Golang data race cause by consurrent map read

Question

I have a server to handle events, this server has a mutex lock and a events table(map structure). When the server receives a new event, it will acquire lock to prevent data race, store this event in the events table, and start a goroutine to monitor this event has done. If I run the program with -race flag, it will output data race.

package main

import (
    "sync"
    "time"
)

type event struct {
    done chan bool
}

type server struct {
    mu     sync.Mutex
    events map[int]*event
}

func main() {
    s := server{}
    s.events = make(map[int]*event)

    for i := 0; i < 10; i++ {
        go func(i int) {
            s.mu.Lock()
            s.events[i] = &event{}
            s.events[i].done = make(chan bool)
            s.mu.Unlock()
            go func() {
                time.Sleep(1 * time.Millisecond)
                <-s.events[i].done
                // server do something.
            }()
        }(i)
    }

    time.Sleep(1 * time.Second)

    for i := 0; i < 10; i++ {
        // event happen.
        s.events[i].done <- true
    }
}

Output

==================
WARNING: DATA RACE
Read at 0x00c00010dd10 by goroutine 14:
  runtime.mapaccess1_fast64()
      c:/go/src/runtime/map_fast64.go:12 +0x0    
  main.main.func1.1()
      C:/SimpleAsyncBFT/race/main.go:29 +0x7c    

Previous write at 0x00c00010dd10 by goroutine 15:
  runtime.mapassign_fast64()
      c:/go/src/runtime/map_fast64.go:92 +0x0    
  main.main.func1()
      C:/SimpleAsyncBFT/race/main.go:24 +0xbe    

Goroutine 14 (running) created at:
  main.main.func1()
      C:/SimpleAsyncBFT/race/main.go:27 +0x1c6

Goroutine 15 (finished) created at:
  main.main()
      C:/SimpleAsyncBFT/race/main.go:22 +0xed

I know adding lock in the monitor goroutine will solve this problem, but will cause deadlock! The done channel is just used to notify the server that the event has been done. If channel was not suitable for this condition, how to achieve this?