Reinitialize virtqueues

[br]

Summary

Reinitialize all the registers

This fixes vtnet operation with virtio-net-device of latest QEMU.

Test Plan

Tested with MMIO bus on FreeBSD/RISC-V:

qemu-system-riscv64 -machine virt -m 2048M -kernel bbl -nographic -netdev tap,ifname=tap0,script=no,id=net0 -device virtio-net-device,netdev=net0 -smp cpus=8

[br]

Teardown virtqueues before RESET in virtio PCI bus driver as well

[bryanv]

I don't follow why qemu clearing "pointers" means that we realloc the virtqueues in the guest. The host virtqueue state is reinitialized as part of the reinit process. Also, what do you believe has "recently" changed in qemu's virtio_reset? The blame shows very little changes in the last several years.

I didn't write the MMIO driver and am not as familiar the spec, but perhaps your issue is because vtmmio_reinit_virtqueue() does not replicate all the steps that vtmmio_alloc_virtqueues() perform. Namely, it is missing

		vtmmio_write_config_4(sc, VIRTIO_MMIO_QUEUE_NUM, size);
		vtmmio_write_config_4(sc, VIRTIO_MMIO_QUEUE_ALIGN,
		    VIRTIO_MMIO_VRING_ALIGN);

that I could see being otherwise reset by qemu because of a reset status.

[br]

Reinitialize virtqueue registers

[br]

In D15821#334539, @bryanv wrote:

I don't follow why qemu clearing "pointers" means that we realloc the virtqueues in the guest. The host virtqueue state is reinitialized as part of the reinit process. Also, what do you believe has "recently" changed in qemu's virtio_reset? The blame shows very little changes in the last several years.

I didn't write the MMIO driver and am not as familiar the spec, but perhaps your issue is because vtmmio_reinit_virtqueue() does not replicate all the steps that vtmmio_alloc_virtqueues() perform. Namely, it is missing

		vtmmio_write_config_4(sc, VIRTIO_MMIO_QUEUE_NUM, size);
		vtmmio_write_config_4(sc, VIRTIO_MMIO_QUEUE_ALIGN,
		    VIRTIO_MMIO_VRING_ALIGN);

that I could see being otherwise reset by qemu because of a reset status.

Thanks. I missed this

[bryanv]

sys/dev/virtio/mmio/virtio_mmio.c

768 ↗

(On Diff #43812)

This seems reasonable, although it might make sense to move this after the virtqueue_reinit to keep the pattern similar to the virtqueue_alloc in vtmmio_alloc_virtqueues.