[virtio] Support modern virtio v2 PCI virtio block

[kallisti5]

It looks like a newer "modern" virtio v2 design is now in use which our drivers don't work with. The PCI device ID's are higher.

Seen on a vultr vm trying to boot Haiku R1/beta3

src/add-ons/kernel/drivers/disk/virtual/virtio_block/virtio_block.cpp likely needs some code path to attach to a "virtio bus" or a "pci bus"

[kallisti5]

example

[kallisti5]

target is R1/Beta4 because if we ever want to run haikuporter builders in cloud environments, solid virtio block support will be critical.

[kallisti5]

https://dev.haiku-os.org/ticket/17239 is related

[kallisti5]

Oddly looking at these devices.. our virtio_pci bus driver checks for the virtio vendor and a device in the ranges:

#define VIRTIO_PCI_VENDORID     0x1AF4
#define VIRTIO_PCI_DEVICEID_MIN 0x1000
#define VIRTIO_PCI_DEVICEID_MAX 0x103F

The virtio block device in this VM is 1af4:1042 which is outside of that MIN/MAX range :-| Does that guide the user to attach directly, or does the min/max need expanded to pick up the devices?

[kallisti5]

Here's what linux looks for:

/* Qumranet donated their vendor ID for devices 0x1000 thru 0x10FF. */
static const struct pci_device_id virtio_pci_id_table[] = {
        { PCI_DEVICE(PCI_VENDOR_ID_REDHAT_QUMRANET, PCI_ANY_ID) },
        { 0 }
};

maybe we're artificially limiting our device id's too much?

[kallisti5]

ah ha.

FreeBSD: sys/dev/virtio/pci/virtio_pci_legacy.c

        if (pci_get_vendor(dev) != VIRTIO_PCI_VENDORID)
                return (ENXIO);

        if (pci_get_device(dev) < VIRTIO_PCI_DEVICEID_MIN ||
            pci_get_device(dev) > VIRTIO_PCI_DEVICEID_LEGACY_MAX)
                return (ENXIO);

Freebsd sys/dev/virtio/pci/virtio_pci_modern.c

       if (pci_get_vendor(dev) != VIRTIO_PCI_VENDORID)
                return (ENXIO);

        if (pci_get_device(dev) < VIRTIO_PCI_DEVICEID_MIN ||
            pci_get_device(dev) > VIRTIO_PCI_DEVICEID_MODERN_MAX)
                return (ENXIO);

        if (pci_get_device(dev) < VIRTIO_PCI_DEVICEID_MODERN_MIN) {
                if (!vtpci_modern_transitional)
                        return (ENXIO);
                devid = pci_get_subdevice(dev);
        } else
                devid = pci_get_device(dev) - VIRTIO_PCI_DEVICEID_MODERN_MIN;

sys/dev/virtio/pci/virtio_pci_var.h:#define VIRTIO_PCI_DEVICEID_MIN	0x1000
sys/dev/virtio/pci/virtio_pci_var.h:#define VIRTIO_PCI_DEVICEID_LEGACY_MAX	0x103F
sys/dev/virtio/pci/virtio_pci_var.h:#define VIRTIO_PCI_DEVICEID_MODERN_MIN	0x1040
sys/dev/virtio/pci/virtio_pci_var.h:#define VIRTIO_PCI_DEVICEID_MODERN_MAX	0x107F

So..

• deviceid 0x1000 - 0x103F is "Virtio Legacy"
• deviceid 0x1040 - 0x107F is "Virtio Modern"

ugh.

[kallisti5]

Per this... it appears qemu may have a way to enable the "modern, virtio v2" interface..

​https://lists.sr.ht/~philmd/qemu/patches/8154

For this reason, the v2 personality is disabled, keeping the legacy
behavior as default. Machine types willing to use v2, can enable it
using MachineClass's compat_props.

I'm guessing this is what Vultr has enabled for some reason.

[kallisti5]

see ​https://review.haiku-os.org/c/haiku/+/4420 for an early "quick" solution. It fixes the missing virtio devices for me locally.

This appears to be a change of default functionality in qemu-6.1.0? I've been surfing the qemu changes but don't see a reference.

[pulkomandy]

Moving various tickets out of the beta4 milestone because no one is actively working on them.

[korli]

fixed in hrev57329