fix(profiling): reset all profiling c++ mutexes on fork [backport 2.19] (#11993)

Backport d855c4a from #11768 to 2.19.

I'm not sure why it took so long to surface this defect, but it turns
out that stack v2 can deadlock applications because not all mutices are
reset.

The repro in #11762 appears to be pretty durable. I need to investigate
it a bit more in order to distill it down into a native stress test we
can use moving forward. In practice, this patch suppresses the noted
behavior in the repro.

## Checklist
- [x] PR author has checked that all the criteria below are met
- The PR description includes an overview of the change
- The PR description articulates the motivation for the change
- The change includes tests OR the PR description describes a testing
strategy
- The PR description notes risks associated with the change, if any
- Newly-added code is easy to change
- The change follows the [library release note
guidelines](https://ddtrace.readthedocs.io/en/stable/releasenotes.html)
- The change includes or references documentation updates if necessary
- Backport labels are set (if
[applicable](https://ddtrace.readthedocs.io/en/latest/contributing.html#backporting))

## Reviewer Checklist
- [x] Reviewer has checked that all the criteria below are met 
- Title is accurate
- All changes are related to the pull request's stated goal
- Avoids breaking
[API](https://ddtrace.readthedocs.io/en/stable/versioning.html#interfaces)
changes
- Testing strategy adequately addresses listed risks
- Newly-added code is easy to change
- Release note makes sense to a user of the library
- If necessary, author has acknowledged and discussed the performance
implications of this PR as reported in the benchmarks PR comment
- Backport labels are set in a manner that is consistent with the
[release branch maintenance
policy](https://ddtrace.readthedocs.io/en/latest/contributing.html#backporting)

Co-authored-by: David Sanchez <[email protected]>

Author: github-actions[bot]

ddtrace/internal/datadog/profiling/dd_wrapper/include/uploader_builder.hpp

@@ -43,6 +43,8 @@ class UploaderBuilder
     static void set_output_filename(std::string_view _output_filename);
 
     static std::variant<Uploader, std::string> build();
+
+    static void postfork_child();
 };
 
 } // namespace Datadog

ddtrace/internal/datadog/profiling/dd_wrapper/src/code_provenance.cpp

@@ -14,9 +14,8 @@ namespace Datadog {
 void
 Datadog::CodeProvenance::postfork_child()
 {
-    get_instance().mtx.~mutex();            // Destroy the mutex
+    // NB placement-new to re-init and leak the mutex because doing anything else is UB
     new (&get_instance().mtx) std::mutex(); // Recreate the mutex
-    get_instance().reset();                 // Reset the state
 }
 
 void

ddtrace/internal/datadog/profiling/dd_wrapper/src/ddup_interface.cpp

@@ -24,6 +24,7 @@ ddup_postfork_child()
     Datadog::Uploader::postfork_child();
     Datadog::SampleManager::postfork_child();
     Datadog::CodeProvenance::postfork_child();
+    Datadog::UploaderBuilder::postfork_child();
 }
 
 void

ddtrace/internal/datadog/profiling/dd_wrapper/src/profile.cpp

@@ -203,6 +203,6 @@ Datadog::Profile::collect(const ddog_prof_Sample& sample, int64_t endtime_ns)
 void
 Datadog::Profile::postfork_child()
 {
-    profile_mtx.unlock();
+    new (&profile_mtx) std::mutex();
     cycle_buffers();
 }

ddtrace/internal/datadog/profiling/dd_wrapper/src/sample.cpp

@@ -408,7 +408,6 @@ Datadog::Sample::push_absolute_ns(int64_t _timestamp_ns)
     return true;
 }
 
-
 bool
 Datadog::Sample::push_monotonic_ns(int64_t _monotonic_ns)
 {

ddtrace/internal/datadog/profiling/dd_wrapper/src/uploader.cpp

@@ -175,5 +175,6 @@ Datadog::Uploader::postfork_parent()
 void
 Datadog::Uploader::postfork_child()
 {
-    unlock();
+    // NB placement-new to re-init and leak the mutex because doing anything else is UB
+    new (&upload_lock) std::mutex();
 }

ddtrace/internal/datadog/profiling/dd_wrapper/src/uploader_builder.cpp

@@ -186,3 +186,10 @@ Datadog::UploaderBuilder::build()
 
     return Datadog::Uploader{ output_filename, ddog_exporter };
 }
+
+void
+Datadog::UploaderBuilder::postfork_child()
+{
+    // NB placement-new to re-init and leak the mutex because doing anything else is UB
+    new (&tag_mutex) std::mutex();
+}

ddtrace/internal/datadog/profiling/stack_v2/src/sampler.cpp

@@ -67,6 +67,11 @@ _stack_v2_atfork_child()
     // so we don't even reveal this function to the user
     _set_pid(getpid());
     ThreadSpanLinks::postfork_child();
+
+    // `thread_info_map_lock` and `task_link_map_lock` are global locks held in echion
+    // NB placement-new to re-init and leak the mutex because doing anything else is UB
+    new (&thread_info_map_lock) std::mutex;
+    new (&task_link_map_lock) std::mutex;
 }
 
 __attribute__((constructor)) void

ddtrace/internal/datadog/profiling/stack_v2/src/thread_span_links.cpp

@@ -53,10 +53,8 @@ ThreadSpanLinks::reset()
 void
 ThreadSpanLinks::postfork_child()
 {
-    // Explicitly destroy and reconstruct the mutex to avoid undefined behavior
-    get_instance().mtx.~mutex();
+    // NB placement-new to re-init and leak the mutex because doing anything else is UB
     new (&get_instance().mtx) std::mutex();
-
     get_instance().reset();
 }
 

releasenotes/notes/fix-profiling-native-mutices-62440b5a3d9d6c4b.yaml

@@ -0,0 +1,5 @@
+---
+fixes:
+  - |
+    profiling: Fixes a bug where profiling mutexes were not cleared on fork in the child process.  This could
+    cause deadlocks in certain configurations.