Go: cmd/compile: uncaught RuntimeError: memory access out of bounds

Created on 26 Mar 2020  路  16Comments  路  Source: golang/go

What version of Go are you using (go version)?

$ go version
go version go1.14.1 darwin/amd64

Does this issue reproduce with the latest release?

Yes

What operating system and processor architecture are you using (go env)?

go env Output

$ go env
GO111MODULE=""
GOARCH="amd64"
GOBIN=""
GOCACHE="/Users/hajimehoshi/Library/Caches/go-build"
GOENV="/Users/hajimehoshi/Library/Application Support/go/env"
GOEXE=""
GOFLAGS=""
GOHOSTARCH="amd64"
GOHOSTOS="darwin"
GOINSECURE=""
GONOPROXY=""
GONOSUMDB=""
GOOS="darwin"
GOPATH="/Users/hajimehoshi/go"
GOPRIVATE=""
GOPROXY="https://proxy.golang.org,direct"
GOROOT="/usr/local/go"
GOSUMDB="sum.golang.org"
GOTMPDIR=""
GOTOOLDIR="/usr/local/go/pkg/tool/darwin_amd64"
GCCGO="gccgo"
AR="ar"
CC="clang"
CXX="clang++"
CGO_ENABLED="1"
GOMOD=""
CGO_CFLAGS="-g -O2"
CGO_CPPFLAGS=""
CGO_CXXFLAGS="-g -O2"
CGO_FFLAGS="-g -O2"
CGO_LDFLAGS="-g -O2"
PKG_CONFIG="pkg-config"
GOGCCFLAGS="-fPIC -m64 -pthread -fno-caret-diagnostics -Qunused-arguments -fmessage-length=0 -fdebug-prefix-map=/var/folders/ht/ky_bwgzs4bd5z1hh02k34x_h0000gn/T/go-build937690981=/tmp/go-build -gno-record-gcc-switches -fno-common"

What did you do?

Build this Go program as Wasm and run it on a browser several times:

package main

import (
        "fmt"
        "time"
)

func main() {
        ch1 := make(chan string)
        ch2 := make(chan string)
        ch3 := make(chan struct{})

        go func() {
                for {
                        time.Sleep(5 * time.Millisecond)
                        ch1 <- "A"
                }
        }()
        go func() {
                for {
                        time.Sleep(7 * time.Millisecond)
                        ch2 <- "B"
                }
        }()
        go func() {
                for {
                        time.Sleep(3 * time.Second)
                        close(ch3)
                }
        }()

loop:
        for {
                select {
                case str := <-ch1:
                        fmt.Println(str)
                case str := <-ch2:
                        fmt.Println(str)
                case <-ch3:
                        fmt.Println("Done")
                        break loop
                }
        }
}

What did you expect to see?

No error (except for 'already exit')

What did you see instead?

Sometimes I saw this error:

Uncaught RuntimeError: memory access out of bounds
    at runtime.__waitq_.dequeue (wasm-function[113]:0xbe09)
    at runtime.chanrecv (wasm-function[109]:0xab34)
    at runtime.chanrecv1 (wasm-function[108]:0xa89b)
    at syscall.fsCall (wasm-function[1636]:0x12510e)
    at syscall.Write (wasm-function[1634]:0x12471b)
    at internal_poll.__FD_.Write (wasm-function[1657]:0x128d1e)
    at os.__File_.Write (wasm-function[1669]:0x129ef2)
    at fmt.Fprintln (wasm-function[1700]:0x131eb1)
    at main.main (wasm-function[1720]:0x14065c)
    at wasm_pc_f_loop (wasm-function[950]:0xc3a5f)


  | runtime.__waitq_.dequeue | @ | 008b857e-113:116
聽 | runtime.chanrecv | @ | 008b857e-109:249
聽 | runtime.chanrecv1 | @ | 008b857e-108:37
聽 | syscall.fsCall | @ | 008b857e-1636:354
聽 | syscall.Write | @ | 008b857e-1634:451
聽 | internal_poll.__FD_.Write | @ | 008b857e-1657:385
聽 | os.__File_.Write | @ | 008b857e-1669:164
聽 | fmt.Fprintln | @ | 008b857e-1700:170
聽 | main.main | @ | 008b857e-1720:471
聽 | wasm_pc_f_loop | @ | 008b857e-950:14
聽 | wasm_export_resume | @ | 008b857e-949:5
聽 | _resume | @ | wasm_exec.js:544
聽 | (anonymous) | @ | wasm_exec.js:308
聽 | setTimeout (async) | 聽 | 聽
聽 | runtime.scheduleTimeoutEvent | @ | wasm_exec.js:306
聽 | runtime.scheduleTimeoutEvent | @ | 008b857e-969:6
聽 | runtime.beforeIdle | @ | 008b857e-158:107
聽 | runtime.findrunnable | @ | 008b857e-564:1278
聽 | runtime.schedule | @ | 008b857e-569:960
聽 | runtime.park_m | @ | 008b857e-572:245
聽 | runtime.mcall | @ | 008b857e-899:77
聽 | runtime.gopark | @ | 008b857e-515:312
聽 | runtime.selectgo | @ | 008b857e-655:2335
聽 | main.main | @ | 008b857e-1720:349
聽 | wasm_pc_f_loop | @ | 008b857e-950:14
聽 | wasm_export_resume | @ | 008b857e-949:5
聽 | _resume | @ | wasm_exec.js:544
聽 | (anonymous) | @ | wasm_exec.js:308
...

Arch-Wasm NeedsInvestigation

Most helpful comment

I can also reproduce it. It seems y.prev in runtime.(*waitq).dequeue is pointing to some garbage, a huge number beyond the address space. And it is also reproducible without the second channel (ch2), with a different error where it panics in with nil channel in chanrecv, called from syscall.fsCall. This channel is clearly non-nil, as it is created a few lines ago. Looks like some kind of memory corruption to me. I haven't figured out the cause, though.

All 16 comments

CC @neelance

I could not confirm this issue on node.js, but could confirm on Chrome browser (Version 80.0.3987.149 (Official Build) (64-bit))

CC @dmitshur

Any updates?

I can also reproduce it. It seems y.prev in runtime.(*waitq).dequeue is pointing to some garbage, a huge number beyond the address space. And it is also reproducible without the second channel (ch2), with a different error where it panics in with nil channel in chanrecv, called from syscall.fsCall. This channel is clearly non-nil, as it is created a few lines ago. Looks like some kind of memory corruption to me. I haven't figured out the cause, though.

This doesn't happen with Go 1.13.9. I think this is a regression. I'll try to bisect.

6becb033341602f2df9d7c55cc23e64b925bbee2 seems the culprit: from this commit, I saw an error, which is different from the original error though:

fatal error: unexpected signal during runtime execution
wasm_exec.js:47 
runtime stack:
wasm_exec.js:47 runtime: unexpected return pc for syscall.fsCall called from 0x160258670d0f8fc0
wasm_exec.js:47 stack: frame={sp:0x156ac0, fp:0x156b58} stack=[0x154fd8,0x156bd8)
wasm_exec.js:47 00000000001569c0:  0000000000038ab1  000000000000002a 
wasm_exec.js:47 00000000001569d0:  000000000013bbc0  0000000010bf0003 <runtime.typedmemmove+3> 
wasm_exec.js:47 00000000001569e0:  0000000000000001  160258670ca48400 
wasm_exec.js:47 00000000001569f0:  0000000000000001  000000000042c360 
wasm_exec.js:47 0000000000156a00:  000000000042c3b0  000000001057001a <runtime.chanrecv+26> 
wasm_exec.js:47 0000000000156a10:  0000000000000000  0000000000156b38 
wasm_exec.js:47 0000000000156a20:  0000000000000007  0000000013240000 <runtime.(*mheap).coalesce.func2+0> 
wasm_exec.js:47 0000000000156a30:  0000000010880010 <runtime.beforeIdle+16>  0000000000000000 
wasm_exec.js:47 0000000000156a40:  0000000000000002  0000000000000007 
wasm_exec.js:47 0000000000156a50:  000000000046b9c0  0000000000000001 
wasm_exec.js:47 0000000000156a60:  1602586700000001  160258670d0f8fc0 
wasm_exec.js:47 0000000000156a70:  0000000000161800  0000000000410058 
wasm_exec.js:47 0000000000156a80:  0000000000000000  0000000000000001 
wasm_exec.js:47 0000000000156a90:  0000000010560002 <runtime.chanrecv1+2>  0000000000410000 
wasm_exec.js:47 0000000000156aa0:  0000000000156b38  0000000000000001 
wasm_exec.js:47 0000000000156ab0:  0000000011e60010 <runtime.casgstatus+16>  000000001626000d <syscall.fsCall+13> 
wasm_exec.js:47 0000000000156ac0: <0000000000410000  0000000000156b38 
wasm_exec.js:47 0000000000156ad0:  000000000042c360  0000000100000001 
wasm_exec.js:47 0000000000156ae0:  0000000100000000  000000000042c3b0 
wasm_exec.js:47 0000000000156af0:  0000000000000000  00000000004105d8 
wasm_exec.js:47 0000000000156b00:  1602586700000007  0000000000410000 
wasm_exec.js:47 0000000000156b10:  0000000200000001  000000000013bbc0 
wasm_exec.js:47 0000000000156b20:  160258670cc8d600  0000000000410000 
wasm_exec.js:47 0000000000156b30:  000000001209005e <runtime.schedule+94>  0000000000000000 
wasm_exec.js:47 0000000000156b40:  0000000000000000  0000000000000000 
wasm_exec.js:47 0000000000156b50: !160258670d0f8fc0 >0000000012ed0000 <runtime.isSystemGoroutine+0> 
wasm_exec.js:47 0000000000156b60:  0000000000400900  0000000000000000 
wasm_exec.js:47 0000000000156b70:  0000000000000000  000000000013bbc0 
wasm_exec.js:47 0000000000156b80:  0000000010590004 <runtime.chanparkcommit+4>  000000000042e0b8 
wasm_exec.js:47 0000000000156b90:  00000000120c001b <runtime.park_m+27>  0000000000400900 
wasm_exec.js:47 0000000000156ba0:  000000000042e0b8  0000000000000001 
wasm_exec.js:47 0000000000156bb0:  000000000013bbc0  00000000134d0001 <runtime.mcall+1> 
wasm_exec.js:47 0000000000156bc0:  0000000000400900  0000000000156c00 
wasm_exec.js:47 0000000000156bd0:  0100000000000000 
wasm_exec.js:47 syscall.fsCall(0x12ed0000, 0x400900, 0x0, 0x0, 0x13bbc0, 0x10590004, 0x42e0b8, 0x120c001b)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/syscall/fs_js.go:513 +0xd
wasm_exec.js:47 
wasm_exec.js:47 goroutine 1 [waiting]:
wasm_exec.js:47 syscall/js.Value.Call(0x7ff800000000000a, 0x32e54, 0x5, 0x452000, 0x6, 0xa, 0x5)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/syscall/js/js.go:326 +0x3
wasm_exec.js:47 syscall.fsCall(0x32e54, 0x5, 0x436cb8, 0x5, 0x5, 0x44e001, 0x430a90, 0x41e008)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/syscall/fs_js.go:496 +0xc
wasm_exec.js:47 syscall.Write(0x1, 0x41e008, 0x2, 0x8, 0x1440000e, 0x12da18, 0x0)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/syscall/fs_js.go:417 +0x11
wasm_exec.js:47 internal/poll.(*FD).Write(0x42c120, 0x41e008, 0x2, 0x8, 0x0, 0x0, 0x0)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/internal/poll/fd_unix.go:268 +0x22
wasm_exec.js:47 os.(*File).write(...)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/os/file_unix.go:276
wasm_exec.js:47 os.(*File).Write(0x40c020, 0x41e008, 0x2, 0x8, 0x436e18, 0x13410000, 0x10770005)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/os/file.go:153 +0xf
wasm_exec.js:47 fmt.Fprintln(0x4a200, 0x40c020, 0x436ec8, 0x1, 0x1, 0x13a20, 0x156c08, 0x0)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/fmt/print.go:265 +0x8
wasm_exec.js:47 fmt.Println(...)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/fmt/print.go:274
wasm_exec.js:47 main.main()
wasm_exec.js:47     /Users/hajimehoshi/test/wasmbug/main.go:38 +0xf
wasm_exec.js:47 
wasm_exec.js:47 goroutine 6 [chan send]:
wasm_exec.js:47 main.main.func1(0x42e060)
wasm_exec.js:47     /Users/hajimehoshi/test/wasmbug/main.go:16 +0x4
wasm_exec.js:47 created by main.main
wasm_exec.js:47     /Users/hajimehoshi/test/wasmbug/main.go:13 +0x5
wasm_exec.js:47 
wasm_exec.js:47 goroutine 7 [sleep]:
wasm_exec.js:47 time.Sleep(0x6acfc0)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/runtime/time.go:247 +0xe
wasm_exec.js:47 main.main.func2(0x42e0c0)
wasm_exec.js:47     /Users/hajimehoshi/test/wasmbug/main.go:21 +0x3
wasm_exec.js:47 created by main.main
wasm_exec.js:47     /Users/hajimehoshi/test/wasmbug/main.go:19 +0x6
wasm_exec.js:47 
wasm_exec.js:47 goroutine 8 [sleep]:
wasm_exec.js:47 time.Sleep(0xb2d05e00)
wasm_exec.js:47     /Users/hajimehoshi/go-code/src/runtime/time.go:247 +0xe
wasm_exec.js:47 main.main.func3(0x42e120)
wasm_exec.js:47     /Users/hajimehoshi/test/wasmbug/main.go:27 +0x3
wasm_exec.js:47 created by main.main
wasm_exec.js:47     /Users/hajimehoshi/test/wasmbug/main.go:25 +0x7

After 6becb033341602f2df9d7c55cc23e64b925bbee2, c2e2296dd7dea68de4a45d8dc82b1920130eb74a is the first commit to see the same error as reported.

CC @ianlancetaylor

I think that somebody familiar with wasm is going to have to dig into this. My contribution here is https://golang.org/cl/171827 and https://golang.org/cl/203890. It sounds like the new timers have broken some garbage collection invariant, but I don't know what it would be. Sorry.

Thanks @hajimehoshi and @ianlancetaylor ! I could probably look into this a little more, but maybe not today. I'm having some issue running wasm on my browser...

@cherrymui Hi, what's going on this? Thanks.

This is the same issue as #38574, please see my analysis there.

Change https://golang.org/cl/230178 mentions this issue: runtime: fix race condition between timer and event handler

Thanks.

I was wondering if this will be backported to 1.14 or not since this is a memory-corruption error.

@hajimehoshi Seems unlikely, because js/wasm is still experimental. See https://github.com/golang/go/wiki/MinorReleases for the policy on backports.

Ah OK, that's unfortunate but makes sense. Thanks!

Was this page helpful?
0 / 5 - 0 ratings