Don't ignore port ranges. For example:
network_port(amanda, udp,10080-10082,s0, tcp,10080-10083,s0)
can be used to generate:
base -A selinux_new_input -p udp --dport 10080:10082 -j SECMARK --selctx system_u:object_r:amanda_server_packet_t:s0
base -A selinux_new_input -p tcp --dport 10080:10083 -j SECMARK --selctx system_u:object_r:amanda_server_packet_t:s0
base -A selinux_new_output -p udp --dport 10080:10082 -j SECMARK --selctx system_u:object_r:amanda_client_packet_t:s0
base -A selinux_new_output -p tcp --dport 10080:10083 -j SECMARK --selctx system_u:object_r:amanda_client_packet_t:s0
Signed-off-by: Topi Miettinen <toiwoton@gmail.com>
Optionally generate Netfilter NFT tables. Sample output:
```#!/usr/sbin/nft -f
flush ruleset
table inet security {
secmark default_input_packet {
"system_u:object_r:server_packet_t:s0"
}
secmark default_output_packet {
"system_u:object_r:client_packet_t:s0"
}
secmark afs_bos_input {
"system_u:object_r:afs_bos_server_packet_t:s0"
}
secmark afs_bos_output {
"system_u:object_r:afs_bos_client_packet_t:s0"
}
...
chain INPUT {
type filter hook input priority 0; policy accept;
ct state new meta secmark set "default_input_packet"
ct state new udp dport 7007 meta secmark set "afs_bos_input"
...
ct state new ct secmark set meta secmark
ct state established,related meta secmark set ct secmark
}
chain FORWARD {
type filter hook forward priority 0; policy accept;
}
chain OUTPUT {
type filter hook output priority 0; policy accept;
ct state new meta secmark set "default_output_packet"
ct state new udp dport 7007 meta secmark set "afs_bos_output"
...
ct state new ct secmark set meta secmark
ct state established,related meta secmark set ct secmark
}
}
```
The labels are applied to TCP and/or UDP as needed. MCS and MLS are
not really handled.
Signed-off-by: Topi Miettinen <toiwoton@gmail.com>
When provisioning a Debian test virtual machine with VirtualBox (using
the main Vagrantfile), vagrant runs commands dhclient and ip from a ssh
session because of this directive:
debian.vm.network "private_network", type: "dhcp"
This triggers:
type=AVC msg=audit(1578749426.820:68): avc: denied { write } for
pid=541 comm="dhclient" path="pipe:[14006]" dev="pipefs" ino=14006
scontext=unconfined_u:unconfined_r:dhcpc_t
tcontext=system_u:system_r:sshd_t tclass=fifo_file permissive=1
type=PROCTITLE msg=audit(1578749426.820:68):
proctitle=2F7362696E2F6468636C69656E74002D34002D76002D69002D72002D
7066002F72756E2F6468636C69656E742E657468312E706964002D6C66002F7661
722F6C69622F646863702F6468636C69656E742E657468312E6C6561736573002D
49002D6466002F7661722F6C69622F646863702F6468636C69656E74362E657468
31
type=AVC msg=audit(1578749427.868:69): avc: denied { read } for
pid=544 comm="ip" path="pipe:[14005]" dev="pipefs" ino=14005
scontext=unconfined_u:unconfined_r:ifconfig_t
tcontext=system_u:system_r:sshd_t tclass=fifo_file permissive=1
type=AVC msg=audit(1578749427.868:69): avc: denied { write } for
pid=544 comm="ip" path="pipe:[14006]" dev="pipefs" ino=14006
scontext=unconfined_u:unconfined_r:ifconfig_t
tcontext=system_u:system_r:sshd_t tclass=fifo_file permissive=1
type=PROCTITLE msg=audit(1578749427.868:69):
proctitle=6970002D34006164647200666C757368006465760065746831006C61
62656C0065746831
Handle this by adding attribute vagrant_provisioning_cmd_type to the
relevant domains, in the policy module specific to the Vagrant test
environments.
Signed-off-by: Nicolas Iooss <nicolas.iooss@m4x.org>
When using Vagrant to run virtual machines with SELinux enabled, several
specific accesses need to be allowed. It does not make much sense to add
the needed rules to the refpolicy, as they are very specific to the use
of Vagrant to provision a virtual machine to test a policy. Therefore,
create a dedicated module to allow the required accesses.
Signed-off-by: Nicolas Iooss <nicolas.iooss@m4x.org>
"is" is for testing two references are the same object. The fact that this
worked is specific to the Python implementation.
Signed-off-by: Chris PeBenito <pebenito@ieee.org>
fc_sort is the only/last build tool that requires a C compiler
Re-implement it in python, so that gcc dependencies can be dropped
The output of the C and the python version differ slightly in the order of equally specific file contexts
old:
/.* system_u:object_r:default_t
/sys(/.*)? system_u:object_r:sysfs_t
/mnt(/[^/]*) -l system_u:object_r:mnt_t
/mnt(/[^/]*)? -d system_u:object_r:mnt_t
/opt/.* system_u:object_r:usr_t
/var/.* system_u:object_r:var_t
/usr/.* system_u:object_r:usr_t
/srv/.* system_u:object_r:var_t
/tmp/.* <<none>>
/run/.* <<none>>
/dev/.* system_u:object_r:device_t
/etc/.* system_u:object_r:etc_t
new:
/.* system_u:object_r:default_t
/sys(/.*)? system_u:object_r:sysfs_t
/mnt(/[^/]*) -l system_u:object_r:mnt_t
/mnt(/[^/]*)? -d system_u:object_r:mnt_t
/dev/.* system_u:object_r:device_t
/etc/.* system_u:object_r:etc_t
/opt/.* system_u:object_r:usr_t
/run/.* <<none>>
/srv/.* system_u:object_r:var_t
/tmp/.* <<none>>
/usr/.* system_u:object_r:usr_t
/var/.* system_u:object_r:var_t
Without this fix, building a custom module in a directory that contains
a file with special characters in its name (e.g. '(') triggers a syntax
error:
$ cat >foo.te <<EOF
module foo 1.0;
require {
class file entrypoint;
type shell_exec_t;
type vmtools_unconfined_t;
}
allow vmtools_unconfined_t shell_exec_t : file entrypoint;
EOF
$ touch "my broken (file)"
$ make -f /usr/share/selinux/devel/Makefile foo.pp
/bin/sh: -c: line 0: syntax error near unexpected token `('
/bin/sh: -c: line 0: `find anaconda-ks.cfg my broken (file) vncserver.strace systemd.strace rhel-server-7.6-x86_64-boot.iso rt_minimal.c vnc.cil foo.te rsyslog tmp virt-install.log evil_banner.sh livemedia.log program.log foo.if rhel7-minimal.ks TestZip.java TestZip.class foo.fc sudoloop foo.pp strace.log -maxdepth 0 -type d'
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1692676
Reported-by: Renaud Métrich <rmetrich@redhat.com>
Suggested-by: Petr Lautrbach <plautrba@redhat.com>
Signed-off-by: Ondrej Mosnacek <omosnace@redhat.com>
Have Makefile include CUSTOM_BUILDOPT in generated build.conf
Update Makefile.devel to pass CUSTOM_BUILDOPT while building module
Signed-off-by: Dave Sugar <dsugar@tresys.com>
The segenxml tool is used to generate documentation regarding the policy
definitions. Its output is an XML file that contains the in-line
comments associated with boolean generation as well as interface
definitions.
With booleans also generated inside templates, this information was
(until now) ignored. Templates such as apache's apache_content_template
which created new booleans were not properly documented, as the
in-template comments were ignored.
In this patch, we will go over module code first and seek template
calls. When a template call is matched, the module code is updated
(expanded) with the template content (while substituting the arguments
to get a proper code listing). Only after all templates have been
expanded we seek the necessary boolean definitions.
Changes since v2:
- Fix BOOLEAN statements to match backtick (`) and tick (') usages as
well
- Fix match for arguments to also include multiple entries ( { ... } )
Changes since v1:
- Also apply the regexp on BOOLEAN to allow generating templated
boolean/tunable documentation
Signed-off-by: Sven Vermeulen <sven.vermeulen@siphos.be>
Some of the templates in the reference policy generate new booleans and
tunables, based on the $1, $2, ... parameters passed on. To allow
segenxml, which generates the necessary documentation on booleans, to
keep track of template-generated booleans as well, we need to allow it
to substitute template calls with the actual template content.
The gentemplates.sh script is a helper script that will extract template
code and store it as files (one file per template). These files are then
later on used by the segenxml tool.
Signed-off-by: Sven Vermeulen <sven.vermeulen@siphos.be>
python3.6 will error out with the message "invalid escape sequence"
in genhomedircon.py. This patch fixes these errors by turning the string
in the into a raw string.
Avoid memory leakages in the fc_sort executable (now passes
all valgrind AND Clang static analyzer tests fine).
Some NULL pointer checks with or without associated error
reporting.
Some white space and comment formatting fixes.
Optimization: avoid unnecessary operations (unnecessary
memory allocation/deallocation and list copying).
Reverts 7821eb6f37 as such
trick is no longer needed, given that all memory leakages
have now been fixed.
This is the sixth version of this patch. Please do not use
the first version as it introduces a serious bug.
For reference, the original issue reported by the Cland
static analyzer is as follows:
support/fc_sort.c:494:6: warning: Potential leak of memory
pointed to by 'head'
malloc(sizeof(file_context_bucket_t));
Signed-off-by: Guido Trentalancia <guido@trentalancia.com>
Acked-by: William Roberts <william.c.roberts@intel.com>
Using the LLVM static analyzer, we get a warning about leaked memory
pointed by bcurrent. Since the warning happens within "main" and we
don't care about deallocating the memory and just call "exit" which
gets rid of the warning.
Strip trailing space(s) from 'NAME' being read from /etc/selinux/config as a trailing space in that name will cause a weird error during the make process for modules built out of tree. This is seen on current RedHat machines which have space in the 'SELINUXTYPE' variable.
Fix a "-Werror=implicit-fallthrough" compiler warning/error on
the switch statement.
This third version (v3) fixes a bug introduced in the first
version and improves the style over the second version.
Signed-off-by: Guido Trentalancia <guido at trentalancia.com>
Every Infiniband network will have a default pkey, so that is labeled.
The rest of the pkey configuration is network specific. The policy allows
access to the default and unlabeled pkeys for sysadm and staff users.
kernel_t is allowed access to all pkeys, which it needs to process and
route management datagrams.
Endports are all unlabeled by default, sysadm users are allowed to
manage the subnet on unlabeled endports. kernel_t is allowed to manage
the subnet on all ibendports, which is required for configuring the HCA.
This patch requires selinux series: "SELinux user space support for
Infiniband RDMA", due to the new ipkeycon labeling mechanism.
Signed-off-by: Daniel Jurgens <danielj@mellanox.com>
Python 3.6 complains about the strings which are used as regular
expression in the support scripts:
File "support/segenxml.py", line 37
INTERFACE = re.compile("^\s*(interface|template)\(`(\w*)'")
^
SyntaxError: invalid escape sequence \s
When compiling modules using support/Makefile.devel (which is installed
in /usr/share/selinux/*/include/Makefile) with "make -j9", the build
fails because tmp/ does not exist.
Add the missing command to create tmp/ when running tmp/%.mod.fc target.
Gentoo bug: https://bugs.gentoo.org/show_bug.cgi?id=530178
When working on fc_sort to try to understand why using /usr/s?bin/... file
contexts has been reported not to work properly [1], I found it frustrating not
to be able to do "tmp/fc_sort my_filecontexts.fc" and see the result printed on
the screen. This patch implements this behavior by making optional the second
argument of fc_sort.
[1] commit 36e2216f8 of contrib repository,
http://git.overlays.gentoo.org/gitweb/?p=proj/hardened-refpolicy.git;a=commit;h=36e2216f82192660d063012e69281f27ba20864b
When running fc_sort on an empty context file, this program uses uninitialized
pointers when accessing to the elements of a list. On my system, it goes in a
very long loop (maybe infinite) because uninitialized fields in malloc'ed
structures happen to contain valid pointers in the heap.
This patch fixes this bug by initializing ->next and ->data fields before they
may be read.
This patch includes the necessary refactoring to support python 3.
Changes since v2
- Do not include contrib submodule (no relevant changes there)
- Update in pyplate to fix a failure with str/unicode in doc generation
Signed-off-by: Sven Vermeulen <sven.vermeulen@siphos.be>
system_r is required by the policy_module macro, which however will
be expanded as empty if the module is built into base.pp. system_r
is defined in the kernel.te, its definition should be moved to the
top of base.conf so that other modules copied earlier into base.conf
than kernel.te could reference system_r in their unconditional block
properly.
Signed-off-by: Harry Ciao <qingtao.cao@windriver.com>
