crossbeam::channel::Receiver::try_recv can block forever if sending thread is blocked about crossbeam

[parasyte]

FWIW, blocking behavior was detected by a user on wasm32 builds (using std::sync::mpsc::Recevier::try_recv()) where the main thread is not allowed to block.

I tracked it down to Mutex contention in https://users.rust-lang.org/t/wasm-winit-wgpu-weird-error-about-waiting-on-the-main-thread-despite-not-waiting-on-the-main-thread/96804/5

[benhansen-io]

Also note that a similar block on the sender happens in the bounded channel.

Stack trace of sender:

#0  0x000055555557d6a8 in core::sync::atomic::atomic_compare_exchange_weak<usize> (dst=0x5555555e2c80, old=110471369, new=110471370, success=core::sync::atomic::Ordering::SeqCst, failure=core::sync::atomic::Ordering::Relaxed)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/core/src/sync/atomic.rs:3278
#1  0x0000555555572252 in core::sync::atomic::AtomicUsize::compare_exchange_weak (self=0x5555555e2c80, current=110471369, new=110471370, success=core::sync::atomic::Ordering::SeqCst, failure=core::sync::atomic::Ordering::Relaxed)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/core/src/sync/atomic.rs:2485
#2  0x00005555555702d9 in crossbeam_channel::flavors::array::Channel<usize>::start_send<usize> (self=0x5555555e2c00, token=0x7fffee3f5930) at /home/ben/.cargo/registry/src/index.crates.io-6f17d22bba15001f/crossbeam-channel-0.5.8/src/flavors/array.rs:175
#3  0x0000555555570a56 in crossbeam_channel::flavors::array::Channel<usize>::send<usize> (self=0x5555555e2c00, msg=42, deadline=...) at /home/ben/.cargo/registry/src/index.crates.io-6f17d22bba15001f/crossbeam-channel-0.5.8/src/flavors/array.rs:342
#4  0x000055555556f283 in crossbeam_channel::channel::Sender<usize>::send<usize> (self=0x7fffee3f5b10, msg=42) at /home/ben/.cargo/registry/src/index.crates.io-6f17d22bba15001f/crossbeam-channel-0.5.8/src/channel.rs:437
#5  0x000055555556ebaa in mpsc_deadlock_reproducer::main::{closure#0} () at src/main.rs:29
#6  0x00005555555760c3 in std::sys_common::backtrace::__rust_begin_short_backtrace<mpsc_deadlock_reproducer::main::{closure_env#0}, ()> (f=...) at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/sys_common/backtrace.rs:134
#7  0x000055555556e993 in std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure#0}<mpsc_deadlock_reproducer::main::{closure_env#0}, ()> () at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/thread/mod.rs:526
#8  0x0000555555568713 in core::panic::unwind_safe::{impl#23}::call_once<(), std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure_env#0}<mpsc_deadlock_reproducer::main::{closure_env#0}, ()>> (self=...)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/core/src/panic/unwind_safe.rs:271
#9  0x0000555555575886 in std::panicking::try::do_call<core::panic::unwind_safe::AssertUnwindSafe<std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure_env#0}<mpsc_deadlock_reproducer::main::{closure_env#0}, ()>>, ()> (data=0x7fffee3f5c90)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/panicking.rs:485
#10 0x0000555555575adb in __rust_try ()
#11 0x00005555555756ee in std::panicking::try<(), core::panic::unwind_safe::AssertUnwindSafe<std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure_env#0}<mpsc_deadlock_reproducer::main::{closure_env#0}, ()>>> (f=...)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/panicking.rs:449
#12 0x0000555555574243 in std::panic::catch_unwind<core::panic::unwind_safe::AssertUnwindSafe<std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure_env#0}<mpsc_deadlock_reproducer::main::{closure_env#0}, ()>>, ()> (f=...)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/panic.rs:140
#13 0x000055555556e40c in std::thread::{impl#0}::spawn_unchecked_::{closure#1}<mpsc_deadlock_reproducer::main::{closure_env#0}, ()> () at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/thread/mod.rs:525
#14 0x00005555555645de in core::ops::function::FnOnce::call_once<std::thread::{impl#0}::spawn_unchecked_::{closure_env#1}<mpsc_deadlock_reproducer::main::{closure_env#0}, ()>, ()> () at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/core/src/ops/function.rs:250
#15 0x00005555555a1865 in alloc::boxed::{impl#45}::call_once<(), dyn core::ops::function::FnOnce<(), Output=()>, alloc::alloc::Global> () at library/alloc/src/boxed.rs:1973
#16 alloc::boxed::{impl#45}::call_once<(), alloc::boxed::Box<dyn core::ops::function::FnOnce<(), Output=()>, alloc::alloc::Global>, alloc::alloc::Global> () at library/alloc/src/boxed.rs:1973
#17 std::sys::unix::thread::{impl#2}::new::thread_start () at library/std/src/sys/unix/thread.rs:108
#18 0x00007ffff7d1844b in ?? () from /usr/lib/libc.so.6
#19 0x00007ffff7d9be40 in ?? () from /usr/lib/libc.so.6

Stack trace of receiver:

#0  0x0000555555568587 in crossbeam_utils::backoff::Backoff::spin (self=0x7fffee1f4754) at /home/ben/.cargo/registry/src/index.crates.io-6f17d22bba15001f/crossbeam-utils-0.8.16/src/backoff.rs:151
#1  0x000055555556fda6 in crossbeam_channel::flavors::array::Channel<usize>::start_recv<usize> (self=0x5555555e2c00, token=0x7fffee1f4828) at /home/ben/.cargo/registry/src/index.crates.io-6f17d22bba15001f/crossbeam-channel-0.5.8/src/flavors/array.rs:293
#2  0x000055555557102a in crossbeam_channel::flavors::array::Channel<usize>::try_recv<usize> (self=0x5555555e2c00) at /home/ben/.cargo/registry/src/index.crates.io-6f17d22bba15001f/crossbeam-channel-0.5.8/src/flavors/array.rs:388
#3  0x000055555556f3a5 in crossbeam_channel::channel::Receiver<usize>::try_recv<usize> (self=0x7fffee1f4ba0) at /home/ben/.cargo/registry/src/index.crates.io-6f17d22bba15001f/crossbeam-channel-0.5.8/src/channel.rs:761
#4  0x000055555556ecbc in mpsc_deadlock_reproducer::main::{closure#1} () at src/main.rs:46
#5  0x00005555555760d9 in std::sys_common::backtrace::__rust_begin_short_backtrace<mpsc_deadlock_reproducer::main::{closure_env#1}, ()> (f=<error reading variable: Cannot access memory at address 0x2>)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/sys_common/backtrace.rs:134
#6  0x000055555556e9c5 in std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure#0}<mpsc_deadlock_reproducer::main::{closure_env#1}, ()> () at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/thread/mod.rs:526
#7  0x0000555555568768 in core::panic::unwind_safe::{impl#23}::call_once<(), std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure_env#0}<mpsc_deadlock_reproducer::main::{closure_env#1}, ()>> (self=...)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/core/src/panic/unwind_safe.rs:271
#8  0x00005555555758fe in std::panicking::try::do_call<core::panic::unwind_safe::AssertUnwindSafe<std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure_env#0}<mpsc_deadlock_reproducer::main::{closure_env#1}, ()>>, ()> (data=0x7fffee1f4c70)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/panicking.rs:485
#9  0x0000555555575adb in __rust_try ()
#10 0x000055555557578e in std::panicking::try<(), core::panic::unwind_safe::AssertUnwindSafe<std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure_env#0}<mpsc_deadlock_reproducer::main::{closure_env#1}, ()>>> (f=...)
    at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/panicking.rs:449
#11 0x0000555555574256 in std::panic::catch_unwind<core::panic::unwind_safe::AssertUnwindSafe<std::thread::{impl#0}::spawn_unchecked_::{closure#1}::{closure_env#0}<mpsc_deadlock_reproducer::main::{closure_env#1}, ()>>, ()> (
    f=<error reading variable: Cannot access memory at address 0x2>) at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/panic.rs:140
#12 0x000055555556e793 in std::thread::{impl#0}::spawn_unchecked_::{closure#1}<mpsc_deadlock_reproducer::main::{closure_env#1}, ()> () at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/std/src/thread/mod.rs:525
#13 0x00005555555645be in core::ops::function::FnOnce::call_once<std::thread::{impl#0}::spawn_unchecked_::{closure_env#1}<mpsc_deadlock_reproducer::main::{closure_env#1}, ()>, ()> () at /rustc/90c541806f23a127002de5b4038be731ba1458ca/library/core/src/ops/function.rs:250
#14 0x00005555555a1865 in alloc::boxed::{impl#45}::call_once<(), dyn core::ops::function::FnOnce<(), Output=()>, alloc::alloc::Global> () at library/alloc/src/boxed.rs:1973
#15 alloc::boxed::{impl#45}::call_once<(), alloc::boxed::Box<dyn core::ops::function::FnOnce<(), Output=()>, alloc::alloc::Global>, alloc::alloc::Global> () at library/alloc/src/boxed.rs:1973
#16 std::sys::unix::thread::{impl#2}::new::thread_start () at library/std/src/sys/unix/thread.rs:108
#17 0x00007ffff7d1844b in ?? () from /usr/lib/libc.so.6
#18 0x00007ffff7d9be40 in ?? () from /usr/lib/libc.so.6

[rc-andres]

We've seen this happen as well using Linux (it hangs for ~10 seconds) using try_recv(). Initially the code we had would do a check for empty and then do recv() and we noticed that hanging, we found that weird so we switched it to try_recv() and saw that it behaves the same way.

In our case the sending thread was the same as the receiver thread but an earlier point (I am not sure why the code was written that way tbh...).

[ibraheemdev]

@rc-andres was your case using a similar thread priority setup?

[ibraheemdev]

See rust-lang/rust#112723 (comment), I think it's unlikely that try_recv can be truly non-blocking.