Alexander Rojas created MESOS-4711:
--------------------------------------
Summary: Race condition in libevent poll implementation causes
crash
Key: MESOS-4711
URL: https://issues.apache.org/jira/browse/MESOS-4711
Project: Mesos
Issue Type: Bug
Components: libprocess
Affects Versions: 0.28.0
Environment: CentOS 6.7 running in VirtualBox
Reporter: Alexander Rojas
The issue first arose in MESOS-3271, but can be reproduced every time by using
the mentioned environment and running:
{noformat}
sudo ./bin/mesos-tests.sh
--gtest_filter="MemoryPressureMesosTest.CGROUPS_ROOT_SlaveRecovery"
--gtest_repeat=1000
{noformat}
The problem can be traced back to
[{{libevent_poll.cpp}}|https://github.com/apache/mesos/blob/3539b7a0e15b594148308319bf052d28b1429b98/3rdparty/libprocess/src/libevent_poll.cpp].
If the event is triggered and the the future associated with the event is
discarded, the situation arises in which
[{{pollCallback()}}|https://github.com/apache/mesos/blob/3539b7a0e15b594148308319bf052d28b1429b98/3rdparty/libprocess/src/libevent_poll.cpp#L33]
starts executing just early enough to finish before
[{{pollDiscard()}}|https://github.com/apache/mesos/blob/3539b7a0e15b594148308319bf052d28b1429b98/3rdparty/libprocess/src/libevent_poll.cpp#L53]
executes. If that happens, {{pollCallback()}} deletes the poll object and
{{pollDiscard()}} is left with a dangling pointer which crashes when it
executes the line {{event_active(ev, EV_READ, 0);}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
