ARMv6 pmap - fix the way how a decision is made that a page is under pv management

[skra]

Summary

PHYS_TO_VM_PAGE() function looks up a page in more segments, but first_page is connected only to the segment associated with vm_page_array. There could be more dynamically created segments for fictitious pages.

Probably, it's not so big problem on architectures where RAM physical address starts from zero. However, it's not a case for many ARM platforms.

On the other hand, AFAIK, fictitious pages under pv management are used only in new drm code.

[kib]

Besides DRM, another consumer of the dynamic managed fictitious page ranges is Xen. I am not sure about state of the Xen/arm port on FreeBSD.

sys/arm/arm/pmap-v6.c

2750 ↗

(On Diff #22153)

the `? TRUE : FALSE``` part is not needed, the comparison operator already produces the bool typed result. You might consider changing return type of is_managed() to bool as well.

[skra]

sys/arm/arm/pmap-v6.c

2750 ↗

(On Diff #22153)

Right. I saw that in this morning. About bool type, is it prefered now to slowly change all occurrences of boolean_t into bool?

[kib]

sys/arm/arm/pmap-v6.c

2750 ↗

(On Diff #22153)

I use bool type for all new code and when modifying old code. The only exception is when boolean_t is inscribed in the KPI and the pass over consumers is not desirable.