1061012 - [2.1][Flame] Recent Nightlies use too much battery power because /system/b2g/b2g and /system/bin/mediaserver seem to take several % cpu all the time

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Performance, defect)

Description

[Olli Pettay [:smaug][[e-mail]]]

[Blocking Requested - why for this release]:
This is a recent major regression, and others have seen this too.

STR, boot to recent nightly build, connect using adb and run top.

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

I can see also ksmd taking several % cpu time.

Comment 2

[:gerard-majax]

(In reply to Olli Pettay [:smaug] from comment #1)
> I can see also ksmd taking several % cpu time.

I'm seeing this on several other devices, namely Desire Z, Keon, Nexus S. Others are seeing it on Hamachi as well.

Comment 3

[Paul Adenot (:padenot)]

Probably my fault (for the mediaserver thing, probably not the rest), I'll send a patch.

Comment 4

[Paul Adenot (:padenot)]

Actually, the fix will come from Gaia.

Comment 5

[Etienne Segonzac (:etienne)]

Attachment: Gaia PR

Alive: I'm pretty sure this was a legacy from the "clipping alternative" that we forgot to remove, let me know if there's another reason why we weren't setVisible(false)'ing the AttentionWindow when closed.

Rik: Thanks for pushing me to fix both issues and not just the WindowManager one, here's a review :)

Comment 6

[(inactive after 6/18) Alive Kuo [:alive]]

Comment on attachment 8482305 [details] [review]
Gaia PR

When attention window will full size enters closed state in appTransitionController, we should not send it to background soon. Insteadly, attentionToaster will perform the toaster transition, and after attentionToaster enters closed state, we will send it to background. This is a cowork of two finite state machines. If we send it to background in 1st finite state machine we will see empty and static toaster in 2nd finite state machine.

The problem should be fixed by sending callscreenWindow to background after creating it. Or we could setVisible(false) in this.hide().

Comment 7

[Etienne Segonzac (:etienne)]

Comment on attachment 8482305 [details] [review]
Gaia PR

Got it, it was indeed an issue with the reloadWindow().
Patch updates!

Comment 8

[(inactive after 6/18) Alive Kuo [:alive]]

Comment on attachment 8482305 [details] [review]
Gaia PR

r=me

Comment 9

[Etienne Segonzac (:etienne)]

https://github.com/mozilla-b2g/gaia/commit/23b367e74b2de316295e49fb012984ba689e3d5b

Thanks!

Comment 10

[Joshua Mitchell (Inactive)]

no need for a regression window if this is already patched and fixed - removing tag.

Comment 11

[Ryan VanderMeulen [:RyanVM]]

Looks like this had a favorable impact on B2G emulator xpcshell runtimes, which were another early indication of problems after the MSG refactor landed.

Comment 12

[Ryan VanderMeulen [:RyanVM]]

Please request Gaia v2.1 approval on this as soon as you get a chance. All patches need approval to land now regardless of blocking status.

Comment 13

[Etienne Segonzac (:etienne)]

Comment on attachment 8482305 [details] [review]
Gaia PR

[Approval Request Comment]
[Bug caused by] (feature/regressing bug #): an unlucky mix of "the mediaserver thing" and bug 927862

[User impact] if declined: big battery life regression

[Testing completed]: Attention Window uses (calls, alarm) and TonePlayer uses in the dialer (keypad, callscreen keypad, end tone)

[Risk to taking this patch] (and alternatives if risky): low, and it's actually adding test coverage.

[String changes made]: none

Comment 14

[Ryan VanderMeulen [:RyanVM]]

v2.1: https://github.com/mozilla-b2g/gaia/commit/01bb4c77f4a3672b9f8d939c125f5724840590ce

Comment 15

[bhavana bajaj [:bajaj]]

Switching the 2.1?->2.1+, on these fixed bugs as these are regression.

Nothing to land here, its just flag-cleanup of 2.1? list. Please Ni me if there is confusion/disagreement.

Comment 17

[Sue]

Hi reporter,
There are CPU infos(use adb shell top -m 5) of my devices, could you help to check if this issue was fixed? Thanks!

Flame 2.1:
User 1%, System 3%, IOW 0%, IRQ 0%
User 5 + Nice 0 + Sys 11 + Idle 287 + IOW 0 + IRQ 0 + SIRQ 0 = 303

  PID PR CPU% S  #THR     VSS     RSS PCY UID      Name
 2469  0   2% R     1   1308K    532K     root     top
  204  0   1% S    57 214368K  68676K     root     /system/b2g/b2g
   98  0   0% S     1      0K      0K     root     ksmd
  994  0   0% S     6   7288K    492K     root     /system/bin/mpdecision
  908  0   0% S    17  87228K  29784K     u0_a908  /system/b2g/b2g

Flame 2.2:
User 0%, System 3%, IOW 0%, IRQ 0%
User 2 + Nice 0 + Sys 10 + Idle 294 + IOW 0 + IRQ 0 + SIRQ 0 = 306

  PID PR CPU% S  #THR     VSS     RSS PCY UID      Name
 5199  0   2% R     1   1308K    532K     root     top
   98  0   0% S     1      0K      0K     root     ksmd
  207  0   0% S    77 265720K  67904K     root     /system/b2g/b2g
    3  0   0% S     1      0K      0K     root     ksoftirqd/0
    8  0   0% S     1      0K      0K     root     migration/0

Comment 18

[Olli Pettay [:smaug][[e-mail]]]

I haven't seen this recently.