Technical article 102444

NFS panics with "PANIC: issue_strat - Unknown device 255/255."

OpenServer 5.0.0OpenServer 5.0.2OpenServer 5.0.4

Written · Last revised

Problem

SCO OpenServer Enterprise System 5.0 panics with the message:

PANIC: issue_strat - Unknown device 255/255

Trying to dump 8095 pages to dumpdev hd (1/41) at block 0, 102 pages per '.'

PANIC String: issue_strat - Unknown device %u/%u

This problem has been reported on SCO OpenServer 5.0.2 with Support Level Supplement (SLS) OSS449A installed. It has also been observed on SCO OpenServer Release 5.0.0 installed with the Networking Supplement 1.0.0 and SLS OSS449A, and SCO OpenServer Release 5.0.4 installed with the Release Supplement 504c (rs504c).

The panic, however, is not seen on all SCO OpenServer 5.0 systems and only occurs under certain conditions. The panic occurs when a single mount of an external filesystem is attempted, or when it has been attempted from a system with a large number (over 255) of incoming NFS connections.

Cause

When NFS is initially installed on an SCO OpenServer system, and the
default number of NFS connections of 256 is set, an entry is created
in the system file "/etc/conf/cf.d/mdevice" which, on rebuild of the
kernel, creates the device files "/dev/nfsd" and "/dev/nfsclnt". If
the number of NFS connections is increased by 256 from the default
setting, NFS uses "extended minor" numbers to handle the extra
devices needed for client connections and creates additional entries
in the system "mdevice" file.  For example:
crw-rw-rw-   1 root  sys    80,  1 Jan 15 17:33 /dev/nfsclnt
crw-rw-rw-   1 root  sys    80,  0 Jan 15 17:33 /dev/nfsd

creates an entry in /etc/conf/cf.d/mdevice as follows:

...
...
nfsd  oci  icb  nfsd    80      80      0       256     -1
...
nfsd  oci  icbM nfsd    87      87      80      256     -1
...
...
                        ^^              ^^
A               B

See mdevice(F) for a detailed explanation of the various fields. In the example above, the block major number of the extended minor entry, with the "M" flag, is set to 87. This happens to be greater in value than the base block major number, that is, the block major number of the first non-extended minor device, set at a value of 80. When NFS is creating extended minor entries, it searches for available major numbers on the system to use. In the example above, the value it found at 87 happens to be greater than 80, A > B.

If, however, the settings had been created as follows, a panic will occur after 255 NFS connections have been made to the system. For example:

...
nfsd  oci  icb  nfsd    87      87      0       256     -1
...
nfsd  oci  icbM nfsd    80      80      87      256     -1
...
                        ^^              ^^
A               B

On kernel rebuild these entries will create the device files as:

crw-rw-rw-   1 root    sys   87,  1 Jan 15 17:50 /dev/nfsclnt
crw-rw-rw-   1 root    sys   87,  0 Jan 15 17:50 /dev/nfsd

Another example where the system has been configured for 1024 connections is as follows:

...
...
nfsd  oci  icb  nfsd    81      81      0       256     -1
...
nfsd  oci  icbM nfsd    77      77      81      256     -1
nfsd  oci  icbM nfsd    78      78      81      512     -1
nfsd  oci  icbM nfsd    79      79      81      768     -1
^^              ^^
A               B

In both of the latter examples, A is less than B. This setting causes a problem in other kernel code in the operating system. These settings may not occur on all systems; it depends on what major numbers are available on the system when NFS is being configured.

Solution

This problem has been reported to SCO Engineering.

If the number of NFS connections is first set to the default number of 256 connections within the scoadmin(ADM) network Configuration Manager, and then is progressively increased by 256 with a kernel relink and rebuild at each stage, this problem does not appear.

Alternately, if the system has already been configured at 1024 NFS connections and the panic occurs, and the mdevice entries look as per the examples above, the major numbers may be switched around so that A is greater than B and the nfsd devices are rebuilt.

For example, the settings above can be switched to:

...
nfsd  oci  icb  nfsd    77      77      0       256     -1
...
nfsd  oci  icbM nfsd    78      78      77      256     -1
nfsd  oci  icbM nfsd    79      79      77      512     -1
nfsd  oci  icbM nfsd    81      81      77      768     -1
^^              ^^
A               B

That is, A has been set greater than B by switching the major device numbers from the extended minor devices to the major device entry.

The original device entries must be removed by executing this command:

# rm /dev/nfs*

The kernel must be relinked and rebuilt to create the new devices.

It has also been observed that this panic can occur when extended minor numbers have not been configured for use. In which case, using netconfig(ADM) to remove and re-add the NFS protocol has been known to resolve this problem.

NOTE:     Use the command line options of the configure(ADM) utility to edit
          the system mdevice file. Keep a record or create backups of the
          device entries and the "/etc/conf/cf.d/mdevice" file.

See also

mdevice(F), configure(ADM).

From the Xinuos technical article database, record 102444. Reproduced with permission. The original is at www.sco.com/ta/102444, which is reachable but blocked from search engines. Xinuos marks this article against a retired product line. That refers to the product’s lifecycle, not to whether the advice still works — plenty of it still applies to running systems.

Next step

Need this fixed rather than explained?

Tell us the release, what the system runs and what it is doing. We will reply by email with what we would do next.

  • You deal with the people who do the work
  • No obligation, and no upgrade required before we will help
  • We tell you plainly if something is not worth doing

Send an enquiry

We reply by email. If a server is down, say so in the first line. You can also write to help@scoexpert.com.