AppArmor was putting the reference to i_private data on its end after
removing the original entry from the file system. However the inode
can aand does live beyond that point and it is possible that some of
the fs call back functions will be invoked after the reference has
been put, which results in a race between freeing the data and
accessing it through the fs.
While the rawdata/loaddata is the most likely candidate to fail the
race, as it has the fewest references. If properly crafted it might be
possible to trigger a race for the other types stored in i_private.
Fix this by moving the put of i_private referenced data to the correct
place which is during inode eviction.
Fixes: c961ee5f21 ("apparmor: convert from securityfs to apparmorfs for policy ns files")
Reported-by: Qualys Security Advisory <qsa@qualys.com>
Reviewed-by: Georgia Garcia <georgia.garcia@canonical.com>
Reviewed-by: Maxime Bélair <maxime.belair@canonical.com>
Reviewed-by: Cengiz Can <cengiz.can@canonical.com>
Signed-off-by: John Johansen <john.johansen@canonical.com>
An unprivileged local user can load, replace, and remove profiles by
opening the apparmorfs interfaces, via a confused deputy attack, by
passing the opened fd to a privileged process, and getting the
privileged process to write to the interface.
This does require a privileged target that can be manipulated to do
the write for the unprivileged process, but once such access is
achieved full policy management is possible and all the possible
implications that implies: removing confinement, DoS of system or
target applications by denying all execution, by-passing the
unprivileged user namespace restriction, to exploiting kernel bugs for
a local privilege escalation.
The policy management interface can not have its permissions simply
changed from 0666 to 0600 because non-root processes need to be able
to load policy to different policy namespaces.
Instead ensure the task writing the interface has privileges that
are a subset of the task that opened the interface. This is already
done via policy for confined processes, but unconfined can delegate
access to the opened fd, by-passing the usual policy check.
Fixes: b7fd2c0340 ("apparmor: add per policy ns .load, .replace, .remove interface files")
Reported-by: Qualys Security Advisory <qsa@qualys.com>
Tested-by: Salvatore Bonaccorso <carnil@debian.org>
Reviewed-by: Georgia Garcia <georgia.garcia@canonical.com>
Reviewed-by: Cengiz Can <cengiz.can@canonical.com>
Signed-off-by: John Johansen <john.johansen@canonical.com>
Add support for the per permission tag index for a given permission
set. This will be used by both meta-data tagging, to allow annotating
accept states with context and debug information. As well as by rule
tainting and triggers to specify the taint or trigger to be applied.
Since these are low frequency ancillary data items they are stored
in a tighter packed format to that allows for sharing and reuse of the
strings between permissions and accept states. Reducing the amount of
kernel memory use at the cost of having to go through a couple if
index based indirections.
The tags are just strings that has no meaning with out context. When
used as meta-data for auditing and debugging its entirely information
for userspace, but triggers, and tainting can be used to affect the
domain. However they all exist in the same packed data set and can
be shared between different uses.
Signed-off-by: John Johansen <john.johansen@canonical.com>
The set of rules on a profile is not dynamically extended, instead
if a new ruleset is needed a new version of the profile is created.
This allows us to use a vector of rules instead of a list, slightly
reducing memory usage and simplifying the code.
Signed-off-by: John Johansen <john.johansen@canonical.com>
The label struct is variable length. While its use in struct aa_profile
is fixed length at 2 entries the variable length member needs to be
the last member in the structure.
The code already does this but the comment has it in the wrong location.
Also add a comment to ensure it stays at the end of the structure.
While we are at it, update the documentation for other profile members
as well.
Signed-off-by: John Johansen <john.johansen@canonical.com>
Fine grained unix mediation in Ubuntu used ABI v7, and policy using
this has propogated onto systems where fine grained unix mediation was
not supported. The userspace policy compiler supports downgrading
policy so the policy could be shared without changes.
Unfortunately this had the side effect that policy was not updated for
the none Ubuntu systems and enabling fine grained unix mediation on
those systems means that a new kernel can break a system with existing
policy that worked with the previous kernel. With fine grained af_unix
mediation this regression can easily break the system causing boot to
fail, as it affect unix socket files, non-file based unix sockets, and
dbus communication.
To aoid this regression move fine grained af_unix mediation behind
a new abi. This means that the system's userspace and policy must
be updated to support the new policy before it takes affect and
dropping a new kernel on existing system will not result in a
regression.
The abi bump is done in such a way as existing policy can be activated
on the system by changing the policy abi declaration and existing unix
policy rules will apply. Policy then only needs to be incrementally
updated, can even be backported to existing Ubuntu policy.
Signed-off-by: John Johansen <john.johansen@canonical.com>
Extend af_unix mediation to support fine grained controls based on
the type (abstract, anonymous, fs), the address, and the labeling
on the socket.
This allows for using socket addresses to label and the socket and
control which subjects can communicate.
The unix rule format follows standard apparmor rules except that fs
based unix sockets can be mediated by existing file rules. None fs
unix sockets can be mediated by a unix socket rule. Where 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.
allow 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.
allow 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
allow unix rw addr=auto,
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.
allow unix addr=none,
If the address component of a rule is not specified then the rule
applies to autobind, abstract and anonymous sockets.
The label on the socket can be compared using the standard label=
rule conditional. Eg.
allow unix addr=@foo peer=(label=bar),
see man apparmor.d for full syntax description.
Signed-off-by: John Johansen <john.johansen@canonical.com>
Previously apparmor has only sent SIGKILL but there are cases where
it can be useful to send a different signal. Allow the profile
to optionally specify a different value.
Signed-off-by: John Johansen <john.johansen@canonical.com>
This is a step towards merging the file and policy state machines.
With the switch to extended permissions the state machine's ACCEPT2
table became unused freeing it up to store state specific flags. The
first flags to be stored are FLAG_OWNER and FLAG other which paves the
way towards merging the file and policydb perms into a single
permission table.
Currently Lookups based on the objects ownership conditional will
still need separate fns, this will be address in a following patch.
Signed-off-by: John Johansen <john.johansen@canonical.com>
In order to speed up the mediated check, precompute and store the
result as a bit per class type. This will not only allow us to
speed up the mediation check but is also a step to removing the
unconfined special cases as the unconfined check can be replaced
with the generic label_mediates() check.
Note: label check does not currently work for capabilities and resources
which need to have their mediation updated first.
Signed-off-by: John Johansen <john.johansen@canonical.com>
aa_label_audit, aa_label_find, aa_label_seq_print and aa_update_label_name
were added by commit
f1bd904175 ("apparmor: add the base fns() for domain labels")
but never used.
aa_profile_label_perm was added by commit
637f688dc3 ("apparmor: switch from profiles to using labels on contexts")
but never used.
aa_secid_update was added by commit
c092921219 ("apparmor: add support for mapping secids and using secctxes")
but never used.
aa_split_fqname has been unused since commit
3664268f19 ("apparmor: add namespace lookup fns()")
aa_lookup_profile has been unused since commit
93c98a484c ("apparmor: move exec domain mediation to using labels")
aa_audit_perms_cb was only used by aa_profile_label_perm (see above).
All of these commits are from around 2017.
Remove them.
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: John Johansen <john.johansen@canonical.com>
unprivileged unconfined can use change_profile to alter the confinement
set by the mac admin.
Allow restricting unprivileged unconfined by still allowing change_profile
but stacking the change against unconfined. This allows unconfined to
still apply system policy but allows the task to enter the new confinement.
If unprivileged unconfined is required a sysctl is provided to switch
to the previous behavior.
Reviewed-by: Georgia Garcia <georgia.garcia@canonical.com>
Signed-off-by: John Johansen <john.johansen@canonical.com>
With the move to permission tables the dfa is no longer a stand
alone entity when used, needing a minimum of a permission table.
However it still could be shared among different pdbs each using
a different permission table.
Instead of duping the permission table when sharing a pdb, add a
refcount to the pdb so it can be easily shared.
Reviewed-by: Georgia Garcia <georgia.garcia@canonical.com>
Signed-off-by: John Johansen <john.johansen@canonical.com>
The cred is needed to properly audit some messages, and will be needed
in the future for uid conditional mediation. So pass it through to
where the apparmor_audit_data struct gets defined.
Reviewed-by: Georgia Garcia <georgia.garcia@canonical.com>
Signed-off-by: John Johansen <john.johansen@canonical.com>
The last usage of PROF_{ADD,REPLACE} were removed by commit 18e99f191a
("apparmor: provide finer control over policy management"). So remove
these two unused macros.
Signed-off-by: GONG, Ruiqi <gongruiqi1@huawei.com>
Signed-off-by: John Johansen <john.johansen@canonical.com>
Bother unconfined and learning profiles use the null profile as their
base. Refactor so they are share a common base routine. This doesn't
save much atm but will be important when the feature set of the
parent is inherited.
Signed-off-by: John Johansen <john.johansen@canonical.com>
Convert profile->rules to a list as the next step towards supporting
multiple rulesets in a profile. For this step only support a single
list entry item. The logic for iterating the list will come as a
separate step.
Signed-off-by: John Johansen <john.johansen@canonical.com>
In preparation for moving from a single set of rules and a single
attachment to multiple rulesets and attachments separate from the
profile refactor attachment information and ruleset info into their
own structures.
Signed-off-by: John Johansen <john.johansen@canonical.com>
Currently permissions are encoded in the dfa accept entries that are
then mapped to an internal permission structure. This limits the
permissions that userspace can specify, so allow userspace to directly
specify the permission table.
Signed-off-by: John Johansen <john.johansen@canonical.com>
Allow the profile to contain a user mode prompt flag. This works similar
to complain mode but will try to send messages to a userspace daemon.
If the daemon is not present or timesout regular informent will occur.
Signed-off-by: John Johansen <john.johansen@canonical.com>
add indexes for label and tag entries. Rename the domain table to the
str_table as its a shared string table with label and tags.
Signed-off-by: John Johansen <john.johansen@canonical.com>
Convert from an unsigned int to a state_t for state position. This is
a step in prepping for the state position carrying some additional
flags, and a limited form of backtracking to support variables.
Signed-off-by: John Johansen <john.johansen@canonical.com>
Remap polidydb dfa accept table from embedded perms to an index, and
then move the perm lookup to use the accept entry as an index into the
perm table. This is done so that the perm table can be separated from
the dfa, allowing dfa accept to index to share expanded permission
sets.
Signed-off-by: John Johansen <john.johansen@canonical.com>
continue permission unification by converting xmatch to use the
policydb struct that is used by the other profile dfas.
Signed-off-by: John Johansen <john.johansen@canonical.com>
John Johansen
Dr. David Alan GilbertandJohn Johansen
GONG, RuiqiandJohn Johansen
Xiu JianfengandJohn Johansen