Jan Janssen [Tue, 4 Jan 2022 12:00:37 +0000 (13:00 +0100)]
boot: Change boot entry sorting
There are a few undesirable properties to how boot entries are
currently sorted.
First, it sorts by entry file name only, which may not correspond
to the title that is shown (for exmaple because it is prefixed by
machine-id). The file ending will also create unexpected ordering
("arch-lts.conf" would come before "arch.conf").
While the list is sorted alphabetically ascending, it is also
lower version/priority first, which is unintuitive. In particular,
a boot-counted entry that is bad (0 tries left) will be at the very
top.
Additionally, the Windows and Mac loaders should be sorted with
the rest of the loaders.
The logs used the service name as the primary log key. But the service name
often needs to contain escape symbols, and the logs are rather hard to read
because of this. Thus the logs are changed to use the path to the source
desktop file. I think this is much more useful because the user will want to
look at the source file too and maybe change it if something goes wrong.
A bit more logging to show which directories we are looking at and why we
skip certain units is added too.
$ rm -rf /tmp/out && mkdir /tmp/out && SYSTEMD_LOG_LEVEL=debug SYSTEMD_LOG_TARGET=console build/systemd-xdg-autostart-generator /tmp/{out,out,out}
Scanning autostart directory "/home/zbyszek/.config/autostart"…
Scanning autostart directory "/etc/xdg/autostart"…
/etc/xdg/autostart/tracker-miner-rss-3.desktop: not generating unit, marked as skipped by generator.
/etc/xdg/autostart/gnome-initial-setup-first-login.desktop: ExecCondition executable gnome-systemd-autostart-condition not found, unit will not be started automatically: No such file or directory
/etc/xdg/autostart/geoclue-demo-agent.desktop: symlinking app-geoclue\x2ddemo\x2dagent@autostart.service in xdg-desktop-autostart.target/.wants…
SELinux enabled state cached to: disabled
Directory "/tmp" already exists, but has mode 0777 that is too permissive (0755 was requested), refusing.
/etc/xdg/autostart/polkit-mate-authentication-agent-1.desktop: symlinking app-polkit\x2dmate\x2dauthentication\x2dagent\x2d1@autostart.service in xdg-desktop-autostart.target/.wants…
/etc/xdg/autostart/mate-settings-daemon.desktop: symlinking app-mate\x2dsettings\x2ddaemon@autostart.service in xdg-desktop-autostart.target/.wants…
/etc/xdg/autostart/user-dirs-update-gtk.desktop: symlinking app-user\x2ddirs\x2dupdate\x2dgtk@autostart.service in xdg-desktop-autostart.target/.wants…
/etc/xdg/autostart/org.freedesktop.problems.applet.desktop: symlinking app-org.freedesktop.problems.applet@autostart.service in xdg-desktop-autostart.target/.wants…
/etc/xdg/autostart/org.gnome.SettingsDaemon.Datetime.desktop: not generating unit, startup phases are not supported.
/etc/xdg/autostart/org.gnome.SettingsDaemon.XSettings.desktop: not generating unit, startup phases are not supported.
/etc/xdg/autostart/org.gnome.SettingsDaemon.DiskUtilityNotify.desktop: symlinking app-org.gnome.SettingsDaemon.DiskUtilityNotify@autostart.service in xdg-desktop-autostart.target/.wants…
/etc/xdg/autostart/gnome-initial-setup-copy-worker.desktop: not generating unit, startup phases are not supported.
/etc/xdg/autostart/org.gnome.Evolution-alarm-notify.desktop: symlinking app-org.gnome.Evolution\x2dalarm\x2dnotify@autostart.service in xdg-desktop-autostart.target/.wants…
/etc/xdg/autostart/tracker-miner-fs-3.desktop: not generating unit, marked as skipped by generator.
/etc/xdg/autostart/orca-autostart.desktop: ExecCondition executable gnome-systemd-autostart-condition not found, unit will not be started automatically: No such file or directory
...
Inspired by https://bugzilla.redhat.com/show_bug.cgi?id=2038750.
The return value from xdg_autostart_service_generate_unit() is ignored by the
caller, so we can do a shortcut return without functional change. This is nicer
because we're now consistently always returning an error if something failed.
meson: no longer skip dependencies when fuzzers are built locally
to make it easier to fuzz code that uses external libraries like libelf/libdw.
The dependencies are skipped on OSS-Fuzz because they aren't available
at runtime if they aren't linked statically. This restriction can safely
be lifted when the fuzzers are built locally with all the dependencies
installed. As far as I know there is at least one fuzz target in the systemd
repository that can benefit from this: https://github.com/systemd/systemd/issues/11018
Julia Kartseva [Fri, 7 Jan 2022 23:02:57 +0000 (15:02 -0800)]
bpf: check if lsm link ptr is libbpf error
BPF_RAW_TRACEPOINT_OPEN is expected to work only on x86 and x86_64,
since BPF trampoline is implemented only on these architectures.
Attach probing by bpf_program__attach_lsm already happens in
`bpf_lsm_supported`. The resulting pointer can store libbpf error and
that is the case for unsupported architectures.
Add libbpf error check to `bpf_lsm_supported` so execution does not
reach the point where unit startup fails.
In the olden days systemd-resolved used dbus and it didn't make sense to start
it before dbus which is started fairly late. But we have mostly ported resolved
over to varlink. The queries from nss-resolve are done using varlink, so name
resolution can work without dbus. resolvectl still uses dbus, so e.g. 'resolvectl
query' will not work, but by starting systemd-resolved earlier we're not making this
any worse.
If systemd-resolved is started after dbus, it registers the name and everything
is fine. If it is started before dbus, it'll watch for the dbus socket and
connect later. So it should be fine to start systemd-resolved earlier. (If dbus
is stopped and restarted, unfortunately systemd-resolved does not reconnect.
This seems to be a small bug: since our daemons know how to watch for
dbus.socket, they could restart the watch if they ever lose the connection. But
this scenario shouldn't happen in normal boot, and restarting dbus is not
supported anyway.)
Moving the start earlier the following advantages:
- name resolution becomes availabe earlier, in particular for synthesized
hostnames even before the network is up.
- basic.target is part of initrd.target, so systemd-resolved will get started
in the initrd if installed. This is required for nfs-root when the server is
specified using a name (https://bugzilla.redhat.com/show_bug.cgi?id=2037311).
bpf: actually skip RestrictFileSystems= when not supported
Units would fail to start, incl. systemd-journald.service and systemd-udevd.service.
Since unit->manager->restrict_fs will be set if and only if we can use it,
we can just check for that and remove the other checks.
Follow-up for 299d9417238e0727a48ebaabb5a9de0c908ec5c8.
Luca Boccassi [Sun, 9 Jan 2022 14:00:25 +0000 (14:00 +0000)]
test: store empty files rather than symlinks for test-fstab-generator
Dangling symlinks get pruned when packaging up the installation
directory. Just store empty files instead, and compare the names
rather than the content for .requires/.wants - the filename is
what is important anyway, the content is ignored.
> [It is] called in the setup of ld-linux-x86-64.so.2 from _dl_sysdep_start.
> My local call stack (with LTO):
>
> #0 init_cpu_features.constprop.0 (/usr/lib64/ld-linux-x86-64.so.2)
> #1 _dl_sysdep_start (/usr/lib64/ld-linux-x86-64.so.2)
> #2 _dl_start (/usr/lib64/ld-linux-x86-64.so.2)
> #3 _start (/usr/lib64/ld-linux-x86-64.so.2)
>
> Looking through the source, I think it's this (links for glibc 2.34):
> - First dl_platform_init calls _dl_x86_init_cpu_features, a wrapper for init_cpu_features.
> - Then init_cpu_features calls get_cet_status.
> - At last, get_cet_status invokes arch_prctl.
Adam Williamson [Wed, 5 Jan 2022 22:07:14 +0000 (14:07 -0800)]
kernel-install: prefer /boot over /boot/efi for $BOOT_ROOT
This restores the preference order from before 9e82a74. The code
previous to that change 'preferred' /boot over /boot/efi; that
commit changed it to check /boot/efi before checking /boot.
Changing this precedence could (and did, for me) have unexpected
effects - it seems safer to leave it how it was.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Markus Weippert [Tue, 4 Jan 2022 12:56:11 +0000 (13:56 +0100)]
homed: stop before stopping dbus
Otherwise, systemd-homed-active.service will fail to deactivate all
homes because homectl can no longer talk to homed if dbus stops first.
As a result, /home cannot be umounted.
Doing this on systemd-homed-active.service instead works as well, but
systemd-homed will exit 1 if dbus is already shut down.
Julia Kartseva [Thu, 6 Jan 2022 00:34:56 +0000 (16:34 -0800)]
bpf: do not freeze if bpf lsm fails to set up
BPF LSM is cgroup unaware and it's set up is happening in core manager.
It occures that the current implementation is too restrictive and causes
pid 1 to freeze.
Instead:
* in bpf_lsm_setup set manager->restrict_fs pointer last,
so it is an indicator that the set up was successful
* check for manager->restrict_fs before applying unit options
If the action failed, we should log about the issue, and continue.
Exiting would bring the graphical session down, which of course is not
appreciated by users.
As documented in previous commits, a non-negative return from the callback
doesn't matter, so the callback is simplified a bit.
It was copy-pasted directly from OSS-Fuzz where it makes sense to
kind of strip binaries to get nice backtraces but when the fuzzers
are built and run locally with gdb it would be nice to have a little
bit more than that.
It was initially discovered in elfutils where I put the same flags
and was surprised when I couldn't run the fuzzer comfortably step
by step, which led to the same change there: https://github.com/google/oss-fuzz/pull/7092
:-)
network: move logging from tc .fill_message to the callers
Structured initialization is used a bit more.
There were two kinds of log messages: about failed size calculations and
about failed appends to the message. I "downgraded" the first type to log_debug,
and moved the latter to the caller. This way there should be at most one high-priority
message.
I also changed sizeof(<type>) to sizeof(var) — there is less chance of select-and-paste
error in the second form.
network: simplify logging in l2tp_create_session() and l2tp_create_tunnel()
All the detailed logging is replaced by a simple "Failed to create netlink message",
which should be enough for the user in the unlikely case that this ever fails.
network: replace detailed netlink append messages with a single generic message
This commit is the first in the series, and they generally follow the same
idea: we had very detailed logging for message append operations which would
only fail either with some type error or intrinsic limit (and then they would
fail everywhere, so this would be noticed during development or in CI), or they
would fail with ENOMEM, in which case the exact location is not very interesting
since this is not repeatable.
I am in general in favour of detailed logging messages, because it helps with
diagnosis of errors, but I think case is an exception. Despite not being very
useful, those messages required a lot of effort, because they were customized
for each and every append operation. In fact some of the messages contained copy
errors. The text of the messages (since they are generally unique) also added up
to a considerable size.
This removes the log messages after each sd_netlink_message_append_*() in
fill_message_create() with a single line in netdev_create(). As described
above, we are just appending fields to a message, so those calls would almost
never fail.
A forgotten 'return' was added in one place.
$ size build/systemd-networkd{.0,}
text data bss dec hex filename 1878634 394016 36 2272686 22adae build/systemd-networkd.0 1842450 394080 36 2236566 222096 build/systemd-networkd
random-seed: hash together old seed and new seed before writing out file
If we're consuming an on-disk seed, we usually write out a new one after
consuming it. In that case, we might be at early boot and the randomness
could be rather poor, and the kernel doesn't guarantee that it'll use
the new randomness right away for us. In order to prevent the new
entropy from getting any worse, hash together the old seed and the new
seed, and replace the final bytes of the new seed with the hash output.
This way, entropy strictly increases and never regresses.
Fixes: https://github.com/systemd/systemd/issues/21983 Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
projects / thirdparty / systemd.git / log
