runtime: map race with *[]byte as index/key returns nil map value when map value is not nil

[andrewhodel]

Go version

go version go1.25.3 linux/amd64

Output of go env in your module/workspace:

AR='ar'
CC='gcc'
CGO_CFLAGS='-O2 -g'
CGO_CPPFLAGS=''
CGO_CXXFLAGS='-O2 -g'
CGO_ENABLED='1'
CGO_FFLAGS='-O2 -g'
CGO_LDFLAGS='-O2 -g'
CXX='g++'
GCCGO='gccgo'
GO111MODULE=''
GOAMD64='v1'
GOARCH='amd64'
GOAUTH='netrc'
GOBIN=''
GOCACHE='/home/zip/.cache/go-build'
GOCACHEPROG=''
GODEBUG=''
GOENV='/home/zip/.config/go/env'
GOEXE=''
GOEXPERIMENT=''
GOFIPS140='off'
GOFLAGS=''
GOGCCFLAGS='-fPIC -m64 -pthread -Wl,--no-gc-sections -fmessage-length=0 -ffile-prefix-map=/tmp/go-build1641070464=/tmp/go-build -gno-record-gcc-switches'
GOHOSTARCH='amd64'
GOHOSTOS='linux'
GOINSECURE=''
GOMOD='/dev/null'
GOMODCACHE='/home/zip/go/pkg/mod'
GONOPROXY=''
GONOSUMDB=''
GOOS='linux'
GOPATH='/home/zip/go'
GOPRIVATE=''
GOPROXY='https://proxy.golang.org,direct'
GOROOT='/home/zip/xyzbots/go'
GOSUMDB='sum.golang.org'
GOTELEMETRY='local'
GOTELEMETRYDIR='/home/zip/.config/go/telemetry'
GOTMPDIR=''
GOTOOLCHAIN='auto'
GOTOOLDIR='/home/zip/xyzbots/go/pkg/tool/linux_amd64'
GOVCS=''
GOVERSION='go1.25.3'
GOWORK=''
PKG_CONFIG='pkg-config'

What did you do?

package main

import (
	"fmt"
	"math/rand"
	"sync"
	"time"
)

func main() {
	var l = make(map[*[]byte]map[string]string)

	var ll = make([]byte, 0)
	var existing_pointer *[]byte = &ll
	l[existing_pointer] = make(map[string]string)

	var m sync.Mutex

	var c = 0
	for {
		if c == 20000 {
			break
		}

		go func() {

			time.Sleep(time.Second * time.Duration(rand.Intn(10)))

			m.Lock()

			var another_pointer *[]byte
			for e := range l {
				// this is the same pointer as existing_pointer
				another_pointer = e
			}

			if l[another_pointer] == nil {
				// this can happen multiple times incorrectly
				fmt.Println("another_pointer is not in l")
				l[another_pointer] = make(map[string]string)
			}

			m.Unlock()

		}()

		c += 1
	}
}

What did you see happen?

fmt.Println("another_pointer is not in l")

This is executed multiple times.

What did you expect to see?

fmt.Println("another_pointer is not in l")

This executed once.

[andrewhodel]

Hardware

15.6" Ноутбук HUAWEI MateBook D 15 i3, Intel Core i3-1115G4 (3.0 ГГц), RAM 8 ГБ, SSD 256 ГБ, Intel UHD Graphics, Windows Home, (53013ERV), Mystic Silver, Российская клавиатура

[Jorropo]

For me this program exit almost instantly after creating all the goroutines which on my machine happens in 100ms since you do not block at the end of main.

I have modified it:

package main

import (
	"fmt"
	"math/rand/v2"
	"sync"
	"time"
)

func main() {
	var l = make(map[*[]byte]map[string]string)

	var ll = make([]byte, 0)
	var existing_pointer *[]byte = &ll
	l[existing_pointer] = make(map[string]string)

	var m sync.Mutex
	var wg sync.WaitGroup

	var c = 0
	for {
		if c == 20000 {
			break
		}

		wg.Add(1)
		go func() {
			defer wg.Done()

			time.Sleep(time.Second * time.Duration(rand.IntN(10)))

			m.Lock()

			var another_pointer *[]byte
			for e := range l {
				// this is the same pointer as existing_pointer
				another_pointer = e
			}

			if l[another_pointer] == nil {
				// this can happen multiple times incorrectly
				fmt.Println("another_pointer is not in l")
				l[another_pointer] = make(map[string]string)
			}

			m.Unlock()

		}()

		c += 1
	}
	wg.Wait()
}

I cannot get a repro either way.

[gabyhelp]

Related Issues

• map: a key-value pair may be GC #38976 (closed)
• go/build: Why is replacing the map key not working as expected? #76805 (closed)
• go/map #59917 (closed)
• sync: Different behavior between sync.Map and normal map paired with unsafe.String #70536 (closed)
• Unexpected Pointer Behavior in `range` Loop Over Maps in Go (Go 1.24.1) #73138 (closed)
• sync: Map CompareAndSwap is not atomic with respect to other map operations in Go 1.24 #70970 (closed)
• gollvm: cannot use unsafe.Pointer as map key type #52861
• runtime: panic in mapaccess2_faststr (suspected data race) #65115 (closed)
• go bug sync.Map #32641 (closed)
• Map protected by mutex caused panic: concurrent map writes #20060 (closed)

_{(Emoji vote if this was helpful or unhelpful; more detailed feedback welcome in this discussion.)}

[andrewhodel]

This was the result of a mutex name change and I thought it was the hardware.

Not a bug.

[andrewhodel]

Still the same problem with certainty of the mutex.

If I track/keep the pointer that is used as the map key in a []*[]byte slice, it works as expected.

[ianlancetaylor]

Please attach a new standalone test case that shows the bug and doesn't immediately exit. Thanks.

[andrewhodel]

Please attach a new standalone test case that shows the bug and doesn't immediately exit. Thanks.

https://go.dev/play/p/BAp1u7JtDkm

package main

import (
	"fmt"
	"math/rand"
	"os"
	"sync"
	"time"
)

func main() {
	var l = make(map[*[]byte]map[string]string)

	var ll = make([]byte, 0)
	var existing_pointer *[]byte = &ll
	l[existing_pointer] = make(map[string]string)

	var m sync.Mutex

	var c = 0
	for {
		if c == 20000 {
			os.Exit(0)
			break
		}

		go func() {

			time.Sleep(time.Second * time.Duration(rand.Intn(10)))

			m.Lock()

			var another_pointer *[]byte
			for e := range l {
				// this is the same pointer as existing_pointer
				another_pointer = e
			}

			if l[another_pointer] == nil {
				// this can happen multiple times incorrectly
				fmt.Println("another_pointer is not in l")
				l[another_pointer] = make(map[string]string)
			}

			m.Unlock()

		}()

		c += 1
	}

	select {}
}

This happens only on the specific architecture (hardware) that is documented in the primary report. It works in the playground.

[randall77]

I cannot reproduce the failure on linux/amd64 with go1.25.3 (or a few other Go versions I tried) using the last posted code.
I also cannot reproduce on darwin/arm64.

How often does the failure reproduce for you? Every run?

Is there anything special about your machine that we should know? Is it real hardware or is it a virtual machine of some sort?

[seankhliao]

what are the actual pointer values (print them out)?