runtime: remove dedicated scavenger thread

A whole thread is too much for background scavenger that sleeps all the time anyway.
We already have sysmon thread that can do this work.
Also remove g->isbackground and simplify enter/exitsyscall.

LGTM=rsc
R=golang-codereviews, rsc
CC=golang-codereviews, khr, rlh
https://golang.org/cl/108640043

7 files changed, 77 insertions, 92 deletions

diff --git a/src/pkg/runtime/heapdump.c b/src/pkg/runtime/heapdump.c
index 3fc54aedbc..1a38087c8d 100644
--- a/src/pkg/runtime/heapdump.c
+++ b/src/pkg/runtime/heapdump.c
@@ -406,7 +406,7 @@ dumpgoroutine(G *gp)
 	dumpint(gp->gopc);
 	dumpint(gp->status);
 	dumpbool(gp->issystem);
-	dumpbool(gp->isbackground);
+	dumpbool(false);  // isbackground
 	dumpint(gp->waitsince);
 	dumpstr(gp->waitreason);
 	dumpint((uintptr)gp->sched.ctxt);
diff --git a/src/pkg/runtime/malloc.h b/src/pkg/runtime/malloc.h
index fb51a9e5dd..f5a2b2a42d 100644
--- a/src/pkg/runtime/malloc.h
+++ b/src/pkg/runtime/malloc.h
@@ -516,7 +516,7 @@ MSpan*	runtime·MHeap_LookupMaybe(MHeap *h, void *v);
 void*	runtime·MHeap_SysAlloc(MHeap *h, uintptr n);
 void	runtime·MHeap_MapBits(MHeap *h);
 void	runtime·MHeap_MapSpans(MHeap *h);
-void	runtime·MHeap_Scavenger(void);
+void	runtime·MHeap_Scavenge(int32 k, uint64 now, uint64 limit);
 
 void*	runtime·persistentalloc(uintptr size, uintptr align, uint64 *stat);
 int32	runtime·mlookup(void *v, byte **base, uintptr *size, MSpan **s);
diff --git a/src/pkg/runtime/mgc0.c b/src/pkg/runtime/mgc0.c
index 86470c182b..d931e31525 100644
--- a/src/pkg/runtime/mgc0.c
+++ b/src/pkg/runtime/mgc0.c
@@ -1095,7 +1095,7 @@ static struct
 static void
 bgsweep(void)
 {
-	g->issystem = 1;
+	g->issystem = true;
 	for(;;) {
 		while(runtime·sweepone() != -1) {
 			sweep.nbgsweep++;
@@ -1109,9 +1109,7 @@ bgsweep(void)
 			continue;
 		}
 		sweep.parked = true;
-		g->isbackground = true;
 		runtime·parkunlock(&gclock, runtime·gostringnocopy((byte*)"GC sweep wait"));
-		g->isbackground = false;
 	}
 }
 
@@ -1685,9 +1683,9 @@ runfinq(void)
 		finq = nil;
 		if(fb == nil) {
 			runtime·fingwait = true;
-			g->isbackground = true;
+			g->issystem = true;
 			runtime·parkunlock(&finlock, runtime·gostringnocopy((byte*)"finalizer wait"));
-			g->isbackground = false;
+			g->issystem = false;
 			continue;
 		}
 		runtime·unlock(&finlock);
diff --git a/src/pkg/runtime/mheap.c b/src/pkg/runtime/mheap.c
index 59be0e093c..a447bbc973 100644
--- a/src/pkg/runtime/mheap.c
+++ b/src/pkg/runtime/mheap.c
@@ -574,13 +574,6 @@ MHeap_FreeSpanLocked(MHeap *h, MSpan *s, bool acctinuse, bool acctidle)
 		runtime·MSpanList_Insert(&h->freelarge, s);
 }
 
-static void
-forcegchelper(Note *note)
-{
-	runtime·gc(1);
-	runtime·notewakeup(note);
-}
-
 static uintptr
 scavengelist(MSpan *list, uint64 now, uint64 limit)
 {
@@ -603,18 +596,20 @@ scavengelist(MSpan *list, uint64 now, uint64 limit)
 	return sumreleased;
 }
 
-static void
-scavenge(int32 k, uint64 now, uint64 limit)
+void
+runtime·MHeap_Scavenge(int32 k, uint64 now, uint64 limit)
 {
 	uint32 i;
 	uintptr sumreleased;
 	MHeap *h;
 	
 	h = &runtime·mheap;
+	runtime·lock(&h->lock);
 	sumreleased = 0;
 	for(i=0; i < nelem(h->free); i++)
 		sumreleased += scavengelist(&h->free[i], now, limit);
 	sumreleased += scavengelist(&h->freelarge, now, limit);
+	runtime·unlock(&h->lock);
 
 	if(runtime·debug.gctrace > 0) {
 		if(sumreleased > 0)
@@ -630,72 +625,15 @@ scavenge(int32 k, uint64 now, uint64 limit)
 static void
 scavenge_m(G *gp)
 {
-	runtime·lock(&runtime·mheap.lock);
-	scavenge(g->m->scalararg[0], g->m->scalararg[1], g->m->scalararg[2]);
-	runtime·unlock(&runtime·mheap.lock);
+	runtime·MHeap_Scavenge(-1, ~(uintptr)0, 0);
 	runtime·gogo(&gp->sched);
 }
 
-static FuncVal forcegchelperv = {(void(*)(void))forcegchelper};
-
-// Release (part of) unused memory to OS.
-// Goroutine created at startup.
-// Loop forever.
-void
-runtime·MHeap_Scavenger(void)
-{
-	uint64 tick, forcegc, limit;
-	int64 unixnow;
-	int32 k;
-	Note note, *notep;
-
-	g->issystem = true;
-	g->isbackground = true;
-
-	// If we go two minutes without a garbage collection, force one to run.
-	forcegc = 2*60*1e9;
-	// If a span goes unused for 5 minutes after a garbage collection,
-	// we hand it back to the operating system.
-	limit = 5*60*1e9;
-	// Make wake-up period small enough for the sampling to be correct.
-	if(forcegc < limit)
-		tick = forcegc/2;
-	else
-		tick = limit/2;
-
-	for(k=0;; k++) {
-		runtime·noteclear(&note);
-		runtime·notetsleepg(&note, tick);
-
-		unixnow = runtime·unixnanotime();
-		if(unixnow - mstats.last_gc > forcegc) {
-			// The scavenger can not block other goroutines,
-			// otherwise deadlock detector can fire spuriously.
-			// GC blocks other goroutines via the runtime·worldsema.
-			runtime·noteclear(&note);
-			notep = &note;
-			runtime·newproc1(&forcegchelperv, (byte*)&notep, sizeof(notep), 0, runtime·MHeap_Scavenger);
-			runtime·notetsleepg(&note, -1);
-			if(runtime·debug.gctrace > 0)
-				runtime·printf("scvg%d: GC forced\n", k);
-		}
-		g->m->locks++;	// ensure that we are on the same m while filling arguments
-		g->m->scalararg[0] = k;
-		g->m->scalararg[1] = runtime·nanotime();
-		g->m->scalararg[2] = limit;
-		runtime·mcall(scavenge_m);
-		g->m->locks--;
-	}
-}
-
 void
 runtime∕debug·freeOSMemory(void)
 {
 	runtime·gc(2);  // force GC and do eager sweep
 
-	g->m->scalararg[0] = -1;
-	g->m->scalararg[1] = ~(uintptr)0;
-	g->m->scalararg[2] = 0;
 	runtime·mcall(scavenge_m);
 }
 
diff --git a/src/pkg/runtime/proc.c b/src/pkg/runtime/proc.c
index 31ff4f55de..ce0f74aa6c 100644
--- a/src/pkg/runtime/proc.c
+++ b/src/pkg/runtime/proc.c
@@ -130,6 +130,15 @@ static bool exitsyscallfast(void);
 static bool haveexperiment(int8*);
 static void allgadd(G*);
 
+static void forcegchelper(void);
+static struct
+{
+	Lock	lock;
+	G*	g;
+	FuncVal	fv;
+	uint32	idle;
+} forcegc;
+
 extern String runtime·buildVersion;
 
 // The bootstrap sequence is:
@@ -200,8 +209,6 @@ runtime·schedinit(void)
 extern void main·init(void);
 extern void main·main(void);
 
-static FuncVal scavenger = {runtime·MHeap_Scavenger};
-
 static FuncVal initDone = { runtime·unlockOSThread };
 
 // The main goroutine.
@@ -247,7 +254,8 @@ runtime·main(void)
 
 	if(g->m != &runtime·m0)
 		runtime·throw("runtime·main not on m0");
-	runtime·newproc1(&scavenger, nil, 0, 0, runtime·main);
+	forcegc.fv.fn = forcegchelper;
+	forcegc.g = runtime·newproc1(&forcegc.fv, nil, 0, 0, runtime·main);
 	main·init();
 
 	if(g->defer != &d || d.fn != &initDone)
@@ -1619,8 +1627,6 @@ void
 
 	p = releasep();
 	handoffp(p);
-	if(g->isbackground)  // do not consider blocked scavenger for deadlock detection
-		incidlelocked(1);
 
 	// Resave for traceback during blocked call.
 	save(runtime·getcallerpc(&dummy), runtime·getcallersp(&dummy));
@@ -1664,9 +1670,6 @@ runtime·exitsyscall(void)
 {
 	g->m->locks++;  // see comment in entersyscall
 
-	if(g->isbackground)  // do not consider blocked scavenger for deadlock detection
-		incidlelocked(-1);
-
 	g->waitsince = 0;
 	if(exitsyscallfast()) {
 		// There's a cpu for us, so we can run.
@@ -2602,7 +2605,7 @@ checkdead(void)
 	runtime·lock(&allglock);
 	for(i = 0; i < runtime·allglen; i++) {
 		gp = runtime·allg[i];
-		if(gp->isbackground)
+		if(gp->issystem)
 			continue;
 		s = gp->status;
 		if(s == Gwaiting)
@@ -2623,10 +2626,23 @@ checkdead(void)
 static void
 sysmon(void)
 {
-	uint32 idle, delay;
-	int64 now, lastpoll, lasttrace;
+	uint32 idle, delay, nscavenge;
+	int64 now, unixnow, lastpoll, lasttrace;
+	int64 forcegcperiod, scavengelimit, lastscavenge, maxsleep;
 	G *gp;
 
+	// If we go two minutes without a garbage collection, force one to run.
+	forcegcperiod = 2*60*1e6;
+	// If a heap span goes unused for 5 minutes after a garbage collection,
+	// we hand it back to the operating system.
+	scavengelimit = 5*60*1e6;
+	lastscavenge = runtime·nanotime();
+	nscavenge = 0;
+	// Make wake-up period small enough for the sampling to be correct.
+	maxsleep = forcegcperiod/2;
+	if(scavengelimit < forcegcperiod)
+		maxsleep = scavengelimit/2;
+
 	lasttrace = 0;
 	idle = 0;  // how many cycles in succession we had not wokeup somebody
 	delay = 0;
@@ -2644,16 +2660,19 @@ sysmon(void)
 			if(runtime·atomicload(&runtime·sched.gcwaiting) || runtime·atomicload(&runtime·sched.npidle) == runtime·gomaxprocs) {
 				runtime·atomicstore(&runtime·sched.sysmonwait, 1);
 				runtime·unlock(&runtime·sched.lock);
-				runtime·notesleep(&runtime·sched.sysmonnote);
+				runtime·notetsleep(&runtime·sched.sysmonnote, maxsleep);
+				runtime·lock(&runtime·sched.lock);
+				runtime·atomicstore(&runtime·sched.sysmonwait, 0);
 				runtime·noteclear(&runtime·sched.sysmonnote);
 				idle = 0;
 				delay = 20;
-			} else
-				runtime·unlock(&runtime·sched.lock);
+			}
+			runtime·unlock(&runtime·sched.lock);
 		}
 		// poll network if not polled for more than 10ms
 		lastpoll = runtime·atomicload64(&runtime·sched.lastpoll);
 		now = runtime·nanotime();
+		unixnow = runtime·unixnanotime();
 		if(lastpoll != 0 && lastpoll + 10*1000*1000 < now) {
 			runtime·cas64(&runtime·sched.lastpoll, lastpoll, now);
 			gp = runtime·netpoll(false);  // non-blocking
@@ -2677,6 +2696,22 @@ sysmon(void)
 		else
 			idle++;
 
+		// check if we need to force a GC
+		if(unixnow - mstats.last_gc > forcegcperiod && runtime·atomicload(&forcegc.idle)) {
+			runtime·lock(&forcegc.lock);
+			forcegc.idle = 0;
+			forcegc.g->schedlink = nil;
+			injectglist(forcegc.g);
+			runtime·unlock(&forcegc.lock);
+		}
+
+		// scavenge heap once in a while
+		if(lastscavenge + scavengelimit/2 < now) {
+			runtime·MHeap_Scavenge(nscavenge, now, scavengelimit);
+			lastscavenge = now;
+			nscavenge++;
+		}
+
 		if(runtime·debug.schedtrace > 0 && lasttrace + runtime·debug.schedtrace*1000000ll <= now) {
 			lasttrace = now;
 			runtime·schedtrace(runtime·debug.scheddetail);
@@ -2751,6 +2786,23 @@ retake(int64 now)
 	return n;
 }
 
+static void
+forcegchelper(void)
+{
+	g->issystem = true;
+	for(;;) {
+		runtime·lock(&forcegc.lock);
+		if(forcegc.idle)
+			runtime·throw("forcegc: phase error");
+		runtime·atomicstore(&forcegc.idle, 1);
+		runtime·parkunlock(&forcegc.lock, runtime·gostringnocopy((byte*)"force gc (idle)"));
+		// this goroutine is explicitly resumed by sysmon
+		if(runtime·debug.gctrace > 0)
+			runtime·printf("GC forced\n");
+		runtime·gc(1);
+	}
+}
+
 // Tell all goroutines that they have been preempted and they should stop.
 // This function is purely best-effort.  It can fail to inform a goroutine if a
 // processor just started running it.
diff --git a/src/pkg/runtime/runtime.h b/src/pkg/runtime/runtime.h
index beafc76637..df2999bbd9 100644
--- a/src/pkg/runtime/runtime.h
+++ b/src/pkg/runtime/runtime.h
@@ -282,8 +282,7 @@ struct	G
 	String	waitreason;	// if status==Gwaiting
 	G*	schedlink;
 	bool	ispanic;
-	bool	issystem;	// do not output in stack dump
-	bool	isbackground;	// ignore in deadlock detector
+	bool	issystem;	// do not output in stack dump, ignore in deadlock detector
 	bool	preempt;	// preemption signal, duplicates stackguard0 = StackPreempt
 	bool	paniconfault;	// panic (instead of crash) on unexpected fault address
 	int8	raceignore;	// ignore race detection events
diff --git a/src/pkg/runtime/time.go b/src/pkg/runtime/time.go
index 9430414cea..97092c655e 100644
--- a/src/pkg/runtime/time.go
+++ b/src/pkg/runtime/time.go
@@ -192,9 +192,7 @@ func timerproc() {
 		if delta < 0 {
 			// No timers left - put goroutine to sleep.
 			timers.rescheduling = true
-			timers.gp.isbackground = 1
 			goparkunlock(&timers.lock, "timer goroutine (idle)")
-			timers.gp.isbackground = 0
 			continue
 		}
 		// At least one timer pending.  Sleep until then.