I currently have no idea whether the test kmod can actually unload
successfully because the vmtest kernels don't support unloading modules.
It might also be useful to test unloading other modules in the future.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
drgn is currently licensed as GPLv3+. Part of the long term vision for
drgn is that other projects can use it as a library providing
programmatic interfaces for debugger functionality. A more permissive
license is better suited to this goal. We decided on LGPLv2.1+ as a good
balance between software freedom and permissiveness.
All contributors not employed by Meta were contacted via email and
consented to the license change. The only exception was the author of
commit c4fbf7e589 ("libdrgn: fix for compilation error"), who did not
respond. That commit reverted a single line of code to one originally
written by me in commit 640b1c011d ("libdrgn: embed DWARF index in
DWARF info cache").
Signed-off-by: Omar Sandoval <osandov@osandov.com>
This will be useful for testing different architectures (e.g., by
creating a rootfs with debootstrap) or different userspace environments.
For example, we can test busybox with something like:
mkdir busyboxroot busyboxroot/bin busyboxroot/tmp
cp "$(which busybox)" busyboxroot/bin
busyboxroot/bin/busybox --install busyboxroot/bin
python3 -Bm vmtest.vm -k '*' -r busyboxroot
Signed-off-by: Omar Sandoval <osandov@osandov.com>
This still isn't enough to be fully generic since since some
architectures don't support kexec_file_load, especially on older
kernels, and some architectures can't use the vmlinuz directly (e.g.,
AArch64 may need it to be decompressed). But, this is at least cleaner
than hard-coding the number for x86-64.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
We don't specifically need BusyBox; we just need a reasonable Linux
userspace, which we can assume is already available on the host, whether
it's coreutils+util-linux, BusyBox, or something else.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
The test command does this, and I always end up doing it when I'm doing
manual testing with the vmtest.vm CLI, so let's just do it by default.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
We currently don't have any tests for the BPF helpers or the
bpf_inspect.py tool. As a result, the latter is broken on newer kernel
versions. Before we can add tests, we need the vmtest kernel to support
BPF.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
For local testing with vmtest, we just want an extracted kernel package,
so save the trouble of compressing the package only to extract it and
allow vmtest.kbuild to output the directory directly.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Linux kernel commit 31cb50b5590f ("kbuild: check static EXPORT_SYMBOL*
by script instead of modpost") (in v5.19-rc1) added this script to the
build process, and the latest vmtest kernel build failed without it.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Since most Linux distros enable this, we should make sure we test it. It
was added for SLUB in Linux kernel commit 2482ddec670f ("mm: add SLUB
free list pointer obfuscation") (in v4.14), so we'll still get test
coverage of the non-hardened codepath while 4.9 is around.
It was also added for SLAB in Linux kernel commit 3404be67bf73
("mm/slab: expand CONFIG_SLAB_FREELIST_HARDENED to include SLAB") (in
v5.9), although that currently doesn't change the in-memory data
structures.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Several have snuck in since the last time I did this in commit
5541fad063 ("Fix some flake8 errors"). Prepare for adding flake8 to
pre-commit by fixing them.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Now that commit d999703f94 ("vmtest: add kernel module build
dependencies to kernel packages") added the files necessary to build a
test kernel module, add the module (currently a stub) and the
scaffolding necessary to build and load it.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
We have several helpers that are difficult to test because there are no
userspace APIs to test them against. Radix trees, slab objects, and
hlists are a few examples. For lists and rbtrees, luckily we can find
instances of those data structures exposed to userspace in some way, but
this is somewhat brittle.
We'd like a way to be able to unit test directly against kernel code
that sets things up in a way that is easy to test against. The easiest
way to do this is with a custom kernel module.
There are two options for how to enable this:
1. Build the custom kernel module as part of the vmtest kernel build and
package it with the vmtest kernel package.
2. Include the artifacts needed for kernel module builds in the vmtest
kernel package, then build the kernel module when running tests.
The latter makes the kernel packages significantly larger (on a build of
v5.18-rc6, 46M -> 53M compressed, 173M -> 217M decompressed), but it has
the huge advantage that it does not require a vmtest kernel rebuild to
add or modify tests. In order to optimize for making it easy to add new
helpers with test cases, this is the approach I chose.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Now `python3 -m vmtest.manage -K --no-build` can be used to get the list
of latest kernel releases that need to be built. Also rename --dry-run
to --no-upload since --no-build is also a dry run in a way.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
If gh.download() in _download_kernel() raises an exception (e.g.,
because of an HTTP 403 error due to rate-limiting), then we will try to
exit the subprocess context managers for zstd and tar.
subprocess.Popen.__exit__() closes stdin if it is a pipe and then waits
for the subprocess. The context managers are exited in reverse order, so
first we'll try to wait for tar. But, that will never exit because its
stdin (piped from zstd) is still open. Fix it by always explicitly
closing zstd's stdin once the tar process is running.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
GitHub release assets don't allow "~" in filenames; they are changed
into ".". This means that we can't take advantage of the fact that ~
version sorts before anything, so add a sub-version to the localversion
which is 1 for the default flavor and 0 for everything else.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
The vmtest build workflow is running out of disk space because the build
directories are big, and we have a lot of them now. Default to deleting
them if uploading packages, default to keeping them in dry run mode, and
add options to choose one or the other explicitly.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
In order to test the upcoming slab helpers (#166) with the different
slab allocators, we need kernels built with SLAB and SLOB in addition to
the default SLUB. We also have some helpers which depend on SMP vs !SMP.
There will likely be other options to test in the future, like RCU
implementations.
A testing matrix of all of these options would be too big. Instead,
introduce the notion of "flavors": groups of tweaks to the vmtest kernel
configuration. For now, we have three flavors: "default" (which uses
SLUB), "alternative" (which uses SLAB), and "tiny" (which is !SMP,
!PREEMPT, and uses SLOB). As long as the options are orthogonal enough
and each one doesn't have many choices, we won't need many more flavors
than that.
I haven't decided yet exactly how to use these flavors. The easiest
thing to do is probably to run the entire Linux kernel test suite on
every flavor of every version. Since the total runtime is dominated by
boot and loading debugging information, this won't be much slower than
only running targeted tests. That will be sorted out later; for now,
this adds the build infrastructure.
While we're bumping the vmtest localversion, let's also add
CONFIG_CC_OPTIMIZE_FOR_SIZE=y.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Disabling SMP is necessary to work around a bug in QEMU's handling of
the capture kernel, but makes the tests run much slower. However, this
bug only appears to manifest when KVM acceleration is disabled, so the
testing harness has been modified to only disable SMP when this is true.
[Omar: use an environment variable instead of touching a file]
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Signed-off-by: Kevin Svetlitski <svetlitski@fb.com>
Now that the vmtest kernel supports kdump, add a script that can be used
to crash and enter the kdump environment on demand. Use that to crash
after running the normal test suite so that we can run tests against
/proc/vmcore. vmcore tests live in their own directory; presently the
only test is a simple sanity check that ensures we can can attach to
/proc/vmcore.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Signed-off-by: Kevin Svetlitski <svetlitski@fb.com>
We can avoid the need for the kexec tool if we load the kdump kernel
ourselves, which is much easier with kexec_file_load(). Add the config
options to enable it.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
We would like to test drgn against kernel core dumps (e.g., for #129).
One option would be to include some vmcore files in the repository and
test against those. But those can be huge, and we'd need a lot of them
to test different kernel versions. Instead, we can run vmtest, enable
kdump, and trigger a crash. To do that, we first need to enable a few
kernel config options.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
These haven't been running in vmtest since they were added. Enable
cgroup2 in vmtest and rework the cgroup tests to create cgroups that we
can test with.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
We're seeing some hangs waiting for entropy when running tests, so let's
enable the virtio-rng module. While we're doing a rebuild, we might as
well enable cgroups so that we can finally run those tests.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
QEMU warns about the default 8k msize
(https://wiki.qemu.org/Documentation/9psetup#msize). I wasn't able to
measure any performance difference, but bump it to 1MiB to silence the
warning.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Mainly unused imports, unused variables, unnecessary f-strings, and
regex literals missing the r prefix. I'm not adding it to the CI linter
because it's too noisy, though.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
I incorrectly assumed that the stable repo would have the mainline tags
soon after they were released, but this is not the case. We also need to
check the mainline repo for tags and fetch them from the mainline repo.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
The argument to download_kernels() is an iterable of kernels, but we're
passing it a single kernel string.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
In Python before 3.8, tempfile.TemporaryDirectory.cleanup() fails when
the directory doesn't exist. Since we rename the temporary download
directory to its final name, this always fails. Switch to using
tempfile.mkdtemp() directly instead.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
As noted by commit 738261290f ("CI: temporarily disable vmtest"),
vmtest was generating too much traffic to the Dropbox shared folder that
hosted vmtest kernels. Instead, we can store kernel packages as GitHub
release assets. Update the code for downloading and uploading vmtest
assets, and also add a scheduled GitHub action to build new kernels
every Monday so I don't have to remember to do it manually.
This also drops vmtest support for 5.6-5.9, which now fail to build with
newer binutils due to the issue fixed in Linux kernel commit
1d489151e9f9 ("objtool: Don't fail on missing symbol table").
Signed-off-by: Omar Sandoval <osandov@osandov.com>
The install directory contains redundant copies of the modules already
in the build tree and built package, so clean it up on success.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Travis CI is no longer offering free open source CI, so migrate to
GitHub Actions. The only downside is that GitHub Actions doesn't support
nested virtualization, but we can work around that by falling back to
slow emulation.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
We currently build with CONFIG_MODULES=n for simplicity. However, this
means that we don't test kernel module support at all. Let's enable
module support. This requires changing how we distribute kernels. Now,
the /lib/modules/$(uname -r) directory (including the vmlinux and
vmlinuz) is bundled up as a tarball. We extract it, then mount it with
VirtFS, and do some extra setup for device nodes. (We lose the ability
to run kernel builds directly, but I've never actually used that
functionality.)
Signed-off-by: Omar Sandoval <osandov@osandov.com>
Black was recently changed to treat a trailing comma as an indicator to
put each item/argument on its own line. We have a bunch of places where
something previously had to be split into multiple lines, then was
edited to fit on one line, but Black kept the trailing comma. Now this
update wants to unnecessarily split it back up. For now, let's get rid
of these commas. Hopefully in the future Black has a way to opt out of
this.
Signed-off-by: Omar Sandoval <osandov@osandov.com>
