`panic: runtime/cgo: misuse of an invalid Handle` (`Code-Hex/vz/v3.changeStateOnObserver`)

[AkihiroSuda]

$ limactl start template://experimental/vz

$ limactl stop vz
INFO[0000] Sending SIGINT to hostagent process 47742    
INFO[0000] Waiting for the host agent and the driver processes to shut down 
INFO[0000] [hostagent] 2023/02/24 17:59:43 tcpproxy: for incoming conn 127.0.0.1:51498, error dialing "192.168.5.15:22": connect tcp 192.168.5.15:22: no route to host 
INFO[0000] [hostagent] 2023/02/24 17:59:53 tcpproxy: for incoming conn 127.0.0.1:51499, error dialing "192.168.5.15:22": connect tcp 192.168.5.15:22: connection was refused 
INFO[0000] [hostagent] Received SIGINT, shutting down the host agent 
INFO[0000] [hostagent] Shutting down the host agent     
INFO[0000] [hostagent] Stopping forwarding "/run/lima-guestagent.sock" (guest) to "/Users/suda/.lima/vz/ga.sock" (host) 
INFO[0000] [hostagent] Shutting down VZ                 
ERRO[0000] [hostagent] dhcp: unhandled message type: RELEASE 
INFO[0001] [hostagent] panic: runtime/cgo: misuse of an invalid Handle 
INFO[0001] [hostagent] goroutine 17 [running, locked to thread]: 
INFO[0001] [hostagent] runtime/cgo.Handle.Value(...)    
INFO[0001] [hostagent]  /usr/local/Cellar/go/1.19.6/libexec/src/runtime/cgo/handle.go:124 
INFO[0001] [hostagent] github.com/Code-Hex/vz/v3.changeStateOnObserver(0x0, 0xc000006801?) 
INFO[0001] [hostagent]  /Users/suda/gopath/pkg/mod/github.com/!code-!hex/vz/v3@v3.0.4/virtualization.go:160 +0xec 

version: e939454

[AkihiroSuda]

Looks like this happens only on stopping a new instance

[balajiv113]

Will check this out

[balajiv113]

Debugged this, this is from upstream only. Will raise a ticket in vz. I tried to do some workaround but none worked.

[jandubois]

Looks like this happens only on stopping a new instance

Yes, looks like that for me too. It also hangs for almost another 3 minutes after the panic() before exiting the process. This seems to be inside the vz package, so I don't think we could capture it and exit right away (maybe not worth doing anyways).

After restarting the instance, I can shut it down normally without a panic, but still get an error from DHCP:

ERRO[0001] [hostagent] dhcp: unhandled message type: RELEASE

[balajiv113]

but still get an error from DHCP:
DHCP error is from gvisor-tap-vsock.

[dmigwi]

I experienced the same error working on a separate product on Mac Os. The crashes also happened severally when I was shutting down the application. After updating the o.s to Monterey the crash seems to have disappeared by I want to understand what exactly happened.

[jandubois]

@dmigwi This issue is about macOS 13 (Ventura) because there is no Virtualization framework support in Lima for earlier macOS version (we only support EFI boot, which was added in Ventura). So if your problem with the vz library was fixed by Monterey, then this is a separate issue.

[dmigwi]

@jandubois Thanks for letting me know that is issue is not yet resolved since Ventura is the latest Mac Os release. A weird thing that happened when I experienced that crash on my product (desktop app) is my mac was weirdly hot and the fan was humming way louder than usual. This led me to believe something in the background or o.s. level was pushing my CPU to its extreme limit.

I tend to believe the golang GC kicked in to clean up memory space it thought wasn't needed anymore since space was kinda being stretched thin then. The communication between golang and cgo is via memory locations inform of integers, so when cgo attempted to resolve the memory location that turned out to reused by something else that crash happened then. Its all documented here: https://golang.design/research/cgo-handle/

My initial hypothesis is that the crash happens when the garbage collector runs and cleans up memory locations its not expected to mess with. I could be wrong on this!

[jandubois]

@dmigwi Please report your issue on the https://github.com/Code-Hex/vz repo if you have additional information to reproduce the problem.

This repo is purely about Lima, so we will not be able to help you with your project here, since you seem to be experiencing a separate issue.