1639275 - Crash in mozilla::widget::Base64UtilsSupport::Encode

Status: RESOLVED FIXED

(GeckoView :: General, defect, P1)

Description

[[ex-Mozilla] Agi Sferro | :agi]

This bug is for crash report bp-4fda00b7-8785-4c54-83a0-0b88f0200519.

A reddit user is seeing frequent crashes with this signature.

https://www.reddit.com/r/firefox/comments/gmqkcp/random_crash_error_from_firefox_nightly_after/fr5gk3u/

Looks like something in jni, aklotz any ideas?

Top 10 frames of crashing thread:

0 libc.so libc.so@0x83134 
1 libc.so libc.so@0x83104 
2 libart.so libart.so@0x4b3c08 
3 libbase.so libbase.so@0xc5b4 
4 libart.so libart.so@0x374a94 
5 libart.so libart.so@0x3c1f40 
6 libxul.so mozilla::jni::ArrayRefBase<_jbyteArray*, signed char>::GetElements const widget/android/jni/Refs.h:819
7 libxul.so _jstring* mozilla::jni::NativeStub<mozilla::java::Base64Utils::Encode_t, mozilla::widget::Base64UtilsSupport, mozilla::jni::Args<mozilla::jni::Ref<mozilla::jni::TypedObject<_jbyteArray*>, _jbyteArray*> const&> >::Wrap<&mozilla::widget::Base64UtilsSupport::Encode> widget/android/jni/Natives.h:690
8 base.odex base.odex@0x1964ec 
9 base.odex base.odex@0x1964ec 

Comment 1

[[ex-Mozilla] Agi Sferro | :agi]

I'm wondering if we have more of these buried under the libc.so@<addr> signature.

Comment 2

[(No longer employed by Mozilla) Aaron Klotz]

Some kind of failure retrieving the length of an array, but that's the best I can do without analyzing the raw dump (which is a non-trivial process).

Comment 3

[[ex-Mozilla] Agi Sferro | :agi]

Thanks, I skimmed our crashes, they're all in mozilla::widget::Base64UtilsSupport under various libc.so signature.

Comment 4

[[ex-Mozilla] Agi Sferro | :agi]

Looks like there's been an uptick recently. I wonder if it runs some java stacks after the jni call.

Comment 5

[[ex-Mozilla] Agi Sferro | :agi]

Looks like we use this mostly in WebPush https://searchfox.org/mozilla-central/search?q=Base64Utils.encode&path=

Comment 6

[[ex-Mozilla] Agi Sferro | :agi]

snorp, ni in case you have ideas.

Comment 7

[James Willcox (:snorp) ([e-mail]) (he/him)]

It looks like we're just missing a null check in the native side.

Comment 8

[James Willcox (:snorp) ([e-mail]) (he/him)]

Attachment: Bug 1639275 - Add some null checking to Base64Utils

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/f121c6c8aeb0
Add some null checking to Base64Utils r=geckoview-reviewers,aklotz

Comment 10

[Bogdan Tara[:bogdan_tara | bogdant]]

https://hg.mozilla.org/mozilla-central/rev/f121c6c8aeb0