This replaces #36209 and implements this via a new builtin, rather than
by shelling out to systemd-dissect.
Replaces: #36209
Replaces: #28894Fixes: #34319Fixes: #33453Fixes: #27897Fixes: #18035
If we cannot find the callout we need in the build dir let's look for it
in $PATH as last resort.
This makes invoke_callout_binary() usable for all binaries we install
into $PATH (as opposed to /usr/lib/systemd), but has no effect
on callout binaries specified with full path.
This is useful, since we soon want to invoke journalctl as a callout.
We generally let gpt-auto-generator write to the "late" generator dir,
so that the explicit fstab-generator can write to the "middle" one. And
I think we should leave it that way. But we also want to override the
generic systemd-fsck-root.service service potentially, and if we'd do
that in the "late" generator dir, it would be pointless, since that's
ordered *after* the static version hence would never be taken into
consdiration.
Hence clean this up: keep writing to the late dir for everything,
except for the fsck stuff.
Currently, there's no nice way to get a key into the dm-verity kernel
keyring unless recompiling the kernel, or enabling SB or buying into
shim. Neither sounds particularly attractive.
hence provide a reasonable fallback: if unlocking with signed roothash
doesn#t work, just try without. maybe the kernel policy allows this,
maybe not. It's worth a try.
DDIs may contain multiple versions of the same OS, or even multiple
OSes. Hence it makes sense to not just pick the "newest", whatever that
might be, but only partitions associated with specific images, or in a
specific version.
Let's a concept for such filtering: a per-designator glob expression
that can be applied to the partition label string, and can be used for
such filtering.
Usecase: when picking UKI belonging to OS image X in version Y, make
sure we only pick a /usr/ partition belonging to X in version Y, and a
root and home partition belonging to X in any version.
This only adds the basic infrastructure, but doesn't actually expose it
anywhere.
So far, we did not support auto-discovery the /usr/ partition at boot.
Change that: on explicit request (i.e. mount.usr=dissect) automatically
discover the /usr/ partition based on our usual dissection logic.
This is useful when booting into storage target mode, where we do not
want to enter a root fs, we just want to stay forever in the initrd,
and hence there's value in not even generating and jobs to find the
rootfs
This introduces root=dissect, which is a lot like root=gpt-auto, but
uses the image dissection logic to find the root partition. It's also
based on the GPT efi var telling us about the root disk, but instead of
just picking some suitable root disk manually it uses the dissection
logic.
This is a big step forward, since it means we can actually make
use of the image policy logic for the root fs too.
Previously if we found a verity signature partition in an image, and the
image policy required "verity" (but did not allow "signature") we'd
refuse the image. This is of course unnecessarily strict: if "verity" is
allowed, we can make use of the verity data, and ignore the signature
data.
hence, relax the rules here: when we pick up a partition and want to
test it against the policy, always consider all "weaker" uses too, maybe
they are allowed if the "stronger" users isn't.
When dissecting an image, let's make use of the Verity data even if we
got told no root hash explicitly: we can simply determine it by
concatenating the data partition uuid with the verity partition uuid.
Of course, on first thought this doesn't really add much: if the root
hash is not pinned from somewhere, this does not guarantee trust in
the image.
However, this is very useful for attestation: if we have the root hash
we can measure it before mounting things, even if we don't actually
authenticate it.
Hence, at best this helps us with attestation, at worst it doesn't improve
security but certainly doesn't hurt it.
Previously, if during early enumeration of the partition table we
figured out we have a verity + verity-sig partition, we determined we are
"ready" to do verity/verity-sig, as reported by the "verity_ready" and
"verity_ready_sig" fields of DissectedImage.
Let's rework this: only determine we are ready in case the VeritySettings
structure is fully populated. Mark verity-sig/verity only as ready once
we actually *load* the verity sig metadata from the special partition.
This is conceptually more correct, as we consider things "ready" only if
we actually have all data for it loaded. It's also preparation for a
later commit that guesses the verity root hash based on what we discover.
So far the gpt-auto symlinks would point to:
1. the disk the ESP/XBOOTLDR is located on
2. or to a loopback device whose filename field is set to "rootdisk"
or "rootdisk.raw"
This makes sense in the initrd. But once we transition to the host this
is quite confusing, since the symlinks might point to a different place
than what we actually ended up transitioning too: the actual backing
device of the root file system might be different from what gpt-auto
found.
Let's clean this up: let's avoid any ambiguities here: let's extend the
rules above with one more rule:
3. if we left the initrd, we'll make gpt-auto point to the selected
root file system, if it otherwise would have been a candidate.
Or in other words, the ID_PART_GPT_AUTO_ROOT_DISK=1 udev property now
always makes sense: in the initrd it points to the future root disk, and
on the host to the actual root disk.
Fixes: #34319
So far the gpt-auto symlinks would point to:
1. the disk the ESP/XBOOTLDR is located on
2. or to a loopback device whose filename field is set to "rootdisk"
or "rootdisk.raw"
This makes sense in the initrd. But once we transition to the host this
is quite confusing, since the symlinks might point to a different place
than what we actually ended up transitioning too: the actual backing
device of the root file system might be different from what gpt-auto
found.
Let's clean this up: let's avoid any ambiguities here: let's extend the
rules above with one more rule:
3. if we left the initrd, we'll make gpt-auto point to the selected
root file system, if it otherwise would have been a candidate.
Or in other words, the ID_PART_GPT_AUTO_ROOT_DISK=1 udev property now
always makes sense: in the initrd it points to the future root disk, and
on the host to the actual root disk.
Fixes: #34319
This replaces home-grown notification method from worker processes to
manager process with sd-notify. Then, we can reuse many library
functions commonly used by many components.
Hopefully, this does not change any effective functionalities.
Lennart Poettering
Yu Watanabe
ZIHCO
victor-ok