This is required for the android.sdk_sysprop_guard data source
implemented in traced's builtin producers to set the system properties
that guard Skia's inititialization of the Perfetto SDK and track_event
data source in SurfaceFlinger/RenderEngine and HWUI apps.
Bug: 279614916
Test: presubmit?
Change-Id: Ia04014336e86672e54c21883533afe70fcb83e84
We are changing the --save-for-bugreport feature and moving
the file opening/write from the traced service to the perfetto
cmdline client.
This is as part of a bigger refactor to simplify the API surface
in view of non-destructive snapshots of trace buffers.
Add matching sepolicies to perfetto.te
Bug: 260112703
Test: atest perfetto_integrationtests --test-filter '*PerfettoCmdlineTest*'
Change-Id: Ic1dd6b1bf3183f6b7fb551859e35cae950676ffb
In order for ART code to call perfetto DataSource::Trace() we need to
wait for all data source instances to have completed their setup. To do
so, we need to know how many of them exist.
This introduces a new sysprop traced.oome_heap_session.count, writeable
by perfetto traced and readable by apps and system_server that can be
used to communicate this.
See go/art-oom-heap-dump for more details
Test: manual, atest HeapprofdJavaCtsTest
Bug: 269246893
Change-Id: Ib8220879a40854f98bc2f550ff2e7ebf3e077756
The feature was superseded by tzdata mainline module(s).
Bug: 148144561
Test: see system/timezone
Test: m selinux_policy
Change-Id: I48d445ac723ae310b8a134371342fc4c0d202300
Merged-In: I48d445ac723ae310b8a134371342fc4c0d202300
Remove references in sepolicy. Leave a few of the types defined since
they're public and may be used in device-specific policy.
Bug: 211461392
Test: build/boot cuttlefish
Change-Id: I615137b92b82b744628ab9b7959ae5ff28001169
Creating a per-user encrypted directory such as /data/system_ce/0 and
the subdirectories in it too early has been a recurring bug. Typically,
individual services in system_server are to blame; system_server has
permission to create these directories, and it's easy to write
"mkdirs()" instead of "mkdir()". Such bugs are very bad, as they
prevent these directories from being encrypted, as encryption policies
can only be set on empty directories. Due to recent changes, a factory
reset is now forced in such cases, which helps detect these bugs;
however, it would be much better to prevent them in the first place.
This CL locks down the ability to create these directories to just vold
and init, or to just vold when possible. This is done by assigning new
types to the directories that contain these directories, and then only
allowing the needed domains to write to these parent directories. This
is similar to what https://r.android.com/1117297 did for /data itself.
Three new types are used instead of just one, since these directories
had three different types already (system_data_file, media_rw_data_file,
vendor_data_file), and this allows the policy to be a bit more precise.
A significant limitation is that /data/user/0 is currently being created
by init during early boot. Therefore, this CL doesn't help much for
/data/user/0, though it helps a lot for the other directories. As the
next step, I'll try to eliminate the /data/user/0 quirk. Anyway, this
CL is needed regardless of whether we're able to do that.
Test: Booted cuttlefish. Ran 'sm partition disk:253,32 private', then
created and deleted a user. Used 'ls -lZ' to check the relevant
SELinux labels on both internal and adoptable storage. Also did
similar tests on raven, with the addition of going through the
setup wizard and using an app that creates media files. No
relevant SELinux denials seen during any of this.
Bug: 156305599
Change-Id: I1fbdd180f56dd2fe4703763936f5850cef8ab0ba
This change adds a neverallow rule in traced.te to limit the processes
that can find tracingproxy_service, the context for TracingServiceProxy.
I wanted to avoid moving the tracingproxy_service definition to public,
so there were a few services that are exempted from this neverallow
rule.
Bug: 191391382
Test: Manually verified that with this change, along with the other
change in this topic, I see no errors when taking a bugreport while a
Traceur trace is running.
Change-Id: I8658df0db92ae9cf4fefe2eebb4d6d9a5349ea89
Configures sepolicy to allow for the new TracingServiceProxy system
services, and to allow Perfetto to access the service.
Bug: 175591887
Test: Validated the service started successfullyy, and invoked via CLI
Change-Id: Idb6438948a9d96063f8455544b97ef66267cde23
Feature description: if a background trace is happening at the
time dumpstate is invoked, the tracing daemon will snapshot
the trace into a fixed path (/data/misc/perfetto-traces/bugreport/).
Dumpstate will attach the trace, if present, to the bugreport.
From a SELinux viewpoint this involves the following permissions:
- Allow dumpstate to exec+trans perfetto --save-for-bugreport
(this will just send an IPC to traced, which will save the trace).
- Allow dumpstate to list, read and unlink the trace file.
- Create a dedicated label for bugreport traces, to prevent that
dumpstate gets access to other traces not meant for bug reporting.
Note that this does NOT allow dumpstate to serialze arbitary traces.
Traces must be marked as "eligible for bugreport" upfront in the
trace config (which is not under dumpstate control), by
setting bugreport_score > 0.
Design doc: go/perfetto-betterbug
Bug: 170334305
Test: manual:
1. start a perfetto trace with bugreport_score > 0
2. adb shell dumpstate
3. check that the bugreport zip contains the trace
Change-Id: I259c3ee9d5be08d6b22c796b32875d7de703a230
This allows us to log metrics from traced to statsd
for failures. This is required for implementation of
go/perfetto-failure-stats.
This matches the CL aosp/1512303 which adds the initial logging to
traced.
This solves the following denied message from logcat:
avc: denied { write } for name="statsdw" scontext=u:r:traced:s0
tcontext=u:object_r:statsdw_socket:s0
Bug: 177215620
Change-Id: I4cc613c3a8f4d75c4a5c232b996f8a6cffd3ba9d
Together with aosp/1282157 this change allows the service to
create trace files in the /data/misc/perfetto-traces folder.
Before this change they needed to be created by the perfetto
cmdline client and pass the FD.
This doesn't work for host tools like Android GPU Inspector
(https://gpuinspector.dev/) which talk to the UNIX socket
over adb forward and cannot pass a FD from the host.
Bug: 153519149
Test: manual: adb shell perfetto --txt -c -
buffers { size_kb: 65536 }
data_sources {
config {
name: "linux.ftrace"
ftrace_config { ftrace_events: "sched_switch" }
}
}
duration_ms: 5000
write_into_file: true
output_path: "/data/misc/perfetto-traces/ttt"
Change-Id: I184329805741654983843e6a29c1fac19a836f59
This concerns the data transfer between an untrusted producer process,
and the tracing service (traced daemon). They communicate over a
combination of a unix socket and shared memory.
Normally, the service creates the shared memory region, and hands it off
to the producer process (see perfetto_producer() macro). This patch
allows for an alternative scheme, where the producer process is allowed
to create the shared memory region, which will then be adopted by the
tracing service. The service already inherently doesn't trust the
producer, so it'll validate that the shared memory is appropriately
sealed before using it.
The immediate use-case is chrome's go/perfetto-startup-tracing-v2. But
this mode has advantages (e.g. being able to write to the shared memory
before connecting) for other producer domains as well.
Bug: 148841422
Change-Id: I90f864b900958792553f0208f4a0041dbf2892cc
Give /data itself a different label to its contents, to ensure that
only init creates files and directories there.
This change originally landed as aosp/1106014 and was reverted in
aosp/1116238 to fix b/140402208. aosp/1116298 fixes the underlying
problem, and with that we can re-land this change.
Bug: 139190159
Bug: 140402208
Test: aosp boots, logs look good
Change-Id: I1a366c577a0fff307ca366a6844231bcf8afe3bf
Give /data itself a different label to its contents, to ensure that
only init creates files and directories there.
Bug: 139190159
Test: aosp boots, logs look good
Change-Id: I3ee654a928bdab3f5d435ab6ac24040d9bdd9abe
Bug: http://b/135139675
Coverage files are written to /data/misc/trace (governed by the
method_trace_data_file selinux type). Allow all domains to access
(create directories, access files) this directory when native coverage
is enabled (by setting NATIVE_COVERAGE to true) in an userdebug or eng
build.
Also relax neverallow constraints to allow access to
method_trace_data_file for native coverage builds.
Test: Build 32-bit cuttlefish with coverage:
m NATIVE_COVERAGE=true COVERAGE_PATHS="*"
and verify that there are no selinux denials in kernel log and
logcat.
Change-Id: I3fe7c77612854b9de7de7a0ddd5cbf44a2f5c21e
This CL introduces allows traced to set the
sys.traceur.trace_end_signal property at the end
of the tracer. In turn that property notifies the
the Traceur app.
This is to allowing Traceur to be killed during
a long-trace and avoid wasting resources making
it a persistent service.
See aosp/886616 for the matching traceur change.
Test: manual
Bug: 116754732
Change-Id: I89e2f02b3f973813ce8ff3507d397a06502f84c1
The test was failing because of a selinux denial.
This adds the required rule.
Test: atest (previously failing CTS test passed)
Change-Id: Ieb99f9ab4c6014a3d0aa1fe6c6fb6b82fa9b7631
This is being done in preparation for the migration from ashmem to
memfd. In order for tmpfs objects to be usable across the Treble
boundary, they need to be declared in public policy whereas, they're
currently all declared in private policy as part of the
tmpfs_domain() macro. Remove the type declaration from the
macro, and remove tmpfs_domain() from the init_daemon_domain() macro
to avoid having to declare the *_tmpfs types for all init launched
domains. tmpfs is mostly used by apps and the media frameworks.
Bug: 122854450
Test: Boot Taimen and blueline. Watch videos, make phone calls, browse
internet, send text, install angry birds...play angry birds, keep
playing angry birds...
Change-Id: I20a47d2bb22e61b16187015c7bc7ca10accf6358
Merged-In: I20a47d2bb22e61b16187015c7bc7ca10accf6358
(cherry picked from commit e16fb9109c)
This requires moving the type declaration of
perfetto traced to public, because iorapd
needs to refer to it.
Denials without this CL:
https://pastebin.com/raw/sxHMeLEU
Bug: 72170747
Test: 1. runcon u:r:iorapd:s0 iorap.cmd.perfetto \
-v --output-proto /data/misc/iorapd/test
2. Check that no selinux denials other than
avc: denied { entrypoint } for path="/system/bin/iorap.cmd.perfetto" dev="sda6" ino=21 scontext=u:r:iorapd:s0 tcontext=u:object_r:system_file:s0 tclass=file permissive=1
show up (this is a side-effect of runcon).
Change-Id: Iacd1ab201fe9fb2a6302dbd528f42f709cbca054
This appears to be the minimum change required to accommodate Traceur
running the detachable Perfetto process.
Bug: 116754732
Test: Started a perfetto trace using --detach and it started
successfully.
Change-Id: I12881ae343389abdcc74af5f11ecbac99b03ef7c
system_file_type is a new attribute used to identify files which exist
on the /system partition. It's useful for allow rules in init, which are
based off of a blacklist of writable files. Additionally, it's useful
for constructing neverallow rules to prevent regressions.
Additionally, add commented out tests which enforce that all files on
the /system partition have the system_file_type attribute. These tests
will be uncommented in a future change after all the device-specific
policies are cleaned up.
Test: Device boots and no obvious problems.
Change-Id: Id9bae6625f042594c8eba74ca712abb09702c1e5
This allows an optimization that consists in the "perfetto" cmdline
client passing directly the file descriptor for the output trace
to traced (as opposite to having traced streaming back the trace
data to "perfetto" and having that one doing the write() into file).
This reduces sensibly the memory traffic and CPU overhead of traces
with a minor change.
Bug: 73625179
Test: builds + perfetto_integrationtests w/ long_trace.cfg
Change-Id: I81f5a230338ced20dc543fd91c5a0bd0e58725f2
The previous selinux rules obtained via audit2allow didn't really
work with the case of apps connecting to the producer socket,
despite all the allow rules being correctly in place.
This was failing our CTS tests.
The reason for the failure (see denials pasted below) is due to
Multi Level Security (for multi-user), which was still preventing
apps form a different level to connect to the traced producer
socket and write to the shmem buffers they get passed back.
This CL tags the objects being accessed as mlstrusted.
CTS tests pass with this CL.
Denials:
avc: denied { write } for pid=8545 comm="traced_probes" name="traced_producer" dev="tmpfs" ino=23629 scontext=u:r:untrusted_app_25:s0:c512,c768 tcontext=u:object_r:traced_producer_socket:s0 tclass=sock_file permissive=1
avc: denied { write } for pid=8545 comm="traced_probes" name="traced_producer" dev="tmpfs" ino=23629 scontext=u:r:untrusted_app_25:s0:c512,c768 tcontext=u:object_r:traced_producer_socket:s0 tclass=sock_file permissive=1
avc: denied { connectto } for pid=8545 comm="traced_probes" path="/dev/socket/traced_producer" scontext=u:r:untrusted_app_25:s0:c512,c768 tcontext=u:r:traced:s0 tclass=unix_stream_socket permissive=1
avc: denied { connectto } for pid=8545 comm="traced_probes" path="/dev/socket/traced_producer" scontext=u:r:untrusted_app_25:s0:c512,c768 tcontext=u:r:traced:s0 tclass=unix_stream_socket permissive=1
avc: denied { write } for pid=8545 comm="traced_probes" path=2F6D656D66643A706572666574746F5F73686D656D202864656C6574656429 dev="tmpfs" ino=104483 scontext=u:r:untrusted_app_25:s0:c512,c768 tcontext=u:object_r:traced_tmpfs:s0 tclass=file permissive=1
Change-Id: I1598bc0b07bf39b8d0420b66caf06a4ca884f383
Bug: 73340039
Test: CtsPerfettoTestCases
Restrictions introduced in vendor init mean that new devices
may not no longer exempt vendor init from writing to system_data_file.
This means we must introduce a new label for /data/vendor which
vendor_init may write to.
Bug: 73087047
Test: build and boot Taimen and Marlin. Complete SUW, enroll fingerprint
No new denials.
Change-Id: I65f904bb28952d4776aab947515947e14befbe34
Perfetto is a performance instrumentation and logging framework,
living in AOSP's /external/pefetto.
Perfetto introduces in the system one binary and two daemons
(the binary can specialize in either depending on the cmdline).
1) traced: unprivileged daemon. This is architecturally similar to logd.
It exposes two UNIX sockets:
- /dev/socket/traced_producer : world-accessible, allows to stream
tracing data. A tmpfs file descriptor is sent via SCM_RIGHTS
from traced to each client process, which needs to be able to
mmap it R/W (but not X)
- /dev/socket/traced_consumer : privilege-accessible (only from:
shell, statsd). It allows to configure tracing and read the trace
buffer.
2) traced_probes: privileged daemon. This needs to:
- access tracingfs (/d/tracing) to turn tracing on and off.
- exec atrace
- connect to traced_producer to stream data to traced.
init.rc file:
https://android-review.googlesource.com/c/platform/external/perfetto/+/575382/14/perfetto.rc
Bug: 70942310
Change-Id: Ia3b5fdacbd5a8e6e23b82f1d6fabfa07e4abc405
