Define a specific block device type for system so that we can
prevent raw writes to the system partition by anything other than
recovery.
Define a specific block device type for recovery so that we
can prevent raw writes to the recovery partition by anything
other than install_recovery or recovery.
These types must be assigned to specific block device nodes
via device-specific policy. This change merely defines the types,
adds allow rules so that nothing will break when the types are assigned,
and adds neverallow rules to prevent adding further allow rules
on these types.
This change does not remove access to the generic block_device type
from any domain so nothing should break even on devices without these
type assignments.
Change-Id: Ie9c1f6d632f6e9e8cbba106f07f6b1979d2a3c4a
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Allow error reporting via the pty supplied by init.
Allow vold to invoke fsck for checking volumes.
Addresses denials such as:
avc: denied { ioctl } for pid=133 comm="e2fsck" path="/dev/pts/0" dev="devpts" ino=3 scontext=u:r:fsck:s0 tcontext=u:object_r:devpts:s0 tclass=chr_file
avc: denied { execute } for pid=201 comm="vold" name="e2fsck" dev="mmcblk0p25" ino=98 scontext=u:r:vold:s0 tcontext=u:object_r:fsck_exec:s0 tclass=file
These denials show up if you have encrypted userdata.
Change-Id: Idc8e6f83a0751f17cde0ee5e4b1fbd6efe164e4c
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Resolves denials such as:
avc: denied { write } for pid=1546 comm="Binder_1" name="/" dev="dm-0" ino=2 scontext=u:r:platform_app:s0:c512,c768 tcontext=u:object_r:asec_apk_file:s0 tclass=dir
This is required to install a forward-locked app.
Change-Id: I2b37a56d087bff7baf82c738896d9563f0ab4fc4
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
The current policy would allow any application that were to
"magically" get a sensitive UID into the coresponding
sensitive domain. Rather then only using UID as an input
selector, require seinfo=platform.
Change-Id: I8a7490ed55bdcd3e4a116aece2c3522b384024ec
Remove the ability of init to execute programs from / or /system
without changing domains. This forces all helper programs and
services invoked by init to be assigned their own domain.
Introduce separate domains for running the helper programs
executed from the fs_mgr library by init. This requires a domain
for e2fsck (named fsck for generality) and a domain for running
mkswap (named toolbox since mkswap is just a symlink to the toolbox
binary and the domain transition occurs on executing the binary, not
based on the symlink in any way).
e2fsck is invoked on any partitions marked with the check mount
option in the fstab file, typically userdata and cache but never
system. We allow it to read/write the userdata_block_device and
cache_block_device types but also allow it to read/write the default
block_device type until we can get the more specific types assigned
in all of the device-specific policies.
mkswap is invoked on any swap partition defined in the fstab file.
We introduce a new swap_block_device type for this purpose, to be
assigned to any such block devices in the device-specific policies,
and only allow it to read/write such block devices. As there seem to be
no devices in AOSP with swap partitions in their fstab files, this does
not appear to risk any breakage for existing devices.
With the introduction of these domains, we can de-privilege init to
only having read access to block devices for mounting filesystems; it
no longer needs direct write access to such devices AFAICT.
To avoid breaking execution of toolbox by system services, apps, or the shell,
we allow all domains other than kernel and init the ability to
run toolbox in their own domain. This is broader than strictly required;
we could alternatively only add it to those domains that already had
x_file_perms to system_file but this would require a coordinated change
with device-specific policy.
Change-Id: Ib05de2d2bc2781dad48b70ba385577cb855708e4
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
In order to support the new goldfish service domains in
a change with the same Change-Id for the build project, we need
the following changes in external/sepolicy:
- /system/bin/logcat needs its own type so that it can be used as an
entrypoint for the goldfish-logcat service. A neverallow rule prevents
us from allowing entrypoint to any type not in exec_type.
- The config. and dalvik. property namespaces need to be labeled
with something other than default_prop so that the qemu-props
service can set them. A neverallow rule prevents us from allowing
qemu-props to set default_prop.
We allow rx_file_perms to logcat_exec for any domain that
was previously allowed read_logd() as many programs will read
the logs by running logcat. We do not do this for all domains
as it would violate a neverallow rule on the kernel domain executing
any file without transitioning to another domain, and as we ultimately
want to apply the same restriction to the init domain (and possibly others).
Change-Id: Idce1fb5ed9680af84788ae69a5ace684c6663974
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Android's native bridge functionality allows an Android native
app written on one CPU architecture to run on a different architecture.
For example, Android ARM apps may run on an x86 CPU.
To support this, the native bridge functionality needs to replace
/proc/cpuinfo with the version from /system/lib/<ISA>/cpuinfo
using a bind mount. See commit ab0da5a9a6860046619629b8e6b83692d35dff86
in system/core.
This change:
1) Creates a new label proc_cpuinfo, and assigns /proc/cpuinfo
that label.
2) Grants read-only access to all SELinux domains, to avoid
breaking pre-existing apps.
3) Grants zygote mounton capabilities for that file, so zygote
can replace the file as necessary.
Addresses the following denial:
avc: denied { mounton } for path="/proc/cpuinfo" dev="proc" ino=4026532012 scontext=u:r:zygote:s0 tcontext=u:object_r:proc:s0 tclass=file
Bug: 17671501
Change-Id: Ib70624fba2baeccafbc0a41369833f76b976ee20
Apps should be able to read the contents of mounted OBBs.
Steps to reproduce:
1) Install com.namcobandaigames.soulcaliburgp (SoulCalibur)
2) Attempt to run the app.
Expected:
App runs successfully.
Actual:
App crashes. See denials below.
This can also be reproduced by running the newly introduced CTS
test in I2018b63b0236ce6b5aee4094e40473315b1948c3
Addresses the following denials:
avc: denied { read } for pid=4133 comm="roidJUnitRunner" name="test1.txt" dev="loop0" ino=23 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:vfat:s0 tclass=file
avc: denied { open } for pid=4133 comm="roidJUnitRunner" name="test1.txt" dev="loop0" ino=23 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:vfat:s0 tclass=file
avc: denied { getattr } for pid=4133 comm="roidJUnitRunner" path="/mnt/obb/f73da56689d166b5389d49ad31ecbadb/test1.txt" dev="loop0" ino=23 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:vfat:s0 tclass=file
avc: denied { search } for name="/" dev="loop0" ino=1 scontext=u:r:untrusted_app:s0 tcontext=u:object_r:vfat:s0 tclass=dir permissive=0
(cherrypick of commit 62083414a4)
Bug: 17633509
Change-Id: I49b722b24c1c7d9ab084ebee7c1e349d8d660ffa
Add levelFrom=user to the entries for apps other than those
that run in the predefined platform UIDs (e.g. system, nfc, radio, ...).
This causes libselinux to assign a per-user category set computed from
the user ID portion of the Linux UID to each app process and its
/data/data/<pkgdir> or /data/user/N/<pkgdir> directory. These
per-user category sets can be seen in the last field of ps -Z output for
apps and ls -Z /data/data or /data/user/N output for the package
directories.
With this applied, apps running on behalf of one user cannot read
or write files created by apps running on behalf of another user,
even if the file is world-readable or -writable. Similar isolation is
enforced over process interactions (including /proc/pid file access),
local socket communications, and System V IPC, as expressed in the
set of constraints defined in the mls configuration. At present,
Binder IPC is not restricted by the mls configuration; if desired,
there is a constraint in the configuration that can be uncommented
to also apply isolation on direct binder IPC, although communication
will still be possible indirectly via the system_server.
Bug: 13507660
Change-Id: I3972f846ff5e7363799ba521f1258d662b18d64e
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
During factory provisioning, some manufacturers may need to pull files
from /factory (label efs_file and bluetooth_efs_file) to collect
device specific identifiers such as the mac address, using commands
similar to the following:
adb shell cat /factory/ssn
adb shell cat /factory/bt/bd_addr.conf
adb shell cat /factory/wifi/mac.txt
adb shell cat /factory/60isn
read-only access to these files is currently disallowed by a
neverallow rule. Relax the rules to allow read-only access to the
shell user if desired.
No new SELinux rules are added or deleted by this change. This is
only a relaxation in what's allowed for vendor specific policy.
Bug: 17600278
Change-Id: I13f33f996c077918dce70a5cff31a87eac436678
Netlink uevent sockets are used by the kernel to inform userspace
when certain events occur, for example, when new hardware is added
or removed. This allows userspace to take some action based on those
messages.
Relax the neverallow rule for NETLINK_KOBJECT_UEVENT sockets.
Certain device specific app domains, such as system_app, may have a
need to receive messages from this socket type.
Continue to neverallow NETLINK_KOBJECT_UEVENT sockets for untrusted_app.
These sockets have been the source of rooting attacks in Android
in the past, and it doesn't make sense to expose this to untrusted_apps.
No new SELinux rules are introduced by this change. This is an
adjustment of compile time assertions only.
Bug: 17525863
Change-Id: I3e538dc8096dc23b9678bcd20e3c1e742c21c967
Introduce separate types for the userdata and cache block
devices so that we can assign them and allow access to them
in device-specific policy without allowing access to any other
block device (e.g. system). These types will only be used if
assigned to device node paths in the device-specific file_contexts
configuration. Otherwise, this change will have no impact - the
userdata and cache block devices will continue to default to block_device
type.
To avoid breakage when these new types are assigned to the userdata
block device, allow access by vold and uncrypt, but auditallow
these accesses to confirm that these are required.
Change-Id: I99d24f06506f51ebf1d186d9c393b3cad60e98d7
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
The kernel driver has been deprecated by the new userspace
driver. Don't continue to allow access to the old driver.
Maintain the labeling on /dev/log/* for now, just in case.
Bug: 13505761
Change-Id: Ibf8ef3af6274ede4262aada9222eaf63f63307b4
Enable labeling apps differently depending on whether they
are running for the primary user / owner or for a secondary user.
Change-Id: I37aa5b183a7a617cce68ccf14510c31dfee4e04d
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
dumpstate and lmkd need to act on apps running at any level.
Various file types need to be writable by apps running at any
level.
Change-Id: Idf574d96ba961cc110a48d0a00d30807df6777ba
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
On 64 bit systems, it's necessary to read the /system/bin executables
elf header to determine if it's a 32 bit or 64 bit executable to
contact the correct debuggerd service.
Bug: 17487122
(cherry picked from commit 04f3d79077)
Change-Id: Ib7835ffac1811a5aef54a250689287c1666720ef
Permits the system server to change keystore passwords for users other
than primary.
(cherrypicked from commit de08be8aa0)
Bug: 16233206
Change-Id: I7941707ca66ac25bd122fd22e5e0f639e7af697e
When using MLS (i.e. enabling levelFrom= in seapp_contexts),
certain domains and types must be exempted from the normal
constraints defined in the mls file. Beyond the current
set, adbd, logd, mdnsd, netd, and servicemanager need to
be able to read/write to any level in order to communicate
with apps running with any level, and the logdr and logdw
sockets need to be writable by apps running with any level.
This change has no impact unless levelFrom= is specified in
seapp_contexts, so by itself it is a no-op.
Change-Id: I36ed382b04a60a472e245a77055db294d3e708c3
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
If the sdcard daemon is restarted (crash or otherwise), one of the first
things it attempts to do is umount the previously mounted /mnt/shell/emulated
fuse filesystem, which is denied by SELinux with the following denial:
sdcard : type=1400 audit(0.0:6997): avc: denied { unmount } for scontext=u:r:sdcardd:s0 tcontext=u:object_r:fuse:s0 tclass=filesystem permissive=0
Allow the operation.
Steps to reproduce:
1) adb shell into the device and su to root
2) run "kill -9 [PID OF SDCARD]
Expected:
sdcard daemon successfully restarts without error message.
Actual:
SELinux denial above, plus attempts to mount a new filesystem
on top of the existing filesystem.
(cherrypicked from commit abfd427a32)
Bug: 17383009
Change-Id: I386bfc98e2b5b32b1d11408f7cfbd6e3c1af68f4
The kernel, when it creates a loop block device, starts a new
kernel thread "loop0" (drivers/block/loop.c). This kernel thread,
which performs writes on behalf of other processes, needs read/write
privileges to the sdcard. Allow it.
Steps to reproduce:
0) Get device with external, removable sdcard
1) Run: "adb install -s foo.apk"
Expected:
APK installs successfully.
Actual:
APK fails to install. Error message:
Vold E Failed to write superblock (I/O error)
loop0 W type=1400 audit(0.0:3123): avc: denied { read } for path="/mnt/secure/asec/smdl1645334795.tmp.asec" dev="mmcblk1p1" ino=528 scontext=u:r:kernel:s0 tcontext=u:object_r:vfat:s0 tclass=file permissive=0
PackageHelper E Failed to create secure container smdl1645334795.tmp
DefContainer E Failed to create container smdl1645334795.tmp
Bug: 17158723
Change-Id: I4aa86e372cc55348f6b8becfa17bd4da583925d4
Remove the CTS specific rule which allows appdomain processes
to view /proc entries for the rest of the system. With this change,
an SELinux domain will only be able to view it's own /proc
entries, e.g. untrusted_app can only view /proc entries for other
untrusted_app, system_app can only view /proc entries for other
system_apps, etc.
/proc contains sensitive information, and we want to avoid
leaking this information between app security domains.
Bug: 17254920
Change-Id: I59da37dde00107a5ab123df3b79a84afa855339f
Add a neverallow rule (compile time assertion) that no SELinux domain
other than init can set default_prop. default_prop is assigned to a
property when no more specific label exists for that property.
This ensures that all properties are labeled properly, and that
no-one (other than init) gets access to unknown properties.
Change-Id: If279960f23737e263d4d1b5face7b5c49cda7ae7
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
Init never uses binder, so allowing binder related operations
for init never makes sense. Disallow all binder opertions for
init.
This change expands on commit a730e50bd9,
disallowing any init binder operation, not just call operations, which
may be accidentally added by blindly running audit2allow.
Change-Id: I12547a75cf68517d54784873846bdadcb60c5112
Addresses the following denial when debuggerd attempts to stat Webview mmap'd
shared relro files on process crash. Full read permissions may not be necessary:
W/debuggerd( 185): type=1400 audit(0.0:97): avc: denied { search } for name="shared_relro" dev="mmcblk0p28" ino=618955 scontext=u:r:debuggerd:s0 tcontext=u:object_r:shared_relro_file:s0 tclass=dir
Bug: 17101854
Change-Id: I11eea85668ba033c554e5aab99b70a454fb75164
Before actual installation, apks are put in a staging area where they are
scanned by a verifier before completing the install flow. This verifier runs as
a priv-app, which is in the untrusted_app domain. Allow untrusted_app
read-access to these files.
Bug: 16515815
Change-Id: Ifedc12a33b1f53b62f45013e7b253dbc79b02a4e
The list of processes comes from frameworks/native/cmds/dumpstate/utils.c.
dumpstate calls dump_backtrace_to_file() for each such process, which
asks debuggerd to dump the backtrace.
Resolves denials such as:
avc: denied { dump_backtrace } for scontext=u:r:dumpstate:s0 tcontext=u:r:surfaceflinger:s0 tclass=debuggerd
avc: denied { dump_backtrace } for scontext=u:r:dumpstate:s0 tcontext=u:r:drmserver:s0 tclass=debuggerd
avc: denied { dump_backtrace } for scontext=u:r:dumpstate:s0 tcontext=u:r:mediaserver:s0 tclass=debuggerd
avc: denied { dump_backtrace } for scontext=u:r:dumpstate:s0 tcontext=u:r:sdcardd:s0 tclass=debuggerd
Change-Id: Idbfb0fef0aac138073b7217b7dbad826a1193098
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
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>
