nfs security update

CentOS Errata and Security Advisory CESA-2009:0955

The nfs-utils package provides a daemon for the kernel NFS server and
related tools, which provides a much higher level of performance than the
traditional Linux NFS server used by most users.

A flaw was found in the nfs-utils package provided by RHBA-2008:0742. The
nfs-utils package was missing TCP wrappers support, which could result in
an administrator believing they had access restrictions enabled when they
did not. (CVE-2008-1376)

This update also includes the following bug fixes:

• the “nfsstat” command now displays correct statistics. In previous
  versions, performing more than 2^31 RPC calls could cause the “nfsstat”
  command to incorrectly display the number of calls as “negative”. This was
  because “nfsstat” printed statistics from /proc/net/rpc/* files as signed
  integers; with this version of nfs-utils, “nfsstat” now reads and prints
  these statistics as unsigned integers. (BZ#404831)

• imapd upcalls now support zero-length reads and perform extra bounds
  checking in gssd and svcgssd. This fixes a bug in previous versions that
  could cause the rpc.imapd daemon to hang when communicating with the
  kernel, which would halt any ID translation services. (BZ#448710)

• tcp_wrappers supported in nfs-utils now allows proper application of
  hosts access rules defined in /etc/hosts.allow and /etc/hosts.deny. (BZ#494585)

• the nfs init script did not check whether SECURE_NFS was set to “yes”
  before starting, stopping, or querying rpc.svcgssd. On systems where
  SECURE_NFS was not set to “yes”, the nfs init script could not start the
  rpc.svcgssd daemon at the “service nfs start” command because the rpcsvcssd
  init script would check the status of SECURE_NFS before starting the
  daemon. However, at the “service nfs stop” or “service nfs restart”
  commands, nfs init script would attempt to stop rpc.svcgssd and then report
  a failure because the daemon was not running in the first place. These
  error messages may have misled end-users into believing that there was a
  genuine problem with their NFS configuration. This version of nfs-utils
  contains a fix backported from Red Hat Enterprise Linux 5. nfs-utils now
  checks the status of SECURE_NFS before the nfs init script attempts to
  start, query or stop rpc.svcgssd and therefore, the irrelevant error
  messages seen previously will not appear. (BZ#470423)

• the nfs init script is now fully compliant with Linux Standard Base Core
  specifications. This update fixes a bug that prevented “/etc/init.d/nfs
  start” from exiting properly if NFS was already running. (BZ#474570)

• /var/lib/nfs/statd/sm is now created with the proper user and group
  whenever rpc.statd is called. In previous versions, some thread stack
  conditions could incorrectly prevent rpc.statd from creating the
  /var/lib/nfs/statd/sm file, which could cause “service nfslock start” to
  fail. (BZ#479376)

All users of nfs-utils should upgrade to this updated package, which
resolves these issues.

Merged security bulletin from advisories:
https://lists.centos.org/pipermail/centos-announce/2009-May/078047.html
https://lists.centos.org/pipermail/centos-announce/2009-May/078058.html

Affected packages:
nfs-utils

Upstream details at:
https://access.redhat.com/errata/RHSA-2009:0955