Re: AW: Colibri imx8qxp: Missing kernel boot module

[Julien Grall <julien@xxxxxxx>]

Hi Daniel,

On 17/08/2020 11:29, Daniel Wagner2 wrote:
> > -----Ursprüngliche Nachricht-----
> > Von: Peng Fan <peng.fan@xxxxxxx>
> > Gesendet: Montag, 17. August 2020 12:02
> > An: Daniel Wagner2 <Daniel.Wagner2@xxxxxxxxxxxxxxxxxx>; xen-
> > users@xxxxxxxxxxxxxxxxxxxx
> > Betreff: RE: Colibri imx8qxp: Missing kernel boot module
> >
> > > Subject: AW: Colibri imx8qxp: Missing kernel boot module
> > >
> > >
> > >
> > > > -----Ursprüngliche Nachricht-----
> > > > Von: Peng Fan <peng.fan@xxxxxxx>
> > > > Gesendet: Montag, 17. August 2020 11:47
> > > > An: Daniel Wagner2 <Daniel.Wagner2@xxxxxxxxxxxxxxxxxx>; Julien Grall
> > > > <julien@xxxxxxx>; xen-users@xxxxxxxxxxxxxxxxxxxx
> > > > Betreff: RE: Colibri imx8qxp: Missing kernel boot module
> > > >
> > > > > Subject: AW: Colibri imx8qxp: Missing kernel boot module
> > > > >
> > > > >
> > > > >
> > > > > > -----Ursprüngliche Nachricht-----
> > > > > > Von: Xen-users <xen-users-bounces@xxxxxxxxxxxxxxxxxxxx> Im
> > > > > > Auftrag von
> > > > > Julien
> > > > > > Grall
> > > > > > Gesendet: Freitag, 14. August 2020 21:46
> > > > > > An: Daniel Wagner2 <Daniel.Wagner2@xxxxxxxxxxxxxxxxxx>; xen-
> > > > > > users@xxxxxxxxxxxxxxxxxxxx; Peng Fan <peng.fan@xxxxxxx>
> > > > > > Betreff: Re: Colibri imx8qxp: Missing kernel boot module
> > > > > >
> > > > > > On 10/08/2020 16:32, Daniel Wagner2 wrote:
> > > > > > > Hello xen-users,
> > > > > > Hello,
> > > > > >
> > > > > > > we are trying to get Xen running on a Toradex "Colibri iMX8X"
> > > > > > > module (see [1] at the bottom of this email), which features
> > > > > > > the
> > > > > > > iMX8 QXP
> > > > > prozessor.
> > > > > > > We found that NXP has a Xen reference implementation [2][3]
> > > > > > > for their MEK Module and tried to port that to the Toradex
> module.
> > > > > > > When booting via the bootscript [4], which is unaltered except
> > > > > > > for the "dom0fdt_file" and "xenhyper_bootargs" variables, we
> > > > > > > get the following
> > > > > > > error:
> > > > > > >
> > > > > > > [...] (See [5] for complete bootlog)
> > > > > > > (XEN) *** LOADING DOMAIN 0 ***
> > > > > > > (XEN) Missing kernel boot module?
> > > > > > > (XEN)
> > > > > > > (XEN) ****************************************
> > > > > > > (XEN) Panic on CPU 0:
> > > > > > > (XEN) Could not set up DOM0 guest OS
> > > > > > > (XEN) ****************************************
> > > > > > > (XEN)
> > > > > > > (XEN) Reboot in five seconds...
> > > > > > >
> > > > > > > For Dom0 we took the linux-toradex kernel source, ran "make
> > > xenconfig"
> > > > > > > [6], which should add Xen-support and rebuilt the Image with
> > > > > > > "make
> > > > > Image".
> > > > > > > For dom0 DTB [7] we copied the dom0 DTS NXP uses for their
> > > > > MEK-Module
> > > > > > > [8] and only adjusted the "bootargs" parameter.
> > > > > > > This Linux Image was able to run after we rebuilt and booted
> > > > > > > it without
> > > > > Xen.
> > > > > > Which Device-Tree did you for boot Linux without Xen?
> > > > > Used the same Device-Tree Binary that we want to use with Xen
> > > > > which is described in [7]. In [7] you can see rows 23 and 24,
> > > > > which include the Devicetree as it is supplied by Toradex.
> > > > > To boot linux with our dom0 DTB, the u-boot variable fdt_file was
> > > > > changed from fsl-imx8qxp-colibri-eval-v3.dtb to
> > > > > fsl-imx8qxp-colibri-eval-v3-dom0.dtb.
> > > >
> > > > Did you use the U-Boot cmd 'xenmmcboot' or 'xennetboot'?
> > > To boot linux without Xen the u-boot cmd 'boot' is used.
> > > To boot Xen we use 'run xenmmcboot' (in an adjusted form to account
> > > for different filenames and bootargs, see [4]).
> > So kernel Image loaded to 0x80a00000, right?
> >
> > Please load kernel image before run xenmmcboot if not.
> The xenmmcboot script [4] line 25 loads the linux image.
> Inserting the values for the variables in this line gives:
> 'fatload mmc 0:1 0x92000000 image'
> So we load the kernel image to 0x92000000. Is this not the right adress?
> How can the right adress, to load the kernel image to, be determined?
With U-boot, you will have to unfortunately determine the address 
manually. I usually use the address provided by U-boot for baremetal 
boot and load both Xen and Linux close to each other.
> Loading it to 0x80a00000 yields the same error ("Missing kernel boot
> module?")
It may be possible that the Device-Tree node is incorrect. Would it be 
possible to dump the Device-Tree node /chosen/ in either U-boot or Xen?
Cheers,

--
Julien Grall