Let's follow our modern style (i.e. return proper errors, use structure
initialization and _cleanup_).
Most importantly: remove state file and FIFO removal from
inhibitor_free() and let's move it to inhibitor_stop().
This makes sure that state files/FIFOs are not removed when the we
terminate logind, i.e. that they can survive logind restarts.
Fixes: #11825
Also make some parsing errors, fatals and others (that just care fore
'decoration') non-fatal.
The single caller of inhibitor_load() didn't log about any errors, hence
let's do this in our function, similar to how this is done in
session_load() already.
For some reason, systemd-logind is trying to handle idle action in one of my containers:
Jun 07 10:28:08 rawhide systemd-logind[42]: System idle. Taking action.
Jun 07 10:28:08 rawhide systemd-logind[42]: Requested operation not supported, ignoring.
But we didn't log what exactly was being done. Let's put the name of the action in messages.
All callers pass either a fixed action, or HANDLE_IGNORE is explicitly filtered
out. Let's remove this case here, because we cannot properly log what opreation
we are ignoring.
Previously, logind's logind-session.h would define prototypes for
logind-session.c and logind-session-dbus.c. Split that out, so that
there's a separate logind-session-dbus.h for that. Similar for seats and
users as well as the manager itself.
This changes no code, just rearranges where protoypes are located.
Most of the operations one can do on sessions so far accepted an empty
session name as a shortcut for the caller's session. This is quite
useful traditionally, but much less useful than it used to be, since
most user code now (rightfully) runs in --user context, not in a
session.
With this change we tweak the logic a bit: we introduce the two special
session and seat names "self" and "auto". The former refers to the
session/seat the client is in, and is hence mostly equivalent to te
empty string "" as before. However, the latter refers to the
session/seat the client is in if that exists, with a fallback of the
user's display session if not. Clients can hence reference "auto"
instead of the empty string if they really don't want to think much
about sessions.
Why "self" btw? Previously, we'd already expose a special dbus object
with the path /org/freedesktop/login1/session/self (and similar for the
seat), matching what the empty string did for bus calls that took a
session name. With this scheme we reuse this identifier and introduce
"auto" in a similar way.
Of course this means real-life seats and sessions can never be named
"self" or "auto", but they aren't anyway: valid seat names have to start
with "seat" anyway, and sessions are generated server-side as either a
numeric value or "c" suffixed with a counter ID.
Fixes: #12399
Interestingly, elect_display_compare() already ordered "user" sessions
before "greeter" sessions, though nothing other than "user" sessions
where ever considered anyway.
Fixes: #12399
This augments the drm/input device management by adding a single method
call for setting the brightness of an "leds" or "backlight" kernel class
device.
This method call requires no privileges to call, but a caller can only
change the brightness on sessions that are currently active, and they
must own the session.
This does not do enumeration of such class devices, feature or range
probing, chnage notification; it doesn't help associating graphics or
input devices with their backlight or leds devices. For all that clients
should go directly to udev/sysfs. The SetBrightness() call is just for
executing the actual change operation, that is otherwise privileged.
Example line:
busctl call org.freedesktop.login1 /org/freedesktop/login1/session/self org.freedesktop.login1.Session SetBrightness ssu "backlight" "intel_backlight" 200
The parameter the SetBrightness() call takes are the kernel subsystem
(i.e. "leds" or "backlight"), the device name, and the brightness
value.
On some hw setting the brightness is slow, and implementation and write
access to the sysfs knobs exposes this slowness. Due to this we'll fork
off a writer process in the background so that logind doesn't have to
block. Moreover, write requestes are coalesced: when a write request is
enqueued while one is already being executed it is queued. When another
write reques is then enqueued the earlier one is replaced by the newer
one, so that only one queued write request per device remains at any
time. Method replies are sent as soon as the first write request that
happens after the request was received is completed.
It is recommended that bus clients turn off the "expect_reply" flag on
the dbus messages they send though, that relieves logind from sending
completion notification and is particularly a good idea if clients
implement reactive UI sliders that send a quick secession of write
requests.
Replaces: #12413
These devices do not become user-accessible this way, but they are
logically assigned to a seat, which makes a lot of sense, since they are
human-facing output devices, and such should belong to one.
When emitting the calendarspec warning we want to see some color.
Follow-up for 04220fda5c.
Exceptions:
- systemctl, because it has a lot hand-crafted coloring
- tmpfiles, sysusers, stdio-bridge, etc, because they are also used in
services and I'm not sure if this wouldn't mess up something.
This is partially a refactoring, but also makes many more places use
unlocked operations implicitly, i.e. all users of fopen_temporary().
AFAICT, the uses are always for short-lived files which are not shared
externally, and are just used within the same context. Locking is not
necessary.
let's use sd_notifyf(). Let's also stop validating the session ID here.
This is the destructor. if it contains a dash, we are already too late
here anyway.
Let's be helpful to static analyzers which care about whether we
knowingly ignore return values. We do in these cases, since they are
usually part of error paths.
This adds support for user to set & get reboot parameter for reboot.
As callee would be next issuing Reboot call same policy checks are being used.
If unit file issuing the reboot action defines RebootArgument (or similar) that
setting takes precedence.
Setting an access mode != 0666 is explicitly supported via -Dgroup-render-mode
In such a case, re-add the uaccess tag.
This is basically the same change that was done for /dev/kvm in
commit fa53e24130 and
ace5e3111c
and partially reverts the changes from
4e15a7343c
When 'nomodeset' is specified, there's no DRM driver to take over from
efifb. This means no device will be marked as a seat master, so gdm will
never find a sufficiently active seat to start on.
I'm not aware of an especially good way to detect this through a proper
kernel API, so check for the word 'nomodeset' on the command line and
allow fbdev devices to be seat masters if found.
For https://bugzilla.redhat.com/show_bug.cgi?id=1683197.
This behaves similar to the "boot into firmware" logic, and also allows
either direct EFI operation (which sd-boot supports and others might
support eventually too) or override through env var.
This extends the reboot-to-firmware logic in logind, so that other than
EFI firmwares could be theoretically support. The scheme is like this:
if you want to support this, set the $SYSTEMD_REBOOT_TO_FIRMWARE=1 env
var for logind. If so, this will override the EFI logic, and cause a
file /run/systemd/reboot-to-firmware file to be created when
reboot-to-firmware is requested. This file has no contents, it's mere
existance indicates a reboot with reboot-to-firmware set.
The idea is that for alternative firmwares a drop-in for logind is added
that sets the env var, in combination with some code run during shutdown
that checks for the file and does the right thing.
User instance of systemd is optional feature and if user@.service
template is masked then administrator most likely doesn't want --user
instances of systemd for logged in users. We don't need to be verbose
about it.
This enables graphical capability for a video adapter of Parallels
virtualization platform (Parallels Desktop for Mac product) which is not
a DRM device at the moment.
This fixes GUI in Fedora 29 guest on Parallels Desktop where gdm now
strictly checks for CanGraphical property of a seat, see [1].
Should be noted that there's no in-kernel driver for Parallels video at
the moment so device matching is done by vid/pid.
[1] https://gitlab.gnome.org/GNOME/gdm/merge_requests/37
This reverts commit 69bd76f2b9.
$DBUS_SESSION_BUS_ADDRESS is again set only if the socket exists.
Quoting https://github.com/systemd/systemd/pull/11327#issuecomment-452019027:
> [setting $DBUS_SESSION_BUS_ADDRESS unconditionally] makes pam_systemd
> incompatible with installations and distributions where dbus was not
> configured with --enable-user-session, and the session dbus-daemon is started
> by autolaunching or dbus-launch (as opposed to dbus.socket). I don't think
> that's wise: using autolaunching or dbus-launch, and disabling or not
> installing dbus.socket and dbus.service on the systemd user instance, is our
> compatibility story for people who still need a D-Bus session bus per X11
> session for whatever reason.
>
> For example, Debian can currently do either way, with a dbus-user-session
> package strongly recommended but not actually mandatory. dbus-user-session
> requires libpam-systemd; if pam_systemd now requires dbus.socket (which is in
> the dbus-user-session package), that's a circular dependency, which we
> normally try hard to avoid.
For systems that use dbus.socket this doesn't matter much, because the
user session is ordered after the user managaer, which pulls in dbus.socket
very early. For example, when logging over ssh:
sshd[20796]: pam_systemd(sshd:session): pam-systemd initializing
sshd[20796]: pam_systemd(sshd:session): Asking logind to create session: uid=1001 pid=20796 service=sshd type=tty class=user desktop= seat= vtnr=0 tty= display= remote=yes remote_user= remote_host=::1
sshd[20796]: pam_systemd(sshd:session): Session limits: memory_max=n/a tasks_max=n/a cpu_weight=n/a io_weight=n/a
systemd[1]: Created slice User Slice of UID 1001.
systemd[1]: Starting User Runtime Directory /run/user/1001...
systemd-logind[1210]: New session 3796 of user guest.
systemd[1]: Started User Runtime Directory /run/user/1001.
systemd[1]: Starting User Manager for UID 1001...
systemd[20805]: pam_systemd(systemd-user:session): pam-systemd initializing
systemd[20805]: Starting D-Bus User Message Bus Socket.
...
systemd[20805]: Reached target Sockets.
systemd[20805]: Reached target Basic System.
systemd[1]: Started User Manager for UID 1001.
systemd[1]: Started Session 3796 of user guest.
sshd[20796]: pam_systemd(sshd:session): Reply from logind: id=3796 object_path=/org/freedesktop/login1/session/_33796 runtime_path=/run/user/1001 session_fd=13 seat= vtnr=0 original_uid=1001
sshd[20796]: pam_unix(sshd:session): session opened for user guest by (uid=0)
Hence, everything in the ssh session is ordered after the user instance.
And in the user instance, services should be orderd after dbus.socket using
inter-unit dependencies. dbus.socket in turns does
systemctl --user set-environment DBUS_SESSION_BUS_ADDRESS=unix:path=%t/bus.
So there should be no race between starting of the dbus socket and our check
if it exists.
The alternative would be to set the "DBUS_SESSION_BUS_ADDRESS=unix:path=%s/bus;autolaunch:".
AFAICT, this would work as well. But I don't see any case where it actually works
better. Since this is an area with many compatiblity concerns, let's stick to
the previous setup which seems to work well.
The message is changed from
Cannot create session: Already running in a session...
to
Not creating session: Already running in a session...
This is more neutral and avoids suggesting a problem.
"Will not create session: ..." was suggested, but it sounds like the action
would have yet to be performed. I think Using present continuous is better.
Fixes#10822 (for good now, I hope).
There's very little lost if the variable is set for a socket that isn't
connectible, but a lot lost (races, ...) if it's not set but the socket exists.
Also, drop the FIXME note, since we don't plan to revert this revert any time
soon.
This reverts commit 2b2b7228bf.
Fixes#11293.
Removing the environment variable causes problems, e.g. Xfce and Chromium and
... don't communicate with the running dbus instance. If they attempt to start their
own instance, things become even more confusing. Those packages could be fixed
one by one, but removing the variable right now is causing too many problems.
This reverts commit edda44605f.
The kernel explicitly supports resuming with a different kernel than the one
used before hibernation. If this is something that shouldn't be supported, the
place to change this is in the kernel. We shouldn't censor something that this
exclusively in the kernel's domain.
People might be using this to switch kernels without restaring programs, and
we'd break this functionality for them.
Also, even if resuming with a different kernel was a bad idea, we don't really
prevent that with this check, since most users have more than one kernel and
can freely pick a different one from the menu. So this only affected the corner
case where the kernel has been removed, but there is no reason to single it
out.
This has been irritating me for quite a while: let's prefix these enum
values with a common prefix, like we do for almost all other enums.
No change in behaviour, just some renaming.
This splits out a bunch of functions from fileio.c that have to do with
temporary files. Simply to make the header files a bit shorter, and to
group things more nicely.
No code changes, just some rearranging of source files.
Similar to the previous commit: in many cases no further fd processing
needs to be done in forked of children before execve() or any of its
flavours are called. In those case we can use FORK_RLIMIT_NOFILE_SAFE
instead.
Ideally, coccinelle would strip unnecessary braces too. But I do not see any
option in coccinelle for this, so instead, I edited the patch text using
search&replace to remove the braces. Unfortunately this is not fully automatic,
in particular it didn't deal well with if-else-if-else blocks and ifdefs, so
there is an increased likelikehood be some bugs in such spots.
I also removed part of the patch that coccinelle generated for udev, where we
returns -1 for failure. This should be fixed independently.
This reverts 5fdf2d51c2, except for one improved
log message.
Fixes#10613.
Checking if resume= is configured is a good idea, but it turns out we cannot do
it reliably:
- the code only supported boot options with sd-boot, and it's not very widely
used. This means that for most systemd we could only check the current
commandline, not the next one.
- Various systems resume without, e.g. Debian has
/etc/initramfs-tools/conf.d/resume in the initramfs.
Making those checks better would be possible with enough effort, but there'll
be always new systems that boot in a slightly different way and we would need
to keep adding new cases. Longer term, we want to rely on autodetecting the
resume partition, and then checks like this will not be necessary at all. It is
quite clear from the number of bug reports that the number of poeple impacted
by this is quite high now, so let's just drop this.
When no sessions are registered on a given VT (anymore), we should always let
the kernel processes VT switching (instead of simply emitting a warning)
otherwise the requests sent by the kernel are simply ignored making the VT
switch requested by users simply impossible.
Even if it shouldn't happen, this case was encountered in issue #9754, so
better to be safe than sorry.
Basically when a session ends, logind notices and restores VT_AUTO so the
kernel takes back VT-switching over.
logind achieves that by watching the process that took control of the session
(via the "TakeControl" D-Bus method), aka "the watched process", which can
be different from the one that initially opened the VT aka "the terminal
controlling process".
In this case the terminal controlling process can exit after the watched one
did and while logind is restoring the VT.
Even if logind took care to re-open the VT in case the VT was already in HUP
state, it wasn't enough because the terminal controlling process could have
exited right after, leaving the VT in HUP state and in VT_PROCESS mode making
further VT-switching impossible.
This patch fixes this situation by forcing logind to become the terminal
controlling process.
Fixes: #9754.
We generally want to close the pager last. This patch closes the pager last,
after the static destuctor calls. This means that they can do logging and such
like during normal program runtime.
This way, we can extend the macro a bit with stuff pulled in from other
headers without this affecting everything which pulls in macro.h, which
is one of our most basic headers.
This is just refactoring, no change in behaviour, in prepartion for
later changes.
In the PAM module we need to suppress LOG_DEBUG messages manually, if
debug logging is not on, as PAM won't do this for us. We did this
correctly for most log messages already, but two were missing. Let's fix
those too.
Fixes: #10822
If suspend-then-hibernate, hybrid-sleep or plain hibernation is
supposed to be execute due to a key press/lid switch but is not
supported, automatically fall back to plain suspend (and log about it).
Fixes: #10558
The three core variables that affect idleness handling are whether we
are docked, whether we are on AC power and whether the lid is closed,
hence let's also expose the third variable on the bus, to make things
nicely debuggable.
It's not, after all, that's what SetRebootToFirmware() is about.
(I was wondering for a moment whether to make this EMITS_CHANGES, but
decided against it, given that the flag actually can be changed
externally to logind too, and we couldn't send out notifications for
that.)
Now that we don't (mis-)use the env file parser to parse kernel command
lines there's no need anymore to override the used newline character
set. Let's hence drop the argument and just "\n\r" always. This nicely
simplifies our code.
Pretty much everything uses just the first argument, and this doesn't make this
common pattern more complicated, but makes it simpler to pass multiple options.
Currently we consider any framebuffer device as enough to have a
valid graphical session, but this might lead to many false postives
like in the case of framebuffer devices that have a linked drm card
which is still in the process of being added, or for vesa fb, and
so it doesn't ensure us that we can have a proper graphical session.
Since these days we normally don't consider anything without a DRM
card able to provide a full graphical session, let's not set this
at this level.
Drivers which can provide a graphical session with the sole fb are
still free to mark any device as `master-of-seat`
Fixes#10435
Let's be more careful with what we serialize: let's ensure we never
serialize strings that are longer than LONG_LINE_MAX, so that we know we
can read them back with read_line(…, LONG_LINE_MAX, …) safely.
In order to implement this all serialization functions are move to
serialize.[ch], and internally will do line size checks. We'd rather
skip a serialization line (with a loud warning) than write an overly
long line out. Of course, this is just a second level protection, after
all the data we serialize shouldn't be this long in the first place.
While we are at it also clean up logging: while serializing make sure to
always log about errors immediately. Also, (void)ify all calls we don't
expect errors in (or catch errors as part of the general
fflush_and_check() at the end.
This makes use of rlimit_nofile_bump() in all tools that access the
journal. In some cases this replaces older code to achieve this, and
others we add it in where it was missing.
Let's be safe than sorry, in particular as logind doesn't set it up
anymore, but user-runtime-dir@.service does, and logind doesn't really
track success of that.
This heavily borrows from @intelfx' PR #5546, but watches all three
units that are associated with a user now: the slice, the user@.service
and user-runtime-dir@.service.
The logic and reasoning behind it is the same though: there's no value
in keeping lingering users around if all their three services are gone.
Replaces: #5546Fixes: #4162
Let's make this a bit prettier, and propagate unexpected access() errors
correctly.
(The callers of this function will suppress them, but it's nicer of they
do that, rather than us doing that twice in both the callers and the
callees)
This is useful so that during shutdown scope units are always terminated
before the mounts necessary for the home directory.
(Ideally we'd also add a similar dependency from the user@.service
instance to the home directory, but this isn't as easy as that service
is defined statically and not dynamically, and hence not easy to modify
dynamically, in particular when it comes to deps)
I think this is a slightly cleaner approach than parsing the
configuration file at multiple places, as this way there's only a single
reload cycle for logind.conf, and that's systemd-logind.service's
runtime.
This means that logind and dbus become a requirement of
user-runtime-dir, but given that XDG_RUNTIME_DIR is not set anyway
without logind and dbus around this isn't really any limitation.
This also simplifies linking a bit as this means user-runtime-dir
doesn't have to link against any code of logind itself.
Instead of managing it explicitly, let's simplify things and rely on
regular Wants=/Requires= dependencies to pull in these units from
user@.service and the session scope, and StopWhenUneeded= to stop these
auxiliary units again. This way, they can be pulled in easily by
unrelated units too.
This simplifies things quite a bit: for each session we now only need to
manage the session scope, and for each user the user@.service, the other
units are not something we need to manage anymore.
This patch also makes sure that if user@.service of a user is masked we
will continue to work, and user-runtime-dir@.service will still be
correctly pulled in, as it is now a dependency of the scope unit.
Fixes: #9461
Replaces: #5546
Let's propagate errors from stopping sessions via seat_stop(). This is
similar to how we propagate such errors in user_stop() for all sessions
associated with a user.
Note that we propagate these errors, but we don't abort the function.
let's make sure we log about every failure
Also, complain about systems where /dev/tty0 exists but
/sys/class/tty/tty0/active does not. Such systems (usually container
environments) are pretty broken as they mount something that is not a VC
to /dev/tty0 and they really shouldn't.
Systems should either have a VC or not, but not badly fake one by
mounting things wildly.
This just adds a warning message, as before we'll simply turn off VC
handling in this case.
Previously this was serialized as part of the user object. This didn't
work however, as we load users first, and sessions seconds and hence
referencing a session from the user load logic cannot work.
Fix this by storing an IS_DISPLAY property along with each session, and
make the session with this set display session when it is loaded.
Let's update things a bit to follow current practices:
- User structure initialization rather than zero-initialized allocation
- Always propagate proper errors from allocation functions
- Use _cleanup_ for freeing objects when allocation fails half-way
- Make destructors return NULL
Also, while we are at it, beef it up, by adding json-seq support (i.e.
https://tools.ietf.org/html/rfc7464). This is particularly useful in
conjunction with jq's --seq switch.
For non-`seat0` seats, attaching a graphics card to a seat can
lead to it getting created. This is because the graphics device
is a "master device" which means that device is a seat-defining
device.
`seat0` may get created, even before the graphics driver is loaded,
though. This is because the graphics driver is loaded
asynchronously at startup, and `seat0` is the primary seat of
system, associated with the system VTs.
When a graphics card is attached to a seat the `CanGraphical`
property on that seat will flip to `true`.
For seats that haven't been created yet (non-`seat0` seats), this
leads to `seat_start` getting called which ultimately causes the
seat to get serialized to `/run/systemd/seats`.
For `seat0`, which is already created, `seat_start` will return
immediately, which means the updated `CanGraphical` state will
never get written to `/run/systemd/seats`.
The end result is that clients querying `sd_seat_can_graphical`
won't get the correct answer for `seat0` in cases where the
graphics device takes a long time to load until some other peice
of seat state is updated.
This commit fixes the problem by calling `seat_save` explicitly
for already running seats at the time a graphics device is
attached.