RHBA-2005:451 - Bug Fix Advisory

Topic

Updated netdump packages that fix a "service netdump start" bug are now
available.

Description

The netdump client sets up the kernel to send crash dumps to a remote
system. Some manual setup is required in order to set up secure dumps. The
netdump server listens to the network for crashed kernels to contact it and
then writes the log and memory dump to /var/crash before asking the crashed
machine to reboot.

If a user manually executes consecutive "service netdump start" commands
without an intervening "service netdump stop" command, the system loses its
netdump capability.

Upon running "service netdump start", the /etc/init.d/netdump script:

(1) generates a random magic number
(2) sends it to the configured netdump-server
(3) passes it to the netconsole module as an argument

The netdump module uses this magic number during subsequent communications
with the netdump-server for validation. If "service netdump start" is run
a second time without an intervening "service netdump stop", a new magic
number is generated, and sent to the netdump-server. Since the netconsole
module is already running as a result of the first "service netdump start",
the modprobe of the netconsole module fails. If a subsequent panic were to
occur, the netconsole module uses the first obsolete magic number, and the
netdump-server rejects the netdump attempt.

In this updated package, a "service netdump start" now stops any currently
running netconsole modules, so that a new magic number can be properly
passed to the newly-invoked netconsole module.

If a user does not specify a MAC address, and the netdump server resides on
a different subnet from the client, then netdump may fail to operate
correctly. This limitation has been removed in this update. Now, the
netdump client attempts to find the first "hop" between the client and
the server, and use the MAC address for this router automatically. This
results in less manual configuration steps necessary to set up a netdump
client.

Users of the netdump package should upgrade to this updated package, which
resolves this issue.

Solution

Before applying this update, make sure that all previously-released
errata relevant to your system have been applied. Use Red Hat
Network to download and update your packages. To launch the Red Hat
Update Agent, use the following command:

up2date

For information on how to install packages manually, refer to the
following Web page for the System Administration or Customization
guide specific to your system:

http://www.redhat.com/docs/manuals/enterprise/

Affected Products

• Red Hat Enterprise Linux Server 3 x86_64
• Red Hat Enterprise Linux Server 3 ia64
• Red Hat Enterprise Linux Server 3 i386
• Red Hat Enterprise Linux Workstation 3 x86_64
• Red Hat Enterprise Linux Workstation 3 ia64
• Red Hat Enterprise Linux Workstation 3 i386
• Red Hat Enterprise Linux Desktop 3 x86_64
• Red Hat Enterprise Linux Desktop 3 i386
• Red Hat Enterprise Linux for IBM z Systems 3 s390x
• Red Hat Enterprise Linux for IBM z Systems 3 s390
• Red Hat Enterprise Linux for Power, big endian 3 ppc

Fixes

• BZ - 90803 - /etc/init.d/netdump start script requires client to be on same subnet as server

CVEs

(none)

References

(none)