Skip to the content.

2025

其他的解读

https://mp.weixin.qq.com/s/snl49e1T0nJ7PNAz2MrYHg https://mp.weixin.qq.com/s/kUq6MvRyyX1aeuSZlPSxgA https://mp.weixin.qq.com/s/tFkSmDOz7rwcVXgYb3yzFg https://mp.weixin.qq.com/s/V09xR2oPdDLuCKbNFpoFzA https://mp.weixin.qq.com/s/0YQTyw5BtCx5mGrezDFtbQ

所有的 slide

https://pretalx.com/kvm-forum-2025/schedule/

The next generation QEMU functional testing framework

https://pretalx.com/kvm-forum-2025/talk/KKAUKN/

QEMU had many test suites (unit tests, “qtests”, “iotests”, TCG-tests, …)

(这里的第一个问题,为什么 Avocado 是需要的) https://github.com/avocado-framework/avocado 只有几百个 star

太细节了,不看了,有需要在看吧

Virtio 2025 state of the union

https://pretalx.com/kvm-forum-2025/talk/JQYA7Z/

QEMU Keynote

https://pretalx.com/kvm-forum-2025/talk/GDPJJP/

QEMU with VirtIO GPU Vulkan Support https://news.ycombinator.com/item?id=42392802

iphone 的支持: https://news.ycombinator.com/item?id=43592409 https://news.ycombinator.com/item?id=44258670

Introduction and KVM Keynote

很多都看不懂

Optimizations

Single-binary: Unify QEMU system binaries per target architecture

为什么,因为有的平台中,一个系统有两个架构,也就是常说的异构:

Qualcomm Snapdragon X Elite ● CPU Oryon (aarch64) ● NPU Hexagon (hexagon)

AMD Zynq Ultrascale+ MP (qemu-system-aarch64 + qemu-system-microblaze) ● CPU Cortex A53 (aarch64) ● MCU Arm Cortex-R5F (arm) ● PMU (Platform Management Unit) Microblaze controller (microblaze)

jq –raw-output < build/compile_commands.json ‘.[].command’ grep -v      
‘CONFIG_USER_ONLY’ grep -v tests/ grep -v qemu-keymap grep -v linux-user  
grep -v qemu-pr-helper sed -e ‘s/.* //’ -e ‘s/”,//’ sort uniq -c sort
-rn grep -v ‘^\s*1 ‘ less    

没有完全理解,不过可以大概知道,这个工作就是编排代码, 不然一个 machine 下只有一个架构的代码

The State of QEMU WebAssembly Port

https://pretalx.com/kvm-forum-2025/talk/EVRL9V/

效果不错啊: https://ktock.github.io/qemu-demo/busybox-arm.html

https://ktock.github.io/qemu-demo/alpine-x86_64.html

https://ktock.github.io/qemu-demo/raspi3ap.html

忽然想到,还可以把什么东西编译出来,放到 browser 中?

基本实现原理: ● TCI is an IR interpreter. Other backends JIT translates the IR to the host arch

● QEMU v10.1 supports compilation of TCI (interpreter) to Wasm

我理解,就是从 tcg 可以被 tcg interpreter 解析,现在 qemu 可以支持将 TCI 构建为 Wasm ,所以相当于 tcg interpreter 可以在网页中运行。

(这个工作很有意思,例如存储,网络如何实现的)

类似的工作是 jslinux

Making io_uring pervasive in QEMU

https://vmsplice.net/~stefan/stefanha-kvm-forum-2025.pdf

  1. 没合并,
  2. 有点水,前面讲了太多关于 io uring 的实现

等 10.2 合并之后,看看如果现在的事件监听还是不是 io uring 吧

似乎想要添加一个操作为 : aio_add_sqe 来替代 aio_set_fd_handler 操作 然后这么等待

while (busy_waiting) {
 …poll…
 if (io_uring_cq_ready(io_uring)) {
 break; /* go process CQEs */
 }
}

现在,如果真的可以用 iouring 作为 epoll ,不就是这种方法等待的吗?

From time to time : QEMU Time Control Redefined

https://pretalx.com/kvm-forum-2025/talk/NBJCSX/

主要考虑 tcg 问题

Exploring VM placement strategies for chiplet architectures

https://pretalx.com/kvm-forum-2025/talk/7EWDQN/

没有太理解,为什么不去直接在 linux kernel 中完成工作,而是通过分析 qemu 的行为, 然后来绑定核心?

GiantVM: A Many-to-one Virtualization System Built Atop the QEMU/KVM Hypervisor

https://pretalx.com/kvm-forum-2025/talk/MDGYZG/

有趣,简直逆天

把多个机器合并为一个 vm

Shadow ioeventfd: Accelerating MMIO in vfio-user with Kernel-Assisted Dispatch

https://pretalx.com/kvm-forum-2025/talk/KTKTG9/

nutanix 在 vfio-user 上继续的工作

Upstreaming NVIDIA vGPU Support: Architecture, Implementation, and Roadmap

https://pretalx.com/kvm-forum-2025/talk/FNPW3D/

还没合并到主线

NVIDIA vGPU Support on Grace Blackwell Superchip: Architecture, Design, Upstreaming Status

https://pretalx.com/kvm-forum-2025/talk/XVV8JS/ https://github.com/zhiwang-nvidia/linux/tree/zhi/vgpu-rfc-v2

感觉没什么干活啊

Windows on Arm on QEMU/KVM: Challenges and Solutions

https://pretalx.com/kvm-forum-2025/talk/BDG7QR/

尝试把 windows arm iso 在 qemu 上启动起来

Improving Windows Hypervisor-Protected Code Integrity (HVCI) Performance on KVM

nutanix 的工作

https://pretalx.com/kvm-forum-2025/talk/JCPZNG/

本站所有文章转发 CSDN 将按侵权追究法律责任,其它情况随意。