When attaching a portable service with extensions, immediately validate
that the os-release and extension-release metadata values match, rather
than letting it fail when the units are started
In some cases it's useful to explicitly generate the JSON field names to
generate for table columns, instead of auto-mangling them from table
header names that are intended for human consumption.
This adds the infra and a test for it.
It's intended to be used by #20544, for the first column, which in text
mode should have an empty header field, but have an explicit name in
json output mode.
With this change, "mkosi build" will automatically build systemd for the
current distro without any further configuration. If people want to do a
cross-distro build by default, they can still create mkosi.default, but I
assume that this is relatively rare.
If people have symlinked mkosi.default to one of the files in .mkosi/, they'll
need to adjust the symlink.
(Building without configuration would always fail, since systemd has many many
required dependencies. I think it's nicer to do the most commonly expected
thing by default, i.e. rebuild for the current distro.)
Mkosi is nowadays packaged for most distros, so recommend installing of distro
packages as the primary installation mechanism.
The pattern was added in 6242cda99d, with the
idea that users will have local configuration files for mkosi and git should
not bother them about those. But let's make this narrower, and only match
files with "local". This way we reduce the risk that some unrelated file
will be ignored by accident.
.gitignore in the parent directory is used, because mkosi apparently tries
to load all files under mkosi.default.d/, without looking at the extension.
This is probably something to fix in mkosi too.
The comments were outdated: at least "nfsnobody" is not used in Fedora since a
few years. So I hope we don't need this anymore. The meson build scripts do
autodetection on their own.
journal_file_verify() uses a set of tmpfs files to create lists
of object positions by type.
The existing code used a bare write() call for every object
position written, incurring a syscall per listed object.
This commit encapsulates the bare file descriptors in FILE *'s
and replaces the bare write with fwrite, buffering the writes so
there's less syscalls.
Cached `journalctl --verify` tests showed a ~8% faster runtime
with this change on a release build, verifying 1.3GiB of
production journals across 16 files.
Previously, the devlink was created based on the priority saved in udev
database. So, we needed to reevaluate devlinks after database is saved.
But now the priority is stored in the symlink under /run/udev/links, and
the loop of devlink creation is controlled with the timestamp of the
directory. So, the double evaluation is not necessary anymore.
By the previous commit, it is not necessary to distinguish if the devlink
already exists. Also, I cannot find any significant advantages of the
previous complecated logic, that is, first try to create directly, and then
fallback to atomically creation. Moreover, such logic increases the chance
of conflicts between multiple udev workers.
This makes devlinks always created atomically. Hopefully, this reduces the
conflicts between the workers.
During creating a symlink to a device node, if another device node which
requests the same symlink is added/removed, `stat_inode_unmodified()`
should always detects that. We do not need to continue the loop
unconditionally.
Previously, we only store device IDs in /run/udev/links, and when
creating/removing device node symlink, we create sd_device object
corresponds to the IDs and read device node and priority from the
object. That requires parsing uevent and udev database files.
This makes link_find_prioritized() get the most prioritzed device node
without parsing the files.
It expects a generic "struct sockaddr", not a "struct sockaddr_un".
Pass the right member of the union.
Not sure why gcc/llvm never complained about this...
When pulling in the SHA256 implementation from glibc, only some of the
coding style was adjusted to ours, other was not. Let's make things a
bit more consistent.
The GUIDs we usually deal with should be considered constant. Hence make
them so. Unfortunately the prototypes for various functions doesn't mark
them as const (but still decorates them with "IN", clarifying they are
input-only), hence we need to cast things at various places. We already
cast in similar fashion in many other cases, hence unify things here in
one style.
Making the EFI_GUID constant (and in particular so when specified in C99
compound literal style) allows compilers to merge multiple instances of
them.
The automatic logic can't always find the original root partition (ex:
if the rootfs is copied to a ext4 fs backed by zram in the initramfs),
so we want to support "/run/systemd/volatile-root" which is a symlink to
the original root partition.
Fix#20610
As we usually do, remove old unix socket before bind.
Note that systemd-udevd-control.socket has `RemoveOnStop=yes`.
So, it is not necessary to be removed when the service exits.
Replaces #20584 and #20588.
Luca Boccassi
lainahai
saikat0511
Yu Watanabe
Dan Streetman
Zbigniew Jędrzejewski-Szmek
Lennart Poettering
Jan Janssen
Vito Caputo
Kristian Klausen
Franck Bui