fix(profiling): guard against invalid access on exit

[KowalskiThomas]

Description

This PR addresses a crash that can occur at process exit if the Sampling thread failed to stop within the allowed time. When this happens, we would previously still free the Sampler object (through implicit destruction of the std::unique_ptr) and the Sampling Thread could then try to read data from e.g. the StringTable which was being or had been deleted.

Note I don't have any proof that the crash is caused by this specific code path, but it does make sense to me because in that case, we would definitely crash. So regardless, it's probably worth "fixing" (even though we're working around and not really properly fixing here -- I just don't see another way).

#0 0x000070c4fae4d0b0 std::_Hashtable<unsigned long, std::pair<unsigned long const, std::string>, std::allocator<std::pair<unsigned long const, std::string> >, std::__detail::_Select1st, std::equal_to<unsigned long>, std::hash<unsigned long>, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, false, true> >::_M_find_before_node(unsigned long, unsigned long const&, unsigned long) const [clone .constprop.0]
#1 0x000070c4fae4d8e5 std::_Hashtable<unsigned long, std::pair<unsigned long const, std::string>, std::allocator<std::pair<unsigned long const, std::string> >, std::__detail::_Select1st, std::equal_to<unsigned long>, std::hash<unsigned long>, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, false, true> >::_M_find_node(unsigned long, unsigned long const&, unsigned long) const [clone .isra.0]
#2 0x000070c4fae4699f StringTable::key(_object*, StringTag)
#3 0x000070c4fae46cb4 Frame::create(EchionSampler&, PyCodeObject*, int) [clone .localalias]
#4 0x000070c4fae46f87 Frame::get(EchionSampler&, PyCodeObject*, int) [clone .localalias]
#5 0x000070c4fae47101 Frame::read(EchionSampler&, _PyInterpreterFrame*, _PyInterpreterFrame**)
#6 0x000070c4fae47226 unwind_frame(EchionSampler&, _object*, FrameStack&, unsigned long) [clone .localalias]
#7 0x000070c4fae496b7 ThreadInfo::unwind(EchionSampler&, _ts*) [clone .localalias]
#8 0x000070c4fae4cb01 ThreadInfo::sample(EchionSampler&, _ts*, long)
#9 0x000070c4fae4cc36 std::_Function_handler<void (_ts*, ThreadInfo&), Datadog::Sampler::sampling_thread(unsigned long)::{lambda(InterpreterInfo&)#1}::operator()(InterpreterInfo&) const::{lambda(_ts*, ThreadInfo&)#1}>::_M_invoke(std::_Any_data const&, _ts*&&, ThreadInfo&)
#10 0x000070c4fae49942 for_each_thread(EchionSampler&, InterpreterInfo&, std::function<void (_ts*, ThreadInfo&)> const&)
#11 0x000070c4fae499da std::_Function_handler<void (InterpreterInfo&), Datadog::Sampler::sampling_thread(unsigned long)::{lambda(InterpreterInfo&)#1}>::_M_invoke(std::_Any_data const&, InterpreterInfo&)
#12 0x000070c4fae46288 for_each_interp(pyruntimestate*, std::function<void (InterpreterInfo&)> const&)
#13 0x000070c4fae49d8d Datadog::Sampler::sampling_thread(unsigned long) [clone .localalias]
#14 0x000070c4fae49ff9 call_sampling_thread(void*)
#15 0x000070c4fdf321f5 start_thread (nptl/nptl/pthread_create.c:442)
#16 0x000070c4fdfb1b40 __clone (sysdeps/unix/sysv/linux/x86_64/clone.S:102)
#17 0x0000000000000000

[datadog-prod-us1-6]

View session in Datadog

Bits Dev status: ✅ Done

CI Auto-fix: Disabled | Enable

Comment @DataDog to request changes

[datadog-official]

I can only run on private repositories.

[cit-pr-commenter-54b7da]

Codeowners resolved as

ddtrace/internal/datadog/profiling/stack/include/sampler.hpp            @DataDog/profiling-python
ddtrace/internal/datadog/profiling/stack/src/sampler.cpp                @DataDog/profiling-python
releasenotes/notes/profiling-fix-segv-sampling-thread-exit-e909f07be050246c.yaml  @DataDog/apm-python