Solaris Common Messages and Troubleshooting Guide

Previous: Chapter 1 About Error Messages

Chapter 2 Alphabetical Message Listing

Messages listed here are in the current Solaris 7 system. Many were also present in earlier Solaris 2 systems.

Messages are listed alphabetically.

Numbers and Symbols

`29a00 illegal instruction`

Cause

When trying to boot a client from a boot/jumpstart server to install/upgrade a workstation, it fails with the following message:

boot net - install
Rebooting with command: net - install
Boot device: /iommu/sbus/ledma@f, 400010/le@f, 8c0000 File and args: -
install
29a00  Illegal Instruction
(0) ok

Action

The problem lies in the /tftpboot directory of the boot server. Confirm that the HOSTID and HOSTID.ARCH files are linked to the correct inetboot file for your architecture. The following is an example of how a symbolic link should look:

# cd /tftpboot
# ls -l 81971904*
81971904 -> inetboot.sun4m.Solaris_2.4
81971904.SUN4M -> inetboot.sun4m.Solaris_2.4

The following is an example of an incorrect entry for a sun4m system:

C753002F -> inetboot.axil4m.Solaris_2.5.1
C753002F.AXIL4M -> inetboot.axil4m.Solaris_2.5.1

If the entries are not correct, remove the entry for the particular client in this directory, using rm_install_client or rm_client commands, and re-add the client with the add_install_client(1M) or add_client command or through Solstice giving the correct architecture.

`451 timeout waiting for input during` `source`

Cause

When sendmail(1M) reads from anything that might time out, such as an SMTP connection, it sets a timer to the value of the r processing option before reading begins. If the read doesn't complete before the timer expires, this message appears and reading stops. (Usually this is during RCPT.) The mail message is then queued for later delivery.

Action

If you see this message often, increase the value of the r processing option in the /etc/mail/sendmail.cf file. If the timer is already set to a large number, look for hardware problems such as poor network cabling or connections.

`550 hostname... Host unknown`

Cause

This sendmail(1M) message indicates that the destination host machine, specified by the address portion after the @ (at-sign), was not found during DNS (Domain Naming System) lookup.

Action

Use the nslookup(1M) command to verify that the destination host exists in that or other domains, perhaps with a slightly different spelling. Failing that, contact the intended recipient and ask for a proper address.

Sometimes this return message indicates that the intended host is down, rather than unknown. If a DNS record contains an unknown alternate host, and the primary host is down, sendmail(1M) returns a "Host unknown" message from the alternate host.¤

For uucp(1C) mail addresses, the "Host unknown" message probably means that the destination hostname is not listed in the /etc/uucp/Systems file.

Technical Notes

¤ This is a known sendmail(1M) version 8.6.7 bug.

`550 username... User unknown`

Cause

This sendmail(1M) message indicates that the intended recipient, specified by the address portion before the @ (at-sign), could not be located on the destination host machine.

Action

Check the e-mail address and try again, perhaps with a slightly different spelling. If this doesn't work, contact the intended recipient and ask for a proper address.

`554 hostname... Local configuration error`

Cause

This sendmail(1M) message usually indicates that the local host is trying to send mail to itself.

Action

Check the value of the $j macro in the /etc/mail/sendmail.cf file to ensure that this value is a fully-qualified domain name.

Technical Notes

When the sending system provides its hostname to the receiving system (in the SMTP HELO command), the receiving system compares its name to the sender's name. If these are the same, the receiving system issues this error message and closes the connection. The name provided in the HELO command is the value of the $j macro.

"A"

Action

The easiest solution is to reduce the size of the parent process environment by unsetting extraneous environment variables. (See the man page for the shell you're using to find out how to list and change your environment variables.) Then run the program again.

Technical Notes

An argument list longer than ARG_MAX bytes was presented to a member of the exec(2) family of system calls.

The symbolic name for this error is E2BIG, errno=7.

`Argument out of domain`

Cause

This is a programming error or a data input error.

Action

Ask the program's author to fix this condition, or supply data in a different format.

Technical Notes

This indicates an attempt to evaluate a mathematical programming function at a point where its value is not defined. The argument of a programming function in the math package (3M) is out of the domain of the function. This could happen when taking the square root, power, or log of a negative number, when computing a power to a non-integer, or when passing an out-of-range argument to a hyperbolic programming function.

To help pinpoint a program's math errors, use the matherr(3M) facility.

The symbolic name for this error is EDOM, errno=33.

`Arguments too long`

Cause

This C shell error message indicates that there are too many arguments after a command. For example, this can happen by invoking rm * in a huge directory. The C shell cannot handle more than 1706 arguments.

Action

Temporarily start a Bourne shell with sh(1) and run the command again. The Bourne shell dynamically allocates command line arguments. Return to your original shell by typing exit.

`assertion failed: string, file name, line int`

Cause

An unexoected condition in the program has happened.

Action

Contact the vendor or author of the program to ask why it failed. If you have the source code for the program, you can look at the file and line number where the assertion failed. This might give you an idea of how to run the program differently.

Technical Notes

Ensure that the mount point is a directory and not a regular file or a symbolic link.

"B"

`Bad address`

Cause

The system encountered a hardware fault in attempting to access a parameter of a programming function.

Action

Check the address to see if it resulted from supplying the wrong device or option to a command. If that is not the problem, contact the vendor or author of the program for an update.

Technical Notes

This error could occur any time a function that takes a pointer argument is passed an invalid address. Because processors differ in their ability to detect bad addresses, on some architectures passing bad addresses can result in undefined behaviors.

The symbolic name for this error is EFAULT, errno=14.

`BAD/DUP FILE I=i OWNER=o MODE=m SIZE=s MTIME=t CLEAR?`

Cause

While checking inode link counts during phase 4, fsck(1M) found a file (or directory) that either does not exist or exists somewhere else.

Action

To clear the inode of its reference to this file or directory, answer yes. With the -p (preen) option, fsck(1M) automatically clears bad or duplicate file references, so answering yes to this question seldom causes a problem.

`Bad file number`

Cause

Generally this is a program error, not a usage error.

Action

Contact the vendor or author of the program for an update.

Technical Notes

Either a file descriptor refers to no open file, or a read(2) (or write(2)) request is made to a file that is open only for writing (or reading).

The symbolic name for this error is EBADF, errno=9.

`block no. BAD I=inode no.`

Cause

Upon detecting an out-of-range block, fsck(1M) prints the bad block number and its containing inode (after I=).

Action

In fsck(1M) phases 2 and 4, you decide whether or not to clear these bad blocks. Before committing to repair with fsck(1M), you could determine which file contains this inode by passing the inode number to the ncheck(1M) command:

# ncheck -iinum filesystem

`BAD_MESSAGE (error code 100) from X.400`

Cause

X.400 software was working fine. Suddenly message exchanges failed in ma_start_delivery(). It was returning an error code of 100 (BAD_MESSAGE).

ma_start_delivery() call fails when trying to exhange file of more than 900 bytes.

Action

Action

The most common cause of this error is overlapping disk partitions. Do not immediately rerun fsck(1M) as suggested by the lines that display after the error message. First make sure that you have a recent backup of the filesystem involved; if not, try to back up the filesystem now using ufsdump(1M). Then run the format(1M) command, select the disk involved, and print out the partition information.

# format
: N
> partition
> print

Note whether the overlap occurs at the beginning or end of the filesystem involved. Then run newfs(1M) with the -N option to print out the filesystem parameters, including the location of backup super-blocks.

# newfs -N /dev/dsk/device

Select a super-block from a non-overlapping area of the disk, but note that in most cases you have only one chance to select the proper replacement super-block, which fsck(1M) soon propagates to all the cylinders. If you select the wrong replacement super-block, data corruption will probably occur, and you will have to restore from backup tapes. After you select a new super-block, provide fsck(1M) with the new master super-block number:

# fsck -o b=NNNN /dev/dsk/device

Technical Notes

Specific reasons for a damaged super-block include: a wrong magic number, out of range NCG (number of cylinder groups) or CPG (cylinders per group), the wrong number of cylinders, a preposterously large super-block size, and trashed values in super-block. These reasons are generally not meaningful because a corrupt super-block is usually extremely corrupt.

`BAD TRAP`

`Broken pipe`

Cause

This condition is often normal, and the message is merely informational (as when piping many lines to the head(1) program). The condition occurs when a write on a pipe does not find a reading process. This usually generates a signal to the executing program, but this message displays when the program ignores the signal.

Action

Check the process at the end of the pipe to see why it exited.

Technical Notes

The symbolic name of this error is EPIPE, errno=32.

`Bus Error`

Cause

A process has received a signal indicating that it attempted to perform I/O to a device that is restricted or that does not exist. This message is usually accompanied by a core dump, except on read-only filesystems.

Action

Use a debugger to examine the core file and determine what program fault or system problem led to the bus error. If possible, check the program's output files for data corruption that might have occurred before the bus error.

Technical Notes

Bus errors can result from either a programming error or device corruption on your system. Some common causes of bus errors are: invalid file descriptors, unreasonable I/O requests, bad memory allocation, misaligned data structures, compiler bugs, and corrupt boot blocks.

"C"

`cannot change passwd, not correct passwd`

Cause

While running yppasswd(1) and trying to change a user's password, it responded with the message cannot change passwd, not correct passwd.

Also getting yppasswd user string does not exist on server console, but running ypcat passwd | grep (user) returns the username. Verified that yppasswdd(1M) is running.

Action

Check the passwd(4) file with pwck(1M) and verify that yppasswdd(1M) is running on the right server. Then verify where the passwd(4) file is located and if changed, check that yppasswdd(1M) has the location in the process line. The password located in /etc/yp should read /usr/lib/yp/rpc.yppasswdd -D /etc/yp. The -D option with the passwd files directory location tells yppasswdd(1M) where to update and verify password changes.

`Cannot exec a shared library directly`

Cause

Attempting to exec(2) a shared library directly.

Technical Notes

The symbolic name for this error is ELIBEXEC, errno=87.

`Cannot find SERVER hostname in network database`

Cause

A brief description: the user is on a different subnet and is running permanent licenses:

ultra1(50)% cc -o hello hello.c
License Error : Cannot find the license server (fry)
in the network database for product(Sun WorkShop Compiler C)
Cannot find SERVER hostname in network database (-14,7)
cc: acomp failed for hello.c
ultra1(51)%

Action

First, make sure that the server is up and running. Second, make sure that the server is in the /etc/hosts file of the client system by typing: ping server name. Third, check the license daemon on teh server to see if it is running. Fourth, make sure that there is an elementary license file on the client:

cd /etc/opt/licenses
more sunpro.loc

Fifth, in the sunpro.loc directory are there only text license files such as sunpro.lic,1? Sixth, on the client check:

 % cd /etc
 % more nsswitch.conf | grep hosts
 hosts:      nis [NOTFOUND=return] files

This says that it will use the nis server to look up the IP address. If it is set first for nis and the /etc/hosts file has the server listed by name, change the line to

hosts:      files nis

Then see if it can find it. If not, try truss and snoop to see what is going on.

`Cannot open FCC file`

Cause

When trying to send mail via Netscape, the message is displayed. The problem is that Netscape is trying to save the outbound message to a file that has been specified by the user, but does not exist.

Action

To correct this problem: go to options Mail and News Preferences, then go to Compose. A template pops up. There is a section that specifies where to save out-going mail and news files. Make sure that these files exist or remove them from the template, if you do not care about logging what messages are sent via Netscape.

`Cannot send after transport endpoint shutdown`

Cause

A request to send data was disallowed because the transport endpoint has already been shut down.

Technical Notes

The symbolic name for this error is ESHUTDOWN, errno=143.

`Can't create public message device (Device busy)`

Cause

This message comes from the lp(1) print scheduler, indicating that it is either extremely busy or hung.

Action

If print jobs are coming out of the printer in question, wait until they are finished and then resubmit this print job. If you see this message again, the lp(1) system is probably hung.

See the message "lp hang" for a procedure to clear the queue.

Technical Notes

If lp(1) is unable to create a device for printer messages, the message FIFO could be already in use, or locked by another print job.

`Can't invoke /etc/init, error int`

Cause

This message can appear while a system is booting, indicating that the init(1M) program is missing or corrupted. Note that /etc/init is a symbolic link to /sbin/init.

Action

Boot the miniroot so you can replace init(1M). Halt the machine by typing Stop-A or by pressing the reset button. Reboot single-user from CD-ROM, the net, or diskette. For example, type boot cdrom -s at the ok prompt to boot from CD-ROM. After the system comes up and gives you a # prompt, mount the device corresponding to the original / partition somewhere, with a command similar to the mount(1M) command below. Then copy the init(1M) program from the miniroot to the original / partition, and reboot the system.

# mount /dev/dsk/c0t3d0s0 /mnt
# cp /sbin/init /mnt/sbin/init
# reboot

If this doesn't work, other files might be corrupted, and you might need to reinstall the entire system.

Technical Notes

The error number is 2 if /sbin/init is missing, or 8 if /sbin/init has an incorrect executable format. This is usually followed by a "panic: icode" message. The system tries to reboot itself, but goes into a loop, because rebooting is impossible without init(1M).

`can't open /dev/rdsk/string: (null): UNEXPECTED INCONSISTENCY`

The symbolic name for this error is ETIMEDOUT, errno=145.

`console login: ^J^M^Q^K^K^P`

Cause

This usually occurs because OpenWindows exited abnormally, leaving the system's keyboard in the wrong mode. The characters that appear when someone attempts to login are garbage transliterations of what someone types.

Action

On a SPARC system: find another machine and remote login to this system, then run this command:

$ /usr/openwin/bin/kbd_mode -a

This puts the console back into ASCII mode. Note that kbd_mode is not a windows program, it just fixes the console mode.

On an x86 system: log in remotely and start, then kill the X server, or reboot the system.

Technical Notes

The usual reason for this problem occurring is an automated script run from cron(1M) that clears out the /tmp directory every so often. Ensure that any such scripts do not remove the /tmp/.X11-pipe or /tmp/.X11-unix directories, or any files in them.

`core dumped`

Cause

A core(4) file contains an image of memory at the time of software failure, and is used by programmers to find the reason for the failure.

Action

To see which program produced a core(4) file, run either the file(1) command or the adb(1)(1) command. The following examples show the output of the file(1) and adb(1) commands on a core file from the dtmail program.

$ file core
core: ELF 32-bit MSB core file SPARC Version 1, from `dtmail'

$ adb core
core file = core -- program `dtmail'
SIGSEGV  11: segmentation violation
^D      (use Control-d to quit the program)

Ask the vendor or author of this program for a debugged version.

Technical Notes

Some signals, such as SIGQUIT, SIGBUS, and SIGSEGV, produce a core dump. See the signal(5) man page for a complete list.

If you have the source code for the program, you can try compiling it with cc -g, and debugging it yourself using dbx or a similar debugger. The where directive of dbx provides a stack trace.

On mixed networks, it can be difficult to discern which machine architecture produced a particular core dump, since adb(1) on one type of system generally cannot read a core(4) file from another type of system, and will produce an "unrecognized file" message. Run adb(1) on various machine architectures until you find the right one.

The term "core" is archaic-- ferrite core memory was supplanted by silicon RAM in the 1970s, although spaceships still employ core memory for its imperviousness to radiation.

`corrupt label - wrong magic number` or `corrupt label - label checksum failed`

Cause

After a power cycle, the machine comes up with error messages saying: corrupt label - label checksum failed or corrupt label - wrong magic number . format(1M) showed:

  0 unassigned    wm       0               0         (0/0/0)          0
  1 unassigned    wm       0               0         (0/0/0)          0
  2     backup    wm       0 - 5460        4.2G    (5460/0/0)   4154160
  3 unassigned    wm       0               0         (0/0/0)          0
  4 unassigned    wm       0               0         (0/0/0)          0
  5 unassigned    wm       0               0         (0/0/0)          0
  6 unassigned    wm       0 - 2730       2.1G       (0/0/0)          0
  7 unassigned 	  wm       2730-5460      2.1G       (0/0/0)          0

The disks were using raw partitions beginning at block 0 (cylinder 0).

The disk label (VTOC) is kept on the block 0 of cylinder 0. The label will eventually get overwritten by database programs using raw partitions if the raw partition begins at cylinder 0. (Unix filesystems avoid this area of the partition.)

Action

The workaround is to go into format and get the backup label using the backup command. Relabel the disk using this backup label. You should then be able to access the disk.

Backup the data on this disk.

Go back to the disk and relabel it, starting the raw partition at cylinder 1 (This looses one cylinder, but prevents corrupting the VTOC).

Label again.

Restore the data from your backup.

`could not grant slave pty`

Cause

User gets the error message: could not grant slave pty when attempting a telnet(1), rlogin(1), or rsh(1) session (anything that requires a shell) or when trying to bring up an x-term.

Action

The user's file permissions were set wrong on /usr/lib/pt_chmod. The user had:

# ls -la /usr/lib/pt_chmod
---s--x--x   1 bin     bin         3120 May  3  1996

The permissions should be:

# ls -la /usr/lib/pt_chmod
---s--x--x   1 root     bin         3120 May  3  1996

Note that the owner should be root, user had bin as the owner. Also note that the setuid bit must be set. Once the user did a chown root pt_chmod, everything worked again.

`Could not initialize tooltalk (tt_open): TT_ERR_NOMP`

Cause

Various desktop tools display or print this message when the ttsession(1) process is not available. The TookTalk service generally tries to restart ttsession(1) if it is not running. So this error indicates that the ToolTalk service is either not installed or is not installed correctly.

Action

Verify that the ttsession(1) command exists in /usr/openwin/bin or /usr/dt/bin. If this command is not present, ToolTalk is not installed correctly. The packages constituting ToolTalk are the runtime SUNWtltk, developer support SUNWtltkd, and the manual pages SUNWtltkm. CDE ToolTalk packages have the same names with ".2" appended.

Technical Notes

The full TT_ERR_NOMP message string reads as follows: "No ttsession(1) is running, probably because tt_open(3) has not been called yet. If this is returned from tt_open(3) it means ttsession(1) could not be started, which generally means ToolTalk is not installed on the system."

`Could not open ToolTalk Channel`

Cause

  # cpio -id -H bar -I /dev/diskette0

It fails with the message.

Action

Kill /usr/sbin/vold by running /etc/init.d/volmgt stop and use the device name /dev/rfd0

`Cross-device link`

Cause

An attempt was made to make a hard link to a file on another device, such as on another filesystem.

Action

Establish a symbolic link using ln -s instead. Symbolic links are permitted across filesystem boundaries.

Technical Notes

The symbolic name for this error is EXDEV, errno=18.

"D"

`data access exception`

Cause

This message can result from running an old version of the operating system that does not support new hardware, or by running an operating system that is not configured for new hardware. It can also result from incorrectly installed DSIMMs or from a disk problem.

Action

Upgrade your operating system to a version that supports the new hardware or machine architecture. For example, upgrading a SPARCstation 2 (with sun4c kernel architecture) to a SPARCstation 20 (with sun4m kernel architecture) requires an operating system upgrade or reconfiguration.

`Data fault`

Cause

This is a kind of bad trap that usually causes a system panic. When this message appears after a bad trap message, a system text or data access fault probably occurred.¤ In the absence of a bad trap message, this message might indicate a user text or data access fault. Data loss is possible if the problem occurs other than at boot time.

Action

Make sure the machine can reboot, then check the log file /var/adm/messages for hints about what went wrong.

¤ See the message "BAD TRAP" for more information.

`Deadlock situation detected/avoided`

Cause

A programming deadlock situation was detected and avoided.

Action

If the system had not detected and avoided a deadlock, a piece of software would have hung. Run the program again. The deadlock might not reoccur.

Technical Notes

This error usually relates to file and record locking, but can also apply to mutexes, semaphores, condition variables, and read/write locks.

The symbolic name for this error is EDEADLK, errno=45.

`Destination address required`

Cause

A required address was omitted from an operation on a transport endpoint. Destination address required.

Technical Notes

The symbolic name for this error is EDESTADDRREQ, errno=96.

`/dev/fd/int: /dev/fd/int: cannot open`

Cause

setuid and setgid shell scripts refuse to run - they simply return an error message similar to /dev/fd/3: /dev/fd/3: cannot open. (The number following /dev/fd/ is not necessarily 3.) The first line of the script properly starts a shell, and the filesystem containing the script is not mounted with the nosuid option.

Running truss on the shell script reveals that a call to open(2) is failing with error number 6 (ENXIO):

open("/dev/fd/3", O_RDONLY)                     Err#6 ENXIO

Action

Setuid and setgid shell scripts use the file descriptors in /dev/fd. The contents of /dev/fd are a File Descriptor Filesystem (fdfs) on Solaris 2 and have no connection with floppy disks!

Ensure that the fdfs is mounted as /dev/fd. The following line should appear in /etc/vfstab exactly like this (with NO initial comment symbol):

fd		-		/dev/fd		fd	-	no	-

before the machine is next rebooted.

It may be possible to remount /dev/fd without rebooting by running the following as root:

# mount fd /dev/fd

If this fails the machine must be rebooted after editing /etc/vfstab as detailed above, before setuid/setgid shell scripts are available.

Some administrators, unaware of what /dev/fd is for, comment out the entry in /etc/vfstab that mounts the fdfs (File Descriptor filesystem). This may go unnoticed until an attempt is made to run a setuid or setgid shell script.

`/dev/rdsk/c0t6d0s2: No such file or directory`

Cause

When attempting to eject a CD-ROM on a Ultra 450 system, the eject cdrom command fails, displaying the error message.

This is because the CD-ROM is on controller 1 not 0. For the eject(1) command, the cdrom "nickname" equates to /dev/rdsk/c0t6d0s2. On an Ultra 450, the CD-ROM is /dev/rdsk/c1t6d0s2. So, using cdrom does not work.

Action

Use the following command instead:

# eject cdrom0

If volume manager (/usr/sbin/vold) is not running, you can use:

# eject /dev/rdsk/c1t6d0s2

Note: Make sure that the front panel of the system is unobstructed so that the CD-ROM tray is not blocked. Otherwise, the eject(1) command appears to hang since the tray is trying to open but is physically blocked.

`Device busy`

Cause

An attempt was made to mount a device that was already mounted or to unmount a device containing an active file (such as an open file, a current directory, a mount point, or a running program). This message also occurs when trying to enable accounting that is already enabled.

Action

To unmount a device containing active processes, close all the files under that mount point, quit any programs started from there, and change directories out of that hierarchy. Then try to unmount again.

Technical Notes

Mutexes, semaphores, condition variables, and read/write locks set this error condition to indicate that a lock is held.

The symbolic name for this error is EBUSY, errno=16.

`device busy`

Cause

You eject cdrom and you receive the message device busy. This could be the result of a number of problems. Below is a list of things that you can check and do to allow ejection of the cd from the device.

Action

Ensure that the current directory is not somewhere in the CD:

 % cd
 %eject cdrom

Step B: As root

# cd /etc/init.d
# ./volmgt stop
# eject cdrom

If this works

# ./volmgt start

If this does not work go to step B.

Step C: As root

# fuser /cdrom

Kill any processes that you feel that you have already terminated. A note of caution: If this is a nfs-mounted cdrom and there are other users who access this drive, make sure you know what process you are killing and why.

# ./volmgt stop
#  ps -ef | grep vold

If vold still exists kill the process.

#  eject cdrom

If this does not work:

#  cd /vol

Make sure that dev, dsk, rdsk, rmt are in the directory. If not, most possibly your /vol directory is corrupt and a reboot may be needed for proper rebuild.

Step D: The last three options are: 1) Reboot. 2) If the cd drive is external to the system, try power cycling the drive and pressing the eject button. 3) If all else fails and the cdrom is external, on the right hand side of the eject button is a small hole into which you can insert a small straight device which forces manual ejection of the caddy.

`/dev/rdsk/string: CAN'T CHECK FILE SYSTEM.`

Cause

The system cannot automatically clean (preen) this filesystem because it appears to be set up incorrectly or is having hard disk problems. This message asks that you run fsck(1M) manually, since data corruption might already have occurred.

Action

Run fsck to clean the filesystem in question. See the message "/dev/rdsk/int: UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY" for proper procedures.

`/dev/rdsk/string: UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY.`

Cause

At boot time the /etc/rcS script runs the fsck(1M) command to check the integrity of filesystems marked "fsck" in /etc/vfstab. If fsck(1M) cannot repair a filesystem automatically, it interrupts the boot procedure and produces this message. When fsck(1M) gets into this state, it cannot repair a filesystem without losing one or more files, so it wants to defer this responsibility to you, the administrator. Data corruption has probably already occurred.

Action

First run fsck -n on the filesystem, to see how many and what type of problems exist. Then run fsck(1M) again to repair the filesystem. If you have a recent backup of the filesystem, you can generally answer "y" to all the fsck(1M) questions. It's a good idea to keep a record of all problematic files and inode numbers for later reference. To run fsck(1M) yourself, specify options as recommended by the boot script. For example:

# fsck /dev/rdsk/c0t4d0s0

Usually the files lost during fsck(1M) repair are these that were created just before a crash or power outage, and they cannot be recovered. If you lose important files, you can recover them from backup tapes.

If you don't have a backup, ask an expert to run fsck(1M) for you.

`Directory not empty`

Cause

The directory operation that was attempted, such as directory removal with rmdir(1), can be performed only on an empty directory.

Action

To remove the directory, first remove all the files that it contains. A quick way to remove a non-empty directory hierarchy is with the rm -r command.

Technical Notes

The symbolic name for this error is ENOTEMPTY, errno=93.

`diskN not unique`

Cause

During boot, system displays disk0 not unique. The error happens before the kernel loads.

Action

There are more than 1 devalias entries for disk0. Use devalias at the OK prompt to see the entries.

To remove the duplicate, run the following command at the OK prompt:

nvunalias disk0

and reset the system.

`Disc quota exceeded`

Cause

The user's disk limit has been exceeded on a user filesystem, usually because a file was just created or enlarged beyond the limit. This almost always refers to a magnetic disk, and not to an optical disc. Any data created after this condition occurs will be lost.

Action

The user can delete files to bring disk usage under the limit, or the server administrator can use the edquota(1M) command to increase the user's disk limit.

Technical Notes

The symbolic name for this error is EDQUOT, errno=49.

`driver is already installed`

Cause

Added the Sunpc 4.1 package and then necessary patches (102924). When trying to run sunpc_install, got the error message. prtconf(1M) shows the driver is not attached, and modinfo(1M) displays 4 modules.

Tried to remove the package and back out the patch. Then reinstalled, but still received the same error.

Action

Sunpc had previously been installed on the system. When removing the package with the pkgrm(1M) command, not all components were removed, because pkgrm(1M) is not aware of changes made by the sunpc_install script.

To resolve this problem it is necessary to remove sections pertaining to Sunpc in the files: /etc/devlink.tab, /etc/driver_aliases, and /etc/rc2.d/S10storekernname, and then reinstall the package.

DUMP: Cannot open dump device `/dev/rdsk/c2t0d0s1': Permission denied

Cause

When using ufsdump(1M) as user sys (UID 3) on a disk drive in a SSA, the ufsdump(1M) command fails with the message.

Action

The permissions on the ssd 'instance path' for disk in an SSA are created with 600 permissions. They should be 0640 for a non-root user to be able to read them. For example,

# ls -lL /dev/rdsk/c2t0d0s1
crw-------   1 root     sys      192,241 Jul 10  1996 /dev/rdsk/c2t0d0s1

Change it so it reads:

crw-r-----   1 root     sys      192,241 Jul 10  1996 /dev/rdsk/c2t0d0s1

You might also want to add the following line

	ssd:* 0640 root sys

to the /etc/minor_perm file, so subsequently added arrays do not have the same problem.

dumptm: Cannot open `/dev/rmt/string': Device busy

Cause

During filesystem backup, the dump program cannot open the tape drive because some other process is holding it open.

Action

Find the process that has the tape drive open, and either kill(1) the process or wait for it to finish.

# ps -ef | grep /dev/rmt
# kill -9 processID

`DUP/BAD I=i OWNER=o MODE=m SIZE=s MTIME=t FILE=f REMOVE?`

The design and implementation of Solaris ISM (Intimate Shared Memory) is what caused the ENOMEM failures, from the Lotus Notes application, besause of the limit reached on the number of shared memory segments that can be attached to a particular process.

The limit occurs because all shared memory segments are attached in the Intimate Shared Memory (ISM) mode courtesy of a system variable they have set in the system file called shmsys:share_page_table.

When a shared memory segment is attached in ISM mode, the OS locks that segment into physical memory and arranges the virtual/physical address mappings such that only one copy of the mapping information is shared amongst all attaching processes. To accomplish this, the OS requires the virtual starting address of the segment be aligned on a 16 Meg (hex 0x1000000) = 16777216-bytes address boundary.

The NULL address lets the system decide what virtual address the segment should be attached at. The system also assigns addresses 0x3000000 apart unless forced to attach addresses at 0x1000000 apart.

Doing a few calculations, a sun4d could create and attach up to 220 1-Meg ISM segments and a sun4m could create and attach up to 235 1-Meg ISM segments, providing the segments were 0x1000000 apart.

Given that ISM is what causes the limit, what can we do about it?

First, the limit only gives Lotus Notes the ability to attach a total of 80 Meg of shared memory total. By increasing the segment size to 10 Meg, as Lotus has already recommended, 8 ISM segments will take care of the load previously needing 80 1-Meg segments. The load could conceivably grow to 800-meg now without running into the ISM addressing limit.

Secondly, we could have turned off the share_page_table (ISM) flag. This would give a sun4m the ability to create in excess of 3000 1-Meg segments. The problem here is that ISM does improve the performance of shared memory accesses and if your customer intends to move up to 2.5.x, ISM is required to get around another set of problems that were discovered with shared memory loads of this kind.

Thirdly, Lotus could change the Notes server so that it kept track of the attach addresses and always attached at 0x1000000 boundary addresses instead of letting the system default to the 0x3000000 address boundary. As I've already shown, this would allow a Notes server to grow to 235 segments on a sun4m.

`error 13`

Cause

Diskless client boot gave NFS mount error 13 Diskless client boot failed with error: nfs mount failed: permission denied error 13

Cause

Diskless client fails to boot, NFS error 13 Diskless client fails to boot giving the error message: nfs error 13

`error 15 initializing`

Error: Error adding OS service Solaris 2.6 sparc sun4u:
inconsistent revision, installed package SUNWpppk revision 3.0.1
does not match revision 11.6.0,REV=1997.07.15.21.46 for sparc
architecture.

This is caused by the optional Solstice PPP 3.0.1 packages from the "Solaris Server Intranet Extension" CD-ROM installed on the system.

Action

The workaround is to remove the PPP 3.0.1 packages and replace them with the PPP packages off of the Solaris 2.6 release CD-ROM. For example:

# pkgrm SUNWlicsw SUNWlit SUNWpppk SUNWpppm SUNWpppr SUNWppps SUNWpppu
:
:  {package remove info}
:
# cd /cdrom/cdrom0/s0/Solaris_2.6/Product
# pkgadd -d . SUNWapppr SUNWapppu SUNWpppk
:
:  {package add info}
:

then use adminsuite to add the OS services, which should then work without error.

NOTE: If the Solstice PPP 3.0.1 package is configured and currently in use on the system, the user should save any of the previously entered PPP configuration info so that it may be restored after the OS services have been installed. (pkgrm(1M)the 3 ppp packages installed from the 2.6 release CD, and pkgadd(1M) all of the ppp packages from the Intranet Extension cdrom again, then redo config.). If the Solstice PPP 3.0.1 package was not used on the system, there is no reason to reinstall it. Use /usr/bin/pkginfo to check the installed packages.

This is documented in Chapter 9 of the Solaris Server Intranet Extension Installation and Release Notes Solaris 2.6 manual.

`Error Host Unknown:`

Cause

On Windows95, running PC-NFSpro2.0, use ping(1M) to reach another computer on the network. ping(1M) returns Host Unknown.

Cause: Name services are not set up correctly.

Action

1: Click the Windows 95 Start button, point to Programs, point to PC-NFSpro, then click Configuration.

2: Click TCP/IP and make sure all settings are entered correctly.

3: If NIS is enabled click Configure NIS and make sure the NIS domain and server names are correct.

Customer tries to mount the /export file system with Volume Manager 2.1.1, getting the message.

Action

Action

Make sure that the software matches the architecture and system you're using. The file(1) command can help you determine the target architecture. If you're using SunOS 4.1 software on a Solaris 2 system, make sure that the Binary Compatibility Package is installed. You can check for it using this command:

$ pkginfo | grep SUNWbcp

Technical Notes

A request was made to execute a file that, although it has the appropriate permissions, does not start with a valid format.

The symbolic name for this error is ENOEXEC, errno=8.

"F"

`failed to initialize adapter`

Cause

Using an Adaptec AHA-154x Cx SCSI HBA during installation of Solaris 2.x x86, you may see the message during the mdb device probe that says failed to initialize adapter after the probe has correctly identified the card. There are a variety of reasons for this error, but in all cases the error is because of misconfiguring the card.

Action

To correct the problem, press Ctrl-a during boot of the computer to enter the 154x BIOS configuration utility. Choose the Configure/View Host Adapter Settings option, then press the F6 key to return the adapter to its factory default settings.

After doing this, reconfigure the adapter per the instructions contained in the x86 Device Configuration Guide or Driver Update Guide if applicable. It is especially important that the adapter be configured to use DMA 6. Note that it defaults to DMA 5 and must be changed.

`fbconsole: ioctl SRIOCSREDIR: Device Busy.`

Cause

When starting Openwindows from the command line, the following error message is echoed on the Solaris "Welcome" screen: fbconsole: ioctl SRIOCSREDIR: Device Busy

Once inside Openwindows, The following message is displayed in the background windows and when starting cmdtool -C:

SYSTEM WARNING: Object 0x340f8, Device busy, ioctl SRIOCSREDIR returned -1, attempt to make tty the console failed (Tty package)

Action

Openwindows was probably started in the background (using the "&"). Exit Openwindows, and run the command in foreground: /usr/openwin/bin/openwin

If this doesn't help, then perhaps some daemon or process is "holding" the console. Type the command: fuser /dev/console.

A list of process id's is returned. Examine these processes to determine if an application has hold of the console (using the ps(1) command, will help).

`fd0: unformatted diskette or no diskette in the drive`

Cause

This message appears on the system console to indicate that the floppy driver fd(4) could not read the label on a diskette. Usually this is either because a new diskette has not yet been formatted, or a formatted diskette has become corrupted. This message often appears along with "read failed" and "bad format" messages after volcheck(1) is run.

Action

If you are certain that the diskette contains no data, run fdformat -d to format the diskette in DOS format. (You can also format a diskette in UFS format if you like, although then it cannot be transported to most other systems.) When the diskette is formatted, you can write on it, if it was not corrupted beyond repair.

`File descriptor in bad state`

Cause

Either a file descriptor refers to no open file or a read request was made to a file that is open only for writing.

Technical Notes

The symbolic name for this error is EBADFD, errno=81.

`File exists`

Cause

The name of an existing file was mentioned in an inappropriate context. For example, establishing a link to an existing file, or overwriting an existing file are not allowed when the csh(1) noclobber option is set.

Action

Look at the names of files in the directory, then try again with a different name or after renaming or removing the existing file.

Technical Notes

The symbolic name for this error is EEXIST, errno=17.

`File locking deadlock`

Cause

This is a programming problem, in some cases unavoidable.

Action

All a user can do is restart the program and hope deadlock does not reoccur.

Technical Notes

In the file locking subsystem, two processes tried to modify some lock at the same time. In the multithreading subsystem, two threads became deadlocked and could not continue. When a program using the threads library encounters this error, it should restart the deadlocked threads.

The symbolic name for this error is EDEADLOCK, errno=56.

`filemgr: mknod: Permission denied`

Cause

File Manager issues this message and fails to come up whenever the /tmp/.removable directory is owned by another user and is not 1777 mode. This can happen, for example, when multiple users share a workstation.

Action

Have the original owner change the mode ((chmod(1)) of this file back to 1777, its default creation mode. Rebooting the workstation also resolves this problem.

Technical Notes

This is a known problem that was fixed in Solaris 2.4.

`File name too long`

Cause

The specified file name has too many characters.

Action

If a file name or path name component is too long, devise a shorter name. If the total path name is longer than PATH_MAX characters, first change to an intermediate directory, then specify a shorter path name. Newly-created data will be lost unless written to another file with a shorter name.

Technical Notes

In a UFS or NFS-mounted UFS filesystem, the length of a path name component exceeds MAXNAMLEN (255) characters, or the total length of the path name exceeds PATH_MAX (1024) characters. In a System V filesystem, the length of a path name component exceeds NAME_MAX (14) characters while no-truncation mode is in effect. These values are defined in the /usr/include/limits.h file.

The symbolic name for this error is ENAMETOOLONG, errno=78.

# fw ctl uninstall
# fw ctl install
# fw fetch localhost

`fwm: no license`

Cause

Firewall-1, version 2.1 and produces the message when the fwstart command is issued or fwm is started from the command line.

There are two possible reasons for this.

The first is when a firewall module is installed without a control station on the same machine, the messages are displayed on the console (under Unix) or in the event log (under WinNT)

The second is that the messages may be legitimate. You may find that fwm has not started and you cannot do some crucial tasks. One problem may be that the license is issued for the wrong hostid.

Action

Check that the license daemon is running on the server. Then...

Case one: Workaround: Ignore the messages. Solution: Upgrade to 2.1c or above.

Case two: To check for a misassigned license, run the command hostid(1). Your hostid is displayed.

Now run the command fw printlic. You will see output like this...

This is FireWall-1 Version 2.1
Type             Expiration Features
id-649f152b	 never	    stdlight

The first field should list the correct hostid. Also check the expiration date and the features. A list of what is included with the features is provided in INFODOC 13215. If you find any inconsistencies, call the Sun License and Password Center and get a license reissued . Have you hostid and serial number ready.

"G"

`giving up`

Cause

"H"

`hang console`

Cause

Console hangs - but all other operations are working, including rlogin(1) and telnet(1). Rebooting the system (via a remote shell) clears the problem.

This problem occurs if another window is opened with the -C option, causing the console to hang. The other window could be another cmdtool window, shelltool window, or even an xterm window. Only one console window can be active at a time.

Action

The window/process that is causing the problem can be located by using the ps(1) command (auxw options may be necessary). The process can then be killed. Eliminate the console window running with -C, and control returns to the 'real' console.

Machine hung in reboot process: when the user is booting the machine, it hangs at checking filesystems.

A possible workaround is: 1. Boot miniroot from tape or CD-ROM, 2. mkdir mnt, 3. Mount the root partition to some mountpoint, i.e. /mnt, 4. Change directory to /mnt/dev, 5. Check to see if the console is located in the, 6. mnt/dev directory, 7. If not, make the device std.: MAKEDEV std, 8. Halt the system and reboot.

`/home/string: No such file or directory`

Cause

An attempt was made to change to a user's home directory, but either that user does not exist or the user's fileserver has not shared (exported) that filesystem.

Action

To check on the existence of a particular user, run the ypmatch(1) or nismatch(1) command, specifying the user name and then the passwd(4) map.

To export filesystems from the remote fileserver, become superuser on that system and run the share(1M) command with the appropriate options. If that system is sharing (exporting) filesystems for the first time, also invoke /etc/init.d/nfs.server start to begin NFS service.

`Host is down`

Cause

A transport connection failed because the destination host was down. For example, mail delivery was attempted over several days, but the destination machine was not available during any of these attempts.

Action

Report this error to the system administrator for the host. If you are the person responsible for this system, check to see if the machine needs repair or rebooting.

Technical Notes

This error results from status information delivered by the underlying communication interface. If there is no known connection to the host, a different message usually results. See "No route to host" for details.

The symbolic name for this error is EHOSTDOWN, errno=147.

`host name configuration error`

To read a mail message containing MIME attachments, use mailtool(1) from Solaris 2.3 or later. If you are running an earlier version of Solaris, rlogin(1) to a later version of Solaris, set the DISPLAY environment variable back to the first system, and run mailtool remotely. If those options prove impossible, ask the originator to send the message again using mailtool(1), or using the CDE dtmail compose File->SendAs->SunMailTool option.

Check that the Ethernet connector is not loose or disconnected. Other possible causes include an open circuit somewhere in the network and noise on the carrier detect line from the transceiver. Use snoop(1M) or a similar program to isolate the problem area, then check the network connectors and transceivers, as needed.

`ifconfig: bad address`

Cause

System fails to boot with error message: ifconfig: bad address. Upon coming up to multiuser ifconfig -a indicate:

le0: flags=863<UP,BROADCAST,NOTRAILERS,RUNNING,MULTICAST> mtu 1500
	inet 0.0.0.0 netmask 0

Once up, command

# ifconfig le0 inet hostname

succeeds, and all is well.

Action

Check /etc/hostname.* for a possible bad entry.

/etc/hosts was linked to /var/named/hosts and /var was a separate file system. Until system comes up in multiuser to mount /var, hostname could not be resolved to proper IP address.

`ifconfig bad address le0`

Cause

Customer installed the recommended 2.5.1 patches. When he booted, rootuser.sh presented the following errors:

ifconfig bad address le0
le0 arp - revarp failed no rarp replies
bad address hme0
hme0 auto-revarp failed: no rarp replies received.

Ip address of interface is set to 0.0.0.0

System fails to: resolve host ip address from /etc/host and no other rarp servers responded to the systems request for it's IP address.

Action

The workaround is to put files before dns [NOT FOUND=return] in /etc/nsswitch.conf

This matches bug number 4040423. Its summary is: if dns [NOTFOUND=return] appears before files in /etc/nsswitch.conf, at boot-time ifconfig complains about bad address. In some cases this can cause the boot to fail.

`If pipe/FIFO, don't sleep in stream head`

Cause

Streams pipe error (not externally visible).

Technical Notes

The symbolic name for this error is ESTRPIPE, errno=92.

`Illegal Instruction`

Cause

Technical Notes

A call to lseek(2) was issued to a pipe. This error condition can also be fixed by altering the program to avoid using lseek(2).

The symbolic name for this error is ESPIPE, errno=29.

`Image Tool: Unable to open XIL Library.`

Cause

This message follows multiple multi-line "XilDefaultErrorFunc" errors, indicating that ImageTool could not locate the X Imaging Library. Many OpenWindows and CDE deskset programs require XIL.

Action

Run pkginfo(1) to determine what packages are installed on the system. If the following packages are not present, install them from CD-ROM or over the net: SUNWxildg, SUNWxiler, SUNWxilow, and SUNWxilrt.

`Inappropriate ioctl for device`

Cause

This is a programming error.

Action

Ask the program's author to fix this condition. The program needs to be changed so it employs a device driver that can accept special character device controls.

Technical Notes

The ioctl(2) system call was given as an argument for a file that is not a special character device. This message replaces the traditional but puzzling "Not a typewriter" message.

The symbolic name for this error is ENOTTY, errno=25.

`INCORRECT BLOCK COUNT I=int (should be int) CORRECT?`

Cause

During phase 1, fsck(1M) determined that the specified inode pointed to a number of bad or duplicate blocks, so the block count should be corrected to the actual number shown.

Action

Generally you can answer yes to this question without harming the filesystem.

`index failed:full:index preceded by saveset name`

Cause

This is a server which has several clients. It seems when the backup kicks off many the savesets are failing with the message listed below.

godzilla                              index failed:full:index      
* godzilla:index 2 retries attempted
* godzilla:index sh: save: not found

Action

Edit the /etc/init.d/networker file and change the nsrexecd startup line to include a -p option to specify the command search path :

(/usr/sbin/nsr/nsrexecd -s masters -p /usr/sbin/nsr )     > /dev/console

`inetd[int]: execv /usr/sbin/in.uucpd: No such file or directory`

Cause

This message indicates that the Internet services daemon inetd(1M) tried to start up the UUCP service without the UUCP daemon existing on the system.

Action

The SUNWbnuu package must be installed before the machine can run UUCP. Run pkgadd(1M) to install this package from the distribution CD-ROM or over the network.

`inetd[int]: string/tcp: unknown service`

Cause

This message indicates that the Internet services daemon inetd(1M) could not locate the TCP service specified after the first colon.

Action

Check the current machine's /etc/services file, and the NIS services map, to see if the service is described. To start this service, add an appropriate entry into the /etc/services file and possibly the services map as well. Note that NIS+ does not consult the local /etc/services file unless you put "files" right after "nisplus" on the services line of the system's /etc/nsswitch.conf file.

If you do not want to start this service, edit the system's /etc/inetd.conf file and delete the entry that tries to start it up.

`inetd[int]: string/udp: unknown service`

Cause

This message indicates that the Internet services daemon inetd(1M) could not locate the UDP service specified after the first colon.

Action

See the message "inetd[int]: string/tcp: unknown service" for a solution.

`inetd: Too many open files`

Cause

This message can appear when someone runs a command from the shell or uses a third-party application. The sar(1) command does not indicate that the system-wide open file limit has been exceeded.

Action

Technical Notes

An asynchronous signal (such as interrupt or quit), which a program was set up to catch, occurred during an internal system call. If execution is resumed after processing the signal, it will appear as if the interrupted programming function returned this error condition, so the program might exit with an incorrect error message.

The symbolic name for this error is EINTR, errno=4.

`Invalid argument`

Cause

Check out www.marimba.com for more details on marimba products.

Another possible solution:

Double-click jws.exe within the: C:\Java-WorkShop\jws\intel-win32\bin\ folder.

`I/O error`

Cause

Some physical Input/Output error has occurred. If the process was writing a file at the time, data corruption is possible.

Action

First find out which device is experiencing the I/O error. If the device is a tape drive, make sure a tape is inserted into the drive. When this error occurs with a tape in the drive, it is likely that the tape contains an unrecoverable bad spot.

If the device is a floppy drive, an unformatted or defective diskette could be at fault. Format the diskette, or obtain a replacement.

If the device is a hard disk drive, you might need to run fsck(1M) and possibly even reformat the disk.

Technical Notes

In some cases this error might occur on a call following the one to which it actually applies.

The symbolic name for this error is EIO, errno=5.

`Is a directory`

Cause

An attempt was made to read or write a directory as if it were a file.

Action

Look at a listing of all the files in the current directory and try again, specifying a file instead of a directory.

Technical Notes

The symbolic name for this error is EISDIR, errno=21.

"J"

`java.lang.UnsatisfiedLinkError:`

Cause

When trying to start Java Workshop 2.0 (or some other Java applications) on Solaris 2, the following error is displayed:

java.lang.UnsatisfiedLinkError: setCursor
        at sun.awt.motif.MComponentPeer.initialize(Compiled Code)
        at sun.awt.motif.MTextAreaPeer.initialize(Compiled Code)
        at sun.awt.motif.MComponentPeer.<init>(Compiled Code)
        at sun.awt.motif.MTextAreaPeer.<init>(Compiled Code)
        at sun.awt.motif.MToolkit.createTextArea(Compiled Code)

Action

The LD_LIBRARY_PATH is probably set up to include a Java lib directory that doesn't quite match the java bin command used. For example, on Solaris 2.6: LD_LIBRARY_PATH = /usr/openwin/lib results in Java Workshop runing properly. But setting: LD_LIBRARY_PATH = /usr/java/lib:/usr/openwin/lib results in the error being displayed, since Java Workshop uses its own version of JDK and the startup process picks up a mixture of versions.

The answer is to not include /usr/java/lib in your LD_LIBRARY_PATH since it is needed only in rare circumstances, such as if you are using the Java Invocation API.

`last message repeated int times`

Cause

This message comes from syslogd(1M), the facility that prints messages on the console and records them in /var/adm/messages. To reduce the log size and minimize buffer usage, syslog collapses any identical messages it sees during a 20 second period, then prints this message with the number of repetitions.

Action

Look above this message to see which message was repeated so often. Then consider the repeated message and take action accordingly. If repeated log entries such as su ... failed appear, consider the possibility of a security breach.

`ld.so.1 fatal: can't set protection on segment`

Cause

Applications have recently begun to fail with this error: ld.so.1 fatal: can't set protection on segment. The failures are random.

Action

This was happening because of the recent introduction of a rogue application that consumed most of the swap space on the system. The other applications, which failed randomly, were doing so because of insufficient swap space to run. The error from ld.so.1 happened because there was no segment on which to set the protections.

`ld.so.1: string: fatal: string: can't open file: errno=2`

Cause

This message is produced in Solaris 2.5.1 and earlier releases. It is not produced in Solaris 2.6, or later releases.

Run the ldd -d command on the application to show its shared object dependencies and symbols that aren't found. Probably your system contains an old version of the shared object that should contain this symbol. Contact the library vendor or author for an update.

Technical Notes

This error does not necessarily occur when you first bring up an application. It could take months to develop, if ordinary use of the application seldom references the undefined symbol.

`le0: Memory error!`

Cause

This message indicates that the network interface encountered an access time-out from the CPU's main memory. There is probably nothing wrong except system overload.

Action

If the system is busy with other processes, this error can occur frequently. If possible, try to reduce the system load by quitting applications or killing some processes.

Technical Notes

The Lance Ethernet chip timed out while trying to acquire the bus for a DVMA transfer. Most network applications wait for a transfer to occur, so generally no data gets lost. However, data transfer might fail after too many time-outs.

`le0: No carrier-- cable disconnected or hub link test disabled?`

Cause

Standalone machines with no Ethernet port connection get this error when the system tries to access the network. If the Ethernet cable is disconnected, SPARC machines with the sun4m architecture usually display this message, whereas machines with the sun4c architecture usually display the "le0: No carrier-- transceiver cable problem" message instead. If the Ethernet cable is connected, this message could result from a mismatch between the machine's NVRAM settings and the Ethernet hub settings.

Action

If this message is continuous, try to save any work to local disk.

When a machine is configured as a networked system, it must be plugged into the Ethernet with a twisted pair J45 connector.

If the Ethernet cable is plugged in, find out whether or not the Ethernet hub does a Link Integrity Test. Then become superuser to check and possibly set the machine's NVRAM. If the hub's Link Integrity Test is disabled, set this variable to false.

# eeprom | grep tpe
tpe-link-test?=true
# eeprom 'tpe-link-test?=false'

The default setting is true. If for some reason tpe-link-test? was set to false, and the hub's Link Integrity Test is enabled, reset this variable to true.

`le0: No carrier-- transceiver cable problem?`

Cause

Standalone machines with no Ethernet port connection get this error when the system tries to access the network.

Action

If this message is continuous, try to save any work to local disk.

When a machine is configured as a networked system, it must be plugged into the Ethernet with either a twisted pair J45 connector or thicknet 10Base-T connector (depending on the building's Ethernet cable type).

Technical Notes

Older workstations have a thicknet connection on the back instead of a twisted pair Ethernet connection, so they require a thicknet to twisted pair transceiver to translate between cabling types.

`level 15 interrupt`

Action

Check the /etc/passwd file and the NIS or NIS+ passwd map on the local system to see if an entry exists for this user. If a user has simply forgotten the password, su(1M) and set a new one with the passwd(1) username command. This command automatically updates the NIS+ passwd map, but with NIS you'll need to coordinate the update with the passwd map.

The "Login incorrect" problem can also occur with older versions of NIS when the user name has more than eight characters. If this is the case, edit the NIS password file, change the user name to have eight or fewer characters, and then remake the NIS passwd map.

If you cannot log in to the system as root, despite knowing the proper password, it is possible that the /etc/passwd file is corrupted. Try to log in as a regular user and su(1M) to root.

If that doesn't work, see the message "su: No shell" and follow most of the instructions given there. Instead of changing the default shell, make the password field blank in /etc/shadow.

`lp hang`

An attempt was made to send a message with mailtool(1) from a directory where the user does not have write permission, and the user's home directory is currently unavailable.

Action

Change to another directory and start mailtool(1) again, or use chmod(1) to change permissions for the directory (if possible).

`mailtool: Could not initialize the Classing Engine`

Cause

When a user runs mailtool(1) on a remote machine, setting the DISPLAY environment back to the local machine, this message might appear inside a dialog box window. The dialog box goes on to say that the Classing Engine must be installed to use Attachments. This problem occurs because rlogin(1) does not propagate the user's environment.

Action

Exit mailtool(1) and set your OPENWINHOME environment variable to /usr/openwin. Then run mailtool(1) again. The error message will not appear, and you will be able to use Attachments.

Technical Notes

Classing Engine is a new name for Tool Talk. Earlier versions of mailtool(1) said Tool Talk: TT_ERR_NOMP instead of Classing Engine.

`Mail Tool is confused about the state of your Mail File.`

Cause

This message appears in a pop-up dialog box whenever you ask mailtool(1) to access messages after another mail reader has modified your inbox. A request follows: "Please Quit this Mail Tool."

Action

Click "Continue" to close the dialog box, then exit mailtool(1). If you continue trying to read mail, messages deleted by the other mail reader will never appear, and mailtool(1) will fail to see any new messages.

`mail: Your mailfile was found to be corrupted (Content-length mismatch).`

Cause

This message comes from mail(1) or mailx(1)(1) whenever it detects messages with a different content length than advertised. The mail(1) program tells you which message might be truncated or might have another message concatenated to it.

Two common causes of content length mismatches are the simultaneous use of different mail readers (such as mail(1) and mailtool(1)), or using a mail reading program (or an editor) that does not update the Content-Length field after altering a message.

Action

The mailx(1) program can usually recover from this error and delineate mail message boundaries correctly. Pay close attention to the message that might be truncated or combined with another message, and to all messages after that one. If a mail file becomes hopelessly corrupted, run it through a text editor to eliminate all Content-Length lines, and ensure that each message has a From (no colon) line for each message, preceded by a blank line.

To avoid mailfile corruption, exit from mailtool(1) without saving changes when you are currently running mail(1) or mailx(1).

`Machine is not on the network`

Cause

This error is Remote File Sharing (RFS) specific. It occurs when users try to advertise, unadvertise, mount, or unmount remote resources while the machine has not properly started a network connect.

Technical Notes

The symbolic name for this error is ENONET, errno=64.

`mailtool: Can't create dead letter: Permission denied`

This message appears in a pop-up dialog box whenever you ask mailtool(1)(1) to access messages after another mail reader has modified your inbox. A request follows: "Please Quit this Mail Tool."

Action

`mail: Your mailfile was found to be corrupted (Content-length mismatch).`

Cause

This message comes from mail(1) or mailx(1) whenever it detects messages with a different content length than advertised. The mail(1) program tells you which message might be truncated or might have another message concatenated to it.

Action

To avoid mailfile corruption, exit from mailtool(1) without saving changes when you are currently running mail(1) or mailx(1).

`file name may contain holes - can't swap on it`.

Cause

A swapfile was created with the command:

# mkfile -nv 50m /ab/swap_50mb

When the user tried to add the file:

# swap -a /ab/swap_50mb

It failed with the message:

/ab/swap_50mb may contain holes - can't swap on it.
/ab/swap_50mb: Error 0

Action

The -n option was supported on SunOS 4, but on SunOS 5 (Solaris 2) -n works only when the file is to be used by NFS. Local swap files cannot be created with the -n option.

`mbuf map full`

Action

`Memory address alignment`

Cause

This message can occur when printing large files on a SPARCprinter attached to a SPARCstation 2.

Action

Replace the SPARCstation 2 CPU with one that is at the most recent dash level.

`memory leaks`

Cause

An application uses up more and more memory, until all swap space is exhausted.

Action

`Message too long`

Cause

A message sent on a transport provider was larger than the internal message buffer or some other network limit.

Technical Notes

The symbolic name for this error is EMSGSIZE, errno=97.

`mount: /dev/dsk/string is already mounted, /string is busy, or...`

Cause

Action

`mount: giving up on: /string`

Cause

Action

If the "RPC: Program not registered" message precedes this one, the requested mount server probably did not share (export) any filesystems, so it has no NFS daemons running. Have the superuser on the mount server share(1M) the filesystem, then run /etc/init.d/nfs.server start to begin NFS service.

If the requested mount server is down or slow to respond, check to see whether the machine needs repair or rebooting.

`mount: mount-point /string does not exist.`

Cause

Someone tried to mount a filesystem onto the specified directory, but there is no such directory.

Action

If this is the directory name you want, run mkdir(1) to create this directory as a mount point.

`mount: the state of /dev/dsk/string is not okay`

Cause

The system was unable to mount the filesystem that was specified because the super-block indicates that the filesystem might be corrupted. This is not an impediment for read-only mounts.

Action

`Multihop attempted`

On the DNS server, the error message is displayed.

Action

This error indicates that an alias (CNAME) is associated with another type of DNS record.

The DNS system allow you to set up an alias to a system using the CNAME record. An example of this is:

alias1		IN CNAME	host1.domain1.

The alias alias1 cannot appear in any other type of record. Only the actual name of the host may be used. So, if you wanted to use this host as a mail exhanger, the record:

alias1		IN MX  10  host2.domain1.

would be illegal and would produce the error.

Instead, you should use

host1		IN MX  10  host2.domain1.

This goes for all types of records, including HINFO and A records.

Also, it is possible to get this error without explicitly setting the left hand side of a record. The DNS system defaults the left hand side to the last given left hand side. So you might have the following in a named database file:

host1	IN A	 123.124.125.126
        IN HINFO Sun Solaris
alias1  IN CNAME host1.domain1.
        IN MX 10 host2.domain1.

In this fragment, there is an implied alias1 in the left hand side of the MX record. If the alias was added after the database was in use for a while, the error would suddenly start showing up. Since the MX record was legal until the CNAME was added in front of it. This example could be fixed either by reversing the order of the MX and CNAME records, or explicitly giving the host1 in the lefthand side of the MX record.

`/net/string: No such file or directory`

Cause

A user tried to change directory (for example with cd(1)) to a network partition on the system specified after /net/, but this host either does not exist or has not shared (exported) any filesystem.

Action

To gain access to files on this system, try rlogin(1).

To export filesystems from the remote system, become superuser on that system and run the share(1M) command with the appropriate options. If that system is sharing filesystems for the first time, also run /etc/init.d/nfs.server start to begin NFS service.

`Network dropped connection because of reset`

Cause

The host you were connected to crashed and rebooted.

Technical Notes

The symbolic name for this error is ENETRESET, errno=129.

`Network is down`

Cause

A transport connection failed because it encountered a dead network.

Action

Report this error to the system administrator for the network. If you are the person responsible for this network, check to see why the network is dead and what repairs are necessary.

Technical Notes

This error results from status information delivered by the underlying communication interface.

The symbolic name for this error is ENETDOWN, errno=127.

`Network is unreachable`

Calendar :Informational - NFS mounted callog file Unsupported.
Your default startup Calendar file appears to be NFS mounted or
a symlink to the same.  This is Not Supported.
			Continue

The following error is displayed in the console window when the Continue button is clicked:

date time host rpc.cmsd[pid]: rpc.cmsd : 
	NFS mounted callog file Not Supported - user@host
date time host rpc.cmsd[pid]: rpc.cmsd : 
	NFS mounted callog file Not Supported - user@host

The calendars would've worked under 2.5.1 or before however.

Action

Action

Check the non-responding NFS server to see whether the machine needs repair or rebooting. Encourage your user community to report such problems quickly but only once.

Should this message appear when booting a diskless client, make sure that the client's /etc/hosts file and the network naming service (NIS, NIS+, or other /etc/hosts files on the network) have been updated.

`NFS server string ok`

Cause

This message is the follow-up to the "NFS server not responding" error. It indicates that the NFS server is back in operation.

Action

When an NFS server first comes up, it is busy fulfilling client requests for a while. Be patient and wait for your client system to respond. Making many extraneous requests only further slows the NFS server response time.

`NFS string failed for server string: error int (string)`

Sep 11 16:34:00 nis_cachemgr: Error in reading NIS cold start file : 
          '/var/nis/NIS_COLD_START'

and nis_cachemgr(1M) isn't there after logging in. Trussing nis_cachemgr(1M) showed that it is reading /var/nis/NIS_COLD_START and immediately reporting an error. Neither reinitializing the client nor copying NIS_COLD_START helps.

Action

This is a timing problem. Put a sleep(1) before the NIS+ initialization in /etc/init.d/rpc, after rpc.bind has been started. rpc.bind is slow about initializing and needs a few extra seconds to get going before nis_cachemgr(1M) starts bugging it.

`No buffer space available`

Cause

An operation on a transport endpoint or pipe was not performed because the system lacked sufficient buffer space or because a queue was full. The target system probably ran out of memory or swap space. Any data written during this condition will probably be lost.

Action

To add more swap area, use the swap -a command on the target system. Alternatively, reconfigure the target system to have more swap space. As a general rule, swap space should be two to three times as large as physical memory.

Technical Notes

The symbolic name for this error is ENOBUFS, errno=132.

`No child processes`

Cause

This message can appear when an application tries to communicate with cooperating process that do not exist.

Action

Restart the parent process so it can create the child processes again. If that doesn't help, this could be the result of a programming error; contact the vendor or author of the program for an update.

Technical Notes

A wait(2) system call was executed by a process that had no existing or unwaited-for child processes. The child processes could have exited prematurely, or might never have been created.

The symbolic name for this error is ECHILD, errno=10.

`No default media available`

Cause

The volume manager issues this message if a user makes an eject(1) request when the drives contain no diskette or CD-ROM to eject.

Action

Insert a diskette or CD-ROM. If the volume manager is confused and there actually is a diskette or CD-ROM in a drive, run volcheck(1) to update the volume manager. If the system remains confused, try booting with the -r option to reconfigure devices.

`No directory! Logging in with home=/`

Cause

The login(1) program could not find the home directory listed in the password file or NIS passwd(4) map, so it deposited the user in the root directory.

Action

Check that the user's home directory is mounted and is owned by and accessible to that user. Perhaps the automounter tried to mount the home directory, but the NFS server did not respond quickly enough. Try listing the files in /home/username. If the NFS server responds to this request, have the user log out and log in again.

It is possible that the automounter daemon is not running. Run the ps(1) command to see if automountd(1M) is present. If not, run the second command; if it appears to be wedged, run both these commands:

# /etc/init.d/autofs stop
# /etc/init.d/autofs start

When the automounter daemon is running, verify that the /etc/auto_master file has a line like this:

/home  auto_home

Verify that the /etc/auto_home file has a line like this:

+auto_home

These entries depend on the NIS auto_home map.

It is also possible that the NFS server has not shared (exported) this /home directory, or that the NFS daemons on the server have disappeared.

`No message of desired type`

Cause

An attempt was made to receive a message of a type that does not exist on the specified message queue. See the msgsnd(2) and msgrcv(2) man pages for details.

Action

This indicates an error in the System V IPC message facility. Generally the message queue is empty or devoid of the desired message type, while IPC_NOWAIT is set.

Cause

A user has attempted to a remote login to the system, and has a valid account name and password, but the shell specified for their account is not available on that system. For example, the seventh field could request the GNU Bourne-again shell /bin/bash, which does not exist on standard Solaris distributions.

Action

If you have a copy of the requested shell, become superuser and install the missing shell on that system. Otherwise, change the user's password file entry (perhaps only in the NIS+ or NIS passwd(4) map) to specify an available shell such as /bin/csh or /bin/ksh.

`No space left on device`

Cause

While writing an ordinary file or creating a directory entry, there was no free space left on the device. The disk, tape, or diskette is full of data. Any data written to that device during this condition will be lost.

Action

Remove unneeded files from the hard disk or diskette until there is space for all the data you are writing. It might be advisable to move some directories onto another filesystem and create symbolic links accordingly. When a tape is full, continue on another one, use a higher density setting, or obtain a higher-capacity tape.

To create multi-volume tapes or diskettes, use the pax(1) or cpio(1) command; tar(1) is still limited to a single volume.

Technical Notes

The symbolic name for this error is ENOSPC, errno=28.

`No such device`

Cause

An attempt was made to apply an operation to an inappropriate device, such as writing to a nonexistent device.

Action

Look in the /devices directory to see why this device does not exist, or why the program expects it to exist. The similar "No such device or address" message tends to indicate I/O problems with an existing device, whereas this message tends to indicate a device that does not exist at all.

Technical Notes

The symbolic name for this error is ENODEV, errno=19.

`No such device or address`

Cause

This error can occur when a tape drive is off-line or when a device has been powered off or removed from the system.

Action

For tape drives, make sure the device is connected, powered on, and toggled online (if applicable). For disk and CD-ROM drives, check that the device is connected and powered on.

With all SCSI devices, ensure that the target switch or dial is set to the number where the system originally mounted it. To inform the system of a change to the target device number, reboot using the -r (reconfigure) option.

Technical Notes

This message results from I/O to a special file's subdevice that either does not exist or that exists beyond the limit of the device.

The symbolic name for this error is ENXIO, errno=6.

`No such file or directory`

Cause

Technical Notes

Action

Look at a listing of all the files in the current directory and try again, specifying a directory instead of a file.

Technical Notes

The symbolic name for this error is ENOTDIR, errno=20.

`Not a stream device`

Cause

A putmsg(2) or getmsg(2) system call was attempted on a file descriptor that is not a STREAMS device.

Technical Notes

The symbolic name for this error is ENOSTR, errno=60.

`Not enough space`

Cause

This message indicates that the system is running many large applications simultaneously, and has run out of swap space (virtual memory). It could also indicate that applications failed without freeing pages from the swap area. Swap space is an area of disk set aside to store portions of applications and data not immediately required in memory. Any data written during this condition will probably be lost.

Action

Reinstall or reconfigure the system to have more swap space. A general rule of thumb is that swap space should be two to three times as large as physical memory. Alternatively, use mkfile(1M) and swap(1M) to add more swap area. This example shows how to add 16 MB of virtual memory in the /usr/swap file (any filesystem with enough free space would work):

# mkfile 16m /usr/swap
# swap -a /usr/swap

To make this automatic at boot time, add the following line to the /etc/vfstab file:

/usr/swap   -   -   swap   -   no   -

Technical Notes

In calling the fork(2), exec(2), sbrk(2), or malloc(3C) routine, a program asked for more memory than the system could supply. This is not a temporary condition; swap space is a system parameter.

The symbolic name for this error is ENOMEM, errno=12.

`not found`

Cause

This message indicates that the Bourne shell could not find the program name given as a command.

Action

Check the form and spelling of the command line. If that looks correct, echo $PATH to see if the user's search path is correct. When communications are garbled, it is possible to unset a search path to such an extent that only built-in shell commands are available. Here is a command to reset a basic search path:

$ PATH=/usr/bin:/usr/ccs/bin:/usr/openwin/bin:.

If the search path looks correct, check the directory contents along the search path to see if programs are missing or if directories are not mounted.

`NOTICE: vxvm: unexpected status on close`

Cause

Everytime system boots (or is shut down), the message is displayed on the console. Sometimes, the following message is also displayed on the console and in the /var/adm/messages file:

	WARNING:
/iommu@0,10000000/sbus@0,10001000/SUNW,soc@2,0/SUNW,pln@a0000000,74127a/ssd@4,2
(ssd22):
	Error for Command: <undecoded cmd 0x35>       Error Level: Fatal
	Requested Block: 0      Error Block: 0
	Vendor: CONNER                  Serial Number: 93081LPT
	Sense Key: Aborted Command
	ASC: 0xb3 (<vendor unique code 0xb3>), ASCQ: 0x0, FRU: 0x0
	WARNING:
/iommu@0,10000000/sbus@0,10001000/SUNW,soc@2,0/SUNW,pln@a0000000,74127a/ssd@4,2
(ssd22): ssd_synchronize_cache failed (5)

Action

In a High Availability system with NVRAM, this would be caused by unprocessed data in a NVRAM cache of the active logical host that has been down and come up later. Because of this, NVRAM should not be used in an HA system. The problem can be solved in this case by getting rid of the NVRAM on the HA system.

In a non-HA system, this can also be caused by stale data in the NVRAM cache. (The example commands below assume the controller for the array is c1.) To fix for a non-HA system:

1. Turn off all fast writes on this array and sync any remaining pending writes.

# ssaadm fast_write -d c1
# ssaadm sync_cache c1

2. When you sync the fast writes to the array, all pending writes are physically made to the disks. Anything that is left in the cache is stale, and thus, it is safe to purge it. Run the command:

# ssaadm purge c1

3. Turn the fast writes for the disks back on. This command MAY be different on your system, depending on the disks on which you want fast writes enabled and the types of fast writes you want:

# ssaadm fast_write -s -e c1

`NOTICE: /string: out of inodes`

Cause

The filesystem specified after the first colon probably contains many small files, exceeding the per-filesystem limit for inodes (file information nodes).

Action

If many small files were created unintentionally, removing them resolves the problem.

Otherwise, follow these steps to increase filesystem capacity for small files. Make several backup copies of the filesystem on different tapes (for safety), then bring the machine down to single-user mode. Use the newfs(1M) command with the -i option to increase inode density for this filesystem. Here is an example:

# newfs -i 1024 /dev/rdsk/partition

Switch user to root and try again.

Technical Notes

The symbolic name for this error is EPERM, errno=1.

`Not supported`

Cause

This version of the system does not support the feature requested, although future versions of the system might provide support.

Action

This is generally not a system message from the kernel, but an error returned by an application. Contact the vendor or author of the application for an update.

Technical Notes

The symbolic name for this error is ENOTSUP, errno=48.

`No utmpx entry`

Cause

During login file system full errors are seen and login fails with the message No utmpx entry.

This is caused by a full file system so that the system has no space to write its utmpx (login info) entry.

Action

To get around this condition the system must be booted into single user mode. Then clear (do not delete) the files: /var/adm/utmp and /var/adm/utmpx. This can be done by typing:

#cat /dev/null > /var/adm/utmp
#cat /dev/null > /var/adm/utmpx

These commands zero out the files but keep it with the correct permissions.

In some cases after clearing these files the /var filesystem may still be full. In this case type:

du -askd /var |sort -nr |more

This will give you a listing of the files from largest to smallest in the /var filesystem. To create space you can zero the files: /var/cron/log, /var/spool/lp/logs, and /var/adm/messages. You can also check /.wastebasket for large files to delete.

`6/04/98 7:27:54 nsrck: SYSTEM error, more space needed to compress [client] index, 8.1 MB required`

Cause

In networker, cannot use the Remove Oldest Cycle feature because the /nsr filesystem is too full to perform a remove. An error message appears in the console window indicating that the file system is full.

Action

1. Stop the networker daemons so that some of the indices can be moved. In SunOS 5, use /etc/init.d/networker stop. In SunOS 4, use ps -ef | grep nsr and kill(1) the processes.

2. Find a filesystem with enough space to move one of the client's indices. Only one of the client's indices should be moved, not the networker server's index. To find the size of a client's index, go to /nsr/index/client name/db and list the contents using ls -l. The data base file can be very large (possibly over 500 MB).

3. Move the contents of a client's index to the other filesystem and check that /nsr has freed the space to use. It may be necessary to unmount and remount /nsr, or even to reboot to designate the space freed by the move as available.

4. Once the space is available, restart the daemons.

5. Go into nwadmin. Under Clients--Indexes, select a client and use Remove Oldest Cycle to free more space.

Use Reclaim Space to reclaim the space from the removed cycles. After a few of the old cycles have been removed, there should be enough space in the filesystem to move the removed client's index back.

6. Stop the daemons, and move the client's index back to /nsr/index/clientname.

7. Restart the daemons. Remove oldest cycles for the client that was just moved.

Tweaking of the browse policy and retention policy may be necessary to prevent this situation from happening in the future.

Other, long term solutions are to add more hard disk and run growfs or move /nsr to a drive with more space on it.

"O"

`Object is remote`

Cause

This error occurs when users try to share a resource that is not on the local machine, or try to mount/unmount a device or path name that is on a remote machine.

Technical Notes

The symbolic name for this error is EREMOTE, errno=66.

`ok`

Cause

This is the OpenBoot PROM monitor prompt. From this prompt, you can boot the system (from disk, CD-ROM, or net), or you can use the go command to continue where you left off.

Action

If you suddenly see this prompt, look at the messages above it to see if the system crashed. If no other messages appear, and you just typed Stop-A or plugged in a new keyboard, type go to continue. You might need to Refresh the window system from its Workspace Menu.

Technical Notes

Never invoke sync from the ok prompt without first running the fsck(1M) command, especially if the filesystem has changed.

`Operation already in progress`

Cause

An operation was attempted on a non-blocking object that already had an operation in progress.

Technical Notes

The symbolic name for this error is EALREADY, errno=149.

`Operation canceled`

Cause

The associated asynchronous operation was canceled before completion.

Technical Notes

The symbolic name for this error is ECANCELED, errno=47.

`operation failed [error 185], unknown group error 0, string`

Cause

When you use admintool to add a user to a newly-created group, admintool issues this error.

Action

Apply patch 101384-05 to fix bug ID 1151837 and to provide a workaround for bug ID 1153087.

`Operation not applicable`

Cause

This error indicates that no system support exists for some function that the application requested.

Action

Ask the system vendor for an upgrade, or contact the vendor or author of the application for an update.

Technical Notes

Action

Action

Change the names of these volumes to read swap1, swap2, etc...

If you still get this message after making the above change, edit the /sbin/swapadd script. Find the line:

c=`$SWAP -l | grep -c '\\<'${special}'\>'`

and change it to

=`$SWAP -l | grep -c ''${special}''`

Also see bug number 1215062 for more information on this behavior.

"P"

`Package not installed`

Cause

This error occurs when users attempt to use a system call from a package which has not been installed.

Technical Notes

The symbolic name for this error is ENOPKG, errno=65.

`panic -boot: Could not mount filesystem`

Cause

The primary problem is that jumpstart gives the following error:

2ec00 RPC: Can't decode result.
whoami RPC call failed with rpc status: 2
panic - boot: Could not mount filesystem.
program terminated
ok

Normally this is a result of the bootparams not being able to get to the install image.

The second problem is that other users have had the same error message, with the additional message:

'Timeout waiting for ARP/RARP packet...'

Action

To solve the first problem:

1. Check how the dfstab(4) (/etc/dfs/dfstab on the install image NFS server) looks:

share -F nfs -o ro,anon=o /jumpstart-dir

2. Run share(1M) command on the installed image NFS server, to make sure it is shared properly.

3. Check /etc/bootparams file on the net install server. Look for entries with incorrect boot path.

4. Make sure that /usr/sbin/rpc.bootparamd is running on the boot server. If necessary, kill and restart it.

5. Check /etc/ethers on the boot server for duplicate or conflicting entries

6. At OK prompt, run test net /test-net and/or watch net /watch-net to test the network connectivity

A workaround for the second problem is to check the nsswitch.conf(4) file. If some of the entries point to NIS such as:

rpc		nis	files
hosts		nis	files
ethers		nis	files
bootparams	files   nis

change all of these entries to files first:

rpc		files 	nis
hosts		files 	nis
ethers		files	nis
bootparams	files	nis

Note - You may have to manually update these files if they do not contain info on the client machine you are trying to jumpstart.

Then remove the client with rm_install_client(1M), remove the contents of tftpboot, and re-add the client:

add_install_client -c /jumpstart-dir/profiles  'client name'  'arch'

`panic: mutex_adaptive_exit`

Cause

CD-ROM boot gets error:panic: mutex_adaptive_exit Documented above under message "panic_mutex". Same SRDB ID.

`Panic`

Cause

A system panics and crashes when a program exercises an operating system bug. Although the crash might seem unfriendly to a user, the sudden stop actually safeguards the system and its data from further corruption.

Along with bringing the operating system to a stop, the panic routine copies the memory contents in use to a dump device, recording critical information about the current state of the CPU from which the panic routine was called.

Because the primary swap device is usually the default dump device, the primary swap device should be large enough to hold a complete image of memory. The system tries to reboot after the memory image is saved.

If the system does not reboot successfully, consider these possibilities:

1. Catastrophic hardware failure, such as faulty memory or a crashed disk;

2. Major kernel configuration faults, such as a buggy device driver;

3. Major kernel tuning errors, such as a too-large value for maxusers;

4. Data corruption, including corruption of the operating system files;

5. Manual intervention is needed, as when fsck(1M) expects answers to its queries.

Action

To find out why a system crashed, you can

1. Look in the /var/adm/message* log files;

Action

Of these methods, using savecore(1M) is the most informative. The savecore(1M) command transfers the system crash dump image generated by the panic routine from the dump device to a file system. The image can then be analyzed with a debugger such as adb(1).

`PARTIALLY ALLOCATED INODE I=int CLEAR?`

Cause

During phase 1, fsck(1M) found that the specified inode was neither allocated nor unallocated. The reason is probably that the system crashed in the middle of a sync(2) or write(2) operation.

Action

Should you answer yes to this question, "UNALLOCATED" messages might result during phase 2, if any directory entries point to this inode. If you are being careful, exit fsck(1M) and run ncheck(1M) (specifying the inode number after the -i option) to determine which file or directory is involved here. You might be able restore this file or directory from another system. It is also possible that fsck(1M) will copy this file to the lost+found directory in a later phase.

`passwd: Changing password for string`

Cause

If you put the following lines into /etc/nsswitch.conf

 passwd:     compat
passwd_compat:     nis

Then when you run passwd, it fails as follows:

server1% passwd
passwd:  Changing password for khh
server1%

NOTE: passwd exits before you get the opportunity to enter a password.

Action

If you read the man page for passwd very carefully, you will see the following:

If all requirements are met, by default, the passwd command will consult /etc/nsswitch.conf to determine in which repositories to perform password update. It searches the passwd(4) and passwd_compat entries. The sources (repositories) associated with these entries will be updated. However, the password update configurations supported are limited to the following 5 cases. Failure to comply with the configurations will prevent users from logging onto the system.

   o passwd: files
   o passwd: files nis
   o passwd: files nisplus
   o passwd: compat (==> files nis)
   o passwd: compat (==> files nisplus)
      passwd_compat: nisplus

rlogin(1) fails on SunOS machine.

Action

1. Check the permissions on in.rlogind on the machine you are trying to connect to. The permissions should look like this:

-rwxr-xr-x  1 root     staff       16384 Jan 20  1994 /usr/etc/in.rlogind

2. Check the login line in the /etc/inetd.conf file. It should look like the following:

login	stream	tcp	nowait	root	/usr/etc/in.rlogind	in.rlogind

3. Check /etc/passwd to see if an invalid login shell has been substituted in the entry for the login id.

`Protocol family not supported`

Cause

The protocol family has not been configured into the system or no implementation for it exists. Used for the Internet protocols.

Technical Notes

The symbolic name for this error is EPFNOSUPPORT, errno=123.

`Protocol not supported`

Cause

The requested networking protocol has not been configured into the system, or no implementation for it exists. (A protocol is a formal description of the messages to be exchanged and the rules to be followed when systems exchange information.)

Action

Verify that the protocol is in the /etc/inet/protocols file and in the NIS protocols map, if applicable. If the protocol is not listed, and you want to permit its use, configure the protocol as documented or as required.

Technical Notes

The symbolic name for this error is EPROTONOSUPPORT, errno=120.

`Protocol wrong type for socket`

Cause

This message indicates either an application programming error, or badly configured protocols.

Action

Make sure that the /etc/protocols file corresponds number-for-number with the NIS protocols(4) map. It it does, ask the vendor or author of the application for an update.

Technical Notes

A protocol was specified that does not support the semantics of the socket type requested. This amounts to a request for an unsupported type of socket. Look at the source code that made this socket request and check that it requested one of the types specified in /usr/include/sys/socket.h.

The symbolic name for this error is EPROTOTYPE, errno=98.

"Q"

`quotactl: open Is a directory`

Cause

Using edquota to set users limits the command displays the error.

edquota updates all quota files that are on a mounted filesystem. So, a directory named quotas causes it to fail.

Action

In one of the mounted filesystems, there is a directory named quotas. To fix the problem, move the directory off the mounted filesystem, rename it, or delete it.

For example: If you have /usr/quotas/old_info The directory /usr/quotas is going to cause edquota to fail. Either move /usr/quotas to /usr/old_quotas or delete the directory.

"R"

`Read error from network: Connection reset by peer`

Cause

This message appears when a user logs remotely into a machine that crashes or s rebooted during the rlogin(1) or rsh(1) session. Any data changes that were not saved are probably lost. Sometimes this message appears only when the user types something, even though the system went down hours before.

Action

Try to rlogin(1) again, perhaps after waiting a few minutes for the system to reboot.

`Read-only file system`

Cause

Files and directories on filesystems that are mounted read-only cannot be changed.

Action

If you only modify these files and directories occasionally, rlogin(1) to the servers from which the filesystems are mounted and change the files or directories there. If you change these files and directories frequently, mount(1M) the filesystems read/write.

Technical Notes

The symbolic name for this error is EROFS, errno=30.

`rebooting...`

Cause

This message appears on the console to indicate that the machine is booting, either after the superuser issued a reboot(1M) command, or after a system panic if the EEPROM's watchdog-reboot? variable is set to true.

Action

Allow the machine to boot itself. In case of a system panic, look above this message for other indications of what went wrong.

`Recipient names must be specified`

Cause

Somebody sent mail without a valid recipient in the To: field, so sendmail(1M) could not deliver the mail message. Using mail(1), the recipient's address might have been specified using spaces or non-alphanumeric characters. The mailtool(1) and mailx(1) commands try to prevent this by issuing "Please specify a recipient" or "No recipients specified" messages instead. If there is at least one valid recipient, each invalid recipient address will generate a "User unknown" message.

Action

Look in the sender's dead.letter file for the automatically saved message, and have the originator send it again, this time specifying a recipient.

`Reset tty pgrp from int to int`

Cause

The C shell sometimes issues this message when it clears away the window process group after the user exits the window system. This can happen when the window system doesn't clean up after itself.

Action

Proceed with your work. This message is only informational.

`Resource temporarily unavailable`

Cause

This indicates that the fork(2) system call failed because the system's process table is full, or that a system call failed because of insufficient memory or swap space. It is also possible that a user is not allowed to create any more processes.

Action

Simply waiting often gives the system time to free resources. However if this message occurs often on a system, reconfigure the kernel and allow more processes. To increase the size of the process table in Solaris 2, increase the value of maxusers in the /etc/system file. The default maxusers value is the amount of main memory in MB, minus 2.

If one user is not allowed to create any more processes, that user has probably exceeded the memory size limit; see the limit(1) man page for details.

Technical Notes

The symbolic name for this error is EAGAIN, errno=11.

`Restartable system call`

Cause

Interrupted system call should be restarted.

Technical Notes

The symbolic name for this error is ESTART, errno=91.

`Result too large`

Cause

This is a programming error or a data input error.

Action

Ask the program's author to fix this condition.

Technical Notes

This indicates an attempt to evaluate a mathematical programming function at a point where its value would overflow or underflow. The value of a programming function in the math package (3M) is not representable within machine precision. This could occur after floating point overflow or underflow (either single or double precision), or after total loss of numeric significance in Bessel functions.

This message can indicate "Result too small" in the case of floating point underflow.

To help pinpoint a program's math errors, use the matherr(3M) facility.

The symbolic name for this error is ERANGE, errno=34.

`rlogin: no directory! connection closed`

Cause

When user tries to remotely login to a machine, he gets the error.

The machine that the customer was trying to rlogin(1) to had permissions of 700 on its root directory. The permissions on root should be 755.

Once the user changed the rootpermissions to 755, he was able to get farther when attempting a rlogin, but it still failed:

Last login: Fri Aug 29 10:24:43 from machinename
no shell
connection closed

Action

The machine that the user was trying to rlogin into had the permissions set to 700 on both the root and /usr/bin directories. For both directories, the permissions should be 775. Once the user changed the permissions to 775, rlogin(1) was successful.

Another possibility is to check the user's >passwd(1) entry in the NIS/NIS+ map. A login shell such as /usr/dist/exe/tcsh or /net/lab/.../csh could cause the failure because of NFS mount permission.

Framing errors are a subset of CRC errors, which are usually caused by anomalies on the physical media. An "alignment/framing error" is a type of CRC error where octet boundaries do not line up.

"S"

`save: SYSTEM error, Arg list too long`

Cause

save fails with the error. The cause of this error is that the data base (index) file for the client is greater than 2 Gbytes. With Solaris 2.6 and SBU 5.0.1 this is no longer a problem.

Action

However, with earlier versions of Solaris you need to bring up nwadmin -> indexes -> select appropriate client -> select appropriate fs -> remove oldes cycle -> reclaim space.

This may have to be repeated a few times to reclaim enough space. The indexes can be recreated later if necessary by using scanner.

`SCSI bus DATA IN phase parity error`

Cause

The most common cause of this problem is unapproved hardware. Some SCSI devices for the PC market do not meet the high I/O speed requirements for the UNIX market. Other possible causes of this problem are improper cabling or termination, and power fluctuations. Data corruption is possible but unlikely to occur, because this parity error prevents data transfer.

Action

Check that all SCSI devices on the bus are Sun-approved hardware. Then verify that all cables measure no longer than six meters, total, and that all SCSI connections are properly terminated. If power fluctuations are occurring, invest in an uninterruptible power supply.

`SCSI transport failed: reason 'reset'`

Cause

This message indicates that the system sent data over the SCSI bus, but the data never reached its destination because of a SCSI bus reset. The most common cause of this condition is conflicting SCSI targets.¤ Data corruption is possible but unlikely to occur, because this failure prevents data transfer.

Action

Verify that all cables measure no longer than six meters, total, and that all SCSI connections are properly terminated. If power surges are a problem, acquire a surge suppressor or uninterruptible power supply.

A machine's internal disk drive is usually SCSI target 3. Make sure that external and secondary disk drives are targeted to 1, 2, or 0, and do not conflict with each other. Also make sure that tape drives are targeted to 4 or 5, and CD drives to 6, avoiding any conflict with each other or with disk drives. If the targeting of the internal disk drive is in question, power off the machine, remove all external drives, turn the power on, and from the PROM monitor run the probe-scsi-all or probe-scsi command.

If SCSI device targeting is acceptable, memory configuration could be the problem, especially for machines with the sun4c architecture. Ensure that high-capacity memory chips (such as 4MB SIMMs) are in lower banks, while lower-capacity memory chips (such as 1MB SIMMs) are in the upper banks.

SPARC systems do not always support third party CD-ROM drives, and may generate a similar "unknown vendor" error message. Check with the CD-ROM vendor for specific configuration requirements.

Some third-party disk drives have a read-ahead cache that interferes with Solaris device drivers. Make sure that any existing read-ahead cache facility is turned off.

`Security exception on host string. USER ACCESS DENIED.`

Cause

When trying to create a user via Adminsuite, placing the home directory on a system remote from the nisplus server, customer gets error message:

Security exception on host hostname. USER ACCESS DENIED.
The user identity (555)username was received, but that user
   is not authorized to execute the requested functionality
      on this system. Is this user a member of an appropriate
        security group on this system ?
            (Function: class directory method create_dir)

User can rsh(1) to the remote machine and create a home directory on the system.

Action

The user was not in the sysadmin group NIS+ tables.

 # niscat group.org_dir | grep sysadmin 
 sysadmin::14:

Add the username to the sysadmin group.

`Segmentation Fault`

Cause

Segmentation faults usually result from programming error. This message is usually accompanied by a core dump, except on read-only filesystems.

Action

To see which program produced a core(4) file, run either the file(1) command or the adb(1) command. The following examples show the output of the file(1) and adb(1) commands on a core file from the dtmail program.

$ file core
core: ELF 32-bit MSB core file SPARC Version 1, from `dtmail'

$ adb core
core file = core -- program `dtmail'
SIGSEGV  11: segmentation violation
^D      (use Control-d to quit the adb rogram)

Ask the vendor or author of this program for a debugged version.

Technical Notes

A process has received a signal indicating that it attempted to access an area of memory that is protected or that does not exist. The two most common causes of segmentation faults are attempting to dereference a null pointer or indexing past the bounds of an array.

`sendmail[]: can't lookup data via name server "dns"` or `sendmail[]: can't lookup data via name server "nis"`

Cause

The following entry in the /etc/nsswitch.conf file: sendmailvars: dns nis files causes the messages to appear in the console window.

Action

The sendmailvars database can only be used with local files and/or NIS+. So, if you do not have this database setup, the default sendmailvars entry should look as follows in the /etc/nsswitch.conf file:

sendmailvars: files

`sendmail[int]: NOQUEUE: SYSERR: net hang reading from string`

Cause

This is a sendmail(1M) message that appears on the console and in the log file /var/adm/messages. If this message occurs once for a particular user, it is possible that a mail message from this user ends with a partial line (having no terminating newline character). If this message appears frequently or at busy times, especially along with other networking errors, it could indicate network problems.

Action

Check the user's mail spool file to see if a message ends without a newline character. If so, talk with the user and determine how to prevent the problem from occurring again. If these messages are the result of network problems, you could try moving the mail spool directory to another machine with a faster network interface.

Technical Notes

During the SMTP receipt of DATA phase, a message-terminating period on a line of its own never arrived, so sendmail(1M) timed out and produced this error.

`Service wouldn't let us acquire selection`

	Your attempt to attach metastate database
	replicas on slice "c?t?d?s?" failed for the
	following reason: Slice c?t?d?s? is too small
	to contain 1 replicas.

This message from the SCSI tape drive appears when Archive tapes generate too many soft (recoverable) errors. It is followed by the advisory "Periodic head cleaning required and/or replace tape cartridge" message. Soft errors are an indication that hard errors could soon occur, causing data corruption.

Action

Technical Notes

The original vnode is no longer valid. The only way to get rid of this error is to force the NFS server and client to renegotiate file handles.

The symbolic name for this error is ESTALE, errno=151.

`statd: cannot talk to statd at string`

Cause

This message comes from the NFS status monitor daemon statd(1M), which provides crash recovery services for the NFS lock daemon lockd(1M). The message indicates that statd(1M) has left old references in the /var/statmon/sm and /var/statmon/sm.bak directories. After a user has removed or modified a host in the hosts database, statd(1M) might not properly purge files in these directories, which results in its trying to communicate with a nonexistent host.

Action

Remove the file named variable (where variable is the hostname) from both the /var/statmon/sm and /var/statmon/sm.bak directories. Then kill the statd daemon and restart it. If that doesn't get rid of the message, kill and restart lockd(1M) as well. If that doesn't work, reboot the machine at your convenience.

`stty: TCGETS: Operation not supported on socket`

Cause

This message results when a user tries to remote copy with rcp(1) or remote shell with rsh(1) from one machine to another, but has an stty(1) command in the remote .cshrc file. This error results in failure of the rcp(1) or rsh(1) command.

Action

The solution is to move the invokation of the stty(1) command to the user's .login (or equivalent) file. Alternatively, execute the stty(1) command in .cshrc only when the shell is interactive. Here is a test to do that:

if ($?prompt) stty ...

Technical Notes

The rcp(1) and rsh(1) commands make a connection using sockets, which do not support stty(1)'s TCGETS ioctl.

`su: No shell`

Cause

This message indicates that someone changed the default login shell for root to a program that is missing from the system. For example, the final colon-separated field in /etc/passwd could have been changed from /sbin/sh to /usr/bin/bash, which does not exist in that location. Possibly an extra space was appended at the end of the line. The outcome is that you cannot login as root or switch user to root, and so cannot directly fix this problem.

Action

The only solution is to reboot the system from another source, then edit the password file to correct this problem. Invoke sync(1M) several times, then halt the machine by typing Stop-A or by pressing the reset button. Reboot single-user from CD-ROM, the net, or diskette, such as by typing boot cdrom -s at the ok prompt.

After the system comes up and gives you a # prompt, mount the device corresponding to the original / partition somewhere, such as with a mount(1M) command similar to the one below. Then run an editor on the newly-mounted system password file (use ed(1) if terminal support is lacking):

# mount /dev/dsk/c0t3d0s0 /mnt
# ed /mnt/etc/passwd

Use the editor to change the password file's root entry to call an existing shell, such as /usr/bin/csh or /usr/bin/ksh.

Technical Notes

To keep the "No shell" problem from happening, habitually use admintool or /usr/ucb/vipw to edit the password file. These tools make it difficult to change password entries in ways that make the system unusable.

`SunPC may NOT run correctly as root`

Cause

With SunPC 4.1 and the 102924 jumbo patch installed: when a user attempts to run SunPC, the following error message is displayed:

SunPC may NOT run correctly as root.
Please run in user mode.
SunPC script is exiting

Yet, the user is not root.

The user's primary group id is probably root. For example:

$ /usr/bin/id
uid=33650(gruff) gid=0(root)

Action

Change the user's primary group to another group, such as 10, and, the user still needs to be in teh root group, add the root group to the user's secondary group list.

`su: 'su root' failed for string on /dev/pts/int`

Cause

Cause

The system reboots automatically. Afterward, the messages file contains System booting after fatal error FATAL.

The message is issued during a reboot after the system detects a hardware error. Things which can cause this are: UPA address parity error, Master queue overflows, DTAG parity errors, E-Cache tag parity errors, and Coherence errors.

Action

Use prtdiag(1M) to help identify failed hardware components. The errors indicate either have a bad CPU module or a bad system board.

`system hang`

Cause

4.1.3C SBUS cards suffer system freeze

`SYSTEM HANGS DURING BOOT`

Cause

When the user boots a system, it hangs after the boot up messages "root on," "swap on," and "dump on." After the system displays these messages, the LEDs will flash and the system hangs.

This is the result of an earlier fsck that deleted devices under the /dev directory. Check for the /dev/console device and if it is missing, make one.

"T"

`tar: /dev/rmt/0: No such file or directory`

Cause

The default tape device /dev/rmt/0, or possibly the device specified by the TAPE environment variable, is not currently connected to the system, is not configured, or its hardware symbolic link is broken.

Action

Action

First run fsck -n on the filesystem, to see how many and what type of problems exist. Then run fsck(1M) again to repair the filesystem. If you have a backup of the filesystem, you can generally answer "y" to all the fsck(1M) questions. It's a good idea to keep a record of all problematic files and inode numbers for later reference. To run fsck(1M) yourself, specify options as recommended by the boot script. For example:

# fsck /dev/rdsk/c0t4d0s0

# telnet firewall-machine
Trying 192.29.174.60 ...
Connected to firewall-machine
Escape character is '^]'.
CheckPoint FireWall-1 authenticated Telnet server running on
firewall-machine
Login: testuser
This gateway does not support Unix Password.

Action

Under Network Objects, edit your Gateway object, Host Properties Auth Schemes and select Unix Password. UNIX Password is not checked by default as it is considered an unsecure method of authentication.

`This mail file has been changed by another mail reader.`

Cause

This message appears in a pop-up dialog box whenever you start mailtool(1) while another mail reader has the inbox locked. A question follows: "Do you wish to ask that mail reader to save the changes?" You are given three choices.

Action

If you choose "Save Changes," mailtool(1) requests the other mail reader to relinquish its lock and write out any changes it has made to your inbox. If you choose "Ignore," mailtool(1) reads your inbox without locking it. If you choose "Cancel" mailtool(1) exits.

`Timeout waiting for ARP/RARP packet`

Cause

This problem can occur while booting from the net, and indicates a network connection problem.

Action

Make sure the Ethernet cable is connected to the network. Check that this system has an entry in the NIS ethers(4) map or locally on the boot server. Then check the IP address of the server and the client to make sure they are on the same subnet. Local /etc/hosts files must agree with one another and with the NIS hosts(4) map.

If those conditions are not causing the problem, go to the system's PROM monitor ok prompt and run test net to test the network connection. (On older PROM monitors, use test-net instead.) If the network test fails, check the Ethernet port, card, fuse, and cable, replacing them if necessary. Also check the twisted pair port to make sure it is patched to the correct subnet.

`Timer expired`

Cause

The timer set for a STREAMS ioctl call has expired. The cause of this error is device specific and could indicate either a hardware or software failure, or perhaps a timeout value that is too short for the specific operation. The status of the ioctl(2) operation is indeterminate. This is also returned in the case of _lwp_cond_timedwait(2) or cond_timedwait(3T).

Technical Notes

The symbolic name for this error is ETIME, errno=62.

`token ring hangs`

Cause

4.1.3C SBUS cards suffer system freeze

`Too many links`

Cause

Ultra system fails to boot with TRAP 3E. The system sometimes also displays bad magic number errors.

This is caused by a bad superblock on the boot disk. Which, in turn, could have beee caused by a SCSI configuration problem.

Action

To fix:

1. Check SCSI bus for illegal configuration, bad cables, and duplicate SCSI addresses;

2. Boot from cdrom in single user.

OK boot cdrom -sw

3. Attempt to fsck(1M) boot disk. This will probably fail with a superblock error.

# fsck /dev/rdsk/device

4. Find out locations of alternate superblocks. BE SURE TO USE AN UPPERCASE -N. For example:

# newfs -N /dev/rdsk/c0t0d0s0
/dev/rdsk/c0t0d0s0:     2048960 sectors in 1348 cylinders of 19 tracks, 
80 sectors 1000.5MB in 85 cyl groups (16 c/g, 11.88MB/g, 5696 i/g)
super-block backups (for fsck -F ufs -o b=#) at:
32, 24432, 48832, 73232, 97632, 122032, 146432, 170832, 195232, 219632,
244032, 268432, 292832, 317232, 341632, 366032, 390432, 414832, 439232,
463632, 488032, 512432, 536832, 561232, 585632, 610032, 634432, 658832,
683232, 707632, 732032, 756432, 778272, 802672, 827072, 851472, 875872,
900272, 924672, 949072, 973472, 997872, 1022272, 1290672, ...

5. Using an alternate superblock, fsck(1M) the disk. You may have to try more than one alternate superblock to get this to work. Pick a couple from the beginning, the middle, and the end.

# fsck -o b=<altblk> /dev/rdsk/c0t0d0s0

6. The boot block is probably bad too. Restore it while we are booted from the cdrom.

# /usr/sbin/installboot /usr/platform/architecture/lib/fs/ufs/bootblk /dev/rdsk/c0t0d0s0

7. Reboot the O.S. Should come up now.

# reboot

"U"

`ufsdump 4mm commands`

Cause

Dump syntax used with autoloader

`umount: warning: /string not in mnttab`

Cause

This message results from the superuser attempting to unmount a filesystem that is not mounted. Subdirectories of filesystems, such as /var, cannot be unmounted.

Action

Run the mount(1M) or df(1M) command to see what filesystems are mounted. If you really want to unmount one of them, specify the existing mount point.

`Unable to connect to license server. Inconsistent encryption code.`

Cause

User receives the following error message. The only thing that had changed was the IP address of the machine.

Action

The IP address defined with ifconfig(1M) must match that in /etc/hosts. That is, if you change the machine's IP address with ifconfig(1M) you must also change the machine's entry in the /etc/hosts file.

For machines with multiple interfaces, you must check and possibly update /etc/hostname.*.

`unable to get pty!`

Cause

When trying to bring up a Terminal window (dtterm) in CDE, a popup appears stating: Unable to get pty!.

Action

This error is because dtterm is not able to open /dev/pts/int (where int is an integer). The reason they cannot open this file is grantpt(3C) failed to change the permissions on the file. grantpt(3C) failed because the binary /usr/lib/pt_chmod is not setuid root. The permissions on /usr/lib/pt_chmod must be 4111.

To restore the correct permissions to pt_chmod, use the following command: (as root)

 
# chmod 4111 /usr/lib/pt_chmod

`Unable to install/attach driver 'string'`

Cause

These messages appear in /var/adm/messages at boot time, when the system tries to load drivers for devices the machine does not have.

Action

Despite the alarmist tone, this message is intended as only informational. You probably don't want all these device drivers, because they make your system kernel larger, requiring more memory.

`Unable to open nwrecover, Error: nwrecover: NSR: please start a server on client_name`

Cause

While trying to open the graphical recovery interface by running nwrecover from the client, the error was displayed.

Action

In this case, multiple networker servers existed and nwrecover could not determine which network server to use for the client.

The server can be specified to the nwrecover command with the -s option.

	nwrecover -c client_name -s server_name

Check for typo the entered e-mail address. Or, the user could be aliased to a non-existant e-mail address in /etc/mail/aliases or the user's .mailrc file.

You cannot mail to a user that has capital letters in its name. sendmail(1M) converts all the capital letters to lowercase before attempting to find the user. Since UNIX is case sensitive, it finds no username on the system with all lowercase letters, so it displays the User unknown message.

Workaround: Make sure all usernames are composed of *only* lowercase letters.

`/usr/dt/bin/rpc.ttdbserverd:Child Status' changed`

Cause

While running CDE, the error in the console or /var/adm/messages file:

Oct 19 04:41:00 darkcastle last message repeated 393 times
Oct 19 04:41:01 darkcastle inetd[120]: /usr/dt/bin/rpc.ttdbserverd:Child Status Changed

Action

The fix is to create the following soft links:

ln -s /usr/openwin/bin/rpc.ttdbserver /usr/dt/bin/rpc.ttdbserver
ln -s /usr/openwin/bin/rpc.ttdbserverd /usr/dt/bin/rpc.ttdbserverd

`/usr/openwin/bin/xinit: connection to X server lost`

Cause

This means that the xinit(1) program, which sets up X11 resources and starts a window manager, failed to locate the X server process. Perhaps the user interrupted window system startup, or exited abnormally from OpenWindows (for example, by killing processes or by rebooting). It is possible that the X server crashed. Data loss is possible in some cases. Depending on process timing, this message might be normal when OpenWindows exits during a system reboot.

Action

The only solution is to exit and restart OpenWindows. You do not need to reboot the system unless it hangs and fails to give you a console prompt. To exit OpenWindows, select Workspace->Exit. To restart OpenWindows, type openwin(1) at the system prompt.

`/usr/ucb/cc: language optional software package not installed`

Cause

When compiling some code for BSD compatibility the error happened after invoking usr/ucb/cc. The unbundled compiler (SPARCworks Professional C product) was installed in /opt.

/usr/ucb/cc is a script which checks for the file /usr/ccs/bin/ucbcc and, if found, invokes it with appropriate library flags for BSD-compatibility compilation.

/usr/ucb/cc is part of the package SUNWscpu. /usr/ccs/bin/ucbcc is supposed to be a symbolic link to /opt/SUNWspro/bin/acc, which is created during installation of the unbundled C compiler, SPROcc.

Action

Verify that you have the essential OS-bundled Developer packages, SUNWscpu, SUNWbtool, and the unbundled C compiler:- SPROcc. However, /usr/ccs/bin/ucbcc was missing on customer's system. Evidently somewhere along the line this link was removed.

Solve the problem by creating a new symbolic link:

   # ln -s /opt/SUNWspro/bin/acc /usr/ccs/bin/ucbcc

Invoke usr/ucb/cc to verify this worked.

Commands used to identify which packages contain particular components involved:

craterlake% grep ucb/cc /var/sadm/install/contents
/usr/ucb/cc f none 0555 bin bin 3084 50323 814621113 *SUNWscpu
craterlake% ls -l /usr/ucb/cc
-r-xr-xr-x   1 bin      bin         3084 Oct 25  1995 /usr/ucb/cc
craterlake% file !$
file /usr/ucb/cc
/usr/ucb/cc:    executable /usr/bin/sh script
craterlake% grep ucbcc /var/sadm/install/contents
/usr/ccs/bin/ucbcc=/opt1/40/SUNWspro/SC4.0/bin/acc s none SPROcc SPROcc.2 SPROcc.5
craterlake% file /usr/ccs/bin/ucbcc
/usr/ccs/bin/ucbcc:     ELF 32-bit MSB executable SPARC Version 1, dynamically linked, stripped
craterlake% ls -l /usr/ccs/bin/ucbcc
lrwxrwxrwx   1 root     other         31 Aug 23  1996 /usr/ccs/bin/ucbcc -> /opt1/40/SUNWspro/SC4.0/bin/acc

`UX: userdel: error: Cannot update system files login cannot be deleted`

Cause

When using userdel to delete a user:

userdel -r userid

If the root (/) filesystem is full, the error is displayed.

Action

Free up space on root (/) filesystem.

"V"

`Value too large for defined data type`

Cause

The user ID or group ID of an IPC object or file system object was too large to be stored in an appropriate member of the caller-provided structure.

Action

Run the application on a newer system, or ask the program's author to fix this condition.

Technical Notes

This error occurs only on systems that support a larger range of user or group ID values than a declared member structure can support. This condition usually occurs because the IPC or file system object resides on a remote machine with a larger value of type uid_t, off_t, or gid_t than that of the local system.

The symbolic name for this error is EOVERFLOW, errno=79.

`Volume Manager reports error:`

Cause

After upgrading from VxVM 2.0 or 2.1 to 2.3, when attempting to run vxva, the volume manager GUI, you get the message:

        Volume Manager reports error:
           Configuration daemon can't speak protocol version

This message indicates that there is a version mismatch between the version of the volume manager daemon (vxconfigd) and the GUI (vxva) you are trying to run. For example, you are running the 2.3 version of vxconfigd, and trying to run an old (2.1) version of vxva.

The most likely reason for this to happen is because the you are using the wrong path for vxva. For vxva versions 2.1 and below, the binary was found in /opt/vxva/bin, but starting with 2.1.1, the location was changed to /opt/SUNWvxva/bin.

If you did not remove the old SUNWvxva package before installing the new 2.3 version (which is a normal thing; you don't NEED to remove the old package), you probably still have the old /opt/vxva/bin in your $PATH, and thus, you're attempting to run the older version of vxva.

Action

Run the newer vxva program: /opt/SUNWvxva/bin/vxva. If that does work and you do not get the error message, simply remove /opt/vxva/bin/vxva from your path statement or remove the old version of vxva and create a symbolic link to the new version with the following two commands:

 
# rm /opt/vxva/bin/vxva 
# ln -s /opt/SUNWvxva/bin/vxva /opt/vxva/bin/vxva

`vxconfigd error: segmentation fault`

Cause

When the system boots, the vxconfid fails to start. It fails with a segmentation fault (core dump).

	vxconfigd error: segmentation fault
	[ vxvm warning: _illegal vminor encountered ]

Action

Check the date on the system using date(1) (/bin/date or /usr/bin/date). If the date on the system is very old (like 1970) or very far out in the future (like 2010), vxconfigd core dumps.

Change the date on the system using /bin/date or /usr/bin/date and the vxconfigd will start like a champ.

`vxvm:vxslicer:ERROR unsupported disk layout`

Cause

When trying to encapsulate a disk you get this error.

Action

You must meet the minimum requirements to encapsulate a disk:

1. You must have two free, zero length, slices on the disk (no cylinders should be assigned to these slices.)

2. You must have two free cylinders on the disk. These 2 cylinders must not be in use by any slice other than slice two.

3. The two free cylinders must be located at the beginning or end of the drive.

"W"

`WARNING: add_spec: No major number for sf`

Cause

The system prints the following warning message while booting:

        SunOS Release 5.5.1 Version Generic_103640-03 [UNIX(R)
        System V Release 4.0]
        Copyright (c) 1983-1996, Sun Microsystems, Inc.
        WARNING: add_spec: No major number for sf

The sf(7D) driver is specific for a Sun Enterprise Network Array (SENA), also known as a "photon".

Action

Action

To reduce the frequency of this message, add this line near the bottom of the /etc/system file and reboot:

set esp:esp_use_poll_loop=0

Action

See "/dev/rdsk/variable: UNEXPECTED INCONSISTENCY" for information about repairing UFS filesystems.

See "THE FOLLOWING FILE SYSTEM(S) HAD AN UNEXPECTED INCONSISTENCY" for information about repairing non-UFS filesystems.

`Watchdog Reset`

Cause

This fatal error usually indicates some kind of hardware problem. Data corruption on the system is possible.

Action

Look for some other message that might help diagnose the problem. By itself, a watchdog reset doesn't provide enough information; because traps are disabled, all information has been lost. If all that appears on the console is an ok prompt, issue the PROM command below to view the final messages that occurred just before system failure:

ok f8002010 wector p

Yes, that word is wector, not vector.

The result is a display of messages similar to those produced by the dmesg(1M) command. These messages can be useful in finding the cause of system failure.

Technical Notes

This message doesn't come from the kernel, but from the OpenBoot PROM monitor, a piece of Forth software that gives you the ok prompt before you boot UNIX. If the CPU detects a trap when traps are disabled (an unrecoverable error), it signals a watchdog. The OpenBoot PROM monitor detects the watchdog, issues this message, and brings down the system.

`Who are you?`

Cause

Many networking programs can print this message, including from(1B), lpr(1B), lprm(1B), mailx(1), rdist(1), sendmail(1M), talk(1), and rsh(1). The command prints this message when it cannot locate a password file entry for the current user. This might occur if a user logged in just before the superuser deleted that user's password entry, or if the network naming service fails for a user who has no entry in the local password file.

Action

If a user's password file entry was accidentally deleted, restore it from backups or from another password file. If a user's login name or user ID was changed, ask that user to logout and login again. If the network naming service failed, check the NIS server(s) and repair or reboot as necessary.

Technical Notes

There is a known problem (bug 1138025) with starting hundreds of rsh(1) processes on another machine. This message appears because rsh(1) hangs while binding to a reserved port, and responds too slowly to interact with the network naming service.

`Window Underflow`

Cause

This message often occurs at boot time, sometimes along with a "Watchdog Reset" error. It comes from the OpenBoot PROM monitor, which was passed a processor trap from the hardware. This error indicates that some program tried to access a SPARC register window that wasn't accessible from the processor.

Action

On some system architectures, specifically sun4c, the problem could be that different capacity memory chips are mixed together. Someone might have placed 1MB SIMMs in the same bank with 4MB SIMMs. If this is so, rearrange the memory chips. Make sure to put higher-capacity SIMMs in the first bank(s), and lower-capacity SIMMs in the remaining bank(s); never mix different capacity SIMMs in the same bank.

The problem could also be that cache memory on the motherboard has gone bad and needs replacement. If main memory is installed correctly, try swapping the motherboard.

Technical Notes

The best way to isolate the problem is to look at the %pc register to see where it got its arguments, and why the arguments were bad. If you can reproduce the condition causing this message, your system vendor might be able to help diagnose the problem.

"X"

`Xlib: extension "GLX" missing on display "0.0"`

Cause

Install OpenGL 1.0 and test the configuration by running /usr/openwin/demo/GL/ogl_install_check which results in following:

# ./ogl_install_check
    Xlib:  extension "GLX" missing on display "0.0".
    Xlib:  extension "GLX" missing on display "0.0".
    Xlib:  extension "GLX" missing on display "0.0".
    can't find visual

Action

First check that the installation has worked correctly by running the package check utility on the runtime package: # pkgchk SUNWglrt. This should result in an error message such as:

	ERROR: /usr/openwin/server/etc/OWconfig
	    file size <187> expected <5423> actual
	    file cksum <14394> expected <27045> actual

(The numbers might be different but there should be only 1 file.) If other errors result, re-install OpenGL, especially the SUNWglrt package.

Assuming that is fine, look at the process owner for the Xsun process using:

 
# ps -aef | grep Xsun | grep -v grep
nobody 20022   225  0 11:36:22 ?        0:34 /usr/openwin/bin/Xsun :0 -nobanner

If the owner is not root, that is most likely the problem since there will be a permission issue loading the graphic pipelines.

If you are are using CDE, ensure that the Xservers file has the form

:0 Local local_uid@console root /usr/openwin/bin/Xsun :0 -nobanner

The Xservers file will be found in /usr/dt/config if you haven't done any customization but more likely /etc/dt/config/. Additional arguments after the -nobanner option are acceptable.

Another way of proving this is to run OpenWindows from the command line as root. It will ensure that the Xsun process is owned by root.

Another possibility is that the system is NOT a Creator 3D. You can only run openGL 1.0 on an Ultra with a Creator 3D graphics card. If you install this application on an Ultra with a Creator framebuffer and NOT a Creator 3D you'll see these same error messages.

`xterm: fatal IO error 32 (Broken Pipe) or KillClient on X server "string:0.0"`

Cause

This means that xterm(1) has lost its connection to the X server. The "0.0" represents the display device, which is usually the console. This message can appear when a user is running xterm and the X server disappears or the client gets shut down. Data loss is possible if applications were killed before saving files.

Action

Try to run the terminal emulator again in a few minutes after the system has rebooted and the window system is running.

`XView warning: Cannot load font set 'string' (Font Package)`

Cause

This message from the XView library warns that a requested font is not installed on the X server. Often multiple warnings are displayed for the same font. The set of available fonts can vary from release to release.

Action

To see which fonts are available on the X server, run the xlsfonts(1) program. Then specify another font name that you see in the output of xlsfonts(1). Sometimes it is possible to locate a similar font from a different vendor.

Technical Notes

There are two packages of X windows fonts: the common but not required fonts (SUNWxwcft), and the optional fonts (SUNWxwoft). Run pkginfo(1) to see if both packages are installed, and add them to the system as you want.

"Y"

`yp_all RPC clnt_call (transport level) failure`

Cause

At random times, a slave NIS server has a problem that causes ypbind(1M) to report: ypserver not responding and the machine must be rebooted. The syslog contains:

Dec 14 07:11:03 rahab syslog: yp_all - RPC clnt_call (transport level) failure:
RPC: Unable to receive; An event requires attention

Action

To workaround, increase the file descriptor limit in the yp startup script, /etc/rc2.d/S71rpc. Add this to the script before ypserv is started:

	ulimit -n 256

`ypbind[int]: NIS server for domain "string" OK`

Cause

This message appears after an "NIS server not responding" message to indicate that ypbind(1M) is able to communicate with an NIS server again.

Action

Proceed with your work. This message is purely informational.

`ypbind[int]: NIS server not responding for domain "string"; still trying`

Cause

This means that the NIS client daemon ypbind(1M) cannot communicate with an NIS server for the specified domain. This message appears when a workstation running the NIS naming service has become disconnected from the network, or when NIS servers are down or extremely slow to respond.

Action

If other NIS clients are behaving normally, check the Ethernet cabling on the workstation that is getting this message. On SPARC machines, disconnected network cabling also produces a series of "no carrier" messages. On x86 machines, the above message might be your only indication that network cabling is disconnected.

If many NIS clients on the network are giving this message, go to the NIS server in question and reboot or repair as necessary. To locate the NIS server for a domain, run the ypwhich(1) command. When the server machine comes back in operation, NIS clients give an "NIS server for domain OK" message.

`ypwhich: can't communicate with ypbind`

Cause

This message from the ypwhich(1) command indicates that the NIS binder process ypbind(1M) is not running on the local machine.

Action

If the system is not configured to use NIS, this message is normal and expected. Configure the system to use NIS if necessary.

If the system is configured to use NIS, but the ypbind(1M) process is not running, invoke the following command to start it up:

# /usr/lib/netsvc/yp/ypbind -broadcast

"Z"

`zsint: silo overflow`

Cause

This message means that the Zilog 8530 character input silo (or serial port FIFO) overflowed before it could be serviced. The zs(7D) driver, which talks to a Zilog Z8530 chip, is reporting that the FIFO (holding about two characters) has been overrun. The number after zs(7D) shows which serial port experienced an overflow:

zs0 - tty serial port 0 (/dev/ttya)
zs1 - tty serial port 1 (/dev/ttyb)
zs2 - keyboard port (/dev/kbd)
zs3 - mouse port (/dev/mouse)

Action

Silo overflows indicate that data in the respective serial port FIFO has been lost. However, consequences of silo overflows might be negligible if the overflows occur infrequently, if data loss is not catastrophic, or if data can be recovered or reproduced. For example, although a silo overflow on the mouse driver (zs3) indicates that the system could not process mouse events quickly enough, the user can perform mouse motions again. Similarly, lost data from a silo overflow on a serial port with a modem connection transferring data using uucp(1C) is recovered when uucp(1C) discovers the loss of data and requests retransmission of the corrupted packet.

Frequent silo overflow messages can indicate a zs(7D) hardware FIFO problem, a serial driver software problem, or abnormal data or system activity. For example, the system ignores interrupts during system panics, so mouse and keyboard activity result in silo overflows.

If the serial ports experiencing silo overflows are not being used, a silo overflow could indicate the onset of a hardware problem.

Technical Notes

Another type of silo overflow is one that occurs during reboot when an HDLC line is connected to any of the terminal ports. For example, an X.25 network could be sending frames before the kernel has been told to expect them. Such overflow messages can be ignored.

Previous: Chapter 1 About Error Messages

Chapter 2 Alphabetical Message Listing

Numbers and Symbols

***** FILE SYSTEM WAS MODIFIED *****

Cause

Action

** Phase 1-- Check Blocks and Sizes

Cause

Action

See Also

** Phase 1b-- Rescan For More DUPS

Cause

Action

See Also

** Phase 2-- Check Pathnames

Cause

Action

See Also

** Phase 3-- Check Connectivity

Cause

Action

See Also

** Phase 4-- Check Reference Counts

Cause

Action

See Also

** Phase 5-- Check Cyl groups

Cause

Action

See Also

29a00 illegal instruction

Cause

Action

451 timeout waiting for input during source

Cause

Action

See Also

550 hostname... Host unknown

Cause

Action

Technical Notes

See Also

550 username... User unknown

Cause

Action

See Also

554 hostname... Local configuration error

Cause

Action

Technical Notes

See Also

"A"

Accessing a corrupted shared library

Cause

Technical Notes

access violation unknown host IP address

Cause

Action

A command window has exited because its child exited.

Cause

Action

Technical Notes

Address already in use

Cause

Technical Notes

Address family not supported by protocol family

Cause

Technical Notes

admintool: Received communication service error 4

Cause

Action

Advertise error

Cause

Technical Notes

answerbook: XView error: NULL pointer passed to xv_set

Cause

Action

See Also

Arg list too long

Cause

Action

`* FILE SYSTEM WAS MODIFIED *`

`** Phase 1-- Check Blocks and Sizes`

`** Phase 1b-- Rescan For More DUPS`

`** Phase 2-- Check Pathnames`

`** Phase 3-- Check Connectivity`

`** Phase 4-- Check Reference Counts`

`** Phase 5-- Check Cyl groups`

`29a00 illegal instruction`

`451 timeout waiting for input during` `source`

`550 hostname... Host unknown`

`550 username... User unknown`

`554 hostname... Local configuration error`

`Accessing a corrupted shared library`

`access violation unknown host IP address`

`A command window has exited because its child exited.`

`Address already in use`

`Address family not supported by protocol family`

`admintool: Received communication service error 4`

`Advertise error`

`answerbook: XView error: NULL pointer passed to xv_set`

`Arg list too long`

`Argument out of domain`

`Arguments too long`

`assertion failed: string, file name, line int`

`Attempting to link in more shared libraries than system limit`

`automountd[int]: server hostname responding`

`automount[int]: name: Not a directory`

`Bad address`

`BAD/DUP FILE I=i OWNER=o MODE=m SIZE=s MTIME=t CLEAR?`

`Bad file number`

`block no. BAD I=inode no.`

`BAD_MESSAGE (error code 100) from X.400`

`bad module/chip at: position`

`Bad request descriptor`

`BAD SUPER BLOCK: string`

`BAD TRAP`

`/bin/sh: file: too big`

`Block device required`

`Boot device: /iommu/sbus/directory/directory/sd@3,0`

`boot error 13 | sun4m`

`Broadcast Message from root (pts/int) on server [date]`