Bug: 65643247
Test: SurfaceFlinger_test passes (except known failures) without selinux
denials
Change-Id: I6ce185f92e5ad64a172da7d7e12167d8da2ebed0

* changes:
  nonplat_property_contexts: use fc_sort vs sort
  plat_property_contexts: use fc_sort vs sort

Test: adb bugreport
Bug: 63629306
Change-Id: I7a366b199ddd0ec303dc25ca8c35764c5d7e3af8

Bug: 65570851
Test: boot sailfish
Change-Id: I008bf5386595c614236de44131afcda7d3fd6d98
Merged-In: I008bf5386595c614236de44131afcda7d3fd6d98
(cherry picked from commit 82ca9c2e)

(This reverts internal commit: 82ca9c2e)
Test: None.

Merged-in: I97ffdd48b64ef5c35267387079204512a093a356
Change-Id: I97ffdd48b64ef5c35267387079204512a093a356

Used to display kernel version in settings app.

avc: denied { read } for name="version" dev="proc"
scontext=u:r:system_app:s0 tcontext=u:object_r:proc_version:s0
tclass=file permissive=0

Bug: 66985744
Test: kernel version now displayed in settings app.
Change-Id: I53f92f63362b900347fd393a40d70ccf5d220d30

sort respects locale settings, so the value of LC_ALL can affect
how sort orders things. This can cause labeling issues.

More information on locale and sort can be found via:
  * locale(1) - man 1 locale
  * sort(1) - man 1 sort
  * https://debbugs.gnu.org/cgi/bugreport.cgi?bug=28539



Rather than deal with this locale setting use fc_sort on
property contexts. This also has the side-effect of
stripping comments, and thus sed can be dropped.

Test: This was tested by:
  * comparing outputs to previous runs
  * compile tested *only*.

Change-Id: I1e1eb4dff76f717b5f82f697e677a108abb69892
Signed-off-by: William Roberts <william.c.roberts@intel.com>

sort respects locale settings, so the value of LC_ALL can affect
how sort orders things. Issues have surfaced when CTS build
servers locale differs from image build server locale. And thus
the prologue of property_contexts differs with what CTS was
expecting.

More information on locale and sort can be found via:
  * locale(1) - man 1 locale
  * sort(1) - man 1 sort
  * https://debbugs.gnu.org/cgi/bugreport.cgi?bug=28539



Rather than deal with this locale setting use fc_sort on
property contexts. This also has the side-effect of
stripping comments, and thus sed can be dropped.

Test: This was tested by:
  * comparing outputs to previous runs
  * booting the x86-64 emulator

Change-Id: I144ef549cc11d9c61849ffc0e1b1b000f1b8d1a8
Signed-off-by: William Roberts <william.c.roberts@intel.com>

Bug: 37916906
Test: Builds 'n' boots.
Change-Id: Ia1d86264446ebecc1ca79f32f11354921bc77668
Merged-In: I208ec6a864127a059fb389417a9c6b259d7474cb

Fixes: 65263013
Test: build
Merged-In: I0ec412481c5990927fcbee7c4303bee2da876210
Change-Id: I0a5b9a80e988fcd16a29807ed83b2c65bba9000f

Run-as is running a command under an app's uid and in its data
directory. That data directory may be accessed through a symlink
from /data/user. So give runas rights to read such a symlink.

Bug: 66292688
Test: manual
Test: CTS JVMTI tests
Change-Id: I0e0a40d11bc00d3ec1eee561b6223732a0d2eeb6

The following commits were cherry-picked from internal master to AOSP,
but to avoid merge-conflicts we'll do a large diff instead of individual
cherry-picks:
521742e9
9aefc916
3686efca
de51e7de
fff3fe2f

Bug: 37916906
Test: angler builds and boots.
Merged-In: Ie010cc12ae866dbb97c387471f433158d3b699f3
Change-Id: I5126ebe88b9c76a74690ecf95851d389cfc22d1f

Bug: 65643247
Test: device boots without denials from bootanim to sysfs and cgroup.
Change-Id: Icf8c45906cb83e1b0a60737d67ae584b9d1b34aa

Bug: 65643247
Test: device boots without denials from rild to proc.
Change-Id: I142a228347ef07266cb612e99c90fb5ec187988a

Bug: 65643247
Test: device boots without denials from bootstat to proc.
Change-Id: Ie31a0488239dbb1614fbcce07540d23afa805b0e

Bug: 65643247
Test: device boots without denials from bootanim to proc.
Change-Id: I0454a2bd4489d7816d82a299f5bc199d6a299ec0

Bug: 62945293
Test: instrumentation, VTS
Change-Id: I7e896b64bf0ee907af21d08f6b78561fadc7f0e3

Change-Id: I88e2887b0691ce3c5018578556abf7c420fe5a1b

Bug: 63600413
Test: VTS, instrumentation, audit2allow
Test: after cherry-pick - it builds
Change-Id: I57c0150a52c13f1ce21f9ae2147e3814aad0fb7e
(cherry picked from commit 567b947d)

Add series of neverallow rules to restrict components from reading or
writing bootloader_boot_reason_prop, system_boot_reason_prop and
last_boot_reason_prop to trusted set of domains.

The policy is that bootloader_boot_reason_prop (ro.boot.bootreason)
has a compliance issue due to the sheer momentum of near unparseable
content as filed by the wide variety (8000 different devices at last
count) bootloaders and is only to be accessible to a series of
responsible system components.  It can be inaccurate as it provides
no means to evaluate a shutdown, likely reporting "cold" (from
initial power up) or the more generic "reboot".

The last_boot_reason_prop (persist.sys.boot.reason) contains
inaccurate information as it is only valid after a controlled reboot
or shutdown.  The value can linger around after less controlled
scenarios.  Since the information could be false, we do not want to
support it as an open API, so we again block access to only
responsible components.

The system_boot_reason_prop (sys.boot.reason) is a canonical boot
reason that takes into account parsing bootloader_boot_reason_prop,
boot_loader_boot_reason_prop and other system and HAL generated hints
to determine a parseable and most accurate reason for the last time
the system was rebooted.

For now the policy for system_boot_reason_prop is to audit users of
the API, and on a need to know basis via device additions to the
selinux rules.  If vendors need their components to access the boot
reason, they need to comply first with CTS tests and spirit with
regards to controlled reboot messaging and in turn read the
system_boot_reason_prop for the canonical information.  It will
contain validated content derived from bootloader_boot_reason_prop
in the scenarios that count.

The controlled reboot APIs include:
- android_reboot(ANDROID_RB_<TYPE>, int flag, const char* reason)
- PowerManagerService.lowLevelShutdown(String reason);
- PowerManagerService.lowLevelReboot(String reason);
- ShutdownThread.shutdown(context, String reason, boolean confirm);
- ShutdownThread.reboot(context, String reason, boolean confirm);
- PowerManager.shutdown(boolean confirm, String reason, boolean wait);
- PowerManager.reboot(String reason);

Any others (including the direct linux reboot syscall) create
problems for generating an accurate canonical boot reason.

Test: compile
Bug: 63736262
Bug: 65686279
Change-Id: I2e5e55bbea1c383c06472eb2989237cfeb852030

* changes:
  Allow sensor hal to use wakelock
  Allow sensor to use gralloc handle and access ion device

labeled /proc/kmsg as proc_kmsg, changed logd's access from proc to
proc_kmsg, and added a compat mapping.

Bug: 65643247
Test: device boots without selinux denials to the newly introduced proc_kmsg
Test: logd-unit-tests passes

Merged-In: I92c9f5694289eb6a94c4d90f14e2de4d46b5228e
Change-Id: I92c9f5694289eb6a94c4d90f14e2de4d46b5228e
(partial CP of commit 528da6fe)

Added permission related to use of wake lock. Wakelock in sensor
HAL is used to gurantee delivery of wake up sensor events before
system go back to sleep.

Bug: 63995095
Test: QCOM and nanohub sensor hal are able to acquire wakelock
      successfuly.

Change-Id: Id4ac3552e18a1cad252017e3dc9ab3d4be8d4ab9

Allow sensor hal to sue gralloc handle and access ion device
so that sensor direct report feature can function correctly when
HardwareBuffer shared memory is used.

Test: SensorDirectReportTest passes without setenforce 0

Change-Id: I2068f6f4a8ac15da40126892e1326e0b90a6576f
Merged-In: I2068f6f4a8ac15da40126892e1326e0b90a6576f

Test: let fs_mgr format a damaged /data partition
Bug: 35219933
Change-Id: I379567772c73e52f532a24acf640c21f2bab5c5b
(cherry picked from commit 6d1ecdcb)

Test: let fs_mgr format a damaged /data partition
Bug: 35219933
Change-Id: If92352ea7a70780e9d81ab10963d63e16b793792
(cherry picked from commit 5f573ab2)