nvme: avoid gcc -Werror=int-to-pointer-cast on 32-bit arch

[rlibby]

Summary

---

There may be a nicer way to spell this.

Test Plan

make buildkernel on amd64 and i386, examine warnings

[imp]

why the crazy vm_offset_t instead of uintptr_t??

sys/dev/nvme/nvme_ctrlr.c

1366–1367

why not (uintptr_t) here and below?

[rlibby]

In D45750#1043538, @imp wrote:

why the crazy vm_offset_t instead of uintptr_t??

Yeah, I think you're right. I waffled on that, I'm happy to make it uintptr_t, I believe they're always the same size.

My drive-by understanding was that in the is_user block, npc->addr is a user address and in that sense vm_offset_t may technically be more appropriate to represent it than a pointer (and vmapbuf casts the pointer to vm_offset_t)... but I'm not proposing we change the vmapbuf prototype for this, and anyway that doesn't apply to the second case.

I'll fix this up with uintptr_t this evening.

[rlibby]

imp: s/vm_offset_t/uintptr_t/