The initial motivation for this is that the "staging" code currently
didn't rewrite the deployment refs, meaning that the staged commit
could be pruned.
Hence first, this new API ensures that deployments also
hold a strong ref to their commit, without relying on the magical
"deployment refs" that we inject. That has always been a weird
artifact of the strict layering separation between OstreeSysroot
and OstreeRepo.
I also plan to change rpm-ostree to start using this API to
hold references to base layers for client-side layering; it also
today generates various refs.
That said, if we still want to support multiple processes
writing to a single repo (as happens on EndlessOS today) we
still need to write refs; perhaps later we could add a concept
of "generators" or something that create refs based on whatever
logic?
Another minor thing this fixes is that we had a printf inside
the library; this propagates the pruned data to the higher level
which can log however it likes.
Closes: #1566
Approved by: jlebon
Prep for reworking how we do sysroot cleanup. We're going to
start doing more lowlevel pruning work there, and I wanted to avoid
duplicating the ref enumeration.
Closes: #1566
Approved by: jlebon
I feel like I'm drowning in a pile of experimental-but-almost-stable
features...
Anyways, since we made the feature opt-in in rpm-ostree in
https://github.com/projectatomic/rpm-ostree/pull/1352
let's mirror that a bit here with an environment variable so people
can play with it more easily.
The tests needed some tweaks; specifically we need to reload the
status fact after making changes. I'm still a bit uncertain
about the Ansible-as-tests.
But we add an upgrade test that uses the new environment variable.
Closes: #1583
Approved by: jlebon
For the same reason we do in the rpm-ostree tests. This also
made sure the test run worked when I was offline on a plane.
Closes: #1583
Approved by: jlebon
OK so I noticed that something was failing and we were missing
`set -xeuo pipefail` in our shells. That of course revealed
the ansible tests didn't actually work - my only defense
here is spending so much time fighting to get it through CI
and trying something new.
Anyways, to make the staged-deploy tests work we need a task
that actually uses `rpm-ostree override` rather than `usroverlay`.
Let's make this a bit saner and have a clean split between
tests that are "shell-script+usroverlay" and "ansible+override".
Closes: #1577
Approved by: jlebon
While we do a `syncfs()` plus `FIFREEZE/THAW` for `/boot`, that
only comes during deployment finalization.
The code here today generally assumes that if the file exists
it's been fully written. So let's do a `fdatasync()` before
we do the `rename()`.
This just came out of looking through the code while working
on deployment staging. In that scenario there's a much larger
window between when we copy the kernel/initramfs and when we
sync `/boot`.
Closes: #1571
Approved by: jlebon
This is easier to `git grep` etc. versus ad-hoc English. Although
we still have some English for the prepare_transaction/commit which
acquire/release in separate phases.
Closes: #1572
Approved by: jlebon
These are further fixes based on running more of the rpm-ostree
test suite.
When dropping the staged deployment, we do need to do the
"post operations" such as bumping the sysroot mtime, so that
clients know something changed. We also need to regenerate
the deployment refs. And of course do a sysroot reload.
Also, add a "base cleanup" after creating a staged deployment
which also regenerates the refs.
Closes: #1570
Approved by: jlebon
There's no reason to do this. I didn't actually hit this problem,
but it's a corner case that just occurred to me while working on
the code.
I think callers should be adapted to skip trying to use staging
if there's no booted deployment.
Closes: #1568
Approved by: jlebon
This was pointed out in a previous PR review; we don't have
a need for the separate variables. Prep for adding an API for
this.
Closes: #1568
Approved by: jlebon
Doing so can break rpm-ostree, which wants to own the cleanup process
to ensure its baselayer refs are generated.
Further, doing the cleanup at shutdown time adds latency. It's also
going to be generally unnecessary as we expect repo pruning to have
been done when writing the refs.
Closes: #1567
Approved by: jlebon
The code has been sitting around for a while but since I disabled
it by default, I doubt anyone is really using it or relying on it.
This patch and turns on locking by default, and also drops the
API which was only public in the experimental API builds.
Conceptually these are two distinct things, and we
may actually want to split up the patches.
I don't think this will break anyone, but it's hard to say for sure.
It's also going to be hard to find out until we actually release
I suspect...
But anyone who is broken should be able to add `locking=false` into
their repo config. On the flip side Endless has been shipping with
this enabled and it is reported to help.
The reason to drop the APIs: I'm a bit concerned about the interactions over time
between libostree's use of the API and any apps that start using it.
For example, if an app specifies a SHARED lock in their code, then
later internally we decide to temporarily grab an `EXCLUSIVE`, but the
app had a second thread/process that was `EXCLUSIVE` already, and
that process was waiting on the first bit of code, then we could
deadlock. I can't think of a real world situation where this would happen
yet though.
We are likely to in the future have say `fsck` take an external lock,
`checkout` grab a shared one, etc.
Closes: #1555
Approved by: jlebon
Today rpm-ostree has some code to run a "sanitycheck" on a deployment.
I had initially deleted that when adapting it to use the staging code,
but I realized it should work fine; we just won't see the merged
config, but that's OK.
When I readded that code it started crashing because we didn't
actually return the new deployment object. We'll gain some coverage
here as I'll land the code to have rpm-ostree use staging, then bump
the rpm-ostree tests here.
Closes: #1559
Approved by: jlebon
The fact that `ostree admin deploy` always itself loaded the
merge kargs masked a bug in the core. Let's change our tests
to not pass any kernel arguments to ensure we cover this.
The new logic in the CLI is a bit subtle, but if you read
carefully is a lot clearer I believe. Basically we have one
of a few "starting points" in the first section, which can
then be further augmented.
Closes: #1558
Approved by: jlebon
Testing out the staged API with rpm-ostree, ostree-prepare-root.service
in the initramfs was failing. Turned out that was because we didn't
have a `root=` kernel argument. Which was because we didn't have
any kernel arguments at all except `ostree=`.
That in turn was because we weren't loading the bootloader config
from the merge deployment.
The serialized deployment data holds the unique identity of
(osname, checksum, deployserial) - look for the real merge deployment
in our deployment list which has the bootloader arguments we need.
This issue was entirely masked by the `ostree admin deploy` command
which itself explicitly loads the merge deployment's kernel arguments
in every case - it never passes the `NULL` default down. A followup
patch will fix that.
Closes: #1558
Approved by: jlebon
Let's only print if the commit isn't already partial; this
addresses a spam of "marking commit partial" from fsck.
Closes: #1548
Approved by: cgwalters
This took a whole lot of experimentation. I hit upon the idea
of doing a `systemctl stop sshd` to avoid the situation where we
might ssh back into the system while it's in the process of shutting
down.
Ultimately the other fix is disabling `ControlMaster`; see
for example: https://github.com/ansible/ansible/issues/17935Closes: #1548
Approved by: cgwalters
Followup to: https://github.com/ostreedev/ostree/pull/1503
After starting some more work on on this in rpm-ostree, it is
actually simpler if the staged deployment just shows up in the list.
It's effectively opt-in today; down the line we may make it the default,
but I worry about breaking things that e.g. assume they can mutate
the deployment before rebooting and have `/etc` already merged.
There's not that many things in libostree that iterate over the deployment
list. The biggest change here is around the
`ostree_sysroot_write_deployments_with_options` API. I initially
tried hard to support a use case like "push a rollback" while retaining
the staged deployment, but everything gets very messy because that
function truly is operating on the bootloader list.
For now what I settled on is to just discard the staged deployment;
down the line we can enhance things.
Where we then have some new gymnastics is around implementing
the finalization; we need to go to some effort to pull the staged
deployment out of the list and mark it as unstaged, and then pass
it down to `write_deployments()`.
Closes: #1539
Approved by: jlebon
If we're running as pid1, avoid printing anything in the normal
success paths as we don't want to affect the physical console by
default; the device may be using a splash screen, etc.
Also cleanup the code a bit to use a single variable
`running_as_pid1`, declare-and-initialize, use the
`bool` type, etc.
Closes: #1531
Approved by: jlebon
In prep for staging work, where we'll need to load the origin
for the staged deployment too.
The function was previously trying to avoid operating on an
instantiated deployment, but the data we need is in the deployment
object at that point.
Closes: #1538
Approved by: jlebon
Prep for handling staged deployments better; if we're not passed
the staged one back, then we just want to delete it but not
touch the bootloader config.
Closes: #1538
Approved by: jlebon
The reason we were returning a hashtable is a bit lost to history,
there's no reason to do so now anyways. Also port to declare-and-initialize
style and add more comments.
Closes: #1538
Approved by: jlebon
Add API to write a deployment state to `/run/ostree/staged-deployment`,
along with a systemd service which runs at shutdown time.
This is a big change to the ostree model for hosts,
but it closes a longstanding set of bugs; many, many people have
hit the "losing changes in /etc" problem. It also avoids
the other problem of racing with programs that modify `/etc`
such as LVM backups:
https://bugzilla.redhat.com/show_bug.cgi?id=1365297
We need this in particular to go to a full-on model for
automatically updated host systems where (like a dual-partition model)
everything is fully prepared and the reboot can be taken
asynchronously.
Closes: https://github.com/ostreedev/ostree/issues/545Closes: #1503
Approved by: jlebon
Add a "hidden command" flag, and use it for `admin instutil` since
I regret adding it, and people should be using the API.
Prep for adding another hidden command as part of staging deployments.
(Down the line we should investigate deduplicating the recursive
command parsing code)
Closes: #1535
Approved by: jlebon
Let's be opinionated now, and our installed/ test story *is*
Ansible/STR. Merge `tests/fedora-str` into `tests/installed/`.
Rework the nondestructive tests into a separate playbook run, and parallelize
them for more efficiency.
The destructive tests are also changed to use Ansible more.
Add a higher level `run.sh` entrypoint and update the `README.md`
with some useful tips.
Closes: #1513
Approved by: jlebon
The main blocker for doing this before was the `goto out` handling
for remounting `/boot`. Handle that by factoring out the bits that
require it to a helper function, and do the C/GError equivalent of
"try/finally".
Not prep for anything right now, just decided to do this since I had the file
open.
Closes: #1515
Approved by: jlebon
For staged deploy, we want to pay the cost of creating copies from
`/usr/etc` → `/etc` at stage time, since it can be expensive. (We
want to minimize time spent during shutdown).
Split it up into two functions; the logic is also simply clearer.
Closes: #1514
Approved by: jlebon
Prep for deployment staging. We had the code to hande "explicit kargs" in one
place, but the "use merge deployment" karg bits mixed in with the "/etc merge"
logic. Those are separate things, and it's better to have karg handling in one
place.
Closes: #1514
Approved by: jlebon
A quick turnaround to include one PR: https://github.com/ostreedev/ostree/pull/1508
"switchroot: Ensure /run/ostree-booted is created even without initramfs"
This fixes ostree when booting without an initramfs. Thanks to @akiernan for the
bug report and helping review the fix! I'm working on enhancing
the test suite, which will help in adding some coverage here.
Also for this release I'm going to avoid adding a "stub" symbol section
to the `-released.sym` file; I don't believe it's necessary.
Closes: #1512
Approved by: jlebon
See https://mail.gnome.org/archives/ostree-list/2018-March/msg00012.html
If ostree-prepare-root is run as pid 1 (i.e we're not using an initramfs), then
anything we write outside the target sysroot (such as `/run/ostree-booted`) will
be lost.
Since `ostree-remount.service` runs fairly early in boot, and is triggered via
`ConditionKernelCommandLine=ostree`, we can just touch the file there in
addition.
Closes: #1508
Approved by: akiernan
We were previously assuming that the host content had duplicates,
which...hopefully it doesn't! We shouldn't rely on that.
Also this test is slow in production and flaky. Let's just test
a single duplicate object.
Closes: #1509
Approved by: jlebon
Support e.g. `-e tests=payload-link`, to choose specific tests for more rapid
iteration, and allow skipping tmpdir cleanup to be able to debug.
Closes: #1509
Approved by: jlebon
This is prep for splitting off "nondestructive" tests which
we can run in parallel from the destructive/invasive ones which
e.g. change the host refspec, do deployments.
The `cd` invocation in `prepare_tmpdir` wasn't working because we were running
it in a subshell. Fix this by dropping the subshell.
Closes: #1509
Approved by: jlebon
Prep for creating more types of tests.
Move copying of `tests/` into the sysinstall-tests rather than `overlay-git`
as not all test types may need that.
Factored out of https://github.com/ostreedev/ostree/pull/1501Closes: #1509
Approved by: jlebon
Of course we only notice these things are broken when the CI breaks.
Also add some comments and flip flatpak to `required: true` since it should be
now.
Closes: #1507
Approved by: jlebon
It's been over a month since 2018.2; we have a few features and various fixes,
and the "stage" work pending which is pretty invasive. Time for a new release!
Closes: #1506
Approved by: jlebon
Ensures it's labeled consistently. Prep for staged deployments which reworks the
logic around when the origin file is written.
Closes: #1505
Approved by: jlebon
Pulling some of this out of stage deploy work. It's generally better as it's
easier to change functions to have multiple callers.
Closes: #1505
Approved by: jlebon
Prep for staged deployments; they won't be in the primary deployment
list, and we want to print them first.
Also pull in some code from rpm-ostree for the red/bold bits and use
that tree-wide.
Update submodule: libglnx
Closes: #1504
Approved by: jlebon
Let's make our `run.sh` generically support any playbook. This is prep for
writing further tests in Ansible. Along with that, rework the Ansible so that
`tests.yml` is a playbook, and then the other bits are just task lists. It's
easier to read.
I also started to add a `use_git_build` variable with the idea that we'll be
able to run these same tests against an upstream image by setting that variable
off.
Closes: #1493
Approved by: jlebon
I was looking at this code in prep for "staging" deployments,
and there are several cleanups to be made here. The first
thing I noticed is that we look for the `ostree=` kernel argument,
but the presence of that should be exactly equivalent to having
`/run/ostree-booted` exist. We just added a member variable for
that, so let's make use of it.
Related to this, we were erroring out if we had the karg but
didn't find a deployment. But this can happen if e.g. one is
using `ostree admin --sysroot` from an ostree-booted system! It's
actually a bit surprising no one has reported this so far; I guess
in the end people are either using non-ostree systems or running
from containers.
Let's add a member variable `root_is_sysroot` that we can use
to determine if we're looking at `/`. Then, our more precise
"should find a booted deployment" state is when both `ostree_booted`
and `root_is_sysroot` are TRUE.
Next, rather than walking all of the deployments after parsing,
we can inline the `fstatat()` while parsing. The mild ugly
thing about this is assigning to the sysroot member variable while
parsing, but I will likely clean that up later, just wanted to avoid
rewriting everything in one go.
Closes: #1497
Approved by: jlebon
We noticed this in a recent PR. While I'm here, also only do
the `find` once, add `-type l` for good measure, and use our
built in `libtest.sh` assertion functions.
Closes: #1494
Approved by: giuseppe
Reusing the way `standard-test-roles` has support for booting
a qcow2 actually gets us to the "VM-in-container" flow. Plus
Ansible over shell script is sometimes nicer.
https://fedoraproject.org/wiki/CI/Tests#Testing_an_Atomic_Host
It's better than what we were doing before for installed tests,
and moreover using Ansible more broadly for testing is going
to align us better with Fedora's CI.
As part of this I split off a "libpaprci" which I intend to maintain
as a "copylib" for a little bit between ostree/rpm-ostree, and then
we'll figure out how to expand from there (maybe some of the patterns
get "baked in" to PAPR for example).
Note the `FAH27-insttests` context moves to the top since it's now
of primary importance, and I expect that we start expanding it.
Closes: #1462
Approved by: jlebon
The original changes here apparently had the *idea* that `--static-deltas-only`
would be useful in general, but we never implemented that. The current
situation where it's ignored unless `--delete-commit` is specified is
very misleading and I can easily see it leading to data loss for people.
Let's error out until we have a chance to make it actually useful.
Related: https://github.com/ostreedev/ostree/issues/1479Closes: #1482
Approved by: giuseppe
I think this got changed in a refactor. We definitely want
to total up the amount of space that *would* be freed even
with `--no-prune` AKA `OSTREE_REPO_PRUNE_FLAGS_NO_PRUNE`.
It's actually a bit terrifying this is apparently the first test case for
the `--no-prune` option...
Closes: https://github.com/ostreedev/ostree/issues/1480Closes: #1483
Approved by: jlebon
The `admin` commandline should be considered a demo; I just added
the `pin` command *mostly* so we could use it for unit tests, although
I can imagine other people using it.
But maintaining completions is a lot of overhead right now, let's not
do it for `admin`.
The other command line options that operate on repos we will definitely maintain
since they're used in releng contexts.
Closes: #1468
Approved by: jlebon
Example user story: Jane rebases her OS to a new major version N, and wants to
keep around N-1 even after a few upgrades for a while so she can easily roll
back. I plan to add `rpm-ostree rebase --pin` to opt-in to this for example.
Builds on the new `libostree-transient` group to store pinning state there.
Closes: https://github.com/ostreedev/ostree/issues/1460Closes: #1464
Approved by: jlebon
The `origin/unlocked` and `origin/override-commit` keys are examples of state
that's really transient; we don't want to maintain them across upgrades. Right
now there are bits for this in both `ostree admin upgrade` as well as in
rpm-ostree.
This new API will slightly clean up both cases, but it's really prep for adding
a concept of deployment "pinning" that will live in the new
`libostree-transient` group.
Closes: #1464
Approved by: jlebon
There are enough fixes here, and there are some potentially larger patches
incoming like wmanley's checkout speedups and the payload link that will need
soak time in master.
Closes: #1455
Approved by: jlebon
Since f4d1334e19 the primary pull code maintains a
maximum queue. In that commit message I said `Note that I kept an assertion.`.
But I think this is wrong since while it covers a lot of the normal cases, if
one is e.g. trying to fetch a ton of refs, the primary pull code doesn't yet
queue those. While it'd be nice to queue those, it isn't worth carrying
extra assertions in the backends that can still trigger.
Closes: https://github.com/ostreedev/ostree/issues/1451Closes: #1453
Approved by: dbnicholson
There are a few cases for knowing whether a commit has identical
content to another commit. Some people want to do a "promotion workflow",
where the content of a commit on a tesitng branch is then "promoted"
to a production branch with `ostree commit --tree=ref`.
Another use case I just hit in rpm-ostree deals with
[jigdo](https://github.com/projectatomic/rpm-ostree/issues/1081) where we're
importing RPMs on both the client and server, and will be using the
content checksum, since the client/server cases inject different metadata
into the commit object.
Closes: https://github.com/ostreedev/ostree/issues/1315Closes: #1449
Approved by: jlebon
Having the `uncompressed-object-cache` directory in `archive` repos by default
is clutter; the functionality should be considered deprecated.
Now we only create the directory if we're doing a checkout with the cache
enabled.
Closes: #1446
Approved by: jlebon
This will allow the text to be used in Wikipedia for example; it
also just makes more sense for documentation than the LGPLv2+.
Closes: #1431Closes: #1432
Approved by: jlebon
When we changed around the kernel location in rpm-ostree, we
started installing the kernel into `/boot` as `modules_object_t`,
and the current policy didn't permit that. For maximum compatibility,
relabel installed kernel/initramfs/dtb as `boot_t`.
https://bugzilla.redhat.com/show_bug.cgi?id=1536991Closes: #1444
Approved by: jlebon
Downstream BZ: https://bugzilla.redhat.com/show_bug.cgi?id=1498281
This came up as a problem with `oci-umount` which was trying to ensure some host
mounts like `/var/lib/containers` don't leak into privileged containers. But
since our `/sysroot` mount wasn't private we also got a copy there.
We should have done this from the very start - it makes `findmnt` way, way less
ugly and is just the obviously right thing to do, will possibly create world
peace etc.
Closes: #1438
Approved by: rhvgoyal
Currently we were parsing `opt_filename` twice...I dug through
the history a bit and it looks like it may have been an accident
from refactoring.
What we're fixing here concretely is that using relative subdirectories
like `--filename somesubdir/foo` broke because we were incorrectly
passing the `somesubdir/` again.
Closes: #1423Closes: #1427
Approved by: jlebon
Prep for further work here. This diff is a bit noisy for the delta bits because
the identation was off originally as well.
Closes: #1424
Approved by: jlebon
In particular I'd like to get the `--copyup` changes out for an rpm-ostree
release that will use them. But there are other good changes here, and let's
keep up a regular release train 🚄 in general.
Closes: #1413
Approved by: jlebon
Previously we were doing e.g. `ot_util_filename_validate()` specifically inline
in dirtree objects, but only *after* writing them into the staging directory (by
default). In (non-default) cases such as not using a transaction, such an object
could be written directly into the repo.
A notable gap here is that `pull-local --untrusted` was *not* doing
this verification, just checksums. We harden that (and also the
static delta writing path, really *everything* that calls
`ostree_repo_write_metadata()` to also do "structure" validation
which includes path traversal checks. Basically, let's try hard
to avoid having badly structured objects even in the repo.
One thing that sucks in this patch is that we need to allocate a "bounce buffer"
for metadata in the static delta path, because GVariant imposes alignment
requirements, which I screwed up and didn't fulfill when designing deltas. It
actually didn't matter before because we weren't parsing them, but now we are.
In theory we could check alignment but ...eh, not worth it, at least not until
we change the delta compiler to emit aligned metadata which actually may be
quite tricky. (Big picture I doubt this really matters much right now
but I'm not going to pull out a profiler yet for this)
The pull test was extended to check we didn't even write a dirtree
with path traversal into the staging directory.
There's a bit of code motion in extracting
`_ostree_validate_structureof_metadata()` from `fsck_metadata_object()`.
Then `_ostree_verify_metadata_object()` builds on that to do checksum
verification too.
Closes: #1412
Approved by: jlebon
While we do protect against path traversal during pull, let's also validate
during checkout; it's a cheap operation and provides good last-mile protection.
Closes: #1412
Approved by: jlebon
I was reading about a recent security issue with both EMC and VMWare:
https://arstechnica.com/information-technology/2018/01/emc-vmware-security-bugs-throw-gasoline-on-cloud-security-fire/
It's a classic path traversal problem, and that made me think more about our
handling of this in libostree. Fortunately of course, not being new to
this rodeo, long ago I *did* consider path traversal. Inside the pull
code, we call `ot_util_filename_validate()`. Also, `fsck` does this too.
I have further followups here, but let's add some test cases for this. I crafted
a repository with a `../` in a dirtree object by patching libostree to inject
it, and that's included as a tarball.
This patch covers the two cases where we do already have checks; pulling
via HTTP, and in `fsck`.
Closes: #1412
Approved by: jlebon
Apparently there testing systems that literally install *all*
packages. Having `ostree-grub2` currently causes grub2 to fail
on a non-ostree managed system. Let's just gracefully exit
if there's no system repository.
https://bugzilla.redhat.com/show_bug.cgi?id=1532668Closes: #1399
Approved by: jlebon
This tripped up the `docbook-dtds` `%post` in my experiments
with doing rpm-ostree for buildroots.
I cloned and built [xfstests](https://git.kernel.org/pub/scm/fs/xfs/xfstests-dev.git)
but haven't yet investigated actually running it.
In the meantime let's do the obvious fix here; we need to distinguish
between "copyup enabled" and "actually did a copyup" in the open path
at least, since if we didn't do a copyup we don't need to re-open.
Closes: #1396
Approved by: jlebon
Colin Walters