[Qemu-devel] -kernel-kqemu

[Qemu-devel] -kernel-kqemu

[Jim C. Brown]

This sounds like an interesting option. Qemu has moved one step closer to
VMware...

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] -kernel-kqemu

[Jim C. Brown]

On Wed, Feb 08, 2006 at 06:04:35PM -0500, Jim C. Brown wrote:
> This sounds like an interesting option. Qemu has moved one step closer to
> VMware...
> 

I should probably add, that this new option (-kernel-kqemu) allows for speeds
very close to VMware because it allows kqemu to virtualize most ring 0 code
(in addition to ring 3 code).

Can't wait to see qvm86 make use of the new API.

> -- 
> Infinite complexity begets infinite beauty.
> Infinite precision begets infinite perfection.
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
> 

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] -kernel-kqemu

[Hetz Ben Hamo]

Hi,

I´m sorry, but I´ve been following the CVS commits and I haven´t
understood this new option (fabrice has not written anything to this
list about it and what it does for the end user).

Could someone please shed some light for me and others who didn´t
understand whats going on please?

Thanks,
Hetz

On 2/9/06, Jim C. Brown <jma5@umd.edu> wrote:
> On Wed, Feb 08, 2006 at 06:04:35PM -0500, Jim C. Brown wrote:
> > This sounds like an interesting option. Qemu has moved one step closer to
> > VMware...
> >
>
> I should probably add, that this new option (-kernel-kqemu) allows for speeds
> very close to VMware because it allows kqemu to virtualize most ring 0 code
> (in addition to ring 3 code).
>
> Can't wait to see qvm86 make use of the new API.
>
> > --
> > Infinite complexity begets infinite beauty.
> > Infinite precision begets infinite perfection.
> >
> >
> > _______________________________________________
> > Qemu-devel mailing list
> > Qemu-devel@nongnu.org
> > http://lists.nongnu.org/mailman/listinfo/qemu-devel
> >
>
> --
> Infinite complexity begets infinite beauty.
> Infinite precision begets infinite perfection.
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>

Re: [Qemu-devel] -kernel-kqemu

[Jim C. Brown]

On Thu, Feb 09, 2006 at 10:46:07AM +0200, Hetz Ben Hamo wrote:
> Hi,
> 
> I?m sorry, but I?ve been following the CVS commits and I haven?t
> understood this new option (fabrice has not written anything to this
> list about it and what it does for the end user).
> 
> Could someone please shed some light for me and others who didn?t
> understand whats going on please?
> 
> Thanks,
> Hetz
> 

-kernel-kqemu virtualizes ring 0 code.

So it basically makes qemu do what VMware does.

IIRC someone reported a 33% speedup with the new option.

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] -kernel-kqemu

[Anthony Liguori]

Jim C. Brown wrote:
> -kernel-kqemu virtualizes ring 0 code.
>
> So it basically makes qemu do what VMware does.
>
> IIRC someone reported a 33% speedup with the new option.
>   
That was me.  That was a 33% speedup on win2k startup time.  kqemu (user 
only) has a negligible impact on win2k startup time which suggests this 
is mostly ring 0 code running which would make it a good benchmark for 
kernel-kqemu performance.

This was a terribly unscientific benchmarking so don't read too much 
into it.

Regards,

Anthony Liguori

Re: [Qemu-devel] -kernel-kqemu

[Phil Krylov]

On 10/02/06, Anthony Liguori <aliguori@us.ibm.com> wrote:
> Jim C. Brown wrote:
> > -kernel-kqemu virtualizes ring 0 code.
> >
> > So it basically makes qemu do what VMware does.
> >
> > IIRC someone reported a 33% speedup with the new option.
> >
> That was me.  That was a 33% speedup on win2k startup time.  kqemu (user
> only) has a negligible impact on win2k startup time which suggests this
> is mostly ring 0 code running which would make it a good benchmark for
> kernel-kqemu performance.
>
> This was a terribly unscientific benchmarking so don't read too much
> into it.

I am experiencing a huge (~5 times)  speedup in compilers and linkers.

Great ;))))

-- Ph.

Re: [Qemu-devel] -kernel-kqemu

[G Portokalidis]

Does the copying of instructions occur in user-space or kernel-space?
I would like to port Argos to use the kernel accelerator, but i'm
unsure this is possible considering the accelerator is not open
source.

If the `parsing' and copying of instructions occurs in user-space, i
could instead of instrumenting instructions, have the necessary
actions packed in a different block and executed in a pseudo-parallel
way.

That will require trapping some instructions (like jmp, call, etc),
which i'm not sure is possible unless you have access to the kernel
module.

Any insight on how the kernel accelerator works?

BTW, Fabrice could you add a link to argos
(http://www.few.vu.nl/argos) on your website? It is a child of qemu
anyway :-)


On 10/02/06, Phil Krylov <phil.krylov@gmail.com> wrote:
> On 10/02/06, Anthony Liguori <aliguori@us.ibm.com> wrote:
> > Jim C. Brown wrote:
> > > -kernel-kqemu virtualizes ring 0 code.
> > >
> > > So it basically makes qemu do what VMware does.
> > >
> > > IIRC someone reported a 33% speedup with the new option.
> > >
> > That was me.  That was a 33% speedup on win2k startup time.  kqemu (user
> > only) has a negligible impact on win2k startup time which suggests this
> > is mostly ring 0 code running which would make it a good benchmark for
> > kernel-kqemu performance.
> >
> > This was a terribly unscientific benchmarking so don't read too much
> > into it.
>
> I am experiencing a huge (~5 times)  speedup in compilers and linkers.
>
> Great ;))))
>
> -- Ph.
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>

Re: [Qemu-devel] -kernel-kqemu

[Troy Benjegerdes]

On Thu, Feb 09, 2006 at 04:01:34PM -0600, Anthony Liguori wrote:
> Jim C. Brown wrote:
> >-kernel-kqemu virtualizes ring 0 code.
> >
> >So it basically makes qemu do what VMware does.
> >
> >IIRC someone reported a 33% speedup with the new option.
> >  
> That was me.  That was a 33% speedup on win2k startup time.  kqemu (user 
> only) has a negligible impact on win2k startup time which suggests this 
> is mostly ring 0 code running which would make it a good benchmark for 
> kernel-kqemu performance.
> 
> This was a terribly unscientific benchmarking so don't read too much 
> into it.
> 
> Regards,
> 
> Anthony Liguori

My win2k guest (with SP4, but not any updates) seemed to hang on startup
with -kernel-kqemu.

(This is today's qemu-cvs with kqemu-1.3.0pre5)

Re: [Qemu-devel] -kernel-kqemu

[Brad Campbell]

Troy Benjegerdes wrote:
> On Thu, Feb 09, 2006 at 04:01:34PM -0600, Anthony Liguori wrote:
>> Jim C. Brown wrote:
>>> -kernel-kqemu virtualizes ring 0 code.
>>>
>>> So it basically makes qemu do what VMware does.
>>>
>>> IIRC someone reported a 33% speedup with the new option.
>>>  
>> That was me.  That was a 33% speedup on win2k startup time.  kqemu (user 
>> only) has a negligible impact on win2k startup time which suggests this 
>> is mostly ring 0 code running which would make it a good benchmark for 
>> kernel-kqemu performance.
>>
>> This was a terribly unscientific benchmarking so don't read too much 
>> into it.
>>
>> Regards,
>>
>> Anthony Liguori
> 
> My win2k guest (with SP4, but not any updates) seemed to hang on startup
> with -kernel-kqemu.

Are you using -m 256 by any chance? I get this result with around that much ram allocated to the 
guest. -m 160 (or less) or -m 384 (or more) works perfectly here..



-- 
"Human beings, who are almost unique in having the ability
to learn from the experience of others, are also remarkable
for their apparent disinclination to do so." -- Douglas Adams

Re: [Qemu-devel] -kernel-kqemu - bug with -m 256

[Dan Sandberg]

Brad Campbell wrote:

> Troy Benjegerdes wrote:
>
>> On Thu, Feb 09, 2006 at 04:01:34PM -0600, Anthony Liguori wrote:
>>
>>> Jim C. Brown wrote:
>>>
>>>> -kernel-kqemu virtualizes ring 0 code.
>>>>
>>>> So it basically makes qemu do what VMware does.
>>>>
>>>> IIRC someone reported a 33% speedup with the new option.
>>>>  
>>>
>>> That was me.  That was a 33% speedup on win2k startup time.  kqemu 
>>> (user only) has a negligible impact on win2k startup time which 
>>> suggests this is mostly ring 0 code running which would make it a 
>>> good benchmark for kernel-kqemu performance.
>>>
>>> This was a terribly unscientific benchmarking so don't read too much 
>>> into it.
>>>
>>> Regards,
>>>
>>> Anthony Liguori
>>
>>
>> My win2k guest (with SP4, but not any updates) seemed to hang on startup
>> with -kernel-kqemu.
>
>
> Are you using -m 256 by any chance? I get this result with around that 
> much ram allocated to the guest. -m 160 (or less) or -m 384 (or more) 
> works perfectly here..
>
>
>
Great tip -thank you!

I have been having the same headache as others: Linux and RectOS guests 
constantly crashing in the very same spot when -kernel-kqemu is enabled.
I just made a quick test adding -m 384 and ReactOS suddenly boots all 
the way with -kernel-kqemu enabled!
My test line was:
qemu.exe -kernel-kqemu -m 384 -L ./bios -boot c -hda c.img

Best regards,
Dan Sandberg

Re: [Qemu-devel] -kernel-kqemu - bug with -m 256 <-please disregard

[Dan Sandberg]

Dan Sandberg wrote:

> Brad Campbell wrote:
>
>> Troy Benjegerdes wrote:
>>
>>> On Thu, Feb 09, 2006 at 04:01:34PM -0600, Anthony Liguori wrote:
>>>
>>>> Jim C. Brown wrote:
>>>>
>>>>> -kernel-kqemu virtualizes ring 0 code.
>>>>>
>>>>> So it basically makes qemu do what VMware does.
>>>>>
>>>>> IIRC someone reported a 33% speedup with the new option.
>>>>>  
>>>>
>>>>
>>>> That was me.  That was a 33% speedup on win2k startup time.  kqemu 
>>>> (user only) has a negligible impact on win2k startup time which 
>>>> suggests this is mostly ring 0 code running which would make it a 
>>>> good benchmark for kernel-kqemu performance.
>>>>
>>>> This was a terribly unscientific benchmarking so don't read too 
>>>> much into it.
>>>>
>>>> Regards,
>>>>
>>>> Anthony Liguori
>>>
>>>
>>>
>>> My win2k guest (with SP4, but not any updates) seemed to hang on 
>>> startup
>>> with -kernel-kqemu.
>>
>>
>>
>> Are you using -m 256 by any chance? I get this result with around 
>> that much ram allocated to the guest. -m 160 (or less) or -m 384 (or 
>> more) works perfectly here..
>>
>>
>>
> Great tip -thank you!
>
> I have been having the same headache as others: Linux and RectOS 
> guests constantly crashing in the very same spot when -kernel-kqemu is 
> enabled.
> I just made a quick test adding -m 384 and ReactOS suddenly boots all 
> the way with -kernel-kqemu enabled!
> My test line was:
> qemu.exe -kernel-kqemu -m 384 -L ./bios -boot c -hda c.img
>
> Best regards,
> Dan Sandberg
>
Sorry!
Monday morning, coffee machine broken and my test far TOO quick since I 
forgot to start the kqemu service first and used a double-click script 
with the result that I missed the warning message that kqemu actually 
wasn´t running.
With -kernel-kqemu enabled I now get the same nasty bug as before 
regardless of -m 160, 256, 384 or 512..
Best regards,
Dan Sandberg

>
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>

Re: [Qemu-devel] -kernel-kqemu

[Fabrice Bellard]

Hi,

I just found the problem related to the -kernel-kqemu bug with win2k and 
'-m 256'. But fixing the bug is another matter !

Fabrice.

Brad Campbell wrote:
> Troy Benjegerdes wrote:
> 
>> On Thu, Feb 09, 2006 at 04:01:34PM -0600, Anthony Liguori wrote:
>>
>>> Jim C. Brown wrote:
>>>
>>>> -kernel-kqemu virtualizes ring 0 code.
>>>>
>>>> So it basically makes qemu do what VMware does.
>>>>
>>>> IIRC someone reported a 33% speedup with the new option.
>>>>  
>>>
>>> That was me.  That was a 33% speedup on win2k startup time.  kqemu 
>>> (user only) has a negligible impact on win2k startup time which 
>>> suggests this is mostly ring 0 code running which would make it a 
>>> good benchmark for kernel-kqemu performance.
>>>
>>> This was a terribly unscientific benchmarking so don't read too much 
>>> into it.
>>>
>>> Regards,
>>>
>>> Anthony Liguori
>>
>>
>> My win2k guest (with SP4, but not any updates) seemed to hang on startup
>> with -kernel-kqemu.
> 
> 
> Are you using -m 256 by any chance? I get this result with around that 
> much ram allocated to the guest. -m 160 (or less) or -m 384 (or more) 
> works perfectly here..

Re: [Qemu-devel] -kernel-kqemu

[Anthony Liguori]

Hetz Ben Hamo wrote:
> Hi,
>
> I´m sorry, but I´ve been following the CVS commits and I haven´t
> understood this new option (fabrice has not written anything to this
> list about it and what it does for the end user).
>
> Could someone please shed some light for me and others who didn´t
> understand whats going on please?
>   
I can take a stab although hopefully Fabrice will post something in the 
near future..

kqemu and qvm86 allow ring 3 code to run on bare metal.  For ring 0 
code, both defer to qemu which does binary translation.

-kernel-kqemu allows portions of ring 0 code to run on bare metal.  
Presumably, this means kqemu now has a block scanner and can determine 
whether or not a block contains sensitive instructions.  It could be 
interpreting those blocks or it may just always return to qemu--it's 
unclear since kqemu is not open source.

The changes in qemu just allow for kqemu to try and execute ring 0 code 
it seems.  All the interesting changes are probably in kqemu itself...

Regards,

Anthony Liguori
> Thanks,
> Hetz
>
> On 2/9/06, Jim C. Brown <jma5@umd.edu> wrote:
>   
>> On Wed, Feb 08, 2006 at 06:04:35PM -0500, Jim C. Brown wrote:
>>     
>>> This sounds like an interesting option. Qemu has moved one step closer to
>>> VMware...
>>>
>>>       
>> I should probably add, that this new option (-kernel-kqemu) allows for speeds
>> very close to VMware because it allows kqemu to virtualize most ring 0 code
>> (in addition to ring 3 code).
>>
>> Can't wait to see qvm86 make use of the new API.
>>
>>     
>>> --
>>> Infinite complexity begets infinite beauty.
>>> Infinite precision begets infinite perfection.
>>>
>>>
>>> _______________________________________________
>>> Qemu-devel mailing list
>>> Qemu-devel@nongnu.org
>>> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>>>
>>>       
>> --
>> Infinite complexity begets infinite beauty.
>> Infinite precision begets infinite perfection.
>>
>>
>> _______________________________________________
>> Qemu-devel mailing list
>> Qemu-devel@nongnu.org
>> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>>
>>     
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>
>   

Re: [Qemu-devel] -kernel-kqemu

[Fabrice Bellard]

Hi,

I will update the documentation about "-kernel-kqemu" soon.

To be short: as some people already noticed, this option allows to run 
user code and most of the kernel code on "bare metal". The result is 
usually a noticable speed up. Only the following guest OSes are 
supported: Linux, Windows 2000 or XP. The installation of Windows 
2000/XP must be run without the -kernel-kqemu option.

I did not test the win32 and x86_64 versions of kqemu yet, but the i386 
version is usable.

Fabrice.

Re: [Qemu-devel] -kernel-kqemu

[Kazu]

Sent: Friday, February 10, 2006 7:01 AM Fabrice Bellard wrote:

> Hi,
>
> I will update the documentation about "-kernel-kqemu" soon.
>
> To be short: as some people already noticed, this option allows to run
> user code and most of the kernel code on "bare metal". The result is
> usually a noticable speed up. Only the following guest OSes are
> supported: Linux, Windows 2000 or XP. The installation of Windows
> 2000/XP must be run without the -kernel-kqemu option.
>
> I did not test the win32 and x86_64 versions of kqemu yet, but the i386
> version is usable.
>

I tested it on Windows XP host. But linux.img in linux-test package and
Win2k don't boot.

A strange thing is that when I used RedHat 7.2 guest, user-kqemu is much
slower than no-kqemu.


On Fedora Core 4 host, Win2k boots fine and is faster than user-kqemu and
no-kqemu. Great!

linux.img from linux-test package also works fine with kernel-kqemu,
user-kqemu on FC4 host.

But FC4-i386-rescuecd.iso guest doesn't boot with kernel-kqemu. It works
with user-kqemu and no-kqemu.
I can see about TSC error messages.

Morphix.iso doesn't boot with kernel-kqemu. It works with user-kqemu and
no-kqemu.
I couldn't see booting logo.

And a strange thing is that when I used Redhat 7.2 guest, user-kqemu is much
slower than no-kqemu.  And kernel-kqemu is much slower than user-kqemu.
It is the same result as on WinXP host.


I seems that Linux guest doesn't work well.

PS. user-kqemu means no option.

Regards,
Kazu

Re: [Qemu-devel] -kernel-kqemu

[Brad Campbell]

Jim C. Brown wrote:
> This sounds like an interesting option. Qemu has moved one step closer to
> VMware...
> 

Win2k-SP4 would not install with -kernel-kqemu, but it runs with it...
All I have to say is..
Oh..My..God.. I'm getting about 60-70% kqemu time. The user interface responsiveness has gone 
through the roof.
I've never used VMware, but this is getting toward what Win4Lin used to be like with Win95..

W00t!


-- 
"Human beings, who are almost unique in having the ability
to learn from the experience of others, are also remarkable
for their apparent disinclination to do so." -- Douglas Adams

Re: [Qemu-devel] -kernel-kqemu

[Christian MICHON]

On 2/9/06, Jim C. Brown wrote:
> This sounds like an interesting option. Qemu has moved one step closer to
> VMware...
>

It hangs my XP host with 100% cpu eaten up, no way to stop qemu,
or kqemu. I have to reboot, and my linux clients freezes very early

Src=CVS (yesterday 09/02/2006)
Host=XP+mingw32
Guest=slax frodo 5.0.6 bootcd (no hdd image)
Kqemu=kqemu-1.3.0pre3 (directly copied .sys into \windows\system32\drivers)

Has this new option been tested only on linux hosts ?

Christian

Re: [Qemu-devel] -kernel-kqemu

[Jim C. Brown]

On Fri, Feb 10, 2006 at 02:29:39PM +0100, Christian MICHON wrote:
> On 2/9/06, Jim C. Brown wrote:
> > This sounds like an interesting option. Qemu has moved one step closer to
> > VMware...
> >
> 
> It hangs my XP host with 100% cpu eaten up, no way to stop qemu,
> or kqemu. I have to reboot, and my linux clients freezes very early
> 
> Has this new option been tested only on linux hosts ?
> 

Yes.

Apparently it doesn't work on w32 hosts yet.

> Christian
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
> 

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.