GitLab

This timestamp needs to be reset when WiFi is disabled. When WiFi
is enabled again, a connectivity scan starts immediately.
Otherwise it can be delayed because of the minimum periodic single
scan interval rule.

Bug: 28986330
Change-Id: I5a01dc267a05791eba34c87e6d9301bd998bf377
Tests: Wifi unit tests, manual test

Change-Id: Ia18913aed38ff5fa7655878340b99e15132a32d1
Fixes: 28970259

Users are unable to change their preference since we've taken out the UI

Changed the setFrequencyBand method to always set the band to 'AUTO'.

b/28906443

Change-Id: I7ebd8aa204f72ecee754e31662d8ebf509ae6063

Bug: 28320003
Change-Id: Ic551cc6d7fcaa94945d2768d7aa2cf705dd510fd

Cut down the max number of lines allowed in Wifi LocalLogs.
Currently LocalLog checks the number of lines in its buffer
instead of total size in number of bytes. The buffer being
used can grow really big if some lines are lengthy.

Bug: 28865103
Change-Id: I7f77de1048bb465b79265b3ecb3a925e3a2659d9
Test: Wifi Framework Unit Tests

Instead of always starting a single scan when the device enters
screen on state, we now keep track of how long it has been since
last periodic single scan and defer the new scan request.

Bug: 28887789
Tests: Wifi Unit Tests, manual tests
Change-Id: Iba5b5b7db897d3227d87f8d989246a1942525b9c

* changes:
  WifiNetworHistory: Don't persist QNS blacklist
  Change clock for time interval/elapsed measurments

Tell WifiConnectivityManager the screen state in case
WifiConnectivityManager missed the screen update event because
it was not started yet.

Bug: 28824622
Tests: Wifi Framework Unit Tests; reproduced the symptom and
       verified it went away after this change.

Change-Id: I33cce9276982f6e77cf89a42ddd14ad00f2e4f46

Bug: 28682351
Change-Id: If9296d862adbd479e8600699f06f4684d377ae6a

QNS blacklists network temporarily by setting the network status. Reset
this value when we bootup so that this not persisted. This is especially
needed as we move away from using wall clock for blacklist time.

Also,
Remove the unused blacklist timestamp key from network history parsing.

BUG: 28824635
Change-Id: Ie64c11f358b2e3389ec419cfcfdde8bb04298dd4
TEST: Compliles & unit-test passes

Fall back to full band scan when the currently connected network is not
present in scan result cache. This could be a transient problem, but we
don't want to cause an error by sending an invalid scan request.

Changes needed for unit-testing this change:
1. Make WifiInfo a real instance so that we can set the values for the
public members directly accessed like tx/rx success rate.
2. Make |getCurrentWifiConfiguration| public so that it can be stubbed
out.

BUG: 28845855
TEST: Compiles & unit-test passes
Change-Id: I6cb2b7ad762b3f8f2ab36d66c4b1fcc39fafffed

1. Use the clock retrieved from WifiInjector in more modules: QNS,
Scanner, WifiConfigManager.
2. Use |elapsedRealtime| for time interval/elapsed measurments. Things
like scan results, config elements still continue to use
|currentTimeMillis|

Also,
1. Had to make |tryEnableQualifiedNetwork| public to prevent mockito from
calling the real method in WifiConfigManager.
2. Move keystore also into WifiInjector.

BUG: 28299258
BUG: 28824635
Change-Id: I91272aa5665abb741da7530d8920d09c8c5ea41f
TEST: Compiles & unit-test passes

We've observed that it's tricky to detect when a connection
fails. Consequently, even if the user is careful to generate
a bugreport while the device is in the failed state, the
bugreport will not include packet fate data for the most
recent connection.

Improve the utility of packet fate, by fetching the most
recent fate on the call to dump(). With this change in place,
we'll get packet fates for the most recent connection
attempt (even if it's still in progress).

BUG=28849396
TEST=unit tests, manual

Manual test
- configure bullhead for open network
- configure bullhead for wpa-2 network, wrong password
- let bullhead fail connecting to wpa-2 network
- let bullhead succeed connecting to open network
$ adb dumpsys > /tmp/dumpsys
$ grep 'fates ---' /tmp/dumpsys
--------------------- Last failed connection fates ----------------------
--------------------- Latest fates ----------------------

Change-Id: Iaeb2584020f70ac5d2f3139bac72ea07a7ddce81

Starts partial scan only when Wifi is connected and traffic
is heavy. This is to work around an issue that WifiInfo
tx/rx rate is not updated to reflect the Wifi state. Ideally
the tx/rx rate should have been reset to 0 when Wifi is
disconnected.

Bug: 28831562
Change-Id: Ifa415b14e2b216654ddd8a2a0c885754da489bef
Test: Wifi Framework Unit Tests, manual tests

Before this, when WifiScanningService processed single scan results from
supplicant that only contained results that did not match a request the
filter code would return null indicating that the results should not be
delivered, causing a NPE later. This change updates the meaning of the
bucket index that the single scan code provides to the filtering code to
instead force it to always include scan results.

Change-Id: I65cd57b14abacec3f407991188c570601d05ac77
Fixes: 28794598

Packet fate reports are timestampped in a different format than
logcat messages. This makes it hard to correlate packet fate
summaries with other system activity.

Resolve this, by adding wall-clock timestamps to packet fate
summaries.

BUG=28701632
TEST=manual test

Manual test:
- set bullhead to non-GMT timezone
- install build on bullhead
- trigger connection failure
- verify that walltime column is in the right ballpark,
  relative to logcat messages

Change-Id: I3aed8d9acca42f0ef0cc4f2f2fbdabdde33ce4c8

Starts partial scan only when Wifi is connected and traffic
is heavy. This is to work around an issue that WifiInfo
tx/rx rate is not updated to reflect the Wifi state. Ideally
the tx/rx rate should have been reset to 0 when Wifi is
disconnected.

Bug: 28831562
Change-Id: Ifa415b14e2b216654ddd8a2a0c885754da489bef
Test: Wifi Framework Unit Tests, manual tests

* changes:
  WifiLogger: use large ringbuffers on eng/userdebug
  WifiService: add build properties abstraction

Bug: 28320003
Change-Id: I4cfdedb1d5c3c29608208bb1c3bda42f4759fb7f

We're having trouble resolving some field reports, due
to insufficient time coverage in the HAL ring buffers.

Increase the ringbuffer size of eng/userdebug builds,
to make it easier to debug issues reported by
dogfooders.

This change will be reverted closer to launch.
(b/28822174 has been filed for the reversion.)

BUG=28729317
TEST=unit tests

Change-Id: I5884dee05db79245d211a457ff44b5e7b96ddbcb

As it turns out, WifiStateMachine isn't the only thing that
wants to change its logging behavior depending on
Build.TYPE.

1) Add an abstraction of Build.TYPE, so that we can
   a) check build type in a common way, and
   b) test the logging behavior across build types
2) Migrate WifiStateMachine to use this new abstraction.

BUG=28822027
TEST=unit tests

Change-Id: I6f4cb00ccb5f3a4b9a5b67e69f8365aab7fd429c

* changes:
  WifiConfigManager: clean up makeChannelList()
  WifiStateMachine: clean up dead code
  WifiConnectivityManagerTest: exponential backoff periodic scan
  Restrict connectivty periodic scan

Before this, when WifiScanningService processed single scan results from
supplicant that only contained results that did not match a request the
filter code would return null indicating that the results should not be
delivered, causing a NPE later. This change updates the meaning of the
bucket index that the single scan code provides to the filtering code to
instead force it to always include scan results.

Change-Id: I65cd57b14abacec3f407991188c570601d05ac77
Fixes: 28794598