WSL2 kernel panic in Hyper-V netvsc receive path with mirrored networking

### Windows Version

Microsoft Windows [Version 10.0.26200.8655]

### WSL Version

2.6.3.0

### Are you using WSL 1 or WSL 2?

- [x] WSL 2
- [ ] WSL 1

### Kernel Version

6.6.87.2-1

### Distro Version

Ubuntu-26.04

### Other Software

`.wslconfig`:

```ini
[general]
instanceIdleTimeout=86400000

[wsl2]
networkingMode=Mirrored
firewall=false
vmIdleTimeout=86400000
distributionStartTimeout=60000

[experimental]
hostAddressLoopback=true
```

Additional environment details:

- WSLg is enabled.
- Docker Engine is installed inside WSL.
- A physical ext4 Linux SSD is attached with `wsl --mount \\.\PHYSICALDRIVE3 --bare`; the distro mounts its ext4 partition at `/mnt/mint` and bind-mounts `/home/mygod` from it.
- There is a recurring user systemd service that launches Chromium inside WSL/WSLg for browser automation.
- Windows-side virtual adapter state before mitigation:

```text
Name IPv4Enabled IPv6Enabled IPv4OperationalState IPv6OperationalState
vEthernet (External Ethernet) True True False False
Wi-Fi True True False False
vEthernet (Default Switch) True True True True
vEthernet (FSE HostVnic) True True True True
```

Relevant Windows adapters:

```text
vEthernet (FSE HostVnic): Hyper-V Virtual Ethernet Container Adapter, Up, 10 Gbps
vEthernet (Default Switch): Hyper-V Virtual Ethernet Adapter, Up, 10 Gbps
Ethernet: Realtek Gaming 2.5GbE Family Controller, Up, 2.5 Gbps
WireGuard Tunnel: Up
VMware VMnet1/VMnet8 adapters: Up
Wi-Fi: Disconnected
```

### Repro Steps

I do not yet have a deterministic minimal reproducer. This is the observed sequence:

1. Run WSL2 with `networkingMode=Mirrored`.
2. Keep the WSL VM running for several days with systemd, Docker, WSLg, and Chromium-based browser automation.
3. At `2026-07-03 03:32:10 -04:00`, a systemd user service started a Chromium run inside WSL.
4. At `2026-07-03 03:32:57 -04:00`, WSL wrote a kernel panic file under `%LOCALAPPDATA%\Temp\wsl-crashes`.
5. The WSL VM terminated/restarted. All open WSL terminal tabs exited and running Linux processes were killed.

The immediately preceding distro journal entries were:

```text
2026-07-03T03:32:10-04:00 systemd[289]: Starting service...
2026-07-03T03:32:11-04:00 yarn[3130739]: yarn run v1.22.22
2026-07-03T03:32:15-04:00 systemd[289]: Started app-org.chromium.Chromium-3130789.scope.
```

There was no Linux shutdown transaction, OOM-killer entry, or systemd poweroff/reboot log before the journal stopped.

### Expected Behavior

WSL2 network packet receive processing should not be able to panic the WSL kernel. A user-space Chromium/network workload should either succeed or fail in user space without terminating the entire WSL VM.

### Actual Behavior

The WSL VM kernel panicked in the Hyper-V network receive path and the VM was terminated/restarted.

User-visible symptoms:

- Existing WSL terminal tabs exited.
- Running Linux processes were killed.
- After the VM restart, the physical disk attachment was gone, so the distro initially got stuck waiting for mounts until the disk was reattached.

Crash artifact:

```text
Temp\wsl-crashes\kernel-panic-1783063977-{f9f13f23-e524-4dd6-9393-1a6883bfb763}.txt
LastWriteTime: 2026-07-03 03:32:57 -04:00
Length: 3956 bytes
```

Important excerpt:

```text
[300418.637096] Call Trace:
[300418.637129]
[300418.637130] ? netvsc_recv_callback+0x32e/0x3c0
[300418.637961] rndis_filter_receive+0x216/0x840
[300418.637967] netvsc_poll+0x2ac/0xba0
[300418.637968] ? ttwu_do_activate+0x64/0x210
[300418.638066] __napi_poll+0x2e/0x1c0
[300418.638561] net_rx_action+0x2a8/0x380
[300418.638563] handle_softirqs+0xf1/0x2e0
[300418.638597] __irq_exit_rcu+0x78/0xd0
[300418.638598] irq_exit_rcu+0x12/0x20
[300418.638599] sysvec_hyperv_callback+0xb4/0xd0
[300418.638609]
...
[300418.639652] Modules linked in: nf_conntrack_netlink xt_nat xt_conntrack xt_MASQUERADE xfrm_user xfrm_algo xt_set ip_set xt_addrtype nft_compat intel_rapl_msr intel_rapl_common kvm_intel kvm irqbypass crc32c_intel sch_fq_codel configfs autofs4 br_netfilter bridge stp llc ip_tables tun
[300418.640463] ---[ end trace 0000000000000000 ]---
[300418.640759] RIP: 0010:memcpy+0x11/0x20
...
[300418.640770] Kernel panic - not syncing: Fatal exception in interrupt
```

Windows System event log at the same time only shows Hyper-V virtual switch NIC/port teardown and recreation. No Windows reboot, sleep/wake transition, storage error, or resource exhaustion event was found around the crash timestamp.

P.S. As a local mitigation, I am considering disabling Receive Segment Coalescing (RSC) on `vEthernet (FSE HostVnic)`, since RSC is currently enabled and the panic occurred in the receive path.

### Diagnostic Logs

I can attach the file:

```text
kernel-panic-1783063977-{f9f13f23-e524-4dd6-9393-1a6883bfb763}.txt
```

I can also provide `journalctl -b -1` output around the timestamp if useful.