David Herrmann [Sat, 20 Sep 2014 16:42:29 +0000 (18:42 +0200)]
login: add public sd_session_get_desktop() API
The desktop brand is stored as DESKTOP variable for sessions. It can be
set arbitrarily by the session owner and identifies the desktop
environment that is running on that session.
David Herrmann [Mon, 22 Sep 2014 10:49:47 +0000 (12:49 +0200)]
bus: align kdbus ioctl parameters to 8byte
All kdbus ioctl arguments must be 8byte aligned. Make sure we use
alloca_align() and _alignas_(8) in all situations where gcc doesn't
guarantee 8-byte alignment.
Note that objects on the stack are always 8byte aligned as we put
_alignas_(8) into the structure definition in kdbus.h.
David Herrmann [Mon, 22 Sep 2014 10:05:16 +0000 (12:05 +0200)]
util: add alloca_align()
The alloca_align() helper is the alloca() equivalent of posix_memalign().
As there is no such function provided by glibc, we simply account for
additional memory and return a pointer offset into the allocated memory to
grant the alignment.
Furthermore, alloca0_align() is added, which simply clears the allocated
memory.
David Herrmann [Sat, 20 Sep 2014 15:47:56 +0000 (17:47 +0200)]
terminal: raise sysview DEVICE_CHANGE events per attachment
Instead of raising DEVICE_CHANGE only per device, we now raise it per
device-session attachment. This is what we want for all sysview users,
anyway, as sessions are meant to be independent of each other. Lets avoid
any external session iterators and just do that in sysview itself.
David Herrmann [Sat, 20 Sep 2014 10:39:59 +0000 (12:39 +0200)]
terminal: forward evdev RESYNC events to linked devices
Whenever we resync an evdev device (or disable it), we should send RESYNC
events to the linked upper layers. This allows to disable key-repeat and
assume some events got dropped.
David Herrmann [Sat, 20 Sep 2014 10:34:43 +0000 (12:34 +0200)]
terminal: always call _enable/_disable on evdev devices
The current pause/resume logic kinda intertwines the resume/pause and
enable/disable functions. Lets avoid that non-obvious behavior and always
make resume call into enable, and pause call into disable, if appropriate.
Tom Gundersen [Sat, 20 Sep 2014 16:12:53 +0000 (18:12 +0200)]
test: udev - restrict nemuric uid's to existing ones
We now verify the existence of uid's before applying them to devicenodes, so change the
test accordingly. We assume that both uid/gid 1 and 2 exist on the test system.
David Herrmann [Sat, 20 Sep 2014 09:43:32 +0000 (11:43 +0200)]
terminal: fix mode sync for connectors
The GETXY ioctls of DRM are usually called twice by libdrm: Once to
retrieve the number of objects, a second time with suitably sized buffers
to actually retrieve all objects. In grdrm, we avoid these excessive calls
and instead just call ioctls with cached buffers and resize them if they
were too small.
However, connectors need to read the mode list via EDID, which is horribly
slow. As the kernel still cannot do that asynchronously (seriously, we
need to fix this!), it has a hack to only do it if count_modes==0. This is
fine with libdrm, as it calls every ioctl twice, anyway. However, we fail
horribly with this as we usually never pass 0.
Fix this by calling into GETCONNECTOR ioctls twice in case we received an
hotplug event. Only in those cases, we need to re-read modes, so this
should be totally fine.
David Herrmann [Sat, 20 Sep 2014 09:41:34 +0000 (11:41 +0200)]
terminal: restructure some logging calls in grdrm
Multiple issues here:
1) Don't print excessive card dumps on each resync. Disable it and make
developers add it themselves.
2) Ignore EINVAL on page-flips. Some cards don't support page-flips, so
we'd print it on each frame. Maybe, at some point, the kernel will add
support to retrieve capabilities for that. Until then, simply ignore
it.
3) Replace the now dropped card-dump with a short message about resyncing
the card.
David Herrmann [Sat, 20 Sep 2014 09:12:44 +0000 (11:12 +0200)]
terminal: grdev: schedule virtual frame events if hw doesn't support it
Whenever we cannot use hardware frame events, we now schedule a virtual
frame event to make sure applications don't have to do this. Usually,
applications render only on data changes, but we can further reduce
render-time by also limiting rendering to vsyncs.
David Herrmann [Sat, 20 Sep 2014 09:11:07 +0000 (11:11 +0200)]
terminal: grdev: raise frame event after DISPLAY_ADD/CHANGE
Whenever a display is added or changed, we suppressed any frame events.
Make sure to raise them manually so we can avoid rendering when handling
anything but FRAME events.
David Herrmann [Sat, 20 Sep 2014 07:48:22 +0000 (09:48 +0200)]
terminal: grdev: treat udev-devices without devnum as hotplug
If we get udev-device events via sysview, but they lack devnum
annotations, we know it cannot be a DRM card. Look through it's parents
and treat it as hotplug event in case we find such a card.
This will treat any new/removed connectors as sub-devices of the real
card, instead of as devices on its own.
David Herrmann [Sat, 20 Sep 2014 07:45:26 +0000 (09:45 +0200)]
terminal: reduce speed of morphing colors in modeset test
The high frequency of the color-morphing is kinda irritating. Reduce it
to a much lower frequency so you can actually look at it longer than few
seconds.
David Herrmann [Sat, 20 Sep 2014 07:44:14 +0000 (09:44 +0200)]
terminal: make drm-connectors first-level devices
So far, we only forward DRM cards via sysview APIs. However, with MST,
connectors can be hotplugged, too. Forward the connectors as first-level
devices via sysview so API users can react to changing DRM connectors.
David Herrmann [Sat, 20 Sep 2014 07:43:16 +0000 (09:43 +0200)]
terminal: forward DEVICE_CHANGE events via sysview
Whe need to react to "change" events on devices, but we want to avoid
duplicating udev-monitors everywhere. Therefore, make sysview forward
change events to the sysview controllers, which can then properly react
to it.
David Herrmann [Sat, 20 Sep 2014 07:29:11 +0000 (09:29 +0200)]
terminal: parse ID_SEAT not only for parents but the device itself
When deciding what seat a device is on, we have to traverse all parents
to find one with an ID_SEAT tag, otherwise, input devices plugged on a
seated USB-hub are not automatically attached to the right seat. But any
tags on the main device still overwrite the tags of the childs, so fix our
logic to check the device itself first, before traversing the parents.
David Herrmann [Fri, 19 Sep 2014 12:48:54 +0000 (14:48 +0200)]
terminal: add systemd-modeset debugging tool
The systemd-modeset tool is meant to debug grdev issues. It simply
displays morphing colors on any found display. This is pretty handy to
look for tearing in the backends and debug hotplug issues.
Note that this tool requires systemd-logind to be compiled from git
(there're important fixes that haven't been released, yet).
David Herrmann [Fri, 19 Sep 2014 12:13:06 +0000 (14:13 +0200)]
terminal: add grdev DRM backend
The grdev-drm backend manages DRM cards for grdev. Any DRM card with
DUMB_BUFFER support can be used. So far, our policy is to configure all
available connectors, but keep pipes inactive as long as users don't
enable the displays on top.
We hard-code double-buffering so far, but can easily support
single-buffering or n-buffering. We also require XRGB8888 as format as
this is required to be supported by all DRM drivers and it is what VTs
use. This allows us to switch from VTs to grdev via page-flips instead of
deep modesets.
There is still a lot room for improvements in this backend, but it works
smoothly so far so more enhanced features can be added later.
David Herrmann [Fri, 19 Sep 2014 12:05:52 +0000 (14:05 +0200)]
terminal: add graphics interface
The grdev layer provides graphics-device access via the
libsystemd-terminal library. It will be used by all terminal helpers to
actually access display hardware.
Like idev, the grdev layer is built around session objects. On each
session object you add/remove graphics devices as they appear and vanish.
Any device type can be supported via specific card-backends. The exported
grdev API hides any device details.
Graphics devices are represented by "cards". Those are hidden in the
session and any pipe-configuration is automatically applied. Out of those,
we configure displays which are then exported to the API user. Displays
are meant as lowest hardware entity available outside of grdev. The
underlying pipe configuration is fully hidden and not accessible from the
outside. The grdev tiling layer allows almost arbitrary setups out of
multiple pipes, but so far we only use a small subset of this. More will
follow.
A grdev-display is meant to represent real connected displays/monitors.
The upper level screen arrangements are user policy and not controlled by
grdev. Applications are free to apply any policy they want.
David Herrmann [Fri, 19 Sep 2014 11:26:39 +0000 (13:26 +0200)]
login: pause devices before acknowledging VT switches
If a session controller does not need synchronous VT switches, we allow
them to pass VT control to logind, which acknowledges all VT switches
unconditionally. This works fine with all sessions using the dbus API,
but causes out-of-sync device use if we switch to legacy sessions that
are notified via VT signals. Those are processed before logind notices
the session-switch via sysfs. Therefore, leaving the old session still
active for a short amount of time.
This, in fact, may cause the legacy session to prepare graphics devices
before the old session was deactivated, and thus, maybe causing the old
session to interfer with graphics device usage.
Fix this by releasing devices immediately before acknowledging VT
switches. This way, sessions without VT handlers are required to support
async session switching (which they do in that case, anyway).
Michal Sekletar [Thu, 24 Jul 2014 08:40:28 +0000 (10:40 +0200)]
socket: introduce SELinuxContextFromNet option
This makes possible to spawn service instances triggered by socket with
MLS/MCS SELinux labels which are created based on information provided by
connected peer.
Implementation of label_get_child_mls_label derived from xinetd.
David Herrmann [Thu, 18 Sep 2014 22:23:42 +0000 (00:23 +0200)]
pty: include linux/ioctl.h for TIOCSIG
TIOCSIG is linux specific, so include the linux ioctl header to make sure
it's defined. We currently rely on some rather non-obvious recursive
includes. Make sure its always defined regardless of the system headers.
When compiling we see this curl warning popping up:
src/journal-remote/journal-upload.c:194:17: warning: call to
‘_curl_easy_setopt_err_error_buffer’ declared with attribute
warning: curl_easy_setopt expects a char buffer of CURL_ERROR_SIZE
as argument for this option [enabled by default]
This patch removes the warning (which occurs twice).
Michael Marineau [Mon, 15 Sep 2014 21:07:39 +0000 (14:07 -0700)]
man: use the escape for "-" in example instead of space.
This sentence can be misread to mean that "\x20" is the escape code for
"-" which is the only character explicitly mentioned. This lead to at
least one user loosing hair over why a mount unit for "/foo/bar-baz"
didn't work. The example escape is arbitrary so lets prevent hair loss.
Robert Milasan [Sat, 13 Sep 2014 13:18:37 +0000 (15:18 +0200)]
udev: fix path for database names on 'change' event
If a device does not have a major/minor number attached, we use different
database names than if it does. On "change" events, we didn't copy the
devnum over, therefore, we used different paths than on 'add' or 'remove'
events (where devnum was properly copied).
Fix this by always copying the devnum into the udev-device.
David Herrmann [Wed, 17 Sep 2014 08:32:49 +0000 (10:32 +0200)]
bus: fix error leak in bus_node_exists()
If we call into user callbacks, we must always propagate possible errors.
Fix bus_node_exists() to do that and adjust the callers (which already
partially propagated the error).
Also speed up that function by first checking for registered enumerators
and/or object-managers.
David Herrmann [Wed, 17 Sep 2014 07:28:09 +0000 (09:28 +0200)]
bus: never respond to GetManagedObjects() on sub-paths
The dbus-spec clearly specifies that GetManagedObjects() should only work
on the root-path of an object-tree. But on that path, it works regardless
whether there are any objects available or not.
We could, technically, define all sub-paths as a root-path of its own
sub-tree. However, if we do that, we enter undefined territory:
Imagine only a fallback vtable is registered. We want
GetManagedObjects() to *NOT* fail with UNKNOWN_METHOD if it is called
on a valid sub-tree of the fallback. On the other hand, we don't want
it to work on arbitrary sub-tree. Something like:
/path/to/fallback/foobar/foobar/foobar/invalid/foobar
should not work.
However, there is no way to know which paths on a fallback are valid
without looking at there registered objects. If no objects are
registered, we have no way to figure it out.
Therefore, we now try to follow the dbus spec by only returning valid data
on registered root-paths. We treat each path as root which was registered
an object-manager on via add_object_manager(). So applications can now
directly control which paths to place an object-manager on.
We also fix the introspection to not return object-manager interfaces on
non-root paths.
Also fixes some dead-code paths initially reported by Philippe De Swert.
David Herrmann [Wed, 17 Sep 2014 07:06:49 +0000 (09:06 +0200)]
sysctl: make --prefix allow all kinds of sysctl paths
Currently, we save arguments passed as --prefix directly and compare them
later to absolute sysctl file-system paths. That is, you are required to
specify arguments to --prefix with leading /proc/sys/. This is kinda
uselesss. Furthermore, we replace dots by slashes in the name, which makes
it impossible to match on specific sysfs paths that have dots in their
name (like netdev names). The intention of this argument is clear, but it
never worked as expected.
This patch modifies --prefix to accept any kind of sysctl paths. It
supports paths prefixed with /proc/sys for compatibility (but drops the
erroneous dot->slash conversion), but instead applies normalize_sysctl()
which turns any name or path into a proper path. It then appends
/proc/sys/ so we can properly use it in matches.
Thanks to Jan Synacek <jsynacek@redhat.com> for catching this!
David Herrmann [Tue, 16 Sep 2014 21:00:26 +0000 (23:00 +0200)]
terminal: remove dead code checking O_WRONLY
We only reject evdev FDs that are O_WRONLY as they're currently pretty
useless. The following check for O_WRONLY is thus never excercised so drop
it.
Thanks to Thomas Andersen (via coverity)!
sd-bus: sd_bus_message_get_errno should only return positive errno
sd_bus_message_get_errno can currently return either a number of
different poitive errno values (from bus-error-mapping), or a negative
EINVAL if passed null as parameter.
The check for null parameter was introduced in 40ca29a1370379d43e44c0ed425eecc7218dcbca
at the same as the function was renamed from bus_message_to_errno and
made public API. Before becoming public the function used to return
only negative values.
It is weird to have a function return both positive and negative errno
and it generally looks like a mistake. The function is guarded by the
--enable-kdbus flags so I wonder if we still have time to fix it up?
It does not have any documentation yet. However, except for a few details
it is just a convenient way to call sd_bus_error_get_errno which is documented
to return only positive errno.
This patch makes it return only positive errno and fixes up the two
calls to the function that tried to cope with both positive and negative
values.
Coverity warned that we have already dereferenced ps->sample before
null-checking it. I suspect that's not really the issue and that
the check is checking the wrong variable.
Likely the oom-check should be on the just allocated ps->sample->next.
libsystemd-network: avoid double-free in error case
Don't manually free 'n' in error path as it's already tagged
_cleanup_free_ and will be freed once it goes out of scope,
leading to double-free in this case.