[mono] Preserve signal handlers during crash chaining by jpnurmi · Pull Request #125835 · dotnet/runtime

jpnurmi · 2026-03-20T14:59:47Z

This change is part of the effort to fix native crash reporting for .NET Android apps (getsentry/sentry-dotnet#3954). We wish to install Sentry's signal handlers before Mono to capture native crashes. However, mono_handle_native_crash unconditionally resets SIGABRT, SIGILL, SIGCHLD, and SIGQUIT to SIG_DFL, even when crash_chaining is enabled, killing the process before the chained handler can run.

Expose static remove_signal_handler as mono_runtime_posix_remove_signal_handler and use it in mono_handle_native_crash to restore pre-Mono signal handlers instead of resetting to SIG_DFL. When signal chaining is disabled, there are no saved handlers, so it falls back to SIG_DFL, same as before.

Includes an Android functional test (CrashChaining) that verifies signal handlers are preserved during crash chaining.

When mono_set_crash_chaining(true) is enabled, mono_handle_native_crash unconditionally resets SIGABRT, SIGILL, SIGCHLD, and SIGQUIT to SIG_DFL. This is problematic on Android where multiple threads may trigger signals concurrently — for example, FORTIFY detecting a destroyed mutex on an HWUI thread raises SIGABRT, but because the handler was reset to SIG_DFL, it kills the process before Mono can chain the original crash to the previous handler (e.g. sentry-native). Guard the signal handler reset with !mono_do_crash_chaining so that when crash chaining is enabled, Mono's handlers remain installed during native crash processing. This allows the crash to be fully handled and chained to any previously installed handlers before the process terminates. The fix has no effect when crash_chaining is disabled (the default), as the existing SIG_DFL reset behavior is preserved. Includes an Android functional test (CrashChaining) that: - Installs a SIGSEGV handler before mono_jit_init (gated by TEST_CRASH_CHAINING env var) to simulate a pre-Mono crash handler - Triggers a native SIGSEGV that Mono chains to the pre-Mono handler - Verifies SIGABRT was not reset to SIG_DFL during crash processing - Without the fix, the test process is killed by SIGABRT (test crashes) - With the fix, the test returns exit code 42 (pass)

dotnet-policy-service · 2026-03-20T15:01:05Z

Tagging subscribers to this area: @akoeplinger, @matouskozak, @simonrozsival
See info in area-owners.md if you want to be subscribed.

Copilot

Pull request overview

This PR fixes Mono’s native-crash handling behavior when crash chaining is enabled so that Mono does not reset key signal handlers to SIG_DFL during mono_handle_native_crash, which can otherwise cause secondary signals (notably on Android) to terminate the process before crash chaining completes.

Changes:

Guard signal-handler resets in mono_handle_native_crash behind !mono_do_crash_chaining.
Add an Android functional test that installs a pre-Mono SIGSEGV handler, triggers a native crash, and validates SIGABRT was not reset to SIG_DFL.
Wire the test via the Android app template using an environment-variable gate (TEST_CRASH_CHAINING).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
src/mono/mono/mini/mini-exceptions.c	Avoids resetting signal handlers to defaults when crash chaining is enabled.
src/tasks/AndroidAppBuilder/Templates/monodroid.c	Adds a crash-chaining functional test path (env-gated) and native test helpers.
src/tests/FunctionalTests/Android/Device_Emulator/CrashChaining/Program.cs	Managed entrypoint invoking the native test and mapping success to exit code 42.
src/tests/FunctionalTests/Android/Device_Emulator/CrashChaining/Android.Device_Emulator.CrashChaining.Test.csproj	New Android functional test project definition + env var injection.

src/tasks/AndroidAppBuilder/Templates/monodroid.c

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

src/tasks/AndroidAppBuilder/Templates/monodroid.c

jpnurmi · 2026-03-23T16:47:09Z

@dotnet-policy-service agree company="Sentry"

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

src/tasks/AndroidAppBuilder/Templates/monodroid.c

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

src/tasks/AndroidAppBuilder/Templates/monodroid.c

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

src/tasks/AndroidAppBuilder/Templates/monodroid.c

jpnurmi · 2026-03-27T09:14:57Z

The fix proposal is simple, but the test gets quite complicated. Sorry about that. :) Let me know if you'd prefer to leave it out, even though then nothing guarantees that it stays fixed in the future...

lateralusX · 2026-03-27T11:43:39Z

src/mono/mono/mini/mini-exceptions.c

+	// that secondary signals (e.g. SIGABRT from FORTIFY on other threads)
+	// don't kill the process with SIG_DFL before we can chain the original
+	// crash to the previous handler.
+	if (!mono_do_crash_chaining) {


Maybe it's better to restore the signal handlers instead of keeping the mono crash handlers active past this point. There is a function in mini-posix.c that removes our registered signal handlers and restore them with what existed before. It is currently a static method, but you could do a mono_runtime_remove_handlers that could remove the signals handled here and restore what was set before mono runtime installed its handlers.

lateralusX · 2026-03-27T12:19:31Z

src/tasks/AndroidAppBuilder/Templates/monodroid.c

 #include <sys/system_properties.h>
 #include <sys/mman.h>
 #include <assert.h>
+#include <setjmp.h>


I'm not too happy to add test specific code into the Android App Builder. Maybe this can be handled differently? Maybe we could add some private property that the tests could use to add a custom source file that gets included in build of monodroid.c? Another option is to use the NativeLibrary item collection and make sure it gets linked into the app. This is what normally used with Native AOT to link custom static libraries together with an app, it is also used together with direct pinvoke. Mono Android App Builder do support direct pinvoke as well for AOT scenarios, but currently don't handle NativeLibrary items collection, so one option could be to piggyback on that. Your library could then use __attribute__((constructor)) to run the signal setup before main gets called and your test symbol should be picked up as well when linking that library.

Extract test-specific native code into a separate source file (test_crash_chaining.c) compiled via the new ExtraNativeSources property. Uses __attribute__((constructor)) to auto-install the pre-mono SIGSEGV handler at .so load time.

Use mono_runtime_posix_remove_signal_handler to restore whatever handlers were installed before Mono, rather than unconditionally resetting to SIG_DFL. This correctly handles crash chaining without needing a special case.

Copilot AI review requested due to automatic review settings March 20, 2026 14:59

github-actions bot added the area-Infrastructure-mono label Mar 20, 2026

dotnet-policy-service bot added the community-contribution Indicates that the PR has been added by a community member label Mar 20, 2026

Copilot started reviewing on behalf of jpnurmi March 20, 2026 15:00 View session

Copilot AI reviewed Mar 20, 2026

View reviewed changes

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

Apply suggestions from code review

3e7cb0f

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Copilot AI review requested due to automatic review settings March 20, 2026 15:26

Use siglongjmp instead of manually skipping the faulting instruction

fcdda4d

Copilot started reviewing on behalf of jpnurmi March 20, 2026 15:27 View session

jpnurmi changed the title ~~[mono] Preserve signal handlers during crash chaining in mono_handle_native_crash~~ [mono] Preserve signal handlers during crash chaining Mar 20, 2026

Copilot AI reviewed Mar 20, 2026

View reviewed changes

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

Use designated initializer instead of memset in signal handler

faf4661

Merge branch 'main' into fix/mono-crash-chaining

b5762a7

Copilot AI review requested due to automatic review settings March 23, 2026 16:49

Copilot started reviewing on behalf of jpnurmi March 23, 2026 16:50 View session

Copilot AI reviewed Mar 23, 2026

View reviewed changes

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

This was referenced Mar 24, 2026

[browser] Wasm.Build.Tests timeout - Timed out after 10s waiting for 'WASM EXIT' #116697

Open

System.Formats.Nrbf.Tests.EdgeCaseTests.CanReadArrayOfAnySize times out on checked coreclr #125994

Closed

Copilot AI mentioned this pull request Mar 24, 2026

Fix CanReadArrayOfAnySize timeout on checked coreclr by skipping the 2 GB case on non-release runtimes #126006

Merged

Merge branch 'main' into fix/mono-crash-chaining

873206e

Merge branch 'main' into fix/mono-crash-chaining

89f6af9

Copilot AI review requested due to automatic review settings March 26, 2026 14:00

Copilot started reviewing on behalf of jpnurmi March 26, 2026 14:02 View session

Copilot AI reviewed Mar 26, 2026

View reviewed changes

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

Update src/tasks/AndroidAppBuilder/Templates/monodroid.c

6a3a54a

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Copilot AI review requested due to automatic review settings March 26, 2026 15:25

Copilot started reviewing on behalf of jpnurmi March 26, 2026 15:26 View session

Copilot AI reviewed Mar 26, 2026

View reviewed changes

src/tasks/AndroidAppBuilder/Templates/monodroid.c Outdated Show resolved Hide resolved

jpnurmi marked this pull request as ready for review March 26, 2026 20:37

jpnurmi requested review from BrzVlad, steveisok and vitek-karas as code owners March 26, 2026 20:37

BrzVlad requested a review from lateralusX March 27, 2026 06:04

jpnurmi mentioned this pull request Mar 27, 2026

Prevent Native/Cocoa SDKs from capturing errors that also get captured by .NET getsentry/sentry-dotnet#3954

Open

lateralusX reviewed Mar 27, 2026

View reviewed changes

jpnurmi added 2 commits March 27, 2026 20:55

Move crash chaining test code out of monodroid.c

4ba4f72

Extract test-specific native code into a separate source file (test_crash_chaining.c) compiled via the new ExtraNativeSources property. Uses __attribute__((constructor)) to auto-install the pre-mono SIGSEGV handler at .so load time.

Restore pre-Mono signal handlers instead of resetting to SIG_DFL

10daad5

Use mono_runtime_posix_remove_signal_handler to restore whatever handlers were installed before Mono, rather than unconditionally resetting to SIG_DFL. This correctly handles crash chaining without needing a special case.

jpnurmi force-pushed the fix/mono-crash-chaining branch from 2b15ad2 to 10daad5 Compare March 27, 2026 20:12

Conversation

jpnurmi commented Mar 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dotnet-policy-service bot commented Mar 20, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

jpnurmi commented Mar 23, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

jpnurmi commented Mar 27, 2026

Uh oh!

lateralusX Mar 27, 2026

Choose a reason for hiding this comment

Uh oh!

lateralusX Mar 27, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

jpnurmi commented Mar 20, 2026 •

edited

Loading