For some reason, the aarch64 6.1 debian kernel was not compiled with
CONFIG_KALLSYMS_ALL=y, and the locations of globals are not available in
kallsyms. To attach breakpoints to these symbols in the test pipeline, we need
to read them from System.map and apply the kaslr offset to get their real
address. The System.map file is not provided in the kernel package by default,
so we need to extract it from the corresponding debug package.
- .github: pull the corresponding debug packages down as well as regular
kernels
- test: attach the perf_event_bp test breakpoint to the modprobe_path address
in kallsyms if present, or by applying the kaslr offset to the System.map
address if not found
- xtask: preferentially extract the System.map file from the debug package, if
available
Bundle handling of Debian kernel archives into xtask so callers can pipe
the raw `.deb` paths straight into `cargo xtask integration-test vm …`.
The driver now extracts each archive into `<cache>/kernel-archives`,
locates the matching `vmlinuz-*`, `lib/modules/*`, and config files, and
feeds those into the initramfs build without requiring the user to
pre-run dpkg/tar. With this in place we drop
`.github/scripts/find_kernels.py`, simplify AGENTS.md/CI instructions to
use `find test/.tmp -name '*.deb'`, remove the gnu-tar requirement we no
longer need, and add `tar` as a workspace dependency for the extractor.
This allows us to run virtualized integration tests on macOS hosts.
Bump Ubuntu to 24.04 because we seem to be getting miscompilation on
x86_64 otherwise (when using `x86_64-linux-musl-gcc`). Add `apt install
liblzma-dev` since it doesn't seem to be present in ubuntu-24.04.
We've recently added an xtask to bpf-linker (aya-rs/bpf-linker#282),
which resulted in multiple binary targets. Therefore, bpf-linker has
to be installed with the following command:
```
cargo install --git https://github.com/aya-rs/bpf-linker.git bpf-linker
```
The last argument (`bpf-linker`) specifies the binary target.
The init module contains a small init system for running our integration
tests against a kernel. While we don't need a full-blown linux distro,
we do need some utilities.
Once such utility is `modprobe` which allows us to load kernel modules.
Rather than create a new module for this utility, I've instead
refactored `init` into `test-distro` which is a module that contains
multiple binaries.
The xtask code has been adjusted to ensure these binaries are inserted
into the correct places in our cpio archive, as well as bringing in the
kernel modules.
Signed-off-by: Dave Tucker <dave@dtucker.co.uk>
- Document the need for external rustfmt invocation.
- Remove reexports.
- Remove `write_to_file`.
- Avoid allocating strings when using bindgen to write bindings.
Turns out this was actually being used through magic.
Remove mips since it is a tier 3 target that is only supported on
nightly and dtolnay/rust-toolchain doesn't seem to handle installing
targets without std.
Exclude xtask since it depends on `ring` which doesn't build on all
targets.
Fix compilation on armv7 which had rotted.
This reverts commit d92fc95c39.
This can be done externally. Do so in CI.
This is an attempt to resolve the inconsistency between CI and local
rustfmt in the generated bindings.
Restore running CI on generated branches; the presence of a PR is
apparently not enough.
It is especially common to run CI on non-main branches in forks. This
shouldn't have any impact on how many CI jobs we run in the main aya
repo if contributors follow a fork-centric workflow.
libbpf is a submodule, so its version is already effectively pinned.
`xtask codegen` also runs `git submodule update` which reverts the
action of updating to `origin/HEAD`. Remove the cruft.
We're seeing 429 from Github trying to download gen_init_cpio, so cache
it using actions cache. Since I'm here add this for kernel images as
well to save time waiting on slow Debian servers.
Unwinding gives us more information, so we shouldn't disable it
globally. It is already disabled for BPF targets via the target configs
in rustc itself.
This complicates the clippy invocation somewhat, so put it in a shell
script for developer as well as CI use.
I did this for arm64 because we'd get a black screen without it but I
have now confirmed that console=ttyAMA0 solves that problem.
I don't remember why I did it for x86.
Instead of relying on Homebrew for macOS (which ships older LLVM
versions) and `apt.llvm.org` for Linux (which often has bugs at the
packaging level), we now use the tarball from Rust CI to provide
`libLLVM`. This ensures it always matches the version used by the latest
Rust nightly.
This removes our reliance on homebrew shipping LLVM versions matching
those used by rustc.
Friday Ortiz
Tamir Duberstein
dependabot[bot]
Michal Rostecki
Dave Tucker
WANG Rui
Michal Rostecki