nfs: patch up MNT_LAZY on sync

[mjg]

Summary

It is a de facto noop, just make it less costly.

quite frankly the code looks like a bug. did you mean to do LK_NOWAIT locking in this case instead?

if the doing nothing is indeed intended, this patch accomplishes it cheaper

[rmacklem]

Yea, it looks like it was coded this way (MNT_LAZY check in loop)
over twenty years ago (I looked in FreeBSD5. Couldn't be bothered
looking at even older code).

I agree it just seems silly to go through the loop doing nothing.

sys/fs/nfsclient/nfs_clvfsops.c

1911

I might stick it after the check for forced dismount
below, just in case something cares about the return
value someday. (The return value appears to be ignored
now.)

[rmacklem]

Just fyi, the buffer cache code I did for NFS way back in
the 1980s was not released as open source, so any code
in FreeBSD's NFS client related to the buffer cache or similar
was written by someone else. When I did the "new" NFS code
with NFSv4 in it, I intentionally cribbed all that stuff out of the
old FreeBSD NFS code, so that it would be "bug compatible"
out of the box.

[rmacklem]

Looks fine to me now.