* changes:
libprocessgroup: Check validity of uid and pid arguments for createProcessGroup
Revert "libprocessgroup: Check validity of uid and pid arguments"
killProcessGroup already checks for validity of the uid and pid input
arguments. Also add this checking to createProcessGroup.
Bug: 333261173
Change-Id: Ie74b9c2716978cb5eb434af12e27945e29af876e
This flag adds "apps" and "system" cgroups underneath the v2 hierarchy
root. Cgroups with UIDs < 10000 (AID_APP_START) will be placed
under "system" and others will be placed under "apps". UIDs under 10000
are reserved for core Android subsystems. This allows us to
apply different cgroup controls collectively to system processes and
normal applications.
Bug: 327480673
Change-Id: I40837dee27a59691f81fef48e66a86c5eacda892
This flag prevents memcg from being mounted as a v1 hierarchy, even if
cgroups.json specifies it should be mounted as v1. It will activate
memcg in the v2 hierarchy even if cgroups.json does not specify that it
should be activated in the v2 hierarchy.
The cgroup_disable=memory kernel command line argument will still
prevent memcg from being activated, as this forced memcg controller
is marked as optional.
Bug: 327480673
Change-Id: Iad2491dd0c1576156ee2346928d041d85af890f0
We already get -Wall and -Werror from the build system, and we do not
want/need -Wexit-time-destructors since it prevents local statics with
non-trivial destructors.
Test: m
Change-Id: I8283bf223404d6c253861d3888c1b720c099386e
* changes:
libprocessgroup: Use pid_t consistently for TIDs
libprocessgroup: Use pid_t consistently for PIDs
libprocessgroup: Use uid_t consistently for UIDs
By using cgroup.kill we don't need to read cgroup.procs at all for
SIGKILLs, which is more efficient and should help reduce CPU contention
and cgroup lock contention. Fallback to cgroup.procs if we encounter an
error trying to use cgroup.kill, but if cgroup.kill fails it's likely
that cgroup.procs will too.
Bug: 239829790
Test: atest StagedRollbackTest:com.android.tests.rollback.host.StagedRollbackTest#testNativeWatchdogTriggersRebootlessApexRollback
Change-Id: I9da67efd00af49b7b4b502fb742c1095d5c7b9e9
This reverts commit a72332f953.
This change was originally reverted due to a bug in the child commit
which caused process group kills not to occur when using cgroup.kill.
Now that has been fixed, bring back this change with the fixed child.
Bug: 301871933
Change-Id: Ia6c74d9b67a8c88aec4812ac4655646934e0d189
This revert was created by Android Culprit Assistant. The culprit was identified in the following culprit search session (http://go/aca-get/91da3c52-9b76-498b-bdbd-a9de7d7ff53b).
Change-Id: I996c595bee9acc15aedaf0a912f67fa027f33cd0
This revert was created by Android Culprit Assistant. The culprit was identified in the following culprit search session (http://go/aca-get/91da3c52-9b76-498b-bdbd-a9de7d7ff53b).
Change-Id: I459265b9c9117d6006c1223947a202505d24c08f
By using cgroup.kill we don't need to read cgroup.procs at all for
SIGKILLs, which is more efficient and should help reduce CPU contention
and cgroup lock contention. Fallback to cgroup.procs if we encounter an
error trying to use cgroup.kill, but if cgroup.kill fails it's likely
that cgroup.procs will too.
Bug: 239829790
Change-Id: I44706faccfb7c4611b512a3642b913f06d30c1dc
In killProcessGroup we currently read cgroup.procs to find processes to
kill, send them kill signals until cgroup.procs is empty, then remove
the cgroup directory. The cgroup cannot be removed until all processes
are dead, otherwise we'll get an EBUSY error from the kernel.
There is a race in the kernel where cgroup.procs can read empty even
though the cgroup is pinned by processes which are still exiting, and
can't be removed yet. [1]
Let's use the populated field of cgroup.events instead of an empty
cgroup.procs file to determine when the cgroup is removable. In
addition to functioning like we expect, this is more efficient because
we can poll on cgroup.events instead of retrying kills and rereading
cgroup.procs every 5ms which should help reduce CPU contention and
cgroup lock contention.
It's still possible that it takes longer for a cgroup to become
unpopulated than our timeout allows, in which case we will fail to
remove the cgroup and leak kernel memory. But this change should help
reduce the probability of that happening.
[1] https://lore.kernel.org/all/CABdmKX3SOXpcK85a7cx3iXrwUj=i1yXqEz9i9zNkx8mB=ZXQ8A@mail.gmail.com/
Bug: 301871933
Change-Id: If7dcfb331f47e06994c9ac85ed08bbcce18cdad7
Provide an additional way to request the block layer to prioritize
foreground I/O over background I/O. This patch prepares for tests with
the mq-deadline I/O scheduler. While CFQ and BFQ support a "weight"
cgroup attribute, the mq-deadline scheduler does not. The mq-deadline
I/O scheduler only supports the I/O priority class for differentiating
I/O priorities.
The io.prio.class attribute is declared optional since this attribute
only exists if CONFIG_BLK_CGROUP_IOPRIO != n in the kernel
configuration.
Bug: 186902601
Change-Id: Iee1004cd0996e32245aff2b51772ef40178e024f
Signed-off-by: Bart Van Assche <bvanassche@google.com>
CL 2828279 doesn't do what it's description says that it does. Making
Service::Stop() work for processes that have been migrated to another v2
cgroup requires changing DoKillProcessGroupOnce(). Hence this CL that
removes the early return statements from DoKillProcessGroupOnce().
Bug: 308900853
Change-Id: Ib798555feeb95a786a619c3d7013c7d5829b01ad
Signed-off-by: Bart Van Assche <bvanassche@google.com>
Revert this CL because it tests whether or not cgroup.procs files are
empty with the stat() system call and because the cgroup filesystem
always reports st_size == 0. Rename RemoveUidCgroups() into
RemoveEmptyUidCgroups().
Change-Id: I4de6f16c814c4b47d8d74c8045f0c1ee71975ac0
Signed-off-by: Bart Van Assche <bvanassche@google.com>
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 changes compared to the previous version of this CL are as follows:
- The JoinCgroup actions for the "io" controller have been left out
since these caused processes to be migrated to the v2 root cgroup.
- The BfqWeight / CfqGroupIdle / CfqWeight settings have been included
in this CL instead of applying these settings as a separate CL.
Change-Id: I67e06ce3462bb1c1345dba78f8d3d655b6519c74
Signed-off-by: Bart Van Assche <bvanassche@google.com>
A JoinCgroup action for a v2 cgroup controller migrates a process or task
from the uid_%d/pid_%d cgroup into another cgroup, e.g. the root cgroup.
This may make services unkillable because Service::Stop() only stops a
service if the uid_%d/pid_%d cgroup still exists when Service::Stop() is
called.
Bug: 309674654
Change-Id: I20b797afdf596125ff5a6ed41cb33fe59b84ac88
Signed-off-by: Bart Van Assche <bvanassche@google.com>
For log messages like the following it is not possible to derive why
this message has been logged:
E libprocessgroup: AddTidToCgroup failed to write '3949'; fd=55: Operation not supported on transport endpoint
Hence include the cgroup path and the tid type in the log message.
Change-Id: I057711fe576b82f6454456b7284186ddeece33c3
Signed-off-by: Bart Van Assche <bvanassche@google.com>
When attempting to remove a cgroup, a ENOENT means this cgroup is
already removed. Treat such errno as success for idempotency.
Test: th
Bug: 308900853
Change-Id: I6ef3c25f03d185194205b3845784d284fdc4d444
PLOG depends on errno being set to a useful value, otherwise it will
print a meaningless error string. A few PLOG call sites occur where
either errno is not set at all, or it is set only some of the time where
there are already PLOG calls closer to where the error occurs. Convert
these PLOG calls to LOG.
Bug: 301871933
Change-Id: Ifa6bd2401f9dd9b84b2506e886336e89bac81bb1
The max_processes calculation is incorrect for KillProcessGroup because
the set of processes in cgroup.procs can differ between the multiple
reads in the implementation. Luckily the exact value isn't very
important because it's just logged. Remove max_processes from the API
and remove the warning about the new behavior in Android 11.
Note that we still always LOG(INFO) that any cgroup is being killed.
Bug: 301871933
Change-Id: I8e449f5089d4a48dbc1797b6d979539e87026f43
Currently we sleep for 5ms before decrementing retries for the last
time. This is a waste of time, so bail out of the loop if the last
rmdir attempt fails.
Change-Id: Ia20840d27592b4eb3d9762647b19c111ff94209f
Current implementation of SetAttributeAction::ExecuteForProcess reuses
SetAttributeAction::ExecuteForTask while not utilizing available uid/pid
information. This results in a call to GetPathForTask() which is an
expensive function due to it reading and parsing /proc/$pid/cgroups.
This can be avoided if we utilize available uid/pid info and the fact
that cgroup v2 attributes share the cgroup v2 hierarchy as process
groups, which use a known path template.
Bug: 292636609
Change-Id: I02e3046bd85d0dfebc68ab444f1796bb54cc69c7
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
ProfileAttribute::Reset does not reset file_v2_name, fix that. Also
provide ProfileAttribute::file_name() to consolidate the code.
Bug: 292636609
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Change-Id: I5b33ca47b4fa5cabf582c8804bd13f72f6e58411
We missed two incorrect specifiers in the previous commit with this same
title.
We use the %d format specificier for uid_t, which maps to
__kernel_uid32_t, which is unsigned. [1] This is undefined behavior
which can lead to paths with negative UIDs when erroneously large
values are passed for uid:
E libprocessgroup: No such cgroup attribute: /sys/fs/cgroup/uid_-89846/cgroup.freeze
Fix it with %u.
[1] https://cs.android.com/search?q=typedef.*__kernel_uid32_t&ss=android%2Fplatform%2Fsuperproject%2Fmain
Change-Id: Ica04b03526bd2e156f026a2797fe9912b259cd9f
We use the %d format specificier for uid_t, which maps to
__kernel_uid32_t, which is unsigned. [1] This is undefined behavior
which can lead to paths with negative UIDs when erroneously large
values are passed for uid:
E libprocessgroup: No such cgroup attribute: /sys/fs/cgroup/uid_-89846/cgroup.freeze
Fix it with %u.
[1] https://cs.android.com/search?q=typedef.*__kernel_uid32_t&ss=android%2Fplatform%2Fsuperproject%2Fmain
Change-Id: Ibb52ba2503e30e2f20770b7d23629167e38d076a
Global UID level cgroup removal was eliminated because of a race
between app launch and app killing using the same directory name. [1]
However isolated app UIDs are assigned sequentially, and are
basically never reused until we wrap around the large range of
isolated UIDs. This leaves thousands of isolated cgroup directories
unused, which consumes kernel memory and increases memory reclaim
overhead. Remove this subset of UID level cgroup directories when
killing process groups.
[1] d0464b0c01
Test: 50 cycle ACT leaves 1000 fewer empty isolated cgroups
Bug: 290953668
Change-Id: If7d2a7b8eec14561a72208049b74ff785ca961bd
This reverts commit 9c0fcbb0f7.
Bug: 261857030
Ignore-AOSP-First: this change is for the U branch only.
Change-Id: Id4bd3494b33dd6bc0644406d599c9bfd597c7435
Signed-off-by: Bart Van Assche <bvanassche@google.com>
This reverts commit 92153fb955.
Bug: 261857030
Ignore-AOSP-First: this change is for the U branch only.
Change-Id: I99417707f0d0b8c7dca3927b6ce9d52436621f4e
Signed-off-by: Bart Van Assche <bvanassche@google.com>
This reverts commit 6ad747ac2d.
Bug: 261857030
Ignore-AOSP-First: this change is for the U branch only.
Change-Id: I93447b71146e6e9297ef49026d90dc4c35b91244
Signed-off-by: Bart Van Assche <bvanassche@google.com>
