v0.7: ethtool stats + per-protocol counters

v0.7

Highlights

• Adds ethtool -S extended statistics for vnet0
• Adds per-protocol TX/RX counters (ARP / IPv4 / IPv6 / other)
• Improves observability via TX EtherType debug logging
• Hardens RX ring behavior under pressure (drops tracked, skb freed)

How to Try

make clean && make
sudo insmod src/vnet_main.ko
sudo ip link set vnet0 up
sudo tcpdump -i vnet0 -e -n arp
ethtool -S vnet0

v0.6 — Driver Statistics & ethtool Support

v0.6 — Driver Statistics & ethtool Support

Highlights

• Added 64-bit TX/RX counters
• Implemented ndo_get_stats64
• Added basic ethtool ops (get_drvinfo, get_link)
• Updated driver metadata and documentation
• Bumped module version to 0.6

Behavior

• ip -s link now shows live driver statistics
• ethtool provides driver info and link status
• All counters update during TX/RX/NAPI activity

This release makes the vnet driver behave more like a real Linux NIC.

v0.5 – Phase 5: NAPI-based RX + Stable Driver Architecture

Phase 5 – NAPI-based RX polling

Highlights

• Added NAPI support to the vnet virtual NIC driver.
• Extended struct vnet_priv with struct napi_struct napi.
• Implemented vnet_napi_poll() to drain the RX ring in poll context with a configurable budget.
• Registered NAPI via netif_napi_add() in vnet_init().
• Enabled NAPI in vnet_open() and disabled it in vnet_stop().
• Updated the RX timer to:
  • Allocate dummy RX packets,
  • Enqueue them into the RX ring,
  • Schedule NAPI via napi_schedule_prep() / __napi_schedule().

Behavior

• RX is no longer processed directly in the timer callback.
• Timer behaves like an RX interrupt: it only enqueues packets and schedules NAPI.
• NAPI poll hands packets to the kernel networking stack using netif_rx().

Status

• Driver has:
  • TX ring (Phase 3)
  • RX ring + RX timer (Phase 4)
  • NAPI-based RX processing (Phase 5)
• Future phases may add ethtool support, stats, and more realistic packet formats.

v0.4 – Phase 4: RX ring and timer

Phase 4 – RX ring and timer-based packet generator

Highlights

• Added RX ring (VNET_RX_RING_SIZE, struct vnet_rx_slot).
• Added RX helpers: vnet_rx_ring_init, vnet_rx_enqueue, vnet_rx_dequeue.
• Implemented vnet_rx_process_all() to drain RX ring (log + free SKBs).
• Added vnet_rx_timer_fn() to simulate hardware RX every 1 second.
• Updated vnet_open() to init TX/RX rings and start RX timer.
• Updated vnet_stop() to stop timer and drain both TX and RX rings.

Status

• RX traffic is synthetic (generated by timer).
• Packets are not yet passed into the networking stack (NAPI planned for Phase 5).

v0.3 – Phase 3: TX ring buffer implementation

Phase 3: TX Ring Buffer Implementation

This release adds a realistic TX ring to the vnet virtual NIC driver.

Highlights

• Introduced VNET_TX_RING_SIZE and struct vnet_tx_slot.
• Added vnet_tx_ring_init(), vnet_tx_enqueue(), vnet_tx_dequeue(), and vnet_tx_complete_all().
• Updated struct vnet_priv to include:
  • TX ring array
  • tx_head / tx_tail indices
• ndo_open now initializes the TX ring.
• ndo_stop drains any remaining packets from the TX ring.
• ndo_start_xmit:
  • Enqueues packets into the TX ring
  • Returns NETDEV_TX_BUSY and stops the queue when the ring is full
  • Simulates immediate TX completion via vnet_tx_complete_all()
• All new code is Doxygen-documented for easier browsing and learning.

Status

• RX path and NAPI are not implemented yet (planned for Phase 4).
• TX completion is synchronous for now (will be moved to async path in later phases).

v0.2 - Phase 2: Minimal net_device skeleton

• Added vnet_main.ko
• register_netdev()
• alloc_etherdev()
• net_device_ops (open, stop, xmit)
• Doxygen documented