This fixes a regression introduced recently, that increased timing jitter
during the startup of the FastMixer and AudioTrack callback threads.
The regression was to make requestPriority() asynchronous as a way to
avoid an apparent priority inversion in system_server. This means that
the target thread could run briefly with the initial priority, before
the new priority takes effect.
This change removes the startup jitter for FastMixer, by making the
requestPriority() synchronous again for that case. It doesn't matter
that this restores the priority inversion involving normal mixer thread,
because it happens during startup of both threads.
The change also removes the startup jitter for the AudioTrack callback
thread, by having the target thread check whether the requestPriority()
has completed yet. If not, the target thread blocks with a timeout
until the priority boost finishes.
Finally, we now log an error message if the expected priority boost
doesn't happen.
Bug:
8698989
Change-Id: Id590e9a274b70ec1ba85b44a585ee37a22e41cbc
int mPreviousPriority; // before start()
SchedPolicy mPreviousSchedulingGroup;
AudioTrackClientProxy* mProxy;
+ bool mAwaitBoost; // thread should wait for priority boost before running
};
class TimedAudioTrack : public AudioTrack
ALOGW("Requested frameCount %u but received frameCount %u", frameCount, temp);
}
frameCount = temp;
+ mAwaitBoost = false;
if (flags & AUDIO_OUTPUT_FLAG_FAST) {
if (trackFlags & IAudioFlinger::TRACK_FAST) {
ALOGV("AUDIO_OUTPUT_FLAG_FAST successful; frameCount %u", frameCount);
+ mAwaitBoost = true;
} else {
ALOGV("AUDIO_OUTPUT_FLAG_FAST denied by server; frameCount %u", frameCount);
// once denied, do not request again if IAudioTrack is re-created
size_t writtenSize;
mLock.lock();
+ if (mAwaitBoost) {
+ mAwaitBoost = false;
+ mLock.unlock();
+ static const int32_t kMaxTries = 5;
+ int32_t tryCounter = kMaxTries;
+ uint32_t pollUs = 10000;
+ do {
+ int policy = sched_getscheduler(0);
+ if (policy == SCHED_FIFO || policy == SCHED_RR) {
+ break;
+ }
+ usleep(pollUs);
+ pollUs <<= 1;
+ } while (tryCounter-- > 0);
+ if (tryCounter < 0) {
+ ALOGE("did not receive expected priority boost on time");
+ }
+ return true;
+ }
// acquire a strong reference on the IMemory and IAudioTrack so that they cannot be destroyed
// while we are accessing the cblk
sp<IAudioTrack> audioTrack = mAudioTrack;
if (old <= 0) {
__futex_syscall4(coldFutexAddr, FUTEX_WAIT_PRIVATE, old - 1, NULL);
}
+ int policy = sched_getscheduler(0);
+ if (!(policy == SCHED_FIFO || policy == SCHED_RR)) {
+ ALOGE("did not receive expected priority boost");
+ }
// This may be overly conservative; there could be times that the normal mixer
// requests such a brief cold idle that it doesn't require resetting this flag.
isWarm = false;
{
}
- virtual int requestPriority(int32_t pid, int32_t tid, int32_t prio)
+ virtual int requestPriority(int32_t pid, int32_t tid, int32_t prio, bool asynchronous)
{
Parcel data, reply;
data.writeInterfaceToken(ISchedulingPolicyService::getInterfaceDescriptor());
data.writeInt32(pid);
data.writeInt32(tid);
data.writeInt32(prio);
- remote()->transact(REQUEST_PRIORITY_TRANSACTION, data, &reply, IBinder::FLAG_ONEWAY);
+ uint32_t flags = asynchronous ? IBinder::FLAG_ONEWAY : 0;
+ remote()->transact(REQUEST_PRIORITY_TRANSACTION, data, &reply, flags);
// fail on exception
if (reply.readExceptionCode() != 0) return -1;
return reply.readInt32();
DECLARE_META_INTERFACE(SchedulingPolicyService);
virtual int requestPriority(/*pid_t*/int32_t pid, /*pid_t*/int32_t tid,
- int32_t prio) = 0;
+ int32_t prio, bool asynchronous) = 0;
};
static const String16 _scheduling_policy("scheduling_policy");
static Mutex sMutex;
-int requestPriority(pid_t pid, pid_t tid, int32_t prio)
+int requestPriority(pid_t pid, pid_t tid, int32_t prio, bool asynchronous)
{
// FIXME merge duplicated code related to service lookup, caching, and error recovery
sp<ISchedulingPolicyService> sps;
}
sleep(1);
}
- return sps->requestPriority(pid, tid, prio);
+ return sps->requestPriority(pid, tid, prio, asynchronous);
}
} // namespace android
// Request elevated priority for thread tid, whose thread group leader must be pid.
// The priority parameter is currently restricted to either 1 or 2.
-int requestPriority(pid_t pid, pid_t tid, int32_t prio);
+// The asynchronous parameter should be 'true' to return immediately,
+// after the request is enqueued but not necessarily executed.
+// The default value 'false' means to return after request has been enqueued and executed.
+int requestPriority(pid_t pid, pid_t tid, int32_t prio, bool asynchronous = false);
} // namespace android
switch(event->type()) {
case CFG_EVENT_PRIO: {
PrioConfigEvent *prioEvent = static_cast<PrioConfigEvent *>(event);
- int err = requestPriority(prioEvent->pid(), prioEvent->tid(), prioEvent->prio());
+ // FIXME Need to understand why this has be done asynchronously
+ int err = requestPriority(prioEvent->pid(), prioEvent->tid(), prioEvent->prio(),
+ true /*asynchronous*/);
if (err != 0) {
ALOGW("Policy SCHED_FIFO priority %d is unavailable for pid %d tid %d; "
"error %d",