Understanding goroutines synchronization

Question

I am trying to understand the golang channels and synchronization. When I run my program with race detector, it results in race detection.

My program:

func main() {
    ch := make(chan int)
    done := make(chan struct{})
    wg := sync.WaitGroup{}

    go func() {
        defer close(ch)
        defer close(done)
        wg.Wait()
        done <- struct{}{}
    }()

    for i := 0; i < 5; i++ {
        x := i
        wg.Add(1)
        go func() {
            defer wg.Done()
            fmt.Println("Value: ", x)
            ch <- x
        }()
    }
    
loop:
    for {
        select {
        case i := <-ch:
            fmt.Println("Value: ", i)
        case <- done:
            break loop
        }
    }
}

Race detector report:

==================
WARNING: DATA RACE
Write at 0x00c000020148 by goroutine 7:
  internal/race.Write()
      /home/linuxbrew/.linuxbrew/Cellar/go/1.16.5/libexec/src/internal/race/race.go:41 +0x125
  sync.(*WaitGroup).Wait()
      /home/linuxbrew/.linuxbrew/Cellar/go/1.16.5/libexec/src/sync/waitgroup.go:128 +0x126
  main.main.func1()
      /home/reddy/code/github.com/awesomeProject/prod.go:106 +0xc4

Previous read at 0x00c000020148 by main goroutine:
  internal/race.Read()
      /home/linuxbrew/.linuxbrew/Cellar/go/1.16.5/libexec/src/internal/race/race.go:37 +0x206
  sync.(*WaitGroup).Add()
      /home/linuxbrew/.linuxbrew/Cellar/go/1.16.5/libexec/src/sync/waitgroup.go:71 +0x219
  main.main()
      /home/reddy/code/github.com/awesomeProject/prod.go:112 +0x124

Goroutine 7 (running) created at:
  main.main()
      /home/reddy/code/github.com/awesomeProject/prod.go:103 +0x104
==================

I am not able to figure out what's going wrong here.

My analysis:

1. wg.Add(1) is incrementing the counter
2. wg.Done() is called at the end of goroutine which decrements the counter
3. ch <- x this should be a blocking call as it's non buffered channel
4. loop should iterate till done channel has some message which happens when the waitgroup counter goes to zero, i.e. all the 5 goroutines published the message
5. once the counter goes to zero, wg goroutine will resume and done is called and once the message is consumed in the main loop, it breaks the loop and should gracefully exit.

Answer 1

The program has a race between the calls to wg.Add and the call to wg.Wait. These calls can happen in any order. The call to wg.Wait does not wait for any of the goroutines when wg.Wait is called before the calls to wg.Add.

Fix by moving the calls to wg.Add before starting the goroutine that calls wg.Wait. This change ensures that the calls to wg.Add happen before the call to wg.Wait.

for i := 0; i < 5; i++ {
    x := i
    wg.Add(1)
    go func() {
        defer wg.Done()
        fmt.Println("Value: ", x)
        ch <- x
    }()
}

go func() {
    defer close(ch)
    defer close(done)
    wg.Wait()
    done <- struct{}{}
}()

The WaitGroup type has code to check for this error when running under the race detector (modeled read, modeled write).

Simplify the code by breaking out of the loop in the main goroutine when ch is closed. The done channel is not needed.

ch := make(chan int)
wg := sync.WaitGroup{}

for i := 0; i < 5; i++ {
    x := i
    wg.Add(1)
    go func() {
        defer wg.Done()
        fmt.Println("Value: ", x)
        ch <- x
    }()
}

go func() {
    wg.Wait()
    close(ch)
}()

for i := range ch {
    fmt.Println("Value: ", i)
}