I like to play around with a pre-emptive linux kernel and Yocto.
As hardware the SAMA5D3x Evalboard + SAMA5D35-CM module is used.
Atmel is providing a meta-atmel layer, which includes the "at91-linux_*.bb" recipe. This recipe builds the kernel for the SAMA5D3x machines.
For using the realtime kernel I should insert the realtime patches and activate them at the kernel config.
I did not found a tutorial how to do this with an existing kernel. My question is:
How/where to modify a given kernel recipe to make it a realtime kernel(preempt-rt)?
My test project is located at the
project dir "/home/user/yocto". It has following content:
"/yocto git clone" ==> Yocto system
"/meta-openembedded" ==> meta embedded recipes
"/meta-atmel" ==> atmels yocto layer
"/meta-atmel/recipes-kernel/linux/linux-at91_4.4.bb" ==> the kernel recipe
"/meta-atmel/conf/machine/sama5d3xek.conf" ==> the machine that run the kernel
"/myTest" ==> my test project
"/myTest/recipes-kernel/linux-at91_%.bbappend" ==> replace the kernel config + add own device tree
"/myTest/recipes-kernel/linux/linux-at91/sama5d3xek/defconfig" ==> my own kernel config
/myTest/recipes-kernel/linux/linux-at91/sama5d3xek/myDev.dts ==> my own device tree
Any ideas/tutorials how to manage to activate the RT-Kernel in Yocto?
In general:
in .../source/poky/recipes-kernel/linux you should find a linux-yocto-rt_X.XX.bb recipe to compile a full preemptive RT kernel.
For meta-atmel you should do:
Download the correct RT patch for your kernel version and apply it using a .bbappend file to your current linux kernel recipe. You could find the patch HERE
Add the patch to your bbappend file (stored in your own layer in the one of accepted direcotries). p.e.: SRC_URI += "file://0001-linux-rt.patch"
Activate preemptive kernel. Manually set CONFIG_PREEMPT=y at defconfig at your layer. Alternativly you can use bitbake virtual/kernel -c menuconfig
Pitfalls at meta-atmel:
the linux-at91_4.4.bb recipe does not care about patch and sublevel of the kernel (p.e. 4.4.66 -> ..). if there is a new version at at91-linux it will go after some time to the meta-atmel layer.
the RT branch of the linux kernel is not provided for every new sublevel
this means constant breaks of your own meta-layer
Related
I am building my Embedded Linux system using Yocto warrior on Ubuntu 18.04. I have my own core image recipe and an initramfs image recipe.
I've been reading the docs ( https://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#var-INITRAMFS_IMAGE ) and various posts on the internet in order to come up with the following in my local.conf:
# Use the INITRAMFS bundled in kernel
#KERNEL_IMAGETYPE = "Image-initramfs-jetson-nano.bin"
#KERNEL_IMAGE_BASE_NAME = "Image-initramfs-jetson-nano.bin"
#INITRAMFS_LINK_NAME = ""
INITRAMFS_NAME = "Initramfs"
INITRAMFS_IMAGE = "tegra-minimal-initramfs"
INITRAMFS_IMAGE_BUNDLE = "1"
These lines do in fact create an initramfs built in version of my Kernel and puts it in the deploy directory by the name Image-Initramfs.bin. It is slightly larger than the Image kernel file that successfully boots. So Yocto ends up building 2 kernels, one with initramfs, and one without.
ubuntu#ip:~/Desktop/jetson-yocto/build$ du -sh tmp/deploy/images/jetson-nano/Image-Initramfs.bin
36M tmp/deploy/images/jetson-nano/Image-Initramfs.bin
ubuntu#ip:~/Desktop/jetson-yocto/build$ du -sh tmp/deploy/images/jetson-nano/Image--4.9+git0+3c02a65d91-r0-jetson-nano-20190729195650.bin
33M tmp/deploy/images/jetson-nano/Image--4.9+git0+3c02a65d91-r0-jetson-nano-20190729195650.bin
The docs say this is accomplished with a secondary compilation path:
Controls whether or not the image recipe specified by INITRAMFS_IMAGE is run through an extra pass (do_bundle_initramfs) during kernel compilation in order to build a single binary that contains both the kernel image and the initial RAM filesystem (initramfs) image. This makes use of the CONFIG_INITRAMFS_SOURCE kernel feature.
Note
Using an extra compilation pass to bundle the initramfs avoids a circular dependency between the kernel recipe and the initramfs recipe should the initramfs include kernel modules. Should that be the case, the initramfs recipe depends on the kernel for the kernel modules, and the kernel depends on the initramfs recipe since the initramfs is bundled inside the kernel image.
The problem is that this initramfs Kernel is not installed by Yocto into the final SD Card image. Only the non-initramfs Kernel is installed. I have not been able to find a Yocto directive/setting on how to make it install the initramfs version instead of the non-initramfs one.
How can I do this?
If you are using wic tool to generate the SD card image then you can add something like below in local.conf
`IMAGE_BOOT_FILES_append = " Image-Initramfs.bin;${KERNEL_IMAGETYPE}"`
however if you are using custom script then you have to provide more information
and customize the SD card generation script.
I want to enable tc command that comes in iproute2 on my linux kernel. My kernel is built using yocto and bitbake.
So, I copied the iproute recipes and whole directory from the following link to try --
https://git.yoctoproject.org/cgit.cgi/poky/plain/meta/recipes-connectivity/iproute2
And included in my yocto build. That picked up recipe and built it all well. But I tc command is still not available on the built kernel.
Question:
What am I missing and how to enable tc in the kernel of a linux image built using Yocto recipes?
You shouldn't need to copy the whole recipe, poky should be in your sources directory. So just reference the recipe in your image. You need both iproute2 and iproute2-tc.
IMAGE_INSTALL += "iproute2 \
iproute2-tc"
Additionally you may need to enable some kernel modules that tc make use of, depending on your needs:
CONFIG_NET_SCHED
CONFIG_NET_SCH_CBQ
CONFIG_NET_SCH_HTB
CONFIG_NET_SCH_HFSC
CONFIG_NET_SCH_ATM
CONFIG_NET_SCH_PRIO
CONFIG_NET_SCH_MULTIQ
CONFIG_NET_SCH_RED
CONFIG_NET_SCH_SFQ
CONFIG_NET_SCH_TEQL
CONFIG_NET_SCH_TBF
CONFIG_NET_SCH_GRED
CONFIG_NET_SCH_DSMARK
CONFIG_NET_SCH_NETEM
CONFIG_NET_SCH_INGRESS
I'm compiling linux for an ARM board. I need to make some customized changes into an existing driver code present in the kernel repository and reload the driver.
I was expecting to find a ".ko" file in the driver directory after doing the make, but no such file exists. Apparently uImage/device tree image compilation doesn't work that way.
Do I need to write my own Makefile for standalone device driver compilation?
It may be a silly question, but sorry I'm pretty new to kernel/device drivers.
EDIT:
I followed the process outlined here: http://odroid.com/dokuwiki/doku.php?id=en:c1_building_kernel
After git checkout and installing the cross-compiler(arm-linux-gnueabihf-gcc 4.9.2), I issue the basic make comands
$ make odroidc_defconfig
$ make -j4
$ make -j4 modules
$ make uImage
All the steps are successful. The last few lines of log look like
KSYM .tmp_kallsyms1.o
KSYM .tmp_kallsyms2.o
LD vmlinux
SORTEX vmlinux
SYSMAP System.map
OBJCOPY arch/arm/boot/ccImage
Kernel: arch/arm/boot/ccImage is ready
Image arch/arm/boot/ccImage.lzo is ready
UIMAGE arch/arm/boot/uImage
Image Name: Linux-3.10.72
Created: Sat Mar 28 22:44:45 2015
Image Type: ARM Linux Kernel Image (lzo compressed)
Data Size: 5459649 Bytes = 5331.69 kB = 5.21 MB
Load Address: 00208000
Entry Point: 00208000
Image arch/arm/boot/uImage is ready
EDIT 2: Path to the driver code
https://github.com/hardkernel/linux/tree/odroidc-3.10.y/drivers/amlogic/efuse
Examining your Makefile
#
# Makefile for eFuse.
#
obj-$(CONFIG_EFUSE) += efuse_bch_8.o efuse_version.o efuse_hw.o efuse.o
We learn that the code can be built as either a loadable module, or permanently linked into the kernel itself.
Examining odroidc_defconfig from branch odroidc-3.10.y-android mentioned in your instructions we find
#
# EFUSE Support
#
CONFIG_EFUSE=y
With the "y" indicating that the code is to be linked into the driver. Had it instead said "m" it would be built as a module.
It's possible you could change that in the kernel config, but it might also cause problems if there's nothing setup to load the module before it is needed.
Likely simply installing the newly built kernel with the code already linked inside (ie, forgetting about the module idea) will work.
Not sure if you are still looking for the answers to this question.
But looking at the Kconfig file in your code, show that -
config EFUSE
bool "EFUSE Driver"
And since all your driver files are compiled with this config, the above config description allows the CONFIG_EFUSE to be 'n' or 'y'. So you can only build static modules (build-in) with this.
All you need to do is change the above description to:
config EFUSE
**tristate** "EFUSE Driver"
and also change the other configs in Kconfig to tristate.
This will allow your driver to be compiled as module once you select the driver as 'M' in your kernel config. Then you should be able to see the ".ko" file corresponding to the driver.
Also do make sure to use EXPORT_SYMBOL(foo) when building the driver as module so that any dependencies are taken care of when loading module symbols.
I'm trying to use a Raspberry Pi board with the images generated by
Buildroot. I'm using the raspberrypi_defconfig configuration, as suggested here, in order to generate the rootfs, the kernel image and all files required to boot the system.
The problem
In order to build a Real Time system I need to apply the Xenomai kernel patch available for kernels 3.5.7, 3.10.32 and 3.14. The list of availables Xenomai patches can be found here. Here are the changes I made in the graphical configuration tool:
Toolchain -> Kernel headers (Manually specified Linux version) -> 3.5.7
Kernel -> Kernel version (custom) -> 3.5.7 as Kernel version
Kernel -> Kernel configutation (Using a defconfig) -> (bcmrpi_quick)
Kernel -> Linux Kernel Extentions -> [*] Adeos/Xenomai Real-time patch
Kernel -> Linux Kernel Extentions -> Path for Adeos patch file (the Relative path to the correct patch)
After the make command I get the following error. Buildroot couldn't find the config file
/buildroot-2013.11/output/build/linux-3.5.7/arch/arm/configs/bcmrpi_quick_defconfig’: No such file or directory
Which defconfig name should I use in the Kernel -> Defconfig name field?
I will appreciate your help with this situation.
I am trying to compile the linux kernel (3.0.0-13) with the Xen dom0 config flags which are not exposed via menuconfig. (Yes, I know that ubuntu provides a 'virtual' flavoured kernel that supports Xen paravirtualization, but that kernel does not seem to boot on my hardware. So, I am trying to compile the 'generic' flavoured ubuntu kernel with the extra Xen config flags since I know that the 'generic' flavour runs on my hardware). Every time that I try to compile my config flags are ignored based on the .config file that is generated and packaged with my kernel binary.
I have tried the following the following:
Downloaded the kernel source using apt-get source linux-image
I have then followed all of the steps from this guide: How to compile a new Ubuntu 11.10 (Oneiric) kernel and performed the following extra steps:
put my own config flags in the config.flavour.xxx file then compiled the linux-image package
Paused the 'debian/rules editconfigs' command immediately after the it runs 'menuconfig' and replaced the build/.config file with my custom .config file then compiled the linux-image package
I have also used the following howto How To Compile A Kernel - The Ubuntu Way and run the following commands on kernel source code that I already had:
edit the .config file to have my config flags
run 'make oldconfig'
run 'make-kpkg clean && fakeroot make-kpkg --initrd --append-to-version=-custom kernel_image kernel_headers'
After every time I have compiled the kernel I have installed the newly compiled linux-image package and have discovered that my config flags are not in the /boot/config-xxx file as I expect.
What am I doing wrong to cause my config flags to be ignored?
What can I do to make sure that my kernel config flags are used when compiling?
Is there some other option than recompiling the kernel to get a Xen dom0 kernel that work for my hardware?
For question 3: Is there another way to get a xen dom0 kernel for my hardware?
Yes.
Although all of the xen documentation says that all stock kernel support xen dom0, what they mean is that the source for all stock kernels now support xen dom0 but that support is turned off in their precompiled binaries.
On debian there is the following package which is a prebuilt linux kernel with the xen dom0 support turned on. Package: linux-image-xen-686
For anyone else who is really looking to compile their own xen dom0 kernel the following site has a good guide: Compiling a Xen Dom0 Kernel for Ubuntu Jaunty
What am I doing wrong to cause my config flags to be ignored?
The root of the issue lies in the first portion of your problem; the Xen dom0 config flags are not exposed via menuconfig
Simply setting them in the .config doesn't mean they'll be activated. You have to consider the dependencies for the config options.
From the linux 3.0 tag at github: https://github.com/torvalds/linux/blob/02f8c6aee8df3cdc935e9bdd4f2d020306035dbe/arch/x86/xen/Kconfig
config XEN_DOM0
def_bool y
depends on XEN && PCI_XEN && SWIOTLB_XEN
depends on X86_LOCAL_APIC && X86_IO_APIC && ACPI && PCI
Are all these depends flags met?
What can I do to make sure that my kernel config flags are used when compiling?
In the beginning stages of the kernel compile process, the .config file is re-written if there are any discrepancies. A good test to make sure your edits will persist is checking if they still exist in your .config file after doing a make menuconfig and saving changes. If after that your flags are still there, you can be sure that your flags are being used.
Is there some other option than recompiling the kernel to get a Xen dom0 kernel that work for my hardware?
Not unless another distribution ships with XEN_DOM0 enabled.