arm64 pmap: do not panic on inability to insert ptp into trie

[alc]

sys/arm64/arm64/pmap.c

5772–5775

There is no need for

if (l2pg != NULL)
        l2pg->ref_count--;

here because this is the kernel pmap. A similar if statement could be eliminated from the amd64 version.

5775–5776

I could eliminate these two lines because !remove_pt does not (temporarily) map the page table page.

[alc]

Rename remove_pt to demote_kl2e to better reflect what it controls.

Convert a panic to a KASSERT.

Deindent some code.

[markj]

Do you have some local modifications to test the ADDR_IS_KERNEL(va) code path in pmap_enter_l2()?

sys/arm64/arm64/pmap.c

5777

Maybe, "kernel L3 page table page" to be a bit more precise?

[kib]

I think a more elaborate commit message would be helpful.

[alc]

See inline comment for an explanation.

[alc]

In D51220#1170373, @markj wrote:

Do you have some local modifications to test the ADDR_IS_KERNEL(va) code path in pmap_enter_l2()?

Yes, I have a debug sysctl procedure that tests this code path.

sys/arm64/arm64/pmap.c

5790–5791

While these operations are no longer needed when we perform pmap_remove_l2(), they are still needed after pmap_remove_l3_range() because it doesn't unmap the L3 PTP.