From 1bbec47bda2740fba22d0bf369d9a1dc6fa059da Mon Sep 17 00:00:00 2001
From: Ti Chi Robot <ti-community-prow-bot@tidb.io>
Date: Wed, 4 Sep 2024 21:37:49 +0800
Subject: [PATCH] schedule: fix datarace in `operator.check` (#8264) (#8579)

close tikv/pd#8263

Signed-off-by: husharp <ihusharp@gmail.com>

Co-authored-by: husharp <ihusharp@gmail.com>
Co-authored-by: Hu# <ihusharp@gmail.com>
---
 server/schedule/operator/operator.go      |  5 +++--
 server/schedule/operator/operator_test.go | 24 +++++++++++++++++++++++
 2 files changed, 27 insertions(+), 2 deletions(-)

diff --git a/server/schedule/operator/operator.go b/server/schedule/operator/operator.go
index 0e760846e64..d4dfb5f929e 100644
--- a/server/schedule/operator/operator.go
+++ b/server/schedule/operator/operator.go
@@ -299,10 +299,11 @@ func (o *Operator) Check(region *core.RegionInfo) OpStep {
 	defer func() { _ = o.CheckTimeout() }()
 	for step := atomic.LoadInt32(&o.currentStep); int(step) < len(o.steps); step++ {
 		if o.steps[int(step)].IsFinish(region) {
-			if atomic.CompareAndSwapInt64(&(o.stepsTime[step]), 0, time.Now().UnixNano()) {
+			current := time.Now()
+			if atomic.CompareAndSwapInt64(&(o.stepsTime[step]), 0, current.UnixNano()) {
 				startTime, _ := o.getCurrentTimeAndStep()
 				operatorStepDuration.WithLabelValues(reflect.TypeOf(o.steps[int(step)]).Name()).
-					Observe(time.Unix(0, o.stepsTime[step]).Sub(startTime).Seconds())
+					Observe(current.Sub(startTime).Seconds())
 			}
 			atomic.StoreInt32(&o.currentStep, step+1)
 		} else {
diff --git a/server/schedule/operator/operator_test.go b/server/schedule/operator/operator_test.go
index 87e44e7a6c6..71fb5e43a9a 100644
--- a/server/schedule/operator/operator_test.go
+++ b/server/schedule/operator/operator_test.go
@@ -18,6 +18,7 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"sync"
 	"sync/atomic"
 	"testing"
 	"time"
@@ -487,3 +488,26 @@ func (suite *operatorTestSuite) TestRecord() {
 	suite.Equal(now, ob.FinishTime)
 	suite.Greater(ob.duration.Seconds(), time.Second.Seconds())
 }
+
+func (suite *operatorTestSuite) TestOperatorCheckConcurrently() {
+	region := suite.newTestRegion(1, 1, [2]uint64{1, 1}, [2]uint64{2, 2})
+	// addPeer1, transferLeader1, removePeer3
+	steps := []OpStep{
+		AddPeer{ToStore: 1, PeerID: 1},
+		TransferLeader{FromStore: 3, ToStore: 1},
+		RemovePeer{FromStore: 3},
+	}
+	op := NewTestOperator(1, &metapb.RegionEpoch{}, OpAdmin|OpLeader|OpRegion, steps...)
+	suite.Equal(core.Urgent, op.GetPriorityLevel())
+	suite.checkSteps(op, steps)
+	op.Start()
+	var wg sync.WaitGroup
+	for i := 0; i < 10; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			suite.Nil(op.Check(region))
+		}()
+	}
+	wg.Wait()
+}