Future of KQEmu

[tpreitzel]

First, I'll try to contain myself, but it won't be easy. I desire to use words that I won't use. The arrogance of some developers just galls me and other users. I only wish Fabrice would fork QEmu and return it to a software virtualization model or send the KVM hacks packing. QEmu was initially a non-hardware model of virtualization that many people used and relied on. Now, it seems that group of users has been abandoned by some developers who have usurped the project and redirected it into a hardware only model. Steve Fosdick, http://lists.gnu.org/archive/html/qemu- ... 00295.html , has a coherent and rational problem with abandoning the software model of virtualization. Since none of the current developers have to lower themselves to a level of using a non-hardware model, they willy nilly thumb their nose at those users who use older hardware. Not one of the responses except for Rebe even considered the necessity of having support within QEmu for a software model of acceleration. Rather, the responses can be read for yourself and range from "We don't maintain it and it's therefore unsupported" to "hack on KQemu yourself and good luck". The arrogance is enough to make anyone puke. I'm seriously thinking of abandoning QEmu because of this attitude. Sickening... stop make excuses and create any software form of acceleration for QEmu. It's needed for reasons already mentioned in the e-mails.

[McStarfighter]

It would be very nice to get an complete answer ...

Is KQemu dead? And if yes: What about the Windows users of Qemu? They can't use KVM and then are forced to run without a good performance, especially on the NT6 Windows versions ...
Or is there a KVM kernel module for Windows in development?

Please answer "us" ...

[hw0023]

Please answer "us" ...

Bump!
/me to

I fight at the moment with kqemu and an atom 330 dual core with instability, low performance and system freeze. I can't say I fail or kqemu, but anyway: It would not be amiss to know about the future plans of this part of qemu.
I use generally qemu/kvm with very good experience so it would be nice to use it with kqemu on hardware without native virtualisation like the actual atom stuff from intel.
Maybe a link to a info or dicussion?

[rowar]

Hi,

I wrote a bug report about this:

https://bugs.launchpad.net/qemu/+bug/493519

[rowar]

I think we have here two different ways and opinions:
- The "traditional" QEMU users who use it on Windows, BSD, Mac OS X, eComstation and Linux with KQEMU mostly for the Desktop.
- The KVM users who use it for mainly for the Data Center Virtualization on special modern Hardware.

A lot of things changed since the beginning of QEMU development:
- Many virtualization software is free now (VMWare Server, -Player, Virtual Box). For the most people there is no need for using QEMU as a free software on the Desktop (except for special reasons).
- Hardware make a lot of improvements. Hardware Virtualization is very common now.
- KVM make big improvements because of the Hardware Virtualization and sponsors.
- The aims of QEMU/KVM development are very different to the beginning now.

The results are:
- Until now the QEMU/KVM developer tryed to fit all aims in one product. But this is not possible in the future. Desktop- and Data Center Virtualization are different things.
- The use of QEMU/KVM getting more and more complicated. Nobody, including the developers, knows all options of QEMU/KVM (see the out-of-dated docs on http://www.qemu.org).

My suggestion: Make a fork.
- Make an easyer to use QEMU-light with KQEMU/KVM for the Desktop only with necessary options based on QEMU 0.11.1.
- Improve the great QEMU/KVM development for the Data Center Virtualization.
- Try to keep both versions partially compatible (same image formats, same important options, ...)

[hw0023]

I think we have here two different ways and opinions:
- The "traditional" QEMU users who use it on Windows, BSD, Mac OS X, eComstation and Linux with KQEMU mostly for the Desktop.
- The KVM users who use it for mainly for the Data Center Virtualization on special modern Hardware.

OK, I agree with you, ...but some people I know (myself included) need a SoHo Server too.
As long hardware with "svm|vmx" is to expensive, loud and hungry, as long kqemu is imho needed.
But I understand the developer standpoint very good. I think in 1 or 2 years KVM compatible hardware would be cheap, silent, green and a standard.
So why fight with an anachronism? I am little sorry about the tons of old hardware, but I have not to decide about other peoples time/work. Maybe Virtualbox can help???

My suggestion: Make a fork.
- Make an easyer to use QEMU-light with KQEMU/KVM for the Desktop only with necessary options based on QEMU 0.11.1.
- Improve the great QEMU/KVM development for the Data Center Virtualization.
- Try to keep both versions partially compatible (same image formats, same important options, ...)

I think this is a good suggestion, but I do not understand enough from qemu/kqemu/kvm to serious talk about it.

Now my kqemu works and the system was amazing fast.
Good work, what a pity to drop this part.

[rowar]

OK, I agree with you, ...but some people I know (myself included) need a SoHo Server too.
As long hardware with "svm|vmx" is to expensive, loud and hungry, as long kqemu is imho needed.
But I understand the developer standpoint very good. I think in 1 or 2 years KVM compatible hardware would be cheap, silent, green and a standard.
So why fight with an anachronism? I am little sorry about the tons of old hardware, but I have not to decide about other peoples time/work. Maybe Virtualbox can help???

VirtuelBox is good for the Desktop not so good for Data Center Virtualisation.

If you need only Linux-Guest try eisXEN http://www.eisxen.org
It is easy to install, easy to use and fast. You can build templates for other Linux distribution too.

I think this is a good suggestion, but I do not understand enough from qemu/kqemu/kvm to serious talk about it.
Now my kqemu works and the system was amazing fast.
Good work, what a pity to drop this part.

Maybe a fork is a bit overkill. But the last QEMU version with KQEMU-support need bugfix support in the future.

I've wrote an how to to buld QEMU 0.11.1 with KQEMU. Sorry, it is in German but you can get a funny translation by google with the link "English" on top:

http://qemu-buch.de/d/QEMU%2BKQEMU_unter_Linux

If you compile QEMU and KQEMU you will independed from the qemu-packets. In Ubuntu 9.10 the packet qemu-kvm (0.11.0) has alredy no kqemu-support!
I think you can use your own compiled QEMU 0.11.1 with KQEMU for some years (like QEMU 0.9.1). Good news: You have not problems with the changing QEMU options

Maybe in two years the hardware with "svm|vmx" will be cheap. (The XP-mode of Windows 7 need hardware with "svm|vmx" too).

[jacek]

Hi. Is "svm|vmx" the same thing as Intel VT-x and AMD-V (pacifica)?
http://en.wikipedia.org/wiki/X86_virtualization.
http://www.channelregister.co.uk/2006/0 ... a_support/
http://ark.intel.com/VTList.aspx
I am very confused because in 2009 you wrote that computers that support hardware virtualization are too expensive. I thought that hardware virtualization is something common and affordable today but after reading this thread I'm not sure. I bought my recent PC in 2007. Its cpu was Athlon 64 x2 3600 (1900MHz) and I always thought that it supports hardware virtualization. There is even option in BIOS to enable/disable Virtualization. Today it is rather low-end machine. I am also suprised that kqemu doesn't need hardware virtualization to accelerate qemu. I used kqemu but never kvm. Is "svm|vmx" some kind of new, expensive hardware virtualization technology that is required by kvm? Please, could you dissolve my doubts about hardware virtualization?

[rowar]

Hi. Is "svm|vmx" the same thing as Intel VT-x and AMD-V (pacifica)?
http://en.wikipedia.org/wiki/X86_virtualization.
http://www.channelregister.co.uk/2006/0 ... a_support/
http://ark.intel.com/VTList.aspx
I am very confused because in 2009 you wrote that computers that support hardware virtualization are too expensive. I thought that hardware virtualization is something common and affordable today but after reading this thread I'm not sure. I bought my recent PC in 2007. Its cpu was Athlon 64 x2 3600 (1900MHz) and I always thought that it supports hardware virtualization. There is even option in BIOS to enable/disable Virtualization. Today it is rather low-end machine. I am also suprised that kqemu doesn't need hardware virtualization to accelerate qemu. I used kqemu but never kvm. Is "svm|vmx" some kind of new, expensive hardware virtualization technology that is required by kvm? Please, could you dissolve my doubts about hardware virtualization?

You can check your computer with this command:

Code: Select all

grep "vmx" /proc/cpuinfo

grep "svm" /proc/cpuinfo


If you get an output your computer support hardware virtualization (see http://qemu-buch.de/d/Installation). In this case you can use kvm. kvm ist faster than kqemu.

The prices for computers are changing and relative. For some years and in some other countries the hardware virtualization is expensive.

[jacek]

Oh I think I understand it now.
In [[/proc/cpuinfo]] VT-x is called "vmx", and AMD-V is called "svm". Am I right?

[rowar]

Oh I think I understand it now.
In [[/proc/cpuinfo]] VT-x is called "vmx", and AMD-V is called "svm". Am I right?

Yes, vmx is Intel VT and svm is AMD-V.