Hello,
I am trying to migrate a STM32MP2 Linux VM from the CAmkES VMM to the
Microkit VMM.
After filling the few platform definitions in vgic.h and vmm.h, I can
boot the Linux VM up to the point where it faults during gic_init in linux:
vmm|ERROR: unexpected memory fault on address: 0x4ac20004
This address happens to maps the GICC interface
The Microkit VMM qemu example in the tutorial I used as a starting point
only maps the GICV interface, so I tried adding the following region to
the Microkit xml port description:
<memory_region name="gic_cpu" size="0x2000" phys_addr="0x4ac20000" />
which cause the Microkit trace failure:
ERROR: object 'frame_mr_gic_cpu_000000000', with paddr
0x00004ac20000..0x00004ac21000 is not in any valid memory region.
Below are the valid ranges of memory to be allocated from:
Valid ranges outside of main memory:
[0x000000000000..0x000040000000)
[0x000040000000..0x000048000000)
[0x000048000000..0x00004a000000)
[0x00004a000000..0x00004a800000)
[0x00004a800000..0x00004ac00000)
[0x00004ac00000..0x00004ac10000)
[0x00004ac11000..0x00004ac12000)
[0x00004ac12000..0x00004ac14000)
[0x00004ac14000..0x00004ac18000)
[0x00004ac18000..0x00004ac20000)
[0x00004ac21000..0x00004ac22000)
...
So there appears to be a hole exactly over 0x4ac20000..0x4ac21000, which
is where the GICC registers live.
Are these “valid ranges” derived directly from the kernel’s device
resources to the root task? before checking the xml memory_region ?
If so, what could cause the kernel to exclude exactly this page?
Since this setup works with the CAmkES VMM, I would not expect a general
device-tree parsing issue,
Is there a way to trace valid device memory ranges construction in
Microkit ?
Thanks for any hints
Best regards,
Christian
