client_secret_path: cli.initConfig: SIGSEGV

[almereyda]

Bug description

When starting a 0.20.0 container (on SQLite), it immediately throws:

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0xb0 pc=0x102e75f]

goroutine 1 [running]:
github.com/juanfont/headscale/cmd/headscale/cli.initConfig()
	/go/src/headscale/cmd/headscale/cli/root.go:59 +0x21f
github.com/spf13/cobra.(*Command).preRun(...)
	/go/pkg/mod/github.com/spf13/[email protected]/command.go:946
github.com/spf13/cobra.(*Command).execute(0x1cc3040, {0x1d075c0, 0x0, 0x0})
	/go/pkg/mod/github.com/spf13/[email protected]/command.go:876 +0x563
github.com/spf13/cobra.(*Command).ExecuteC(0x1cc1f00)
	/go/pkg/mod/github.com/spf13/[email protected]/command.go:1044 +0x3bd
github.com/spf13/cobra.(*Command).Execute(...)
	/go/pkg/mod/github.com/spf13/[email protected]/command.go:968
github.com/juanfont/headscale/cmd/headscale/cli.Execute()
	/go/src/headscale/cmd/headscale/cli/root.go:101 +0x25
main.main()
	/go/src/headscale/cmd/headscale/headscale.go:42 +0x235

This is the container running on GKE on Co-OS nodes.

To Reproduce

Run the 0.20.0 container on Kubernetes.

[almereyda]

This also happens with 0.19.0, and then with pc=0x102e31f. The rest is the same.

With 0.18.0 it now also appears at 0x102cdbf.

I'm able to reproduce the exact same errors locally.

As this happens in cli.initConfig(), this is most likely a parsing error of the configuration file? Let's inspect it.

[almereyda]

This has something to do with client_secret_path. The error does not occur when this directive is not used, but client_secret given instead.

[almereyda]

Related #1127

[traxys]

Happened to me, it seems it was due to invalid permissions on the client_secret_path, changing the owner of that file fixed it.

[almereyda]

Thank you!

I believe we can close here, then.

This is not a use case I need to support anymore.