RISC-V superpage support, part 2/6.

[markj]

Summary

Superpage support for pmap_page_test_mappings(), pmap_protect(),
pmap_unwire().

[mhorne063_gmail.com]

sys/riscv/riscv/pmap.c

3673–3674

Does this comment still apply? I'm not sure what the difference is between dirty and modified is in this context, or where the read/write check is happening.

3683–3690

You may wish to convert this to bool as you did in pmap_unwire(), and same with the above signature. Not necessary as part of this patch however.

[jhb]

sys/riscv/riscv/pmap.c

2360

Should this be a pmap_invalidate_range() of the original sva and eva (you'd have to save the original sva somewhere)? Presumably at some point we could have pmap_invalidate_range() apply some logic to use pmap_invalidate_all() if there were too many pages? (Right now for bbl you might as well always use invalidate_all, but someday that might not be true.)

[markj]

Address Mitchell's comments.

[markj]

sys/riscv/riscv/pmap.c

2360

That's probably the right long-term direction, but I'd rather not do it in this change, as I'm just following amd64's example and I have no good way of testing the change (pmap_invalidate_range() just executes sfence.vma today).

3673–3674

Dirty and modified mean the same thing in this context, they're used interchangeably. (amd64 has a "modified" bit whereas riscv has a "dirty" bit.) Indeed, the comment should be removed, it was copied from arm64.

3683–3690

Thanks, might as well do it here.

[markj]

sys/riscv/riscv/pmap.c

2360

(BTW, there was some discussion along these lines in D12725.)

[markj]

Committed in r344106.