Re: How to use designware-root-port and designware-root-host devices ?

[Arthur Tumanyan]

Hi,

I just tried to run mcimx7d-sabre machine this way:

${HOME}/cosim/usr/local/bin/qemu-system-arm -M mcimx7d-sabre -m 2G \
-kernel ${HOME}/cosim-arm/buildroot/output/images/uImage \
        --initrd ${HOME}/cosim-arm/buildroot/output/images/rootfs.cpio.gz \
-nographic \
-net nic -net user 

and it just prints this and do nothing: qemu-system-arm: warning: nic imx.enet.1 has no peer

Based on what I see in the mcimx7d-sabre.c , it configures just very basic things, no PCIe at all (may be I'm wrong ;) )

Is there any idea what goes wrong here ? Maybe someone has experience with running this machine ?

Before starting to create my own virt machine I would like to know I'm not missing the existing ways to run this.

Thanks in advance,

Arthur

On Thu, Jun 20, 2024 at 11:37 PM Arthur Tumanyan <arthurtumanyan@gmail.com> wrote:

Thanks for the answers, I could move forward a bit more. I'm going/I need to to create a "virt" machine with designware PCI controller for simulation purposes. Will get back with progress in case anyone is interested in results. Thank you again for your time and support.
Arthur

On Thu, Jun 20, 2024, 23:05 Peter Maydell <peter.maydell@linaro.org> wrote:

On Thu, 20 Jun 2024 at 18:34, Thomas Huth <thuth@redhat.com> wrote:
>
> On 20/06/2024 10.28, Arthur Tumanyan wrote:
> >  From the other hand the device is declared as non pluggable:
> > dc->user_creatable = false;
>
> Well, that means that you cannot use those with "-device". They can only be
> instantiated via the code that creates the machine.
>
> > Can you please help me to use designware-root-host/port devices ?
>
> It seems like the i.MX7 SABRE machine is using this device, so instead of
> "-M virt", you could have a try with "-M mcimx7d-sabre" (and a kernel that
> supports this machine) instead.

Right -- these devices are the PCIe controller that's used on the i.MX7
and i.MX6 SoCs, and they're automatically created when you use a machine
type that uses one of those SoCs. The "virt" board doesn't use that
PCIe controller, it uses the "generic PCIe bridge" TYPE_GPEX_HOST
(and you automatically get a PCIe controller when you use the virt board).
You can't change the PCIe controller type of a QEMU machine from
the command line, you have to configure the guest to use the controller
the machine type provides.

thanks
-- PMM