miner: fix possible exit crash due to race in stop

If a thread asks to stop the miner, m_stop will be set, and
that thread will wait to join. If the main thread is exiting
at that time, it will ask the miner to stop, but the miner
will claim it's not mining and early out since m_stop is
set. This will cause the database and other things to get
shutdown. If the miner happens to find a block at that time,
it will try to call core, and crash.

Instead, lock and check whether any threads are currently
in m_threads, since they'll only be cleared once the threads
are joined. Moreover, since we lock, the second thread will
have to wait for the first one to have finished. Calling
join twice on a thread seems fine as per pthread_join(3).
This commit is contained in:
moneromooo-monero 2019-03-14 16:54:03 +00:00
parent 1d1a02e9f9
commit 39f000b394
No known key found for this signature in database
GPG key ID: 686F07454D6CEFC3

View file

@ -432,14 +432,15 @@ namespace cryptonote
{
MTRACE("Miner has received stop signal");
if (!is_mining())
CRITICAL_REGION_LOCAL(m_threads_lock);
bool mining = !m_threads.empty();
if (!mining)
{
MTRACE("Not mining - nothing to stop" );
return true;
}
send_stop_signal();
CRITICAL_REGION_LOCAL(m_threads_lock);
// In case background mining was active and the miner threads are waiting
// on the background miner to signal start.