QEMU Launcher + 32 bit QEMU pet and sfs

[mikeb]

Ok this is not over yert.

Firstly retried with LHP on the atom.

The output of free lacks the Total: output ...this buggers loading an sfs to tmpfs for my loader and I suspect also for sfs load which also failed.
This is the scenario with an sfs downloaded to say /root and then loaded...it has to be moved to either ram or a device to be inserted..I usually go for ram.
So a time waster there but this time it got truly loaded manually....yay it tries to run now.

I tested the launcher which failed the first time ... although kvm.ko loaded apparently it also requires kvm-amd or kvm-intel to work. Since the atom apparently does not support kvm this fails... I did however get it to run but of course booting Lucid was a slow custard stirring operation which takes about 5 to 10 minutes to boot...lol. I do have a kvm supporting 32 bit netbook and if slax 7 supports kvm I might still get the joy of KVM ...note kqemu flies on the same machines...but i lament.

Point here is I did get some real testing done which brings up some changes needed.

1.... i cannot confirm if the kvm-intel or kvm-amd load automatically when kvm does...someone needs to confirm the behavior as I may not be able to .
It definitely means I need to alter the test for valid kvm.

2....The kill session needs modding to handle the various potential binary in use..easy peasy.

As for sfs loading on lighthouse into ram... well that's its buggy combination...out of my domain

Once 1. is verified I will make a fresh pet..

mike

[gcmartin]

HI @MikeB

... The output of free lacks the Total: output ...this buggers loading an sfs to tmpfs for my loader and I suspect also for sfs load which also failed.
This is the scenario with an sfs downloaded to say /root and then loaded...it has to be moved to either ram or a device to be inserted..I usually go for ram.
So a time waster there but this time it got truly loaded manually....yay it tries to run now. ...

For LH64 use: If you have placed the SFS in the ISO's root ("/", not "/root"), you should have seen it at pristine boot time along with any other SFS that are present.

For example: I have LH64 running. I open Menu>Multimedia>IsoMaster to a LH64 ISO. In its top pane, I navigate and select the SFS I want to Add and click Add where it shows up now in the bottom pane. Close-Save creates a new ISO. Booting that ISO, Live, I am presented during boot the option to have any/all SFSs added into my desktop system automatically without any user intervention, at all.

The Atom, Yes, in those, Intel decided not to include virtualization to cut cost of production. They also did not include hyperthreading in many, as well.

Virtualization support once system boots. I have ONLY AMD processors. When I boot FD/LH64, I have found that the kvm modules are loaded..."yeh". So, several weeks ago, I stopped tested for the presence (# lsmod | grep kvm).

Hope this helps

[Ted Dog]

kqemu flies

are you going to be able to put this in .. My AMD laptop has the 'support' turned off in bios but exists in hardware ( I checked before buy, but did not expect W8 said to force it off )

How soon till new update.. I am working on getting visualization running and heard VB ignores bios flags if hardware checks linux and QEMU with kvm does not. If kqemu still works that would be nice.

[gcmartin]

Hi @Ted Dog

... My AMD laptop has the 'support' turned off in bios but exists in hardware ...

I have a AMD X2 laptop. The information in the Guide was verified using that laptop. VM guest fly! Just use those instructions; only 3 steps no matter which Approach you select. I recommend the Launcher Approach.

My lappie is older than yours and cannot imagine why MS would demand it be turned off??? They are BIG on Hyper-V, just as we are on KVM. Find the feature (sound like you already have) and turn it on:, your VM guest will make good use in performance.

If there is something misleading or inaccurate in the Guide, please share to improve for others.

KVM, QEMU and this Puppy Linux utility makes everyone who follows to test and support the Puppy development community who build their distros for community use. Any discrepancies your report would be helpful

Here to help

[Ted Dog]

can't be turned on without hacking boot firmware.. I checked and have the haxor but not willing to fudge with it if all it takes is to ignore it on linux side. VBox workaround was just to comfirm hardware and not trust bios flag. W8 had it turn off on all manufacturers who would sale the preinstalled laptops. If would wa t to do this by the machines designed and upsale w8 server crap.

[Ted Dog]

got it to boot on version 2 of qemu launcher and dog slow.. acts frozen most of the time. Never saw a desktop. kqemu v1 or kvm neither would sbow anything only v2.1 without kvm shows a window and appearance of booting

[mikeb]

Ok kqemu was a kernel module that could execute some of the emulators functions directly on the real cpu thus providing significant speedup.
Kvm if i understand it is a kernel module that can use virtualization functions added to some cpu. Since they are working differently one is not really a substitute for the other.
It also appears that once kvm was supported then kqemu was dropped loosing acceleration for non kvm machines so the only way to use it is to use a pre kvm build of qemu. Kqemu is just another module built from source for the kernel in question. The source itself was a bit dependent on the version of qemu.
My qemu is version 0.9 and I have acceleration for kernels 2.6.28 thru 2.6.33 on any machine for example which is good for testing and the odd ugly encounters with IE demanding software.

Hope that clarifies things. The gui is only a front end...it up to the system/package as to what works and does not.

and yes without kvm support or a kqemu module then things tend to run like cold custard.

Ok going to see if I can get 32 bit kvm working on this centrino duo to hone the detection.

mike

[gcmartin]

@Ted Dog
What did the Guide's command show before you tested?

Can you post results to this:

Code: Select all

# egrep --color=always  '(vmx|svm)' /proc/cpuinfo
and this
# egrep flags --color=always  /proc/cpuinfo

No vmx or svm flag means there is no virtualization present for system to use. This in turn, means that the QEMU-KVM will not work.

Thus, KQEMU is one of the other features from QEMU group that can be employed for improving virtual PC performance when the CPU does NOT provide virtualization. Seem that I remember that some other QEMU features, also, may be used when virtualization hardware is not present.

Here to help

[Ted Dog]

ok I even readup on that months prior to your writeup. so I read it twice up the cpu does support svm even when firmware misreported the lack of support. That is my real issue. There is a way someone used to pull firmware flip the bit and reflash changed firmware but I do not think that is wise.
People are reporting older versions of QEMU with kqemu work well. Have yet to find 64 version of such with modern kernel.

[gcmartin]

Happy to report running a HOST test that I've just concluded to see its behavior. I am running a VM Guest in a VM Guest with the latter giving great performance.

This was done using Approach #2 in the Guide

Scenario

1. Running LH64-602b with QEMU V2.0 and QEMU Launcher V0.2.2
2. On the above HOST, I Launched a VM Guest PC running Slacko64 and a SWAP partition
3. Added QEMU V2.0 and QEMU Launcher V0.2.2 PETs to Slacko64
4. In the above Guest, I Launched a VM Guest PC running FATDOG

This is my test of a KVM Puppy running a VM Guest (FATDOG) in one VM Guest (Slacko64) on a VM Host (LH64) to observe PC system performance.

The benefit to this test is that I can test package installation on a guest and see if that impacts another guest's operation. Or I could remaster a Guest and test the remaster inside the Guest. There are, of course, other possibilities for test, too.

All running rather well

Others may see value and ideas for their own personal needs/usage with the flexibility this provides.

Here to help

[mikeb]

OK the lenovo needs a bios update to enable its kvm support.. not something I tend to rush into doing.

It does appear then kwm and the associated cpu module is loaded automatically so will simply change it to detect those. KVM did load itself in slax 7 even though bios support was missing so kvm-intel failed.

Would be interesting to see the current state of using kqemu since kvm is limited to certain hardware....friendly bios...system built to include kvm (what is the windows situation...I have qemu and kqemu running on that no problem? After all windows users might be more interested in testing or running linux virtually rather than existing linux users.)

Will update the pet later today.

mike

[mikeb]

Ok pet updated to 0.3

only shows kvm if kvm_intel or kvm_amd are loaded and assumes this happens automatically if support is present.
If not acceleration is available the gui offers to make a cup of coffee and a roll for the user while he or she waits for the guest to boot.

fixed the kill session function.

added choice of how many cpu cores to use

Since I get the feeling kqemu may return to the equation I will see about adding it back in in hopefully a non confusing way so end up with a one gui does all result.

But first needs testing with current changes

Mike

[gcmartin]

Hummm,... I have no way of evaluating either the Intel nor the EFI issues reported as I have neither in my arsenal of 64bit laps and desktops. But I am wondering if there is something consistent that could be recommended to users of such. You and @Ted Dog are certainly opening an area of concern where a thread discussion should be opened to address what is going on. This has to be going on for others with this technology as well. As I have read the comments from each of you, I wonder if there is a doorway created for, in this case, Microsoft which exposes the BIOS for OS manipulation when necessary???.

If I had either of these, I would start that discussion, but in the absence ...

Here to help

[mikeb]

Well appears to be something that needs enabling in bios OR a bios update needed in some cases for older hardware.

Cannot comment on UEFI issues though obviously being able to boot other operating systems is not microsoft policy.

I did read of instances where enabling kvm abilities affects other things like sound.

I did try -smp with kqemu but that lost the full acceleration in that case and reverted to having 2 generic pentiums with partial speedup. By the sounds of it kvm gives the full monty.

mike

[step]

I have just completed a quick test of version 3 on a Core Duo 2.4 GHz with 4GB RAM running FatDog64 631, kvm_intel, Qemu2.0 pet virtualizing 8GB vHDD, 2 cores, 2048MB memory, L.A.S.S.I.E.2.0 (a Precise 5.7.1 derivative). Everything went well and performance is snappy. Host CPU varies from 20% to 80+%. Inside the VM I tested web surfing, wget, ROX, sfs loading, Desktop Globe app. Didn't test audio/video.

[mikeb]

Thanks for the feedback... obviously some aspects of qemu are out of the scope of this interface.

General...-smp refers to the simulater guest cpu(s) so my wording is wrong..I will amend.
As for kqemu It seems it can only fullly accelerate one cpu and reverts to more emulated multi cpu for 2 or more.

mike

[gcmartin]

Thanks for the ability to define a virtual Guest PC with more than 1 core! The earlier report I shared has had another performance bump in the Slacko64 Guest as, now with the multiple cores, allows Slacko64 better desktop performance threading, while its launched Guest is running. Guest virtual PCs are "flying" comparatively.

This tool is a great assist for design and setup of the virtual PCs that ISO/distro boot into.
Your tool, QEMU Launcher, makes it easy for others to understand the creating an "internal PC" on the running desktop. Thx!!!

[mikeb]

Ok another update now we seems to have something useable.

version 0.4

• Added a box for additional parameters as suggested earlier
  Merged in kqemu support so should give the appropriate option for KVM or kqemu or no acceleration.
  Clarified smp wording.

Mike

[neerajkolte]

Hi mikeb,
The GUI is improving nicely.
New users might find it actually easier than virtualbox.

Thanks

-Neeraj

[mikeb]

Ok did realise I had had a mental blank which was sorted out after compiling QEMU.

Package updated to 0.5....

• Added choice of virtual machine taken from the included binaries (doh!)
  Added help button which opens qemu main doc in default browser.
  Removed the kill button since the gui has that anyway..or you can simply close the window...same effect. Also it is possible if somewhat masochistic to run several virtual machines so one kill button seemed a little inappropriate.

By the way how do you escape from full screen mode?

Also I built 2.0.0 on Lucid (32 bit) so wondered if that might be useful ... thing is I did not have opengl or SLD enabled/present which makes it more generic but I wondered if there was a significant performance hit without them?

If its useful I could make a sfs and pet including the gui here...

mike

mike