systemd user instance assumed same controllers are available to it as to
PID 1. That is not true generally, in v1 (legacy, hybrid) we don't delegate any
controllers to anyone and in v2 (unified) we may delegate only subset of
controllers.
The user instance would fail silently when the controller cgroup cannot
be created or the controller cannot be enabled on the unified hierarchy.
The changes in 7b63961415 ("cgroup: Swap cgroup v1 deletion and
migration") caused some attempts of operating on non-delegated
controllers to be logged.
Make the user instance first check what controllers are availble to it
and narrow operations only to these controllers. The original checks are
kept in place.
Note that daemon-reexec needs to be invoked in order to update the set
of unabled controllers after a change.
Fixes: #18047Fixes: #17862
The function controller_is_accessible() doesn't do really much in case
of the unified hierarchy. Move common parts into cg_get_path_and_check
and make controller check v1 specific. This is refactoring only.
Docker doesn't set $container, so it cannot be detected that way. Instead, we
check for presence of /.dockerinit, which it creates. Podman does set
$container, but some Red Hat images (in particular, Fedora images) override
$container to equal "oci". So to correctly detect Podman containers, we check
for presence of /run/.containerenv, which is created by Podman and is now the
official way to get information about the container from within the container.
Fixes https://github.com/systemd/systemd/issues/15393
We'd return -1 (-EPERM), even though we have a general rule to use real errno
values. The particular case that caught my attention was:
$ sudo udevadm control -l asdf
Failed to parse log priority 'asdf': Operation not permitted
... but "git grep 'r =.*_from_string' src/" return 110 hits. Confusingly, some
of the _from_string functions already return a proper errno value, so not all
of those are broken, but probably quite a few.
As suggested in https://github.com/systemd/systemd/pull/11484#issuecomment-775288617.
This does not touch anything exposed in src/systemd. Changing the defines there
would be a compatibility break.
Note that tests are broken after this commit. They will be fixed in the next one.
There may be situations where a cgroup should be protected from killing
or deprioritized as a candidate. In FB oomd xattrs are used to bias oomd
away from supervisor cgroups and towards worker cgroups in container
tasks. On desktops this can be used to protect important units with
unpredictable resource consumption.
The patch allows systemd-oomd to understand 2 xattrs:
"user.oomd_avoid" and "user.oomd_omit". If systemd-oomd sees these
xattrs set to 1 on a candidate cgroup (i.e. while attempting to kill something)
AND the cgroup is owned by root, it will either deprioritize the cgroup as
a candidate (avoid) or remove it completely as a candidate (omit).
Usage is restricted to root owned cgroups to prevent situations where an
unprivileged user can set their own cgroups lower in the kill priority than
another user's (and prevent them from omitting their units from
systemd-oomd killing).
sd-boot has a copy of a subset of codes from libbasic. This makes
sd-boot share the code with libbasic, and dedup the code.
Note, startswith_no_case() is dropped from sd-boot, as
- it is not used,
- the previous implementation is not correct,
- gnu-efi does not have StrniCmp() or so.
Let's tighten the logic behind path_extract_filename() a bit: first of
all, refuse all cases of invalid paths with -EINVAL. More importantly
though return a recognizable error when a valid path is specified that
does not contain any filename. Specifically, "/" will now result in
-EADDRNOTAVAIL.
This changes API, but none of the existing callers care about the return
value, hence the change should be fine.
CLI tools may be used in a script. E.g., a script for monitoring a
service may use `systemctl`. Previously, if the kernel command line has
e.g. systemd.log-level=debug, then systemctl in the script produces
debugging logs when the script is invoked by a .service unit, but does
not when the script is running in a terminal. Then,
https://github.com/systemd/systemd/pull/18281#discussion_r561697482,
> I expect users to be (negatively) surprised.
In the previous commit, $SYSTEMD_EXEC_PID= is introduced. Then, we can
now detect whether a command is directly invoked by systemd or through
a script. Let's skip reading the kernel command line when a command is
invoked through a script.
This effectively reverts 41d1f469cf.
Before this, e.g., `networkctl reload` invoked by `systemctl reload systemd-networkd.service`
does not produce debugging logs even if systemd.log-level=debug is set. This fixes
the issue.
The target is update-syscall-tables, so let's call the script
update-syscall-tables.sh to reduce the cognitive overhead when
trying to find the right file.
Upstream uses .text, but this is rather unusual. Let's use .txt as the usual
suffix for text files. This tells various editors and such that the file should
be treated as plain text. I also want to a script to summarize license status,
and having an easy-to-recognize suffix makes this easier.
Even though many of those scripts are very simple, it is easier to include
the header than to try to say whether each of those files is trivial enough
not to require one.
If the file was always generated on the fly, the header would be pointless.
But since we distribute it, it should be there. C.f.
a0e150b2f4.
This was forgotten in 35b42e5600.
Follow-up for ece852c845.
This addresses the following comments by the Lennart:
---
hmm, so this now does two access() calls for the case where the fd is
not requested, and opens things up for races (theoretically, …). now,
the access() code path was in place for optimization, but if an optimization
is less sexy than the original (and i think it is less sexy, since more
than one syscall, and non-atomic), i think we shouldn't do the optimization.
maybe we should just always use open(O_PATH) now, and then fstat() it to
check if regular file, and then access_fd() it for checking if its executable.
Lennart Poettering
Yu Watanabe
Michal Koutný
Luca Boccassi
Zbigniew Jędrzejewski-Szmek
Sergey Bugaev
Anita Zhang
Deepak Rawat
Susant Sahani
Jiehong