[Qemu-devel] booting a linux kernel from NFS

[Qemu-devel] booting a linux kernel from NFS

[Markus Schiltknecht]

Hi,

I get this error, when I don't give qemu a 'hda':

A disk image must be given for 'hda' when booting a Linux kernel.

Why is that? Isn't it completely legal to start without a IDE drive? For 
example with:

qemu \
   -kernel linux-test/bzImage-2.6.18 \
   -append "root=/dev/nfs nfsroot=172.20.0.1:/testing_root \
            ip=172.20.0.10::172.20.0.1:255.255.0.0:node1:eth0:off \
            ide2=noprobe ide3=noprobe ide4=noprobe" \
   -net nic,vlan=0,model=ne2k_isa \
   -net tap,vlan=0,script=/etc/qemu-ifup -m 10


Regards

Markus

Re: [Qemu-devel] booting a linux kernel from NFS

[Rob Landley]

On Tuesday 10 October 2006 11:46 am, Markus Schiltknecht wrote:
> Hi,
> 
> I get this error, when I don't give qemu a 'hda':

Use "/dev/zero".  (Several people have suggested that qemu should default 
to /dev/zero when you give it a kernel but don't give it a hard drive, but 
last I checked it still wasn't doing that.)

> A disk image must be given for 'hda' when booting a Linux kernel.
> 
> Why is that?  Isn't it completely legal to start without a IDE drive?

Because the way -kernel works is to create a fake bootsector internally, 
present it to the bios as the start of hda, and tell the bios to boot from 
that.  If you haven't got an hda, it gets confused.  (It probably shouldn't, 
since it's not actually _using_ it, but it does.)

Note that if you mmap /dev/zero it counts as an anonymous mapping, so feeding 
in /dev/zero for hda works just fine here.

Rob
-- 
"Perfection is reached, not when there is no longer anything to add, but when 
there is no longer anything to take away." - Antoine de Saint-Exupery

Re: [Qemu-devel] booting a linux kernel from NFS

[Markus Schiltknecht]

Rob Landley wrote:
> Because the way -kernel works is to create a fake bootsector internally, 
> present it to the bios as the start of hda, and tell the bios to boot from 
> that.  If you haven't got an hda, it gets confused.  (It probably shouldn't, 
> since it's not actually _using_ it, but it does.)
> 
> Note that if you mmap /dev/zero it counts as an anonymous mapping, so feeding 
> in /dev/zero for hda works just fine here.

Thank you, that works. And it's always good to know how it works.

Regards

Markus