- getprop failures are not only a system-as-root device issue, as it now also happens on OnePlus 5T in recovery when ROM is Pie Open Beta
- this appears to come down to a libc mismatch between recovery and ROM resulting in direct property checks failing but getprop without arguments printing all properties working
- revert most of https://github.com/osm0sis/AnyKernel2/commit/b1524da099efa716616a8eb3357876bca6a73540 since this should remain resolved with the new implementation
- keep the old check and file_getprop fallback for devices without getprop entirely
- add new getprop fallback that parses the full getprop output
- some custom ROMs on some devices (e.g. Lineage on Pocophone) bind mount modules in /system over modules on /vendor, so remount rw at the bind target to ensure the bind source is also remounted and replaced
- toybox mount doesn't accept the /. prefixed paths from find so ensure we use busybox for mounts
Thanks franciscofranco and topjohnwu for the tips!
- adapted from https://github.com/nathanchance/AnyKernel2/commit/3dfba3dcabff3cf1f859276e1851ae35891d8801
- works around odd behaviors in TWRP on system-as-root devices, but retains working fallback using /system/system/build.prop if getprop also fails
- simplify unmount_all stderr redirection
Please see the above linked commit for further details.
Thanks to nathanchance for reporting, the device-specific workaround, and much testing
- remove "kernel" from "Installing" wording since AK2 is also used on recoveries, and multi-partition Treble devices and this could be misleading
- fix getprop build.prop fallback on /system/system devices
- place version text in a file named "version" in the root of the zip and it will display after the kernel.string when flashing
- useful in conjunction with build scripts, i.e. a generated version file not actually part of the kernel AK2 repo
After working on the OnePlus 5/T with custom ROMs on Android 8.1.0, I
have noticed that several of them have modules in /system/lib/modules
with a context of "u:object_r:vendor_file:s0", which goes against the
assumption made in commit 6b11373 ("update-binary: Change how we handle
modules"). Because the context is changed, there is an SELinux denial
when trying to load the module during init.
While it would be nice for ROMs to fix that on their own, I suspect
they are not entirely at fault. Google's documentation does state that
all modules should be loaded from vendor whenever possible so I would
guess Google decided to force *.ko objects to have a vendor_file
context. This is all speculation as I haven't cared to do research into
the subject.
Easiest way around this is to just worry about applying contexts to new
modules. cp does not override contexts so if the module was already
present, we can assume it has the correct context to begin with.
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
As of Oreo, Google is requiring that kernel modules be used whenever
possible and that their location be in /vendor. With Lineage 15.1,
devices without a separate /vendor partition will use /system/vendor,
which symlinked to /vendor.
To make this more flexible and future proof, allow developers to specify
where each module needs to go by setting the path in the modules folder.
If a module needs to go to /system/vendor/lib/modules, then the module
should be placed in modules/system/vendor/lib/modules. Contexts and
permissions will still be updated automatically.
Android documentation: https://source.android.com/devices/architecture/kernel/modular-kernels
Lineage 15.1 commit: https://github.com/LineageOS/android_vendor_lineage/commit/bf38db2a5e358a5449458b3dee2dd9d4015b8147
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
- hide some expected stderr for AK2 mods with no additional ramdisk files
- fix U-Boot support by passing correct repacked ramdisk filename
- fix multi-compression support during repack derp
- fix cleanup on successful flash not working due to double match from cleanuponabort
- fix busybox getting permissions on some devices where chmod does not accept -R to recurse
- use busybox for more possibly non-standard broken recovery shell commands
- silence some more generally expected/inconsequential stderr
- symlinks under /system/lib/modules were getting broken, so follow them instead to replace the target file, avoiding issues with multiple versions of the same module and running out of free space (e.g. OnePlus devices link /system/lib/modules/wlan.ko to /system/lib/modules/qca_cld3/qca_cld3_wlan.ko)
- set contexts on copied modules (following symlinks) to system_file to avoid SELinux issues where unlabeled module files might not be allowed to load
- to make it easier to diagnose issues with patches, image creation or flashing, adding "-debugging" to the end of the AK2 zip filename will enable the creation of a tar.gz of the entire /tmp directory after the installation process (but before cleanup, of course), this includes recovery.log, boot.img, boot-new.img and any other remnants to allow more thorough, direct debugging at a later time
- if it does not exist yet (i.e. hasn't been extracted) then assume cleanuponabort=1 since there wouldn't be anything to see on a failed extraction
- hide any missing file errors from file_getprop here now that we've taken care of that case
"Some older scripts may still use ‘.’ in place of the ‘:’ separator. POSIX 1003.1-2001 does not require support for that, but for backward compatibility GNU chown supports ‘.’ so long as no ambiguity results. New scripts should avoid the use of ‘.’ because it is not portable, and because it has undesirable results if the entire owner‘.’group happens to identify a user whose name contains ‘.’."
Source: http://gnu.org/software/coreutils/manual/html_node/chown-invocation.html#chown-invocation.
- fixes flashing slowdown in TWRP 3.0.0: TWRP3 (currently) appears to have a bugged busybox/ln command which takes an excessive amount of time, no reason not to use our own instead
- fail on recoveries where symlinking our busybox environment doesn't work (CM Recovery on some devices)
- use the included busybox for awk since toybox is lacking (CM Recovery)
- fix fallback broken unzip detection now that we create a subdirectory prior
- probably best to execute the bump/loki/whatever script using our included ash shell instead of the recovery sh.
- add busybox to the zip to provide a uniform expected environment
- set up busybox symlinks in /tmp and execute anykernel.sh from a shell created with these at the front of the path, overriding the recovery applets
- fix copy of any subdirectories under modules/
- adjust display of modules installation text to match the others
- add support for custom user scripts (at /data/custom_boot_image_patch.sh - same as latest SuperSU zips) to automatically execute before reflashing the rebuilt boot image for bump, loki etc. modifications
- initial unzip verification to catch it early on broken recoveries/zips
- display the matching name during the devicecheck while installing
- mount /system ro for the off-chance getprop relies on build.prop (older recoveries), and only remount rw if init.d or modules need to be setup/installed to fix/workaround TWRP's "mount only as ro" feature being finicky and allow /system to remain untouched on devices supporting systemless root.
- add signing documentation to README
- fix insert_line to perform more as expected by adding a newline automatically
- add updated arm mkbootimg+unpackbootimg binaries from my AIK Build Farm, see my repos for changes
- rewrite backend to use proper exit codes from anykernel.sh now that it is shell
- fix ui_print and specific error messages by passing the outfd to anykernel.sh from the backend
- add error messages for unpacking and repacking ramdisk, and make resulting image size check into its own error message
- fix for cpio overwriting AK2 ramdisk files on some recoveries by using a tmp directory
- add further ramdisk unpack verification by checking that there are actually files in the ramdisk directory prior to moving the tmp files back
- add insert_file function to make multi-line additions easier
- add patch_fstab function to greatly simplify mount option/flag changes (thanks @Khaon for the idea)
It seems like the " in Line 6 are not enough, unzip still refused to unpack files inside a directory with a space in it's name.
The new " fix that and now allow us to flash zips inside folders with spaces in their names.
- rewrite update-script as a shell script update-binary replacement, making the zip architecture independent (tools are still dependent)
- add fall-back to included zImage to allow for ramdisk-only injections
- add support for up to 5 device.name properties for variants
- add support for custom dtb swapping (falls back to included if none found in zip root)
- fix ramdisk extraction when there is none in the zip (ie. nothing to be added)
- add exit codes to trigger the abort in the updater-script if something goes wrong
- add ui_print function to write error message directly back to recovery
- compare size of dumped image and new image and abort if the new image is oversized for the partition
osm0sis