add NFSv4 support for va_birthtime

[rmacklem]

Summary

There is a NFSv4 file attribute called TimeCreate
that can be used for va_birthtime.
This patch enables support for setting this attribute
in the server.

It also allows the client to
acquire and set the attribute for a NFSv4
server that supports the attribute.

Test Plan

Tested via the stat(1) command and futimens(2)
system call for an NFSv4 mount between a
patched FreeBSD client and server.
A mount to an unpatched FreeBSD server
did not set the attribute, as expected and
the Linux client that does not use TimeCreate
still functioned as expected (no change in
behaviour).

[markj]

Thanks, this looks good to me.

sys/fs/nfsclient/nfs_clport.c

419

Does this actually need to be conditional? Note that vop_stdstat() also initializes va_birthtime to the same dummy value of {.tv_sec = -1, .tv_nsec = 0}.

[rmacklem]

sys/fs/nfsclient/nfs_clport.c

419

Yea, I'm "on the fence" about this. As you note, it isn't currently needed.
I did it this way in case the dummy value in vop_stdstat() changes
someday. (ie The dummy in the NFS client does not need to be
changed as well, if the dummy value in vop_stdstat() changes.)

[markj]

sys/fs/nfsclient/nfs_clport.c

419

We should probably make the default value a named constant to avoid this problem. I believe anyone changing the default would have to audit all filesystems anyway (msdosfs poses the same problem for instance), so I'm not sure it's worth coding defensively here, but it's a minor detail.

[rmacklem]

sys/fs/nfsclient/nfs_clport.c

419

In a sense it is. VNOVAL, although it is just -1 and is
applied to the tv_sec field.

I agree that doing the assignment unconditionally
makes sense, so I'll commit it that way.