1981751 - [Android 5.1 / API 22 with ARM v7] Crash in [@ FileID::ElfClassBuildIDNoteIdentifier]

Status: RESOLVED FIXED

(Core :: Gecko Profiler, defect, P3)

Description

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

6 crash reports for Firefox for Android 143.0a1 with ARM v6, oldest build ID is 20250725093757

The crash stacks are corrupted.

Crash report: https://crash-stats.mozilla.org/report/index/d7bb6a9b-33d6-4394-b30b-b94a30250807

Reason:

SIGSEGV / SEGV_MAPERR

Top 8 frames:

0  libmozglue.so  FileID::ElfClassBuildIDNoteIdentifier(void const*, unsigned int, std::__ndk1:...  mozglue/baseprofiler/core/shared-libraries-linux.cc:369
0  libmozglue.so  dl_iterate_callback(dl_phdr_info*, unsigned int, void*)  mozglue/baseprofiler/core/shared-libraries-linux.cc:761
1  linker  linker@0x1385
2  libsentry.so  libsentry.so@0x32
3  libmozglue.so  SharedLibraryInfo::GetInfoForSelf()  mozglue/baseprofiler/core/shared-libraries-linux.cc:866
4  libmozglue.so  <.plt ELF section in libmozglue.so>
5  libmozglue.so  <.plt ELF section in libmozglue.so>
6  eglsubAndroid.so  eglsubAndroid.so@0x800e

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Timing and stack look maybe related to bug 1867280?

Comment 2

[Ted Campbell [:tcampbell]]

I guess we are dropping support for API 22 later this year, so eventually this won't be an issue. Maybe Nazim does have some quick ideas here though

Comment 3

[Nazım Can Altınova [:canova]]

Huh, that's weird. This is definitely a regression from bug 1773313. I think it's better to move this to the gecko profiler component.

I don't really have a quick fix to the crash, it normally should happen. It seems like some parts of the note section are not mapped correctly?

OTOH, you're right about API 22. Hopefully that will not be an issue soon.

Comment 4

[BugBot [:suhaib / :marco]]

The product::component has been changed since the backlog priority was decided, so we're resetting it.
For more information, please visit BugBot documentation.

Comment 5

[BugBot [:suhaib / :marco]]

Set release status flags based on info from the regressing bug 1773313

Comment 6

[Markus Stange [:mstange]]

I think this is caused by the custom linker passing null as the base address. For example we return null from SystemElf::GetBase

Comment 7

[Nazım Can Altınova [:canova]]

Yeah, it seems like you're right, thanks. I'm still unsure when we use SystemElf::GetBase though and if we should fix it too? But at least I have enough information to fix this issue.

Comment 8

[Nazım Can Altınova [:canova]]

Oh, it looks like we have a __wrap_dl_iterate_phdr here that we only use when the Android version is below 23. Apparently we are not using the custom linker when it's version 23 and above (bug 1291377). That explains why we are seeing this in version 22.

Comment 9

[Nazım Can Altınova [:canova]]

Attachment: Bug 1981751 - Fix crash in dl_iterate_callback when dlpi_addr is null r?mstange

Skip entries with null base addresses in dl_iterate_callback to prevent
segmentation faults when SystemElf::GetBase() returns nullptr. This
prevents crashes on Android API 22 ARM where the custom linker can
pass null as the base address.

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/5030bcac89e3
https://hg.mozilla.org/integration/autoland/rev/94cb1c6dab4c
Fix crash in dl_iterate_callback when dlpi_addr is null r=mstange,profiler-reviewers

Comment 11

[Norisz Fay [:noriszfay]]

https://hg.mozilla.org/mozilla-central/rev/94cb1c6dab4c