To be able to run systemd in a Type=notify transient unit, the notify
socket can't be bind mounted to /run/systemd/notify as systemd in the
transient unit wants to use that as its own notify socket which
conflicts with systemd on the host.
Instead, for sandboxed units, let's bind mount the notify socket to
/run/host/notify as documented in the container interface. Since we
don't guarantee a stable location for the notify socket and insist users
use $NOTIFY_SOCKET to get its path, this is safe to do.
To be able to run systemd in a Type=notify transient unit, the notify
socket can't be bind mounted to /run/systemd/notify as systemd in the
transient unit wants to use that as its own notify socket which conflicts
with systemd on the host.
Instead, for sandboxed units, let's bind mount the notify socket to
/run/host/notify as documented in the container interface. Since we don't
guarantee a stable location for the notify socket and insist users use
$NOTIFY_SOCKET to get its path, this is safe to do.
Recently, PrivateUsers=identity was added to support mapping the first
65536 UIDs/GIDs from parent to the child namespace and mapping the other
UID/GIDs to the nobody user.
However, there are use cases where users have UIDs/GIDs > 65536 and need
to do a similar identity mapping. Moreover, in some of those cases,
users want a full identity mapping from 0 -> UID_MAX.
To support this, we add PrivateUsers=full that does identity mapping for
all available UID/GIDs.
Note to differentiate ourselves from the init user namespace, we need to
set up the uid_map/gid_map like:
```
0 0 1
1 1 UINT32_MAX - 1
```
as the init user namedspace uses `0 0 UINT32_MAX` and some applications
- like systemd itself - determine if its a non-init user namespace based
on uid_map/gid_map files.
Note systemd will remove this heuristic in running_in_userns() in
version 258 (https://github.com/systemd/systemd/pull/35382) and uses
namespace inode. But some users may be running a container image with
older systemd < 258 so we keep this hack until version 259 for version
N-1 compatibility.
In addition to mapping the whole UID/GID space, we also set
/proc/pid/setgroups to "allow". While we usually set "deny" to avoid
security issues with dropping supplementary groups
(https://lwn.net/Articles/626665/), this ends up breaking dbus-broker
when running /sbin/init in full OS containers.
Fixes: #35168Fixes: #35425
When trying to run dbus-broker in a systemd unit with PrivateUsers=full,
we see dbus-broker fails with EPERM at `util_audit_drop_permissions`.
The root cause is dbus-broker calls the setgroups() system call and this
is disallowed via systemd's implementation of PrivateUsers= by setting
/proc/pid/setgroups = deny. This is done to remediate potential privilege
escalation vulnerabilities in user namespaces where an attacker can remove
supplementary groups and gain access to resources where those groups are
restricted.
However, for OS-like containers, setgroups() is a pretty common API and
disabling it is not feasible. So we allow setgroups() by setting
/proc/pid/setgroups to allow in PrivateUsers=full. Note security conscious
users can still use SystemCallFilter= to disable setgroups() if they want
to specifically prevent this system call.
Fixes: #35425
This introduces a new unit condition check: that matches if a specific
kmod module is allowed. This should be generally useful, but there's one
usecase in particular: we can optimize modprobe@.service with this and
avoid forking out a bunch of modprobe requests during boot for the same
kmods.
Checking if a kernel module is loaded is more complicated than just
checking if /sys/module/$MODULE/ exists, since kernel modules typically
take a while to initialize and we must check that this is complete (by
checking if the sysfs attr "initstate" is "live").
This PR allows an option for systemd exec units to enable UTS namespaces
but not restrict changing hostname via seccomp. Thus, units can change
hostname without affecting the host. This is useful for OS-like
containers running as units where they should have freedom to change
their container hostname if they want, but not the host's hostname.
Fixes: #30348
Previously if one service specified the same unit as their
success and failure handler we bailed out of resolving the triggering unit
even though it is still unique.
The values that were previously hardcoded in sd-varlink.c are now defined
in new varlink_set_info_systemd() and that function is called everywhere
where we create a server.
This allows an option for systemd exec units to enable UTS namespaces
but not restrict changing hostname via seccomp. Thus, units can change
hostname without affecting the host.
Fixes: #30348
Migrating ProtectHostname to enum will set the stage for adding more
properties like ProtectHostname=private in future commits.
In addition, we add PrivateHostnameEx property to dbus API which uses
string instead of boolean.
Recently, PrivateUsers=identity was added to support mapping the first
65536 UIDs/GIDs from parent to the child namespace and mapping the other
UID/GIDs to the nobody user.
However, there are use cases where users have UIDs/GIDs > 65536 and need
to do a similar identity mapping. Moreover, in some of those cases, users
want a full identity mapping from 0 -> UID_MAX.
Note to differentiate ourselves from the init user namespace, we need to
set up the uid_map/gid_map like:
```
0 0 1
1 1 UINT32_MAX - 1
```
as the init user namedspace uses `0 0 UINT32_MAX` and some applications -
like systemd itself - determine if its a non-init user namespace based on
uid_map/gid_map files. Note systemd will remove this heuristic in
running_in_userns() in version 258 and uses namespace inode. But some users
may be running a container image with older systemd < 258 so we keep this
hack until version 259.
To support this, we add PrivateUsers=full that does identity mapping for
all available UID/GIDs.
Fixes: #35168
Follow-up for c26948c6da
We only want to get rid of cred mount here, and RuntimeDirectory=
is documented to be retained for SERVICE_EXITED state.
Fixes#35427
let's just check the debug invocation boolean, and not recheck the
restart mode again. It's mostly redundant (because the boolean should
not have been become true if the restart mode was not set accordingly).
Moreover, i think we might want to eventually allow a manual way to
enable debug invocation mode, and hence this pointless checking would
become a problem.
Also, we never check the restart mode again in other cases, hence we
shouldn't here either.
This partially reverts the commit 405be62f05
"tree-wide: refuse enumerated device with ID_PROCESSING=1".
Otherwise, when systemd-udev-trigger.service is (re)started just before
daemon-reexec, which can be easily happen on systemd package update, then
udev database files for many devices may have ID_PROCESSING=1 property,
thus devices may not be enumerated on daemon-reexec. That causes many
units especially mount units being deactivated after daemon-reexec.
Fixes#35329.
Currently, get_fixed_user() employs USER_CREDS_SUPPRESS_PLACEHOLDER,
meaning home path is set to NULL if it's empty or root. However,
the path is also used for applying WorkingDirectory=~, and we'd
spuriously use the invoking user's home as fallback even if
User= is changed in that case.
Let's instead delegate such suppression to build_environment(),
so that home is proper initialized for usage at other steps.
shell doesn't actually suffer from such problem, but it's changed
too for consistency.
Alternative to #34789
We use the $WATCHDOG_USEC variable for two very closely uses: as part of
the sd_watchdog_enabled() protocol for implementing service watchdogs.
And as part of the protocol between the service manager and
systemd-shutdown across the PID 1 execve() transition during shutdown.
Apparently some exitrds tools got confused by the latter use. Let's
address that by setting $WATCHDOG_PID to 1, in accordance to the
sd_watchdog_enabled() protocol to make clear this is only intended for
PID 1 and nothing else.
Replaces: #35135
Since v256 we completely fail to boot if v1 is configured. Fedora 41 was just
released with v256.7 and this is probably the first major exposure of users to
this code. It turns out not work very well. Fedora switched to v2 as default in
F31 (2019) and at that time some people added configuration to use v1 either
because of Docker or for other reasons. But it's been long enough ago that
people don't remember this and are now very unhappy when the system refuses to
boot after an upgrade.
Refusing to boot is also unnecessarilly punishing to users. For machines that
are used remotely, this could mean somebody needs to physically access the
machine. For other users, the machine might be the only way to access the net
and help, and people might not know how to set kernel parameters without some
docs. And because this is in systemd, after an upgrade all boot choices are
affected, and it's not possible to e.g. select an older kernel for boot. And
crashing the machine doesn't really serve our goal either: we were giving a
hint how to continue using v1 and nothing else.
If the new override is configured, warn and immediately boot to v1.
If v1 is configured w/o the override, warn and wait 30 s and boot to v2.
Also give a hint how to switch to v2.
https://bugzilla.redhat.com/show_bug.cgi?id=2323323https://bugzilla.redhat.com/show_bug.cgi?id=2323345https://bugzilla.redhat.com/show_bug.cgi?id=2322467https://www.reddit.com/r/Fedora/comments/1gfcyw9/refusing_to_run_under_cgroup_01_sy_specified_on/
The advice is to set systemd.unified_cgroup_hierarchy=1 (instead of removing
systemd.unified_cgroup_hierarchy=0). I think this is easier to convey. Users
who are understand what is going on can just remove the option instead.
The caching is dropped in cg_is_legacy_wanted(). It turns out that the
order in which those functions are called during early setup is very fragile.
If cg_is_legacy_wanted() is called before we have set up the v2 hierarchy,
we incorrectly cache a true answer. The function is called just a handful
of times at most, so we don't really need to cache the response.
This new setting allows unsharing the pid namespace in a unit. Because
you have to fork to get a process into a pid namespace, we fork in
systemd-executor to get into the new pid namespace. The parent then
sends the pid of the child process back to the manager and exits while
the child process continues on with the rest of exec_invoke() and then
executes the actual payload.
Communicating the child pid is done via a new pidref socket pair that is
set up on manager startup.
We unshare the PID namespace right before the mount namespace so we
mount procfs correctly. Note PrivatePIDs=yes always implies MountAPIVFS=yes
to mount procfs.
When running unprivileged in a user session, user namespace is set up first
to allow for PID namespace to be unshared. However, when running in
privileged mode, we unshare the user namespace last to ensure the user
namespace does not own the PID namespace and cannot break out of the sandbox.
Note we disallow Type=forking services from using PrivatePIDs=yes since the
init proess inside the PID namespace must not exit for other processes in
the namespace to exist.
Note Daan De Meyer did the original work for this commit with Ryan Wilson
addressing follow-ups.
Co-authored-by: Daan De Meyer <daan.j.demeyer@gmail.com>
The names of these conflict with macros from efi.h that we'll move
to efi-fundamental.h in a later commit. Let's avoid the conflict by
getting rid of these helpers. Arguably this also improves readability
by clearly indicating we're passing arbitrary strings and not constants
to the macros when we invoke them.
When an exec directory is shared between services, this allows one of the
service to be the producer of files, and the other the consumer, without
letting the consumer modify the shared files.
This will be especially useful in conjunction with id-mapped exec directories
so that fully sandboxed services can share directories in one direction, safely.
By default mount(8), umount(8), swapon(8) and swapoff(8) should run with
with the SMACK label inherited from systemd rather than the default one
meant for services.
Fixes: aa5ae9711e
Follow-up-for: 20bbf5ee4c
Let's make ConfigurationDirectory= a bit less "special-casey", by hiding
the fact that it's the only per-service dir we do not do chown()ing for
inside of a new EXEC_DIRECTORY_TYPE_SHALL_CHOWN() helper.
This applies the existing SocketUser=/SocketGroup= options to units
defining a POSIX message queue, bringing them in line with UNIX
sockets and FIFOs. They are set on the file descriptor rather than
a file system path because the /dev/mqueue path interface is an
optional mount unit.
This commit adds two settings private and strict to
the ProtectControlGroups= property. Private will unshare the cgroup
namespace and mount a read-write private cgroup2 filesystem at /sys/fs/cgroup.
Strict does the same except the mount is read-only. Since the unit is
running in a cgroup namespace, the new root of /sys/fs/cgroup is the unit's
own cgroup.
We also add a new dbus property ProtectControlGroupsEx which accepts strings
instead of boolean. This will allow users to use private/strict via dbus
and systemd-run in addition to service files.
Note private and strict fall back to no and yes respectively if the kernel
doesn't support cgroup2 or system is not using unified hierarchy.
Fixes: #34634
This commit refactors ProtectControlGroups= from using a boolean
in the dbus/execute backend to using an enum. There is no functional
change but this will allow adding new non-boolean values (e.g. strict,
private) a la PrivateHome.
The goal of RestartMode=direct is to make restarts invisible
to dependents, so auto restart jobs shouldn't bring them down
at all. So far we only skipped going through failed/dead states
in service_enter_dead(), i.e. the unit would never be considered
dead. But when constructing restart transaction, the stop job
would be propagated to dependents. Consider the following 2 units:
dependent.target:
[Unit]
BindsTo=a.service
After=a.service
a.service:
[Service]
ExecStart=bash -c 'sleep 100 && exit 1'
Restart=on-failure
RestartMode=direct
Before this commit, even though BindsTo= isn't triggered since
a.service never failed, when a.service auto-restarts, dependent.target
is also restarted. Let's suppress it by using JOB_REPLACE instead of
JOB_RESTART_DEPENDENCIES in service_enter_restart().
Fixes#34758
The example above is subtly different from the original report,
to illustrate that the new behavior makes sense for less exotic
use cases too.
Yu Watanabe
Daan De Meyer
Luca Boccassi
Mike Yuan
Ryan Wilson
Luca Boccassi
Lennart Poettering
Ronan Pigott
davjav
Nils K
Zbigniew Jędrzejewski-Szmek
David Tardon
Ivan Kruglov
Andres Beltran
Łukasz Stelmach
David Michael