Add SPDX tags and consistently use the same license text.

Incorporate further review feedback.

In D38195#867987, @imp wrote:

I think this is good. Have you tested on a stable machine?

In D38191#867986, @markj wrote:

cp already preserves holes in files when it uses copy_file_range(). At least, the generic implementation used by UFS and ZFS does. So do we really need a new option? To give consistent behaviour, it seems like it should be sufficient to look for a hole somewhere in the file and refuse to use mmap() if one is present.

Bump year

Explain why only amd64. Remove POSIXRULES.

Fix xo_finish() check

BTW, I'm also tempted to drop the -p ${POSIXRULES} part, which (with POSIXRULES=America/New_York) effectively means “when in doubt, assume you're in New York”. The hardcoded default has the same effect, there is no value in enshrining it in the Makefile.

A little background: by default, old zic produces zoneinfo files containing both 32-bit and 64-bit data, while new zic produces only 64-bit data. Dynamic i386 binaries should work correctly since they will be using a 32-bit version of new libc, but static i386 binaries will not work.

Incorporate further review feedback.

Use EXIT_SUCCESS / EXIT_FAILURE.

Fix further locking issues and add a test.

In D38036#864270, @imp wrote:

This looks fine, but I'd suggest a better commit message explaining the Coverity issues would be helpful

In D38038#864275, @imp wrote:

A better commit message would be useful here. Why do you need to ignore the return value from remove?

Catch up with VFS changes.

More locking issues.

Use IGNORE_FreeBSD_X.

Use uidfix after fixing release.

Remove empty MAKE_ARGS.

Address locking issues.

First round of review.

In D37753#858845, @debdrup wrote:

Shouldn't this be in section 4, since it's a driver and not related to file formats?

In D37753#858820, @markj wrote:

Is it possible to add some basic regression tests under tests/sys/fs/tarfs?

Spell Klara, Inc. consistently.

Remove Buenos Aires Convention verbiage.