diff options
author | Wladimir J. van der Laan <laanwj@gmail.com> | 2016-11-18 13:08:30 +0100 |
---|---|---|
committer | Wladimir J. van der Laan <laanwj@gmail.com> | 2016-11-18 13:10:52 +0100 |
commit | 12519bf62b8c49b1c1744eca6ea5b3162a61f962 (patch) | |
tree | 71bb2fd2c5b5cb823bb7c0be6098085ce81f3ad8 /src | |
parent | a8b2a82618bec43977a65c076f691ec1cddd54be (diff) |
test: Fix use-after-free in scheduler tests
Make a copy of the boost time-point to wait for, otherwise the head of
the queue may be deleted by another thread while this one is waiting,
while the boost function still has a reference to it.
Although this problem is in non-test code, this is not an actual problem
outside of the tests because we use the thread scheduler with only one
service thread, so there will never be threads fighting at the head of
the queue.
The old boost fallback escapes this problem because it passes a scalar
value to wait_until instead of a const object reference.
Found by running the tests in LLVM-4.0-master asan.
Diffstat (limited to 'src')
-rw-r--r-- | src/scheduler.cpp | 7 |
1 files changed, 4 insertions, 3 deletions
diff --git a/src/scheduler.cpp b/src/scheduler.cpp index 52777b61f9..27c03f154c 100644 --- a/src/scheduler.cpp +++ b/src/scheduler.cpp @@ -54,9 +54,10 @@ void CScheduler::serviceQueue() #else // Some boost versions have a conflicting overload of wait_until that returns void. // Explicitly use a template here to avoid hitting that overload. - while (!shouldStop() && !taskQueue.empty() && - newTaskScheduled.wait_until<>(lock, taskQueue.begin()->first) != boost::cv_status::timeout) { - // Keep waiting until timeout + while (!shouldStop() && !taskQueue.empty()) { + boost::chrono::system_clock::time_point timeToWaitFor = taskQueue.begin()->first; + if (newTaskScheduled.wait_until<>(lock, timeToWaitFor) == boost::cv_status::timeout) + break; // Exit loop after timeout, it means we reached the time of the event } #endif // If there are multiple threads, the queue can empty while we're waiting (another |