Use vmspace->vm_stacktop in place of sv_usrstack in more places.

[jhb]

Summary

Obtained from: CheriBSD

[markj]

sys/kern/kern_resource.c

779

I think this actually needs to be round_page(p->p_vmspace->vm_stacktop).

[jhb]

sys/compat/linux/linux_mmap.c

179

This one might also need the roundup? I guess the stack limit includes the sub-page offset?

[markj]

sys/compat/linux/linux_mmap.c

179

Yes in principle, though we don't enable stack randomization for Linux processes. Indeed, the stack limit includes the sub-page offset. Though now I wonder if it'd be better to simply store the rounded stack address. I don't think the precise value is used except for setting the initial stack pointer.

[markj]

sys/compat/linux/linux_mmap.c

179

This isn't easy to do without introducing a new field, since PROC_PS_STRINGS() needs the un-rounded stack top. So for now I think we should just add rounding to this diff.

[jhb]

sys/compat/linux/linux_mmap.c

179

In CheriBSD we have p_psstrings still even after merging the PROC_PS_STRINGS change, but I have removed a local p_usrstack field we used to have and am using vm_stacktop to replace p_usrstack (and these changes are inspired by changes in CheriBSD to remove p_usrstack in favor of vm_stacktop).

[jhb]

More round_page().