blob: bcbe7f88a0567db54ceb5f24dd22ff8ba72525b1 [file] [log] [blame]
// Copyright 2022 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package runtime
import "runtime/internal/atomic"
// gcCPULimiter is a mechanism to limit GC CPU utilization in situations
// where it might become excessive and inhibit application progress (e.g.
// a death spiral).
//
// The core of the limiter is a leaky bucket mechanism that fills with GC
// CPU time and drains with mutator time. Because the bucket fills and
// drains with time directly (i.e. without any weighting), this effectively
// sets a very conservative limit of 50%. This limit could be enforced directly,
// however, but the purpose of the bucket is to accommodate spikes in GC CPU
// utilization without hurting throughput.
//
// Note that the bucket in the leaky bucket mechanism can never go negative,
// so the GC never gets credit for a lot of CPU time spent without the GC
// running. This is intentional, as an application that stays idle for, say,
// an entire day, could build up enough credit to fail to prevent a death
// spiral the following day. The bucket's capacity is the GC's only leeway.
//
// The capacity thus also sets the window the limiter considers. For example,
// if the capacity of the bucket is 1 cpu-second, then the limiter will not
// kick in until at least 1 full cpu-second in the last 2 cpu-second window
// is spent on GC CPU time.
var gcCPULimiter gcCPULimiterState
type gcCPULimiterState struct {
lock atomic.Uint32
enabled atomic.Bool
bucket struct {
// Invariants:
// - fill >= 0
// - capacity >= 0
// - fill <= capacity
fill, capacity uint64
}
// overflow is the cumulative amount of GC CPU time that we tried to fill the
// bucket with but exceeded its capacity.
overflow uint64
// gcEnabled is an internal copy of gcBlackenEnabled that determines
// whether the limiter tracks total assist time.
//
// gcBlackenEnabled isn't used directly so as to keep this structure
// unit-testable.
gcEnabled bool
// transitioning is true when the GC is in a STW and transitioning between
// the mark and sweep phases.
transitioning bool
// assistTimePool is the accumulated assist time since the last update.
assistTimePool atomic.Int64
// idleMarkTimePool is the accumulated idle mark time since the last update.
idleMarkTimePool atomic.Int64
// idleTimePool is the accumulated time Ps spent on the idle list since the last update.
idleTimePool atomic.Int64
// lastUpdate is the nanotime timestamp of the last time update was called.
//
// Updated under lock, but may be read concurrently.
lastUpdate atomic.Int64
// lastEnabledCycle is the GC cycle that last had the limiter enabled.
lastEnabledCycle atomic.Uint32
// nprocs is an internal copy of gomaxprocs, used to determine total available
// CPU time.
//
// gomaxprocs isn't used directly so as to keep this structure unit-testable.
nprocs int32
// test indicates whether this instance of the struct was made for testing purposes.
test bool
}
// limiting returns true if the CPU limiter is currently enabled, meaning the Go GC
// should take action to limit CPU utilization.
//
// It is safe to call concurrently with other operations.
func (l *gcCPULimiterState) limiting() bool {
return l.enabled.Load()
}
// startGCTransition notifies the limiter of a GC transition.
//
// This call takes ownership of the limiter and disables all other means of
// updating the limiter. Release ownership by calling finishGCTransition.
//
// It is safe to call concurrently with other operations.
func (l *gcCPULimiterState) startGCTransition(enableGC bool, now int64) {
if !l.tryLock() {
// This must happen during a STW, so we can't fail to acquire the lock.
// If we did, something went wrong. Throw.
throw("failed to acquire lock to start a GC transition")
}
if l.gcEnabled == enableGC {
throw("transitioning GC to the same state as before?")
}
// Flush whatever was left between the last update and now.
l.updateLocked(now)
l.gcEnabled = enableGC
l.transitioning = true
// N.B. finishGCTransition releases the lock.
//
// We don't release here to increase the chance that if there's a failure
// to finish the transition, that we throw on failing to acquire the lock.
}
// finishGCTransition notifies the limiter that the GC transition is complete
// and releases ownership of it. It also accumulates STW time in the bucket.
// now must be the timestamp from the end of the STW pause.
func (l *gcCPULimiterState) finishGCTransition(now int64) {
if !l.transitioning {
throw("finishGCTransition called without starting one?")
}
// Count the full nprocs set of CPU time because the world is stopped
// between startGCTransition and finishGCTransition. Even though the GC
// isn't running on all CPUs, it is preventing user code from doing so,
// so it might as well be.
if lastUpdate := l.lastUpdate.Load(); now >= lastUpdate {
l.accumulate(0, (now-lastUpdate)*int64(l.nprocs))
}
l.lastUpdate.Store(now)
l.transitioning = false
l.unlock()
}
// gcCPULimiterUpdatePeriod dictates the maximum amount of wall-clock time
// we can go before updating the limiter.
const gcCPULimiterUpdatePeriod = 10e6 // 10ms
// needUpdate returns true if the limiter's maximum update period has been
// exceeded, and so would benefit from an update.
func (l *gcCPULimiterState) needUpdate(now int64) bool {
return now-l.lastUpdate.Load() > gcCPULimiterUpdatePeriod
}
// addAssistTime notifies the limiter of additional assist time. It will be
// included in the next update.
func (l *gcCPULimiterState) addAssistTime(t int64) {
l.assistTimePool.Add(t)
}
// addIdleTime notifies the limiter of additional time a P spent on the idle list. It will be
// subtracted from the total CPU time in the next update.
func (l *gcCPULimiterState) addIdleTime(t int64) {
l.idleTimePool.Add(t)
}
// update updates the bucket given runtime-specific information. now is the
// current monotonic time in nanoseconds.
//
// This is safe to call concurrently with other operations, except *GCTransition.
func (l *gcCPULimiterState) update(now int64) {
if !l.tryLock() {
// We failed to acquire the lock, which means something else is currently
// updating. Just drop our update, the next one to update will include
// our total assist time.
return
}
if l.transitioning {
throw("update during transition")
}
l.updateLocked(now)
l.unlock()
}
// updatedLocked is the implementation of update. l.lock must be held.
func (l *gcCPULimiterState) updateLocked(now int64) {
lastUpdate := l.lastUpdate.Load()
if now < lastUpdate {
// Defensively avoid overflow. This isn't even the latest update anyway.
return
}
windowTotalTime := (now - lastUpdate) * int64(l.nprocs)
l.lastUpdate.Store(now)
// Drain the pool of assist time.
assistTime := l.assistTimePool.Load()
if assistTime != 0 {
l.assistTimePool.Add(-assistTime)
}
// Drain the pool of idle time.
idleTime := l.idleTimePool.Load()
if idleTime != 0 {
l.idleTimePool.Add(-idleTime)
}
if !l.test {
// Consume time from in-flight events. Make sure we're not preemptible so allp can't change.
//
// The reason we do this instead of just waiting for those events to finish and push updates
// is to ensure that all the time we're accounting for happened sometime between lastUpdate
// and now. This dramatically simplifies reasoning about the limiter because we're not at
// risk of extra time being accounted for in this window than actually happened in this window,
// leading to all sorts of weird transient behavior.
mp := acquirem()
for _, pp := range allp {
typ, duration := pp.limiterEvent.consume(now)
switch typ {
case limiterEventIdleMarkWork:
fallthrough
case limiterEventIdle:
idleTime += duration
case limiterEventMarkAssist:
fallthrough
case limiterEventScavengeAssist:
assistTime += duration
case limiterEventNone:
break
default:
throw("invalid limiter event type found")
}
}
releasem(mp)
}
// Compute total GC time.
windowGCTime := assistTime
if l.gcEnabled {
windowGCTime += int64(float64(windowTotalTime) * gcBackgroundUtilization)
}
// Subtract out all idle time from the total time. Do this after computing
// GC time, because the background utilization is dependent on the *real*
// total time, not the total time after idle time is subtracted.
//
// Idle time is counted as any time that a P is on the P idle list plus idle mark
// time. Idle mark workers soak up time that the application spends idle.
//
// On a heavily undersubscribed system, any additional idle time can skew GC CPU
// utilization, because the GC might be executing continuously and thrashing,
// yet the CPU utilization with respect to GOMAXPROCS will be quite low, so
// the limiter fails to turn on. By subtracting idle time, we're removing time that
// we know the application was idle giving a more accurate picture of whether
// the GC is thrashing.
//
// Note that this can cause the limiter to turn on even if it's not needed. For
// instance, on a system with 32 Ps but only 1 running goroutine, each GC will have
// 8 dedicated GC workers. Assuming the GC cycle is half mark phase and half sweep
// phase, then the GC CPU utilization over that cycle, with idle time removed, will
// be 8/(8+2) = 80%. Even though the limiter turns on, though, assist should be
// unnecessary, as the GC has way more CPU time to outpace the 1 goroutine that's
// running.
windowTotalTime -= idleTime
l.accumulate(windowTotalTime-windowGCTime, windowGCTime)
}
// accumulate adds time to the bucket and signals whether the limiter is enabled.
//
// This is an internal function that deals just with the bucket. Prefer update.
// l.lock must be held.
func (l *gcCPULimiterState) accumulate(mutatorTime, gcTime int64) {
headroom := l.bucket.capacity - l.bucket.fill
enabled := headroom == 0
// Let's be careful about three things here:
// 1. The addition and subtraction, for the invariants.
// 2. Overflow.
// 3. Excessive mutation of l.enabled, which is accessed
// by all assists, potentially more than once.
change := gcTime - mutatorTime
// Handle limiting case.
if change > 0 && headroom <= uint64(change) {
l.overflow += uint64(change) - headroom
l.bucket.fill = l.bucket.capacity
if !enabled {
l.enabled.Store(true)
l.lastEnabledCycle.Store(memstats.numgc + 1)
}
return
}
// Handle non-limiting cases.
if change < 0 && l.bucket.fill <= uint64(-change) {
// Bucket emptied.
l.bucket.fill = 0
} else {
// All other cases.
l.bucket.fill -= uint64(-change)
}
if change != 0 && enabled {
l.enabled.Store(false)
}
}
// tryLock attempts to lock l. Returns true on success.
func (l *gcCPULimiterState) tryLock() bool {
return l.lock.CompareAndSwap(0, 1)
}
// unlock releases the lock on l. Must be called if tryLock returns true.
func (l *gcCPULimiterState) unlock() {
old := l.lock.Swap(0)
if old != 1 {
throw("double unlock")
}
}
// capacityPerProc is the limiter's bucket capacity for each P in GOMAXPROCS.
const capacityPerProc = 1e9 // 1 second in nanoseconds
// resetCapacity updates the capacity based on GOMAXPROCS. Must not be called
// while the GC is enabled.
//
// It is safe to call concurrently with other operations.
func (l *gcCPULimiterState) resetCapacity(now int64, nprocs int32) {
if !l.tryLock() {
// This must happen during a STW, so we can't fail to acquire the lock.
// If we did, something went wrong. Throw.
throw("failed to acquire lock to reset capacity")
}
// Flush the rest of the time for this period.
l.updateLocked(now)
l.nprocs = nprocs
l.bucket.capacity = uint64(nprocs) * capacityPerProc
if l.bucket.fill > l.bucket.capacity {
l.bucket.fill = l.bucket.capacity
l.enabled.Store(true)
l.lastEnabledCycle.Store(memstats.numgc + 1)
} else if l.bucket.fill < l.bucket.capacity {
l.enabled.Store(false)
}
l.unlock()
}
// limiterEventType indicates the type of an event occurring on some P.
//
// These events represent the full set of events that the GC CPU limiter tracks
// to execute its function.
//
// This type may use no more than limiterEventBits bits of information.
type limiterEventType uint8
const (
limiterEventNone limiterEventType = iota // None of the following events.
limiterEventIdleMarkWork // Refers to an idle mark worker (see gcMarkWorkerMode).
limiterEventMarkAssist // Refers to mark assist (see gcAssistAlloc).
limiterEventScavengeAssist // Refers to a scavenge assist (see allocSpan).
limiterEventIdle // Refers to time a P spent on the idle list.
limiterEventBits = 3
)
// limiterEventTypeMask is a mask for the bits in p.limiterEventStart that represent
// the event type. The rest of the bits of that field represent a timestamp.
const (
limiterEventTypeMask = uint64((1<<limiterEventBits)-1) << (64 - limiterEventBits)
limiterEventStampNone = limiterEventStamp(0)
)
// limiterEventStamp is a nanotime timestamp packed with a limiterEventType.
type limiterEventStamp uint64
// makeLimiterEventStamp creates a new stamp from the event type and the current timestamp.
func makeLimiterEventStamp(typ limiterEventType, now int64) limiterEventStamp {
return limiterEventStamp(uint64(typ)<<(64-limiterEventBits) | (uint64(now) &^ limiterEventTypeMask))
}
// duration computes the difference between now and the start time stored in the stamp.
//
// Returns 0 if the difference is negative, which may happen if now is stale or if the
// before and after timestamps cross a 2^(64-limiterEventBits) boundary.
func (s limiterEventStamp) duration(now int64) int64 {
// The top limiterEventBits bits of the timestamp are derived from the current time
// when computing a duration.
start := int64((uint64(now) & limiterEventTypeMask) | (uint64(s) &^ limiterEventTypeMask))
if now < start {
return 0
}
return now - start
}
// type extracts the event type from the stamp.
func (s limiterEventStamp) typ() limiterEventType {
return limiterEventType(s >> (64 - limiterEventBits))
}
// limiterEvent represents tracking state for an event tracked by the GC CPU limiter.
type limiterEvent struct {
stamp atomic.Uint64 // Stores a limiterEventStamp.
}
// start begins tracking a new limiter event of the current type. If an event
// is already in flight, then a new event cannot begin because the current time is
// already being attributed to that event. In this case, this function returns false.
// Otherwise, it returns true.
//
// The caller must be non-preemptible until at least stop is called or this function
// returns false. Because this is trying to measure "on-CPU" time of some event, getting
// scheduled away during it can mean that whatever we're measuring isn't a reflection
// of "on-CPU" time. The OS could deschedule us at any time, but we want to maintain as
// close of an approximation as we can.
func (e *limiterEvent) start(typ limiterEventType, now int64) bool {
if limiterEventStamp(e.stamp.Load()).typ() != limiterEventNone {
return false
}
e.stamp.Store(uint64(makeLimiterEventStamp(typ, now)))
return true
}
// consume acquires the partial event CPU time from any in-flight event.
// It achieves this by storing the current time as the new event time.
//
// Returns the type of the in-flight event, as well as how long it's currently been
// executing for. Returns limiterEventNone if no event is active.
func (e *limiterEvent) consume(now int64) (typ limiterEventType, duration int64) {
// Read the limiter event timestamp and update it to now.
for {
old := limiterEventStamp(e.stamp.Load())
typ = old.typ()
if typ == limiterEventNone {
// There's no in-flight event, so just push that up.
return
}
duration = old.duration(now)
if duration == 0 {
// We might have a stale now value, or this crossed the
// 2^(64-limiterEventBits) boundary in the clock readings.
// Just ignore it.
return limiterEventNone, 0
}
new := makeLimiterEventStamp(typ, now)
if e.stamp.CompareAndSwap(uint64(old), uint64(new)) {
break
}
}
return
}
// stop stops the active limiter event. Throws if the
//
// The caller must be non-preemptible across the event. See start as to why.
func (e *limiterEvent) stop(typ limiterEventType, now int64) {
var stamp limiterEventStamp
for {
stamp = limiterEventStamp(e.stamp.Load())
if stamp.typ() != typ {
print("runtime: want=", typ, " got=", stamp.typ(), "\n")
throw("limiterEvent.stop: found wrong event in p's limiter event slot")
}
if e.stamp.CompareAndSwap(uint64(stamp), uint64(limiterEventStampNone)) {
break
}
}
duration := stamp.duration(now)
if duration == 0 {
// It's possible that we're missing time because we crossed a
// 2^(64-limiterEventBits) boundary between the start and end.
// In this case, we're dropping that information. This is OK because
// at worst it'll cause a transient hiccup that will quickly resolve
// itself as all new timestamps begin on the other side of the boundary.
// Such a hiccup should be incredibly rare.
return
}
// Account for the event.
switch typ {
case limiterEventIdleMarkWork:
gcCPULimiter.addIdleTime(duration)
case limiterEventIdle:
gcCPULimiter.addIdleTime(duration)
sched.idleTime.Add(duration)
case limiterEventMarkAssist:
fallthrough
case limiterEventScavengeAssist:
gcCPULimiter.addAssistTime(duration)
default:
throw("limiterEvent.stop: invalid limiter event type found")
}
}