| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|
|
|
|
|
|
|
|
|
|
| |
Goal reached: stage2 produced for raspi-arm64 + debian-amd64
merged into one 'dual-2GiB-01.img' using 'loop_rootfs'
were bootable on
- raspi4
- qemu amd64
- Intel NUK amd64 i6 w/ EFI (vfat) and other partitions
The 2G test image was created using scripts/make_dualboot_image.sh
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
partition: fstype + 'a magic file id'
First fstype partition is not good enough, especially if having a vfat EFI partitions
and perhaps other systems installed.
Hence find_partition:
- for all blk-partitions
- for all matching desired fstype, mounted
- if 'loop_rootfs.id' exists, done!
To ensure logging won't interfere with returned echo'ed strings (stdout),
"overloaded" the log facility with 'mylog_.." using stderr output.
|
|
|
|
| |
faster inflater
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
'make_dualboot_image'
'make_dualboot_image' creates a new raw image file:
- creating one vfat partition covering whole size
- formating partition with vfat
- provision GRUB boot-code (expecting 'grub' folder on vfat partition)
- copy custom files to vfat partition
using image file handling via NBD (updated imagetool.sh).
The custom files are sourced from two sdcard diskimages folder
generated with our pi-gen branch:
- raspi arm64
- debian amd64
Simply copied both content into one merge folder,
passed as last argument 'provisioning-dir' to this script.
+++
Grub file handling takes care of both boot-code image locations,
446 bytes in MBR and the MBR gap from [512 bytes - partition-start[
|
| |
|
| |
|
| |
|
|
|
|
| |
/etc/resolv.conf will be defined via dhcpcd directly via dhcp
|
|
|
|
| |
provided!
|
|
|
|
|
|
| |
- don't use UUIDs but file-id
- preload modules via core.img (within MBR-gap)
- Don't use installation specific load.cfg, remove it.
|
|
|
|
| |
mapping /etc/fstab
|
| |
|
|
|
|
|
|
|
|
|
|
| |
functions from new lib_nbd_img_handling)
imagetool.sh
- new connect* and disconnect
- mount/umount stateless, multiple usage
- lib_nbd_img_handling used as base implementation for qcow2_handling and imagetool.sh
|
| |
|
| |
|
|
|
|
| |
on_chroot (otherwise no arm64 emulation)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
loop_rootfs using grub (WIP)
- Generic packages: nn-packages[-RELEASE] (using the optional '-RELEASE' variant)
- System packages: nn-packages-sys-(raspi|debian)[-RELEASE] (using the optional '-RELEASE' variant),
allow to distinguish not only for the RELEASE but the actual underlying distribution.
- On debian amd64, all required usb-storage etc modules are loaded for loop_rootfs/initrd
- loop_rootfs works on Raspi (Proprietary Bootloader) as well as on Debian (Grub Bootloader)
- loop_rootfs uses the first vfat block-device partition for /boot
- /boot will be mounted by loop_rootfs and handed over to new root-mount-point,
this avoids issues of using an undefined /boot partition (but the first)
- rotatelog_init_rootfs: Rotates /boot/init_rootfs.log if having more than 500 lines
- FIX export-image: Copy the rootfs.img to the exported full image,
which allows testing the image directly via qemu .. etc.
|
|
|
|
|
|
|
|
| |
using a simple rootfs image
Tested compilation of direct_bt for aarch64, resulting in correct binaries in just ~5 minutes.
Issues: Running the Java unit tests fails sadly, might need to investigate OpenJDK11 + qemu-aarch64.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
before resize2fs.
The install image's partition has been downsized to match the shrunken rootfs for efficient storage and copy to sdcard.
Hence 'resize2fs_once' must also resize the parition to the end of the sdcard, then resize2fs for maximum size.
|
|
|
|
| |
for 2nd parition)
|
|
|
|
| |
for custom tasks
|
|
|
|
| |
Cleaned README.md
|
|
|
|
| |
selected (default); Disable apt-cache for all configurations
|
|
|
|
| |
networking fs environment
|
|
|
|
| |
(expect default man and /usr/share)
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
exists
Testing with bullseye, we see some divergence, i.e. python3 for python[2] and certain packages are no more available.
|
|
|
|
| |
'has_journal' for writable image
|
|
|
|
| |
.. etc
|
|
|
|
|
|
| |
initial writable rootfs
Adding/renaming build scripts accordingly
|
| |
|
|
|
|
| |
powertop
|
| |
|
|
|
|
| |
machine to rebuild initrd.img
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Ensure loop_rootfs log output persists across boots and hence fallback,
otherwise we can't validate the 'fallback reboot' stage.
We also need to document its actions.
This takes 'quiet' into account, i.e.
- if 'quiet' or 'log_output' via 'debug' is set: just redirect to file
- else use 'tee' to copy to console as well
Further a potential 'debug' 'log_output',
i.e. '/run/initramfs/initramfs.debug',
is made persistent (appended) to '/boot/init_debug.log'.
Since Busybox shell has no Process Substitution (Bash),
we manually create a PIPE and spawn the tee or print process.
Further `tee_date` and `add_date` is being used
to print the UTC timestamp upfront.
|
|
|
|
|
|
| |
Since no console is available, fsck would fail anyways.
Tested w/ ext4 rootfs file
|
| |
|
|
|
|
|
|
| |
integrity check if 'squashfs'
Also have initrd loop_rootfs check_fsfile() be more verbose for documentation.
|
|
|
|
| |
kernel
|
| |
|
|
|
|
|
|
| |
over all files
zip name: ${IMG_FILENAME}${IMG_SUFFIX}.sdcard.zip
|