1718356 - Crash in [@ OOM | large | mozalloc_abort | alloc::raw_vec::{{impl}}::reserve::do_reserve_and_handle<T>]

Status: RESOLVED WORKSFORME

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

Description

[Andrew McCreight [:mccr8]]

Maybe Fission related. (DOMFissionEnabled=1)

Crash report: https://crash-stats.mozilla.org/report/index/f36b21b4-d5fc-43c9-a011-78a7c0210619

MOZ_CRASH Reason: MOZ_CRASH()

Top 10 frames of crashing thread:

0 mozglue.dll mozalloc_abort memory/mozalloc/mozalloc_abort.cpp:33
1 mozglue.dll mozalloc_handle_oom memory/mozalloc/mozalloc_oom.cpp:51
2 xul.dll gkrust_shared::oom_hook::hook toolkit/library/rust/shared/lib.rs:134
3 xul.dll std::alloc::rust_oom ../53cb7b09b00cbea8754ffb78e7e3cb521cb8af4b//library/std/src/alloc.rs:331
4 xul.dll alloc::alloc::__alloc_error_handler::__rg_oom ../53cb7b09b00cbea8754ffb78e7e3cb521cb8af4b//library/alloc/src/alloc.rs:397
5 xul.dll _rust_alloc_error_handler 
6 xul.dll alloc::alloc::handle_alloc_error ../53cb7b09b00cbea8754ffb78e7e3cb521cb8af4b//library/alloc/src/alloc.rs:366
7 xul.dll alloc::raw_vec::{{impl}}::reserve::do_reserve_and_handle<webrender::hit_test::HitTestClipNode, alloc::alloc::Global> ../53cb7b09b00cbea8754ffb78e7e3cb521cb8af4b/library/alloc/src/raw_vec.rs:325
8 xul.dll webrender::hit_test::add_clips gfx/wr/webrender/src/hit_test.rs:519
9 xul.dll webrender::scene_building::SceneBuilder::build_item gfx/wr/webrender/src/scene_building.rs:1182

Looks like a WebRender OOM that showed up in the 20210618092056 build. I'm assuming that it is probably a signature shift, but who knows maybe it isn't.

Comment 1

[Chris Peterson [:cpeterson]]

Looks like a WebRender OOM that showed up in the 20210618092056 build. I'm assuming that it is probably a signature shift, but who knows maybe it isn't.

Here's the changelog for build 20210618092056:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=2f69dc403886dd1c721cec815b00ec5bfca4b393&tochange=d70c3d846c395717c429cd3f3433457f5f92dced

There were four WebRender changes in that build:

Bug 1715515 - Part 1. Switch Windows 32 and mingwclang tests to WebRender.
Bug 1459526 - Handle full-range video in Webrender.
Bug 1717028 - Fix webrender render target debug view.
Bug 1716971 - Enable webrender on Android on NVIDIA Tegra GPUs.

All of these crashes are on Windows, so the Android bug is probably not relevant. All of these crashes are on amd64, so the Windows 32-bit test bug is probably not relevant.

There is also a bug related to graphics memory pressure, which might be related since this bug is about WebRender OOMs:

Bug 1716579 - Minimize memory usage of windows that haven't been rendered for a long time. r=gfx-reviewers,bradwerth

Comment 2

[Chris Peterson [:cpeterson]]

This doesn't look like a Fission crash even though the crash report in comment 0 has DOMFissionEnabled=1. Adding [not-a-fission-bug] whiteboard tag to hide this bug from Fission bug triage queries.

Comment 3

[Nicolas Silva [:nical]]

There seem to have been a recent uptick in the volume of OOM crashes while reserving memory for vectors when building clips on the SceneBuilder thread (for example https://crash-stats.mozilla.org/report/index/c86acec7-2860-46be-a83c-097960211026#tab-details ).
Glenn could that be related to your recent changes?

Comment 4

[Glenn Watson [:gw]]

It's possible, but I think unlikely.

None of the changes I've made recently should be allocating more memory than previously, so unless there is a bug in one of those patches or something else unexpected going on, I wouldn't expect it to cause that.

Comment 5

[Bob Hood [:bhood]]

If this behavior resurfaces, please update this report or open a new one.