This is part 2 of 2 of the refcounting refactor for IpSecService
resources.
Switched ManagedResources to use RefcountedResource structure for
managing reference counts and eventual cleanup. Further, resource arrays
and quota management have been aggregated into a UserRecord for better
isolation. UID access checking has been similarly moved into the
UserRecordTracker, and resourceId checking has been rolled into
RefcountedResourceArray's accessor methods.
Bug: 63409385
Test: CTS, all unit tests run on aosp_marlin-eng, new tests added
Change-Id: Iee52dd1c9d2583bb6bfaf65be87569e9d50a5b63
This patch adds (but does not enable the usage of) RefcountedResource
objects to IpSecService, with tests to ensure correct function. This is
patch 1 of a series of patches to refactor the resource management
systems in IpSecService.
RefcountedResource objects allow for management of acyclical dependency
trees, ensuring eventual cleanup when resources are no longer used. This
cleanup may be triggered by binder death or by explicit user action.
Bug: 63409385
Test: New tests written in IpSecServiceRefcountedResourceTest,
explicitly testing the RefcountedResource class
Change-Id: Ib5be7482b2ef5f1c8dec9be68f15e90d8b3aba6d
Throughout the IPsec code (API, system server, netd) we use "reserve"
SPI and "allocate" SPI interchangeably. This renames to use "allocate"
everywhere for self-consistency and consistency with the kernel
(ALLOCSPI).
In javadoc, I am leaving the word "reserve" in several places because it
is still an accurate description of how the method behaves.
Bug: 69128142
Test: TreeHugger should be enough
Change-Id: I8ea603b4612303b0393beef04032671fa53d2106
Improve the Validation of IpSecAlgorithm by
explicitly checking the length in addition to
the truncation length (previously an oversight).
In addition, we now check the lengths during
un-parceling, which will catch someone maliciously
manually building a parcel and passing it, bypassing
the checks in the constructor.
Bug: 68780091
Test: runtest -x IpSecAlgorithmTest.java
Change-Id: I8172762617264d34f47d5144336464510f07a701
This patch does some light refactoring in MacAddress to prepare for
exposing MacAddress in the public api:
- documention is improved
- some method names are renamed
- a toSafeString method is added
- a padding bug in the conversion methods outputting strings for
mac addresses is fixed
Bug: 69390696
Test: runtest frameworks-net
Change-Id: I399a97dabc2dfa8df9c5518c8b12484e43ca05c9
Also removes netId field from inside NetworkEvent and stop populating
the network_id field of the NetworkEvent field, deprecating it.
Bug: 34901696
Test: runtest frameworks-net,
manually looking at $ adb shell dumpsys connmetrics
Change-Id: I2c36860f976898883714f72f4d18e06da250c6a6
Added notes that keymat length must include 32 bits of salt.
Bug: 68672051
Test: Frameworks-net unit tests & IpSecManager CTS tests run
Change-Id: I0ae0c5be8a45b2374783b3bd1fa8bf930f15e687
This patch changes instrumentation of default networks and default
network events:
- stop logging events for default network transitions,
but instead consistently log one event per continuous segment
when one given network was the default, including logging an
event for when there is no default network.
- keep a separate rolling buffer of DefaultNetworkEvent for
dumpsys and bug reports.
These changes allow to simplify post aggregation of default network
event metrics by removing any need to do time series processing.
Instead, metrics and counters can be implemented withouth any ambiguity
by following the recipe:
% of x = sum(duration | x = true) / sum (all durations)
where x can be various conditions such as:
- the default network was validated
- the default network was WiFi
- the default network was IPv6
- there was no default network
- ...
Most importantly, this new logging scheme allows to measure much more
reliably:
- the % of the time that a device had Internet, in the sense that the
default network was validated.
- the time transitions between default networks, keyed by previous and
new transports/link layer, which allows to derive wakelock durations
and wakelock power costs from default network switches.
This patch also simplifies the dumpsys interface of the connmetrics
service and reduces the commands to three:
- "flush" for metrics upload.
- "proto" for printing buffered event in text proto format.
- "list" for listing all events and statistics.
Bug: 34901696
Bug: 65700460
Test: runtest frameworks-net
Change-Id: I0521f1681a60cca07ac3bfd5741d64ce44de4cdd
The "roaming" state of a network really belongs on NetworkCapabilities
instead of being published through NetworkInfo.isRoaming(). One major
reason is to support developers creating NetworkRequests for a
non-roaming network.
Watch for any capability changes that network statistics are
interested in (either metered or roaming) and notify it to perform
an update pass; fixes bug where we previously only triggered on
roaming changes.
Fix bug in VPNs where metered/roaming capabilities of underlying
networks weren't being propagated; this was probably preventing
some jobs from running over unmetered networks, and causing other
jobs to run over roaming networks! Also passes along link bandwidth
information from underlying networks, and propegates any changes
to underlying networks.
Fix race condition by reading prevNc inside lock. Utility methods
correctly calculate min/max link bandwidth values.
Test: bit FrameworksNetTests:android.net.,com.android.server.net.,com.android.server.connectivity.,com.android.server.ConnectivityServiceTest
Bug: 68397798, 16207332
Change-Id: I3e1a6544c902bf3a79356b72d3616af1fd2b0f49
