Let's keep the comments together with the lines they describe, this is easier
to read and probably more maintainable in the long run.
Also, () → {} for idiomaticity.
Users may use rules that refer to binaries e.g. in /opt or /usr/local,
and those directories may be separate mount points. We don't need the
binfmt rules in early boot, so let's delay the service so that we can
rely on the full local filesystem being visible.
Fixes#21178.
This new helper converts libc style syscall return values into
systemd-kernel (actually: kernel style) negative errno values.
It's implemented as macro-like inline function, and propagates return
values >= 0 as themselves and returns -errno for negative error returns.
THis is supposed to be little more than syntactic sugar so that we can
reduce a lot of (short, but still) boilerplate code whever we convert
libc style error handling into our own.
As discussed here:
https://github.com/systemd/systemd/pull/21326#discussion_r748413537
There are two ways to deactivate a home dir: the official way if someone
calls "homectl deactivate" or something similar. Or the
error-case/side-effect code path, where we unwind exactly what we
already set up.
The .undo_dm field is supposed to tell us whether we shall still detach
the DM device. We forgot to update it in the official code path. Add
that. (We did it correctly in the error case codepath already).
(The issue is minor: if we forget to reset it in the clean code path,
then we'll try to unmount again what is already unmounted in the error
code paths, but since we ignore errors there — to not "stack" errors –
we never noticed and there's no bad effect of it)
smb3 is an alias for modern CIFS, not the old SMB fs (see kmod alias list, i.e. modinfo fs-smb3).
(the old smbfs has long been removed from the kernel actually, it's dead
and obsolete)
There are a bunch of filesystems that actually are just aliases for
other filesystems. So far we listed them as "ungrouped", suggesting they
should be added to some file system group. But that's not really
something needed, since they are after all not file systems in their own
right, but aliases only (and usually legacy at that).
hence, let's hide them from display (but debug log about them)
devtmpfs and cpuset are not actual filesystems of their own. cpuset used
to be but is now an alias for cgroupsfs. devtmpfs is the same as tmpfs
as its just a "named superblock", i.e. a specific instance of tmpfs, but
not a file system of its own.
This group shall cover file systems whose job is to make stuff that
isn't actually storing anything in itself, and isn't just an API file
system, but allows exposing stuff with special semantics in the VFS.
1. @anonymous → for file systems that aren't real file systems but
simply the backing for sockets, pipes and other "anonymous" fds.
2. @security → for the various MAC security file systems.
After reading https://simonbyrne.github.io/notes/fastmath/ I think we
should drop -ffast-math. The JSON code actually looks for NaN, so the
fact it becomes unreliable kinda sucks.
Moreover, we don't do any number crunching. We use floating point fields
only sporadical for trivial math. Hence the optimization is entirely
unnecessary.
Let's first move the home dir to a new mount point that is only visible
in our own private namespace. Then, do FITRIM and stuff there, so that
we know the regular userspace can't interfere with that, and we know
that the home fs is not used anymore.
(This will become even more important once we add auto-grow/auto-shrink
for home dirs)
Both linux kernel kernel and iproute2 uses int32 type for a link group
attribute and -1 has a special meaning, so setting it to 4294967295
would make it -1 in the linux kernel (and ip link cmd).
When networkd detects an unmanaged link, then the state is changed in
the following order:
pending -> initialized -> unmanaged
The "initialized" state was added by bd08ce5615.
Zbigniew Jędrzejewski-Szmek
Lennart Poettering
Yu Watanabe
Evgeny Vereshchagin
dependabot[bot]
Slava Bacherikov
Luca Boccassi
Taiki Sugawara