Re: [Xenomai] x86/ipipe: force lazy FPU management

Re: [Xenomai] x86/ipipe: force lazy FPU management

[Jan Kiszka]

Hi Philippe,

this commits looks fishy: the original code already disabled eager mode
reliably because of

if (cpu_has_xsaveopt && eagerfpu != DISABLE)
                        ^^^^^^^^^^^^^^^^^^^

right after the original disabling under I-pipe. Instead, the new code
now keeps xfeatures_mask untouched as it performs the disabling too
late. Could you clarify this?

Jan

-- 
Siemens AG, Corporate Technology, CT RDA ITP SES-DE
Corporate Competence Center Embedded Linux

Re: [Xenomai] x86/ipipe: force lazy FPU management

[Philippe Gerum]

On 06/14/2017 11:30 AM, Jan Kiszka wrote:
> Hi Philippe,
> 
> this commits looks fishy: the original code already disabled eager mode
> reliably because of
> 
> if (cpu_has_xsaveopt && eagerfpu != DISABLE)
>                         ^^^^^^^^^^^^^^^^^^^
> 
> right after the original disabling under I-pipe. Instead, the new code
> now keeps xfeatures_mask untouched as it performs the disabling too
> late. Could you clarify this?
> 

Messed up, aka -ENOBRAIN triggered. This was a spurious backport of the
fix I did for 4.9, as the former location of the eagerfpu fixup
introduced a weird bug. The fpu switch init code there postpones the
update of the eager FPU capability until the very end of the routine.

Commit reverted, I'll be issuing a patch within minutes.

Thanks for the heads up.

-- 
Philippe.