xdg-app passed this a filename directly, and in this case it should be
used as is. This regressed to always look for "superblock" in the same
directory as the passed in filename.
https://bugzilla.gnome.org/show_bug.cgi?id=762617
Some Docker layers are just metadata in the `layer.json`. If one is
mapping Docker layers to OSTree commits, one needs to create a dummy
root directory, because OSTree doesn't support metadata-only commits.
Let's just push that logic down here because it's easier than special
casing it in higher levels.
One of the design goals with deltas was not just wire efficiency,
but also having all the data up front about how much data would
be transferred before starting.
Let's expose that better by adding a `dry-run` option to the pull API.
This requires static deltas to be useful. Basically we simply call
the progress callback once with the data from the superblock.
For a production release repository, most OS vendors would want
to just always use static deltas. Add the ability for the pulls to
require it.
(I think I'll also add a summary key for this actually in addition,
so the repo manager can force it too)
If ostree is run in a test setup where it operates as root in a tmp
directory, it might cause issues to flag the deployments as immutable.
The test harness might simply be doing an `rm -rf` (effectively the case
for gnome-desktop-testing-runner), which will then fail.
We add a new debug option to the ostree_sysroot object using GLib's
GDebugKey functionality to allow our tests to communicate to ostree that
we don't want immutable deployments.
This is a more flexible version of the previous
ostree_repo_write_archive_to_mtree() which took a file reference.
This has an extensible options structure, and in particular
now supports `ignore_unsupported_content`.
I plan to use this for importing Docker images which contain device
nodes. (There's no reason for container images to have those, so
we'll just ignore them).
Also here, just like the export variant, the caller is responsible for
setting up libarchive.
I was going to add new API for importing, and it was really confusing
that what I think of now as import and export both had "write" in the
name. It's just clearer to talk about the direction.
At the same time, include `Export` in the options structure.
This isn't an ABI break as the API isn't in a release.
I was getting a compilation error with the GCC hardening flags which
look for a missing mode with `O_CREAT`. The right fix here is to drop
`O_CREAT`, as truncate() should throw `ENOENT` if the file doesn't
exist.
I don't know why we didn't do this a long time ago. This extends the
pull API to allow grabbing a specific commit, and will set the branch
to it. There's some support for this in the deploy engine, but there
are a lot of reasons to support it for raw pulls (such as subset
mirroring cases).
In fact I'm thinking we should also have the override-version logic
here too.
NOTE: One thing I debated here is inventing a new syntax on the
command line. Git doesn't seem to have this functionality (probably
because it'd be rarely used). The '@' character at least doesn't
conflict with anything.
Anyways, I wanted this for some other test cases. Without this,
writing tests that go between different commits is more awkward as one
must generate the content in one repo, then pull downstream, then
generate more content, then pull again. But now I can just keep track
of commit IDs and do exactly what I want without synchronizing the
tests.
At the moment I'm looking at using rpm-ostree to manage RPM inputs
which can then be converted into Docker images. It's most convenient
if we can stream directly out of libostree rather than doing a
checkout + tar combination.
There are also backup/debugging etc. reasons to implement `export` as
well.
While it's not strictly tied to OSTree, let's move
https://github.com/cgwalters/rofiles-fuse in here because:
- It's *very* useful in concert with OSTree
- It's tiny
- We can reuse OSTree's test, documentation, etc. infrastructure
One thing to consider also is that at some point we could experiment
with writing a FUSE filesystem for OSTree. This could internalize a
better equivalent of `--link-checkout-speedup`, but on the other hand,
the cost of walking filesystem trees for these types of operations is
really quite small.
But if we did decide to do more FUSE things in OSTree, this is a step
towards that too.
Now we display stats on the individual parts, such as the blob size
and the number of each type of opcode. Most interesting to me is
things like how many bsdiff opcodes there are vs new objects, etc.
We had code to deal with opening/checksumming/decompressing static
deltas in a few places. I'd like to teach `ostree static-delta show`
how to display more information, and this will allow it to just use
`_ostree_static_delta_part_open()` too.
Right now though, almost all of the details of deltas are private, so
we can't do the "honest thing" and have the command line just use the
shared library.
Eventually some of this should appear in the API, but for now add
command line which is useful for debugging.
I'd like to incrementally convert all of `ostree-repo*.c` to
fd-relative usage, so that we can sanely introduce
`ostree_repo_new_at()` which doesn't involve GFile.
This one is medium risk, but passes the test suite.
The original intention here was that we'd keey around a copy of the
file so that grub2 could eventually learn how to do atomic updates by
checking for a "fully written" marker in the *new* file, and if it
didn't exist, falling back to grub2.cfg.old.
I haven't yet proposed that upstream, but we might as well stop
deleting the file since it's useful as a backup at least.
Reported-by: Gatis Paeglis
This is a better followup to dc9239dd7b
since I wanted to do fsync-less checkouts in rpm-ostree too, and
replicating the "turn off fsync temporarily" was in retrospect just a
hack.
We can simply add a boolean to the checkout options.
https://github.com/GNOME/ostree/pull/172
Originally, a lot of the `fsync()` calls here were added for the
wrong reason - I was chasing a bug that ended up being the extlinux
bootloader not parsing 64 bit ext4 filesystems. But since it looked
like corruption, I tried adding a lot more `fsync()` calls.
All we should have to do is use `syncfs()`. If that doesn't work,
it's a kernel bug.
I'm making this change because skipping the individual fsyncs can be a
major performance win - it's easier for the FS to optimize, we do more
in parallel, etc.
https://bugzilla.gnome.org/show_bug.cgi?id=757117
A fast way to generate new OSTree content using an existing
tree is to checkout (as hard links), add/replace files, then
call `ostree_repo_scan_hardlinks()`, then commit.
But `ostree_repo_scan_hardlinks()` scans the entire repo, which
can be slow if you have a lot of content.
All we really need is a mapping of (device,inode) -> checksum
just for the objects we checked out, then use that mapping
for commits.
This patch adds API so that callers can create a mapping via
`ostree_repo_devino_cache_new()`, then pass it to
`ostree_repo_checkout_tree_at()` which will populate it, and then
`ostree_repo_write_directory_to_mtree()` can consume it.
I plan to use this in rpm-ostree for package layering work.
Notes:
- The old `ostree_repo_scan_hardlinks()` API still works.
- I tweaked the cache to be a set with the checksum colocated with
the key, to avoid a separate malloc block per entry.
https://github.com/GNOME/ostree/pull/167
The logic for checking which bootversion to use tries to access
sysroot->bootversion if the user didn't specify an explicit bootversion
on the command-line nor through the env var. However, at that point, the
sysroot object is not yet initialized, so it will always return 0, even
when it's 1.
This would cause e.g. `grub2-mkconfig` to have no output for the BLS
entries whenever the entries were under `/boot/loader.1`.
Related: RHBZ1293986
The tmp directory is lazily created for each fetcher instance, since
it may require superuser permissions and some instances only need
_ostree_fetcher_request_uri_to_membuf() which keeps everything in
memory buffers.
This is a continuation of earlier work to drop the individual fsync on
files/directories in favor of relying on `syncfs()` for speed.
As part of that cleanup, I'm porting it to be fd-relative.
I feel relatively confident about this change given that this area of
the code has notable test suite coverage, although that code runs as
non-root.
I'm porting the deployment code to be fd-relative, but part of the
logic was using `GFile` to talk to `OstreeRepoFile` to determine the
"bootcsum" (boot config checksum) before checking out the file tree.
We can avoid having both code paths by checking out the tree first,
then looking at it on the filesystem.
Extract existing code from ostree_repo_prune and add an argument COMMIT,
that controls which commit purge. If not set, the old behavior is kept.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Downloads and prints a remote summary file and any signatures in an
easy-to-read format, or alternatively with the --raw option, prints
the summary GVariant data directly.
https://bugzilla.gnome.org/show_bug.cgi?id=759250
Given the previous commit, which isolates SoupSession in a separate
thread, it should be safe to start pushing a temporary main context
for synchronous requests again.
This partially reverts 84fe2ff, which partially reverted 9f3d586.
Related to https://bugzilla.gnome.org/show_bug.cgi?id=753336
Move the SoupSession to a separate thread with its own isolated main
context and main loop. All interaction with the SoupSession occurs
by way of idle sources attached to the session's main context, which
execute on the session's thread.
This should solve the problem of running an asynchronous fetch request
synchronously by pushing a new thread-default main context and iterating
a main loop until the request completes. Prior to this, the new thread-
default main context would interfere with the SoupSession's own async
processing.
A lot of effort here just to avoid touching SoupSession directly in
ostree_fetcher_new(). The reason will become apparent in subsequent
commits.
Note this introduces generated enum/flags GTypes using glib-mkenums.
I could have just made the property type as plain integer, but doing
properties right will henceforth be easier now that the automake-fu
is established.
This way two pulls will not use the same tmpdir and accidentally
overwrite each other. However, consecutive OstreeFetchers will reuse
the tmpdirs, so that we can properly resume downloading large objects.
https://bugzilla.gnome.org/show_bug.cgi?id=757611
Concurrent pulls break since we're sharing the staging directory for
all transactions in the repo. This makes us use a per-transaction directory.
However, in order for resumes to work we first look for existing
staging directories and try to aquire an exclusive lock for them. If
we can't find any staging directory or they are all already locked,
then we create a new one.
https://bugzilla.gnome.org/show_bug.cgi?id=757611
This creates a subdirectory of the tmp dir with a selected prefix,
and takes a lockfile to ensure that nobody else is using the same directory.
However, if a directory with the same prefix already exists and is
not locked that is used instead.
The later is useful if you want to support some kind of resumed operation
on the tmpdir.
touch reused dirs
https://bugzilla.gnome.org/show_bug.cgi?id=757611
Bison is a well known external dependency, so just require it.
Including the generated content in git means it may or may not
be regenerated based randomly on timestamps, etc.
Also use `$(AM_V_GEN)` so we get prettier output.
Previously we were just ignoring this, which hid a bug in
an earlier commit that generated them.
Also change the `commit` program to use both APIs - this
involves extra code, but not too much.
This way, reverting the fix with this on top caused the test suite to
fail. Adding an active test for this would need a custom test program
using the C API, or adding a cmdline flag to the client, neither of
which quite seemed worth it.
Use the parse-datetime module from gnulib, and adapt it to not require
other modules as portability is not really an issue for us.
DATE can be specified in different formats, such as: "-1 week", "last
monday", "1 week ago".
Include the generated .c file in the repository so to not add another
dependency to Bison.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
ostree_repo_write_commit_with_time() converts the timestamp to
big-endian byte order.
ostree_repo_write_commit() was also doing this when calling
ostree_repo_write_commit_with_time(), resulting in a corrupted
commit object (timestamp bytes were backwards).
Recent regression in 14ffd7022a
Just to make copy-and-paste a little easier, as I often use this command
immediately before rebasing.
e.g.
# ostree remote refs fedora-atomic
fedora-atomic:fedora-atomic/f23/x86_64/docker-host
fedora-atomic:fedora-atomic/f23/x86_64/testing/docker-host
^^^^^^^^^^^^^^ (this part is new)
# rpm-ostree rebase fedora-atomic:fedora-atomic/f23/x86_64/testing/docker-host
Do not delete a .commitmeta file after removing the last metadata entry.
This way a client will pull the empty .commitmeta file and overwrite old
metadata as expected.
https://bugzilla.gnome.org/750459
ostree_checksum_bytes_peek() can return NULL if the checksum has an
incorrect length (most likely from disk corruption) but most callers
are not prepared to handle this and would likely crash.
Use ostree_checksum_bytes_peek_validate() instead, which sets a
GError on an invalid checksum.
It may have a different meaning, and the usage screen is not helpful.
Print the usage screen only when the command is not found.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
This is very useful for the inline-parts case, as you can then include
detached signatures in a single file representing the commit.
It is not as important for the generic pull case, as the detached
metadata is only a single small file. Additionally the detached
metadata is not content referenced and may change after the static
delta file was created, so we need to pull the latest version anyway.
If you pass a diriectory it will look for the "superblock" child, otherwise
it will use the file as the superblock. I need this in xdg-app to be able
to install any filename as a bundle.
Also renames OstreeRepoTrustedContentBareCommit to
OstreeRepoContentBareCommit so that it can be used by both.
This will be needed when we introduce checksum verification of objects
in static deltas.
In this mode the parts are stored in the metadata of the main delta
superblock file. This can be useful if you want a single-file delta
for easy transport, or for http in the case the delta is very small.
When a commit is deleted and the repo is configured to use tombstone
commits, create one. Delete the tombstone file only if the commit is
pulled again.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Add a new object type: OSTREE_OBJECT_TYPE_TOMBSTONE_COMMIT that is
used when a commit was intentionally removed.
If the remote repository doesn't use tombstone commits, do not fail on
a missing commit (change 0b795785dd).
When the remote repository uses tombstones, if a commit cannot be
found, check if the tombstone file is present and fail if it is not
present.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
There might be a race here in that we create new symlink files *after*
calling `syncfs`, and they are not guaranteed to end up on disk.
Rework the code so that we create symlinks before, and then only
rename them after (and `fsync()` the directory for good measure).
Additional-fixes-by: Giuseppe Scrivano <gscrivan@redhat.com>
Tested-by: Giuseppe Scrivano <gscrivan@redhat.com>
This still needs verification that we're fixing a real bug; but I'm
fairly confident this won't make the fsync situation worse.
https://bugzilla.gnome.org/show_bug.cgi?id=755595
Adds an entry to the origin file to force the OstreeSysrootUpgrader to
pull and deploy the given checksum.
[origin]
override-commit=CHECKSUM
If the option is not given, any such entry is explicitly removed from
the origin file to ensure we upgrade to the latest available commit.
Upgrader now looks for an "override-commit" key in the origin file
with a commit checksum, which causes the upgrader to pull and deploy
the specified commit rather than the latest available commit on the
origin refspec.
The current code checks if /boot/uEnv.txt is a symlink to
decice if sysroot requires u-boot support. Why this is bad:
There are 2 ways to provide a custom env to u-boot from user space:
1) A compiled binary that is sourced from u-boot.
2) A text file (usually /uEnv.txt) that is imported into env from u-boot.
The current OSTree u-boot integration code was designed with the 1st
case in mind.
Many bootscripts provided by an embedded device vendors expect
to find uEnv.txt in the top level directory, it is often hardcoded
when building u-boot and is difficult to change later on. Or in other
cases it is stored in read-only memory so changing it would require
re-flushing boot loader with a new env. So the issue here is that
OSTree's and vendor uEnv.txt want to exist and the same path and OSTree
would throw away any changes added to /uEnv.txt by user on the next
upgrade/deploy.
This patch "hides" away the OSTree's env file loader/uEnv.txt from users
who are used to edditing uEnv.txt at the top level directory. Now to add
OSTree support on such boards you can simply add a custom logic in uEnv.txt
that loads ostree env from /loader/uEnv.txt
This change is backward compatible with the previous ostree releases and
solves the issue described in:
https://bugzilla.gnome.org/show_bug.cgi?id=755787
Using `commit_subject` instead of `arg` is clearer as it can refer to
a directory, archive or ref.
This is just an aesthetic change in the source code, having no impact
anywhere else.
There is already a check that the destination object does not
exist in all other cases when processing an incoming static delta.
However, the bspatch case would still try to run and fail. Add
an analogous check to that case as well.
https://bugzilla.gnome.org/show_bug.cgi?id=756260
zlib can return LZMA_BUF_ERROR, which indicates that either
the input or output buffer has size zero. This case should cause
the correct error to be passed back from g_converter_convert
to expand the relevant buffer. Since this error is ambiguous
as to which buffer is too small, an explicit check on the
output buffer size is added as well.
https://bugzilla.gnome.org/show_bug.cgi?id=756260
I was working on a different test, and ended up being very confused at
the behavior where removing the last deployment didn't remove the last
`ostree/X/X/X` ref pointing to its commit.
There's no reason to special case the last undeployment AFAIK, and the
existing code handles this.
Track outstanding HTTP requests in a table for easier debugging.
Also fixes a bug discussed in https://bugzilla.gnome.org/755224
where the outstanding request counter was not decremented in the
event of an error, which could result in the fetcher hitting its
max request limit and locking up.
The bug is fixed by removing the request struct from the table in
pending_uri_free(), which is always called regardless of error,
so the outstanding request count is always accurate.
Have OstreeFetcherPendingURI be the GTask's task_data and pass the GTask
around in queues and callback closures. The reference counting before
was a little confusing and this helps clarify it, at least to me.
OstreeFetcherPendingURI no longer needs its own reference count.
Had a rare situation where I had no libsoup development files, so I
took the opportunity to fix the build errors. Ugly, but works now.
Would be nice if libsoup could be a hard dependency since we rarely
ever test a configuration without it.
To support deploying older commits:
ostree pull <remote> <checksum>
ostree admin deploy <checksum>
Prior to this, the deploy command garbage collected <checksum> since
there's no ref pointing to it, and then ostree_sysroot_deploy_tree()
fails because it can't find the <checksum> commit.
https://bugzilla.gnome.org/732526
New public function works like ostree_sysroot_cleanup() EXCEPT FOR
pruning the repository.
Under the hood, add _ostree_sysroot_piecemeal_cleanup() which takes
flags to better control what files are cleaned up. Both public cleanup
functions are now wrappers for _ostree_sysroot_piecemeal_cleanup() with
different flags.
xdg-app was hanging for me with v2015.8, but worked with v2015.7.
I narrowed things down to the GMainLoop/context commit, in which
we started pushing a temporary main context for synchronous
requests internally.
That's never really going to work with libsoup - there needs
to be a single main context which works on the socket. Furthermore,
clients couldn't get progress messages that way.
For *other* internal uses where we added APIs that talk to the remote
repo, we cleanly push a temporary main context.
(Note that I kind of snuck in a change here around the GError handling
in pulls that isn't strictly related but came up in testing)
I noticed xdg-app was looping trying to fetch 1427 refs. We
don't want to do that unless asked to.
(And also, we need to make static delta requests async)
There's no reason to keep them hidden. I have a hard policy that
OSTree should *not* be used to carry secrets. Things like host ssh
private keys should be set up out of band by an OS-external
configuration mechanism such as kickstart, cloud-init, etc.
We also assume that hiding binaries is not very useful as most
attackers would be able to find them on the Internet or (for
subscribed content) acting as a customer.
This fixes a bug with mirroring because we changed to take the
unmodified upstream objects rather than uncompress <-> recompress.
https://bugzilla.gnome.org/show_bug.cgi?id=748959
Now that the computed similar objects are all regular files,
get_unpacked_unlinked_content should never be called on any other
object type. Assert that this is true instead of silently succeeding.
_ostree_delta_compute_similar_objects should not output symlinks.
Previously, a symlink in the "from" commit could be matched to a
real file in the "to" commit, since nothing was filtering symlinks
on the "from" side. This led to failures running the bzdiff
algorithm.
On systems with slow disks, the recursive scanning of directories can
be expensive -- it takes upwards of 2 minutes on our systems. This can
block the main loop for such a long time that it allows the download to
time out...
As such, move all the scanning of objects to a queue, processed from
an idle, to make sure that we don't block the main loop when scanning.
https://bugzilla.gnome.org/show_bug.cgi?id=753336
First of all, what we were doing with having GMainLoop in the internal
APIs is wrong. Synchronous APIs should always create their own main
context and not iterate the caller's. Doing the latter creates
potential for evil reentrancy issues. Sync API should block, async
API is for not blocking.
Now that's out of the way, fix the pull code to do the clean
```
while (termination_condition (state))
g_main_context_iteration (mainctx, TRUE);
```
model for looping. This is a lot easier to understand and ultimately
more reliable than having other code call `g_main_loop_quit()`, as the
loop condition is in exactly one place.
We can also remove the idle source which only fired once.
Note we have to add a hack here to discard the synchronous session and
create a new one which we only use async.
https://bugzilla.gnome.org/show_bug.cgi?id=753336
ostree_repo_prepare_transaction() should always be matched with a call
to either ostree_repo_commit_transaction() or
ostree_repo_abort_transaction().
Since ostree_repo_pull_with_options() does not call
ostree_repo_abort_transaction() on errors, the OstreeRepo instance will
hit an assertion when it's re-used later for another attempt, such as
when the update is driven by an external component through libostree and
network temporarily goes down.
This commit simply always calls ostree_repo_abort_transaction() in the
exit path of ostree_repo_pull_with_options(), since the function is safe
to call even when we're not in a transaction, and that matches e.g. what
ostree-sysroot-cleanup.c does.
The new API permits to query a remote repository summary file and
retrieve the list of available refs.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Returns a GFile for the default system root, which is usually the root
directory unless overridden by the OSTREE_SYSROOT environment variable
(which is mainly intended for testing).
libsoup will cache sessions, so it might be the case that we get a
reused session when pulling from the same repo multiple times in one
process.
In this case we were leaking signal connections, which caused
callbacks into freed memory with bad consequences.
Fix it by tying the signal connection to the object lifetime.
I did a quick audit pass through the pull code. What I focused on the
most is the case where `gpg-verify-summary=true`, and in particular
where `gpg-verify=false` too. This should be a valid and secure
configuration.
The primary change here is to error out very quickly if either
`summary` or `summary.sig` are 404. Previously, we'd only error out
if we were processing deltas.
Expand the existing test case to cover this, plus invalid summary and
invalid sig. (The test case was failing with current git master too).
It allows to specify whether GPG verification for the summary file is
enabled for a specific repository.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Works like "ostree refs" but fetches refs from a remote repo.
This depends on the remote repo having a summary file, but any repo
being served over HTTP *ought* to have one.
This may not be the best idea for general usage, but the only use case
for metalinks currently is fetching a summary file and those are pretty
small. Far more convenient to return the file content in a GBytes.
Colin Walters
Alexander Larsson
Jonathan Lebon
Matthew Barnes
Giuseppe Scrivano
John Hiesey
Daniel Drake
Gatis Paeglis
Leandro Santiago
Jasper St. Pierre
Cosimo Cecchi