Based on experiments, we fount out 128 (128 x global readahead window
size = 16mb) is the optimal multiple to boost up read speeds for the
sequentially accessed files with POSIX_FADV_SEQUENTIAL.
Bug: 195311558
Test: check the /sys/fs/f2fs/<userdata partition>/seq_file_ra_mul value
Change-Id: I7563ad6e47b9ab76ae7fe36978d0e5970a7490e8
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Due to the work done for b/156305599 ("Ensure no process except vold can
create directories like /data/system_ce/0"), the SELinux policy now
enforces that vold is the only process that can write to directories
that contain per-user encrypted subdirectories. This is essential to
prevent bugs where directories that are supposed to be encrypted get
created too early so are not actually encrypted as intended.
However, this only works when SELinux is in enforcing mode. When
SELinux is in permissive mode, only DAC is enforced, and the file modes
allow other processes to write to many of these directories. That
allows system_server to break things once again.
Therefore, remove the write bit from the file modes so that write access
is always denied to processes that don't have CAP_DAC_OVERRIDE. This is
not as strong a restriction as the SELinux policy, which still applies
independently, but it does keep out system_server by itself.
Also remove the sticky bit from /data/misc_ce and /data/misc_de, since
there is no reason for it. (It probably was originally copied from
/data/misc, which might need it. But misc_{ce,de} don't need it.)
Bug: 285239971
Test: Booted Cuttlefish
Change-Id: I1213a4d18c5f851acf213d786400d79d73777ed0
Merging as a separate CL due to a log showing up
related to this on hwasan (is a prebuilt pulling
this in?)
Bug: 276813155
Test: boot cf
Change-Id: I19f7fc51c937d0eb1ee17781fc5d201a0972c4b0
Bionic requires random numbers to init the shadow call stack. Those
numbers are obtained via the syscall getrandom (non-blocking) and will
fallback to /dev/urandom if the former fails.
When loading pKVM modules, we are so early in the boot process that the
only source of entropy for the linux RNG are the architecture random
number generators... which might be available on some platforms. Without
any source of entropy, the only way of generating a random number is to
try to generate some, which is what the bionic fallback expects via
urandom.
As a consequence, add the urandom node to the initramfs.
Bug: 274876849
Change-Id: I164b08f026a238dad9f27a345bdef96717f2aa74
Set the user explicitly.
For boringssl self-test, changed to 'nobody' since
this test doesn't require permissions.
Bug: 276813155
Test: boot, check can 'nobody' can still write to kmesg.
Change-Id: I32f7134e83183bd054bffbb22d412d7a2dc0ad09
There are multiple use cases in Android for which background writes need
to be controlled via the cgroup mechanism. The cgroup mechanism can only
control background writes if both the blkio and memcg controllers are
mounted in the v2 cgroup hierarchy. Hence this patch that migrates the
blkio controller from the v1 to the v2 cgroup hierarchy.
The blkio controller has been marked as optional since not all Android
kernels enable this controller (CONFIG_BLK_CGROUP).
This patch increases the TOTAL_BOOT_TIME for devices with a 4.19 kernel
(redfin) from 18.9 s to 20 s. This patch does not affect the boot time
for devices with a 5.10 or 5.15 kernel.
This patch increases the time spent in CgroupMap::ActivateControllers()
by 25 microseconds in Cuttlefish on an x86-64 CPU.
CgroupMap::ActivateControllers() is called by Service::Start().
Bug: 213617178
Test: Cuttlefish and various phones
Change-Id: I3c07c1be84c3feb277b7d7003652d5d3b57c6541
Signed-off-by: Bart Van Assche <bvanassche@google.com>
Bug: 260366497
Bug: 264600011
Test: The correct label is assigned to dir after taking reboot
Test: Both system_server and dumpstate can access it
Change-Id: Icecbb59ddf936088aa3873bf1b143a08f035fefe
Added SPDX-license-identifier-Apache-2.0 to:
rootdir/Android.mk
Bug: 68860345
Bug: 151177513
Bug: 151953481
Test: m all
Change-Id: I19aaea76a932cc928bbc178c01a33fdc98b0cf16
Grant read/write access to anyone for /data/misc/wmtrace folder on
debuggable builds, it's further protected by the selinux policy.
This is to allow systemui process to write proto logs to the same folder
on device as WindowManager, both can contribute to the transitions like
PiP, Split-Screen and etc.
Bug: 251513116
Test: adb shell dumpsys activity service SystemUIService \
WMShell protolog [start | stop]
Change-Id: Ice57efa17c61d132b02c0a11a762c24d772bd90a
This file contains a description of dev nodes added to the CPIO archive
for the Android ramdisks. /dev/null is a security requirement for
bionic, /dev/console is needed so the kernel can set-up stdout stderr
and stdin before running /init.
Bug: 254835242
Change-Id: I111e2db53fabd63d070b8e9ab9c52faebf484ab3
A critical shutdown service is one that stays on right until the system
is rebooted. In order to be able to capture kernel messages right until
reboot this is required, otherwise after the console service is
terminated some messages can be lost.
Test: Reboot and verify messages show up on serial further down the reboot process
Change-Id: Iea58b5a76afe45b3346803021e3be81742b02ea0
There are multiple use cases in Android for which background writes need
to be controlled via the cgroup mechanism. The cgroup mechanism can only
control background writes if both the blkio and memcg controllers are
mounted in the v2 cgroup hierarchy. Hence this patch that migrates the
blkio controller from the v1 to the v2 cgroup hierarchy.
This patch increases the TOTAL_BOOT_TIME for devices with a 4.19 kernel
(redfin) from 18.9 s to 20 s. This patch does not affect the boot time
for devices with a 5.10 or 5.15 kernel.
This patch increases the time spent in CgroupMap::ActivateControllers()
by 25 microseconds in Cuttlefish on an x86-64 CPU.
CgroupMap::ActivateControllers() is called by Service::Start().
Bug: 213617178
Test: Cuttlefish and various phones
Change-Id: I490740e1c9ee4f7bb5bb7afba721a083f952c8f2
Signed-off-by: Bart Van Assche <bvanassche@google.com>
The folder is used for temporary files of virtualizationservice, with
a subfolder for each running VM. This wil continue to be the case but
each subfolder will be populated by a different instance of virtmgr,
running under the UID of the client (as opposed to system UID of
virtualizationservice).
To this end, change the permission mask of the root folder from 0770 to
0775. This gives non-system UIDs the permission to search the root
folder. This is necessary for the clients to be able to search their
own subfolder. It does not give them permission to read other
subfolders as those will be owned by different client UIDs.
Bug: 245727626
Test: atest -p packages/modules/Virtualization:avf-presubmit
Change-Id: Ie6e3be601ccb3b385f70bcf5b31bf8fff3aff8bc
Defaulting Android to limit memlock to 64KB. This will help preventing
pages from being swapped until the app is killed it's memory will stay
resident. CTS test is enforced only in U+ devies.
Bug: 201797650
Test: Added new test to verify we are memlock at or under 64KB
Change-Id: I5a9e9da12f6df5a056ee47d0593c13e9c779e054
If a service doesn't specify any capabilities in it's definition in the
.rc file, then it will inherit all the capabilities from the init.
Although whether a process can use capabilities is actually controlled
by selinux (so inheriting all the init capabilities is not actually a
security vulnerability), it's better for defense-in-depth and just
bookkeeping to explicitly specify that boringssl_self_test doesn't need
any capabilities
The list of capabilities was obtained via:
```
$ adb pull /sys/fs/selinux/policy /tmp/selinux.policy
$ sesearch --allow -s boringssl_self_test -c capability,capability2 /tmp/selinux.policy
```
Bug: 249796710
Test: device boots
Test: presubmit
Change-Id: I866222e2325e59d7e39d00db59df7b83efc657d9
We will continue to restrict access to /dev/kvm and /dev/vhost-vsock with SELinux.
Bug: 245727626
Test: atest -p packages/modules/Virtualization:avf-presubmit
Change-Id: Id4f3e19c18a51bc51e6363d6ffde31c1032cf967
If the framework is restarting (and cannot yet aquire
wakelocks to block suspend). Take a kernel wakelock
to allow the system to make sufficient progress before
autosuspend can be triggered.
The wakelock is later disable when the framework has
and invokeds enableAutosuspend() on the suspend service.
Bug: 255898234
Test: adb shell "echo mem > /sys/power/state && killall system_server"
Change-Id: Id8cff6564ef05d8c22a8264c51dd313263cb6a9d
Early processes can't rely on APEXes anyway. We don't need to run
linkerconfig.
This helps to reduce the storage usage (no
/system/bin/bootstrap/linkerconfig) and the boottime (not running
linkerconfig).
If we need more complicated linker config even for early processes, then
we could generate it at build-time and use it like recovery version.
Bug: 262330207
Bug: 260982509
Test: MicrodroidAppTest
Test: device boots
Change-Id: Iceca5ffdb1655fd94e90b0091f439bd22130185e
Current period of this is 3 sec and it is used when Perfetto profiling is running on Android. Without Perfetto profiling, it doesn't affect the system at all. However, 3 sec doesn't provide enough granularity to understand F2FS I/O behaviors. To make F2FS I/O profiling ftrace effective, set the ftrace period to 1 sec.
Test: check f2fs iostat_period_ms sysfs node value
Change-Id: I2d418795613dfbd1aea6c4f13c9a39af3deb1c4d
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Remove LLNDK libraries from system required libs as those libraries will
be appended to the configuration from the build.
Bug: 251782700
Test: Cuttlefish build and boot succeeded
Change-Id: I81d508a5e15a9dd1919935f07569271609738710
Create a mirror directory for misc_ce and misc_de storage by bind
mounting the respective directories. This is done for the defaul null
volume only, and other volumes are handled at a later staged.
When an SDK sandbox process is spawned and data isolation needs to
occur, the sdksandbox directories present in the misc directories will
be used to bind mount from, after tmpfs is mounted on the original.
Bug: 214241165
Test: atest SdkSandboxStorageHostTest
Change-Id: Icb1dc7d7fbd53a5c3853acf2f9d4d75b278d7295
Merged-In: Icb1dc7d7fbd53a5c3853acf2f9d4d75b278d7295
Revert submission 2291455-localhost_v6
Reason for revert: b/258627476
Reverted Changes:
Icd11fab47:[Test] Update test to comply with etc/hosts change...
Ibcee52a14:Add ::1 to localhost in etc/hosts
Bug: 258627476
Change-Id: I42c8b0aebdbd8b73a90606c0a999f25d8d071cd1
This reverts commit 1eb3394e9c.
Reason for revert: b/244406239, we've migrated to a sysprop(apex.all.ready) instead of an event.
Change-Id: Iae54df241257e3a3dcad4e54fdbf9dd14e9814de
The "sys.init.perf_lsm_hooks" is set on TestPerEventSelinux and it
is before early-init, but it need trigger by queue_property_triggers
and it is after late-init (zygote start on late-init).
The property is ready on load_bpf_programs, make sure
bpf_attach_tracepoint() is available on zygote start.
Bug: 257102190
Test: cat /sys/fs/bpf/map_time_in_state_uid_time_in_state_map
Change-Id: I5aa102df54b82e1584882800e93efd06ccf61c16
Remove the 'slave' mount flag that was added by commit ef9275223c
(https://r.android.com/2095463) because it doesn't actually do anything
in this context. MS_SLAVE can only be used to change the propagation
type of an existing mount, and the kernel ignores it if MS_BIND is also
specified, due to the way the various high-level operations that the
mount() system call can do are prioritized.
The reason that the /data/user/0 mount gets propagated into /data_mirror
anyway is because the /data mount has the "shared" propagation type. In
the above-mentioned commit I had assumed the default Linux mount
semantics, but actually Android applies the "shared" propagation type to
everything (see SetupMountNamespaces() in init/mount_namespace.cpp).
Test: Booted Cuttlefish and verified (via /proc/self/mountinfo) that
/data/data is still bind-mounted to both /data/user/0 and
/data_mirror/data_ce/null/0.
Bug: 156305599
BYPASS_INCLUSIVE_LANGUAGE_REASON=commit message mentioning removed code
Change-Id: Idc45d8dcb3a21d4e8e2e72f4d4dda7286f898127
CtsNativeVerifiedBootTestCases is currently flaky due to race conditions
between verity_update_state and the test running.
Moving the call to verity_update_state before zygote-start should fix
the test.
Bug: 253033920
Test: Boot Android and check that partitions.system.verified.hash_alg
has a non-empty value
Signed-off-by: Nathan Huckleberry <nhuck@google.com>
Change-Id: I9d252b0b6d74ed784ec2ffe091de2db53c5f45ba
The library provides an interface to interface with the
ConnectivityNative service, and implement port blocking APIs.
Bug: 179733303
Test: atest connectivity_native_test
Change-Id: I86018bfeb60d031faee818e5df469f02ebe32707
I noticed that the zygote64 and zygote64_32 files
had gotten slightly out of sync as a result of change
I3aad4b4b1d2f54db9e7ba86db8a655d8552bad0a. Merge the zygote64_32 changes
into zygote64, and to prevent this from happening again, replace the
64-bit zygote declaration in zygote64_32 with an import from zygote64.
Change-Id: I7fcceeb22b722c2164b9acf0b517a32ce34731fd
