Implement virtio-net device #34

jimmylu890303 · 2024-06-27T18:47:06Z

Implementation of Virtio-net Device

Defined struct virtio_net_dev to represent the virtio-net device.
Implemented virtio_net_init to initialize the TAP device for communication.
Introduced virtio_net_init_pci to register the virtio-net device on PCI, configure virtqueues.
Implemented functions to handle available events for RX and TX virtqueues.
Added functions to notify the guest OS about used descriptors after packet processing.

Makefile Modification

Added rules to compile virtio-net.c into virtio-net.o, ensuring correct compilation and linkage of virtio-net functionality within the project.

Kernel Networking Options Enabled

Enabled necessary kernel configuration options (CONFIG_NET, CONFIG_INET, CONFIG_VIRTIO, CONFIG_VIRTIO_NET, CONFIG_NETDEVICES) to support networking functionality.
Enabled specific networking commands required for testing the virtio-net device.

Structural Updates

Added struct virtio_net_dev entry within struct vm_t to facilitate the integration and management of the virtio-net device.

Additional Changes

Called virtio_net_init and virtio_net_init_pci in vm_arch_init_platform_device and main respectively, ensuring initialization of the virtio-net device before the virtual machine starts.

Makefile

configs/busybox.config

configs/linux-x86.config

src/arch/x86/desc.h

src/arch/x86/vm.c

src/virtio-net.c

src/virtio-net.h

src/virtio-pci.h

src/main.c

src/virtio-net.c

jimmylu890303 · 2024-07-06T14:04:50Z

I have pushed a newer version. Please take a look.

For testing the function, you can use the command below:

Host OS :

sudo ip link delete br0
sudo brctl addbr br0
sudo ip addr add 10.0.0.1/24 dev br0
sudo ip route add default via 10.0.0.1 dev br0
sudo ip link set br0 up
sudo ip link set tap0 master br0
sudo ip link set tap0 up

Guest OS :

ip addr add 10.0.0.2/24 dev eth0
ip link set eth0 up
ip route add default via 10.0.0.1

src/arch/x86/vm.c

RinHizakura

I'm good with the changes now. Thanks for all the kindly reply.

Let' see if @jserv has any extra comment.

jimmylu890303 · 2024-07-10T12:07:30Z

I'm good with the changes now. Thanks for all the kindly reply.

Let' see if @jserv has any extra comment.

Thanks for your review !

jserv

Rebase the latest master branch because the commit "Bump linux version to 6.1.91" was accidentally checked in as a non-functional change. See #35

jserv

Indent all files using clang-format. That is, run find src -name '*.[ch]' | xargs clang-format-12 -i. Do replace clang-format-12 with clang-format version 12 or later.

jimmylu890303 · 2024-07-20T05:34:18Z

I have pushed the latest version with the following updates:

I rebased the latest master branch to remove the non-functional commit "Bump linux version to 6.1.91" as discussed in issue Fail to run kvm-host with Linux in version 6.1.91 #35.
I have reformatted all relevant files using clang-format. Specifically, I ran the command find src -name '*.[ch]' | xargs clang-format-12 -i with clang-format version 12.

jserv · 2024-07-21T07:06:29Z

I have reformatted all relevant files using clang-format. Specifically, I ran the command find src -name '*.[ch]' | xargs clang-format-12 -i with clang-format version 12.

Warning: No newline at end of file src/virtio-net.c
See https://medium.com/@alexey.inkin/how-to-force-newline-at-end-of-files-and-why-you-should-do-it-fdf76d1d090e

Makefile

src/virtio-net.c

jserv

Use git rebase -i to rework the git commits.
The following commits can be squashed and refined:

Define VIRTIO_NET_IRQ 14
Define VIRTIO_PCI_DEVICE_ID_NET macro
Define macro in virtio-net.h
Define struct virtio_net_dev and functions in virtio-net.h

Git commits are snapshots around logical units of change. Over time, commits should tell a story of the history of your repository and how it came to be the way that it currently is.
See https://github.blog/2022-06-30-write-better-commits-build-better-projects/

You don't have to mention the file names in the subject of git commit messages. Instead, summarize the functional changes.

- Define `VIRTIO_NET_IRQ` macro for the `notify_used` virtq callback function to send IRQ - Define `VIRTIO_PCI_DEVICE_ID_NET` (0x1041) according to the Virtio 1.1 specification - Define `VIRTIO_NET_PCI_CLASS` macro according to the PCI specification with class code (0x020000) for Ethernet Controller - Define struct `virtio_net_dev` and operation functions for the virtio-net device

src/virtio-net.c

jserv · 2024-07-21T13:40:57Z

src/main.c

@@ -93,6 +93,8 @@ int main(int argc, char *argv[])
        return throw_err("Failed to load initrd");
    if (diskimg_file && vm_load_diskimg(&vm, diskimg_file) < 0)
        return throw_err("Failed to load disk image");
+    if (vm_enable_net(&vm) < 0)


If virtio-net is not enabled, it still makes sense to initialize and launch this virtual machine via KVM. Thus, instead of returning, we can proceed.

I changed throw_err to fprintf. If vm_enable_net failed to enable virtio-net device, it will now handle the situation by logging an error message.

if (vm_enable_net(&vm) < 0) fprintf(stderr, "Failed to Enable Virtio-Net Device\n");

Refine the message as following:

Failed to enable virtio-net device.

Ok ! Already done

- Defined struct `virtio_net_dev` to represent the virtio-net device. - Implemented `virtio_net_init` to initialize the TAP device and prepare it for communication. - Added `virtio_net_init_pci` for registering the virtio-net device to PCI, setting up virtqueues. - Implemented functions to enable and handle available events for RX and TX virtqueues. - Added functions to notify the guest OS about used descriptors after packet processing.

This commit removes the initialization of vq->info.notify_off value in the virtio-net implementation. The reason for this change is that during the execution of RX and TX enable callback functions, the `vm_ioeventfd_register` function registers the ioeventfd at the same address repeatly. Therefore, the notify_off value is now initialized directly within the virtqueue enable callback functions.

This commit adds a struct `virtio_net_dev` entry within the struct `vm_t` structure to facilitate the implementation of the virtio-net device. This addition is necessary to integrate and manage the virtio-net device within the broader context of the virtual machine.

jserv

Unify the commit "Implement vm_enable_net API" and "Compile virtio-net.c into virtio-net.o during build process" into a single one since they are tended to the same purpose.

The vm_enable_net function to facilitate the integration of additional I/O devices in the future. The function initializes the virtio-net device by setting up the tap device using virtio_net_init and subsequently registers the virtio-net device with the PCI bus. This function is invoked during the VM initialization session.

Enabled kernel configuration options required for networking functionality,including `CONFIG_NET`,`CONFIG_INET`,`CONFIG_VIRTIO`, `CONFIG_VIRTIO_NET`,`CONFIG_NETDEVICES` etc. Additionally, enabled specific networking commacands necessary for testing the virtio-net device.

jserv · 2024-07-21T18:10:36Z

Thank @jimmylu890303 for contributing!