Deadlock in LDClientIdentify

[juancampa]

Is this a support request? No

Describe the bug This SDK can deadlock because a couple of rwlocks are being acquired in reverse order.

The two rwlocks involved are:

1. sharedUserLock
2. clientLock

The deadlock occurs if LDClientIdentify runs concurrently with LDi_readstream or LDi_fetchfeaturemap. Because:

• LDClientIdentify acquires (1) and then (2)
• LDi_readstream acquires (2) and then (1). Same with LDi_fetchfeaturemap

If both functions are running simultaneously, there's a probability of a deadlock ☠️

Here's an instance where we found that happening:

Thread 19 (Thread 0x7f09e3fff700 (LWP 1041)):
#0  0x00007f0a22a7a57f in futex_abstimed_wait (private=<optimized out>, abstime=0x0, expected=3, futex_word=0x7f0a1b7fe4a8 <globalContext+40>) at ../sysdeps/unix/sysv/linux/futex-internal.h:172
#1  0x00007f0a22a7a57f in __pthread_rwlock_rdlock_full (abstime=0x0, rwlock=0x7f0a1b7fe4a0 <globalContext+32>) at pthread_rwlock_common.c:446
#2  0x00007f0a22a7a57f in __GI___pthread_rwlock_rdlock (rwlock=0x7f0a1b7fe4a0 <globalContext+32>) at pthread_rwlock_rdlock.c:27
#3  0x00007f0a1b4aad83 in LDi_rwlock_rdlock_imp () at <redacted>
#4  0x00007f0a1b4bf601 in LDi_readstream () at <redacted>
#5  0x00007f0a1b4b51fa in LDi_bgfeaturestreamer () at <redacted>
#6  0x00007f0a22a756db in start_thread (arg=0x7f09e3fff700) at pthread_create.c:463
#7  0x00007f0a2279e61f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

Thread 1 (Thread 0x7f0a239d3780 (LWP 908)):
#0  0x00007f0a22a7aa74 in futex_abstimed_wait (private=<optimized out>, abstime=0x0, expected=2, futex_word=0x4ad0b08) at ../sysdeps/unix/sysv/linux/futex-internal.h:172
#1  0x00007f0a22a7aa74 in __pthread_rwlock_wrlock_full (abstime=0x0, rwlock=0x4ad0b00) at pthread_rwlock_common.c:803
#2  0x00007f0a22a7aa74 in __GI___pthread_rwlock_wrlock (rwlock=0x4ad0b00) at pthread_rwlock_wrlock.c:27
#3  0x00007f0a1b4aae45 in LDi_rwlock_wrlock_imp () at <redacted>
#4  0x00007f0a1b4af2b9 in LDClientIdentify () at <redacted>
#5  0x00007f0a1b4c18d2 in LDClientCPP::identify(LDUser*) () at <redacted>
#6  0x00007f0a1b25a3c1 in app::SetUser(Napi::CallbackInfo const&) () at <redacted>
#7  0x00007f0a1b29a760 in Napi::details::CallbackData<void (*)(Napi::CallbackInfo const&), void>::Wrapper(napi_env__*, napi_callback_info__*) () at <redacted>
#8  0x00000000009eaa8d in v8impl::(anonymous namespace)::FunctionCallbackWrapper::Invoke(v8::FunctionCallbackInfo<v8::Value> const&) ()
#9  0x0000000000c262eb in v8::internal::MaybeHandle<v8::internal::Object> v8::internal::(anonymous namespace)::HandleApiCallHelper<false>(v8::internal::Isolate*, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::FunctionTemplateInfo>, v8::internal::Handle<v8::internal::Object>, v8::internal::BuiltinArguments) ()
#10 0x0000000000c27896 in v8::internal::Builtin_Impl_HandleApiCall(v8::internal::BuiltinArguments, v8::internal::Isolate*) ()
#11 0x0000000000c27f16 in v8::internal::Builtin_HandleApiCall(int, unsigned long*, v8::internal::Isolate*) ()
#12 0x0000000001448ef9 in Builtins_CEntry_Return1_DontSaveFPRegs_ArgvOnStack_BuiltinExit ()
#13 0x00000000013df942 in Builtins_InterpreterEntryTrampoline ()
...

To reproduce Be unlucky enough to have the above two functions running at the same time

Expected behavior Not to deadlock

Logs N/A

SDK version 2.4.4

Language version, developer tools C/C++

OS/platform Ubuntu 16.04

Additional context N/A

[cwaldren-ld]

Thank you, @juancampa, for the detailed report. We will investigate this.

Filed internally as 154701.

[cwaldren-ld]

Hi @juancampa , the 2.4.5 release should fix the deadlock. Please re-test and let me know if the issue is resolved.

[cwaldren-ld]

I'm going to go ahead and close this issue as the fix has been released. If you still have any problems, please feel free to open another issue.

Thanks!