QEMU 5.1 Release Brings Many Improvements To This Open-Source Virtualization Component

[phoronix]

Phoronix: QEMU 5.1 Release Brings Many Improvements To This Open-Source Virtualization Component

QEMU 5.1 is now available for this important piece of the open-source Linux virtualization stack...

http://www.phoronix.com/scan.php?page=news_item&px=QEMU-5.1-Released

[Terrablit]

Some of these features are really neat.

The various secure/protected memory and secrets features will help a lot in keeping the public cloud secure and free. It's just small steps to a future where the host can't read VM data and applications on the VM can't escape the hypervisor containment. It's not super exciting, I guess, unless you pay attention to all the different security breaches that occur regularly.

Persistent Memory Region (PMR) support from the NVMe 1.4 specification.

Persistent memory features are going to go a long way to improving performance and stability.

Improved HVF acceleration support on Apple macOS.

Hypervisor acceleration is always neat. I'm looking forward to the day when QEMU becomes the most common non-commercial hypervisor on all OSes. And hypervisors are more likely to be baked into commercial OSes for various new features, so this lets people use a consistent interface without conflicting with OS support.

With this and eventual virtualized 3D graphics support, cross-platform hypervisor products like Virtualbox have a good chance of being buried. Which, as an Oracle product, can only be a good thing in the long run.

Support for Zstd compression for QCOW2 images when using compress_type=zstd as a creation option.

Zstd compression support becoming more standardized is definitely a good thing. And QCOW2 definitely needs some love. I can't recall which half-assed feature bothered me about the format support at the moment, but just having people work on it means it's not going to bitrot.

And the extra support for ARM, RISC-V, AVR and POWER cpus make it even more versatile than before. It'd be nice if motherboard manufacturers started adding more user-accessible buses like I2C and SPI and everybody worked on virtualizing them. That could mean a lot for microcontroller product development. Imagine developing your weird-ass peripherals and RGB shit in QEMU.

[Jabberwocky]

Some of these features are really neat.

With this and eventual virtualized 3D graphics support, cross-platform hypervisor products like Virtualbox have a good chance of being buried. Which, as an Oracle product, can only be a good thing in the long run.

I will celebrate it when that happens.

Zstd compression support becoming more standardized is definitely a good thing. And QCOW2 definitely needs some love. I can't recall which half-assed feature bothered me about the format support at the moment, but just having people work on it means it's not going to bitrot.

And the extra support for ARM, RISC-V, AVR and POWER cpus make it even more versatile than before. It'd be nice if motherboard manufacturers started adding more user-accessible buses like I2C and SPI and everybody worked on virtualizing them. That could mean a lot for microcontroller product development. Imagine developing your weird-ass peripherals and RGB shit in QEMU.

Copy on write has always been very useful to me, it has been worth the performance hit. I don't know how many people use encryption or snapshots. I have in the past but do not anymore.

Having user-accessible buses in QEMU would be awesome not only for developing but for testing it too. Wondering if a USB to I2C could be used as a workaround, there's a ton of open source examples floating around for example: https://github.com/harbaum/I2C-Tiny-USB

[Jabberwocky]

Almost forgot.

HyperV breaks my guest (Windows 10) when I enable it with host-passthrough on a Ryzen 2700X (nested virt is configured and working in other guests).

Funny thing is I almost broke the guest OS, Windows 10 did not want to go into safe mode. I had some trouble disabling HyperV as my guest locked up soon after booting. I was able to disable it easily after setting my CPU to EPYC. I don't know why Windows would require you to start normally before you can choose to go into safe mode. >.<

Same problem here: https://www.reddit.com/r/qemu_kvm/co...008_dont_know/

PS: I wonder if it's related to some wierd MS policy where they don't want you to nest HyperV and are trying to break it on purpose? I need to read up on it more. I don't really need HyperV, I was just messing around with Android emulation in Visual Studio.