Profile Head¶

The profile head consists of a required name that is unique and optional attachment conditionals and control flags.

Name

The name of the profile is its identifier. It is what is displayed during introspection (eg. ps -Z), and defines how the profile is referenced by policy rules for any policy interaction via ipc or domain changes. It is recommended that the name be kept short and have meaning for the application it is being applied eg. firefox for the firefox web browser or its functional role eg. log_admin.

If the name is an applications full absolute path name eg. /usr/bin/firefox and an exec attachment conditional is not specified the name is also used as the profile's exec attachment conditional. This use however has been deprecated and is discouraged as it makes for long names that can make profile rules difficult to understand, and may not be fully displayed by some introspection tools.

Attachment Conditionals

The attachment conditionals are used during profile changes to determine whether a profile is a match for the proposed profile transition. The attachment conditionals are optional, how and when they are applied is determined by the specific condition(s) used.

When attachment conditionals are used, the attachment conditionals for all profiles in the namespace will be evaluated. The profile with the set of attachments that result in the best match will become the new profile after a transition operation. Attachments that don't match will result in the profile not being available for transition.

If no conditionals are specified the profile will only be used if a transition explicitly specifies the profile name.

Exec Attachment Conditional

The exec attachment conditional governs how closely the profile matches an executable program. This conditional is only used during an exec operation when the matching exec rule specifies either a px or cx (or their derivatives) transition type. The exec attachment conditional will also be used by tasks that are unconfined as they use a pix transition rule.

If there are no attachment matches then it is up to the exec rule to determine what happens (fail or a fallback option).

Note: see profile Name for information around using the profile name as an attachment conditional.

Exec attachment conditionals can contain variable names and pattern matching. They use a longest left match heuristic to deterime the winner in the case of multiple matches at run time. The exact implementation of this resolution is kernel specific and has improved over time, while retaining backwards compatibility. If the heuristic can not determine a winner between multiple matches the exec will be denied.

Extended Attributes Attachment Conditional

AppArmor profiles have the ability to target files based on their xattr(7) values in addition to their path. For example, the following profile matches files in /usr/bin with the attribute "security.apparmor" and value "trusted":

  /usr/bin/* xattrs(security.apparmor="trusted") {
    # ...
  }

See apparmor_xattrs(7) for further details.

Flags

The profile flags allow modifying the behavior of the profile. If a profile flag is specified it takes priority over any conflicting flags that have been specified by rules in the profile body.

Profile Mode

The profile mode allow controlling the enforcement behavior of the profile rules.

If no mode is specified the profile defaults to enforce mode.

enforce For a given action, if the profile rules do not grant permission the action will be denied, with an EACCES or EPERM error code returned to userspace, and the violation will be logged with a tag of the access being DENIED.

kill This is a variant of enforce mode where in addition to returning EACCES or EPERM for a violation, the task is also sent a signal to kill it.

complain For a given action, if the profile rules do not grant permission the action will be allowed, but the violation will be logged with a tag of the access being ALLOWED.

default_allow This mode changes the default behavior of apparmor from default deny to default allow. When default_allow is specified the resulting profile will allow operations that the profile does not have a rule for. This mode is similar to unconfined but allows for allow and deny rules, specifying audit, and domain transitions. Profiles in this mode may be be reported as being in enforce mode or allow mode when introspected from the kernel.

Note: default_allow is similar and for many profiles will be equivalent to specifying an allow all, rule in the profile. The default_allow flag does not provide all the same option that the allow all, rule provides.

unconfined This mode allows a task confined by the profile to behave as though it is unconfined. The unconfined behavior can be later changed to confinement by using profile replacement. This mode should not be used under regular deployment but can be useful during debugging and some system initialization scenarios.

This mode is similar to default_allow and may be emulated by default_allow in kernels that no longer support a true unconfined mode. It does not generally allow for specifying deny rules, or allow rules that override the default behavior, except in a few custom kernels where unconfined restricts a few operations. It relies on special customized behavior of the unconfined profile in the kernel and as such should only be used for debugging.

Note: true unconfined is being phased out, with unconfined becoming a replaceable profile. As such unconfined mode will be emulated by a special profile compiled with the default_allow flag in newer kernels.

prompt This mode allows task mediation to send an up call to userspace to ask for a decision when there isn't a rule covering the permission request. If userspace does not respond then the access will be denied.

Audit Mode

The audit mode allows control of how AppArmor messages are are logged to the audit system.

audit This flag causes all actions whether allowed or denied to be logged.

Misc modes

mediate_deleted This forces AppArmor to mediate deleted files as if they still exist in the file system.

attach_disconnected This forces AppArmor to attach disconnected objects to the task's namespace and mediate them as though they are part of the namespace. WARNING this mode is unsafe and can result in aliasing and access to objects that should not be allowed. Its intent is a debug and policy development tool.

attach_disconnected.path=ABS PATH Like attach_disconnected, but attach disconnected objects to the supplied path instead of the root of the namespace.

chroot_relative This forces file names to be relative to a chroot and behave as if the chroot is a mount namespace.

debug This flag allows turning on kernel debug messages on a per profile basis. It works in conjunction with other kernel debug flags to control what messages will be output. Its effect is kernel dependent, and it should never appear in policy except when trying to debug kernel or policy problems.

interruptible Enables interrupts for prompt upcall to userspace.

kill.signal=SIGNAL This changes the signal that will be sent by AppArmor when in kill mode or a kill rule has been violated.

Access Modes¶

File permission access modes consists of combinations of the following modes:

r

- read

w

- write -- conflicts with append

a

- append -- conflicts with write

ux

- unconfined execute

Ux

- unconfined execute -- scrub the environment

px

- discrete profile execute

Px

- discrete profile execute -- scrub the environment

cx

- transition to subprofile on execute

Cx

- transition to subprofile on execute -- scrub the environment

ix

- inherit execute

pix

- discrete profile execute with inherit fallback

Pix

- discrete profile execute with inherit fallback -- scrub the environment

cix

- transition to subprofile on execute with inherit fallback

Cix

- transition to subprofile on execute with inherit fallback -- scrub the environment

pux

- discrete profile execute with fallback to unconfined

PUx

- discrete profile execute with fallback to unconfined -- scrub the environment

cux

- transition to subprofile on execute with fallback to unconfined

CUx

- transition to subprofile on execute with fallback to unconfined -- scrub the environment

deny x

- disallow execute (in rules with the deny qualifier)

m

- allow PROT_EXEC with mmap(2) calls

l

- link

k

- lock

Access Modes Details¶

r - Read mode

Allows the program to have read access to the file or directory listing. Read access is required for shell scripts and other interpreted content.

w - Write mode

Allows the program to have write access to the file. Files and directories must have this permission if they are to be unlinked (removed.) Write mode is not required on a directory to rename or create files within the directory.

This mode conflicts with append mode.

a - Append mode

Allows the program to have a limited appending only write access to the file. Append mode will prevent an application from opening the file for write unless it passes the O_APPEND parameter flag on open.

The mode conflicts with Write mode.

ux - Unconfined execute mode

Allows the program to execute the program without any AppArmor profile being applied to the program.

This mode is useful when a confined program needs to be able to perform a privileged operation, such as rebooting the machine. By placing the privileged section in another executable and granting unconfined execution rights, it is possible to bypass the mandatory constraints imposed on all confined processes. For more information on what is constrained, see the apparmor(7) man page.

WARNING 'ux' should only be used in very special cases. It enables the designated child processes to be run without any AppArmor protection. 'ux' does not scrub the environment of variables such as LD_PRELOAD; as a result, the calling domain may have an undue amount of influence over the callee. Use this mode only if the child absolutely must be run unconfined and LD_PRELOAD must be used. Any profile using this mode provides negligible security. Use at your own risk.

Incompatible with other exec transition modes and the deny qualifier.

Ux - unconfined execute -- scrub the environment

'Ux' allows the named program to run in 'ux' mode, but AppArmor will invoke the Linux Kernel's unsafe_exec routines to scrub the environment, similar to setuid programs. (See ld.so(8) for some information on setuid/setgid environment scrubbing.)

WARNING 'Ux' should only be used in very special cases. It enables the designated child processes to be run without any AppArmor protection. Use this mode only if the child absolutely must be run unconfined. Use at your own risk.

Incompatible with other exec transition modes and the deny qualifier.

px - Discrete Profile execute mode

This mode requires that a discrete security profile is defined for a program executed and forces an AppArmor domain transition. If there is no profile defined then the access will be denied.

WARNING 'px' does not scrub the environment of variables such as LD_PRELOAD; as a result, the calling domain may have an undue amount of influence over the callee.

Incompatible with other exec transition modes and the deny qualifier.

Px - Discrete Profile execute mode -- scrub the environment

'Px' allows the named program to run in 'px' mode, but AppArmor will invoke the Linux Kernel's unsafe_exec routines to scrub the environment, similar to setuid programs. (See ld.so(8) for some information on setuid/setgid environment scrubbing.)

Incompatible with other exec transition modes and the deny qualifier.

cx - Transition to Subprofile execute mode

This mode requires that a local security profile is defined and forces an AppArmor domain transition to the named profile. If there is no profile defined then the access will be denied.

WARNING 'cx' does not scrub the environment of variables such as LD_PRELOAD; as a result, the calling domain may have an undue amount of influence over the callee.

Incompatible with other exec transition modes and the deny qualifier.

Cx - Transition to Subprofile execute mode -- scrub the environment

'Cx' allows the named program to run in 'cx' mode, but AppArmor will invoke the Linux Kernel's unsafe_exec routines to scrub the environment, similar to setuid programs. (See ld.so(8) for some information on setuid/setgid environment scrubbing.)

Incompatible with other exec transition modes and the deny qualifier.

ix - Inherit execute mode

Prevent the normal AppArmor domain transition on execve(2) when the profiled program executes the named program. Instead, the executed resource will inherit the current profile.

This mode is useful when a confined program needs to call another confined program without gaining the permissions of the target's profile, or losing the permissions of the current profile. There is no version to scrub the environment because 'ix' executions don't change privileges.

Incompatible with other exec transition modes and the deny qualifier.

Profile transition with inheritance fallback execute mode

These modes attempt to perform a domain transition as specified by the matching permission (shown below) and if that transition fails to find the matching profile the domain transition proceeds using the 'ix' transition mode.

  'Pix' == 'Px' with fallback to 'ix'
  'pix' == 'px' with fallback to 'ix'
  'Cix' == 'Cx' with fallback to 'ix'
  'cix' == 'cx' with fallback to 'ix'
    

Incompatible with other exec transition modes and the deny qualifier.

Profile transition with unconfined fallback execute mode

These modes attempt to perform a domain transition as specified by the matching permission (shown below) and if that transition fails to find the matching profile the domain transition proceeds using the 'ux' transition mode if 'pux', 'cux' or the 'Ux' transition mode if 'PUx', 'CUx' is used.

  'PUx' == 'Px' with fallback to 'Ux'
  'pux' == 'px' with fallback to 'ux'
  'CUx' == 'Cx' with fallback to 'Ux'
  'cux' == 'cx' with fallback to 'ux'
    

Incompatible with other exec transition modes and the deny qualifier.

deny x - Deny execute

For rules including the deny modifier, only 'x' is allowed to deny execute.

The 'ix', 'Px', 'px', 'Cx', 'cx' and the fallback modes conflict with the deny modifier.

Directed profile transitions

The directed ('px', 'Px', 'pix', 'Pix', 'pux', 'PUx') profile and subprofile ('cx', 'Cx', 'cix', 'Cix', 'cux', 'CUx') transitions normally determine the profile to transition to from the executable name. It is however possible to specify the name of the profile that the transition should use.

The name of the profile to transition to is specified using the '->' followed by the name of the profile to transition to. Eg.

  /bin/** px -> profile,
    

Incompatible with other exec transition modes.

m - Allow executable mapping

This mode allows a file to be mapped into memory using mmap(2)'s PROT_EXEC flag. This flag marks the pages executable; it is used on some architectures to provide non-executable data pages, which can complicate exploit attempts. AppArmor uses this mode to limit which files a well-behaved program (or all programs on architectures that enforce non-executable memory access controls) may use as libraries, to limit the effect of invalid -L flags given to ld(1) and LD_PRELOAD, LD_LIBRARY_PATH, given to ld.so(8).

l - Link mode

Allows the program to be able to create a link with this name. When a link is created, the new link MUST have a subset of permissions as the original file (with the exception that the destination does not have to have link access.) If there is an 'x' rule on the new link, it must match the original file exactly.

k - lock mode

Allows the program to be able lock a file with this name. This permission covers both advisory and mandatory locking.

leading OR trailing access permissions

File rules can be specified with the access permission either leading or trailing the file glob. Eg.

  rw /**,               # leading permissions
  /** rw,               # trailing permissions
    

When leading permissions are used further rule options and context may be allowed, Eg.

  l /foo -> /bar,       # lead 'l' link permission is equivalent to link rules
    

Link rules¶

Link rules allow specifying permission to form a hard link as a link target pair. If the subset condition is specified then the permissions to access the link file must be a subset of the profiles permissions to access the target file. If there is an 'x' rule on the new link, it must match the original file exactly.

Eg.

  /file1  r,
  /file2  rwk,
  /link*  rw,
  link subset /link* -> /**,

The link rule allows linking of /link to both /file1 or /file2 by name however because the /link file has 'rw' permissions it is not allowed to link to /file1 because that would grant an access path to /file1 with more permissions than the 'r' permissions the profile specifies.

A link of /link to /file2 would be allowed because the 'rw' permissions of /link are a subset of the 'rwk' permissions for /file1.

The link rule is equivalent to specifying the 'l' link permission as a leading permission with no other file access permissions. When this is done the link rule options can be specified.

The following link rule is equivalent to the 'l' permission file rule

  link /foo -> bar,
  l /foo -> /bar,

File rules that specify the 'l' permission and don't specify the extend link permissions map to link rules as follows.

  /foo l,
  l /foo,
  link subset /foo -> /**,

Comments¶

Comments start with # and may begin at any place within a line. The comment ends when the line ends. This is the same comment style as shell scripts.

Capabilities¶

The only capabilities a confined process may use may be enumerated; for the complete list, please refer to capabilities(7). Note that granting some capabilities renders AppArmor confinement for that domain advisory; while open(2), read(2), write(2), etc., will still return error when access is not granted, some capabilities allow loading kernel modules, arbitrary access to IPC, ability to bypass discretionary access controls, and other operations that are typically reserved for the root user.

Network Rules¶

AppArmor supports simple coarse grained network mediation. The network rule restrict all socket(2) based operations. The mediation done is a coarse-grained check on whether a socket of a given type and family can be created, read, or written. Network netlink(7) rules may only specify type 'dgram' and 'raw'.

AppArmor network rules are accumulated so that the granted network permissions are the union of all the listed network rule permissions.

AppArmor network rules are broad and general and become more restrictive as further information is specified.

eg.

 network,               #allow access to all networking
 network tcp,           #allow access to tcp
 network inet tcp,      #allow access to tcp only for inet4 addresses
 network inet6 tcp,     #allow access to tcp only for inet6 addresses
 network netlink raw,   #allow access to AF_NETLINK SOCK_RAW

Network permissions

Network rule permissions are implied when a rule does not explicitly state an access list. By default if a rule does not have an access list all permissions that are compatible with the specified set of local and peer conditionals are implied.

The create, bind, listen, shutdown, getattr, setattr, getopt, and setopt permissions are local socket permissions. They are only applied to the local socket and can't be specified in rules that have a peer conditional. The accept permission applies to the combination of a local and peer socket. The connect, send, and receive permissions are peer socket permissions.

Mediation of inet/inet6 family

AppArmor supports fine grained mediation of the inet and inet6 families by using the ip and port conditionals. The ip conditional accepts both IPv4 and IPv6 using the regular representation of four octets separated by '.' for IPv4 and eight groups of four hexadecimal numbers separated by ':' for IPv6. Contiguous leading zeros can be replaced by '::' once. On a connected socket, the sender and receiver don't need to be specified in the recvfrom and sendto system calls. In that case, and with unbounded sockets, the IP address is none, or unknown. Unknown or Unbound IP addresses are represented in policy by the 'none' keyword. When the ip conditional is omitted, then all IP addresses will be allowed: IPv4, IPv6 and none. If INADDR_ANY or in6addr_any is used, then the ip conditional can be omitted or they can be represented by:

 network ip=::,         #allow in6addr_any
 network ip=0.0.0.0;    #allow INADDR_ANY

The network rules support the specification of local and remote IP addresses and ports.

 network ip=127.0.0.1 port=8080,
 network peer=(ip=10.139.15.23 port=8081),
 network ip=fd74:1820:b03a:b361::cf32 peer=(ip=fd74:1820:b03a:b361::a0f9),
 network port=8080 peer=(port=8081),
 network ip=127.0.0.1 port=8080 peer=(ip=10.139.15.23 port=8081),

Mount Rules¶

AppArmor supports mount mediation and allows specifying filesystem types and mount flags. The syntax of mount rules in AppArmor is based on the mount(8) command syntax. Mount rules must contain one of the mount, remount or umount keywords, but all mount conditions are optional. Unspecified optional conditionals are assumed to match all entries (eg, not specifying fstype means all fstypes are matched). Due to the complexity of the mount command and how options may be specified, AppArmor allows specifying conditionals three different ways:

1.

If a conditional is specified using '=', then the rule only grants permission for mounts matching the exactly specified options. For example, an AppArmor policy with the following rule:

    mount options=ro /dev/foo -E<gt> /mnt/,
    

Would match:

    $ mount -o ro /dev/foo /mnt
    

but not either of these:

    $ mount -o ro,atime /dev/foo /mnt
    $ mount -o rw /dev/foo /mnt
    

2.

If a conditional is specified using 'in', then the rule grants permission for mounts matching any combination of the specified options. For example, if an AppArmor policy has the following rule:

    mount options in (ro,atime) /dev/foo -> /mnt/,
    

all of these mount commands will match:

    $ mount -o ro /dev/foo /mnt
    $ mount -o ro,atime /dev/foo /mnt
    $ mount -o atime /dev/foo /mnt
    

but none of these will:

    $ mount -o ro,sync /dev/foo /mnt
    $ mount -o ro,atime,sync /dev/foo /mnt
    $ mount -o rw /dev/foo /mnt
    $ mount -o rw,noatime /dev/foo /mnt
    $ mount /dev/foo /mnt
    

3.

If multiple conditionals are specified in a single mount rule, then the rule grants permission for each set of options. This provides a shorthand when writing mount rules which might help to logically break up a conditional. For example, if an AppArmor policy has the following rule:

    mount options=ro options=atime
    

both of these mount commands will match:

    $ mount -o ro /dev/foo /mnt
    $ mount -o atime /dev/foo /mnt
    

but this one will not:

    $ mount -o ro,atime /dev/foo /mnt
    

Note that separate mount rules are distinct and the options do not accumulate. For example, these AppArmor mount rules:

    mount options=ro,
    mount options=atime,

are not equivalent to either of these mount rules:

    mount options=(ro,atime),
    mount options in (ro,atime),

To help clarify the flexibility and complexity of mount rules, here are some example rules with accompanying matching commands:

mount,

the 'mount' rule without any conditionals is the most generic and allows any mount. Equivalent to 'mount fstype=** options=** ** -> /**'.

mount /dev/foo,

allow mounting of /dev/foo anywhere with any options. Some matching mount commands:

    $ mount /dev/foo /mnt
    $ mount -t ext3 /dev/foo /mnt
    $ mount -t vfat /dev/foo /mnt
    $ mount -o ro,atime,noexec,nodiratime /dev/foo /srv/some/mountpoint
    

mount options=ro /dev/foo,

allow mounting of /dev/foo anywhere, as read only. Some matching mount commands:

    $ mount -o ro /dev/foo /mnt
    $ mount -o ro /dev/foo /some/where/else
    

mount options=(ro,atime) /dev/foo,

allow mount of /dev/foo anywhere, as read only and using inode access times. Some matching mount commands:

    $ mount -o ro,atime /dev/foo /mnt
    $ mount -o ro,atime /dev/foo /some/where/else
    

mount options in (ro,atime) /dev/foo,

allow mount of /dev/foo anywhere using some combination of 'ro' and 'atime' (see above). Some matching mount commands:

    $ mount -o ro /dev/foo /mnt
    $ mount -o atime /dev/foo /some/where/else
    $ mount -o ro,atime /dev/foo /some/other/place
    

mount options=ro /dev/foo, mount options=atime /dev/foo,

allow mount of /dev/foo anywhere as read only, and allow mount of /dev/foo anywhere using inode access times. Note this is expressed as two different rules. Matches:

    $ mount -o ro /dev/foo /mnt/1
    $ mount -o atime /dev/foo /mnt/2
    

mount -> /mnt/**,

allow mounting anything under a directory in /mnt/**. Some matching mount commands:

    $ mount /dev/foo1 /mnt/1
    $ mount -o ro,atime,noexec,nodiratime /dev/foo2 /mnt/deep/path/foo2
    

mount options=ro -> /mnt/**,

allow mounting anything under /mnt/**, as read only. Some matching mount commands:

    $ mount -o ro /dev/foo1 /mnt/1
    $ mount -o ro /dev/foo2 /mnt/deep/path/foo2
    

mount fstype=ext3 options=(rw,atime) /dev/sdb1 -> /mnt/stick/,

allow mounting an ext3 filesystem in /dev/sdb1 on /mnt/stick as read/write and using inode access times. Matches only:

    $ mount -o rw,atime /dev/sdb1 /mnt/stick
    

mount options=(ro, atime) options in (nodev, user) /dev/foo -> /mnt/,

allow mounting /dev/foo on /mmt/ read only and using inode access times or allow mounting /dev/foo on /mnt/ with some combination of 'nodev' and 'user'. Matches only:

    $ mount -o ro,atime /dev/foo /mnt
    $ mount -o nodev /dev/foo /mnt
    $ mount -o user /dev/foo /mnt
    $ mount -o nodev,user /dev/foo /mnt
    

Message Queue rules¶

AppArmor supports mediation of POSIX and SYSV message queues.

AppArmor Message Queue permissions are implied when a rule does not explicitly state an access list. By default, all Message Queue permissions are implied.

AppArmor Message Queue permissions become more restricted as further information is specified. Policy can be specified by determining its access mode, type, label, and message queue name.

Regarding access modes, 'r' and 'read' are used to read messages from the queue. 'w' and 'write' are used to write to the message queue. 'create' is used to create the message queue, and 'open' is used to get the message queue identifier when the queue is already created. 'delete' is used to remove the message queue. The access modes to get and set attributes of the message queue are 'setattr' and 'getattr'.

The type of the policy can be either 'posix' or 'sysv'. This information is relevant when the message queue name is not specified, and when specified can be inferred by the queue name, since message queues' name for posix must start with '/', and message queues' key for SYSV must be a positive integer.

The policy label is the label assigned to the message queue when it is created.

The message queue name can be either a string starting with '/' if the type is POSIX, or a positive integer if the type is SYSV. If the type is not specified, then it will be inferred by the queue name.

Example AppArmor Message Queue rules:

    # Allow all Message Queue access
    mqueue,
    # Explicitly allow all Message Queue access,
    mqueue (create, open, delete, read, write, getattr, setattr),
    # Explicitly deny use of Message Queue
    deny mqueue,
    # Allow all access for POSIX queue of name /bar
    mqueue type=posix /bar,
    # Allow create permission for a SYSV queue of label foo
    mqueue create label=foo 123,

User Namespace Rules¶

User namespaces are part of many sandboxing and containerization solutions. They provide a way for a non-system root process to be root within the container. Unfortunately this opens up attack surface in the kernel and has been part of several exploit chains. As such AppArmor can be used to restrict the creation of user namespaces to select processes.

User namespace permission are implied when a rule does not explicitly state an access list. The rule becomes more restrictive as further information is specified.

Note: user namespace creation may be restricted so that it is not available to unprivieged unconfined processes. If this is the case any process trying to create user namespaces will require a profile that allows the necessary permissions.

create

Allow creation of user namespaces.

Example userns rules:

  # Allow all userns perms
  userns,
  # Allow creation of a userns
  userns create,

IO_URing Rules¶

AppArmor supports mediation of the new Linux high speed IO interface. There is limited mediation at this time to just a few permissions at the moment.

IO Uring permission are implied when a rule does not explicitly state an access list. The rule becomes more restrictive as further information is specified.

Note: io_uring access may be restricted so that it is not available to unprivileged unconfined processes. If this is the case any process trying to use io_uring will require a profile that allows the necessary io_uring permissions.

sqpoll

All the task confined by the profile to spawn a io_uring polling thread.

override_creds

Grants the task confined by the profile to override (change) its credentials to the specified label, when executing an io_uring operation.

Example IO_URING rules:

  # Allow io_uring operations
  io_ring,
  # Allow creation of a polling thread
  io_uring sqpoll,
  # Allow task to override credentials during io_uring operation
  io_uring override_creds label=new_creds,

Pivot Root Rules¶

AppArmor mediates changing of the root filesystem through the pivot_root(2) system call. The syntax of 'pivot_root' rules in AppArmor is based on the pivot_root(2) system call parameters with the notable exception that the ordering is reversed. The path corresponding to the put_old parameter of pivot_root(2) is optionally specified in the 'pivot_root' rule using the 'oldroot=' prefix.

AppArmor 'pivot_root' rules can specify a profile transition to occur during the pivot_root(2) system call. Note that AppArmor will only transition the process calling pivot_root(2) to the new profile.

The paths specified in 'pivot_root' rules must end with '/' since they are directories.

Here are some example 'pivot_root' rules:

    # Allow any pivot
    pivot_root,
    # Allow pivoting to any new root directory and putting the old root
    # directory at /mnt/root/old/
    pivot_root oldroot=/mnt/root/old/,
    # Allow pivoting the root directory to /mnt/root/
    pivot_root /mnt/root/,
    # Allow pivoting to /mnt/root/ and putting the old root directory at
    # /mnt/root/old/
    pivot_root oldroot=/mnt/root/old/ /mnt/root/,
    # Allow pivoting to /mnt/root/, putting the old root directory at
    # /mnt/root/old/ and transition to the /mnt/root/sbin/init profile
    pivot_root oldroot=/mnt/root/old/ /mnt/root/ -> /mnt/root/sbin/init,

PTrace rules¶

AppArmor supports mediation of ptrace(2). AppArmor PTrace rules are accumulated so that the granted PTrace permissions are the union of all the listed PTrace rule permissions.

AppArmor PTrace permissions are implied when a rule does not explicitly state an access list. By default, all PTrace permissions are implied.

The trace and tracedby permissions govern ptrace(2) while read and readby govern certain proc(5) filesystem accesses, kcmp(2), futexes (get_robust_list(2)) and perf trace events.

For a ptrace operation to be allowed the profile of the tracing process and the profile of the target task must both have the correct permissions. For example, the profile of the process attaching to another task must have the trace permission for the target task's profile, and the task being traced must have the tracedby permission for the tracing process' profile.

Example AppArmor PTrace rules:

    # Allow all PTrace access
    ptrace,
    # Explicitly allow all PTrace access,
    ptrace (read, readby, trace, tracedby),
    # Explicitly deny use of ptrace(2)
    deny ptrace (trace),
    # Allow unconfined processes (eg, a debugger) to ptrace us
    ptrace (readby, tracedby) peer=unconfined,
    # Allow ptrace of a process running under the /usr/bin/foo profile
    ptrace (trace) peer=/usr/bin/foo,

Signal rules¶

AppArmor supports mediation of signal(7). AppArmor signal rules are accumulated so that the granted signal permissions are the union of all the listed signal rule permissions.

AppArmor signal permissions are implied when a rule does not explicitly state an access list. By default, all signal permissions are implied.

For the sending of a signal to be allowed, the profile of the sending process and the profile of the target task must both have the correct permissions. For example, the profile of a process sending a signal to another task must have the send permission for the target task's profile, and the task receiving the signal must have a receive permission for the sending process' profile.

Example AppArmor signal rules:

    # Allow all signal access
    signal,
    # Explicitly deny sending the HUP and INT signals
    deny signal (send) set=(hup, int),
    # Allow unconfined processes to send us signals
    signal (receive) peer=unconfined,
    # Allow sending of signals to a process running under the /usr/bin/foo
    # profile
    signal (send) peer=/usr/bin/foo,
    # Allow checking for PID existence
    signal (receive, send) set=("exists"),
    # Allow us to signal ourselves using the built-in @{profile_name} variable
    signal peer=@{profile_name},
    # Allow two real-time signals
    signal set=(rtmin+0 rtmin+32),

DBus rules¶

AppArmor supports DBus mediation. The mediation is performed in conjunction with the DBus daemon. The DBus daemon verifies that communications over the bus are permitted by AppArmor policy.

AppArmor DBus rules are accumulated so that the granted DBus permissions are the union of all the listed DBus rule permissions.

AppArmor DBus rules are broad and general and become more restrictive as further information is specified. Policy may be specified down to the interface member level (method or signal name), however the contents of messages are not examined.

Some AppArmor DBus permissions are not compatible with all AppArmor DBus rules. The 'bind' permission cannot be used in message rules. The 'send' and 'receive' permissions cannot be used in service rules. The 'eavesdrop' permission cannot be used in rules containing any conditionals outside of the 'bus' conditional.

'r' and 'read' are synonyms for 'receive'. 'w' and 'write' are synonyms for 'send'. 'rw' is a synonym for both 'send' and 'receive'.

AppArmor DBus permissions are implied when a rule does not explicitly state an access list. By default, all DBus permissions are implied. Only message permissions are implied for message rules and only service permissions are implied for service rules.

Example AppArmor DBus rules:

    # Allow all DBus access
    dbus,
    # Explicitly allow all DBus access,
    dbus (send, receive, bind),
    # Deny send/receive/bind access to the session bus
    deny dbus bus=session,
    # Allow bind access for a particular name on any bus
    dbus bind name=com.example.ExampleName,
    # Allow receive access for a particular path and interface
    dbus receive path=/com/example/path interface=com.example.Interface,
    # Deny send/receive access to the system bus for a particular interface
    deny dbus bus=system interface=com.example.ExampleInterface,
    # Allow send access for a particular path, interface, member, and pair of
    # peer names:
    dbus send
         bus=session
         path=/com/example/path
         interface=com.example.Interface
         member=ExampleMethod
         peer=(name=(com.example.ExampleName1|com.example.ExampleName2)),
    # Allow receive access for all unconfined peers
    dbus receive peer=(label=unconfined),
    # Allow eavesdropping on the system bus
    dbus eavesdrop bus=system,
    # Allow and audit all eavesdropping
    audit dbus eavesdrop,

Unix socket rules¶

AppArmor supports fine grained mediation of unix domain abstract and anonymous sockets. Unix domain sockets with file system paths are mediated via file access rules.

Abstract unix domain sockets is a nonportable Linux extension of unix domain sockets, see unix(7) for more information.

Unix socket address paths

The sun_path component (aka the socket address) of a unix domain socket is specified by the

  addr=

conditional. If an address conditional is not specified as part of a rule then the rule matches both abstract and anonymous sockets.

In apparmor the address of an abstract unix domain socket begins with the @ character, similar to how they are reported (as paths) by netstat -x. The address then follows and may contain pattern matching and any characters including the null character. In apparmor null characters must be specified by using an escape sequence \000 or \x00. The pattern matching is the same as is used by file path matching so * will not match / even though it has no special meaning with in an abstract socket name. Eg.

  unix addr=@*,

Autobound unix domain sockets have a unix sun_path assigned to them by the kernel, as such specifying a policy based address is not possible. The autobinding of sockets can be controlled by specifying the special auto keyword. Eg.

  unix addr=auto,

To indicate that the rule only applies to auto binding of unix domain sockets. It is important to note this only applies to the bind permission as once the socket is bound to an address it is indistinguishable from a socket that have an addr bound with a specified name. When the auto keyword is used with other permissions or as part of a peer addr it will be replaced with a pattern that can match an autobound socket. Eg. For some kernels

  unix rw addr=auto,

is transformed to

  unix rw addr=@[a-f0-9][a-f0-9][a-f0-9][a-f0-9][a-f0-9],

It is important to note, this pattern may match abstract sockets that were not autobound but have an addr that fits what is generated by the kernel when autobinding a socket.

Anonymous unix domain sockets have no sun_path associated with the socket address, however it can be specified with the special none keyword to indicate the rule only applies to anonymous unix domain sockets. Eg.

  unix addr=none,

If the address component of a rule is not specified then the rule applies to autobind, abstract and anonymous sockets.

Unix socket permissions

Unix domain socket rules are accumulated so that the granted unix socket permissions are the union of all the listed unix rule permissions.

Unix domain socket rules are broad and general and become more restrictive as further information is specified. Policy may be specified down to the socket address (aka sun_path) and label level. The content of the communication is not examined.

Unix socket rule permissions are implied when a rule does not explicitly state an access list. By default if a rule does not have an access list all permissions that are compatible with the specified set of local and peer conditionals are implied.

The create, bind, listen, shutdown, getattr, setattr, getopt, and setopt permissions are local socket permissions. They are only applied to the local socket and can't be specified in rules that have a peer component. The accept permission applies to the combination of a local and peer socket. The connect, send, and receive permissions are peer socket permissions.

Only the peer socket permissions will be applied to rules that don't specify permissions and contain a peer component.

Example Unix domain socket rules:

  # Allow all permissions to unix sockets
  unix,
  # Explicitly allow all unix permissions
  unix (create, listen, accept, connect, send, receive, getattr, setattr, setopt, getopt),
  # Explicitly deny unix socket access
  deny unix,
  # Allow create and use of abstract and anonymous sockets for profile_name
  unix peer=(label=@{profile_name}),
  # Allow receiving via unix sockets from unconfined
  unix (receive) peer=(label=unconfined),
  # Allow getattr and shutdown on anonymous sockets
  unix (getattr, shutdown) addr=none,
  # Allow SOCK_STREAM connect, receive and send on an abstract socket @bar
  # with peer running under profile '/foo'
  unix (connect, receive, send) type=stream peer=(label=/foo,addr="@bar"),
  # Allow accepting connections from and receiving from peer running under
  # profile '/bar' on abstract socket '@foo'
  unix (accept, receive) addr=@foo peer=(label=/bar),

Abstract unix domain sockets autobind

Abstract unix domain sockets can autobind to an address. The autobind address is a unique 5 digit string of decimal numbers, eg. @00001. There is nothing that prevents a task from manually binding to addresses with a similar pattern so it is impossible to reliably identify autobind addresses from a regular address.

Interaction of network rules and fine grained unix domain socket rules

The coarse grained networking rules can be used to control unix domain sockets as well. When fine grained unix domain socket mediation is available the coarse grained network rule is mapped into the equivalent unix socket rule.

E.G.

    network unix,  =>  unix,
    network unix stream,   =>  unix stream,

Fine grained mediation rules however can not be losslessly converted back to the coarse grained network rule; e.g.

   unix bind addr=@example,

Has no exact match under coarse grained network rules, the closest match is the much wider permission rule of

   network unix,

change_profile rules¶

AppArmor supports self directed profile transitions via the change_profile api. Change_profile rules control which permissions for which profiles a confined task can transition to. The profile name can contain apparmor pattern matching to specify different profiles.

  change_profile -> **,

The change_profile api allows the transition to be delayed until when a task executes another application. If an exec rule transition is specified for the application and the change_profile api is used to make a transition at exec time, the transition specified by the change_profile api takes precedence.

The Change_profile permission can restrict which profiles can be transitioned to based off of the executable name by specifying the exec condition.

  change_profile /bin/bash -> new_profile,

The restricting of the transition profile to a given executable at exec time is only useful when then current task is allowed to make dynamic decisions about what confinement should be, but the decision set needs to be controlled. A list of profiles or multiple rules can be used to specify the profiles in the set. Eg.

  change_profile /bin/bash -> {new_profile1,new_profile2,new_profile3},

An exec rule can be used to specify a transition for the executable, if the transition should be allowed even if the change_profile api has not been used to select a transition for those available in the change_profile rule set. Eg.

  /bin/bash Px -> new_profile1,
  change_profile /bin/bash -> {new_profile1,new_profile2,new_profile3},

The exec mode dictates whether or not the Linux Kernel's unsafe_exec routines should be used to scrub the environment, similar to setuid programs. (See ld.so(8) for some information on setuid/setgid environment scrubbing.) The safe mode sets up environment scrubbing to occur when the new application is executed and unsafe mode disables AppArmor's requirement for environment scrubbing (the kernel and/or libc may still require environment scrubbing). An exec mode can only be specified when an exec condition is present.

  change_profile safe /bin/bash -> new_profile,

Not all kernels support safe mode and the parser will downgrade rules to unsafe mode in that situation. If no exec mode is specified, the default is safe mode in kernels that support it.

all rule¶

The all rule is used to add a generic rule for all supported rule types. This is useful when policy wants to define a black list instead of white list, but can also be useful to add an access qualifier to all rules.

Eg. Black list

  allow all,
  # begin blacklist
  deny file,
  deny unix,

Eg. Adding audit qualifier

  audit access all,

rlimit rules¶

AppArmor can set and control the resource limits associated with a profile as described in the setrlimit(2) man page.

The AppArmor rlimit controls allow setting of limits and restricting changes of them and these actions can be audited. Enforcement of the set limits is handled by the standard kernel enforcement mechanism for rlimits and will not result in an audited apparmor message if the limit is enforced.

If a profile does not have an rlimit rule associated with a given rlimit then the rlimit is left alone and regular access, including changing the limit, is allowed. However if the profile sets an rlimit then the current limit is checked and if greater than the limit specified in the rule it will be changed to the specified limit.

AppArmor rlimit rules control the hard limit of an application and ensure that if the hard limit is lowered that the soft limit does not exceed the hard limit value.

Eg.

  set rlimit data <= 100M,
  set rlimit nproc <= 10,
  set rlimit nice <= 5,

Variables¶

AppArmor's policy language allows embedding variables into file rules to enable easier configuration for some common (and pervasive) setups. Variables may have multiple values assigned, but any variable assignments must be made before the start of the profile.

The parser will automatically expand variables to include all values that they have been assigned; it is an error to reference a variable without setting at least one value. You can use empty quotes ("") to explicitly add an empty value.

At the time of this writing, the following variables are defined in the provided AppArmor policy:

  @{HOME}
  @{HOMEDIRS}
  @{multiarch}
  @{pid}
  @{pids}
  @{PROC}
  @{securityfs}
  @{apparmorfs}
  @{sys}
  @{tid}
  @{run}
  @{XDG_DESKTOP_DIR}
  @{XDG_DOWNLOAD_DIR}
  @{XDG_TEMPLATES_DIR}
  @{XDG_PUBLICSHARE_DIR}
  @{XDG_DOCUMENTS_DIR}
  @{XDG_MUSIC_DIR}
  @{XDG_PICTURES_DIR}
  @{XDG_VIDEOS_DIR}

These are defined in files in /etc/apparmor.d/tunables and are used in many of the abstractions described later.

You may also add files in /etc/apparmor.d/tunables/home.d for site-specific customization of @{HOMEDIRS}, /etc/apparmor.d/tunables/multiarch.d for @{multiarch} and /etc/apparmor.d/tunables/xdg-user-dirs.d for @{XDG_*}.

The special @{profile_name} variable is set to the profile name and may be used in all policy.

Alias rules¶

AppArmor also provides alias rules for remapping paths for site-specific layouts. They are an alternative form of path rewriting to using variables, and are done after variable resolution. Alias rules must occur within the preamble of the profile. System-wide aliases are found in /etc/apparmor.d/tunables/alias, which is included by /etc/apparmor.d/tunables/global. /etc/apparmor.d/tunables/global is typically included at the beginning of an AppArmor profile.

Globbing (AARE)¶

File resources and other parameters accepting an AARE may be specified with a globbing syntax similar to that used by popular shells, such as csh(1), bash(1), zsh(1).

*

can substitute for any number of characters, excepting '/'

**

can substitute for any number of characters, including '/'

?

can substitute for any single character excepting '/'

[abc]

will substitute for the single character a, b, or c

[a-c]

will substitute for the single character a, b, or c

[^a-c]

will substitute for any single character not matching a, b or c

{ab,cd}

will expand to one rule to match ab, one rule to match cd

Can also include variables.

@{variable}

will expand to all values assigned to the given variable.

When AppArmor looks up a directory the pathname being looked up will end with a slash (e.g., /var/tmp/); otherwise it will not end with a slash. Only rules that match a trailing slash will match directories. Some examples, none matching the /tmp/ directory itself, are:

/tmp/*

Files directly in /tmp.

/tmp/*/

Directories directly in /tmp.

/tmp/**

Files and directories anywhere underneath /tmp.

/tmp/**/

Directories anywhere underneath /tmp.

Rule Qualifiers¶

There are several rule qualifiers that can be applied to permission rules. Rule qualifiers can modify the rule and/or permissions within the rule.

allow

Specifies that permissions requests that match the rule are allowed. This is the default value for rules and does not need to be specified. Conflicts with the deny qualifier.

audit

Specifies that permissions requests that match the rule should be recorded to the audit log.

deny

Specifies that permissions requests that match the rule should be denied without logging. Can be combined with 'audit' to enable logging. Conflicts with the allow qualifier.

owner

Specifies that the task must have the same euid/fsuid as the object being referenced by the permission check.

Qualifier Blocks

Rule Qualifiers can be applied to multiple rules at a time by grouping the rules into a rule block.

  audit {
     /foo r,
     network,
  }

#include mechanism¶

AppArmor provides an easy abstraction mechanism to group common access requirements; this abstraction is an extremely flexible way to grant site-specific rights and makes writing new AppArmor profiles very simple by assembling the needed building blocks for any given program.

The use of '#include' is modelled directly after cpp(1); its use will replace the '#include' statement with the specified file's contents. The leading '#' is optional, and the '#include' keyword can be followed by an option conditional 'if exists' that specifies profile compilation should continue if the specified file or directory is not found.

#include "/absolute/path" specifies that /absolute/path should be used. #include "relative/path" specifies that relative/path should be used, where the path is relative to the current working directory. #include <magic/path> is the most common usage; it will load magic/path relative to a directory specified to apparmor_parser(8). /etc/apparmor.d/ is the AppArmor default.

The supplied AppArmor profiles follow several conventions; the abstractions stored in /etc/apparmor.d/abstractions/ are some large clusters that are used in most profiles. What follows are short descriptions of how some of the abstractions are used.

abstractions/audio

Includes accesses to device files used for audio applications.

abstractions/authentication

Includes access to files and services typically necessary for services that perform user authentication.

abstractions/base

Includes files that should be readable and writable in all profiles.

abstractions/bash

Includes many files used by bash; useful for interactive shells and programs that call system(3).

abstractions/consoles

Includes read and write access to the device files controlling the virtual console, sshd(8), xterm(1), etc. This abstraction is needed for many programs that interact with users.

abstractions/fonts

Includes access to fonts and the font libraries.

abstractions/gnome

Includes read and write access to GNOME configuration files, as well as read access to GNOME libraries.

abstractions/kde

Includes read and write access to KDE configuration files, as well as read access to KDE libraries.

abstractions/kerberosclient

Includes file access rules needed for common kerberos clients.

abstractions/nameservice

Includes file rules to allow DNS, LDAP, NIS, SMB, user and group password databases, services, and protocols lookups.

abstractions/perl

Includes read access to perl modules.

abstractions/user-download

abstractions/user-mail

abstractions/user-manpages

abstractions/user-tmp

abstractions/user-write

Some profiles for typical "user" programs will use these include files to describe rights that users have in the system.

abstractions/wutmp

Includes write access to files used to maintain wtmp(5) and utmp(5) databases, used with the w(1) and associated commands.

abstractions/X

Includes read access to libraries, configuration files, X authentication files, and the X socket.

Some of the abstractions rely on variables that are set in files in the /etc/apparmor.d/tunables/ directory. These variables are currently @{HOME} and @{HOMEDIRS}. Variables cannot be set in profile scope; they can only be set before the profile. Therefore, any profiles that use abstractions should either #include <tunables/global> or otherwise ensure that @{HOME} and @{HOMEDIRS} are set before starting the profile definition. The aa-autodep(8) and aa-genprof(8) utilities will automatically emit #include <tunables/global> in generated profiles.

Feature ABI¶

The feature abi tells AppArmor which feature set the policy was developed under. This is important to ensure that kernels with a different feature set don't enforce features that the policy doesn't support, which can result in unexpected application failures.

When policy is compiled both the kernel feature abi and policy feature abi are consulted to build a policy that will work for the system's kernel.

If the kernel supports a feature not supported by the policy then policy will be built so that the kernel does NOT enforce that feature.

If the policy supports a feature not supported by the kernel the compile may downgrade the rule with the feature to something the kernel supports, drop the rule completely, or fail the compile.

If the policy abi is specified as kernel then the running kernel's abi will be used. This should never be used in shipped policy as it can cause system breakage when a new kernel is installed.

ABI compatibility with AppArmor 2.x

AppArmor 3 remains compatible with AppArmor 2.x by detecting when a profile does not have a feature ABI specified. In this case the policy compile will either apply the pinned feature ABI as specified by the config file or the command line, or if neither of those are applied by using a default feature ABI.

It is important to note that the default feature ABI does not support new features added in AppArmor 3 or later.