File Systems and Non-Global Zones

This section provides information about file system issues on an Oracle Solaris system with zones installed. Each zone has its own section of the file system hierarchy, rooted at a directory known as the zone root. Processes in the zone can access only files in the part of the hierarchy that is located under the zone root. The chroot utility can be used in a zone, but only to restrict the process to a root path within the zone. For more information about chroot, see chroot(1M).

The -o nosuid Option

The -o nosuid option to the mount utility has the following functionality:

• Processes from a setuid binary located on a file system that is mounted using the nosetuid option do not run with the privileges of the setuid binary. The processes run with the privileges of the user that executes the binary.

  For example, if a user executes a setuid binary that is owned by root, the processes run with the privileges of the user.

• Opening device-special entries in the file system is not allowed. This behavior is equivalent to specifying the nodevices option.

This file system-specific option is available to all Oracle Solaris file systems that can be mounted with mount utilities, as described in the mount(1M) man page. In this guide, these file systems are listed in Mounting File Systems in Zones. Mounting capabilities are also described. For more information about the -o nosuid option, see “Accessing Network File Systems (Reference)” in Oracle Solaris Administration: Network Services.

Mounting File Systems in Zones

When file systems are mounted from within a zone, the nodevices option applies. For example, if a zone is granted access to a block device (/dev/dsk/c0t0d0s7) and a raw device (/dev/rdsk/c0t0d0s7) corresponding to a UFS file system, the file system is automatically mounted nodevices when mounted from within a zone. This rule does not apply to mounts specified through a zonecfg configuration.

Options for mounting file systems in non-global zones are described in the following table. Procedures for these mounting alternatives are provided in Configuring, Verifying, and Committing a Zone and Mounting File Systems in Running Non-Global Zones.

Any file system type not listed in the table can be specified in the configuration if it has a mount binary in /usr/lib/fstype/mount.

To mount file system types other than HSFS and NFS from inside the non-global zone, also add the file system type to the configuration by using the zonecfg fs-allowed property.

Allowing file system mounts other than the default might allow the zone administrator to compromise the system.

global# pkg -R /tank/zones/my-zone/root \
install system/file-system/ufs

For more information, see How to Configure the Zone, Mounting File Systems in Running Non-Global Zones, and the mount(1M) man page.

Unmounting File Systems in Zones

The ability to unmount a file system will depend on who performed the initial mount. If a file system is specified as part of the zone's configuration using the zonecfg command, then the global zone owns this mount and the non-global zone administrator cannot unmount the file system. If the file system is mounted from within the non-global zone, for example, by specifying the mount in the zone's /etc/vfstab file, then the non-global zone administrator can unmount the file system.

Security Restrictions and File System Behavior

There are security restrictions on mounting certain file systems from within a zone. Other file systems exhibit special behavior when mounted in a zone. The list of modified file systems follows.

AutoFS

Autofs is a client-side service that automatically mounts the appropriate file system. When a client attempts to access a file system that is not presently mounted, the AutoFS file system intercepts the request and calls automountd to mount the requested directory. AutoFS mounts established within a zone are local to that zone. The mounts cannot be accessed from other zones, including the global zone. The mounts are removed when the zone is halted or rebooted. For more information on AutoFS, see How Autofs Works in Oracle Solaris Administration: Network Services.

Each zone runs its own copy of automountd. The auto maps and timeouts are controlled by the zone administrator. You cannot trigger a mount in another zone by crossing an AutoFS mount point for a non-global zone from the global zone.

Certain AutoFS mounts are created in the kernel when another mount is triggered. Such mounts cannot be removed by using the regular umount interface because they must be mounted or unmounted as a group. Note that this functionality is provided for zone shutdown.

MNTFS

MNTFS is a virtual file system that provides read-only access to the table of mounted file systems for the local system. The set of file systems visible by using mnttab from within a non-global zone is the set of file systems mounted in the zone, plus an entry for root (/) . Mount points with a special device that is not accessible from within the zone, such as /dev/rdsk/c0t0d0s0, have their special device set to the same as the mount point. All mounts in the system are visible from the global zone's /etc/mnttab table. For more information on MNTFS, see Mounting and Unmounting Oracle Solaris File Systems in Oracle Solaris 11.1 Administration: Devices and File Systems.

NFS

NFS mounts established within a zone are local to that zone. The mounts cannot be accessed from other zones, including the global zone. The mounts are removed when the zone is halted or rebooted.

From within a zone, NFS mounts behave as though mounted with the nodevices option.

The nfsstat command output only pertains to the zone in which the command is run. For example, if the command is run in the global zone, only information about the global zone is reported. For more information about the nfsstat command, see nfsstat(1M).

PROCFS

The /proc file system, or PROCFS, provides process visibility and access restrictions as well as information about the zone association of processes. Only processes in the same zone are visible through /proc.

Processes in the global zone can observe processes and other objects in non-global zones. This allows such processes to have system-wide observability.

From within a zone, procfs mounts behave as though mounted with the nodevices option. For more information about procfs, see the proc(4) man page.

LOFS

The scope of what can be mounted through LOFS is limited to the portion of the file system that is visible to the zone. Hence, there are no restrictions on LOFS mounts in a zone.

UFS, UDFS, PCFS, and other storage-based file systems

When using the zonecfg command to configure storage-based file systems that have an fsck binary, such as UFS, the zone administrator must specify a raw parameter. The parameter indicates the raw (character) device, such as /dev/rdsk/c0t0d0s7. The zoneadmd daemon automatically runs the fsck command in preen mode (fsck -p), which checks and fixes the file system non-interactively, before it mounts the file system. If the fsck fails, zoneadmd cannot bring the zone to the ready state. The path specified by raw cannot be a relative path.

It is an error to specify a device to fsck for a file system that does not provide an fsck binary in /usr/lib/fs/fstype/fsck. It is also an error if you do not specify a device to fsck if an fsck binary exists for that file system.

For more information, see The zoneadmd Daemon and the fsck(1M) command.

ZFS

In addition to the default dataset described in File Systems Mounted in Zones, you can add a ZFS dataset to a non-global zone by using the zonecfg command with the add dataset resource. The dataset is visible and mounted in the non-global zone, and also visible in the global zone. The zone administrator can create and destroy file systems within that dataset, and modify the properties of the dataset.

The zoned attribute of zfs indicates whether a dataset has been added to a non-global zone.

# zfs get zoned tank/sales
NAME          PROPERTY    VALUE      SOURCE
tank/sales    zoned       on         local

Each dataset that is delegated to a non-global zone through a dataset resource is aliased. The dataset layout is not visible within the zone. Each aliased dataset appears in the zone as if it were a pool. The default alias for a dataset is the last component in the dataset name. For example, if the default alias is used for the delegated dataset tank/sales, the zone will see a virtual ZFS pool named sales. The alias can be customized to be a different value by setting the alias property within the dataset resource.

A dataset named rpool exists within each non-global zone's zonepath dataset. For all non-global zones, this zone rpool dataset is aliased as rpool.

my-zone# zfs list -o name,zoned,mounted,mountpoint
NAME                ZONED  MOUNTED  MOUNTPOINT
rpool                  on       no  /rpool
rpool/ROOT             on       no  legacy
rpool/ROOT/solaris     on      yes  /
rpool/export           on       no  /export
rpool/export/home      on       no  /export/home

Dataset aliases are subject to the same name restrictions as ZFS pools. These restrictions are documented in the zpool(1M) man page.

If you want to share a dataset from the global zone, you can add an LOFS-mounted ZFS file system by using the zonecfg command with the add fs subcommand. The global administrator or a user granted the appropriate authorizations is responsible for setting and controlling the properties of the dataset.

For more information on ZFS, see Chapter 9, Oracle Solaris ZFS Advanced Topics, in Oracle Solaris 11.1 Administration: ZFS File Systems.

Non-Global Zones as NFS Clients

Zones can be NFS clients. Version 2, version 3, and version 4 protocols are supported. For information on these NFS versions, see Features of the NFS Service in Oracle Solaris Administration: Network Services. .

The default version is NFS version 4. You can enable other NFS versions on a client by using one of the following methods:

• You can use sharectl(1M) to set properties. Set NFS_CLIENT_VERSMAX=number so that the zone uses the specified version by default. See Setting Up NFS Services in Oracle Solaris Administration: Network Services. Use the procedure How to Select Different Versions of NFS on a Client in Managing Network File Systems in Oracle Solaris 11.1.

• You can manually create a version mount. This method overrides sharectl setting. See Setting Up NFS Services in Oracle Solaris Administration: Network Services. Use the procedure How to Select Different Versions of NFS on a Client in Managing Network File Systems in Oracle Solaris 11.1.

Use of mknod Prohibited in a Zone

Note that you cannot use the mknod command documented in the mknod(1M) man page to make a special file in a non-global zone.

Traversing File Systems

A zone's file system namespace is a subset of the namespace accessible from the global zone. Unprivileged processes in the global zone are prevented from traversing a non-global zone's file system hierarchy through the following means:

• Specifying that the zone root's parent directory is owned, readable, writable, and executable by root only

• Restricting access to directories exported by /proc

Note that attempting to access AutoFS nodes mounted for another zone will fail. The global administrator must not have auto maps that descend into other zones.

Restriction on Accessing A Non-Global Zone From the Global Zone

After a non-global zone is installed, the zone must never be accessed directly from the global zone by any commands other than system backup utilities. Moreover, a non-global zone can no longer be considered secure after it has been exposed to an unknown environment. An example would be a zone placed on a publicly accessible network, where it would be possible for the zone to be compromised and the contents of its file systems altered. If there is any possibility that compromise has occurred, the global administrator should treat the zone as untrusted.

Any command that accepts an alternative root by using the -R or -b options (or the equivalent) must not be used when the following are true:

• The command is run in the global zone.

• The alternative root refers to any path within a non-global zone, whether the path is relative to the current running system's global zone or the global zone in an alternative root.

An example is the -R root_path option to the pkgadd utility run from the global zone with a non-global zone root path.

The list of commands, programs, and utilities that use -R with an alternative root path include the following:

• auditreduce

• bart

• installf

• localeadm

• makeuuid

• metaroot

• pkg

• prodreg

• removef

• routeadm

• showrev

• syseventadm

The list of commands and programs that use -b with an alternative root path include the following:

• add_drv

• pprosetup

• rem_drv

• roleadd

• update_drv

• useradd