* The defconfigs from the RPi Kernel set `powersave` as the default
CPU governor, which is a bad idea as that reduces performance by
setting the CPU frequency to the minimum one.
* In case of a Yocto build user-space is not configured by default
to change the CPU governor, so `powersave` will remain the CPU
governor and it will slow down everything.
* I submitted a fix for those defconfigs to upstream here:
https://github.com/raspberrypi/linux/pull/5666
However, we shouldn't wait until that is accepted and backported
to all the stable branches (if ever is).
* This patch unsets the selection of `powersave` as the default CPU
governor.
* Then the CPU governor that will be selected as default after this
patch will be the Linux's default one that is either `ondemand` or
`schedutil` depending on the Kernel version.
Signed-off-by: Carlos Alberto Lopez Perez <clopez@igalia.com>
Especially with systemd its seen that psplash-start service starts
before /dev/fb0 is created by kernel which results in
[FAILED] Failed to start Start psplash boot splash screen.
this is quite frequent race now with kernel 6.1
Add device dependency on sys-devices-platform-gpu-graphics-fb0.device
via a unit file drop-in
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Leon Anavi <leon.anavi@konsulko.com>
Update kernel to version 5.15.92 to fix Bluetooth and Wi-Fi issues
on 64-bit images for Raspberry Pi Zero 2.
Signed-off-by: Leon Anavi <leon.anavi@konsulko.com>
wic images depend on the kernel device trees, and therefore should
depend on virtual/kernel:do_deploy to make sure these are present in the
deploy directory.
Most of the time, this dependency is satisfied indirectly since a rootfs
image will depend on the kernel, but add it explicitly for the cases
where it is not.
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
(cherry picked from commit 482d864b8f1af84915ed6a9641e80af4e49a1f63)
When I was trying to drive a lcd(ili9881) with pwm backlight control.
I encountered with the problem in #1180.
After I check the kernel config of both raspbian OS and yocto.I find
the config of drm is different between them.
Detail of difference:
In raspbian OS, kernel conifg of drm is m while it's y in yocto.
So I change the config of drm to m in yocto too.And the pwm works fine now.
Signed-off-by: bigbearishappy <953308023@qq.com>
Prepending new definitions should be preferred to assignment, as it is
simpler and more conducive to defining new machine configurations that
reuse these configurations.
Signed-off-by: Zachary T Welch <zach@aquabyte.ai>
Otherwise the shell snippet fails with `bad syntax` when `CMDLINE`
contains special characters like `${...}` (useful to insert a U-Boot
variable in `cmdline.txt`).
Signed-off-by: Lluis Campos <lluis.campos@northern.tech>
(cherry picked from commit c57b464b88769c1336695e48e1eb8f35e258b9ae)
This patch had been merged into u-boot upstream. Its commit is
29f7d05a347a("dm: core: Move ofdata_to_platdata() call earlier").
So, remove the randundant patch to avoid building error.
Signed-off-by: Meng Li <Meng.Li@windriver.com>
rpi-test-image is based on core-image-base but includes more packages
that this layer provides. In this way we can have CI test more recipe
updates.
Signed-off-by: Andrei Gherzan <andrei.gherzan@huawei.com>
Drop the 0001-include-sys-types.h-explicitly-for-getting-caddr_t-d.patch
patch as it was fixed in this version[1].
This switches the source from pypi to git (github) as the new version
wasn't published yet.
Lastly, we use the LICENSE.txt file for LIC_FILES_CHKSUM so we don't
have to deal with context lines in the README file.
[1] https://github.com/metachris/RPIO/pull/41
Signed-off-by: Andrei Gherzan <andrei.gherzan@huawei.com>
The 43438 WLAN firmware doesn't include the clm_blob anymore which now
comes as a separate file: cyfmac43430-sdio.clm_blob.
The 43455 WLAN firmware now comes with a minimal build too. We make sure
that it is packaged and we dropped the associated README. In the future
we might want to introduce a build configuration for switching between
the two builds. For now we only deploy it on target and let users deal
with the symlink accordinly.
Signed-off-by: Andrei Gherzan <andrei.gherzan@huawei.com>
This updates the recipe to the head of the master branch. Upstream
shifted the license section in the README file so this patch changes
LIC_FILES_CHKSUM accordingly.
Signed-off-by: Andrei Gherzan <andrei.gherzan@huawei.com>
This version is now released as GPLv3. It also adds an example that is
not a simple binary anymore - it is provided as a tiny library and an
associated tool (spiram/spiram_test). Restructure the compilation and
installation of all examples to cater for this case.
Signed-off-by: Andrei Gherzan <andrei.gherzan@huawei.com>
m
This is so that uboot fit images have an entry point that matches the
LOADADDR that everything is expecting.
Signed-off-by: AJ Bagwell <anthony.bagwell@hivehome.com>
OE-Core's ffmpeg has moved to 5.x version whose headers are not directly
compatible with what omxplayer expects anymore. Therefore use captured
version of ffmpeg and point the header search paths accordingly.
Add missing dependency on alsa-lib for libasound headers
Fixes build with master/kirkstone
Signed-off-by: Khem Raj <raj.khem@gmail.com>
After switching the SRC_URI for firmware, this adjustment was needed to
let this recipe find /opt/vc path in correct place
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The `rtc` machine feature is "considered" by default for all the
machines we provide until now. This can confuse people who would want an
RTC device attached to a RaspberryPi board[1]. Clarify this via
documentation.
[1] https://github.com/agherzan/meta-raspberrypi/issues/1010
Signed-off-by: Andrei Gherzan <andrei.gherzan@huawei.com>
