VirtualBox 6.0 Beta 3 Released: Enables VMSVGA Device By Default, OCI Improvements

[phoronix]

Phoronix: VirtualBox 6.0 Beta 3 Released: Enables VMSVGA Device By Default, OCI Improvements

Oracle's VirtualBox 6.0 multi-platform virtualization software continues inching closer to release...

http://www.phoronix.com/scan.php?page=news_item&px=VirtualBox-6.0-Beta-3

[Danniello]

"VirtualBox 6.0 Beta 3 enables support for the "VMSVGA" graphics device by default. VMSVGA is VMware's SVGA II graphics adapter."

VMware graphic driver is supporting DirectX 10 acceleration in Windows guest VM.
Silly question - will it be working in VirtualBox too? I guess not - it would be too awesome...
But why? DirectX is emulated on other level than graphic driver?

[Chewi]

I didn't used to care about this much as I only used VirtualBox as a headless backend for testing Chef cookbooks with Kitchen and Vagrant. Now in my new job, I'm using CentOS 7 on the desktop as a VirtualBox guest under Windows 10. Not ideal but it hasn't been quite as painful as I expected. Chrome (or rather Vivaldi in my case) is able to use very little hardware acceleration though so these improvements sound welcome. Having said that, I suspect they might not be much use under CentOS 7?

[wswartzendruber]

I remember telling them over ten years ago that Gallium3D was the future and that they should probably start work on a graphics driver. They said they didn't see the point. I guess this is their endgame.

[cl333r]

I remember telling them over ten years ago that Gallium3D was the future and that they should probably start work on a graphics driver. They said they didn't see the point. I guess this is their endgame.

True but OTOH doing whatever any user suggests would be foolish.

[Xaero_Vincent]

Oh wow... Virtualbox using SVGA II is news. I thought they were against using that? Great news, though. SVGA II is lightyears ahead. However, Virgl is even further ahead I believe as far as OpenGL and ES compliance goes.

[ThoreauHD]

On a usability standpoint, I've noticed that 5.2.22 stable doesn't handle the newer kernels on ubuntu 16.04/18.04 very well aka it freezes/segfaults on vm-tools. Migrating to 6 beta resolves that issue.

[andre30correia]

the version is old, 17.3 or something, only opengl 2.1 and dx 9. I teste it with windows 10 some benchmarks and have a lot of bugs. Crashs, no render at all, not even close to vmware player

[SteamPunker]

Not to look a given horse in its mouth, but if you think about it, this is a typical FY move towards the Open Source Software community that Oracle is known for.

Think about it: For years, people have complained about the VirtualBox graphics guest driver not being part of the mainline Linux kernel. The reason for this was because the source code for it was bloated. Oracle never made any serious effort to dust it off and get it mainlined. The Linux kernel developers didn't mince any words about it and went as far as to call the guest driver source code "tainted crap". As a result of this, VirtualBox users who preferred to work with newer kernels would often have to deal with having to patch and recompile the guest drivers in order to make it work on their systems.

Then, Hans de Goede of Red Hat finally decided to do something about it and rewrote the VirtualBox guest video driver under the name vboxsf, basically doing Oracle's work for them. He improved the driver drastically, even reducing the code base from 52,861 lines of code down to 7,275 lines of code. This finally made its quality acceptable enough to be mainlined in the Linux kernel.

And *now* they finally add support for the VMWare guest driver, which had already been part of the mainline Linux kernel for years and also in much better shape with regards to 3D acceleration based on Gallium3D and all of that good stuff. Why didn't Oracle do this earlier? Partial (incomplete?) support for the VMWare guest driver had already been in the VirtualBox code base for years. Why didn't they dust this off earlier, so Hans de Goede wouldn't have had to go through the effort of cleaning up their crap in the first place?

Like I said, this is typical of Oracle. They did something similar to this with OpenOffice. Like VirtualBox, they obtained ownership of it when they acquired Sun Microsystems, and then they let it wither on the vine, until the open source community finally got fed up with its lack of progress and forked it to LibreOffice. And as soon as Oracle noticed that LibreOffice started to thrive as a fork, undergoing much faster development, they decided to donate the OpenOffice code base, as well as the rights to its name, not to the Document Foundation that manages the LibreOffice project, but to the Apache Foundation. As a result, OpenOffice now has an incompatible license, so it can no longer be reunified with the LibreOffice project, which, let's be honest, has a stupid name, and could really use having the OpenOffice name. Purely a dick move.

And then there is Solaris and ZFS. ZFS is open source, but they refuse to relicense it to GPL2, so it can become part of Linux. At the same time, they were behind the BTRFS project, which basically reinvents the wheel unnecessarily. Just relicense ZFS already, Oracle!

Similar examples of screwing over the open source community, seemingly out of some kind of spite, or perhaps "just because they can", are Hudson/Jenkins and MySQL/MariaDB.

Thank Goodness Sun relicensed Java under the GPL before Oracle got its paws on them. Now the AdoptOpenJDK initiative is managing a fork of OpenJDK which has been gaining a lot of traction in a short period of time. That's a promising development, given the relevance of Java in the world of backend web development.

I'm sorry to say, but Oracle is such an incredibly shitty company.

I'm frankly surprised and disappointed that VirtualBox hasn't been forked yet by the open source community, like OpenOffice, MySQL, Solaris, ZFS and OpenJDK. With every major release I enthusiastically check the release notes, only to see my hope dashed with a list of underwhelming and mediocre improvements and minor features. And in this regard, at least the Oracle employees working on VirtualBox are at least quite honest when you read their answers in the VirtualBox forums. They repeatedly make it quite clear that the priority of features is driven by demand from their paying customers. I know there are other open source hypervisors out there, notably Xen and KVM, but none of them offer cross-platform host support or the desktop guest performance that VirtualBox does. Only VMWare Workstation competes with VirtualBox in this field, but it's not open source. VirtualBox has so much potential in the right hands! It already shares quite a bit of code with the QEMU project and most (if not all) of the premium features in the closed-source Oracle extension pack could be reimplemented by porting over code from QEMU. Also, better support for legacy sound cards could be ported from the DOSBox project.

Yes, VirtualBox has a mechanism that allows it to be extended using third-party extension packs. And I've looked into this, but unfortunately, it's poorly documented and it's not trivial to compile a cross-platform extension pack. I've been unable to find proper documentation for this.

People in the Open Source Software community: please fork VirtualBox already! Perhaps someone should just start a web site and take it from there...

[man-walking]

Does VirtualBox still have that performance bug in multicore (2+) guests that make them perfom slower (jerky) then single core ones ?
It's some years that I don't use it, I mean VirtualBox linux version to be used with Windows guests.