The growth of binaries over time and the inclusion of new features
filled the available boot partition space, so that the kernel+initrd
almost couldn't fit twice anymore as required for updates. We employed
workarounds such as wrapper scripts for ignition, afterburn and other
binaries so that they are loaded from /usr. However, this was still not
enough and we would have to do the same for (network) kernel modules and
firmware. To avoid making this ever more complex we can use a dedicated
initrd focused on loading the full initrd from /usr and then this full
initrd can use dracut as before and even drop all the workarounds we
accumulated.
Generate a minimal initrd to use instead of the full bootengine initrd.
The bootengine initrd gets stored as squashfs on /usr. The minimal
initrd still includes the early_cpio for amd64 microcode updates.
We have a fixed list of modules or module directories to include, only
focused on loading /usr and any emergency console interaction. This
requires also checking for module dependencies to copy over.
The busybox, veritysetup, and kmod binaries are needed and get their
required libraries resolved and copied over. They are not static and
use shared libraries which should be ok for now. The resulting vmlinuz
file is 27 MB for amd64, down from ~60 MB, so we have enough room to
include more kernel modules and so on for the next years while we also
grow the boot partition and wait for users to redeploy until we can rely
on a larger boot partition and eventually drop the minimal initrd again.
Pulls in https://github.com/flatcar/bootengine/pull/110 for the
minimal initrd script and https://github.com/flatcar/seismograph/pull/12
for making the device mapper discovery for the "rootdev" command more
reliable.
This also requied a backport of a kernel patch from 2017 that exposes
the PARTUUID in the /sys uevent file.
Co-authored-by: James Le Cuirot <jlecuirot@microsoft.com>
Signed-off-by: Kai Lueke <kailuke@microsoft.com>
Python warns about an invalid escaped sequence (\s) within a string -
make the string raw, so python stops treating backslashes as special
characters.
Signed-off-by: Krzesimir Nowak <knowak@microsoft.com>
We invoke zip inside the BUILD_DIR and tell it to output the
compressed file somewhere in BUILD_DIR. This works if BUILD_DIR is an
absolute path, but breaks when it's relative. Since we already are in
BUILD_DIR, tell zip to output the file in the current directory
instead.
Signed-off-by: Krzesimir Nowak <knowak@microsoft.com>
OS-dependent sysexts that ship kernel modules, usually also ship the
files in /usr/lib/modules/*-flatcar/modules.XXX When multiple such
sysexts get activated, depmod files from just one sysext win and other
kernel modules cannot be loaded using modprobe. We get around this by
removing the depmod files from every sysext with kernel modules.
Instead, we set up modprobe hook, which dynamically runs depmod in a
temporary directory on every sysext kernel module activation.
Signed-off-by: Daniel Zatovic <daniel.zatovic@gmail.com>
The pxe disk is a cpio file that, among other things, contains a
squashfs image.
The image has contents of `/usr` directory, so the image's toplevel
directories are `lib`, `lib64`, `share` and so on. The first fix was
to change the `/usr/share/flatcar/update.conf` path in pseudofile
listing to `/share/flatcar/update.conf`. Otherwise mksquashfs started
complaining that `usr` directory does not exists in the image, so some
of the pseudofiles won't be installed.
Second fix is still related to the same file. It already exists in the
image, so the pseudofile wanting to be installed there won't be,
because mksquashfs stopped liking overwriting the files already
present in the image. I added `-e share/flatcar/update.conf` to avoid
adding the file into the image, so that pseudofile can be created
there. It was actually a suggestion printed by mksquashfs.
Following what was done for incus. Removes the subuid/gid from sysext.
Adding subuid/subgid could be done through Ignition:
```yaml
version: 1.1.0
variant: flatcar
storage:
files:
- path: /etc/subuid
append:
- inline: |
core:1065536:65536
- path: /etc/subgid
append:
- inline: |
core:1065536:65536
```
Related: flatcar/Flatcar#1733
This create more issues than it solves:
* override existing subuid / subgid
* not flexible for the end user
* it has to be created only once (while tmpfiles always try to create
those files)
I think Flatcar should not be responsible to create this and it should
be documented on how to do it through Ignition:
```yaml
version: 1.1.0
variant: flatcar
storage:
files:
- path: /etc/subuid
append:
- inline: |
root:1065536:65536
- path: /etc/subgid
append:
- inline: |
root:1065536:65536
```
Signed-off-by: Mathieu Tortuyaux <mtortuyaux@microsoft.com>
I thought cpio was always creating the output directory automatically,
but it was silently failing. It would only extract the next rootfs when
run a subsequent time.
Signed-off-by: James Le Cuirot <jlecuirot@microsoft.com>
The final part of the script differed only the name of the qemu binary
to execute and in network device driver (virtio-net-pci on amd64 vs
virtio-net-device on arm64). virtio-net-pci seems to be working also
on arm64, so simplify the code to avoid repetition.
There's no need to differentiate between amd64 and arm64 boards here
any more. This also adds bootindex=1 option to the -device flag, so we
can pass more secondary disks without affecting the boot order.
This version writes fewer temporary files and tries cpio multiple times
for concatenated archives again.
Signed-off-by: James Le Cuirot <jlecuirot@microsoft.com>
I couldn't take it anymore! The launcher script could not handle paths
outside the script's own directory, and it was driving me crazy. Now
only the default values are relative to the script's directory. Given
paths are relative to the current directory and absolute paths work as
you would expect.
Signed-off-by: James Le Cuirot <jlecuirot@microsoft.com>
On Linux >= 6.10, the first rootfs is an extra ghost rootfs of 336K,
that has a corrupted CPIO.
To overcome this issue, do not fail on `cpio --extract`.
Setting a profile in a newly created sysroot when building native
toolchains broke after an eselect update. Apparently eselect gets the
path to the coreos-overlay repository and then prefixes it with
ROOT. Since ROOT was set to /build/<arch>-usr, the resulting patch was
wrong. Fix this by telling eselect where to find our make.profile
symlink in new sysroot by setting PORTAGE_CONFIGROOT to
/build/<arch>-usr and where to find our profiles by setting ROOT (and
SYSROOT, because it must match ROOT) to /.
