livie
/
phoenix-firestorm
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

DRTVWR-476: Use LLThreadSafeQueue, not boost::fibers::buffered_channel. 

We've observed buffered_channel::try_push() hanging, which seems very odd. Try
our own LLThreadSafeQueue instead.
master
Nat Goodspeed 2019-11-14 16:45:39 -05:00
parent af35391114
commit bf8aea5059
1 changed files with 11 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
indra/llmessage/llcoproceduremanager.cpp
View File

@ -33,7 +33,7 @@
#include <chrono>
#include <boost/fiber/buffered_channel.hpp>
#include "llthreadsafequeue.h"
#include "llexception.h"
#include "stringize.h"
@ -105,10 +105,7 @@ private:
CoProcedure_t mProc;
};
// we use a buffered_channel here rather than unbuffered_channel since we want to be able to
// push values without blocking,even if there's currently no one calling a pop operation (due to
// fiber running right now)
typedef boost::fibers::buffered_channel<QueuedCoproc::ptr_t> CoprocQueue_t;
typedef LLThreadSafeQueue<QueuedCoproc::ptr_t> CoprocQueue_t;
// Use shared_ptr to control the lifespan of our CoprocQueue_t instance
// because the consuming coroutine might outlive this LLCoprocedurePool
// instance.
@ -289,7 +286,7 @@ LLCoprocedurePool::LLCoprocedurePool(const std::string &poolName, size_t size):
<< LL_ENDL;
// This should ensure that all waiting coprocedures in this
// pool will wake up and terminate.
pendingCoprocs->close();
pendingCoprocs->pushFront({});
}
return false;
});
@ -319,14 +316,13 @@ LLUUID LLCoprocedurePool::enqueueCoprocedure(const std::string &name, LLCoproced
LLUUID id(LLUUID::generateNewID());
LL_INFOS("CoProcMgr") << "Coprocedure(" << name << ") enqueuing with id=" << id.asString() << " in pool \"" << mPoolName << "\" at " << mPending << LL_ENDL;
auto pushed = mPendingCoprocs->try_push(boost::make_shared<QueuedCoproc>(name, id, proc));
// We don't really have a lot of good options if try_push() failed,
auto pushed = mPendingCoprocs->tryPushFront(boost::make_shared<QueuedCoproc>(name, id, proc));
// We don't really have a lot of good options if tryPushFront() failed,
// perhaps because the consuming coroutine is gummed up or something. This
// method is probably called from code called by mainloop. If we toss an
// llcoro::suspend() call here, we'll circle back for another mainloop
// iteration, possibly resulting in being re-entered here. Let's avoid that.
LL_ERRS_IF(pushed != boost::fibers::channel_op_status::success, "CoProcMgr")
<< "Enqueue failed because queue is " << int(pushed) << LL_ENDL;
LL_ERRS_IF(! pushed, "CoProcMgr") << "Enqueue failed" << LL_ENDL;
++mPending;
return id;
@ -338,23 +334,18 @@ void LLCoprocedurePool::coprocedureInvokerCoro(
LLCoreHttpUtil::HttpCoroutineAdapter::ptr_t httpAdapter)
{
QueuedCoproc::ptr_t coproc;
boost::fibers::channel_op_status status;
for (;;)
{
{
LLCoros::TempStatus st("waiting for work for 10s");
status = pendingCoprocs->pop_wait_for(coproc, std::chrono::seconds(10));
LLCoros::TempStatus st("waiting for work");
coproc = pendingCoprocs->popBack();
}
if (status == boost::fibers::channel_op_status::closed)
if (! coproc)
{
// close() pushes an empty pointer to signal done
break;
}
if(status == boost::fibers::channel_op_status::timeout)
{
LL_INFOS_ONCE() << "pool '" << mPoolName << "' stalled." << LL_ENDL;
continue;
}
// we actually popped an item
--mPending;
@ -384,5 +375,5 @@ void LLCoprocedurePool::coprocedureInvokerCoro(
void LLCoprocedurePool::close()
{
mPendingCoprocs->close();
mPendingCoprocs->pushFront({});
}