tunnel.waitGroup.Add() causes a panic under certain circumstances with the following message: sync: negative WaitGroup counter, here is an example stack trace:
goroutine 671916 [running]:
sync.(*WaitGroup).Add(0x21b1084?, 0xc008350701?)
sync/waitgroup.go:64
sync.(*WaitGroup).Done(...)
sync/waitgroup.go:89
github.com/jpillora/chisel/share/tunnel.(*waitGroup).Done(0xc00413f810?)
github.com/jpillora/chisel@v1.10.0/share/tunnel/wg.go:20
github.com/jpillora/chisel/share/tunnel.(*waitGroup).DoneAll(...)
github.com/jpillora/chisel@v1.10.0/share/tunnel/wg.go:26
github.com/jpillora/chisel/share/tunnel.(*Tunnel).BindSSH.func1()
github.com/jpillora/chisel@v1.10.0/share/tunnel/tunnel.go:79We are seeing this problem in real world production workloads. The problem is that Add() is not atomic, in the sense that an intermediate state can be observed by DoneAll() (after the atomic.AddInt32() call but before the inner waitgroup Add() call).
tunnel.waitGroup.Add()causes a panic under certain circumstances with the following message:sync: negative WaitGroup counter, here is an example stack trace:We are seeing this problem in real world production workloads. The problem is that
Add()is not atomic, in the sense that an intermediate state can be observed byDoneAll()(after theatomic.AddInt32()call but before the inner waitgroupAdd()call).