Also enable global reading of kernel policy file. Motivation for this is to
allow read access to the kernel version of the binary selinux policy.
Change-Id: I1eefb457cea1164a8aa9eeb7683b3d99ee56ca99
Fix two neverallow rules that yield Invalid SELinux context
warnings from the CTS SELinuxTest.
For transitions from app domains, we only need to check
{ domain -appdomain } (i.e. domains other than app domains),
not ~appdomain (i.e. all types other than app domains). Otherwise
SELinuxTest tries to generate contexts with the r role and
non-domain types for testing since the target class is process,
and such contexts are invalid.
For keeping file_type and fs_type exclusive, we only need to
check associate permission, not all filesystem permissions, as
only associate takes a file type as the source context. Otherwise
SELinuxTest tries to generate contexts with the r role and
non-domain types for testing filesystem permissions other than
associate, since the source of such checks is normally a process
context.
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
(cherry picked from commit 21ada26dae)
Change-Id: I3346584da9b89f352864dcc30dde06d6bf42e98e
Fix two neverallow rules that yield Invalid SELinux context
warnings from the CTS SELinuxTest.
For transitions from app domains, we only need to check
{ domain -appdomain } (i.e. domains other than app domains),
not ~appdomain (i.e. all types other than app domains). Otherwise
SELinuxTest tries to generate contexts with the r role and
non-domain types for testing since the target class is process,
and such contexts are invalid.
For keeping file_type and fs_type exclusive, we only need to
check associate permission, not all filesystem permissions, as
only associate takes a file type as the source context. Otherwise
SELinuxTest tries to generate contexts with the r role and
non-domain types for testing filesystem permissions other than
associate, since the source of such checks is normally a process
context.
Change-Id: I6c2f63f4786d75294a6938613ba14b64212fc802
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Currently, dex2oat runs in the installd sandbox, and has
all the SELinux capabilities that installd does. That's too
excessive.
dex2oat handles untrusted user data, so we want to put it in
it's own tighter sandbox.
Bug: 15358102
Change-Id: I08083b84b9769e24d6dad6dbd12401987cb006be
Rename sdcard_internal/external types to fuse and vfat
respectively to make it clear that they are assigned to any
fuse or vfat filesystem by default (absent a context= mount option)
and do not necessarily represent the SDcard.
The sdcard_type attribute is still assigned to both types and
can still be used in allow rules to permit access to either the
internal or external SDcard.
Define type aliases for the old names to preserve compatibility
on policy reload and for device-specific policies that may not yet
be updated.
Change-Id: I8d91a8c4c1342b94e4f1bb62ca7ffd2ca4b06ba1
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Many of the neverallow rules have -unconfineddomain. This was
intended to allow us to support permissive_or_unconfined(), and
ensure that all domains were enforcing at least a minimal set of
rules.
Now that all the app domains are in enforcing / confined, there's
no need to allow for these exceptions. Remove them.
Change-Id: Ieb29872dad415269f7fc2fe5be5a3d536d292d4f
Remove the auditallow statements from app.te and
binderservicedomain.te which were causing log spam.
Change-Id: If1c33d1612866df9f338e6d8c19d73950ee028eb
Add keystore_key class and an action for each action supported
by keystore. Add policies that replicate the access control that
already exists in keystore. Add auditallow rules for actions
not known to be used frequently. Add macro for those domains
wishing to access keystore.
Change-Id: Iddd8672b9e9b72b45ee208e6eda608cc9dc61edc
sdcard_internal is assigned to fuse mounts while sdcard_external
is assigned to vfat mounts by genfs_contexts. Originally we
allowed access to both via the sdcard_type attribute, and access
via both means was required. IIUC however, in 4.4 and later,
SDcard access should always occur via the fuse mount and we can
drop access to sdcard_external.
I think we can do the same for all domains except sdcardd. However,
I cannot test this as the Nexus devices do not have external SDcard
support.
Also wondering if we should rename sdcard_internal type to fuse
and sdcard_external type to vfat to more clearly represent their
meaning, since one accesses the external SDcard via the fuse mount now.
Change-Id: Ie44221e9eea90e627a48df5398c456b86293f724
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
The following commits added support for runtime resource overlays.
New command line tool 'idmap'
* 65a05fd56dbc9fd9c2511a97f49c445a748fb3c5
Runtime resource overlay, iteration 2
* 48d22323ce39f9aab003dce74456889b6414af55
Runtime resource overlay, iteration 2, test cases
* ad6ed950dbfa152c193dd7e49c369d9e831f1591
During SELinux tightening, support for these runtime resource
overlays was unknowingly broken. Fix it.
This change has been tested by hackbod and she reports that
everything is working after this change. I haven't independently
verified the functionality.
Test cases are available for this by running:
* python frameworks/base/core/tests/overlaytests/testrunner.py
Change-Id: I1c70484011fd9041bec4ef34f93f7a5509906f40
Remove /data/dalvik-cache/profiles from domain. Profiling information
leaks data about how people interact with apps, so we don't want
the data to be available in all SELinux domains.
Add read/write capabilities back to app domains, since apps need to
read/write profiling data.
Remove restorecon specific rules. The directory is now created by
init, not installd, so installd doesn't need to set the label.
Change-Id: Ic1b44009faa30d704855e97631006c4b990a4ad3
Originally we used the shell domain for ADB shell only and
the init_shell domain for the console service, both transitioned
via automatic domain transitions on sh. So they originally
shared a common set of rules. Then init_shell started to be used
for sh commands invoked by init.<board>.rc files, and we switched
the console service to just use the shell domain via seclabel entry
in init.rc. Even most of the sh command instances in init.<board>.rc
files have been converted to use explicit seclabel options with
more specific domains (one lingering use is touch_fw_update service
in init.grouper.rc). The primary purpose of init_shell at this point
is just to shed certain permissions from the init domain when init invokes
a shell command. And init_shell and shell are quite different in
their permission requirements since the former is used now for
uid-0 processes spawned by init whereas the latter is used for
uid-shell processes spawned by adb or init.
Given these differences, drop the shelldomain attribute and take those
rules directly into shell.te. init_shell was an unconfined_domain(),
so it loses nothing from this change. Also switch init_shell to
permissive_or_unconfined() so that we can see its actual denials
in the future in userdebug/eng builds.
Change-Id: I6e7e45724d1aa3a6bcce8df676857bc8eef568f0
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
As of sepolicy commit a16a59e2c7
(https://android-review.googlesource.com/94580), adf_device
and graphics_device have the exact same security properties.
Merge them into one type to avoid a proliferation of SELinux
types.
Change-Id: Ib1a24f5d880798600e103b9e14934e41abb1ef95
This is to accomodate migration to (and ongoing support of) a
new installed-app file topology, in which APK files are placed
in /data/app/$PACKAGE-rev/, there is a canonical-path symlink
/data/app/$PACKAGE/ -> /data/app/$PACKAGE-rev/, and the native
libraries exist not under a top-level /data/app-lib/$PACKAGE-rev
hard directory, but rather under /data/app/$PACKAGE/lib (when
referenced by canonical path).
Change-Id: I4f60257f8923c64266d98aa247bffa912e204fb0
NDK r8c and below induced text relocations into every NDK
compiled shared library. (https://code.google.com/p/android/issues/detail?id=23203).
For compatibility, we need to support shared libraries with text relocations
in them.
Addresses the following error / denial:
06-02 13:28:59.495 3634 3634 W linker : libCore.so has text relocations. This is wasting memory and prevents security hardening. Please fix.
<4>[ 57.430677] type=1400 audit(1401740939.756:13): avc: denied { execmod } for pid=3634 comm=".playandlearnhd" path="/data/app-lib/com.adobe.air-2/libCore.so" dev="mmcblk0p28" ino=32745 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:system_data_file:s0 tclass=file
Steps to reproduce:
1) Install Adobe AIR (https://play.google.com/store/apps/details?id=com.adobe.air)
2) Install PBS Parents Play & Learn (https://play.google.com/store/apps/details?id=air.org.pbskids.playandlearnhd)
3) Attempt to run Play & Learn app
Expected:
App runs
Actual:
App crashes with error above.
Bug: 15388851
Change-Id: I88bfd72b2abf2407803da0209d2313c8210c6663
Remove /data/security and setprop selinux.reload_policy access
from unconfineddomain, and only add back what is needed to
init (system_server already gets the required allow rules via
the selinux_manage_policy macro).
init (via init.rc post-fs-data) originally creates /data/security
and may later restorecon it. init also sets the property (also from
init.rc post-fs-data) to trigger a reload once /data is mounted.
The system_server (SELinuxPolicyInstallReceiver in particular) creates
subdirectories under /data/security for updates, writes files to these
subdirectories, creates the /data/security/current symlink to the update
directory, and sets the property to trigger a reload when an update bundle
is received.
Add neverallow rules to ensure that we do not allow undesired access
to security_file or security_prop.
This is only truly meaningful if the support for /data/security policies
is restored, but is harmless otherwise.
Also drop the persist.mmac property_contexts entry; it was never used in
AOSP, only in our tree (for middleware MAC) and is obsolete.
Change-Id: I5ad5e3b6fc7abaafd314d31723f37b708d8fcf89
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
As suggested in https://android-review.googlesource.com/95966 , remove
various syslog_* from unconfined. SELinux domains which want to use
syslog_* can declare it themselves.
Change-Id: I7a8335850d1b8d3463491b4ef8c657f57384cfa4
Allow the shell user to see the dmesg output. This data is already
available via "adb bugreport", but isn't easy to access.
Bug: 10020939
Change-Id: I9d4bbbd41cb02b707cdfee79f826a39c1ec2f177
Define a domain and appropriate access rules for shared RELRO files
(used for loading the WebView native library). Any app is permitted to
read the files as they are public data, but only the shared_relro
process is permitted to create/update them.
Bug: 13005501
Change-Id: I9d5ba9e9eedb9b8c80fe6f84a3fc85a68553d52e
On userdebug / eng builds, Android supports the concept of app wrapping.
You can run an app wrapped by another process. This is traditionally used
to run valgrind on apps, looking for memory leaks and other problems.
App wrapping is enabled by running the following command:
adb shell setprop wrap.com.android.foo "TMPDIR=/data/data/com.android.foo logwrapper valgrind"
Valgrind attempts to mmap exec /system/bin/app_process, which is being denied
by SELinux. Allow app_process exec.
Addresses the following denial:
<4>[ 82.643790] type=1400 audit(16301075.079:26): avc: denied { execute } for pid=1519 comm="memcheck-arm-li" path="/system/bin/app_process32" dev="mmcblk0p25" ino=61 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:zygote_exec:s0 tclass=file
Bug: 15146424
Change-Id: I65394938c53da9252ea57856d9f2de465bb30c25
Commit: 7ffb997207 added protection against low
memory mapping for all domains, a superset of appdomain. Remove the same,
redundant neverallow rule from appdomain.
Change-Id: Ia41c02763f6b5a260c56d10adfbab649d9f3f97c
Should no longer be required due to restorecon_recursive of /data
by init.rc (covers /data/dalvik-cache and /data/app-lib) and due to
restorecon_recursive of /data/data by installd (covers /data/data
directories).
Change-Id: Icb217c0735852db7cca8583e381264ef8cd8839c
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
ADF is a modern replacement for fbdev.
ADF's device nodes (/dev/adf[X]), interface nodes
(/dev/adf-interface[X].[Y]), and overlay engine nodes
(/dev/adf-overlay-engine[X].[Y]) are collectively used in similar
contexts as fbdev nodes. Vendor HW composers (via SurfaceFlinger) and
healthd will need to send R/W ioctls to these nodes to prepare and
update the display.
Ordinary apps should not talk to ADF directly.
Change-Id: Ic0a76b1e82c0cc1e8f240f219928af1783e79343
Signed-off-by: Greg Hackmann <ghackmann@google.com>
We were using system_data_file for the /data/data directories of
system UID apps to match the DAC ownership of system UID shared with
other system files. However, we are seeing cases where files created
in these directories must be writable by other apps, and we would like
to avoid allowing write to system data files outside of these directories.
So introduce a separate system_app_data_file type and assign it.
This should also help protect against arbitrary writes by system UID
apps to other system data directories.
This resolves the following denial when cropping or taking a user photo
for secondary users:
avc: denied { write } for path="/data/data/com.android.settings/cache/TakeEditUserPhoto2.jpg" dev="mmcblk0p28" ino=82120 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:system_data_file:s0 tclass=file
avc: denied { write } for path="/data/data/com.android.settings/cache/CropEditUserPhoto.jpg" dev="mmcblk0p30" ino=602905 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:system_data_file:s0 tclass=file
Bug: 14604553
Change-Id: Ifa10e3283b07f6bd6ecc16eceeb663edfd756cea
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Commit 3fbc536dfd allowed untrusted
app to read radio data files passed via binder, but didn't allow
write access. Write access is needed when sending MMS messages.
Steps to reproduce:
1) have some photos on the device
2) Launch messaging app
3) Attach a MMS (Picture, capture video, capture picture, audio recording etc..)
4) Send
EXPECTED RESULTS:
No crash
OBSERVED RESULTS:
- Messaging crashes on sending MMS
- messages are stuck in sending state
Additional details:
05-05 10:14:01.196 2457 2457 W Binder_3: type=1400 audit(0.0:20): avc: denied { write } for path="/data/data/com.android.providers.telephony/app_parts/PART_1399310041183_temp.jpg" dev="mmcblk0p23" ino=604417 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:radio_data_file:s0 tclass=file
05-05 10:14:01.202 27809 28219 E JavaBinder: !!! FAILED BINDER TRANSACTION !!!
05-05 10:14:01.203 27809 28219 E PduPersister: Failed to open Input/Output stream.
05-05 10:14:01.203 27809 28219 E PduPersister: java.io.FileNotFoundException: Failed opening content provider: content://mms/part/4
05-05 10:14:01.203 27809 28219 E PduPersister: at android.content.ContentResolver.openAssetFileDescriptor(ContentResolver.java:966)
05-05 10:14:01.203 27809 28219 E PduPersister: at android.content.ContentResolver.openOutputStream(ContentResolver.java:674)
05-05 10:14:01.203 27809 28219 E PduPersister: at android.content.ContentResolver.openOutputStream(ContentResolver.java:650)
05-05 10:14:01.203 27809 28219 E PduPersister: at com.google.android.mms.pdu.PduPersister.persistData(PduPersister.java:837)
05-05 10:14:01.203 27809 28219 E PduPersister: at com.google.android.mms.pdu.PduPersister.persistPart(PduPersister.java:761)
05-05 10:14:01.203 27809 28219 E PduPersister: at com.google.android.mms.pdu.PduPersister.persist(PduPersister.java:1398)
05-05 10:14:01.203 27809 28219 E PduPersister: at com.android.mms.data.WorkingMessage.createDraftMmsMessage(WorkingMessage.java:1577)
05-05 10:14:01.203 27809 28219 E PduPersister: at com.android.mms.data.WorkingMessage.sendMmsWorker(WorkingMessage.java:1431)
05-05 10:14:01.203 27809 28219 E PduPersister: at com.android.mms.data.WorkingMessage.access$700(WorkingMessage.java:82)
05-05 10:14:01.203 27809 28219 E PduPersister: at com.android.mms.data.WorkingMessage$2.run(WorkingMessage.java:1228)
05-05 10:14:01.203 27809 28219 E PduPersister: at java.lang.Thread.run(Thread.java:818)
05-05 10:14:01.221 27809 28219 E AndroidRuntime: FATAL EXCEPTION: WorkingMessage.send MMS
05-05 10:14:01.221 27809 28219 E AndroidRuntime: Process: com.android.mms, PID: 27809
05-05 10:14:01.221 27809 28219 E AndroidRuntime: java.lang.NullPointerException: Attempt to invoke virtual method 'java.lang.String android.net.Uri.getLastPathSegment()' on a null object reference
05-05 10:14:01.221 27809 28219 E AndroidRuntime: at android.content.ContentUris.parseId(ContentUris.java:85)
05-05 10:14:01.221 27809 28219 E AndroidRuntime: at com.android.mms.model.SlideshowModel.finalResize(SlideshowModel.java:691)
05-05 10:14:01.221 27809 28219 E AndroidRuntime: at com.android.mms.data.WorkingMessage.sendMmsWorker(WorkingMessage.java:1448)
05-05 10:14:01.221 27809 28219 E AndroidRuntime: at com.android.mms.data.WorkingMessage.access$700(WorkingMessage.java:82)
05-05 10:14:01.221 27809 28219 E AndroidRuntime: at com.android.mms.data.WorkingMessage$2.run(WorkingMessage.java:1228)
05-05 10:14:01.221 27809 28219 E AndroidRuntime: at java.lang.Thread.run(Thread.java:818)
05-05 10:14:01.222 659 5253 W ActivityManager: Force finishing activity com.android.mms/.ui.ComposeMessageActivity
Bug: 14562421
Change-Id: Iba6914eeec4bf0c8c04ee83584327a4824c0a9a9
To see whether we can safely remove these allow rules on unlabeled files
since we now have restorecon_recursive /data in init.rc to fully relabel
legacy userdata partitions, audit all accesses on such files.
Exclude the init domain since it performs the restorecon_recursive /data
and therefore will read unlabeled directories, stat unlabeled files,
and relabel unlabeled directories and files on upgrade. init may also
create/write unlabeled files in /data prior to the restorecon_recursive
/data being called.
Exclude the kernel domain for search on unlabeled:dir as this happens
during cgroup filesystem initialization in the kernel as a side effect
of populating the cgroup directory during the superblock initialization
before SELinux has set the label on the root directory.
Change-Id: Ieb5d807f529db9a4bf3e6c93e6b37c9648c04633
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
I9b8e59e3bd7df8a1bf60fa7ffd376a24ba0eb42f added a profiles
subdirectory to /data/dalvik-cache with files that must be
app-writable. As a result, we have denials such as:
W/Profiler( 3328): type=1400 audit(0.0:199): avc: denied { write } for name="com.google.android.setupwizard" dev="mmcblk0p28" ino=106067 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:dalvikcache_data_file:s0 tclass=file
W/Profiler( 3328): type=1300 audit(0.0:199): arch=40000028 syscall=322 per=800000 success=yes exit=33 a0=ffffff9c a1=b8362708 a2=20002 a3=0 items=1 ppid=194 auid=4294967295 uid=10019 gid=10019 euid=10019 suid=10019 fsuid=10019 egid=10019 sgid=10019 fsgid=10019 tty=(none) ses=4294967295 exe="/system/bin/app_process" subj=u:r:untrusted_app:s0 key=(null)
W/auditd ( 286): type=1307 audit(0.0:199): cwd="/"
W/auditd ( 286): type=1302 audit(0.0:199): item=0 name="/data/dalvik-cache/profiles/com.google.android.setupwizard" inode=106067 dev=b3:1c mode=0100664 ouid=1012 ogid=50019 rdev=00:00 obj=u:object_r:dalvikcache_data_file:s0
We do not want to allow untrusted app domains to write to the
existing type on other /data/dalvik-cache files as that could be used
for code injection into another app domain, the zygote or the system_server.
So define a new type for this subdirectory. The restorecon_recursive /data
in init.rc will fix the labeling on devices that already have a profiles
directory created. For correct labeling on first creation, we also need
a separate change to installd under the same change id.
Bug: 13927667
Change-Id: I4857d031f9e7e60d48b8c72fcb22a81b3a2ebaaa
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
This change folds the shared_app, media_app, and release_app
domains into untrusted_app, reducing the set of app domains down
to just distinct domains for the fixed UID apps (e.g. system_app, bluetooth,
nfc, radio), a single domain for apps signed by the platform key
(platform_app), and a single domain for all other apps (untrusted_app).
Thus, SELinux only distinguishes when already distinguished by a predefined
Android ID (AID) or by the platform certificate (which get the signature-only
Android permissions and thus may require special OS-level accesses).
It is still possible to introduce specific app domains for specific
apps by adding signer and package stanzas to mac_permissions.xml,
but this can be done on an as-needed basis for specialized apps that
require particular OS-level permissions outside the usual set.
As there is now only a single platform app domains, get rid of the
platformappdomain attribute and platform_app_domain() macro. We used
to add mlstrustedsubject to those domains but drop this since we are not
using MLS in AOSP presently; we can revisit which domains need it if/when
we use MLS.
Since we are dropping the shared, media, and release seinfo entries from
seapp_contexts, drop them from mac_permissions.xml as well. However,
we leave the keys.conf entries in case someone wants to add a signer
entry in the future for specific apps signed by those keys to
mac_permissions.xml.
Change-Id: I877192cca07360c4a3c0ef475f016cc273e1d968
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
This appears to have been created to allow untrusted_app to
access DownloadProvider cache files without needing to allow
open access to platform_app_data_file. Now that platform_app_data_file
is gone, there is no benefit to having this type.
Retain a typealias for download_file to app_data_file until
restorecon /data/data support is in place to provide compatibility.
This change depends on:
https://android-review.googlesource.com/#/c/87801/
Change-Id: Iab3c99d7d5448bdaa5c1e03a98fb6163804e1ec4
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
The original concept was to allow separation between /data/data/<pkgdir>
files of "platform" apps (signed by one of the four build keys) and
untrusted apps. But we had to allow read/write to support passing of
open files via Binder or local socket for compatibilty, and it seems
that direct open by pathname is in fact used in Android as well,
only passing the pathname via Binder or local socket. So there is no
real benefit to keeping it as a separate type.
Retain a type alias for platform_app_data_file to app_data_file until
restorecon /data/data support is in place to provide compatibility.
Change-Id: Ic15066f48765322ad40500b2ba2801bb3ced5489
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
We already have neverallow rules for all domains about
loading policy, setting enforcing mode, and setting
checkreqprot, so we can drop redundant ones from netd and appdomain.
Add neverallow rules to domain.te for setbool and setsecparam
and exclude them from unconfined to allow fully eliminating
separate neverallow rules on the :security class from anything
other than domain.te.
Change-Id: I0122e23ccb2b243f4c5376893e0c894f01f548fc
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Coalesce a number of allow rules replicated among multiple
app domains.
Get rid of duplicated rules already covered by domain, appdomain,
or platformappdomain rules.
Split the platformappdomain rules to their own platformappdomain.te
file, document them more fully, and note the inheritance in each
of the relevant *_app.te files.
Generalize isolated app unix_stream_socket rules to all app domains
to resolve denials such as:
avc: denied { read write } for pid=11897 comm="Binder_2" path="socket:[203881]" dev="sockfs" ino=203881 scontext=u:r:release_app:s0 tcontext=u:r:untrusted_app:s0 tclass=unix_stream_socket
avc: denied { getattr } for pid=11990 comm=4173796E635461736B202334 path="socket:[203881]" dev="sockfs" ino=203881 scontext=u:r:release_app:s0 tcontext=u:r:untrusted_app:s0 tclass=unix_stream_socket
avc: denied { getopt } for pid=11990 comm=4173796E635461736B202334 scontext=u:r:release_app:s0 tcontext=u:r:untrusted_app:s0 tclass=unix_stream_socket
avc: denied { read write } for pid=6890 comm="Binder_10" path="socket:[205010]" dev="sockfs" ino=205010 scontext=u:r:release_app:s0 tcontext=u:r:media_app:s0 tclass=unix_stream_socket
avc: denied { getattr } for pid=11990 comm=4173796E635461736B202334 path="socket:[205010]" dev="sockfs" ino=205010 scontext=u:r:release_app:s0 tcontext=u:r:media_app:s0 tclass=unix_stream_socket
avc: denied { getopt } for pid=11990 comm=4173796E635461736B202334 scontext=u:r:release_app:s0 tcontext=u:r:media_app:s0 tclass=unix_stream_socket
Change-Id: I770d7d51d498b15447219083739153265d951fe5
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
A number of binder_call rules are duplicated by other rules
written in terms of attributes/sets (e.g. appdomain, binderservicedomain).
Get rid of the duplicates.
Also use binder_use() in racoon.te rather than manually writing the
base rule for communicating with the servicemanager.
Change-Id: I5a459cc2154b1466bcde6eccef253dfcdcb44e0a
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Change I6dacdc43bcc1a56e47655e37e825ee6a205eb56b switched
the keystore to using binder instead of a socket, so this
socket type and rules have been unused for a while. The type
was only ever assigned to a /dev/socket socket file (tmpfs) so
there is no issue with removing the type (no persistent files
will have this xattr value).
Change-Id: Id584233c58f6276774c3432ea76878aca28d6280
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
There is some overlap between socket rules in app.te and the net.te rules,
but they aren't quite identical since not all app domains presently include
the net_domain() macro and because the rules in app.te allow more permissions
for netlink_route_socket and allow rawip_socket permissions for ping.
The current app.te rules prevent one from ever creating a non-networked app
domain. Resolve this overlap by:
1) Adding the missing permissions allowed by app.te to net.te for
netlink_route_socket and rawip_socket.
2) Adding net_domain() calls to all existing app domains that do not already
have it.
3) Deleting the redundant socket rules from app.te.
Then we'll have no effective change in what is allowed for apps but
allow one to define app domains in the future that are not allowed
network access.
Also cleanup net.te to use the create_socket_perms macro rather than *
and add macros for stream socket permissions.
Change-Id: I6e80d65b0ccbd48bd2b7272c083a4473e2b588a9
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
In order to prevent Zygote descriptors from leaking into the child
environment, they should be closed by the forked-off child process
before the child switches to the application UID. These changes close
the descriptors via dup2(), substituting a descriptor open to
/dev/null in their place; this allows the Zygote Java code to close
the FileDescriptor objects cleanly.
This is a multi-project change: dalvik, art, libcore, frameworks/base,
and external/sepolicy are affected. The CLs need to be approved
together, lest the build break or the software fail to boot.
Bug: 12114500
Change-Id: Ie45ddf6d661a1ea8570cd49dfea76421f2cadf72
- Add write_logd, read_logd and control_logd macros added along
with contexts for user space logd.
- Specify above on domain wide, or service-by-service basis
- Add logd rules.
- deprecate access_logcat as unused.
- 'allow <domain> zygote:unix_dgram_socket write;' rule added to
deal with fd inheritance. ToDo: investigate means to allow
references to close, and reopen in context of application
or call setsockcreatecon() to label them in child context.
Change-Id: I35dbb9d5122c5ed9b8c8f128abf24a871d6b26d8
When setting a static wallpaper on multiple devices the
following denials were encountered.
avc: denied { getattr } for pid=1775 comm="llpaper_chooser" path="/data/system/users/0/wallpaper" dev="mmcblk0p23" ino=104679 scontext=u:r:shared_app:s0 tcontext=u:object_r:wallpaper_file:s0 tclass=file
avc: denied { getattr } for pid=799 comm="ndroid.systemui" path="/data/system/users/0/wallpaper" dev="mmcblk0p23" ino=104679 scontext=u:r:platform_app:s0 tcontext=u:object_r:wallpaper_file:s0 tclass=file
avc: denied { getattr } for pid=1909 comm=4173796E635461736B202332 path="/data/system/users/0/wallpaper" dev="mmcblk0p28" ino=586422 scontext=u:r:release_app:s0 tcontext=u:object_r:wallpaper_file:s0 tclass=file
platform_app access is from the SystemUI app whereas the other
denials are from the Launcher that is used on the particular
device. For instance, Launcher2 triggers the shared_app denial
whereas release_app (used by Launcher3) triggers the other denial.
Because of this, add the rule to all appdomains.
The static wallpaper is still set without this change. Just add
the rule to avoid the noise in the logs.
Change-Id: Ida84d1695d52379d67b87318403f629fd07109a4
Signed-off-by: rpcraig <rpcraig@tycho.ncsc.mil>
Re-purpose the existing bluetooth_socket type, originally
for /dev/socket/bluetooth used by bluetoothd in the old
bluetooth stack, for sockets created by bluedroid under
/data/misc/bluedroid, and allow mediaserver to connect
to such sockets. This is required for playing audio
on paired BT devices.
Based on b/12417855.
Change-Id: I24ecdf407d066e7c4939ed2a0edb97222a1879f6
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
This is required for the grouper sepolicy, where we must allow
bluetooth domain to write to the base sysfs type due to a kernel bug.
Change-Id: I14b0530387edce1097387223f0def9b59e4292e0
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Otherwise all domains can create/write files that are executable
by all other domains. If I understand correctly, this should
only be necessary for app domains executing content from legacy
unlabeled userdata partitions on existing devices and zygote
and system_server mappings of dalvikcache files, so only allow
it for those domains.
If required for others, add it to the individual
domain .te file, not for all domains.
Change-Id: I6f5715eb1ecf2911e70772b9ab4e531feea18819
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
ping in Android no longer requires any additional privileges beyond
the caller. Drop the ping domain and executable file type entirely.
Also add net_domain() to shell domain so that it can create and
use network sockets.
Change-Id: If51734abe572aecf8f510f1a55782159222e5a67
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
tmpfs_domain() macro defines a per-domain type and
allows access for tmpfs-backed files, including ashmem
regions. execute-related permissions crept into it,
thereby allowing write + execute to ashmem regions for
most domains. Move the execute permission out of tmpfs_domain()
to app_domain() and specific domains as required.
Drop execmod for now we are not seeing it.
Similarly, execute permission for /dev/ashmem crept into
binder_use() as it was common to many binder using domains.
Move it out of binder_use() to app_domain() and specific domains
as required.
Change-Id: I66f1dcd02932123eea5d0d8aaaa14d1b32f715bb
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
execmem permission controls the ability to make an anonymous
mapping executable or to make a private file mapping writable
and executable. Remove this permission from domain (i.e.
all domains) by default, and add it explicitly to app domains.
It is already allowed in other specific .te files as required.
There may be additional cases in device-specific policy where
it is required for proprietary binaries.
Change-Id: I902ac6f8cf2e93d46b3a976bc4dabefa3905fce6
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
system_server and app domains need to map dalvik-cache files with PROT_EXEC.
type=1400 msg=audit(13574814.073:132): avc: denied { execute } for pid=589 comm="system_server" path="/data/dalvik-cache/system@priv-app@SettingsProvider.apk@classes.dex" dev="mmcblk0p30" ino=684132 scontext=u:r:system_server:s0 tcontext=u:object_r:dalvikcache_data_file:s0 tclass=file
Apps need to map cached dex files with PROT_EXEC. We already allow this
for untrusted_app to support packaging of shared objects as assets
but not for the platform app domains.
type=1400 audit(1387810571.697:14): avc: denied { execute } for pid=7822 comm="android.youtube" path="/data/data/com.google.android.youtube/cache/ads1747714305.dex" dev="mmcblk0p30" ino=603259 scontext=u:r:platform_app:s0 tcontext=u:object_r:platform_app_data_file:s0 tclass=file
Change-Id: I309907d591ea6044e3e6aeb57bde7508e426c033
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
The closure of /dev/socket/zygote occurs in the zygote child
process, after Zygote has dropped privileges and changed
SELinux domains. In Google's internal tree, socket closures
are following a different path, which is causing getopt/getattr
to be used on the file descriptor. This is generating a large
number of denials.
Allow the operations for now. getopt/getattr are fairly harmless.
Long term, we shouldn't be performing these operations on the
zygote socket.
Addresses the following denials:
18.352783 type=1400 audit(1386374111.043:7): avc: denied { getattr } for pid=682 comm="ndroid.systemui" path="socket:[9287]" dev="sockfs" ino=9287 scontext=u:r:platform_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
18.353088 type=1400 audit(1386374111.043:8): avc: denied { getopt } for pid=682 comm="ndroid.systemui" path="/dev/socket/zygote" scontext=u:r:platform_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
18.833251 type=1400 audit(1386374111.524:9): avc: denied { getattr } for pid=761 comm="d.process.acore" path="socket:[9287]" dev="sockfs" ino=9287 scontext=u:r:shared_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
18.833557 type=1400 audit(1386374111.524:10): avc: denied { getopt } for pid=761 comm="d.process.acore" path="/dev/socket/zygote" scontext=u:r:shared_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
19.042419 type=1400 audit(1386374111.734:11): avc: denied { getattr } for pid=806 comm="d.process.media" path="socket:[9287]" dev="sockfs" ino=9287 scontext=u:r:media_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
19.042724 type=1400 audit(1386374111.734:12): avc: denied { getopt } for pid=806 comm="d.process.media" path="/dev/socket/zygote" scontext=u:r:media_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
19.182830 type=1400 audit(1386374111.874:14): avc: denied { getattr } for pid=825 comm="putmethod.latin" path="socket:[9287]" dev="sockfs" ino=9287 scontext=u:r:untrusted_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
19.183105 type=1400 audit(1386374111.874:15): avc: denied { getopt } for pid=825 comm="putmethod.latin" path="/dev/socket/zygote" scontext=u:r:untrusted_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
19.235473 type=1400 audit(1386374111.924:16): avc: denied { getattr } for pid=840 comm="ndroid.settings" path="socket:[9287]" dev="sockfs" ino=9287 scontext=u:r:system_app:s0 tcontext=u:r:zygote:s0 tclass=unix_stream_socket
Bug: 12061011
Change-Id: Ie1ec7636185aba7954656802e5eed735f49830c9
Add the necessary rules to support dumpstate.
Start off initially in permissive until it has more testing.
Dumpstate is triggered by running "adb bugreport"
Change-Id: Ic17a60cca1f6f40daa4f2c51e9ad6009ef36cfbd
Various third party apps come with their own binaries that they write out to
their sandbox directories and then execute, e.g.:
audit(1386527439.462:190): avc: denied { execute_no_trans } for pid=1550 comm="Thread-79" path="/data/data/com.cisco.anyconnect.vpn.android.avf/app_bin/busybox" dev="mmcblk0p23" ino=602891 scontext=u:r:untrusted_app:s0:c39,c256 tcontext=u:object_r:app_data_file:s0:c39,c256 tclass=file
While this is not ideal from a security POV, it seems necessary to support for
compatibility with Android today.
Split out the execute-related permissions to a separate allow rule as it
only makes sense for regular files (class file) not other kinds of files
(e.g. fifos, sockets, symlinks), and use the rx_file_perms macro.
Move the rule to untrusted_app only so that we do not permit system apps
to execute files written by untrusted apps.
Change-Id: Ic9bfe80e9b14f2c0be14295c70f23f09691ae66c
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Confine run-as (but leave permissive for now) and add
other allow rules required for the use of run-as and ndk-gdb
functionality.
Change-Id: Ifae38233c091cd34013e98830d72aac4c4adcae0
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Allow apps to communicate with each other via pipes.
In particular, this fixes a bug where printing from Chrome wasn't
working.
STEPS TO REPRODUCE:
1. Launch Chrome
2. From menu tap print and observe
OR
1. Launch Drive, Select any file (*.txt, *.doc. *.pdf.........)
2. Select print
Addresses the following denials:
<5>[ 122.352797] type=1400 audit(1386363998.374:18): avc: denied { write } for pid=3786 comm=4173796E635461736B202332 path="pipe:[19164]" dev="pipefs" ino=19164 scontext=u:r:untrusted_app:s0 tcontext=u:r:release_app:s0 tclass=fifo_file
<5>[ 123.248363] type=1400 audit(1386363999.264:19): avc: denied { getattr } for pid=2677 comm=".android.chrome" path="pipe:[19164]" dev="pipefs" ino=19164 scontext=u:r:untrusted_app:s0 tcontext=u:r:release_app:s0 tclass=fifo_file
<5>[ 123.248620] type=1400 audit(1386363999.264:20): avc: denied { write } for pid=3308 comm="ChildProcessMai" path="pipe:[19164]" dev="pipefs" ino=19164 scontext=u:r:isolated_app:s0 tcontext=u:r:release_app:s0 tclass=fifo_file
Bug: 12032455
Change-Id: Ic1cb5c1d42596f5a8fc3fe82fcbfe47aa43a7d6c
Remove unconfined_domain() from the bluetooth app domain,
restore the rules from our policy, and move the neverallow
rule for bluetooth capabilities to bluetooth.te.
Make the bluetooth domain permissive again until it has
received sufficient testing.
Change-Id: I3b3072d76e053eefd3d0e883a4fdb7c333bbfc09
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Shell domain can transition to other domains for runas, ping, etc.
Change-Id: If9aabb4f51346dc00a89d03efea25499505f278d
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
This change synchronizes the AOSP set of neverallow rules for
app domains with our own. However, as we exclude unconfineddomain
from each neverallow rule, it causes no breakage in the AOSP policy.
As app domains are confined, you will need to either adjust the
app domain or the neverallow rule according to your preference.
But our policy builds with all of these applied with all app domains
confined.
Change-Id: I00163d46a6ca3a87e3d742d90866300f889a0b11
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
This is a follow-up CL to the extraction of "system_app" domain
from the "system" domain which left the "system" domain encompassing
just the system_server.
Since this change cannot be made atomically across different
repositories, it temporarily adds a typealias "server" pointing to
"system_server". Once all other repositories have been switched to
"system_server", this alias will be removed.
Change-Id: I90a6850603dcf60049963462c5572d36de62bc00
The comment says that apps can read downloaded files, but the
file_type_auto_trans() macro expands to permit create/write access.
Also we don't need a type transition when staying in the same type
as the parent directory so we only truly need allow rules here.
Hence, we remove file_type_auto_trans() altogether, and add an allow
rule for search access to the directory. If create/write access is
truly required, then we can just change the allow rules to use
rw_dir_perms and create_file_perms.
Change-Id: Icd71c9678419442cfd8088317317efd4332f9b4a
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
r_dir_file(appdomain, isolated_app) was in both app.te and isolated_app.te;
delete it from isolated_app.te.
binder_call(appdomain, isolated_app) is a subset of binder_call(appdomain, appdomain); delete it.
Change-Id: I3fd90ad9c8862a0e4dad957425cbfbc9fa97c63f
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
This is my first attempt at creating an enforcing SELinux domain for
apps, untrusted_apps, and isolated_apps. Much of these rules are based on the
contents of app.te as of commit 11153ef349
with extensive modifications, some of which are included below.
* Allow communication with netd/dnsproxyd, to allow netd to handle
dns requests
* Allow binder communications with the DNS server
* Allow binder communications with surfaceflinger
* Allow an app to bind to tcp/udp ports
* Allow all domains to read files from the root partition, assuming
the DAC allows access.
In addition, I added a bunch of "neverallow" rules, to assert that
certain capabilities are never added.
This change has a high probability of breaking someone, somewhere.
If it does, then I'm happy to fix the breakage, rollback this change,
or put untrusted_app into permissive mode.
Change-Id: I83f220135d20ab4f70fbd7be9401b5b1def1fe35
app.te covers a lot of different apps types (platform_app, media_app,
shared_app, release_app, isolated_app, and untrusted_app), all
of which are going to have slightly different security policies.
Separate the different domains from app.te. Over time, these
files are likely to grow substantially, and mixing different domain types
is a recipe for confusion and mistakes.
No functional change.
Change-Id: Ida4e77fadb510f5993eb2d32f2f7649227edff4f
For unlabeled files, revert to DAC rules. This is for backwards
compatibility, as files created before SELinux was in place may
not be properly labeled.
Over time, the number of unlabeled files will decrease, and we can
(hopefully) remove this rule in the future.
To prevent inadvertantly introducing the "relabelto" permission, add
a neverallow domain, and add apps which have a legitimate need to
relabel to this domain.
Bug: 9777552
Change-Id: I71b0ff8abd4925432062007c45b5be85f6f70a88
This prevents denials from being generated by the base policy.
Over time, these rules will be incrementally tightened to improve
security.
Change-Id: I4be1c987a5d69ac784a56d42fc2c9063c402de11
/data/app-private is used when making an
app purchase or forward locking. Provide a
new label for the directory as well as the
tmp files that appear under it.
Change-Id: I910cd1aa63538253e10a8d80268212ad9fc9fca5
Signed-off-by: rpcraig <rpcraig@tycho.ncsc.mil>
Read access to /dev/log/* is no longer restricted.
Filtering on reads is performed per-uid by the kernel logger driver.
Change-Id: Ia986cbe66b84f3898e858c60f12c7f3d63ac47cf
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
This feels like a hidden bug- it shouldn't be trying to
stat everything under /cache anyways- but allowing for now.
Change-Id: Ib5ddfbb408c9f0b6c6218c78a678fcdb09360ccd
Just allow them unconditionally for compatibility.
Change-Id: I85b56532c6389bdfa25731042b98d8f254bd80ee
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Two new types are introduced:
sdcard_internal
sdcard_external
The existing type of sdcard, is dropped and a new attribute
sdcard_type is introduced.
The boolean app_sdcard_rw has also been changed to allow for
controlling untrusted_app domain to use the internal and external
sdcards.
Change-Id: Ic7252a8e1703a43cb496413809d01cc6cacba8f5
The binder_transfer_binder hook was changed in the kernel, obsoleting
the receive permission and changing the target of the transfer permission.
Update the binder-related policy to match the revised permission checking.
Change-Id: I1ed0dadfde2efa93296e967eb44ca1314cf28586
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Policy covers:
* backup_data_file type for labeling all
files/dirs under /data dealing with
backup mechanism.
* cache_backup_file type for labeling all
files/dirs under /cache dealing with
backup mechanism. This also covers the
the use of LocalTransport for local archive
and restore testing.
* the use of 'adb shell bmgr' to initiate
backup mechanism from shell.
* the use of 'adb backup/restore' to archive
and restore the device's data.
Change-Id: I700a92d8addb9bb91474bc07ca4bb71eb4fc840e
Signed-off-by: rpcraig <rpcraig@tycho.ncsc.mil>
Generic init.rc allows any process to use
socket tagging. Adjust app policy to ensure
that any app can read from the misc device.
Change-Id: I4076f0fbc1795f57a4227492f6bfc39a4398ffa5
Signed-off-by: rpcraig <rpcraig@tycho.ncsc.mil>
