Age | Commit message (Collapse) | Author |
|
|
|
Try autodetect the EFI directory instead of hardcoding it to /boot/efi.
This is needed when /boot is the EFI directory and no encryption is
used.
|
|
|
|
Previously, setup-disk could fail when installing over an existing
identical partition table. The additions ensure that a clean disklabel
is created and existing partitions are wiped.
Signed-off-by: Edin Taric <unicorn@regrow.earth>
|
|
Ref: https://gitlab.alpinelinux.org/alpine/alpine-conf/-/merge_requests/106
Fixes https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10476
|
|
Use an option to setup-disk to write the selected diskmode instead of
using an undocumented /tmp/alpine-install-diskmode.out
Adjust setup-alpine to use this, and make sure that we write the
diskmode when installing to a mounted root.
fixes https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10444
|
|
Only give warning about MBR on x86 and x86_64 and when in legacy bios
mode. (syslinux or grub).
Make sure grub is installed if needed, by using select_bootloader_pkg.
|
|
Some modules wants non-existing firmware like b43 and b43legacy
fixes https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10530
|
|
|
|
|
|
Consider my laptop with linux-lts on UEFI with Secure Boot:
System.map-lts.....4 MiB
initramfs-lts.....36 MiB
vmlinuz-lts........8 MiB
linux-lts.efi.....44 MiB (signed Unified Kernel Image)
linux-lts.efi.bak 44 MiB (backup of the previous)
Total............136 MiB
256 MiB might be enough, but what if the user wants to experiment with
more kernel variants? It would be still too tight. On the other side,
if the user has only 1 GiB disk (e.g. ncopa uses 1GiB images for testing),
then even 256 MiB would be too much.
Originally, I just increased the default to 512 MiB, but ncopa pointed
out that it's too much for small disks and he personally uses 1GiB
disks (images) for testing. For this reason, I chose a more
sophisticated approach that takes into account the size of the disk.
* 512 MiB if disk > 8 GiB
* 264 MiB if disk > 2 GiB
* 160 MiB otherwise
And the size is never lower than the approximate minimal allowed size of
the FAT32 filesystem for the disk's block size (32 MiB for 512 bytes
blocks, 260 MiB for 4096 bytes blocks).
Issue reported by Andrey Maslyuk <andrey@maslyuk.net> (via email).
|
|
0 is not a safe fallback for a block size.
|
|
VMD is Intel's Volume Management Device. Some Intel-based machines
have a BIOS/UEFI option to present a NVME SSD via either VMD or as
a "native" NVME device.
If Alpine is installed on a NVME device when it is presented via
VMD then once the Alpine ISO/USB boots the 'vmd' kernel module will
be loaded by /etc/init.d/hwdrivers and Alpine will see a
/dev/nvme0n1 device to which setup-alpine can install successfully.
However, once the newly installed system is then booted from the
NVME device the initramfs' init will fail to find the rootfs to
mount/boot from as, without the 'vmd' module being loaded, no NVME
device will be visible.
This MR ensures that the 'vmd' kernel module is added to the cmdline
modules list whenever the 'vmd' has been automatically loaded by
hwdrivers init.d when booting Alpine prior to running
setup-alpine/setup-disk.
This MR works together with
https://gitlab.alpinelinux.org/alpine/mkinitfs/-/merge_requests/112
(which ensures that 'vmd' present in the initramfs).
|
|
|
|
|
|
fixes use with yash.
|
|
and add test from missing block device
|
|
we dont dont that for real when testing
|
|
|
|
Use an env var SETUP_DISK_TESTFUNC to run individual functions for unit
testing purposes.
|
|
not used since commit 427a14cb8d80 (setup-disk: remove grsec traces)
|
|
prefix /sys/... with $ROOT so we can test it
|
|
Some shells may apply word splitting after expanding variable in
`local a=$1`. I know just about yash, but since `local` is supposed to
be a builtin *command*, the "correct" behaviour of common shells is
actually a special case for `local` - it doesn't apply for other
commands. See https://osdn.net/projects/yash/ticket/46041.
|
|
'echo -n' and 'echo -e' are not portable, not all commonly used shells
support both of them (in the same way). 'echo -e' is not even defined
in POSIX.
https://pubs.opengroup.org/onlinepubs/009604599/utilities/echo.html:
> New applications are encouraged to use printf instead of echo.
> ...
> Conforming applications that wish to do prompting without <newline>s
> or that could possibly be expecting to echo a -n, should use the
> printf utility derived from the Ninth Edition system.
|
|
Currently, /boot becomes 34 MB, which is too small.
Fixes https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10536
|
|
Make code slightly more readable. No functional changes
|
|
Fix a hang that happens with interactively selecting 'none' with efi
mode.
Fixes commit 851247d96340 (setup-disk: calculate size of ESP parition
for grub)
|
|
fixes https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10501
|
|
|
|
If you pass -m none, just return immediately: For non-interactive diskless setups.
|
|
fixes https://gitlab.alpinelinux.org/alpine/aports/-/issues/13839
|
|
fixes https://gitlab.alpinelinux.org/alpine/aports/-/issues/13861
|
|
Retry create luks device if the entered password does not match rather
than exit with error.
fixes https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10514
|
|
Fail early if disk is too small or similar
|
|
some devices, like NVME may have a physical block size of 4096, in which
case FAT32 needs minimum 260MB.
Use physical_block_size to calculate the size for ESP
|
|
Grub's efi files are just a few MB, so we can reduce the size of the
ESP. We also enforce FAT32, which requires minimum 32MB. Apparently many
UEFI firmware implementations only support FAT32.
https://wiki.osdev.org/EFI_System_Partition#Format
|
|
|
|
fixes https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10512
|
|
Make sure we add raid to initfs when we have encrypted raid.
|
|
Some nvme devices has problems getting detected by nlplug-finds. Work
around it by always load nvme early when needed.
|
|
fixes https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10498
|
|
|
|
|
|
|
|
|
|
we have nto supported grsec for a while. Remove everything about grsec
and PAX.
|
|
It is more common than not to prefer tmpfs for an Alpine installation.
|
|
https://www.raspberrypi.com/documentation/computers/config_txt.html#model-filters
Related: alpine/aports!31953
|
|
Signed-off-by: flavorplus <flavorplus@users.noreply.github.com>
|
|
Commit ec0c7a7 changed the sed expression for PREFIX injection but
forgot to update the scripts.
|