ifmond /usr/sbin/ifor_pmd /usr/sbin/in.nbd /usr/sbin/in.snmpd /usr/sbin/initialize /usr/sbin/ldsysdump /usr/sbin/ping /usr/sbin/pkgadd /usr/sbin/pkgcat /usr/sbin/pkginstall /usr/sbin/rpcbind /usr/sbin/rtpm /usr/sbin/shutdown /usr/sbin/trpt /var/spool/dist/nsccdi.package