Re: [Qemu-devel] [RFC PATCH] target/s390x/cpu_models: Set some additional feature bits for the "qemu" CPU

[Aurelien Jarno]

On 2017-05-18 03:55, Thomas Huth wrote:
> On 17.05.2017 18:49, David Hildenbrand wrote:
> > On 17.05.2017 17:35, Thomas Huth wrote:
> > 2. I would recommend to not enable STFLE for now. Why?
> > 
> > It is/was an indication that the system is running on a z9 (and
> > implicitly has the basic features). This was not only done because
> > people were lazy, but because this bit was implicitly connected to other
> > machine properties.
> 
> Uh, that's ugly!
> 
> > One popular example is the "DAT-enhancement facility 2". It introduced
> > the "LOAD PAGE TABLE ENTRY ADDRESS" instruction, but no facility bit was
> > introduced. SO there is no way to check if the instruction is available
> > and actually working.
> 
> Does the Linux kernel use this instruction at all? I just grep'ed
> through the kernel sources and did not find it. If the Linux kernel does
> not use it, I think we should ignore this interdependency and just
> provide the STFLE feature bit to the guest - since recent Linux kernels
> depend on it.
> 
> > Please note that we added a feature representation for this facility,
> > because this would allow us later on to at least model removal of such a
> > facility (if HW actually would drop it) on a CPU model level.
> 
> What about STFLE bit 78, according to my version of the POP, it says:
> 
> "The enhanced-DAT facility 2 is installed in the
>  z/Architecture architectural mode."

No that is different. IBM has chosen very confusing names for the DAT
related facilities:
- DAT-Enhancement Facility 1 => bits 3, 4 & 5
- DAT-Enhancement Facility 2 
- Enhanced-DAT Facility 1    => bit 8
- Enhanced-DAT Facility 2    => bit 78

Aurelien

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
address@hidden                 http://www.aurel32.net