This effectively reverts 66425daf2c.
The commit assumes that if the network interface has multifunctions,
then the function fields of the pci devices under the same PCI bridge
device are unique.
But it seems not, at least on some setups. See issue #28929.
Let's revert the change, and always refuse to set slot base naming if
a PCI bridge is detected.
Fixes#28929.
Even more than with the previous commit, this is not a trivial function
and there's no reason to believe this will actually be inlined nor that
it would be beneficial.
The function isn't necessarily fast (it's O(n)), and there's no reason
to have it defined as inline function, since it's neither fast, nor
entirely trivial.
This is preparation for #28891, which adds a bunch more helpers around
"struct iovec", at which point this really deserves its own .c/.h file.
The idea is that we sooner or later can consider "struct iovec" as an
entirely generic mechanism to reference some binary blob, and is the
go-to type for this purpose whenever we need one.
Before the split, it made sense to assert, as checks were on setup.
But now these come from deserialization, and the fuzzer hits the
asserts, so simply return an error instead.
We have a test where we compare the results from nftw() and our own
resurce_dit_at(). nftw() skips a dangling symlink when running under mkosi and
the test fails. I don't understand why nftw() does that, but in our code we
don't need to test and care about the details of nftw(), which we don't use,
outside of the one test, so let's just skip symlinks in the test.
Closes#29603.
With coverage builds we disable Protect{Home,System}= via a service.d
dropin in /etc, which has, unfortunately, higher priority than our
transient systemd-run stuff. Let's just skip the affected tests in that
case instead of making the test setup even more complicated.
When building with mkosi I would get the following:
[1/477] Generating version.h with a custom command
fatal: detected dubious ownership in repository at '/work/src'
To add an exception for this directory, call:
git config --global --add safe.directory /work/src
and then the tag would be generated as 'v254-'. This is obviously some problem
with the setup, but we should handle this gracefully. Let's fall back to 'v254'
instead.
In the case where we have a repo but no tags, use --dirty=^ too, as in the case
with tags.
I tested four cases:
- normal checkout
- checkout with .git removed
- checkout with .git chowned to root
- checkout wiht all tags removed
As reproducing it is actually pretty easy, with the benefit of hindsight:
~# systemd-run -P -p MountImages="/this/should/definitely/not/exist.img:/run/img2\:3:nosuid" false
Running as unit: run-u42.service
free(): double free detected in tcache 2
This list is owned by ExecContext, which is cleaned up when sd-executor
fails, but it is also cleaned up when namespace setup exits, so we get
a double free.
Fixes https://github.com/systemd/systemd/issues/29610
Follow-up for bb5232b6a3
If there is no libseccomp support, just ban the entire syscall instead
so wrappers will fall back to older, supported syscalls.
Also reflect all of this in `test-seccomp.c`.
varlink_collect is meant to be used when the client is willing to wait for the reply from the varlink method, much like varlink_call.
However, unlike varlink_call, it allows the client to collect all "more" replies that may be sent by method before the "final" reply is enqueued.
It aggregates all of these replies into a json variant array that it returns to the client.
Lennart Poettering
Luca Boccassi
Daan De Meyer
Joerg Behrmann
Yu Watanabe
Zbigniew Jędrzejewski-Szmek
Frantisek Sumsal
Mike Yuan
Arseny Maslennikov
Arthur Shau