Martin Pitt [Sun, 1 Feb 2015 23:19:29 +0000 (00:19 +0100)]
tests: determine rootlibdir from pkg-config
Prefer asking pkg-config for the rootlibdir, to also support systems with
--enable-split-usr. Fall back to the hardcoded /usr/lib/systemd if that fails.
test-dhcp-client: open a real udp socket for test execution
./test-dhcp-client would attempt to operate fd 0, i.e. stdin.
For example, './test-dhcp-client </dev/null' would fail with EPERM
because /dev/null cannot be used with epoll.
util: rework strappenda(), and rename it strjoina()
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.
bootchart: Ensure that systemd is the init called after using bootchart
When booting with systemd-bootchart, default to call the systemd binary
rather than the init binary on disk, which might be another init system.
Collecting data only works with booting systemd.
Sangjung Woo [Mon, 2 Feb 2015 11:25:02 +0000 (20:25 +0900)]
tmpfiles: Remove unnecessary blank line when configured with "--disable-resolved"
This patch removes unnecessary blank line in
/usr/lib/tmpfiles.d/etc.conf when configured with "--disable-resolved".
(i.e. ENABLE_RESOLVED is not defined)
Drop all capabilities:
1. prctl(PR_SET_KEEPCAPS, keep_capabilities != 0) // 0 when we drop all
capabilities
2. setresuid() // bye bye capabilities
3. Add CAP_SETPCAP // fails because we have no capabilities
4. Reduce capability bounding set
5. Drop capabilities
6. prctl(PR_SET_KEEPCAPS, 0)
Capabilites should always be kept after setresuid() so that the capability
bounding set can be reduced.
We must be careful not to leave PR_SET_KEEPCAPS on. We could use the
setresuid() call to drop capabilities, but the rules when capabilities
are dropped are fairly complex, since a transition to non-zero uid must
happen. Let's instead keep the capabilities during setresuid(), and drop
them later.
In the test, p is a path to a directory, always absolute. dent->d_name
is a single path component, so they cannot be equal. The comparison
was wrong also for other reasons: D type supports globs, so direct
comparisons using streq are not enough.
Services which are not crucial to system bootup, and have Type=oneshot
can effectively "hang" the system if they fail to complete for whatever
reason. To allow the boot to continue, kill them after a timeout.
In case of systemd-journal-flush the flush will continue in the background,
and in the other two cases the job will be aborted, but this should not
result in any permanent problem.
Tom Gundersen [Fri, 30 Jan 2015 18:54:01 +0000 (19:54 +0100)]
networkd: dhcp-server - start as soon as addresses have been set
We would otherwise wait for the interface to be completely configured, which
could take considerable time with IPv4LL. As a result nspawn was very slow
at obtaining IP addresses.
Tom Gundersen [Thu, 29 Jan 2015 06:26:58 +0000 (07:26 +0100)]
sd-rtnl: don't fail event handler when callback fails
As in sd-bus, simply log at debug level when a callback fails, but don't fail the event handler.
Otherwise any error returned by any callback will disable the rtnl event handler. We should
only do that on serious internal errors in sd-rtnl that we know cannot be recovered from.
Program received signal SIGSEGV, Segmentation fault.
0x000055555557c9be in main (argc=2, argv=0x7fffffffe4d8) at src/core/main.c:1832
1832 arg_shutdown_watchdog = m->shutdown_watchdog;
(gdb) bt
(gdb) bt full
m = 0x0
Peter Hutterer [Wed, 28 Jan 2015 04:05:01 +0000 (14:05 +1000)]
hwdb: add a touchpad hwdb
Currently used to tag devices in the new Lenovo *50 series and the X1 Carbon
3rd. These laptops re-introduced the physical trackpoint buttons that were
missing from the *40 series but those buttons are now wired up to the
touchpad.
The touchpad now sends BTN_0, BTN_1 and BTN_2 for the trackpoint. The same
button codes were used in older touchpads that had dedicated scroll up/down
buttons. Input drivers need to work around this and thus know what they're
dealing with.
For the previous gen we introduced INPUT_PROP_TOPBUTTONPAD in the kernel, but
the resulting mess showed that these per-device quirks should really live in
userspace.
The list currently includes the X1 Carbon 3rd PNPID, others will be added as
get to know which PNPID they have.
Martin Pitt [Wed, 28 Jan 2015 17:14:01 +0000 (18:14 +0100)]
logind: handle closing sessions over daemon restarts
It may happen that you have several sessions with the same VT:
- Open a session c1 which leaves some processes around, and log out. The
session will stay in State=closing and become Active=no.
- Log back in on the same VT, get a new session "c2" which is State=active and
Active=yes.
When restarting logind after that, the first session that matches the current
VT becomes Active=yes, which will be c1; c2 thus is Active=no and does not get
the usual polkit/device ACL privileges.
Restore the "closing" state in session_load(), to avoid treating all restored
sessions as State=active. In seat_active_vt_changed(), prefer active sessions
over closing ones if more than one session matches the current VT.
Finally, fix the confusing comment in session_load() and explain it a bit
better.
Martin Pitt [Wed, 28 Jan 2015 12:57:47 +0000 (13:57 +0100)]
rules: clean up stale CD drive mounts after ejection
Ejecting a CD with the hardware drive button only causes a change uevent, but
the device node stays around (just without a medium). Pick up these uevents and
mark the device as SYSTEMD_READY=0 on ejection, so that systemd stops the
device unit and consequently all mount units on it.
On media insertion, mark the device as SYSTEMD_READY=1 again.
Martin Pitt [Wed, 28 Jan 2015 12:53:25 +0000 (13:53 +0100)]
core/mount: add dependencies to dynamically mounted mounts too
Add unit dependencies for dynamic (i. e. not from fstab) mounts. With that,
mount units properly bind to their underlying device, and thus get
automatically stopped/unmounted when the underlying device goes away.
This cleans up stale mounts from unplugged devices.
Thanks to Lennart Poettering for pointing out the fix!
core: output unit status output strings to console, only if we actually are changing unit state
Unit _start() and _stop() implementations can fail with -EAGAIN to delay
execution temporarily. Thus, we should not output status messages before
invoking these calls, but after, and only when we know that the
invocation actually made a change.
Martin Pitt [Wed, 28 Jan 2015 07:00:28 +0000 (08:00 +0100)]
sysv-generator: Re-fix .sh suffix handling
Commit 4e48855534 caused the .sh suffix to be stripped from the original
"filename", which caused the generated units to call the wrong init.d script.
Only use the .sh stripped file name for comparing with Provides:, not for
generating the Exec*= lines.
core: if two start jobs for the same swap device node are queued, only dispatch one of them at a time
If two start jobs for two seperate .swap device nodes are queued, which
then turns out to be referring to the same device node, refuse
dispatching more than one of them at the same time.
This should solve an issue when the same swap partition is found via GPT
auto-discovery and via /etc/fstab, where one uses a symlink path, and
the other the raw devce node. So far we might have ended up invoking
mkswap on the same node at the very same time with the two device node
names.
With this change only one mkswap should be executed at a time. THis
mkswap should have immediate effect on the other swap unit, due to the
state in /proc/swaps changing, and thus suppressing actual invocation of
the second mkswap.