Before:
$ systemctl preset getty@.service
Failed to preset unit, file /etc/systemd/system/getty.target.wants/getty@tty1.service
already exists and is a symlink to ../../../../usr/lib/systemd/system/getty@.service.
After:
$ systemctl preset getty@.service
Created symlink /etc/systemd/system/getty.target.wants/getty@tty1.service,
pointing to /usr/lib/systemd/system/getty@.service.
We don't really care where the symlink points to. For example, it might point
to /usr/lib or /etc, and systemd will always load the unit from /etc in
preference to /usr/lib. In fact, if we make a symlink like
/etc/systemd/system/multi-user.target.wants/b.service -> ../a.service, pid1
will still start b.service. The name of the symlink is the only thing that
matters, as far as systemd is concerned. For humans it's confusing when the
symlinks points to anything else than the actual unit file. At the very least,
the symlink is supposed to point to a file with the same name in some other
directory. Since we don't care where the symlink points, we can always replace
an existing symlink.
Another option I considered would be to simply leave an existing symlink in
place. That would work too, but replacing the symlink with the expected value
seems more intuitive.
Of course those considerations only apply to .wants and .requires. Symlinks
created with "link" and "alias" are a separate matter.
Fixes#3056.
Fixes#1892.
Previously:
Failed to enable unit: Invalid argument
Now:
Failed to enable unit, file /etc/systemd/system/ssh.service already exists.
It would be nice to include the unit name in the message too. I looked into
this, but it would require major surgery on the whole installation logic,
because we first create a list of things to change, and then try to apply them
in a loop. To transfer the knowledge which unit was the source of each change,
the data structures would have to be extended to carry the unit name over into
the second loop. So I'm skipping this for now.
Fixes#2191:
$ systemctl --root=/ enable sddm
Created symlink /etc/systemd/system/display-manager.service, pointing to /usr/lib/systemd/system/sddm.service.
$ sudo build/systemctl --root=/ enable gdm
Failed to enable unit, file /etc/systemd/system/display-manager.service already exists and is a symlink to /usr/lib/systemd/system/sddm.service.
$ sudo build/systemctl --root= enable sddm
$ sudo build/systemctl --root= enable gdm
Failed to enable unit: File /etc/systemd/system/display-manager.service already exists and is a symlink to /usr/lib/systemd/system/sddm.service.
(I tried a few different approaches to pass the error information back to the
caller. Adding a new parameter to hold the error results in a gigantic patch
and a lot of hassle to pass the args arounds. Adding this information to the
changes array is straightforward and can be more easily extended in the
future.)
In case local installation is performed, the full set of errors can be reported
and we do that. When running over dbus, only the first error is reported.
With any masked unit that would that would be enabled by presets, we'd get:
test@rawhide $ sudo systemctl preset-all
Failed to execute operation: Unit file is masked.
test@rawhide $ sudo systemctl --root=/ preset-all
Operation failed: Cannot send after transport endpoint shutdown
Simply ignore those units:
test@rawhide $ sudo systemctl preset-all
Unit xxx.service is masked, ignoring.
If the error code ever leaks (we print the strerror error instead of providing
our own), the message for ESHUTDOWN is "Cannot send after transport endpoint
shutdown", which can be misleading. In particular it suggest that some
mishandling of the dbus connection occured. Let's change that to ERFKILL which
has the advantage that a) it sounds implausible as actual error, b) has the
connotation of disabling something manually.
This allows dropping all user configuration and reverting back to the vendor
default of a unit file. It basically undoes what "systemctl edit", "systemctl
set-property" and "systemctl mask" do.
The SysV compat code checks whether there's a native unit file before looking
for a SysV init script. Since the newest rework generated units will show up in
the unit path, and hence the checks ended up assuming that there always was a
native unit file for each init script: the generated one.
With this change the generated unit file directory is suppressed from the
search path when this check is done, to avoid the confusion.
Move the search path check from the SysV service compat support into install.c
so that we can reuse the usual algorithm instead of rolling a private loop for
this.
Now, that the search path logic knows the unit path for transient units we also
can introduce an explicit unit file state "transient" that clarifies to the
user what kind of unit file he is encountering.
Previously, we'd execute some operations with the root prefix applied, while
others without (which was a bug). Clean this up: all paths are now prefixed
properly with the root path, and we strip it off when necessary.
(Of course, an alternative option would be to strictly pass around paths
without the prefix prepended and only prepend it right before hitting the disk,
however, I am came to the conclusion this would result in more code.)
We use the root directory parameter while putting together the LookupPaths
structure, hence let's also store it in the structure as-is. That way we can
drop a parameter from half of the functions in install.c
Also, let's move the validation of the root paths into lookup_paths_init() so
that we can drop even more code from install.c
Previously, we had two enums ManagerRunningAs and UnitFileScope, that were
mostly identical and converted from one to the other all the time. The latter
had one more value UNIT_FILE_GLOBAL however.
Let's simplify things, and remove ManagerRunningAs and replace it by
UnitFileScope everywhere, thus making the translation unnecessary. Introduce
two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking
if we are running in one or the user context.
Now that we store the generator directories in LookupPaths we can use this to
intrdouce a new unit file state called "generated", for units in these
directories.
Fixes: #2348
A long time ago – when generators where first introduced – the directories for
them were randomly created via mkdtemp(). This was changed later so that they
use fixed name directories now. Let's make use of this, and add the genrator
dirs to the LookupPaths structure and into the unit file search path maintained
in it. This has the benefit that the generator dirs are now normal part of the
search path for all tools, and thus are shown in "systemctl list-unit-files"
too.
All callers of create_symlink(), such as install_info_symlink_wants(), expect
that to return > 0 if it actually did something, and then return that number.
unit_file_enable() uses that to determine if any action was done
(carries_install_info != 0) and if not, show a "The unit files have no
[Install] section" warning.
Return 1 instead of 0 in the two code paths of create_symlink() when the link
was created or replaced with a new value.
This fixes getting a bogus "No [Install] section" warning when enabling a unit
with full path, like "systemctl enable /some/path/myunit.service".
Previously, the %u, %U, %s and %h specifiers would resolve to the user
name, numeric user ID, shell and home directory of the user configured
in the User= setting of a unit file, or the user of the manager instance
if no User= setting was configured. That at least was the theory. In
real-life this was not ever actually useful:
- For the systemd --user instance it made no sense to ever set User=,
since the instance runs in user context after all, and hence the
privileges to change user IDs don't even exist. The four specifiers
were actually not useful at all in this case.
- For the systemd --system instance we did not allow any resolving that
would require NSS. Hence, %s and %h were not supported, unless
User=root was set, in which case they would be hardcoded to /bin/sh
and /root, to avoid NSS. Then, %u would actually resolve to whatever
was set with User=, but %U would only resolve to the numeric UID of
that setting if the User= was specified in numeric form, or happened
to be root (in which case 0 was hardcoded as mapping). Two of the
specifiers are entirely useless in this case, one is realistically
also useless, and one is pretty pointless.
- Resolving of these settings would only happen if User= was actually
set *before* the specifiers where resolved. This behaviour was
undocumented and is really ugly, as specifiers should actually be
considered something that applies to the whole file equally,
independently of order...
With this change, %u, %U, %s and %h are drastically simplified: they now
always refer to the user that is running the service instance, and the
user configured in the unit file is irrelevant. For the system instance
of systemd this means they always resolve to "root", "0", "/bin/sh" and
"/root", thus avoiding NSS. For the user instance, to the data for the
specific user.
The new behaviour is identical to the old behaviour in all --user cases
and for all units that have no User= set (or set to "0" or "root").
Some distributions use alias unit files via symlinks in /usr to cover
for legacy service names. With this change we'll allow "systemctl
enable" on such aliases.
Previously, our rule was that symlinks are user configuration that
"systemctl enable" + "systemctl disable" creates and removes, while unit
files is where the instructions to do so are store. As a result of the
rule we'd never read install information through symlinks, since that
would mix enablement state with installation instructions.
Now, the new rule is that only symlinks inside of /etc are
configuration. Unit files, and symlinks in /usr are now valid for
installation instructions.
This patch is quite a rework of the whole install logic, and makes the
following addional changes:
- Adds a complete test "test-instal-root" that tests the install logic
pretty comprehensively.
- Never uses canonicalize_file_name(), because that's incompatible with
operation relative to a specific root directory.
- unit_file_get_state() is reworked to return a proper error, and
returns the state in a call-by-ref parameter. This cleans up confusion
between the enum type and errno-like errors.
- The new logic puts a limit on how long to follow unit file symlinks:
it will do so only for 64 steps at max.
- The InstallContext object's fields are renamed to will_process and
has_processed (will_install and has_installed) since they are also
used for deinstallation and all kinds of other operations.
- The root directory is always verified before use.
- install.c is reordered to place the exported functions together.
- Stricter rules are followed when traversing symlinks: the unit suffix
must say identical, and it's not allowed to link between regular units
and templated units.
- Various modernizations
- The "invalid" unit file state has been renamed to "bad", in order to
avoid confusion between UNIT_FILE_INVALID and
_UNIT_FILE_STATE_INVALID. Given that the state should normally not be
seen and is not documented this should not be a problematic change.
The new name is now documented however.
Fixes#1375, #1718, #1706
Instead, let the caller do that. Fix this by moving masked unit messages
into the caller, by returning a clear error code (ESHUTDOWN) by which
this may be detected.
There are more than enough calls doing string manipulations to deserve
its own files, hence do something about it.
This patch also sorts the #include blocks of all files that needed to be
updated, according to the sorting suggestions from CODING_STYLE. Since
pretty much every file needs our string manipulation functions this
effectively means that most files have sorted #include blocks now.
Also touches a few unrelated include files.
- Rely everywhere that we use abs() on the error code passed in anyway,
thus don't need to explicitly negate what we pass in
- Never attach synthetic error number information to log messages. Only
log about errors we *receive* with the error number we got there,
don't log any synthetic error, that don#t even propagate, but just eat
up.
- Be more careful with attaching exactly the error we get, instead of
errno or unrelated errors randomly.
- Fix one occasion where the error number and line number got swapped.
- Make sure we never tape over OOM issues, or inability to resolve
specifiers
Commit aedd401 introduced new unit file state, UNIT_FILE_INDIRECT. Unit file is
said to have indirect state if it contains [Install] section which has only
Also= directive. Thus, if enable of such unit file is requested then some other
unit file gets enabled.
Whether or not unit file is in indirect state can be determined by calling
unit_file_can_install. Function unit_file_get_list populates list of unit files
present in given lookup location. So far it did call unit_file_can_install in a
way that would prevent finding out about unit files in indirect state. Such unit
file would be incorrectly marked as static.
Fixes following assertion in test-install,
Assertion 'p->state == s' failed at src/test/test-install.c:59, function main(). Aborting.
[1] 26868 abort (core dumped) ./test-install
Maybe there is some left-over value stored in r from previous function
call. Let's make sure we always return consistent error code when we reach end of
the function body.
Fixes following crash of test-install,
Assertion 'r == 0' failed at src/test/test-install.c:52, function main(). Aborting.
[1] 11703 abort (core dumped) ./test-install
All other types exported from install.h should be namespaces like this,
hence namespace InstallInfo the same way.
Also, remove external forward definition of UnitFileScope type.
A variety of changes:
- Make sure all our calls distuingish OOM from other errors if OOM is
not the only error possible.
- Be much stricter when parsing escaped paths, do not accept trailing or
leading escaped slashes.
- Change unit validation to take a bit mask for allowing plain names,
instance names or template names or an combination thereof.
- Refuse manipulating invalid unit name
like:
src/shared/install.c: In function ‘unit_file_lookup_state’:
src/shared/install.c:1861:16: warning: ‘r’ may be used uninitialized in
this function [-Wmaybe-uninitialized]
return r < 0 ? r : state;
^
src/shared/install.c:1796:13: note: ‘r’ was declared here
int r;
^
With debugging on, sysv-generator would print the full set of
lookup paths for *every* sysv script.
While at it, pass LookupPaths as a pointer in sysv-generator,
and constify it everywhere.
This patch removes includes that are not used. The removals were found with
include-what-you-use which checks if any of the symbols from a header is
in use.
After all it is now much more like strjoin() than strappend(). At the
same time, add support for NULL sentinels, even if they are normally not
necessary.
Using:
find . -name '*.[ch]' | while read f; do perl -i.mmm -e \
'local $/;
local $_=<>;
s/(if\s*\([^\n]+\))\s*{\n(\s*)(log_[a-z_]*_errno\(\s*([->a-zA-Z_]+)\s*,[^;]+);\s*return\s+\g4;\s+}/\1\n\2return \3;/msg;
print;'
$f
done
And a couple of manual whitespace fixups.
As a followup to 086891e5c1 "log: add an "error" parameter to all
low-level logging calls and intrdouce log_error_errno() as log calls
that take error numbers", use sed to convert the simple cases to use
the new macros:
find . -name '*.[ch]' | xargs sed -r -i -e \
's/log_(debug|info|notice|warning|error|emergency)\("(.*)%s"(.*), strerror\(-([a-zA-Z_]+)\)\);/log_\1_errno(-\4, "\2%m"\3);/'
Multi-line log_*() invocations are not covered.
And we also should add log_unit_*_errno().
f7101b7368 copied some logic to prevent enabling masked units, but
also added a check which causes attempts to enable templated units to
fail. Since we know the logic beyond this check will properly handle
units which truly do not exist, we can rely on the unit file state
comparison to suffice for expressing the intent of f7101b7368.
ref: https://bugs.archlinux.org/task/42616
That hashmap_move_one() currently cannot fail with -ENOMEM is an
implementation detail, which is not possible to guarantee in general.
Hashmap implementations based on anything else than chaining of
individual entries may have to allocate.
hashmap_move_one will not fail with -ENOMEM if a proper reservation has
been made beforehand. Use reservations in install.c.
In cgtop.c simply propagate the error instead of asserting.
This patch adds a transient user unit directory under
`$XDG_RUNTIME_DIR/systemd/user/` and stores transient user-instance
units (such as those created by `systemd-run --user`) under there
instead of putting them in $XDG_CONFIG_HOME/systemd/user/.
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=67331
It is redundant to store 'hash' and 'compare' function pointers in
struct Hashmap separately. The functions always comprise a pair.
Store a single pointer to struct hash_ops instead.
systemd keeps hundreds of hashmaps, so this saves a little bit of
memory.
This patch modifies unit_file_get_list which will now return
hashmap of structures where f->path is *without* root_dir prefix.
This change should be ok, because current code either does not use
root_dir at all or calls basename() on the f->path.
Previously systemctl died with message
-bash-4.2# systemctl --root /rawhi list-unit-files
(src/systemctl/systemctl.c:868) Out of memory.
in the case that no unit files were found in the --root
or the directory did not exist.
So lets return ENOENT in the case that --root does not exist
and empty list in the case that there are no unit files.
String which ended in an unfinished quote were accepted, potentially
with bad memory accesses.
Reject anything which ends in a unfished quote, or contains
non-whitespace characters right after the closing quote.
_FOREACH_WORD now returns the invalid character in *state. But this return
value is not checked anywhere yet.
Also, make 'word' and 'state' variables const pointers, and rename 'w'
to 'word' in various places. Things are easier to read if the same name
is used consistently.
mbiebl_> am I correct that something like this doesn't work
mbiebl_> ExecStart=/usr/bin/encfs --extpass='/bin/systemd-ask-passwd "Unlock EncFS"'
mbiebl_> systemd seems to strip of the quotes
mbiebl_> systemctl status shows
mbiebl_> ExecStart=/usr/bin/encfs --extpass='/bin/systemd-ask-password Unlock EncFS $RootDir $MountPoint
mbiebl_> which is pretty weird
There is a small number of the places in sources where we don't check
asprintf() return code and assume that after error the function
returns NULL pointer via the first argument. That's wrong, after
error the content of pointer is undefined.
Special care is needed so that we get an error message if the
file failed to parse, but not when it is missing. To avoid duplicating
the same error check in every caller, add an additional 'warn' boolean
to tell config_parse whether a message should be issued.
This makes things both shorter and more robust wrt. to error reporting.
This restores the original root handling logic that was present prior to
112cfb18 when path expansion moved to path_strv_canonicalize_absolute.
That behavior partially went away in 12ed81d9.
Alternatively all users of conf_files_list* could be updated to
concatenate the paths themselves as unit_file_query_preset did but since
no user needs the un-concatenated form that is pointless duplication.
The DefaultInstance= name is used when enabling template units when only
specifying the template name, but no instance.
Add DefaultInstance=tty1 to getty@.service, so that when the template
itself is enabled an instance for tty1 is created.
This is useful so that we "systemctl preset-all" can work properly,
because we can operate on getty@.service after finding it, and the right
instance is created.
The new "systemctl preset-all" command may now be used to put all
installed units back into the enable/disable state the vendor/admin
encoded in preset files.
Also, introduce "systemctl --preset-mode=enable-only" and "systemctl
--preset-mode=disable-only" to only apply the enable or only the disable
operations of a "systemctl preset" or "systemctl preset-all" operation.
"systemctl preset-all" implements this RFE:
https://bugzilla.redhat.com/show_bug.cgi?id=630174
int unit_file_mask(...) in ./src/shared/install.c calls
get_config_path(...) which can in 4 error cases return without setting
"ret", and thus "prefix" can be uninitialized when unit_file_mask(...)
finishes (which it does directly after the error is returned from
get_config_path(...)).
Running systemctl enable/disable/set-default/... with the --root
option under strace reveals that it accessed various files and
directories in the main fs, and not underneath the specified root.
This can lead to correct results only when the layout and
configuration in the container are identical, which often is not the
case. Fix this by adding the specified root to all file access
operations.
This patch does not handle some corner cases: symlinks which point
outside of the specified root might be interpreted differently than
they would be by the kernel if the specified root was the real root.
But systemctl does not create such symlinks by itself, and I think
this is enough of a corner case not to be worth the additional
complexity of reimplementing link chasing in systemd.
Also, simplify the code in a few places and remove an hypothetical
memory leak on error.
create_symlink() do not check the return value of unlink(), this may
confuse the user.
Before the unlink() call we check the 'force' argument. If it is not set
we fail with -EEXIST, otherwise we unlink() the file, therefore the next
symlink() should not fail with -EEXIST (do not count races...).
However since callers may not have appropriate privileges to unlink()
the file we lose the -EPERM or any other errno code of unlink(), and
return the -EEXIST of the next symlink(). Fix this by checking unlink()
results.
Before:
$ systemctl --force --root=~/container-03 set-default multi-user.target
Failed to set default target: File exists
After:
$ systemctl --force --root=~/container-03 set-default multi-user.target
Failed to set default target: Permission denied
safe_close() automatically becomes a NOP when a negative fd is passed,
and returns -1 unconditionally. This makes it easy to write lines like
this:
fd = safe_close(fd);
Which will close an fd if it is open, and reset the fd variable
correctly.
By making use of this new scheme we can drop a > 200 lines of code that
was required to test for non-negative fds or to reset the closed fd
variable afterwards.
Even if the lower-leveld dbus1 protocol calls it "serial", let's expose
the word "cookie" for this instead, as this is what kdbus uses and since
it doesn't imply monotonicity the same way "serial" does.
The only problem is that libgen.h #defines basename to point to it's
own broken implementation instead of the GNU one. This can be fixed
by #undefining basename.
Pass on the line on which a section was decleared to the parsers, so they
can distinguish between multiple sections (if they chose to). Currently
no parsers take advantage of this, but a follow-up patch will do that
to distinguish
[Address]
Address=192.168.0.1/24
Label=one
[Address]
Address=192.168.0.2/24
Label=two
from
[Address]
Address=192.168.0.1/24
Label=one
Address=192.168.0.2/24
Label=two
This patch converts PID 1 to libsystemd-bus and thus drops the
dependency on libdbus. The only remaining code using libdbus is a test
case that validates our bus marshalling against libdbus' marshalling,
and this dependency can be turned off.
This patch also adds a couple of things to libsystem-bus, that are
necessary to make the port work:
- Synthesizing of "Disconnected" messages when bus connections are
severed.
- Support for attaching multiple vtables for the same interface on the
same path.
This patch also fixes the SetDefaultTarget() and GetDefaultTarget() bus
calls which used an inappropriate signature.
As a side effect we will now generate PropertiesChanged messages which
carry property contents, rather than just invalidation information.
Previously the specifier calls could only indicate OOM by returning
NULL. With this change they will return negative errno-style error codes
like everything else.
Assertion 'p' failed at src/shared/path-util.c:51, function path_get_file_name(). Aborting.
The unit file could not be found, and i->path would not be set.
In 02b9e969 a code path was added which attempts to remove symlinks
to a nonexistent (removed) unit file. This worked OK in case of
non-instance services, but broke in the case of instance services.
Behaviour wrt. to instance units is changed in the way that 02b9e969
changed it for non-instance units: it is now possible to remove
instance symlinks to a template unit that has been removed.
This patch isn't a full fix, because the behaviour wrt. to enabling
and disabling instance units is still broken: e.g it is possible to
start autovt@tty5.service, but it is not possible to enable it,
because autovt@.service is a symlink, and on the other hand, disabling
getty@tty5.service removes all symlinks to getty@.service, which is
wrong too. But segfaults make bad pr, so let's add at least this
partial fix for now.
Before, "systemctl reenable getty@tty1.service" would fail with:
Failed to issue method call: File exists
To fix this, reimplement "reenable" explicitly as a disable followed by
an enable.
This is shorter and is how the man page documents its behavior.
Before, one the unit file was deleted, install_context_for_removal()
would refuse to look for symlinks. But we can remove dangling symlinks
anyway.
In principle, package installation/deinstallation scripts should do
that before the unit is uninstalled, but they don't always do. Also,
a user might have added additional symlinks manually.
https://bugs.freedesktop.org/show_bug.cgi?id=62395
Patch resolves the problem that 'systemctl is-enabled' does
not work for templated units.
Without this patch, systemctl is-enabled something@abc.service
returned "No such file or directory", because it first checked
if /usr/lib/systemd/system/something@abc.service, etc. exists.
If systemctl is-enabled is called for templated units, this
check should be omitted and it should search for symlinks in
the .wants dirs right away.
This patch fixes the broken behaviour and resolves
https://bugs.freedesktop.org/show_bug.cgi?id=55318.
[zj: fixed the patch to still check for broken symlinks and
masked instances. Also removed untrue assumptions from
the patch description.]
The information about the unit for which files are being parsed
is passed all the way down. This way messages land in the journal
with proper UNIT=... or USER_UNIT=... attribution.
'systemctl status' and 'journalctl -u' not displaying those messages
has been a source of confusion for users, since the journal entry for
a misspelt setting was often logged quite a bit earlier than the
failure to start a unit.
Based-on-a-patch-by: Oleksii Shevchuk <alxchk@gmail.com>
Make sure we compare errno against positive error codes.
The ones in hwclock.c and install.c can have an impact, the
rest are unlikely to be hit or in code that isn't widely
used.
Also check that errno > 0, to help gcc know that we are
returning a negative error code.
Before, we would initialize many fields twice: first
by filling the structure with zeros, and then a second
time with the real values. We can let the compiler do
the job for us, avoiding one copy.
A downside of this patch is that text gets slightly
bigger. This is because all zero() calls are effectively
inlined:
$ size build/.libs/systemd
text data bss dec hex filename
before 897737 107300 2560 1007597 f5fed build/.libs/systemd
after 897873 107300 2560 1007733 f6075 build/.libs/systemd
… actually less than 1‰.
A few asserts that the parameter is not null had to be removed. I
don't think this changes much, because first, it is quite unlikely
for the assert to fail, and second, an immediate SEGV is almost as
good as an assert.
This allows one templated unit to refer to another templated unit
at installation time.
Examples:
> grep WantedBy ~/.config/systemd/user/mpop@.timer
WantedBy=services@%i.target
> srv disable mpop@iit.timer
rm '/home/alxchk/.config/systemd/user/services@iit.target.wants/mpop@iit.timer'
> srv enable mpop@iit.timer
ln -s '/home/alxchk/.config/systemd/user/mpop@.timer' '/home/alxchk/.config/systemd/user/services@iit.target.wants/mpop@iit.timer'
Based-on-patch-by: Oleksii Shevchuk <alxchk@gmail.com>
Note: I did s/MANAGER/SYSTEMD/ everywhere, even though it makes the
patch quite verbose. Nevertheless, keeping MANAGER prefix in some
places, and SYSTEMD prefix in others would just lead to confusion down
the road. Better to rip off the band-aid now.
When looking for symlinks, it doesn't make sense to error-out if
the directory is missing. The user might delete an empty directory.
This check caused test-unit-file to fail when run before installation.
Previously generated units were always placed at the end of the search
path. With this change there will be three unit dirs instead of one, to
place generated entries at the beginning, in the middle and at the end
of the search path:
beginning: for units that need to override all configuration, regardless
of user or vendor. Example use: system-update-generator uses this to
temporarily redirect default.target.
middle: for units that need to override vendor configuration, but not
vendor configuration. Example use: /etc/fstab should override vendor
supplied configuration (think /tmp), but should not override native user
configuration.
end: does not override anything but is available as well. Possible usage
might be to convert D-Bus bus service files to native units but allowing
vendor supplied native units to win.