914728 - Make ProcessPriorities/OomScoreAdjusts and KillUnderMB two different sets and enlarge OomScoreAdjusts

Status: RESOLVED FIXED

(Firefox OS Graveyard :: General, defect)

Comment 1

[Gabriele Svelto [:gsvelto] (PTO)]

I had prepared a patch for bug 886278 that made the OomScoreAdjusts relative to the amount of total memory instead of absolute values. If there's interest in it I can post it here (it was a WIP).

Comment 2

[Alan Huang]

(In reply to Gabriele Svelto [:gsvelto] from comment #1)
> I had prepared a patch for bug 886278 that made the OomScoreAdjusts relative
> to the amount of total memory instead of absolute values. If there's
> interest in it I can post it here (it was a WIP).

Oh, I remember that from leo. We indeed need bug 886278. I just came from partner side last week. Currently our parameters in b2g.js cause their LMK cannot prevent oom-killer raises too early. Because we don't have bug 886278 landed yet, so I asked partners to adjust them with chip vendor first :)

BTW, what I want to do here is enlarge OomScoreAdjusts instead of killUnderMBs, see bug 914541. So I think this would effect you work to adjust killUnderMBs on bug 886278.

Kan-ru and I are in Oslo this week but we just left our hosting partner side. We could discuss bug 822325, 886278, 914541 and this tomorrow :)

Comment 3

[Alan Huang]

Attachment: Don't crash on KillerUnderMB not found, enlarge OomScoreAdjusts

Hi Gabriele,
  Can you help to review this? Thanks.

Before:
  oom_adj min_free
        0  4096 KB
        1  5120 KB
        2  6144 KB
        3  7168 KB
        4  8192 KB
        6 20480 KB
After:
  oom_adj min_free
        0  4096 KB
        1  5120 KB
        2  6144 KB
        6  7168 KB
        8  8192 KB
       10 20480 KB

I also tested this by adding addtional OomScoreAdjust without KillUnderMB for bug 914541, which is one of our purpose here.

Comment 4

[Gabriele Svelto [:gsvelto] (PTO)]

Comment on attachment 802580 [details] [diff] [review]
Don't crash on KillerUnderMB not found, enlarge OomScoreAdjusts

Review of attachment 802580 [details] [diff] [review]:
-----------------------------------------------------------------

r=me with the small nit below.

::: hal/gonk/GonkHal.cpp
@@ +1088,5 @@
>      // The LMK in kernel silently malfunctions if we assign the parameters
>      // in non-increasing order, so we add this assertion here. See bug 887192.
>      MOZ_ASSERT(oomScoreAdj > lowerBoundOfNextOomScoreAdj);
>      MOZ_ASSERT(killUnderMB > lowerBoundOfNextKillUnderMB);
> +    

nit: trailing whitespace

Comment 5

[Alan Huang]

Attachment: Don't crash on KillerUnderMB not found, enlarge OomScoreAdjusts, v2

Remove trailing whitespaces. r=gsvelto

Comment 6

[Alan Huang]

Try in progress:
try: -p ics_armv7a_gecko,panda,unagi

https://tbpl.mozilla.org/?tree=Try&rev=b30f5df0d7a8

Comment 7

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/b2g-inbound/rev/a96f92dc0088

Comment 8

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/a96f92dc0088