libprocstat: Drop zfs_defs.c hack, including its _KERNEL define

[jrtc27]

Summary

Now that we have a _WANT_ZNODE we can use that instead of defining
_KERNEL, and we're able to move the code back into zfs.c using a real
znode_t pointer.

Whilst here, tidy the includes.

[jrtc27]

Trim and sort CFLAGS

[imp]

Thanks

lib/libprocstat/Makefile

71

What a long list of args. Any chance there's a make variable for all this?

lib/libprocstat/zfs.c

45

Why move it? Just tidying or was the old location now breaking? The move is fine, but I'm curious.

[jrtc27]

lib/libprocstat/Makefile

71

Of course not :) don't go looking for how many similar copies of this there are. The real answer is probably in part at least that it needs to vary between uses...

lib/libprocstat/zfs.c

45

Just tidying as a good citizen

[kib]

lib/libprocstat/zfs.c

82–83

Is this comment still relevant?

[jrtc27]

lib/libprocstat/zfs.c

82–83

Yes :( OpenZFS's SPL shadows sys/vnode.h with its own that doesn't include ours; IIRC the end result is struct vnode is an incomplete type here. OpenZFS is still a pile of hacks in the name of trying to pretend both its supported OSes are Solaris, rather than using its own namespaced abstractions...

[imp]

lib/libprocstat/zfs.c

82–83

Indeed. The OpenZFS integration is its own special kind of hell. Too much is inlined, so too many things need to know about other things and the namespace abstractions are just barely up to the task, but here's one of the holes for sure.

[kib]

lib/libprocstat/zfs.c

82–83

Your rant is better than the cryptic comment above, I suggest you to put in there. Perhaps in the separate commit to not delay this one.

Also, in commit above, if kept, vnode.h should be replaced by freebsd/sys/sys/vnode.h or such.

[jrtc27]

Update comment

[kib]

Thanks