[FlareCoding]

Summary

• The virtio PCI transport lived inside the network driver: capability parsing, the status handshake, feature negotiation, queue setup and MSI-X routing. A second virtio device would have had to copy all of it.
• The transport moves into a virtio_pci_driver base class in drivers/virtio, next to the shared PCI definitions and the virtqueue. virtio-net derives from it and behaves as before.

---

Note

Medium Risk
Refactors kernel virtio-net bring-up and MMIO/virtqueue setup without intended behavior change; regressions would affect network attach and PCI feature negotiation.

Overview
Extracts the virtio modern PCI transport (capability walk, BAR mapping, status handshake, feature negotiation, virtqueue registration, MSI-X routing helpers, ISR ack, device reset) from virtio_net into a new virtio_pci_driver base under drivers/virtio. virtio-net now derives from that class and keeps only net-specific attach/queue buffer logic, calling start_transport(), setup_queue(), set_driver_ok(), and device_config() instead of inlined transport code.

Shared virtio pieces are relocated and renamed: generic PCI definitions stay in virtio_pci.h (net-only constants/structs move to virtio_net.h), and the virtqueue implementation moves from drivers/net to drivers/virtio/virtqueue.*. Feature negotiation in the base class now takes required vs optional feature masks; virtio-net treats MAC, STATUS, and VERSION_1 as optional and reads negotiated bits from the returned mask. Link-up checks use VIRTIO_NET_S_LINK_UP instead of a raw status bit.

^{Reviewed by Cursor Bugbot for commit 63d1c32. Bugbot is set up for automated code reviews on this repo. Configure here.}