Re: Re: [Qemu-devel] kqemu 0.6.2 and XP

[Ben Taylor]

"Andreas Bollhalder" <address@hidden>
> 
> I tried the math on my WinXP host with Win2k guest.
> 
> When using the following math with my 8192MB QCOW image, QEMU refuse
> to start:
> 
> 1 Block = 512 Bytes
> 
> Cylinders = Blocks / (Heads * Sectors)
> 
> (kBytes * 2) / (16 * 63)
> (8388608 kBytes * 2) / (16 * 63)
> 
> For 8192MB Image: -hdachs 16644,16,63

Yeah. Sorry about that. You probably want to use
something like 64 heads and 63 or 255 sectors/cylinder
in this case.  I expect that 16000 heads is probably 
a bit much. IIRC, Solaris tends to barf on big
volumes that have really big c/h/s translations.

  
> I brute forced the cylinder value, QEMU will start with a lower or
> equal value of 16383 cylinders.

Very interesting.  Wonder if that's some of ide limit.

> 
> I made my own math based on this:
> 
> Cylinders = Blocks / (Heads * (Sectors + 1)) - 1
> 
> (kBytes * 2) / (16 * 64) - 1
> (8388608 kBytes * 2) / (16 * 64) - 1
> 
> For 8192MB Image: -hdachs 16383,16,63

well, you should be able to make it ~4100 cylinders
just by increasing the heads from 16 to 64.

> 
> 
> I'm realy confused about what's right or not. I created the image with
> the QEMU Manager. Can anyone explain this ???

I think the large number of cylinders is a *bad* thing.
Sorry if I don't have a more technical term. It's not
from looking at the code, but from field experience when
dealing with really big Luns on a SAN.

Ben
> 
> Andreas
> 
> 
> Ben Taylor wrote:
> 
> > I forgot one very important part.  The calculation needs
> > to be <img size in bytes> / (heads * sectors per cylinder 
> > * 512 bytes) = cylinders. so in your case, the calculation is close
> 
> > well, the default sector size is 512, and you've done
> > your sizing in number of 1k blocks, your calculations was
> > close.  It should be:
> >
> >   blocks     secs/block      hds     sec/cyl
> >
> > (2 000 000   *   2 )    /  (   16   *   63 ) = 3968
> 
> > The first time I did the math, my 2g partition went
> > to 8g partition when I used the -hdachs parameters
> > cause I really screwed up the math.  (it appears if
> > you mess your c-h-s values, your image since may
> > change. :-)
> 
> > HTH, and sorry for the confusion.
> 
> > Ben
> 
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>