This adds a small service "systemd-portabled" and a matching client
"portablectl", which implement the "portable service" concept.
The daemon implements the actual operations, is PolicyKit-enabled and is
activated on demand with exit-on-idle.
Both the daemon and the client are an optional build artifact, enabled
by default rhough.
Let's shorten the code a bit, and unify how we forward executoin from
the machine functions that take a machine name as first argument to the
matching functions on the machine object.
This adds directories in /etc and /run to the search paths for OS
images. While it doesn't make much sense to actually place huge disk
images there, it's good enough for symlinks to those.
The main reason for supporting this is that this allows us to neatly
symlink portable image files located outside of the search path into the
search path when attaching them, so that attaching them also means they
are discoverable properly for all commands.
This new flag indicates whether the image object was found in the search
paths using the usual algorithm, or was instantiated by path.
This is useful for code that wants to know whether an image may be
referenced by its shortened name or must be specified by its full name.
Let's rework error handling a bit in image_find() and friends: when we
can't find an image, return -ENOENT rather than 0. That's better as
before we violated the usual rule in our codebase that return parameters
are initialized when the return value is >= 0 and otherwise not touched.
This also makes enumeration and validation a bit more strict: we'll only
accept ".raw" as suffix for regular files, and filter out this suffix
handling on directories/subvolumes, where it makes no sense.
We need to chop off the .raw suffix from the files we find before we can
test it against the hashmap. Hence do that.
And while we are at it, we can pass the pretty name into image_make(),
since we already have it properly formatted.
This distuingishes two different classes of images, one for the purpose
of npsawn-like containers, i.e. "machines", and one for portable
services.
This distinction is mostly about search paths. We look for machine
images in /var/lib/machines and for portable images in
/var/lib/portables.
This adds fozr new flags:
- If CONF_FILES_DIRECTORY is specified conf_file_list() and friends
will look for directories only.
- Similar CONF_FILES_REGULAR means we'll look only for regular files.
- If CONF_FILES_BASENAME is specified the resulting list will contain
only the basenames of all discovered files or directories, not the
full paths.
- If CONF_FILES_FILTER_MASKED is specified the resulting list will have
masked entries removed (i.e. those symlinked to /dev/null and
suchlike)
These four flags are useful for discovering portable service profile
information.
While we are at it, also improve a couple of other things:
- More debug logging
- use path_hash_ops instead of string_hash_ops when putting together the
path lists
When we look into a portable service image it might contain the unit
files in split-usr directories rather than merged-usr directories as on
the host. Hence, let#s add a flag that checking all dirs can be forced.
Most our other parsing functions do this, let's do this here too,
internally we accept that anyway. Also, the closely related
load_env_file() and load_env_file_pairs() also do this, so let's be
systematic.
We already have a flag for creating a new mount namespace for the child.
Let's add an extension to that: a new FORK_MOUNTNFS_SLAVE flag. When
used in combination will mark all mounts in the child namespace as
MS_SLAVE so that the child can freely mount or unmount stuff but it
won't leak into the parent.
We already do this kind of validation in nspawn when we operate on a
plain directory, let's also do this on raw images under the same
condition: that we are about too boot the image. Also, do this when we
are about to read OS metadata from it.
This call creates an fd from another fd containing the same data.
Specifically, repeated read() on the returned fd should return the same
data as the original fd. This call is useful when we want to copy data
out of disk images and suchlike, and want to be pass fds with the data
around without having to keep the disk image continously mounted.
The implementation tries to be somewhat smart and tries to prefer
memfds/pipes over files in /tmp or /var/tmp based on the size of the
data, but has appropropriate fallbacks in place.
Externally it's an uint64_t anyway, and internally we most just
initialize it to physical_memory() which returns uint64_t, hence there's
exactly zero value in using it as size_t internally. Hence, let's fix
that, and use uint64_t everywhere.
Otherwise querying the preset status of a unit to the user instance gives
incorrect results since in this case the scope used by the manager is
UNIT_FILE_USER.
sysexits.h has:
#define EX_CANTCREAT 73 /* can't create (user) output file */
EX_DATAERR is a copy-paste error from the previous sentence, which is
correct.
If we log to the pty that is configured as stdin/stdout/stderr of the
container too early we risk filling it up in full before we start
processing the pty from the parent process, resulting in deadlocks.
Let's hence keep a copy of the original tty we were started on before
setting up stdin/stdout/stderr, so that we can log to it, and keep using
it as long as we can.
Since the kernel's pty internal buffer is pretty small this actually
triggered deadlocks when we debug logged at lot from nspawn's child
processes, see: https://github.com/systemd/systemd/pull/9024#issuecomment-390403674
With this change we won't use the pty at all, only the actual payload we
start will, and hence we won't deadlock on it, ever.
Let's make use of log_set_open_when_needed() in nspawn too, i.e. at the
point where we close logging because we are about to rearrange fds,
let's automatically reopen the logging fds when we need them, the same
way as we do that in the service manager. This makes things simpler and
more robust.
Lennart Poettering