* Create common kernel config for Rockchip aarch64 boards
* Green: drop kernel option already included in main config
* Move rockchip RNG patchset to common folder
* Odroid-m1 has no board specific patches now
* Use separate path for v6.1.y and v6.6.y kernel config fragments
Since we're now maintaining Linux configs for two different versions,
it may happen that we want to add some options only to one of the
versions. While the Kconfig might figure the invalid options itself,
our config checking tooling would spam us with warnings. This commit
splits the configs to two directories. This pattern is used only for
the common fragments, more specific ones are usually sharing the same
Linux version anyway.
* Add back options removed in v6.6.y to v6.1.y kernel config fragments
* Linux: Update kernel 6.6.15
* Update buildroot packages to work with Linux 6.6
* Fix top-level and pc patches of linux
* Update tinker patch series
* Drop Odroid M1 patches
M1 is now supported in upstream.
* Update Hardkernel patches
Needed larger refactoring because of 379ae64609c7a3301b60483eb65bd8bc78f76328
* Update Green patches
* Update Odroid XU4 patches
Removing the TMU patch/hack for now, need to check if it's still needed.
If it is indeed, then it needs slighter rewrite.
* Move Rockchip RNG patches to M1 and Green dirs
* Update rtl88x2bu package to fix build
* Update gasket package to fix build
* Fix eq3_char_loop build
* Use fan53555 instead of custom rk860x driver
* Fix kernel base configs and fragments after 6.6 update
Mostly removed options that have been removed between releases. Only
a few options have been renamed, then there's bunch of options that
had dependencies added so they are available only on some architectures,
which are not those that we're using.
* Remove deprecated regulator-compatible from Green DTS
Enabling CONFIG_EXPERT, which was a dependency of some options we try
to set by our config fragments, had a side-effect of toggling some other
options, most importantly the framebuffer console support. Enable the
options found by diffing old and new kernel configs.
Fixes#3112
* Remove all non-existing kernel config symbols
* Remove unapplied x86 Intel sound options
These are missing various subsystem dependencies and were never in fact
enabled, assuming they're rather exotic and removing them completely.
* Add missing dependencies, adjust tristate values, remove nonsense
* Use KERNEL_LZ4 only for non-aarch64
Since aarch64 doesn't use self-extracting kernel:
https://patchwork.kernel.org/project/linux-arm-kernel/patch/20190119185540.20526-1-tobias.johannes.klausmann@mni.thm.de/
* Extract PCI options to device-support-pci fragment (renamed from device-support-pcie)
RPi 4+ should use this fragment too, since CM4 has PCIe support.
* Rename RPi's kernel-32b fragment to kernel-armv7
Compress firmware files from linux-firmware using ZSTD algorithm.
This should grant us some more space to add more firmwares and should
not have any major performance impact, because firmwares are not accessed
frequently.
Includes buildroot submodule bump:
* buildroot 07e08e01b2...8a75878da4 (1):
> linux-firmware: add option for firmware files compression
Preemptively enable larger set of WiFi drivers for all platforms and add more firmwares for them with the aim to harmonize WiFi device support among all boards and to have implicit support of devices that users might want to use. Targets `generic_aarch64`, `generic_x86_64` and `ova` also include options and firmwares for cards that are using PCI/PCIe bus - support for these is in a separate config fragment.
Especially the `generic_x86_64` is currently very tight with the rootfs space, so I had to do some triaging and select only sensible drivers and firmwares - especially archaic PCMCIA devices or devices not supporting only 802.11g or lower standards were among the first that I removed during the triaging - we can consider enabling those but this time on an someone's explicit need to have them enabled.
This closes#2815 and replaces large part of #2761, also potentially addresses (at least) these: #2806, #2783, #2841, #2776, #2725, #2600
-------------
* Remove WiFi options from generic and board kernel config fragments
* Enable MMC in OVA kernel
This is needed for SDIO drivers to work. Use the same options as we
currently use for generic_x86_64.
* Add CRYPTO_MICHAEL_MIC to the common kernel config
This is requirement for TKIP and is a dependency of ATH11K driver.
* Add kernel config fragments with wireless cards support
* Add firmwares for WiFi cards
* Enable more Bluetooth device drivers
Clean up all kernel configs and fragments from non-existing kernel
options, invalid choice values and choices that trigger warnings
during kernel package configuration.
Here's an example of warnings triggered for Yellow:
.config:8531:warning: override: MODULE_COMPRESS_NONE changes choice state
.config:8536:warning: override: ZSWAP_COMPRESSOR_DEFAULT_LZ4 changes choice state
.config:8537:warning: override: ZSWAP_ZPOOL_DEFAULT_ZSMALLOC changes choice state
.config:8543:warning: override: CPU_FREQ_DEFAULT_GOV_ONDEMAND changes choice state
.config:8717:warning: override: reassigning to symbol CGROUP_HUGETLB
.config:8767:warning: symbol value 'm' invalid for XFRM
.config:8852:warning: symbol value 'm' invalid for MEDIA_CONTROLLER_DVB
.config:8972:warning: symbol value 'm' invalid for SND_HDA_I915
There were also some options that are set in our or default configs
but end up patched by `KCONFIG_(DIS|EN)ABLE_OPT` in package makefiles,
these options are now explicitly set in our fragments too. For example
this was toggled for `generic_aarch64`:
CONFIG_DEFAULT_SECURITY_APPARMOR n -> y
CONFIG_DEFAULT_SECURITY_DAC y -> n
CONFIG_GCC_PLUGINS y -> n
The only goal of this commit is to make sure no warnings appear
anymore and the resulting kernel configs remain unchanged. This will
allow us to create tools for sanity checks of our kernel config
overrides. There is one single change in `ova` config resulting from
previously invalid `m` option for a bool value:
-# CONFIG_9P_FS_POSIX_ACL is not set
+CONFIG_9P_FS_POSIX_ACL=y
* Use kernel local version for HAOS compiled Linux kernel
Use the local version config option to add "haos" to the system's Linux
kernel version to indicate the kernel is built specifically for Home
Assistant OS. This makes sure to overwrite any other local version (e.g.
provided by Raspberry Pi kernel's defconfig) and makes it easier to
verify something is running on HAOS since the string will be visible in
any Container using `uname -a`.
* Add dash in front
Currently `CONFIG_OVERLAY_FS_METACOPY` and
`CONFIG_OVERLAY_FS_REDIRECT_DIR` kernel options are enabled but not
preferred by Docker. The metadata copy feature is disabled by default,
and also not actively used by the overlayfs2 driver (see
https://github.com/moby/moby/commit/2c3d1f7b4be754abf4c3a169aa19db0792847fc6).
So the metadata copy config is not really problematic per se. However,
it enables the redirect_dir feature. And a kernel which has the
redirect_dir feature compiled in also enables it by default. This
actually makes the overlayfs2 driver to fallback to naive diff, which
is, from what I understand, slower than the overlayfs native diff (see
also
https://github.com/moby/moby/commit/49c3a7c4bac2877265ef8c4eaf210159560f08b4).
The Docker daemon is also reporting this on startup:
Not using native diff for overlay2, this may cause degraded performance
for building images: kernel has CONFIG_OVERLAY_FS_REDIRECT_DIR enabled
Currently `CONFIG_OVERLAY_FS_METACOPY` is enabled, and it also enables
`CONFIG_OVERLAY_FS_REDIRECT_DIR`. There was already a previous attempt
to disable the latter (see #2067).
Disable both configs explicitly until Docker is able to use them.
Enable fully preemptible kernel (low-latency desktop) configuration for
Home Assistant. Home Assistant can be considered as a soft real-time
system, where a lower latency is preferred over throughput.
A few tests using the rt_test development add-on didn't show measurable
improvements, but this could be due to rather synthetic test.
Currently some platform use voluntary preemptible kernel, and some fully
preemptible. So besides improving latency, this also aims to synchronize
the settings across all platforms.
Also make sure that debugging is not enable as it can have high runtime
overhead according to Kconfig.
* Enable Multi-Gen LRU
Multi-Gen LRU should improve performance under memory pressure. This is
especially useful for embedded platforms where memory is scarce.
* Add service to configure Multi-Gen LRU
Use min_ttl_ms of 1 which is the least aggressive in terms of lag. Since
we are a server application, we can tune trashing prevention with a
higher acceptable lag.
* Add multiple routes support in NetworkManager
Support multiple routes to the same network learned via Router
Information Option. With this change, the kernel will have multiple
routing table entries to a given Thread network. The routes gateway
won't be updated with every new RIO any longer since every gateway
has its own entry.
* Enable IPv6 router reachability probing
Currently router reachability probing is disabled since HAOS enables
IPv6 forwarding and the necessary kernel options are not enabled. With
this change router reachability probing is enabled even though we are
a router on our own.
Note that Linux commit ea659e077528 ("[IPV6] ROUTE: Do not enable router
reachability probing in router mode.") by default disabled this
behavior. But since we are acting as a router as well as a host device,
we rather want this reachability probing.
See also: https://lore.kernel.org/netdev/b9182b02829b158d55acc53a0bcec1ed667b2668.1680000784.git.stefan@agner.ch/T/#u
* Use zswap instead of swap in zram
This requires a swap file which will get generated automatically on
startup.
* Fix file size and free disk space comparison
* Set zswap factor to 33%
* Set vm.swappiness to 1
Decrease swapping to a minimum. This is also recommended for database
work loads by the MariaDB documentation. In practice it causes the least
amount of writes to disk when under memory pressure, while still making
swap available when needed.
It seems that Raspberry Pi enabled Multi-Gen LRU by default. By my
testing, it performs worse in some situation. Add it by default for all
platforms, but disable it by default for now.
Tim L
Jan Čermák
Stefan Agner
Jan Čermák
Jens Maus
Irene Cheng Chi
Adam Duff