This is the same as fd_reopen() except that it takes the original file
descriptor by reference and closes it if the reopen succeeded, replacing
it with the reopened file descriptor.

This is quite useful for the "I need to make this writable file
descriptor read-only before sealing something" case.

Signed-off-by: Allison Karlitskaya <[email protected]>

Add a new member to CopyFlags to request copying of fs-verity status.
We already call FS_IOC_GETFLAGS in the copy code, and it tells us if we
have verity support.  If not, there is no change.

If we do notice a file with fs-verity enabled, and if CopyFlags
requested that we copy this, we request the descriptor from the source
file and use it to setup fs-verity on the destination, using the same
parameters.

Signatures don't seem to be a particularly well-loved (or used) feature
of fs-verity and we don't bother to query them here.  Support for that
could be added later, if desired.

This change means that, with the correct combination of flags, we might
end up calling `fsync()` on a read-only file descriptor.  This is
permitted by POSIX and supported on Linux.

Nothing uses this yet.

Signed-off-by: Allison Karlitskaya <[email protected]>

Add a new `MkfsFlags` enum and use it to replace the existing `quiet`
and `discard` booleans on `make_filesystem()`.

This introduces a strange approach to converting booleans into flags
values: multiplication.  This is done because:

  bool a, b;
  ..., FLAG_A * a | FLAG_B * b, ...

is a lot less visually noisy than

  ..., (a ? FLAG_A : 0) | (b ? FLAG_B : 0), ...

(Note that the extra brackets are required only in the ternary case
because `|` has lower precedence than `*` but higher than `?:`.)

It also works even if the original value assigned to the `bool` was
something other than 1 or 0.  I was also unable to get either `gcc` or
`clang` to produce different behaviour or complain about it across a
wide range of `-std` and `-W` options.

Signed-off-by: Allison Karlitskaya <[email protected]>

Add an fsverity flag to MkfsFlags and use it to pass the `-O verity`
option when creating an ext4 or f2fs filesystem: they share the same
argument for this.

The only other filesystem that currently supports fs-verity is btrfs and
it doesn't require a flag to be enabled when creating the filesystem.

Nothing uses this yet.

Signed-off-by: Allison Karlitskaya <[email protected]>

We currently convert the source:target pairs of the `CopyFiles=` lines
in `repart.d` files into a pairwise strv.  This works great if the only
thing that can be specified is a source and a target, but we're about to
add a flags field.

Let's start by making this a bit more explicit: we now turn each
`CopyFiles=` line into a `CopyFilesLine` struct.  We keep an array of
those in the `Partition` now, instead of the strv.

So far this is a whole lot of added complexity for nothing, but it's
necessary for the next step.

Signed-off-by: Allison Karlitskaya <[email protected]>

We currently pass the CopyFlags that we use to populate the temporary
directory in the form of a constant at each of the copy_tree_at() call
sites.  De-duplicate that and move it into the `CopyFilesLine` struct,
initializing it from the parser.

Add our first non-constant flag: `fsverity=`.  This can be set to `off`
(the default) or `copy`, in which case we copy the fs-verity state from
the source files.

This arrangement is amenable to the introduction of more flags to
`CopyFiles=` lines, if we want to add them in the future.

Update the `repart.d(5)` manpage.

Closes systemd#35352

Signed-off-by: Allison Karlitskaya <[email protected]>