1194532 - [Aries] Phone is hot after being on for ~2 days, mdnsd is at 45% CPU constantly

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Performance, defect)

Description

[James Hobin [:hobinjk]]

Attachment: sample-top-output.txt

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:43.0) Gecko/20100101 Firefox/43.0
Build ID: 20150813030208

Steps to reproduce:

Used phone for two days, mostly for calls, texting, and internet browsing


Actual results:

Phone began to be constantly hot, with /system/bin/mdnsd listed as the likely culprit with its constant use of 45-50% CPU


Expected results:

mdnsd should not take over the phone's processor, more generally the phone should be a normal temperature.

Comment 1

[ralf tauscher]

I can confirm this on my aries running ver2.5 

git commit gaia
2015-08-14 18:55 

build indentifier
20150811022743

Comment 2

[ralf tauscher]

its not there right after reboot. I don't know how to reproduce it.

does anybody know, which "app" invokes mdnsd?

Comment 3

[Cervantes Yu [:cyu] [:cervantes]]

I think you encounter https://bugzilla.mozilla.org/show_bug.cgi?id=1175387#c1
We need to fix the buggy mdnsd.

Comment 4

[Cervantes Yu [:cyu] [:cervantes]]

(In reply to Cervantes Yu [:cyu] [:cervantes] from comment #3)
> I think you encounter https://bugzilla.mozilla.org/show_bug.cgi?id=1175387#c1
We saw this on one aries-kk device and it's the same problem.
> We need to fix the buggy mdnsd.

Comment 5

[Cervantes Yu [:cyu] [:cervantes]]

I think I can place a check to the FD before it's added to the FDSet for select(). I'll work on a quick fix for this.

Comment 6

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Not sure if this is related: bug 1172383

Comment 7

[Cervantes Yu [:cyu] [:cervantes]]

On the phone that observes a CPU spin we observe that mdnsd leaks file descriptors. The leak results from:

* assert(close(socket)) that is omitted on a release build, and
* the lack of fclose() to a open file.

The leaks are fixed on mdnsresponder for lollipop. We need to upgrade mdnsresponder to the lollipop branch.

Comment 8

[Cervantes Yu [:cyu] [:cervantes]]

I'll dogfood the lollipop version of mdnsd to see if it fixes the CPU spin. If it's fixed then I'll update the repo manifest to upgrade mdnsresponder.

Comment 9

[Cervantes Yu [:cyu] [:cervantes]]

mdnsd works well so far. The device is still coll given that I switched wifi networks and turned on/off tethering, which trigger the FD leak.

Comment 10

[Cervantes Yu [:cyu] [:cervantes]]

Attachment: Upgrade mdnsd to fix several FD leaks that causes CPU spin

Comment 11

[Michael Wu [:mwu]]

Comment on attachment 8653242 [details] [review]
Upgrade mdnsd to fix several FD leaks that causes CPU spin

I wonder if Android suffers from the same problem.　Does it shut off mdnsd?

Comment 12

[Cervantes Yu [:cyu] [:cervantes]]

(In reply to Michael Wu [:mwu] from comment #11)
> Comment on attachment 8653242 [details] [review]
> Upgrade mdnsd to fix several FD leaks that causes CPU spin
> 
> I wonder if Android suffers from the same problem.　Does it shut off mdnsd?

Yes, at least xiaomi suffers the problem. openwrt also has this problem. A search of "mdnsd battery drain" returns a lot of reports. I don't think it shut off mdnsd.

Comment 13

[Ryan VanderMeulen [:RyanVM]]

Master: https://github.com/mozilla-b2g/b2g-manifest/commit/c84b4f58c6949cf3d4a2e6a354027282c1826679

Comment 14

[Cervantes Yu [:cyu] [:cervantes]]

Naoki, will this change be installed with OTA, or we need to find another way have it installed to the dogfood devides?

Comment 15

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

I don't think this change will be in OTA.
We need a way to FOTA.

Comment 16

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

FOTA pushed to internal users.