Sometimes we need to fix minor issues like changing the key or fixing other small problem on live OS. We can ship this as a part of BSP package, but its handling should be made easy.
This downloads script from CDN, verify its signature and executes after apt upgrade starts installing packages.
* Record patch output to syslog
* Enable armbian-live-patch as additional service and run patch mechanism at startup
* Additional security check
> tl,dr: add 4 small-ish UEFI-like kernels, with Microsoft patches & fixes, for use with Microsoft WSL2 on x86/arm64 and 6.1.y/6.6.y
- the boards are UEFI derivatives, using a common `microsoft` vendor include to modify
- `KERNELPATCHDIR`/`LINUXFAMILY` (for now, we don't want those patches in regular UEFI builds / .debs)
- disable `EXTRAWIFI` (kernel is for a VM, will never have wifi so doesn't need any drivers)
- `LINUXCONFIG`, so we can use Microsoft's own monolithic kernel, required for WSL2 (their initrd is a mistery)
- really, what we're mostly interested right now are the kernels (in the future we might have an "Armbian" WSL2 app in the Microsoft Store)
- `current` `6.1.y`:
- rebased from https://github.com/microsoft/WSL2-Linux-Kernel/tree/linux-msft-wsl-6.1.y onto real 6.1.y
- using Microsoft's `.config` exactly (monolithic, there are no `=m`'s)
- `edge` `6.6.y`:
- also from https://github.com/microsoft/WSL2-Linux-Kernel/tree/linux-msft-wsl-6.1.y but rebased onto 6.6.y
- using updated Microsoft's `.config` (monolithic, there are no `=m`'s)
- dropped 2 of 6.1.y's patches that were actually upstreamed in the meantime:
- `mm-page_reporting-Add-checks-for-page_reporting_order-param` - mainlined in https://lore.kernel.org/all/1664517699-1085-2-git-send-email-shradhagupta@linux.microsoft.com/
- `hv_balloon-Add-support-for-configurable-order-free-page-reporting` - mainlined in https://lore.kernel.org/all/1664517699-1085-3-git-send-email-shradhagupta@linux.microsoft.com/
- drop the `arm64: hyperv: Enable Hyper-V synthetic clocks/timers` patch, since it causes asm breakage on 6.6.y
- a shame, but I tried and can't fix it myself - @kelleymh ?
- add my own patch to fix:
- `1709-drivers-hv-dxgkrnl-restore-uuid_le_cmp-removed-from-upstream-in-f5b3c341a.patch` due to https://lore.kernel.org/all/20230202145412.87569-1-andriy.shevchenko@linux.intel.com/ landing in 6.6
- `1710-drivers-hv-dxgkrnl-adapt-dxg_remove_vmbus-to-96ec29396-s-reality-void-return.patch` to adapt to
https://lore.kernel.org/all/TYCP286MB2323A93C55526E4DF239D3ACCAFA9@TYCP286MB2323.JPNP286.PROD.OUTLOOK.COM/
* rockchip64: refresh kernel config of current and edge branch
* rockchip64: enable ARM_PSCI_CPUIDLE for current and edge branch
---------
Co-authored-by: Igor <igor@armbian.com>
Co-authored-by: Igor Pecovnik <igor.pecovnik@gmail.com>
* fix: vim3: blacklist simpledrm kernel module, as it conflicts with panfrost
* fix: meson-g12b: blacklist simpledrm kernel module, as it conflicts with panfrost
* fix: meson-g12a/meson-sm1: blacklist simpledrm kernel module, as it conflicts with panfrost
---------
Co-authored-by: Igor <igor@armbian.com>
- this should make drivers hash consistent, at the expense of being moar tiresome
- _any_ changes at _any_ patches or drivers-related bash code will cause _all_ kernels to be rebuilt
- opposed to "some changes caused all kernels to be rebuilt"
- function defined in FETCH_FROM_REPO_CALLBACK_IF_REF_MUTABLE is called if reference is mutable
- which is anything that's not `commit:` or `tag:`
- and really there's no guarantee that `tag:` is immutable, but lets assume
- also, for extension style hook `post_uboot_custom_postprocess`, don't do it 3 times, just once
- this commit will cause rebuild of all u-boots
- and that's a good thing, many custom changes in family code (eg ddr change in rk322x) were being ignored
- thanks @paolosabatino
* Change conf to csc
* Bring back to stable
* Remove BPI Pro
* Move media kernel to csc
* Move Opi prime back as its used as a test subject
* Retire Clearfog pro
* Remove Rock64
* Retire Helios4 too
* Move nanopi r5c from csc to wip as we have a volonteer
* Promote recent Khadas boards.
* Change status of Opi one as maintainer responded
Winter, Spring, Summer, Autumn.
This can only apply in northern hemisphere countries. Southern hemisphere is the opposite months.
First quarter, second quarter, third quarter, fourth quarter are universal options.
- From https://github.com/edk2-porting/edk2-rk3588/releases/tag/v0.9.1
- For example, for `rock-5b_UEFI_Release_v0.9.1.img`
- `UEFI_EDK2_BOARD_ID` is `rock-5b`
- **Important**: this has no effect unless optional extension is enabled
- to test this out:
- make sure to read carefully the instructions at https://github.com/edk2-porting/edk2-rk3588/blob/master/README.md
- suppose you previously built the regular u-boot version with:
- `./compile.sh BOARD=rock-5b BRANCH=legacy RELEASE=jammy`
- now you can build the UEFI version with:
- `./compile.sh BOARD=rock-5b BRANCH=legacy RELEASE=jammy EXT=uefi-edk2-rk3588`
- write the produced image to SD or eMMC and boot it
- use normally, or enter "UMS" mode by selecting that option in the grub menu
- You can write image to eMMC, boot it, enter UMS, write it again to NVMe, reboot, press <ESC> in UEFI and boot from NVMe
- Also works if .img is written to eg USB stick, and UEFI edk2 is separately deployed to SPI flash
- **Important**: make _absolutely_ sure you are able to force Maskrom mode (by shorting pins, pressing buttons, etc) before writing an UEFI image to eMMC; UEFI has no RockUSB ("Loader mode" support) at all. The new "initramfs-usb-gadget-ums" extension hopes to address this, but it is not guaranteed to work.
- this extension is _100% optional_ and shouldn't adversely affect any builds if not enabled
- requires `UEFI_EDK2_BOARD_ID` to be set in board file, so we know which UEFI/edk2 build to use
- it finds the latest edk2 version from GitHub automatically (currently `v0.9.1`)
- it downloads (and caches) the correct edk2 build image automatically
- if forces certain aspects of the image:
- must use GPT partitioning
- must NOT use a separate /boot partition
- it _disables_ the building and deploying of u-boot _completely_ (thus blobs etc are from edk2)
- it creates a GPT `"uboot"` partition pointing to edk2's FIT, required by SPL
- this extension:
- automatically enables 'grub-with-dtb'
- automatically enable 'initramfs-usb-gadget-ums', to compensate for lack of ums/rockusb since we dont have u-boot anymore
- this optional extension adds an initramfs script that:
- enumerates and filters all block devices
- exposes each device as an UMS (USB Mass Storage) in an USB Gadget
- loops forever with info (board never boots)
- the idea here is to compensate for UEFI's lack of "ums" or "rockusb" mode that's present in u-boot
- it also allows to expose USB/NVMe devices that might or not be detected by bootloader, if the kernel works
- this allows user to enter UMS mode from GRUB
- for usage with the `initramfs-usb-gadget-ums` extension
- has no effect unless `initramfs-usb-gadget-ums` extension is enabled together with `grub-with-dtb`
