Unfortunately this needs a new binary to do the mount because there's just
too many special steps to outsource this to systemd-mount:
- EPERM needs to be treated specially
- UserRuntimeDir= setting must be obeyed
- SELinux label must be adjusted
This allows user@.service to be started independently of logind.
So 'systemctl start user@nnn' will start the user manager for user nnn.
Logind will start it too when the user logs in, and will stop it (unless
lingering is enabled) when the user logs out.
Fixes#7339.
This removes the UserTasksMax= setting in logind.conf. Instead, the generic
TasksMax= setting on the slice should be used. Instead of a transient unit we
use a drop-in to tweak the default definition of a .slice. It's better to use
the normal unit mechanisms instead of creating units on the fly. This will also
make it easier to start user@.service independently of logind, or set
additional settings like MemoryMax= for user slices.
The setting in logind is removed, because otherwise we would have two sources
of "truth": the slice on disk and the logind config. Instead of trying to
coordinate those two sources of configuration (and maintainer overrides to
both), let's just convert to the new one fully.
Right now now automatic transition mechanism is provided. logind will emit a
hint when it encounters the setting, but otherwise it will be ignored.
Fixes#2556.
Files which are installed as-is (any .service and other unit files, .conf
files, .policy files, etc), are left as is. My assumption is that SPDX
identifiers are not yet that well known, so it's better to retain the
extended header to avoid any doubt.
I also kept any copyright lines. We can probably remove them, but it'd nice to
obtain explicit acks from all involved authors before doing that.
This reworks the SELinux and SMACK label fixing calls in a number of
ways:
1. The two separate boolean arguments of these functions are converted
into a flags type LabelFixFlags.
2. The operations are now implemented based on O_PATH. This should
resolve TTOCTTOU races between determining the label for the file
system object and applying it, as it it allows to pin the object
while we are operating on it.
3. When changing a label fails we'll query the label previously set, and
if matches what we want to set anyway we'll suppress the error.
Also, all calls to label_fix() are now (void)ified, when we ignore the
return values.
Fixes: #8566
When we are attempting to create directory somewhere in the bowels of /var/lib
and get an error that it already exists, it can be quite hard to diagnose what
is wrong (especially for a user who is not aware that the directory must have
the specified owner, and permissions not looser than what was requested). Let's
print a warning in most cases. A warning is appropriate, because such state is
usually a sign of borked installation and needs to be resolved by the adminstrator.
$ build/test-fs-util
Path "/tmp/test-readlink_and_make_absolute" already exists and is not a directory, refusing.
(or)
Directory "/tmp/test-readlink_and_make_absolute" already exists, but has mode 0775 that is too permissive (0755 was requested), refusing.
(or)
Directory "/tmp/test-readlink_and_make_absolute" already exists, but is owned by 1001:1000 (1000:1000 was requested), refusing.
Assertion 'mkdir_safe(tempdir, 0755, getuid(), getgid(), MKDIR_WARN_MODE) >= 0' failed at ../src/test/test-fs-util.c:320, function test_readlink_and_make_absolute(). Aborting.
No functional change except for the new log lines.
Let's replace usage of fputc_unlocked() and friends by __fsetlocking(f,
FSETLOCKING_BYCALLER). This turns off locking for the entire FILE*,
instead of doing individual per-call decision whether to use normal
calls or _unlocked() calls.
This has various benefits:
1. It's easier to read and easier not to forget
2. It's more comprehensive, as fprintf() and friends are covered too
(as these functions have no _unlocked() counterpart)
3. Philosophically, it's a bit more correct, because it's more a
property of the file handle really whether we ever pass it on to another
thread, not of the operations we then apply to it.
This patch reworks all pieces of codes that so far used fxyz_unlocked()
calls to use __fsetlocking() instead. It also reworks all places that
use open_memstream(), i.e. use stdio FILE* for string manipulations.
Note that this in some way a revert of 4b61c87511.
This adds uid_is_system() and gid_is_system(), similar in style to
uid_is_dynamic(). That a helper like this is useful is illustrated by
the fact that test-condition.c didn't get the check right so far, which
this patch fixes.
In https://bugzilla.redhat.com/show_bug.cgi?id=1486859 error messages appera:
Sep 06 19:09:07 ld92.e.math.uh.edu audit[21482]: AVC avc: denied { read } for pid=21482 comm="systemd-logind" name="dbus-1" dev="tmpfs" ino=5548194 scontext=system_u:system_r:systemd_logind_t:s0 tcontext=unconfined_u:object_r:session_dbusd_tmp_t:s0 tclass=dir permissive=0
Sep 06 19:09:07 ld92.e.math.uh.edu systemd-logind[21482]: Failed to remove runtime directory /run/user/8664: Permission denied
But it's not clear which of the two rm_rf's is the source. Let's make
them different.
As a follow-up for db3f45e2d2 let's do the
same for all other cases where we create a FILE* with local scope and
know that no other threads hence can have access to it.
For most cases this shouldn't change much really, but this should speed
dbus introspection and calender time formatting up a bit.
Let's use chase_symlinks() everywhere, and stop using GNU
canonicalize_file_name() everywhere. For most cases this should not change
behaviour, however increase exposure of our function to get better tested. Most
importantly in a few cases (most notably nspawn) it can take the correct root
directory into account when chasing symlinks.
We don't have plural in the name of any other -util files and this
inconsistency trips me up every time I try to type this file name
from memory. "formats-util" is even hard to pronounce.
The parsing functions for [User]TasksMax were inconsistent. Empty string and
"infinity" were interpreted as no limit for TasksMax but not accepted for
UserTasksMax. Update them so that they're consistent with other knobs.
* Empty string indicates the default value.
* "infinity" indicates no limit.
While at it, replace opencoded (uint64_t) -1 with CGROUP_LIMIT_MAX in TasksMax
handling.
v2: Update empty string to indicate the default value as suggested by Zbigniew
Jędrzejewski-Szmek.
v3: Fixed empty UserTasksMax handling.
This adds the boolean RemoveIPC= setting to service, socket, mount and swap
units (i.e. all unit types that may invoke processes). if turned on, and the
unit's user/group is not root, all IPC objects of the user/group are removed
when the service is shut down. The life-cycle of the IPC objects is hence bound
to the unit life-cycle.
This is particularly relevant for units with dynamic users, as it is essential
that no objects owned by the dynamic users survive the service exiting. In
fact, this patch adds code to imply RemoveIPC= if DynamicUser= is set.
In order to communicate the UID/GID of an executed process back to PID 1 this
adds a new "user lookup" socket pair, that is inherited into the forked
processes, and closed before the exec(). This is needed since we cannot do NSS
from PID 1 due to deadlock risks, However need to know the used UID/GID in
order to clean up IPC owned by it if the unit shuts down.
config_parse_user_tasks_max() was incorrectly accepting percentage value
between 1 and 99. Update it to accept 0% and 100%. This brings it in line
with TasksMax handling in systemd.
Let's change from a fixed value of 12288 tasks per user to a relative value of
33%, which with the kernel's default of 32768 translates to 10813. This is a
slight decrease of the limit, for no other reason than "33%" sounding like a nice
round number that is close enough to 12288 (which would translate to 37.5%).
(Well, it also has the nice effect of still leaving a bit of room in the PID
space if there are 3 cooperating evil users that try to consume all PIDs...
Also, I like my bikesheds blue).
Since the new value is taken relative, and machined's TasksMax= setting
defaults to 16384, 33% inside of containers is usually equivalent to 5406,
which should still be ample space.
To summarize:
| on the host | in the container
old default | 12288 | 12288
new default | 10813 | 5406
The deserialize_timestamp_value() is renamed timestamp_deserialize() to be more
consistent with dual_timestamp_deserialize()
And add the NULL check back on realtime and monotonic
manager_{start,stop}_{slice,scope,unit} functions had an optional job
output parameter. But all callers specified job, so make the parameter
mandatory, add asserts. Also extract common job variable handling to
a helper function to avoid duplication.
Avoids gcc warning about job being unitialized.
GLIB has recently started to officially support the gcc cleanup
attribute in its public API, hence let's do the same for our APIs.
With this patch we'll define an xyz_unrefp() call for each public
xyz_unref() call, to make it easy to use inside a
__attribute__((cleanup())) expression. Then, all code is ported over to
make use of this.
The new calls are also documented in the man pages, with examples how to
use them (well, I only added docs where the _unref() call itself already
had docs, and the examples, only cover sd_bus_unrefp() and
sd_event_unrefp()).
This also renames sd_lldp_free() to sd_lldp_unref(), since that's how we
tend to call our destructors these days.
Note that this defines no public macro that wraps gcc's attribute and
makes it easier to use. While I think it's our duty in the library to
make our stuff easy to use, I figure it's not our duty to make gcc's own
features easy to use on its own. Most likely, client code which wants to
make use of this should define its own:
#define _cleanup_(function) __attribute__((cleanup(function)))
Or similar, to make the gcc feature easier to use.
Making this logic public has the benefit that we can remove three header
files whose only purpose was to define these functions internally.
See #2008.
If the last reference to a user is released, we queue stop-jobs for the
user-service and slice. Only once those are finished, we drop the
user-object. However, if a new session is opened before the user object is
fully dropped, we currently incorrectly re-use the object. This has the
effect, that we get stale sessions without a valid "systemd --user"
instance.
Fix this by properly allowing user_start() to be called, even if
user->stopping is true.
Just like user->slice, there is no reason to store the unit name in /run,
nor should we allocate it dynamically on job instantiation/removal. Just
keep it statically around at all times and rely on user->started ||
user->stopping to figure out whether the unit exists or not.
Few changes to user_new() and user_free():
- Use _cleanup_(user_freep) in constructor
- return 'int' from user_new()
- make user_free() deal with partially initialized objects
- keep reverse-order in user_free() compared to user_new()
- make user_free() return NULL
- make user_free() accept NULL as no-op
Currently, we allocate user->slice when starting a slice, but we never
release it. This is incompatible if we want to re-use a user object once
it was stopped. Hence, make sure user->slice is allocated statically on
the user object and use "u->started || u->stopping" as an indication
whether the slice is actually available on pid1 or not.
Lets not pretend we support changing XDG_RUNTIME_DIR via logind state
files. There is no reason to ever write the string into /run, as we
allocate it statically based on the UID, anyway. Lets stop that and just
allocate the runtime_path in "struct User" at all times.
We keep writing it into the /run state to make sure pam_systemd of
previous installs can still read it. However, pam_systemd is now fixed to
allocate it statically as well, so we can safely remove that some time in
the future.
Last but not least: If software depends on systemd, they're more than free
to assume /run/user/$uid is their runtime dir. Lets not require sane
applications to query the environment to get their runtime dir. As long as
applications know their login-UID, they should be safe to deduce the
runtime dir.
This new setting configures the TasksMax= field for the slice objects we
create for each user.
This alters logind to create the slice unit as transient unit explicitly
instead of relying on implicit generation of slice units by simply
starting them. This also enables us to set a friendly description for
slice units that way.
