That's why the syscall was invented, so let's use it. Just noticed while reading
the code while working on another patch.
Closes: #1270
Approved by: jlebon
I was going to fix a bug in the static deltas code and I noticed
we were missing `pull-test.sh` coverage for bareuseronly 🙈.
Obviously fixing this requires duplicating some of the bits we have in
`basic-test.sh`; need to hoist that into `libtest.sh`. For now though let's get
the coverage.
Closes: #1270
Approved by: jlebon
This isn't perfect, but at least we fix an error-overwrite error, and in
practice `ostree admin unlock` isn't wrapped by `rpm-ostree` yet, so spew to
stderr is OK.
Closes: https://github.com/ostreedev/ostree/issues/1273Closes: #1279
Approved by: guyshapiro
For checkouts that are on the same device, for regular files we can simply
"adopt" existing files. This is useful in the "build from subtrees" pattern that
happens with e.g. `rpm-ostree install` as well as flatpak and gnome-continuous.
New files are things like an updated `ldconfig` cache, etc. And particularly for
`rpm-ostree` we always regenerate the rpmdb, which for e.g. this workstation is
`61MB`.
We probably should have done this from the start, and instead had a `--copy`
flag to commit, but obviously we have to be backwards compatible.
There's more to do here - the biggest gap is probably for `bare-user` repos,
which are often used with things like `rpm-ostree compose tree` for host
systems. But we can do that later.
Closes: #1272
Approved by: jlebon
I was working on "adopt" and hit corruption; turns out we were missing the
`--canonical-permissions` arg in this existing test.
(Need to abstract all of this more)
Closes: #1272
Approved by: jlebon
I was working on a patch to do build on the work done to
import content objects async to do the same for metadata, but right
now we basically rely on writing them first to do the GPG verification
when scanning.
Things will be cleaner for that if we can pass the commit object directly into
`scan_commit_object()` and consistently use `gpg_verify_unwritten_commit()`.
We're careful here to continue to do it both ways (but at most one time), to
account for the case where a bad commit has been pulled and written - we need to
keep failing GPG verification there.
Closes: #1269
Approved by: jlebon
Prep for a later patch to do GPG verification before writing commit objects;
`_ostree_repo_gpg_verify_with_metadata()` already handles this, and so dropping
this gives us consistent error messages.
Closes: #1269
Approved by: jlebon
Let's react to `Ctrl-C` faster here. Noticed while I was doing an update on my
desktop and playing with cancellation.
Closes: #1266
Approved by: jlebon
This works around an (IMO) SpiderMonkey bug - it tries to
clean up in a shared library destructor, but doesn't install a
`pthread_atfork()` handler to unset its state.
Closes: https://github.com/ostreedev/ostree/issues/1262Closes: #1264
Approved by: dbnicholson
This ends up a lot better IMO. This commit is *mostly* just
`s/glnx_close_fd/glnx_autofd`, but there's also a number of hunks like:
```
- if (self->sysroot_fd != -1)
- {
- (void) close (self->sysroot_fd);
- self->sysroot_fd = -1;
- }
+ glnx_close_fd (&self->sysroot_fd);
```
Update submodule: libglnx
Closes: #1259
Approved by: jlebon
It's no longer called directly by the pull code, so make it static.
The goal here is to have the pull and local-fs commit paths use higher level
more efficient APIs, and eventually make those APIs public.
Closes: #1257
Approved by: jlebon
This simplifies a lot of code; the header function was structured
to write to an input stream, but many callers only wanted the checksum,
so it's simpler (and error-free) to simply allocate a whole buffer
and checksum that.
For the callers that want to write it, it's also still simpler to allocate the
buffer and write the whole thing rather than having this function do the
writing.
A lot of the complexity here again is a legacy of the packfile code, which is
dead.
This is prep for faster regfile commits where we can avoid `G{In,Out}putStream`.
Closes: #1257
Approved by: jlebon
Nothing was using the `bytes_written` data (we always discard partially written
tmpfiles), so simplify everything by dropping it. Further, we always passed an
offset of `0`, so drop that argument too. (I believe that this was previously
used by the "pack files" code that we deleted long ago)
Second, we had an unnecessary internal wrapper for this function; drop that too.
Closes: #1257
Approved by: jlebon
The faster (OpenSSL/GnuTLS) code lived in a `GInputStream` wrapper, and that
adds a lot of weight (GObject + vtable calls). Move it into a simple
autoptr-struct wrapper, and use it in the metadata path, so we're
now using the faster checksums there too.
This also drops a malloc there as the new API does hexdigest in place to a
buffer.
Prep for more work in the commit path to avoid `GInputStream` for local file
commits, and ["adopting" files](https://github.com/ostreedev/ostree/pull/1255).
Closes: #1256
Approved by: jlebon
For many cases of commit, we can actually optimize things by simply "adopting"
the object rather than writing a new copy. For example, in rpm-ostree package
layering.
We can only make that optimization though if we take ownership of the file. This
commit hence adds an API where a caller tells us to do so. For now, that just
means we `unlink()` the files/dirs as we go, but we can now later add the
"adopt" optimization.
Closes: #1255
Approved by: jlebon
What the deltas code is doing is weird/unfortunate. The name
`ot_variant_read()` conflicts too much with `ot_variant_read_fd()`.
Since nothing else uses it, move it into the deltas code.
Closes: #1254
Approved by: jlebon
A lot of the libostree code is honestly too complex for its
own good (this is mostly my fault). The way we do HTTP writes
is still one of those. The way the fetcher writes tempfiles,
then reads them back in is definitely one of those.
Now that we've dropped the "partial object" bits in:
https://github.com/ostreedev/ostree/pull/1176 i.e. commit
https://github.com/ostreedev/ostree/commit/0488b4870e80ef575d8b0edf6f2a9e5ad54bf4df
we can simplify things a lot more by having the fetcher
return an `O_TMPFILE` rather than a filename.
For trusted archive mirroring, we need to enable linking
in the tmpfiles directly.
Otherwise for at least content objects they're compressed, so we couldn't link
them in. For metadata, we need to do similar logic to what we have around
`mmap()` to only grab a tmpfile if the size is large enough.
Closes: #1252
Approved by: jlebon
Buried in this large patch is a logical fix:
```
- if (!map)
- return glnx_throw_errno_prefix (error, "mmap");
+ if (map == (void*)-1)
+ return glnx_null_throw_errno_prefix (error, "mmap");
```
Which would have helped me debug another patch I was working
on. But it turns out that actually correctly checking for
errors from `mmap()` triggers lots of other bugs - basically
because we sometimes handle zero-length variants (in detached
metadata). When we start actually returning errors due to
this, things break. (It wasn't a problem in practice before
because most things looked at the zero size, not the data).
Anyways there's a bigger picture issue here - a while ago
we made a fix to only use `mmap()` for reading metadata from disk
only if it was large enough (i.e. `>16k`). But that didn't
help various other paths in the pull code and others that were
directly doing the `mmap()`.
Fix this by having a proper low level fs helper that does "read all data from
fd+offset into GBytes", which handles the size check. Then the `GVariant` bits
are just a clean layer on top of this. (At the small cost of an additional
allocation)
Side note: I had to remind myself, but the reason we can't just use
`GMappedFile` here is it doesn't support passing an offset into `mmap()`.
Closes: #1251
Approved by: jlebon
I'm regretting a bit having the `guint8*csum` variant of checksums
except for the serialized form. Once we start doing processing
it's easier to just have it remain hex.
Do an on-stack conversion for the metadata scanning function; this
drops a malloc and also just looks nicer.
Also add some long-awaited function comments to the two.
Closes: #1240
Approved by: jlebon
I was reading the pull code for the last release, and spotted
a bug in commit f923c2e1ea - in
the case where the ref doesn't exist, we don't set an error,
tripping an assertion in the main code.
The previous code wanted the ref to always exist, so just flip back the boolean
for "ignore noent". I moved the `g_strchomp()` just into the HTTP path - if a
local repo is corrupted in this way it's something to fix in that repo.
Closes: #1238
Approved by: pwithnall
I was trying to do a change for rpm-ostree to use
`OSTREE_REPO_COMMIT_MODIFIER_FLAGS_CANONICAL_PERMISSIONS`
for container builds with `bare-user-only,` but hit an assertion here
ultimtely because we weren't setting `standard::type`.
Rather than hand-rolling `GFileInfo` creation, use the stat buffer conversion
code which is more robust and used in multiple places already.
Closes: #1227
Approved by: jlebon
In particular I'd like to get the copy fix in, since it might affect users for
the keyring bits.
Update submodule: libglnx
Closes: #1225
Approved by: jlebon
I noticed this while debugging why I was seeing "2 metadata objects" fetched for
a different PR. I knew 1 was detached meta, but the other turned out to be this.
There's no reason to request a delta if the ref is unchanged.
Closes: #1220
Approved by: jlebon
I was trying to debug this while working on another PR. Add `echo ok` lines and
also use `assert_file_has_content` more directly; there's no reason to use an
intermediate `grep` since if it fails it won't print the original file.
Closes: #1220
Approved by: jlebon
So the output isn't all intermingled. I just pushed a commit
to add `--log-directory`, so we need to build it from git master
for now.
Closes: #1218
Approved by: jlebon
Followup for recent work in commits:
- 8a7a359709
- 1a9a473580
Keep track of how many objects we imported, and print that for `ostree
pull-local` (also do this even if noninteractive, like we did for `pull`).
In implementing this at first I used separate variables for import
from repo vs import from localcache, but that broke some of the
tests that checked those values.
It's easier to just merge them; we know from looking at whether or not
`remote_repo_local` is set whether or not we were doing a "HTTP pull with
localcache" versus a true `pull-local` and can use that when rendering status.
Closes: #1219
Approved by: jlebon
Our CI uses default Docker, which has SELinux labeling but is rather
evil in returning `EOPNOTSUPP` to any attempts to set `security.selinux`,
even if to the same value.
The previous fire 🔥 for this was: https://github.com/ostreedev/ostree/pull/759
The `bare` repo mode really only makes sense as uid 0, so our installed
test framework is a good match for this. However, the unit tests *do*
work in a privileged container even as non-root, and *also* should
work on SELinux-disabled systems. So let's teach the test framework
how to skip in those situations.
I tested this both in a priv container (my default builder) and an unpriv
container (like our CI).
At the same time, start executing the `test-basic.sh` from an installed test,
so we get better coverage than before.
This is just the start - all of the sysroot tests really need the
same treatment.
Closes: #1217
Approved by: jlebon
I now think commit fab1e113db was a mistake;
because it breaks the mental model that at least I'd built up that "local repos
don't have checksums verified, HTTP does".
For example, a problem with this is (with that mental model in place) it's easy
for people who set up mirrors like this to then do local pulls, and at that
point we've done a deployment with no checksum verification.
Further, since then we did PR #671 AKA commit 3d38f03 which is really most of
the speed hit.
So let's switch the default even for this case to doing checksum verification,
and add `ostree pull --http-trusted`. People who are in situations where they
know they want this can find it and turn it on.
Closes: https://github.com/ostreedev/ostree/issues/1211Closes: #1212
Approved by: jlebon
Rather than carrying two booleans, just convert `OstreeRepoPullFlags`
into `OstreeRepoImportFlags`. This allows us to drop an internal
wrapper function and just directly call `_ostree_repo_import_object()`.
This though reveals that our mirroring import path doesn't check the
`OSTREE_REPO_PULL_FLAGS_UNTRUSTED` flag...it probably should.
Prep for further work.
Closes: #1212
Approved by: jlebon
Make the "local repo" processing conditional the same as the "localcache" bits;
this is really just a de-indent. Also add some comments. Prep for further work.
Closes: #1212
Approved by: jlebon
This fixes up the last of the embarassing bits I saw from
the stack trace in:
https://github.com/ostreedev/ostree/issues/1184
We had a hardlink fast path, but that doesn't apply across
devices, which occurs in two notable cases:
- Installer ISO with local repo
- Tools like pungi that copy the repo to a local snapshot
Obviously there are a lot of subtleties here around things like the
bare-user-only conversions as well as exactly what data we copy. I think to get
better test coverage we may want to add `pull-local --no-hardlink` or so.
Closes: #1197
Approved by: jlebon
For the old `OSTREE_REPO_MODE_ARCHIVE_Z2`. Use it mostly tree
wide except for the repo finder tests (to avoid conflicting with
some outstanding PRs).
Just noted another user coming in some of those tests and wanted to do a
cleanup.
Closes: #1209
Approved by: jlebon
We added a `.dir-locals.el` in commit: 9a77017d87
There's no need to have it per-file, with that people might think
to add other editors, which is the wrong direction.
Closes: #1206
Approved by: jlebon
Such an evil bug 🙈. I was just reading an strace trying to figure out what was
going on, and noticed we had the `XXXXXX` in the lockfile name. It was only
after that I realized that that this might *be* the cause of the skopeo issue.
This is another case where we definitely need more test coverage of things that
actually use the API multiple times in process; might look at dusting off the
work for the rpm-ostree test.
Closes: https://github.com/ostreedev/ostree/issues/1196Closes: #1204
Approved by: jlebon
We keep occasionally regressing this so let's start covering it now. I'm
intentionally not running the tests since that would likely entail a lot more
conditionalizing pain.
Closes: #1194
Approved by: jlebon
While opening a repo we've recorded the device/inode for a while; use it to
avoid calling `linkat()` during object import if we know it's going to fail.
Closes: #1193
Approved by: jlebon
Conceptually `ostree-repo-pull.c` should be be written using
just public APIs; we theoretically support building without HTTP
for people who just want to use the object store portion and
do their own fetching.
We have some nontrivial behaviors in the pull layer though; one
of those is the "bareuseronly" verification. Make a new internal
API that accepts flags, move it into `commit.c`. This
is prep for further work in changing object import to support
reflinks.
Closes: #1193
Approved by: jlebon
There are use cases for not syncing at all; think build cache repos, etc. Let's
be consistent here and make sure if fsync is disabled we do no sync at all.
I chose this opportunity to add tests using the shiny new strace fault
injection. I can forsee using this for a lot more things, so I made
the support for detecting things generic.
Related: https://github.com/ostreedev/ostree/issues/1184Closes: #1186
Approved by: jlebon
Make the equivalent of a `GOnce` or Rust `lazy_static!` detecting this and share
it between the two callers.
Prep for a future similar patch for strace fault injection.
Closes: #1186
Approved by: jlebon
I saw in a stack trace that the main thread was calling `exit()` even while
worker threads were alive and doing sha256/write/fsync etc. for objects.
The stack trace was a SEGV as the main thread was calling into library
`atexit()` handlers and we were a liblz4 destructor:
```
#0 0x00007f2db790f8d4 _fini (liblz4.so.1)
#1 0x00007f2dbbae1c68 __run_exit_handlers (libc.so.6)
```
(Why that library has a destructor I don't know offhand, can't find
it in the source in a quick look)
Anyways, global library destructors and worker threads continuing simply don't
mix. Let's wait for our outstanding operations before we exit. This is also a
good idea for projects using libostree as a shared library, as we don't want
worker threads outliving operations.
Our existing pull corruption tests exercise coverage here.
I added a new `caught-error` status boolean to the progress API, and use it the
commandline to tell the user that we're waiting for outstanding ops.
Closes: #1185
Approved by: jlebon
We have a lot of layers of abstraction here; let's fold in the `trusted`
conditional into the call, since that's all the public API we're using does
anyways.
Prep for a future patch around object copying during imports.
Closes: #1187
Approved by: jlebon
In 5c940987e7 / #646 we
added `--retain-branch-depth`; this adds a symmetric
`--only-branch` for the case where a repo owner just
wants to prune a specific branch.
The implementation here is pretty straightforward; we
just walk all refs and inject the equivalent of
`--retain-branch-depth=$ref=-1` if they're *not* in
`--only-branch`.
Closes: https://github.com/ostreedev/ostree/issues/1115Closes: #1127
Approved by: jlebon
Update libglnx, which is mostly port the repo stagedir code
to the new tmpdir API. This turned out to require some
libglnx changes to support de-allocating the tmpdir ref while
still maintaining the on-disk dir.
Update submodule: libglnx
Closes: #1172
Approved by: jlebon
Doing this in prep for libglnx tmpdir porting, but I think we should also do
this because the partial fetch code IMO was never fully baked; among other
things it was never integrated into the scheme we came up with for "boot id
sync" that we use for complete/staged objects.
There's a lot of complexity here that while we have some coverage for, I think
we need to refocus on the core functionality. The libcurl backend doesn't have
an equivalent to this today.
In particular for small objects, this is simply overly complex. The downside is
clearly for large objects like FAH's 61MB initramfs; not being able to resume
fetches of those is unfortunate.
In practice though, I think most people should be using deltas, and we need to
make sure deltas work for large objects anyways.
Further ultimately the peer-to-peer work should help a lot for people
with truly unreliable connections.
Closes: #1176
Approved by: jlebon
I was looking at fixing an `rpm-ostree livefs` bug where we need to replace
`/usr/lib/passwd`. It's obviously bad if that temporarily disappears 😉. My plan
is to do a subpath checkout of just `/usr/lib/{passwd,group}`.
Make this atomic (i.e. file always exists) by changing the logic to create a
temporary link in repo/tmp, then rename() it into place.
A bonus here is we kill one of the very few (only?) non-error-cleanup i.e.
"non-linear" `goto`s in the ostree codebase.
Closes: #1171
Approved by: jlebon
It turns out that librpm automatically merges identical files between
distinct packages, and this occurs in practice with Fedora today between
`chkconfig` and `initscripts` for exmaple.
Since we added this for rpm-ostree, we basically want to do what librpm does,
let's change the semantics to do a merge. While we're here rename
to `UNION_IDENTICAL`.
Closes: #1156
Approved by: jlebon
We use the "exhaustive enum" pattern (i.e. no `default:`) in some places so
we're forced to touch all users when adding cases.
Closes: #1167
Approved by: peterbaouoft
The new libglnx `glnx_mkdtempat()` uses autocleanups, which
is inconvenient for this use case where we *don't* want autocleanups.
Since we don't need it to be fd-relative, just directly invoke
`g_mkdtemp_full()` which is fine for this use case.
Prep for updating libglnx.
Closes: #1161
Approved by: jlebon
There was only one tricky bit here around the ownership of the lines; I made use
of `g_steal_pointer()` to consistently track ownership, and converted to a `for`
loop while still preserving the loop logic around the last entry.
Closes: #1154
Approved by: jlebon
Steal some code from flatpak for this, which allows porting a few more things to
new style. I started on a public API version of this but was trying to roll some
other things into it and it snowballed. Let's do this version since it's easy
for now.
While here I changed things so that `generate_deployment_refs()` now just uses
`_set_ref_immediate()` rather than requring a txn.
Also, AFAICS there was no test coverage of `generate_deployment_refs()`; I tried
commenting it out and at least `admin-test.sh` passed. Add some coverage of this
- I verified that with this commenting out bits of that function cause the test
to fail.
Closes: #1132
Approved by: jlebon
I resisted trying to do anything invasive here like fd-relative porting as our
coverage is weak. But this was all straightforward porting to decl-after-stmt
style.
Closes: #1153
Approved by: jlebon
There were some important ones there like a random `syncfs()`. The remaining
users are mostly blocked on the "fstatat enoent" case, I'll wait to port those.
Closes: #1150
Approved by: jlebon
A lof of the functions here are async and have nontrivial exits, but these ones
are all sync were straightforward ports.
Not prep for anything, just chipping away at porting.
Closes: #1146
Approved by: jlebon
We have `ot_ensure_unlinked_at()` for the "ignore ENOENT" case, and
`glnx_unlinkat()` otherwise. Port all in-tree callers to one or the other as
appropriate.
Just noticed an unprefixed error in the refs case and decided to do a tree-wide
check.
Closes: #1142
Approved by: jlebon
I'd mostly been skipping the GPG functions due to lack of autoptr for a few
things, but I noticed these bits were straightforward.
Closes: #1136
Approved by: jlebon
The vast majority of invocations of `ot_gpgme_error_to_gio_error()` were paired
with `g_prefix_error()`; let's combine them for the same reason we do
`glnx_throw_errno_prefix()`. For the few cases that don't we might as well add
some prefix.
I also changed it to `return FALSE` in prep for more style porting.
Closes: #1135
Approved by: jlebon
I was reading the pull-local command docs and realized it was somewhat unclear
that `--untrusted` *only* applied to local repo pulls; in other words that we
always treat non-local pulls as untrusted.
Tweak the docstring, and add tests that verify this explicitly.
Closes: #1130
Approved by: jlebon
However, they weren't showing up in the output HTML and I have
no idea why; I looked at what we're doing and it looks close enough
to what's going on in `GDBusConnection` that I was using as a reference.
I'm not going to spend a lot of time to debug it right now.
Closes: #1140
Approved by: jlebon
Followup from previous patch - we can now centralize the sysroot loading.
Besides the obvious cleanup value, this is also prep for dropping an
`ostree_sysroot_get_path()` user.
Closes: #1123
Approved by: jlebon
This is exactly analogous to the `ostree init` case where
we have `OSTREE_BUILTIN_FLAG_NO_REPO` to avoid trying to load
a repo when we're creating one.
Let's avoid the pointless sysroot for `init-fs`; among other
things this will then let us do `ostree_sysroot_load()` inside
the argument parsing, and drop it from every other user.
Closes: #1123
Approved by: jlebon
Rather than calling `ostree_sysroot_get_path()`, which I'd like to deprecate for
the same reason as `ostree_repo_get_path()`.
Closes: #1123
Approved by: jlebon
Colin Walters