[Qemu-devel] qemu-fast kernel patch question

[Qemu-devel] qemu-fast kernel patch question

[Lindsay Mathieson]

Hi All - great project, have managed to instal win98 and its usable 
under linux - thats a miracle.

I'm using qemu 0.6.0 (website binaries) and Mandrake Official 10, kernel 
2.6.3-15mdk and want to modify it for the fast option. I have 
include/asm/page.h, which looks slightly different to what is described 
but essentially the same. However there is no vmlinux.lds anywhere to 
modify.

Any suggestions ?

Thanks - Lindsay

Re: [Qemu-devel] qemu-fast kernel patch question

[Lindsay Mathieson]

Lindsay Mathieson wrote:

> I'm using qemu 0.6.0 (website binaries) and Mandrake Official 10, 
> kernel 2.6.3-15mdk and want to modify it for the fast option. I have 
> include/asm/page.h, which looks slightly different to what is 
> described but essentially the same. However there is no vmlinux.lds 
> anywhere to modify.


I think I figured it out - I'm meant to apply the 
linux-2.6-qemu-fast.patch patch ? am compiling it now, we'll see if it 
works in a couple of hours ....

Which lead to my next dumb question - the fast patch for the 2.6 kernel, 
I'm assuming its to allow qemu-fast to run under linux 2.6, not for 
linux 2.6 to be used as a guest under qemu-fast, the docs seemed a bit 
ambigous on this.

Thanks - lindsay

Re: [Qemu-devel] qemu-fast kernel patch question

[Mulyadi Santosa]

Hello

> Which lead to my next dumb question - the fast patch for the 2.6 kernel,
> I'm assuming its to allow qemu-fast to run under linux 2.6, not for
> linux 2.6 to be used as a guest under qemu-fast, the docs seemed a bit
> ambigous on this.

The correct meaning is...the "qemu-fast patch" means to be applied on guest 
kernel so the guest's kernel memory layout won't introduce any conflict when 
qemu-fast is doing direct MMU operation on host

Hope it helps 

regards

Mulyadi

Re: [Qemu-devel] qemu-fast kernel patch question

[Lindsay Mathieson]

[-- Attachment #1: Type: text/plain, Size: 654 bytes --]

Mulyadi Santosa wrote:

>Hello
>
>  
>
>>Which lead to my next dumb question - the fast patch for the 2.6 kernel,
>>I'm assuming its to allow qemu-fast to run under linux 2.6, not for
>>linux 2.6 to be used as a guest under qemu-fast, the docs seemed a bit
>>ambigous on this.
>>    
>>
>
>The correct meaning is...the "qemu-fast patch" means to be applied on guest 
>kernel so the guest's kernel memory layout won't introduce any conflict when 
>qemu-fast is doing direct MMU operation on host
>
>Hope it helps 
>
>  
>
Yup, thanks - I don't need to patch my host kernel to run qemu-fast then.

Should qemu-fast with with Win98 then ?

Thanks - Lindsay

[-- Attachment #2: Type: text/html, Size: 1087 bytes --]

Re: [Qemu-devel] qemu-fast kernel patch question

[Mulyadi Santosa]

Hello

> Yup, thanks - I don't need to patch my host kernel to run qemu-fast then.
> Should qemu-fast with with Win98 then ?

Do you mean: 
1. applying qemu-fast patch to Win98 guest? --> Not a chance :-)
2. running qemu-fast on top of Win98 host? Well, I haven't tried it 
personally. But, if you can run qemu-fast on Win98 host, I doubt if qemu-fast 
can do direct MMU access. The reason is: to do direct MMU access, the 
developers need to know to request that into host kernel. On Linux or other 
open source kernel, we can find it out, but on closed kernel, this would be 
difficult.....

The conclusion: just run the usual "qemu" on top of Win98 host

regards

Mulyadi

Re: [Qemu-devel] qemu-fast kernel patch question

[Johannes Martin]

Hi,

On Sun, 25 Jul 2004, Mulyadi Santosa wrote:
> Do you mean:
> 1. applying qemu-fast patch to Win98 guest? --> Not a chance :-)
> 2. running qemu-fast on top of Win98 host? Well, I haven't tried it
> personally. But, if you can run qemu-fast on Win98 host, I doubt if qemu-fast
> can do direct MMU access. The reason is: to do direct MMU access, the
> developers need to know to request that into host kernel. On Linux or other
> open source kernel, we can find it out, but on closed kernel, this would be
> difficult.....
>
> The conclusion: just run the usual "qemu" on top of Win98 host

Here's a quote from qemu-tech.html:
  ---
  2.10 MMU emulation

  For system emulation, QEMU uses the mmap() system call to emulate the
  target CPU MMU. It works as long the emulated OS does not use an area
  reserved by the host OS (such as the area above 0xc0000000 on x86
  Linux).
  ---

As far as I understand this, the qemu kernel patch makes the guest kernel
use a different reserved area than the linux host. So it shouldn't matter
which kernel I apply the patch to, right? If I patch the (linux) host
kernel, I should be able to boot any unpatched (linux) guest kernel, if I
don't patch the (linux) host kernel, I have to patch my (linux) guest
kernels. Or do we need a patch-aware qemy-fast to run on a patched host?

Assuming that OS XYZ does not use the area around 0xc0000000, it should
work inside qemu-fast, right? And if it does use that area, it should work
inside qemu-fast on a patched host?

Maybe one of the wizards could clarify this...

Thanks
	Johannes

Re: [Qemu-devel] qemu-fast kernel patch question

[Jim C. Brown]

On Sun, Jul 25, 2004 at 12:24:16PM +0200, Johannes Martin wrote:
> Hi,
> 
> Here's a quote from qemu-tech.html:
>   ---
>   2.10 MMU emulation
> 
>   For system emulation, QEMU uses the mmap() system call to emulate the
>   target CPU MMU. It works as long the emulated OS does not use an area
>   reserved by the host OS (such as the area above 0xc0000000 on x86
>   Linux).
>   ---
> 
> As far as I understand this, the qemu kernel patch makes the guest kernel
> use a different reserved area than the linux host. So it shouldn't matter
> which kernel I apply the patch to, right? If I patch the (linux) host
> kernel, I should be able to boot any unpatched (linux) guest kernel, if I
> don't patch the (linux) host kernel, I have to patch my (linux) guest
> kernels. Or do we need a patch-aware qemy-fast to run on a patched host?
> 
> Assuming that OS XYZ does not use the area around 0xc0000000, it should
> work inside qemu-fast, right? And if it does use that area, it should work
> inside qemu-fast on a patched host?
> 
> Maybe one of the wizards could clarify this...
> 
> Thanks
> 	Johannes
> 

You could patch the host not to use that area, but then it would have to use
another area ... thus you'd have to patch the guest anyways, regardless of the
host.

Fabrice has said that the eventual idea is to have qemu-fast detect which areas
are not accessable via mmap() and use the softmmu to emulate for only those
areas. The rest would be accessed via the faster mmap(). So, if you patch
qemu-fast in the right way, you might not have to patch the guest or the host. :)

> 
> _______________________________________________
> 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] qemu-fast kernel patch question

[Mulyadi Santosa]

Hello Jim :-)

First thank you to help correcting my logics :-)

> You could patch the host not to use that area, but then it would have to
> use another area ... thus you'd have to patch the guest anyways, regardless
> of the host.
Agree....why I don't think in such a logic before...hrrrrrr :-)

> Fabrice has said that the eventual idea is to have qemu-fast detect which
> areas are not accessable via mmap() and use the softmmu to emulate for only
> those areas. The rest would be accessed via the faster mmap(). So, if you
> patch qemu-fast in the right way, you might not have to patch the guest or
> the host. :)

So, that would be running mmap() scanning on early boot of Qemu, am  I right? 
from there, Qemu can take decision which area is free and which is already 
taken by host....Hm....looks like Qemu needs additional kernel module 
inserted to host's kernel to help Qemu on "memory layout" negotiation....:-)

regards

Mulyadi