Skip to content

Avoid leaking uv_async_t in Dispatcher::Deactivate() #1131

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: master
Choose a base branch
from

Conversation

mszabo-wikia
Copy link

On disconnect, Dispatcher::Deactivate() calls uv_close() to close its active uv_async_t instance and nulls out the corresponding field, but never frees the underlying memory.

To reproduce, run the following with node-rdkafka and librdkafka compiled with ASAN:

const { KafkaConsumer } = require('../');

const consumer = new KafkaConsumer({
  'group.id': 'kafka',
  'metadata.broker.list': 'localhost:9092',
}, {});

consumer.connect({ timeout: 2000 }, function (err) {
  if (err) {
    console.error('Error connecting to Kafka:', err);
    return;
  }

  consumer.disconnect();
})

This should report:

Direct leak of 128 byte(s) in 1 object(s) allocated from:
    #0 0x7f049c3aa647 in operator new(unsigned long) ../../../../src/libsanitizer/asan/asan_new_delete.cpp:99
    #1 0x7f0471d872a1 in NodeKafka::Callbacks::Dispatcher::Activate() ../src/callbacks.cc:69
    #2 0x7f0471e75c06 in NodeKafka::Workers::KafkaConsumerConnect::HandleOKCallback() ../src/workers.cc:579
    #3 0x7f0471dd679c in Nan::AsyncWorker::WorkComplete() ../node_modules/nan/nan.h:2008
    #4 0x7f0471dd679c in Nan::AsyncExecuteComplete(uv_work_s*) ../node_modules/nan/nan.h:2365
    #5 0x7f0471dd679c in Nan::AsyncExecuteComplete(uv_work_s*, int) ../node_modules/nan/nan.h:2369
    #6 0x18bb75c in uv__work_done ../deps/uv/src/threadpool.c:329
    #7 0x18bf0b2 in uv__async_io ../deps/uv/src/unix/async.c:176
    #8 0x18d3b2a in uv__io_poll ../deps/uv/src/unix/linux.c:1485
    #9 0x18bfdd6 in uv_run ../deps/uv/src/unix/core.c:447
    #10 0xbc9be5 in node::SpinEventLoopInternal(node::Environment*) (/usr/bin/node+0xbc9be5)
    #11 0xd1d920 in node::NodeMainInstance::Run(node::ExitCode*, node::Environment*) [clone .part.0] (/usr/bin/node+0xd1d920)
    #12 0xd1e38c in node::NodeMainInstance::Run() (/usr/bin/node+0xd1e38c)
    #13 0xc710be in node::Start(int, char**) (/usr/bin/node+0xc710be)
    #14 0x7f049bdf0d09 in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x23d09)

The libuv documentation says it's only safe to free the underlying memory in a close callback passed to uv_close(), or after such a callback has returned. So, use a unique_ptr with a custom deleter to accomplish this.

On disconnect, Dispatcher::Deactivate() calls uv_close() to close its
active uv_async_t instance and nulls out the corresponding field,
but never frees the underlying memory.

To reproduce, run the following with node-rdkafka and librdkafka
compiled with ASAN:
```js
const { KafkaConsumer } = require('../');

const consumer = new KafkaConsumer({
  'group.id': 'kafka',
  'metadata.broker.list': 'localhost:9092',
}, {});

consumer.connect({ timeout: 2000 }, function (err) {
  if (err) {
    console.error('Error connecting to Kafka:', err);
    return;
  }

  consumer.disconnect();
})
```

This should report:
```
Direct leak of 128 byte(s) in 1 object(s) allocated from:
    #0 0x7f049c3aa647 in operator new(unsigned long) ../../../../src/libsanitizer/asan/asan_new_delete.cpp:99
    Blizzard#1 0x7f0471d872a1 in NodeKafka::Callbacks::Dispatcher::Activate() ../src/callbacks.cc:69
    Blizzard#2 0x7f0471e75c06 in NodeKafka::Workers::KafkaConsumerConnect::HandleOKCallback() ../src/workers.cc:579
    Blizzard#3 0x7f0471dd679c in Nan::AsyncWorker::WorkComplete() ../node_modules/nan/nan.h:2008
    Blizzard#4 0x7f0471dd679c in Nan::AsyncExecuteComplete(uv_work_s*) ../node_modules/nan/nan.h:2365
    Blizzard#5 0x7f0471dd679c in Nan::AsyncExecuteComplete(uv_work_s*, int) ../node_modules/nan/nan.h:2369
    Blizzard#6 0x18bb75c in uv__work_done ../deps/uv/src/threadpool.c:329
    Blizzard#7 0x18bf0b2 in uv__async_io ../deps/uv/src/unix/async.c:176
    Blizzard#8 0x18d3b2a in uv__io_poll ../deps/uv/src/unix/linux.c:1485
    Blizzard#9 0x18bfdd6 in uv_run ../deps/uv/src/unix/core.c:447
    Blizzard#10 0xbc9be5 in node::SpinEventLoopInternal(node::Environment*) (/usr/bin/node+0xbc9be5)
    Blizzard#11 0xd1d920 in node::NodeMainInstance::Run(node::ExitCode*, node::Environment*) [clone .part.0] (/usr/bin/node+0xd1d920)
    Blizzard#12 0xd1e38c in node::NodeMainInstance::Run() (/usr/bin/node+0xd1e38c)
    Blizzard#13 0xc710be in node::Start(int, char**) (/usr/bin/node+0xc710be)
    Blizzard#14 0x7f049bdf0d09 in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x23d09)
```

The [libuv documentation](https://docs.libuv.org/en/v1.x/handle.html#c.uv_close)
says it's only safe to free the underlying memory in a close callback
passed to uv_close(), or after such a callback has returned. So, use a
unique_ptr with a custom deleter to accomplish this.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant