> Fixing that problem in tmpfs_subr.c::tmpfs_mem_info() is pretty
> simple, see diff #1.  With this diff, my previous example with df's
> and swapctl -a/-d works as I'd expect.
> 
> Furthermore, fixing that function to not look like it needs to know
> uvm internals (use only uvmexp.*), do much less work and therefore
> also to kill the concerns behind the XXX comments at the top, is also
> pretty simple, see diff #2.
> 
> A slightly different approach would be to use uvmexp.inactive rather
> than uvmexp.filepages.  I think I prefer this, actually, though I'd
> like to watch it in action a little longer, it might need something
> slightly different.

some random points:
- not all filepages are reclaimable.
  (eg. wired pages, mfs)
- uvmexp.inactive is not appropriate as it includes anonymous pages.
- this kind of calculation should be done in uvm, not tmpfs.
  (cf. uvm_reclaimable)
- you seem to misuse uvmexp values in diff #2.
- if you take this route, eventually you need to account buffer cache,
  pool idle pages, etc.

YAMAMOTO Takashi