995597 - [tarako]When b2g killed by LMK/OOM, b2g need quick restart without boot animation

Status: RESOLVED DUPLICATE of bug 999953

(Firefox OS Graveyard :: General, defect)

Description

[James Zhang (Spreadtrum)]

When b2g killed by OOM, b2g need quick restart without boot animation.
Our QA summit a lot of critical issue because it seems system restart.

Comment 1

[[:fabrice] Fabrice Desré]

We could suppress the boot animation, but that will not make us start faster. This animation runs off the main thread: https://mxr.mozilla.org/mozilla-central/source/widget/gonk/libdisplay/BootAnimation.cpp#575

Comment 2

[[:fabrice] Fabrice Desré]

And by supressing the boot animation, I mean doing that only on "restarts".

Comment 3

[thomas tsai]

Is b2g killed by OOM a normal use case? Or does killing b2g by OOM only happen in Monkey test?

Comment 4

[James Zhang (Spreadtrum)]

(In reply to thomas tsai from comment #3)
> Is b2g killed by OOM a normal use case? Or does killing b2g by OOM only
> happen in Monkey test?

It not only happen in monkey test. Sometimes QA found it restart and submit critical bug in my bugzilla.

Comment 5

[James Zhang (Spreadtrum)]

see my bugzilla
http://bugzilla.spreadtrum.com/bugzilla/show_bug.cgi?id=299310

Comment 6

[James Zhang (Spreadtrum)]

LMK and OOM also can kill b2g, it likes system crash and restart.
We should protect b2g and nuwa process.
Why not kill foreground apps and preallocated process?

04-12 10:24:09.265 <4>0[  643.664651] lowmem_shrink select 83 (b2g), adj 0, size 28068, to kill
04-12 10:24:09.265 <4>0[  643.664745] lowmem_shrink send sigkill to 83 (b2g), adj 0, size 28068

Comment 7

[James Zhang (Spreadtrum)]

Attachment: slog_20140412103121_sp6821a_userdebug.zip

b2g killed by LMK log

Comment 8

[James Zhang (Spreadtrum)]

Can I change kernel CONFIG_SLAB to CONFIG_SLOB for shrink memory?

Comment 9

[thomas tsai]

Hi James: what's the impact between slab and slob? how confident in your kernel team can evaluate this change?

Comment 10

[thomas tsai]

We have to fix why b2g is killed. At that moment, all low priority processes should be killed and the memory is still not enough, so b2g is finally killed. We have to confirm this scenario.
And the project time is getting short, is it too risky to change kernel's slab and slob?

Comment 11

[Keven Kuo [:kkuo]]

+1

(In reply to thomas tsai from comment #10)
> We have to fix why b2g is killed. At that moment, all low priority processes
> should be killed and the memory is still not enough, so b2g is finally
> killed. We have to confirm this scenario.
> And the project time is getting short, is it too risky to change kernel's
> slab and slob?

Comment 12

[Joe Cheng [:jcheng] (please needinfo)]

triage: b2g should not be killed, lets focus on why b2g is killed , not blocking

Comment 13

[John Hammink]

Duping this to 999953, as fixing the other fixes this.   I'm watching the other bug as well and plan to  test both STRs