Simon McVittie [Mon, 30 Jul 2018 15:51:01 +0000 (16:51 +0100)]
Skip test-pull-repeated during CI
This test is expected to fail a small proportion of the time. During
the build of ostree 2018.7-1 in Debian, it seems we were unlucky on
s390x. Non-deterministic tests are also problematic for autopkgtest,
where they can gate migration of our dependencies like GLib, so skip
this test unless the caller has opted-in to non-deterministic tests.
It would be appropriate to enable this test in environments where
failures can easily be retried and are not disruptive to other
packages.
Signed-off-by: Simon McVittie <smcv@debian.org>
Gbp-Pq: Topic debian
Gbp-Pq: Name Skip-test-pull-repeated-during-CI.patch
Colin Walters [Fri, 6 Dec 2024 19:57:19 +0000 (14:57 -0500)]
man: Note semantics combining `root.transient` with `composefs.enabled`
It's all quite confusing having to reason about both the pre-composefs
ostree and the composefs version. But hopefully soon we more firmly
leave behind that first legacy.
Jonathan Lebon [Sat, 14 Dec 2024 13:47:21 +0000 (08:47 -0500)]
lib/deploy: error out if composefs enabled but unsupported
If composefs was explicitly requested (`enabled = true`) but libostree
was not compiled with composefs support, error out at deploy time. This
matches the logic in `ostree-prepare-root`.
Jonathan Lebon [Sat, 14 Dec 2024 13:45:19 +0000 (08:45 -0500)]
libostree/deploy: enable composefs by default
The composefs libostree integration has been supported for a while now
and is actively in use in various ostree/bootc-based systems. Let's
turn it on by default.
This has no effect if composefs support is not compiled in. Note also
that this does not change the default value of the `composefs.enabled`
tristate to `true`. The default is still `maybe`, but the deploy API
will now also create composefs images for `maybe`.
The reason for doing it this way is so that systems upgrading from
old libostree versions (which may either not have composefs support or
may have composefs-related bugs) will still be able to upgrade and not
trip `ostree-prepare-root` in the new deployment (which allows missing
composefs images for `maybe`).
We may in the future change the default value to `true`.
See also: https://github.com/ostreedev/ostree/issues/2867
Colin Walters [Wed, 27 Nov 2024 02:15:23 +0000 (21:15 -0500)]
core: Always sort incoming xattrs
When recomputing selinux attrs during commit, we weren't sorting,
which could cause various issues like fsck failures.
This is a big hammer; change things so we always canonicalize
(i.e. sort) the incoming xattrs when creating a file header
and directory metadata.
I think almost all places in the code were already keeping
things sorted, but it's better to ensure correctness first.
If we ever have some performance issue (I'm doubtful) we
could add something like `_ostree_file_header_known_canonicalized`
or so.
Closes: https://github.com/ostreedev/ostree/issues/3343 Signed-off-by: Colin Walters <walters@verbum.org>
Colin Walters [Mon, 4 Nov 2024 18:48:56 +0000 (13:48 -0500)]
prepare-root: Fix composefs docs
In practice in ostree-sysroot-deploy.c we only react to having
`composefs = yes`; the docs mention `maybe` but that never did
anything.
The value is wrong in the code too, but I'm not touching
that here to avoid conflating changes - the main thing to fix
is the docs because here `maybe == no`.
Colin Walters [Wed, 30 Oct 2024 14:07:26 +0000 (10:07 -0400)]
checkout: Only verify digest if repo requires fsverity
Fixes a regression from the previous commit; in
the case where the target repo doesn't have composefs in
signed mode there's no reason to verify the digest
at checkout time because we aren't verifying it at
boot time either.
The regression is in cases that use rpm-ostree e.g.
where as of recently we unconditionally add the composefs
digest, but for e.g. FCOS we aren't deploying with fsverity
enabled.
Closes: https://github.com/ostreedev/ostree/issues/3330 Signed-off-by: Colin Walters <walters@verbum.org>
Simon McVittie [Thu, 31 Oct 2024 10:54:13 +0000 (10:54 +0000)]
tests: Work around GPG 2.2.45 error behaviour when revoking an expired key
In GPG 2.2.45, a diagnostic message about the only trusted key having
already expired causes this import to produce exit status 2, but the
import still succeeds (the key is still revoked).
Colin Walters [Sun, 27 Oct 2024 14:20:29 +0000 (10:20 -0400)]
deploy: Don't recompute verity checksums if not enabled
This fixes a truly horrific performance bug when
composefs is enabled, but fsverity is not supported
by the filesystem. We'd fall back to doing *userspace*
checksumming of all files at deployment time which was absolutely
not expected or required.
There's really an immense amount of technical debt
here, such as the confusion between `ex-integity.composefs`
vs the prepare-root config, how we handle "torn" states
where some objects don't have verity enabled but some do,
etc.
The ostree composefs state has two modes:
- signed: We need to enforce fsverity
- unsigned: Best effort resilience
So we fix this by making the deploy path to make verity
"opportunistic" - if the ioctl gives us the data, then we
add it to the composefs.
However, this code path is also invoked when we're
computing the expected composefs digest to inject
as commit metadata, and *that* API must work regardless
of whether the target repo has fsverity enabled as
it may operate on a build server.
One lucky thing in all of this: When I went to add
the "checkout composefs" API I added a stub `GVariant`
for options extensibility, which we now use.
Colin Walters [Wed, 18 Sep 2024 17:41:59 +0000 (13:41 -0400)]
curl: Make socket callback during cleanup into no-op
Because curl_multi_cleanup may invoke callbacks, we effectively have
some circular references going on here. See discussion in
https://github.com/curl/curl/issues/14860
Basically what we do is the socket callback libcurl may invoke into a no-op when
we detect we're finalizing. The data structures are owned by this object and
not by the callbacks, and will be destroyed below. Note that
e.g. g_hash_table_unref() may itself invoke callbacks, which is where
some data is cleaned up.
Dan Nicholson [Sun, 15 Sep 2024 19:59:36 +0000 (13:59 -0600)]
workflow/docs: Fix deployments
A couple fixes to make PRs and non-PRs work correctly:
* In a conditional expression, `true` or `false` are returned unless you
terminate both sides in a ternary. That was causing 2 strings to be
suffixed with `false` instead of an empty string.
* For a PR, we do actually want to cancel in progress runs since there's
no danger of breaking an in progress deployment.
* For PRs, just use the same `github-pages-pr` name for the artifact.
The important part is that it's not called `github-pages` where an in
progress deployment could pick it up. Otherwise it can use the same
name all the time.
Colin Walters [Tue, 20 Aug 2024 20:18:54 +0000 (16:18 -0400)]
lib/traverse: Fix minor memory leak
I was trying to check something with `-fsanitize=address`
and it warned about this memory leak. It's...subtle, basically
we were leaking when the same commit was added to the hash table.
But unfortunately fixing that then complicates ownership
over the return value; what we really want to use here is
`g_hash_table_steal_all_keys` but RHEL 9.4 is still rocking
`glib2-2.68.4` so we can't use it.
(Rust would mean we wouldn't have leaked anything here in the
first place...)
Dan Nicholson [Fri, 30 Aug 2024 00:19:30 +0000 (18:19 -0600)]
switchroot: Stop making /sysroot mount private
Back in 2b8d586c5, /sysroot was changed to be a private mount so that
submounts of /var do not propagate back to the stateroot /var. That's
laudible, but it makes /sysroot different than every other shared mount
in the root namespace. In particular, it means that submounts of
/sysroot do not propagate into separate mount namespaces.
Rather than make /sysroot private, make /var a slave+shared mount so
that it receives mount events from /sysroot but not vice versa. That
achieves the same effect of preventing /var submount events from
propagating back to /sysroot while allowing /sysroot mount events to
propagate forward like every other system mount. See
mount_namespaces(7)[1] and the linux shared subtrees[2] documentation
for details on slave+shared mount propagation.
When /var is mounted in the initramfs, this is accomplished with
mount(2) syscalls. When /var is mounted after switching to the real
root, the mount propagation flags are applied as options in the
generated var.mount unit. This depends on a mount(8) feature that has
been present since util-linux 2.23. That's available in RHEL 7 and every
non-EOL Debian and Ubuntu release. Applying the propagation from
var.mount fixes a small race, too. Previously, if a /var submount was
added before /sysroot was made private, it would have propagated back
into /sysroot. That was possible since ostree-remount.service orders
itself after var.mount but not before any /var submounts.
Dan Nicholson [Fri, 30 Aug 2024 11:53:10 +0000 (05:53 -0600)]
tests: Add mount propagation test
This tests the current behavior of making /sysroot a private mount so
that submounts on /var do not propagate back to /sysroot. It also shows
how submounts of /sysroot do not propagate into separate mount
namespaces for the same reason.
Simon McVittie [Wed, 31 Aug 2022 08:44:06 +0000 (09:44 +0100)]
d/control, d/rules: Disable gjs-based tests on armel
It's looking as though mozjs128, and therefore the next gjs release,
will not support armel due to its lack of atomic instructions causing
relatively basic features to be disabled.
projects / ostree.git / log