1015602 - division by zero crash or bin-unused and heap-chunks too huge (jemalloc3)

Status: RESOLVED FIXED

(Core :: Memory Allocator, defect)

Description

[Jan Beich]

Attachment: workaround

Bug 1006769 Part 1 added memory reporters for |chunksize| and |bin_unused| (new) fields of jemalloc_stats_t. Calculating the former is trivial but the latter may need upstream help or poking with dalloc bins stats.

Explicit Allocations

32,885.73 MB (100.0%) -- explicit
├──32,803.28 MB (99.75%) -- heap-overhead
│  ├──32,801.46 MB (99.74%) ── bin-unused
│  └───────1.82 MB (00.01%) ++ (3 tiny)
└──────82.45 MB (00.25%) ++ (22 tiny)

Other Measurements 

 57.72 MB ── heap-allocated
3,366,480 ── heap-chunks
  0.00 MB ── heap-chunksize
 59.54 MB ── heap-committed
 61.00 MB ── heap-mapped
    3.15% ── heap-overhead-ratio

Comment 1

[Eric Rahm [:erahm]]

Comment on attachment 8428252 [details] [diff] [review]
workaround

Review of attachment 8428252 [details] [diff] [review]:
-----------------------------------------------------------------

lgtm!

Comment 2

[Jan Beich]

I've wanted to update the comment but there's probably not much point. Except for file/function_name it already carries enough information. Adding a wrong hint is worse.

  // We could get this value out of base.c::base_pages, but that really should
  // be an upstream change, so don't worry about it for now.

Comment 3

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/57dd204266ba

Comment 4

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/57dd204266ba

Comment 5

[James Zhang (Spreadtrum)]

Do we need land it on v1.3t? It blocks bug 1006769.

Comment 6

[[:fabrice] Fabrice Desré]

 https://hg.mozilla.org/releases/mozilla-b2g28_v1_3t/rev/fd539ef9f590

Comment 7

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

1.3T already fixed. marking 1.3T+