Occasionally, ConnectivityServiceTest fails with "BUG: only one
idle handler allowed". I have not been able to reproduce this
consistently, but code inspection reveals an unsafe access to
mIdleHandler inside queueIdle. Wrap that in a synchronized block.
Change-Id: I27307e2e55fa8d937d7f043bd436894091c3c667
1. Override WakeupMessage with an implementation that uses
sendEmptyMessageDelayed. This allows us to replace a
6-second sleep with a 150ms wait.
2. Change waitFor()'s polling interval from 100ms to 50ms.
With these changes ConnectivityServiceTest goes from ~9s to ~3s.
Change-Id: Id78aab9cded90a5b17f09d7f437904e179808dd2
NetworkStatsService will register data usage requests
and keep data usage stats scoped to the request.
There are different types of data usage requests
- scoped to a set of NetworkTemplate; these are restrictred to
device owners and carrier apps and allow the caller to monitor
all activity on the specified interfaces.
- scoped to all uids visible to the user, if the user has
android.Manifest.permission#PACKAGE_USAGE_STATS permission.
The set of uids may change over time, so we keep track of that.
- scoped to a set of uids given by the caller, granted that
the caller has access to those uids.
- scoped to the caller's own data usage. This doesn't require
PACKAGE_USAGE_STATS.
Bug: 25812785
Change-Id: Ie11f35fc1f29d0dbe82f7fc924b169bb55c76708
Whether a network is deemed roaming or not was already being tracked
as part of the NetworkIdentitySet, so the underlying data store
already tracks roaming and native data separately. However, this data
was being aggregated together in NetworkStatsCollection#getSummary,
since the NetworkIdentitySet is converted to an iface name for the
purposes of matching, and the iface name will be identical whether or
not the iface is considered roaming. Now it is separated.
Also fixes a long-standing bug in NetworkIdentitySet where an identity
read from a saved file would always be considered roaming == false,
even if it wasn't at the time it was written.
Bug: 25813438
Change-Id: I11ab5b51182ed8da7af8fde468df065f9fdc3dad
This also creates a hidden api for the captive portal server calculation
so that the Setup Wizard can use this as well.
bug:13246857
Change-Id: I4dfd0916df97cfce13252c7cc15f7bd05ed95f77
Current usage of NetworkSpecifier: network factory will match a request
if the request has either a (1) empty network specifier, or (2) a
network specifier which is identical to that of the network factory.
Note: 'matching' w.r.t. network specifier - all other matching rules
are still in effect.
Change: add rule (3) or the network specifier of the network factory
is the special string (which is defined as "*" and which user-facing
network requests aren't allowed to use).
Rationale: allows on-demand network creation.
Example:
- Can specify a Wi-Fi NetworkRequest with NetworkSpecifier="ssid"
- It will match a Wi-Fi network factory specifying NetworkSpecifier="*"
- That network factory will bring up a Wi-Fi network, connecting to the
specified SSID.
- Once the network is created it will create a NetworkAgent which will
now have a NetworkSpecifier matching that of the request (not the
match-all special string!)
That final step of making sure that the NetworkAgent matches the request
and not the match-all is critical to delivering any subsequent callbacks
correctly. I.e. your network will only get callbacks which match it.
Bug: 26192833
Change-Id: I49e3b492e0bb48a3f6e9a34e3f94f0e1cf89741f
Currently, access to network usage history and statistics requires a
signature|privileged permission, an AppOps bit (associated with the
PACKAGE_USAGE_STATS permission), or device/profile ownership. Once
access is granted via one of these mechanisms, it generally applies to
any UID running in the same user as the caller.
This CL expands access as follows:
-Any app can access its own usage history with no extra requirements.
-Carrier-privileged applications can access usage history for the
entire device.
-Device owners can access per-UID breakdowns for usage. Previously
they could access the summary for the whole device, but not the
individual breakdowns.
We simplify the permission model by defining three access levels -
DEFAULT (own app only), USER (all apps in the same user), and DEVICE
(all apps on the device), and propagate these levels throughout.
Finally, this CL fixes an apparent bug in
NetworkStatsSerice#hasAppOpsPermissions - if the AppOp bit was in
MODE_DEFAULT, hasAppOpsPermission would always return false instead of
falling back to the PackageManager permission check.
Bug: 25812859
Bug: 25813856
Change-Id: Ic96e0776e2a4215a400163872acea1ededfaced9
Breakages:
-ag/574873 - Renders testReportXtOverDev obsolete as this is no longer
a supported mode. Test has been removed.
-ag/600223 - Tests were sending a CONNECTIVITY_ACTION bcast to trigger
a call to updateIfaces(), but the listener was removed.
Tests now call forceUpdateIfaces() directly.
-ag/648284 - Calls to get VPN info were not mocked.
Change-Id: I309f2b5d006549104cb1d3cb83e99363dd6dac16
The Alarm Manager now supports a set() variant that takes a listener
callback to invoke at alarm trigger time rather than a PendingIntent.
This is much lower overhead and has guaranteed low delivery latency
from the trigger time. The tradeoff is that the app must be running
*continuously* from the time the alarm is set to the time it is
delivered. If the app exits for any reason before the alarm fires,
the listener becomes invalid and the alarm will be dropped. This is
more or less equivalent to setting an alarm with a broadcast
PendingIntent that matches only a runtime-registered receiver.
The app's alarm listener can be any object that implements the new
AlarmManager.OnAlarmListener interface and implements its onAlarm()
method. There is no data delivered at alarm trigger time: whatever
state needs to be associated with the specific alarm instance should
simply be packaged inside the OnAlarmListener instance.
An alarm using OnAlarmListener can request that the onAlarm() method
be called on an arbitrary handler. If the program passes 'null' for
this parameter when setting the alarm, the callback occurs on the
application's main Looper thread.
Cherry-picked from a75b36178d
Bug 20157436
Change-Id: I2eb030a24efdd466a2eee1666c5231201b43684b
The Alarm Manager now supports a set() variant that takes a listener
callback to invoke at alarm trigger time rather than a PendingIntent.
This is much lower overhead and has guaranteed low delivery latency
from the trigger time. The tradeoff is that the app must be running
*continuously* from the time the alarm is set to the time it is
delivered. If the app exits for any reason before the alarm fires,
the listener becomes invalid and the alarm will be dropped. This is
more or less equivalent to setting an alarm with a broadcast
PendingIntent that matches only a runtime-registered receiver.
The app's alarm listener can be any object that implements the new
AlarmManager.OnAlarmListener interface and implements its onAlarm()
method. There is no data delivered at alarm trigger time: whatever
state needs to be associated with the specific alarm instance should
simply be packaged inside the OnAlarmListener instance.
An alarm using OnAlarmListener can request that the onAlarm() method
be called on an arbitrary handler. If the program passes 'null' for
this parameter when setting the alarm, the callback occurs on the
application's main Looper thread.
Bug 20157436
Change-Id: I2eb030a24efdd466a2eee1666c5231201b43684b
This currently fails in many different ways, but it tells us what
to fix.
Bug: 22606153
Bug: 23884210
Change-Id: If2e5ee0a8d7b26cad67d3d566ed5b1383e0db096
1. Make TestNetworkCallback a bit smarter and rename it to
SingleUseNetworkCallback. This allows us to get rid of all the
calls to TestNetworkCallback#getConditionVariable.
2. Delete the commented out code that used to test a
ConnectivityService model that has not been used since KK.
3. Remove unused imports, etc.
Bug: 22606153
Change-Id: I81a2d0b970d19e5f4515490d8c2f88d416445fa1
Requests without NET_CAPABILITIES_INTERNET and just the default network
capabilities should not be marked restricted. Without this fix apps
can hit permissions exceptions if they inadvertently make requests
without NET_CAPABILITIES_INTERNET.
Bug:23164917
Change-Id: I4c7136821315bcb05dfc42ffbc505a5d4f6109e6
With this change:
1. NOT_RESTRICTED should be removed from NetworkRequests that bring up
special restricted carrier networks (e.g. IMS, FOTA).
2. NetworkRequests without NOT_RESTRICTED require CONNECTIVITY_INTERNAL
permission to register
3. Binding sockets to networks without NOT_RESTRICTED requires
CONNECTIVITY_INTERNAL permission
Bug:21637535
Change-Id: I5991d39facaa6b690e969fe15dcbeec52e918321
The methods startUsingNetworkFeature, stopUsingNetworkFeature and
requestRouteToHost were @removed in all the M preview builds, but
internal and external developers have noted that this imposes
additional burden for applications that need to work across
multiple platform versions because it causes compile-time errors.
We switched from @removed back to @deprecated to avoid these
problems. In order to effectively deprecate these methods, which
are error-prone and insecure, make them throw
UnsupportedOperationException if the app's target SDK is M or
above.
Because there are still one or two places in system code that use
these APIs, exempt Process.SYSTEM_UID and the OMA-DM client from
the check for now.
Bug: 22728205
Change-Id: I790bd32f3aa8067cbb625962a209bb9232f4b58c
If a network no longer satisfies a NetworkRequest, send the onLost
NetworkCallback. If it was a real request (not listen) then update
the NetworkFactories.
To test this change I created a little infrastructure to fake
different Internet connectivity probe results during tests. This
allowed me to rewrite some of ConnectivityServiceTest's logic for
validating networks. This brought to light a couple issues that
I had to address to keep tests passing:
1. testUnlingeringDoesNotValidate was relying on a bad side-effect
of my old method of ConnectivityServiceTest's logic for
validating networks, so I rewrote the test.
2. ConnectivityService was not sending out NetworkCallbacks for
WiFi when Cellular was validated. I'm including a fix for this
in this CL also.
Bug:22220234
Change-Id: I29314f38189817f8b2561a213c4f9e8522696663
Previously, once a network validated, for the purposes of comparing networks
to select the default network, we always considered it validated.
With this change if a network later fails to validate, we'll take this latest
validation result into account. This means if WiFi and cellular are up
(e.g. if we recently switched from cellular->WiFi, and cellular is now
lingering) and both are validated, but for some reason WiFi fails a validation,
cellular will become the default network connection.
Bug:20896761
Change-Id: I858aa10c1aaec5cd9032067f960963409107bdb1
Reduce the duplication of some logic so when falling back to Cellular
when WiFi fails to validate is enabled, there's less chance for bugs
and failures:
1. De-duplicate several Network vs NetworkRequest matching functions
2. Remove the very tricky nascent logic by adding a simple "lingering" bit.
Bug:20896761
Change-Id: I21da9e827eec9cfd6835fcaa650192b9186ed053
It's not clear what it means to request a network with a mutable
NetworkCapability like NET_CAPABILITY_VALIDATED or
NET_CAPABILITY_CAPTIVE_PORTAL. Presently requesting such a network
would fail in a number of different ways:
1. The NetworkFactories would fail to match the request against their
filter which doesn't include stateful NetworkCapabilities.
2. If the NetworkFactories did match, they'd bring up networks to try
and satisfy the requests, but the networks would not have any
mutable NetworkCapabilities initially so they'd be reaped.
Because of these problems it's safest to simply disallow these
requests.
Bug: 21343774
Change-Id: I56303242b81d39b370b8d5d1e32059bfcfc25949
We used to only remove requests that we'd acted on but that's
just wrong.
Also adds test case which exposed the problem but passes with the fix.
bug:20731384
Change-Id: I581a005560cc71167b857abf2452769399a9e1b7
Overlapping the NetIDs can cause the ConnectivityService instance under test
to inadvertently use real networks, for example when NetworkMonitor attempts
to validate a network. This fixes test hangs when run on devices with
active internet connections.
Change-Id: I5e1898953f0117b9f75beccac4a52ae2db173567
Add idle mode support to the alarm manager. Introduce
a new concept of flags associated with alarms to tell
the alarm manager how to treat the alarm -- they allow
everything from the alarm that will bring us out of idle
mode, to alarms that are allowed when idle or should
also bring us out of idle. The standalone boolean is
now also a flag.
(Note there is currently no protection from user space
setting the flags however it wants; I will be working
on that in a follow-up change.)
When in idle mode, the alarm manager pushes all alarms
that shouldn't execute during that time over to a
separate list that is not executed until out of idle.
To help with this, I reworked a bit how Alarm objects
are managed, so that when rebatching or moving between
lists we don't have to allocated new objects but can
just use the same existing instance.
Also tweaked the sync manager to deal with idle mode,
which currently just means doing the same thing as when
low on storage -- turning off sync.
Add new ACTION_CHARGING and ACTION_DISCHARGING broadcasts
that apps can listen for to know when the device is actively
charging and discharging. These are better than the old
POWER_CONNECTED and POWER_DISCONNECTED ones because we only
report charging when we actually see that there is enough
power being provided to charge the battery (and will report
discharging if there is not enough power).
The job controller uses these new actions for scheduling
jobs that want to run while plugged in. Removed the
"stable charging" stuff while doing so, since the new
charging state serves as an even better signal for that.
Introduced two new process states: FOREGROUND_SERVICE and
TOP_SLEEPING. This will allow us to treat foreground services
specially (such as still allowing network access to them for
background music playback) while not mixing them together with
whatever happens to be the top activity while the device is
asleep.
Also some other small cleanup here and there.
Change-Id: I7a9808b578bad6f50deb8e1baf919298512a0d3a
There are some cases where multiple subscriber identities (IMSI)
should be treated as "merged together" from a data usage
perspective. This is done by extending the template used for
matching purposes to support multiple subscribers.
Then, when we query historical usage or set network policies, we
normalize the matching template to merge to any other identities
that should be included. When normalizing, the "lowest" identity
is always used for equality and storage purposes, which allows
identities to come and go over time.
This change also fixes data usage recording for multi-SIM devices
by passing along the concrete subscriber identity for each network
interface. Also correctly create default policies for multi-SIM
devices. This change also drops setPolicyDataEnable() until it can
be wired up to the right underlying NetworkAgent. (This means we
still bring up the network, and then rely on iptables rules to block
traffic when over the limit, instead of proactively disabling the
connection.)
Bug: 18012787
Change-Id: If6acf32009fdfea2b836f5aff8e2f3e5e0248b4a
The previous code retrieved information from the legacy tracker multiple
times for each user query, leading to race conditions where the info
could've changed between the calls.
Refactors the handling of legacy data types in ConnectivityService and
unifies call paths so that APIs that deal with legacy data types
(NetworkInfo and int/networkType) and newer types (such as Network) go
through common code paths, using NetworkState to hold all the necessary
data pieces. This enables follow-on bug fixes to getActiveNetworkInfo().
The changes are limited to public "query" APIs (those that retrieve some
network information or the other). More details about the specific
changes and their rationale can be found in the code review thread.
Bug: 17460017
Change-Id: I656ee7eddf2b8cace5627036452bb5748043406c
Removing ConnectivityService.NetworkFactory. This requires disabling
the ConnectivityServiceTest, but that's been broken since we stopped
using NetworkStateTrackers anyway.
Change-Id: I9b86bd37eb9d018c40f60dca5b00d62c36d4e3ad
Adds a new kind of alarm that represents an alarm clock and
a way to query the next scheduled alarm clock.
Deprecates Settings.System.NEXT_ALARM_FORMATTED.
Bug: 14589952
Change-Id: I297eeeff36d07adcda010afac183d0f5ee37dc99
This is necessary so that the framework can know whether an IPv6
address is likely to be usable (i.e., if it's global scope and
preferred). Also, it will simplify the address notification
methods in INetworkManagementEventObserver, which currently take
the address, the flags, and the scope as separate arguments.
1. Add flags and scope to the class and update the unit test.
Use the IFA_F_* and RT_SCOPE_* constants defined by libcore.
Since most callers don't know about flags and scope, provide
constructors that default the flags to zero and determine the
scope from the address. Addresses notified by the kernel will
have these properly set. Make multicast addresses invalid.
Update the class documentation.
2. Provide an isSameAddressAs() method that compares only the
address and prefix information between two LinkAddress
objects. This is necessary because an interface can't have
two addresses with the same address/prefix but different
flags.
3. Update LinkProperties's addLinkAddress and removeLinkAddress
to identify existing addresses to add/remove using
isSameAddressAs instead of implicit equals(). Specifically:
- If addLinkAddress is called with an address that is already
present, the existing address's flags and scope are updated.
This allows, for example, an address on an interface to go
from preferred to deprecated when it expires, without it
having to be removed and re-added.
- If removeLinkAddress is called with an address that is
present but with different flags, it deletes that address
instead of failing to find a match.
4. Update the INetworkManagementEventObserver address
notification methods to take just a LinkAddress instead of
LinkAddress, flags, and scope. While I'm at it, change the
order of the arguments for consistency with the other
functions in the interface.
Change-Id: Id8fe0f09a7e8f6bee1ea3b52102178b689a9336e
Currently address{Updated,Removed} pass in the address as a
string such as "fe80::1/64". Use LinkAddresses instead, since
that's what it is.
This makes the code more robust in the unlikely case that netd
passes in an invalid string. In the future we can move flags and
scope into the LinkAddress itself and simplify the code further.
Bug: 9180552
Change-Id: I66599f9529cf421caa7676fdd0141bb110b8589e
These are sent if the device receives IPv6 Router Advertisements
with DNS server configuration options. Currently, nothing listens
to them; in a future change we will use them as IPv6 DNS servers.
[Cherry-pick of af70f6c614]
Bug: 9180552
Change-Id: I05000c0cd3867a68ab390102e8470b6912a9d3aa
