[Libre-soc-dev] [OpenPOWER-HDL-Cores] https://bugzilla.kernel.org/show_bug.cgi?id=213257

[Luke Kenneth Casson Leighton]

On Sat, May 29, 2021 at 1:21 AM Benjamin Herrenschmidt <
benh at kernel.crashing.org> wrote:

>
> You man in qemu ?


no, this is not qemu.  if it was qemu, we would have raised a qemu
bugreport.

this is an entirely new project that makes **direct**use of kvm
virtualisation,
**directly** calling the IOCTL for kvm single-step in the linux kernel.

https://git.libre-soc.org/?p=kvm-minippc.git;a=blob;f=main.c;h=42890f62dc8b04c3c1fa5e65e3138139c7708d58;hb=d55aaafa9a078ecb1d7420d3f54faf9679e865cb#l442

You have MSR:FP set ?
>

yes.

https://git.libre-soc.org/?p=kvm-minippc.git;a=blob;f=main.c;h=42890f62dc8b04c3c1fa5e65e3138139c7708d58;hb=d55aaafa9a078ecb1d7420d3f54faf9679e865cb#l341

the reason why the bugreport is being raised is because *there is no
failure*
*when running in non-single-step mode*.

when the guest program is run in "straight" (non-debugging,
non-single-step) mode
it produces the *correct results*.

when the guest program is run in KVM IOCTL SINGLE STEP mode, *all FP
registers are zero*.

hence raising a linux kernel bug.

l.