New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 728769 link

Starred by 2 users

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 3
Type: Bug



Sign in to add a comment

autotest: wifi_client: don't force-disable Wifi power-save mode

Project Member Reported by briannorris@chromium.org, Jun 1 2017

Issue description

Currently, all Wifi-related autotests will disable Wifi power-save mode before running. See:

server/cros/network/wifi_client.py WifiClient::__init__

        # Turn off powersave mode by default.
        self.powersave_switch(False)

This seems to have been done in 2013 here:

https://gerrit.chromium.org/gerrit/46537
commit 6866e09a5c03ad98eca832f30f8277b3bddb4eb6
Author: Christopher Wiley <wiley@chromium.org>
Date:   Tue Mar 26 12:41:56 2013 -0700

   autotest: Cleanup bad powersave state before network_WiFiMatFunc/007CheckPS


It claims that "we fail future test runs because power save remains on".

Previously, this might have been reasonable to do, as we left power-save disabled when plugged into AC, and only enabled power-save when on battery. Thus, there was some production use case of "power-save off", and therefore, measuring (for example) performance with power-save disabled is an understandable choice.

However, as of  bug 689648 , we unconditionally enable Wifi power-save mode for all devices. IMO, this removes any argument that we should qualify our devices with power-save disabled -- if power-save is a source of bugs / unacceptable performance regressions / test failures, then it seems likely such bugs will hit real users. We're not doing anyone a favor by disabling this mode.

This bug should track the effort to re-enable power-save in our test environment. If there are any tests that fail today only when we enable power-save mode, we should investigate and fix them.
 
Owner: satyat@chromium.org
Status: Assigned (was: Available)
Satya, are you interested in this? It shouldn't be too complicated of a CL. Probably want to run a device or two through the test suites with this changed, and make sure harpreet@ is aware, for any potential test regressions.

Comment 2 by satyat@google.com, Aug 11 2017

Power save mode is enabled in CL: https://chromium-review.googlesource.com/c/604753/1.

Tests that break, and their respective reports are 
1) network_WiFi_GTK: crbug/740198
2) network_WiFi_DisconnectReason.ap_send_chan_switch: crbug/754909
Owner: ----
Status: Available (was: Assigned)
With the latest patch set + workarounds, this fails reliably on Kevin + Whirlwind, in network_WiFi_Regulatory:

/tmp/test_that_results_hnRjsQ/results-1-network_WiFi_Regulatory/network_WiFi_Regulatory   FAIL: Client never lost connectivity

There are probably other failures.

I think Satya noted some potential problems with that test already, so we should investigate whether we need to fixup or deprecate that test too.
I also see failures here:

/tmp/test_that_results_wR4whO/results-1-network_WiFi_BgscanBackoff.wifi_bgscan_backoff/network_WiFi_BgscanBackoff [  FAILED  ]
/tmp/test_that_results_wR4whO/results-1-network_WiFi_BgscanBackoff.wifi_bgscan_backoff/network_WiFi_BgscanBackoff   FAIL: Significant difference in rtt due to bgscan: 531.6 > 8.8 + 200

On first blush, BgscanBackoff sounds like it could be a real problem with the DUT. Is it reasonable to get these results with background scan:

08/15 17:20:37.114 DEBUG|             utils:0280| [stdout] 1000 packets transmitted, 999 received, 0% packet loss, time 100548ms
08/15 17:20:37.114 DEBUG|             utils:0280| [stdout] rtt min/avg/max/mdev = 1.629/12.820/531.551/33.750 ms, pipe 6

vs. these without:

08/15 17:21:09.250 DEBUG|             utils:0280| [stdout] 100 packets transmitted, 100 received, 0% packet loss, time 9942ms
08/15 17:21:09.250 DEBUG|             utils:0280| [stdout] rtt min/avg/max/mdev = 2.620/8.803/10.893/1.156 ms

?

The test seems repeatable too.
Filed this for Kevin/Marvell 8997 issues with network_WiFi_BgscanBackoff:
https://issuetracker.google.com/64771701
Project Member

Comment 6 by sheriffbot@chromium.org, Aug 20

Labels: Hotlist-Recharge-Cold
Status: Untriaged (was: Available)
This issue has been Available for over a year. If it's no longer important or seems unlikely to be fixed, please consider closing it out. If it is important, please re-triage the issue.

Sorry for the inconvenience if the bug really should have been left as Available.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
Cc: -pstew@chromium.org -cernekee@chromium.org
Components: -Test Infra>Client>ChromeOS>Test
Labels: -Pri-1 Pri-3
Owner: briannorris@chromium.org
Status: Assigned (was: Untriaged)

Sign in to add a comment