At almost all places if diskseq is not supported we encode this as
diskseq zero. But in two places we got the check for that wrong,
assuming it was UINT64_MAX.
Fix that.
This allows us to install everything in the same dnf command instead
of having to use a prepare script to run dnf from within the image.
This is a hack until mkosi supports release specific dropin files.
Let's make sure we're testing unprivileged builds properly. Usage
of SourceFileTransfer= and SourceFileTransferFinal= are removed as
they were dropped by mkosi. SourceFileTransfer=mount is now the
default in mkosi so behavior for the build script is unchanged. We
stop copying sources in the final image until mkosi adds support
for virtiofs.
This is just a workaround. Once we drop gnu-efi, the arm build system
for EFI binaries should be changed to use the arm-none-eabi toolchain,
which should not exhibit this behavior.
Follow-up for 4453ebe4db.
With the feature list all dandified, the most important part of the
output, i.e. the project name and version, are less visible.
These units are also present in the initrd, so instead of an assert,
just use a condition so they are skipped where they need to be skipped.
Fixes https://github.com/systemd/systemd/issues/26358
This does what 404803e6ca did for the
monotonic timestamp getter, but for the realtime timestamp.
It also also makes the return value optional, exactly as for the
monotonic timestamp logic.
If a service defines Delegate=yes, its subcgroup won't inherit the
LogFilterPatterns= option, because the option is stored on the unit's
cgroup attributes, not on the subcgroup.
Fixed by using the unit's cgroup attributes instead.
From a given cgroup path, cg_path_get_unit() allows to retrieve the
unit's name. Although, this removes the path to the unit's cgroup,
preventing the result to be used to fetch xattrs.
Introduce cg_path_get_unit_path() which provides the path to the unit's
cgroup. This function behave similarly to cg_path_get_unit() (checking
the validity and escaping the unit's name).
If we fail to get the necessary information, let's just not print that
part of the message. 'n/a' looks pretty ugly.
I used a bunch of ternary operators instead of seperate log lines because
with two components that might or might not be there, we need four different
combinations.
Also, the unit name doesn't need to be quoted, it's always printable.
We would print "Current command vanished from the unit file, execution of
the command list won't be resumed." as a warning, but most of the time there
is nothing to resume, because a unit has just one command. So let's detect
the case where the command that was active is the last command in the sequence
and skip the warning.
I was considering how to store the information that the command is last. An
important consideration is not to use a format that would confuse older versions
of systemd. (It wouldn't be a big problem if older systemd just refused the
new serialization, since we require systemd to be newer, but we should avoid
the case where the deserialization is "successful", but actually incorrect.)
Similarly, the deserialization from the old systemd must not confuse new systemd.
For this command, we have a list of arguments at the end, so just adding a
new field either in the middle or at the end is problematic because it's hard
to ensure that we don't mix up the positional and variable arguments.
We actually need to store just one bit of information, so '+' is prefixed on
the index of the last command and used by new systemd to skip the warning.
When deserializing from older systemd, '+' is not present, so we detect all
commands as "not last", and still emit the warning, so we err on the side of
caution. If the user were to deserialize from newer to older systemd, nothing
untoward would happen, because the '+' is ignored. (Users shouldn't do this,
but we know that this occasionally happens with initrds or exitrds and package
downgrades.)
Let's ensure that entry seqnums remain stable and monotonic across the
entire runtime of the system, even if local storage is turned off. Let's
do this by maintainer a counter file in /run/ which we mmap() and
wherein we maintain the counter from early-boot on till late shutdown.
This takes inspiration of the kernel-seqnum file we already maintain
like that that tracks which kmsg messages we already processed. In fact,
we reuse the same code for maintaining it.
This should allow the behaviour entry seqnums to be more predictable, in
particular when journal local storage is turned off. Previously, we'd
maintain the seqnum simply by always bumping it to the maximum of the
last written entry seqnum plus one, and the biggest seqnum so far
written to the journal file on disk. If we'd never write a file on disk,
or if no journal file was existing during the initrd→seqnum transition
we'd completely lose the current seqnum position during daemon restarts
(such as the one happening during the switch-root operation).
This also will cause a journal file rotation whenever we try to write to
a journal file with multiple sequence number IDs, so that we know that
from early boot trhough the entire runtime we'll have stable sequence
numbers that do not jump, and thus can be used to determine "lost"
messages.
