Hi, Rick,

It appears that panic occurs in acpi_atomic_read.  I think the most 
likely cause is that the acpi_generic_address is not pre-mapped.  Can 
you try the patch attached?

It will print registers mapped and accessed.  To use it, run the 
following command line before workload.

dmesg | grep GHES

Then try to find something like

GHES: gar accessed: x, xxxx

in kernel log when panic occurs.

Best Regards,
Huang Ying