On 5 April 2018 at 14:34, KONRAD Frederic <frederic.kon...@adacore.com> wrote: > On 04/05/2018 03:31 PM, Philippe Mathieu-Daudé wrote: >> Luckily in my case I can workaround with a 4K SRAM, but it took me hours >> to track this down until Paolo helped me.
> I seems to remember having the same pain with the mmio execution > stuff. It triggered a bad ram pointer as far a I remember. I think this would once have ended up in report_bad_exec() which prints a more-or-less helpful message. However for CPUs which now implement the unassigned-access or transaction-failed hooks you'll get a guest exception instead now, I guess. Incidentally we really ought to fix the interaction of mmio-exec and migration, rather than just leaving it as it is (disabling migration for xilinx_qspips device). thanks -- PMM
