Unsoundness when starting an isolate per thread #1467

dev-ardi · 2024-04-19T16:01:20Z

Is it supposed to be sound to instantiate one isolate per thread?
If it's not the API shouldn't allow it.

If it is, you have data races as reported by ThreadSanitizer at least when creating and dropping the isolates.

There are more data races reported by it.

use std::sync::Once;

fn main() {
    fn init_isolate() {
        // Init isolate
        let isolate = &mut v8::Isolate::new(Default::default());
        let scope = &mut v8::HandleScope::new(isolate);
        let context = v8::Context::new(scope);
        let scope = &mut v8::ContextScope::new(scope, context);
    }

    static START: Once = Once::new();

    START.call_once(|| {
        let platform = v8::new_default_platform(0, false).make_shared();
        v8::V8::initialize_platform(platform);
        v8::V8::initialize();
    });

    for _ in 0..16 {
        std::thread::spawn(move || {
            init_isolate();
        });
    }
}

This code can be ran with RUSTFLAGS=-Zsanitizer=thread cargo +nightly run -r -Zbuild-std --target x86_64-unknown-linux-gnu

The text was updated successfully, but these errors were encountered:

mmastrac · 2024-04-19T17:34:07Z

Do you mind providing the output of the data races? v8 may have some undocumented ordering requirements that we may need to encode via mutexes.

I assume these don't happen when you serialize them on different threads via mutex?

dev-ardi · 2024-04-19T19:18:19Z

I assume these don't happen when you serialize them on different threads via mutex?

What do you mean by this? Serialize what? Lock what?

The output of TSan:

==================
WARNING: ThreadSanitizer: data race (pid=182953)
  Atomic read of size 1 at 0x5644b1136a18 by thread T16:
    #0 pthread_mutex_lock <null> (tests+0x117f1a)
    #1 v8::base::OS::GetRandomMmapAddr() <null> (tests+0x1a339a)
    #2 tests::main::init_isolate::hdb6a8a0e21dd9715 <null> (tests+0x19b574)
    #3 std::sys_common::backtrace::__rust_begin_short_backtrace::ha9ac64f218ace66d (.llvm.4541231759099217939) <null> (tests+0x19bdbf)
    #4 std::panicking::try::do_call::h443b026a56148bf6 (.llvm.4541231759099217939) <null> (tests+0x19c00f)
    #5 __rust_try.llvm.4541231759099217939 <null> (tests+0x19c561)
    #6 std::panicking::try::h0d28ae24534cbb2c <null> (tests+0x19be48)
    #7 core::ops::function::FnOnce::call_once$u7b$$u7b$vtable.shim$u7d$$u7d$::hfb10caf8482df762 tests.c69f0d4a92970f3-cgu.2 (tests+0x19ce40)
    #8 _$LT$alloc..boxed..Box$LT$F$C$A$GT$$u20$as$u20$core..ops..function..FnOnce$LT$Args$GT$$GT$::call_once::h6f76cea8842f91a2 <null> (tests+0x1577b2d)
    #9 std::sys::pal::unix::thread::Thread::new::thread_start::hc367969c21712737 std.207e8f97ea6a6eb5-cgu.05 (tests+0x1552e3c)

  Previous write of size 1 at 0x5644b1136a18 by thread T22:
    #0 pthread_mutex_init <null> (tests+0x117c0f)
    #1 v8::base::CallOnceImpl(std::Cr::atomic<unsigned char>*, std::Cr::function<void ()>) <null> (tests+0xa8c7ff)
    #2 tests::main::init_isolate::hdb6a8a0e21dd9715 <null> (tests+0x19b574)
    #3 std::sys_common::backtrace::__rust_begin_short_backtrace::ha9ac64f218ace66d (.llvm.4541231759099217939) <null> (tests+0x19bdbf)
    #4 std::panicking::try::do_call::h443b026a56148bf6 (.llvm.4541231759099217939) <null> (tests+0x19c00f)
    #5 __rust_try.llvm.4541231759099217939 <null> (tests+0x19c561)
    #6 std::panicking::try::h0d28ae24534cbb2c <null> (tests+0x19be48)
    #7 core::ops::function::FnOnce::call_once$u7b$$u7b$vtable.shim$u7d$$u7d$::hfb10caf8482df762 tests.c69f0d4a92970f3-cgu.2 (tests+0x19ce40)
    #8 _$LT$alloc..boxed..Box$LT$F$C$A$GT$$u20$as$u20$core..ops..function..FnOnce$LT$Args$GT$$GT$::call_once::h6f76cea8842f91a2 <null> (tests+0x1577b2d)
    #9 std::sys::pal::unix::thread::Thread::new::thread_start::hc367969c21712737 std.207e8f97ea6a6eb5-cgu.05 (tests+0x1552e3c)

  Location is global 'v8::base::(anonymous namespace)::rng_mutex' of size 48 at 0x5644b1136a10 (tests+0x29fea18)

  Thread T16 (tid=183014, running) created by main thread at:
    #0 pthread_create <null> (tests+0x11641b)
    #1 std::sys::pal::unix::thread::Thread::new::ha89e7bb49291c670 <null> (tests+0x1552c51)
    #2 std::thread::spawn::h625baca1ceeb02e1 <null> (tests+0x19c94b)
    #3 tests::main::h20db97e3753fa6e0 tests.c69f0d4a92970f3-cgu.0 (tests+0x19b3df)
    #4 std::sys_common::backtrace::__rust_begin_short_backtrace::h556a55e1ea49f791 tests.c69f0d4a92970f3-cgu.1 (tests+0x19bd8f)
    #5 std::rt::lang_start::_$u7b$$u7b$closure$u7d$$u7d$::h873a0e41b53764be (.llvm.4541231759099217939) <null> (tests+0x19bdfd)
    #6 std::panicking::try::do_call::h9deba7566458d303 std.207e8f97ea6a6eb5-cgu.07 (tests+0x155da49)
    #7 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #8 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #9 std::panicking::try::do_call::he5dfd3bef74c0515 (.llvm.6907422618784771728) <null> (tests+0x155dc1b)
    #10 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #11 std::panicking::try::h536db1b3ee4a40d5 <null> (tests+0x155d968)
    #12 std::rt::lang_start_internal::h6189102e9d939b31 <null> (tests+0x15578e6)
    #13 main <null> (tests+0x19bc32)
    #14 __libc_start_call_main <null> (libc.so.6+0x2a10d) (BuildId: 224c8d53cec4f2afe0c167f14dc41e087c39dfba)
    #15 __libc_start_call_main <null> (libc.so.6+0x2a10d) (BuildId: 224c8d53cec4f2afe0c167f14dc41e087c39dfba)

  Thread T22 (tid=183020, running) created by main thread at:
    #0 pthread_create <null> (tests+0x11641b)
    #1 std::sys::pal::unix::thread::Thread::new::ha89e7bb49291c670 <null> (tests+0x1552c51)
    #2 std::thread::spawn::h625baca1ceeb02e1 <null> (tests+0x19c94b)
    #3 tests::main::h20db97e3753fa6e0 tests.c69f0d4a92970f3-cgu.0 (tests+0x19b3df)
    #4 std::sys_common::backtrace::__rust_begin_short_backtrace::h556a55e1ea49f791 tests.c69f0d4a92970f3-cgu.1 (tests+0x19bd8f)
    #5 std::rt::lang_start::_$u7b$$u7b$closure$u7d$$u7d$::h873a0e41b53764be (.llvm.4541231759099217939) <null> (tests+0x19bdfd)
    #6 std::panicking::try::do_call::h9deba7566458d303 std.207e8f97ea6a6eb5-cgu.07 (tests+0x155da49)
    #7 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #8 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #9 std::panicking::try::do_call::he5dfd3bef74c0515 (.llvm.6907422618784771728) <null> (tests+0x155dc1b)
    #10 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #11 std::panicking::try::h536db1b3ee4a40d5 <null> (tests+0x155d968)
    #12 std::rt::lang_start_internal::h6189102e9d939b31 <null> (tests+0x15578e6)
    #13 main <null> (tests+0x19bc32)
    #14 __libc_start_call_main <null> (libc.so.6+0x2a10d) (BuildId: 224c8d53cec4f2afe0c167f14dc41e087c39dfba)
    #15 __libc_start_call_main <null> (libc.so.6+0x2a10d) (BuildId: 224c8d53cec4f2afe0c167f14dc41e087c39dfba)

SUMMARY: ThreadSanitizer: data race (/home/ardi/repos/tests/target/x86_64-unknown-linux-gnu/release/tests+0x117f1a) in pthread_mutex_lock
==================
==================
WARNING: ThreadSanitizer: data race (pid=182953)
  Atomic read of size 1 at 0x5644b1139f78 by thread T16:
    #0 pthread_mutex_lock <null> (tests+0x117f1a)
    #1 v8::internal::CodeRangeAddressHint::GetAddressHint(unsigned long, unsigned long) <null> (tests+0x28e736)
    #2 tests::main::init_isolate::hdb6a8a0e21dd9715 <null> (tests+0x19b574)
    #3 std::sys_common::backtrace::__rust_begin_short_backtrace::ha9ac64f218ace66d (.llvm.4541231759099217939) <null> (tests+0x19bdbf)
    #4 std::panicking::try::do_call::h443b026a56148bf6 (.llvm.4541231759099217939) <null> (tests+0x19c00f)
    #5 __rust_try.llvm.4541231759099217939 <null> (tests+0x19c561)
    #6 std::panicking::try::h0d28ae24534cbb2c <null> (tests+0x19be48)
    #7 core::ops::function::FnOnce::call_once$u7b$$u7b$vtable.shim$u7d$$u7d$::hfb10caf8482df762 tests.c69f0d4a92970f3-cgu.2 (tests+0x19ce40)
    #8 _$LT$alloc..boxed..Box$LT$F$C$A$GT$$u20$as$u20$core..ops..function..FnOnce$LT$Args$GT$$GT$::call_once::h6f76cea8842f91a2 <null> (tests+0x1577b2d)
    #9 std::sys::pal::unix::thread::Thread::new::thread_start::hc367969c21712737 std.207e8f97ea6a6eb5-cgu.05 (tests+0x1552e3c)

  Previous write of size 1 at 0x5644b1139f78 by thread T22:
    #0 pthread_mutex_init <null> (tests+0x117c0f)
    #1 v8::internal::CodeRange::InitReservation(v8::PageAllocator*, unsigned long) <null> (tests+0x28eff5)
    #2 tests::main::init_isolate::hdb6a8a0e21dd9715 <null> (tests+0x19b574)
    #3 std::sys_common::backtrace::__rust_begin_short_backtrace::ha9ac64f218ace66d (.llvm.4541231759099217939) <null> (tests+0x19bdbf)
    #4 std::panicking::try::do_call::h443b026a56148bf6 (.llvm.4541231759099217939) <null> (tests+0x19c00f)
    #5 __rust_try.llvm.4541231759099217939 <null> (tests+0x19c561)
    #6 std::panicking::try::h0d28ae24534cbb2c <null> (tests+0x19be48)
    #7 core::ops::function::FnOnce::call_once$u7b$$u7b$vtable.shim$u7d$$u7d$::hfb10caf8482df762 tests.c69f0d4a92970f3-cgu.2 (tests+0x19ce40)
    #8 _$LT$alloc..boxed..Box$LT$F$C$A$GT$$u20$as$u20$core..ops..function..FnOnce$LT$Args$GT$$GT$::call_once::h6f76cea8842f91a2 <null> (tests+0x1577b2d)
    #9 std::sys::pal::unix::thread::Thread::new::thread_start::hc367969c21712737 std.207e8f97ea6a6eb5-cgu.05 (tests+0x1552e3c)

  Location is global 'v8::internal::(anonymous namespace)::GetCodeRangeAddressHint()::object' of size 80 at 0x5644b1139f78 (tests+0x2a01f78)

  Thread T16 (tid=183014, running) created by main thread at:
    #0 pthread_create <null> (tests+0x11641b)
    #1 std::sys::pal::unix::thread::Thread::new::ha89e7bb49291c670 <null> (tests+0x1552c51)
    #2 std::thread::spawn::h625baca1ceeb02e1 <null> (tests+0x19c94b)
    #3 tests::main::h20db97e3753fa6e0 tests.c69f0d4a92970f3-cgu.0 (tests+0x19b3df)
    #4 std::sys_common::backtrace::__rust_begin_short_backtrace::h556a55e1ea49f791 tests.c69f0d4a92970f3-cgu.1 (tests+0x19bd8f)
    #5 std::rt::lang_start::_$u7b$$u7b$closure$u7d$$u7d$::h873a0e41b53764be (.llvm.4541231759099217939) <null> (tests+0x19bdfd)
    #6 std::panicking::try::do_call::h9deba7566458d303 std.207e8f97ea6a6eb5-cgu.07 (tests+0x155da49)
    #7 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #8 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #9 std::panicking::try::do_call::he5dfd3bef74c0515 (.llvm.6907422618784771728) <null> (tests+0x155dc1b)
    #10 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #11 std::panicking::try::h536db1b3ee4a40d5 <null> (tests+0x155d968)
    #12 std::rt::lang_start_internal::h6189102e9d939b31 <null> (tests+0x15578e6)
    #13 main <null> (tests+0x19bc32)
    #14 __libc_start_call_main <null> (libc.so.6+0x2a10d) (BuildId: 224c8d53cec4f2afe0c167f14dc41e087c39dfba)
    #15 __libc_start_call_main <null> (libc.so.6+0x2a10d) (BuildId: 224c8d53cec4f2afe0c167f14dc41e087c39dfba)

  Thread T22 (tid=183020, running) created by main thread at:
    #0 pthread_create <null> (tests+0x11641b)
    #1 std::sys::pal::unix::thread::Thread::new::ha89e7bb49291c670 <null> (tests+0x1552c51)
    #2 std::thread::spawn::h625baca1ceeb02e1 <null> (tests+0x19c94b)
    #3 tests::main::h20db97e3753fa6e0 tests.c69f0d4a92970f3-cgu.0 (tests+0x19b3df)
    #4 std::sys_common::backtrace::__rust_begin_short_backtrace::h556a55e1ea49f791 tests.c69f0d4a92970f3-cgu.1 (tests+0x19bd8f)
    #5 std::rt::lang_start::_$u7b$$u7b$closure$u7d$$u7d$::h873a0e41b53764be (.llvm.4541231759099217939) <null> (tests+0x19bdfd)
    #6 std::panicking::try::do_call::h9deba7566458d303 std.207e8f97ea6a6eb5-cgu.07 (tests+0x155da49)
    #7 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #8 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #9 std::panicking::try::do_call::he5dfd3bef74c0515 (.llvm.6907422618784771728) <null> (tests+0x155dc1b)
    #10 __rust_try.llvm.6907422618784771728 <null> (tests+0x155ed61)
    #11 std::panicking::try::h536db1b3ee4a40d5 <null> (tests+0x155d968)
    #12 std::rt::lang_start_internal::h6189102e9d939b31 <null> (tests+0x15578e6)
    #13 main <null> (tests+0x19bc32)
    #14 __libc_start_call_main <null> (libc.so.6+0x2a10d) (BuildId: 224c8d53cec4f2afe0c167f14dc41e087c39dfba)
    #15 __libc_start_call_main <null> (libc.so.6+0x2a10d) (BuildId: 224c8d53cec4f2afe0c167f14dc41e087c39dfba)

SUMMARY: ThreadSanitizer: data race (/home/ardi/repos/tests/target/x86_64-unknown-linux-gnu/release/tests+0x117f1a) in pthread_mutex_lock
==================
ThreadSanitizer: reported 2 warnings

I got even more data races in my real code, when I actually used the isolates. This lead to a segfault.

mmastrac · 2024-04-20T16:29:49Z

I assume these don't happen when you serialize them on different threads via mutex?
What do you mean by this? Serialize what? Lock what?

Serialization and locking refer to mutual exclusion, ie: std::sync::Mutex.

We can certainly fix this or accept a PR, but we would need to understand if this means:

Only one isolate can be instantiated at any given time, or
The first isolate instantiation must happen on its own, while further instantiations may be in parallel.

dev-ardi · 2024-04-22T13:04:34Z

Serialization and locking refer to mutual exclusion, ie: std::sync::Mutex.

Oh serializing as in running serially, I thought you meant serializing as in marshalling :P

Indeed it seems like only the first isolate must be started on its own

dev-ardi · 2024-04-22T13:25:32Z

This is also a good moment to revise the API. I'll expand on another issue though

dev-ardi · 2024-04-22T13:31:33Z

I'll work on this

linrongbin16 · 2024-09-13T05:32:11Z

I think I also meet such issue (or requirement).

I want to use a tokio::spawn to run the js runtime in a non-main thread. But the compiler tells me the v8::Isolate doesn't implement the std::marker::Send so it cannot be sent between threads.

Without the Send trait, seems I cannot use it in either tokio::spawn, tokio::blocking_spawn or even std::thread::spawn.

Could you help me with that?

devsnek · 2024-09-13T06:05:27Z

This is correct, v8::Isolate is not a thread safe type. Lockers (#643) would allow this.

linrongbin16 · 2024-09-13T10:05:23Z

I checked the PR #1411, seems there is some hard issue in the design of the Global, which cannot be easily solve in a short time.

dev-ardi · 2024-09-13T10:38:37Z

What you can do in this case is instead of putting it in a task, you just put it in a plain old thread, and then communicate back and forth with channels.

It's wouldn't be a good idea to put them in a task in the first place since they would be blocking most of the time (well, depending on your workload).

dev-ardi · 2024-09-13T10:43:08Z

For everyone else finding this unsoundness bug here's a work around, you just need to use a OnceLock:
https://github.com/denoland/rusty_v8/pull/1470/files#diff-5968ad1a2c8f5b1977d1dc3776ad078298f49e52a8a7bc137ecb46d6a45fc8a9R314

linrongbin16 · 2024-09-13T11:34:43Z

For everyonelse finding this unsoundness bug here's a work around, you just need to use a OnceLock: https://github.com/denoland/rusty_v8/pull/1470/files#diff-5968ad1a2c8f5b1977d1dc3776ad078298f49e52a8a7bc137ecb46d6a45fc8a9R314

Thanks for sharing! One more question: do you have any idea, does V8 natively support async/await keyword? or the runtime need to implement it? and how to implement?

Update: I searched async/await in ECMAScript https://ecma-international.org/publications-and-standards/standards/ecma-262/, seems it's already been mentioned several times. I guess it's already natively supported. But still wonder how does it work when rusty_v8 running in a single thread. Does the V8 Context switches between different async functions inside a single thread?

devsnek · 2024-09-13T14:05:56Z

Please open a new issue or discussion for new topics 😅

dev-ardi mentioned this issue Apr 22, 2024

api improvement #1470

Closed

linrongbin16 mentioned this issue Sep 13, 2024

Block: How to run a js runtime along with tokio task? rsvim/rsvim#112

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Unsoundness when starting an isolate per thread #1467

Unsoundness when starting an isolate per thread #1467

dev-ardi commented Apr 19, 2024 •

edited

Loading

mmastrac commented Apr 19, 2024 •

edited

Loading

dev-ardi commented Apr 19, 2024

mmastrac commented Apr 20, 2024 •

edited

Loading

dev-ardi commented Apr 22, 2024 •

edited

Loading

dev-ardi commented Apr 22, 2024

dev-ardi commented Apr 22, 2024

linrongbin16 commented Sep 13, 2024

devsnek commented Sep 13, 2024

linrongbin16 commented Sep 13, 2024 •

edited

Loading

dev-ardi commented Sep 13, 2024

dev-ardi commented Sep 13, 2024

linrongbin16 commented Sep 13, 2024 •

edited

Loading

devsnek commented Sep 13, 2024

Unsoundness when starting an isolate per thread #1467

Unsoundness when starting an isolate per thread #1467

Comments

dev-ardi commented Apr 19, 2024 • edited Loading

mmastrac commented Apr 19, 2024 • edited Loading

dev-ardi commented Apr 19, 2024

mmastrac commented Apr 20, 2024 • edited Loading

dev-ardi commented Apr 22, 2024 • edited Loading

dev-ardi commented Apr 22, 2024

dev-ardi commented Apr 22, 2024

linrongbin16 commented Sep 13, 2024

devsnek commented Sep 13, 2024

linrongbin16 commented Sep 13, 2024 • edited Loading

dev-ardi commented Sep 13, 2024

dev-ardi commented Sep 13, 2024

linrongbin16 commented Sep 13, 2024 • edited Loading

devsnek commented Sep 13, 2024

dev-ardi commented Apr 19, 2024 •

edited

Loading

mmastrac commented Apr 19, 2024 •

edited

Loading

mmastrac commented Apr 20, 2024 •

edited

Loading

dev-ardi commented Apr 22, 2024 •

edited

Loading

linrongbin16 commented Sep 13, 2024 •

edited

Loading

linrongbin16 commented Sep 13, 2024 •

edited

Loading