Give CValidationInterface Support for calling notifications on the CScheduler Thread #10179

Merged
merged 7 commits into from Jul 11, 2017
View
@@ -215,6 +215,19 @@ void Shutdown()
fFeeEstimatesInitialized = false;
}
+ // FlushStateToDisk generates a SetBestChain callback, which we should avoid missing
+ FlushStateToDisk();
+
+ // After there are no more peers/RPC left to give us new data which may generate
+ // CValidationInterface callbacks, flush them...
+ GetMainSignals().FlushBackgroundCallbacks();
+
+ // Any future callbacks will be dropped. This should absolutely be safe - if
+ // missing a callback results in an unrecoverable situation, unclean shutdown
+ // would too. The only reason to do the above flushes is to let the wallet catch
+ // up with our current chain to avoid any strange pruning edge cases and make
+ // next startup faster by avoiding rescan.
+
{
LOCK(cs_main);
if (pcoinsTip != NULL) {
@@ -251,6 +264,7 @@ void Shutdown()
}
#endif
UnregisterAllValidationInterfaces();
+ GetMainSignals().UnregisterBackgroundSignalScheduler();
@ryanofsky

ryanofsky Jul 10, 2017

Contributor

In commit "Give CMainSignals a reference to the global scheduler"

It seems like the right place for this unregister call would be earlier in Shutdown(), after scheduler thread is cancelled and the last signal is sent, for consistency with the register call, which is made when the scheduler thread is started.

This would let you flush the background queue when the signal scheduler is destroyed and not need separate FlushBackgroundCallbacks and EmptyQueue calls made in advance.

Also this would bring shutdown code closer to an raii-style ordering, which should make cleanup easier in the future and would make the various object lifetimes a little easier to understand now.

@TheBlueMatt

TheBlueMatt Jul 11, 2017

Contributor

We cant unregister the background scheduler from the validation interface until we're sure nothing is gonna generate anymore callbacks (so, really, after the pcoinsTip/etc deletions). If we want to mirror the initialization order, we'd have to move it even further down, not up, as de-init in RAII order would be after wallet deltion.

#ifdef ENABLE_WALLET
for (CWalletRef pwallet : vpwallets) {
delete pwallet;
@@ -1203,6 +1217,8 @@ bool AppInitMain(boost::thread_group& threadGroup, CScheduler& scheduler)
CScheduler::Function serviceLoop = boost::bind(&CScheduler::serviceQueue, &scheduler);
threadGroup.create_thread(boost::bind(&TraceThread<CScheduler::Function>, "scheduler", serviceLoop));
+ GetMainSignals().RegisterBackgroundSignalScheduler(scheduler);
+
/* Start the RPC server already. It will be started in "warmup" mode
* and not really process calls already (but it will signify connections
* that the server is there and will be ready later). Warmup mode will
@@ -12,6 +12,7 @@
#include "rpc/server.h"
#include "rpcconsole.h"
#include "test/testutil.h"
+#include "test/test_bitcoin.h"
#include "univalue.h"
#include "util.h"
@@ -35,24 +36,15 @@ void RPCNestedTests::rpcNestedTests()
{
// do some test setup
// could be moved to a more generic place when we add more tests on QT level
- const CChainParams& chainparams = Params();
- RegisterAllCoreRPCCommands(tableRPC);
tableRPC.appendCommand("rpcNestedTest", &vRPCCommands[0]);
ClearDatadirCache();
std::string path = QDir::tempPath().toStdString() + "/" + strprintf("test_bitcoin_qt_%lu_%i", (unsigned long)GetTime(), (int)(GetRand(100000)));
QDir dir(QString::fromStdString(path));
dir.mkpath(".");
ForceSetArg("-datadir", path);
//mempool.setSanityCheck(1.0);
- pblocktree = new CBlockTreeDB(1 << 20, true);
- pcoinsdbview = new CCoinsViewDB(1 << 23, true);
- pcoinsTip = new CCoinsViewCache(pcoinsdbview);
- InitBlockIndex(chainparams);
- {
- CValidationState state;
- bool ok = ActivateBestChain(state, chainparams);
- QVERIFY(ok);
- }
+
+ TestingSetup test;
SetRPCWarmupFinished();
@@ -145,13 +137,5 @@ void RPCNestedTests::rpcNestedTests()
QVERIFY_EXCEPTION_THROWN(RPCConsole::RPCExecuteCommandLine(result, "rpcNestedTest(abc,,)"), std::runtime_error); //don't tollerate empty arguments when using ,
#endif
- UnloadBlockIndex();
- delete pcoinsTip;
- pcoinsTip = nullptr;
- delete pcoinsdbview;
- pcoinsdbview = nullptr;
- delete pblocktree;
- pblocktree = nullptr;
-
fs::remove_all(fs::path(path));
}
View
@@ -139,3 +139,69 @@ size_t CScheduler::getQueueInfo(boost::chrono::system_clock::time_point &first,
}
return result;
}
+
+bool CScheduler::AreThreadsServicingQueue() const {
+ return nThreadsServicingQueue;
+}
+
+
+void SingleThreadedSchedulerClient::MaybeScheduleProcessQueue() {
+ {
+ LOCK(m_cs_callbacks_pending);
+ // Try to avoid scheduling too many copies here, but if we
+ // accidentally have two ProcessQueue's scheduled at once its
+ // not a big deal.
+ if (m_are_callbacks_running) return;
+ if (m_callbacks_pending.empty()) return;
+ }
+ m_pscheduler->schedule(std::bind(&SingleThreadedSchedulerClient::ProcessQueue, this));
+}
+
+void SingleThreadedSchedulerClient::ProcessQueue() {
+ std::function<void (void)> callback;
+ {
+ LOCK(m_cs_callbacks_pending);
+ if (m_are_callbacks_running) return;
+ if (m_callbacks_pending.empty()) return;
+ m_are_callbacks_running = true;
+
+ callback = std::move(m_callbacks_pending.front());
+ m_callbacks_pending.pop_front();
+ }
+
+ // RAII the setting of fCallbacksRunning and calling MaybeScheduleProcessQueue
+ // to ensure both happen safely even if callback() throws.
+ struct RAIICallbacksRunning {
+ SingleThreadedSchedulerClient* instance;
+ RAIICallbacksRunning(SingleThreadedSchedulerClient* _instance) : instance(_instance) {}
+ ~RAIICallbacksRunning() {
+ {
+ LOCK(instance->m_cs_callbacks_pending);
+ instance->m_are_callbacks_running = false;
+ }
+ instance->MaybeScheduleProcessQueue();
+ }
+ } raiicallbacksrunning(this);
+
+ callback();
+}
+
+void SingleThreadedSchedulerClient::AddToProcessQueue(std::function<void (void)> func) {
+ assert(m_pscheduler);
+
+ {
+ LOCK(m_cs_callbacks_pending);
+ m_callbacks_pending.emplace_back(std::move(func));
+ }
+ MaybeScheduleProcessQueue();
+}
+
+void SingleThreadedSchedulerClient::EmptyQueue() {
+ assert(!m_pscheduler->AreThreadsServicingQueue());
+ bool should_continue = true;
+ while (should_continue) {
+ ProcessQueue();
+ LOCK(m_cs_callbacks_pending);
+ should_continue = !m_callbacks_pending.empty();
+ }
+}
View
@@ -14,6 +14,8 @@
#include <boost/thread.hpp>
#include <map>
+#include "sync.h"
+
//
// Simple class for background tasks that should be run
// periodically or once "after a while"
@@ -41,7 +43,7 @@ class CScheduler
typedef std::function<void(void)> Function;
// Call func at/after time t
- void schedule(Function f, boost::chrono::system_clock::time_point t);
+ void schedule(Function f, boost::chrono::system_clock::time_point t=boost::chrono::system_clock::now());
// Convenience method: call f once deltaSeconds from now
void scheduleFromNow(Function f, int64_t deltaMilliSeconds);
@@ -69,6 +71,9 @@ class CScheduler
size_t getQueueInfo(boost::chrono::system_clock::time_point &first,
boost::chrono::system_clock::time_point &last) const;
+ // Returns true if there are threads actively running in serviceQueue()
+ bool AreThreadsServicingQueue() const;
+
private:
std::multimap<boost::chrono::system_clock::time_point, Function> taskQueue;
boost::condition_variable newTaskScheduled;
@@ -79,4 +84,30 @@ class CScheduler
bool shouldStop() { return stopRequested || (stopWhenEmpty && taskQueue.empty()); }
};
+/**
+ * Class used by CScheduler clients which may schedule multiple jobs
+ * which are required to be run serially. Does not require such jobs
+ * to be executed on the same thread, but no two jobs will be executed
+ * at the same time.
+ */
+class SingleThreadedSchedulerClient {
+private:
+ CScheduler *m_pscheduler;
+
+ CCriticalSection m_cs_callbacks_pending;
+ std::list<std::function<void (void)>> m_callbacks_pending;
+ bool m_are_callbacks_running = false;
+
+ void MaybeScheduleProcessQueue();
+ void ProcessQueue();
+
+public:
+ SingleThreadedSchedulerClient(CScheduler *pschedulerIn) : m_pscheduler(pschedulerIn) {}
+ void AddToProcessQueue(std::function<void (void)> func);
+
+ // Processes all remaining queue members on the calling thread, blocking until queue is empty
+ // Must be called after the CScheduler has no remaining processing threads!
+ void EmptyQueue();
+};
+
#endif
@@ -62,6 +62,12 @@ TestingSetup::TestingSetup(const std::string& chainName) : BasicTestingSetup(cha
pathTemp = GetTempPath() / strprintf("test_bitcoin_%lu_%i", (unsigned long)GetTime(), (int)(InsecureRandRange(100000)));
fs::create_directories(pathTemp);
ForceSetArg("-datadir", pathTemp.string());
+
+ // Note that because we don't bother running a scheduler thread here,
+ // callbacks via CValidationInterface are unreliable, but that's OK,
+ // our unit tests aren't testing multiple parts of the code at once.
+ GetMainSignals().RegisterBackgroundSignalScheduler(scheduler);
+
mempool.setSanityCheck(1.0);
pblocktree = new CBlockTreeDB(1 << 20, true);
pcoinsdbview = new CCoinsViewDB(1 << 23, true);
@@ -88,6 +94,8 @@ TestingSetup::~TestingSetup()
UnregisterNodeSignals(GetNodeSignals());
threadGroup.interrupt_all();
threadGroup.join_all();
+ GetMainSignals().FlushBackgroundCallbacks();
+ GetMainSignals().UnregisterBackgroundSignalScheduler();
UnloadBlockIndex();
delete pcoinsTip;
delete pcoinsdbview;
View
@@ -10,6 +10,7 @@
#include "key.h"
#include "pubkey.h"
#include "random.h"
+#include "scheduler.h"
#include "txdb.h"
#include "txmempool.h"
@@ -53,6 +54,7 @@ struct TestingSetup: public BasicTestingSetup {
fs::path pathTemp;
boost::thread_group threadGroup;
CConnman* connman;
+ CScheduler scheduler;
TestingSetup(const std::string& chainName = CBaseChainParams::MAIN);
~TestingSetup();
Oops, something went wrong.