Open Bug 1647269 Opened 4 years ago Updated 2 years ago

mozalloc_abort MOZ_CRASH() on Android

Categories

(Core :: Graphics: ImageLib, defect, P3)

Unspecified
Android
defect

Tracking

()

Performance Impact none
Tracking Status
firefox79 --- affected

People

(Reporter: petru, Unassigned)

Details

(Keywords: crash)

Crash Data

Seen on Fenix - https://github.com/mozilla-mobile/fenix/issues/10385 while accessing a certain website - https://musicgenretree.org/chart.html.

It seems like a relatively heavy website but with more reports for the OOM | small crash signature for other various websites I think there may actually be a problem in our mozalloc.

Crash Signature: https://crash-stats.mozilla.org/signature/?product=Fenix&signature=OOM%20%7C%20small → [OOM | small]
Crash Signature: [OOM | small] → [@ OOM | small]
Keywords: crash

This is unlikely to be due to a problem with jemalloc. It is more likely that there's a memory leak somewhere.

The usual first step for this kind of issue would be to get a report from about:memory, but I'm not sure if you can do that on Fenix and it probably wouldn't work in a case like this where it crashes very quickly. Maybe some kind of profiler could get information about what the allocation stacks are?

Component: Memory Allocator → Performance

On desktop, that website uses about 3MB of DOM/JS stuff and about 100MB of image stuff:

123.97 MB (100.0%) -- explicit
├───99.36 MB (80.15%) -- images
│ ├──99.36 MB (80.15%) -- content/raster/used
│ │ ├──99.35 MB (80.14%) -- progress=10f/image(2500x10000, https://www.musicgenretree.org/1111%20Essential%20Recordings%20of%20Music%20(LowRes).png)
│ │ │ ├──95.37 MB (76.93%) -- unlocked/types=2000/surface(2500x10000)
│ │ │ │ ├──95.37 MB (76.93%) ── decoded-nonheap

So, maybe it is doing something with images that is causing issues.

The crash report in the linked GitHub issue is bp-bab030a4-6da6-4fb4-b49a-b0fe30200622. The OOM allocation size is 262,144 bytes, which is a bit large, but not too huge. There's no other memory related information in the crash report as far as I can see.

Hey aosmond, is this the sort of thing that Imagelib should track? It looks like we're OOM'ing after decoding some pretty large images?

Flags: needinfo?(aosmond)
Whiteboard: [qf-]
Webcompat Priority: --- → ?

Please file separate bugs for individual websites that are experiencing low memory crashes. They aren't necessarily the same underlying issue, and combining multiple things into a single bug can be confusing.

The linked github issue talks about scrolling not being smooth on a particular website, which is different from an out of memory crash. Please file a separate bug for that.

Webcompat Priority: ? → ---
Flags: needinfo?(aosmond)
Performance Impact: --- → -
Whiteboard: [qf-]
Component: Performance → ImageLib
Severity: -- → S3
Priority: -- → P3
You need to log in before you can comment on or make changes to this bug.