[Buildroot] [PATCH 2 of 2 v2] Add xenomai real-time Framework to buildroot
Arnout Vandecappelle
arnout at mind.be
Thu Sep 15 22:58:49 UTC 2011
On Wednesday 14 September 2011 14:56:15, Thomas De Schampheleire wrote:
[snip]
> diff --git a/linux/linux-ext-xenomai.mk b/linux/linux-ext-xenomai.mk
> new file mode 100644
> --- /dev/null
> +++ b/linux/linux-ext-xenomai.mk
> @@ -0,0 +1,29 @@
> +##################################################
> +# Linux Adeos/Xenomai extensions
> +#
> +# Patch the linux kernel with xenomai extension
> +##################################################
> +
> +ifeq ($(BR2_LINUX_KERNEL_EXT_XENOMAI),y)
> +# Add dependency to xenomai (user-space) which provide ksrc part
> +LINUX_DEPENDENCIES += xenomai
> +
> +# Adeos patch version
> +XENOMAI_ADEOS_PATCH = $(call qstrip,$(BR2_LINUX_KERNEL_EXT_XENOMAI_ADEOS_PATCH))
> +ifeq ($(XENOMAI_ADEOS_PATCH),)
> +XENOMAI_ADEOS_OPT = --default
> +else
> +XENOMAI_ADEOS_OPT = --adeos=$(XENOMAI_ADEOS_PATCH)
> +endif
> +
> +# Prepare kernel patch
> +define XENOMAI_PREPARE_KERNEL
> + $(XENOMAI_DIR)/scripts/prepare-kernel.sh \
> + --linux=$(LINUX_DIR) \
> + --arch=$(BR2_ARCH) \
> + $(XENOMAI_ADEOS_OPT)
> +endef
> +
> +LINUX_POST_EXTRACT_HOOKS += XENOMAI_PREPARE_KERNEL
> +
> +endif #BR2_LINUX_EXT_XENOMAI
Don't we usually limit the ifeq() block to the LINUX_DEPENDENCIES and
LINUX_POST_EXTRACT_HOOKS lines?
Shouldn't it be _POST_PATCH_HOOKS rather than _POST_EXTRACT_HOOKS? Not
that it matters much in the case of the kernel, but it's more consistent.
> diff --git a/package/Config.in b/package/Config.in
> --- a/package/Config.in
> +++ b/package/Config.in
> @@ -487,6 +487,10 @@
> endif
> endmenu
>
> +menu "Real-Time"
> +source "package/xenomai/Config.in"
> +endmenu
> +
> menu "Shell and utilities"
> source "package/at/Config.in"
> if BR2_PACKAGE_BUSYBOX_SHOW_OTHERS
> diff --git a/package/xenomai/Config.in b/package/xenomai/Config.in
> new file mode 100644
> --- /dev/null
> +++ b/package/xenomai/Config.in
> @@ -0,0 +1,78 @@
> +config BR2_PACKAGE_XENOMAI
> + bool "Xenomai Userspace"
> + help
> + Real-Time Framework for Linux
> + http://www.xenomai.org
I think most of the README should be put here rather than as a README.
People only look at a package's README if they're modifying the buildroot
integration of that package.
At the very least, the help text should refer to the README.
> +
> +config BR2_PACKAGE_XENOMAI_VERSION
> + depends on BR2_PACKAGE_XENOMAI
> + string "Custom Xenomai version"
> + help
> + Manualy select Xenomai Version.
> + Take care of compatibility with the Linux kernel.
I think this could be improved as follows:
Manually select Xenomai version. If left empty, the default version
will be used.
Make sure that the selected version has a patch for your selected Linux
kernel. If it does not, download and select a patch manually with
BR2_LINUX_KERNEL_EXT_XENOMAI_ADEOS_PATCH, in the
Linux Kernel -> Linux Kernel Extensions menu.
[I don't mention which version is the default so we don't have to update
the version number in two places. Actually, contrary to what PeterK has
suggested, I would put a default in the config and nothing in the makefile.]
[BTW here's a nice example why I prefer the patch and the Xenomai
userspace to sit close to each other in the config. Referring to a
completely different menu in this way is not nice IMHO.]
> +
> +# Manually select CPU (use it if not detected)
Comment is redundant
> +config BR2_PACKAGE_XENOMAI_CPU_TYPE_MANUAL
> + bool "Manually Select Xenomai CPU"
> + depends on (BR2_PACKAGE_XENOMAI && BR2_arm)
> + help
> + This option allows you to select Xenomai CPU in a list. Use it only
> + if CPU is not auto-detected by Buildroot ! (i.e. your board is not
> + described)
If this config is not set, it defaults to generic. Also this config isn't used
in the makefile. Doesn't it make more sense to remove this config and just
put generic as the default in the following choice?
> +
> +choice
> + prompt "Xenomai ARM CPU type"
> + depends on BR2_PACKAGE_XENOMAI_CPU_TYPE_MANUAL
> + help
> + Manually select CPU type for Xenomai.
Improved help text:
On this architecture, Xenomai needs to be configured for a specific
sub-architecture. Select the appropriate sub-architecture from the list.
[I don't know what this auto-detection mentioned in the _MANUAL
config means...]
> +
> + config BR2_PACKAGE_XENOMAI_CPU_AT91RM9200
> + bool "Atmel AT91RM9200"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_AT91SAM926X
> + bool "Atmel AT91SAM926X"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_INTEGRATOR
> + bool "ARM Ltd. Integrator"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_IXP4XX
> + bool "Intel IXP4XX (XScale)"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_IMX
> + bool "Freescale i.MX (MX1/MXL)"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_IMX21
> + bool "Freescale i.MX21"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_MX2
> + bool "Freescale MXC/MX2"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_MX3
> + bool "Freescale MXC/MX3"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_SA1100
> + bool "StrongARM SA1100"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_S3C2410
> + bool "S3C2410"
> +
> + config BR2_PACKAGE_XENOMAI_CPU_GENERIC_ARM
> + bool "Generic ARM"
> +
> +endchoice
> +
> +# Now set CPU type. We force manually selected type even if detected in case
> +# user takes the risk.
> +config BR2_PACKAGE_XENOMAI_CPU_TYPE
> + string
> + depends on BR2_PACKAGE_XENOMAI && BR2_arm
> + default "at91rm9200" if BR2_PACKAGE_XENOMAI_CPU_AT91RM9200
> + default "at91sam926x" if BR2_PACKAGE_XENOMAI_CPU_AT91SAM926X
> + default "integrator" if BR2_PACKAGE_XENOMAI_CPU_INTEGRATOR
> + default "ixp4xx" if BR2_PACKAGE_XENOMAI_CPU_IXP4XX
> + default "generic" if BR2_PACKAGE_XENOMAI_CPU_GENERIC_ARM
> + default "imx" if BR2_PACKAGE_XENOMAI_CPU_IMX
> + default "imx21" if BR2_PACKAGE_XENOMAI_CPU_IMX21
> + default "mx2" if BR2_PACKAGE_XENOMAI_CPU_MX2
> + default "mx3" if BR2_PACKAGE_XENOMAI_CPU_MX3
> + default "s3c2410" if BR2_PACKAGE_XENOMAI_CPU_S3C2410
> diff --git a/package/xenomai/README b/package/xenomai/README
> new file mode 100644
> --- /dev/null
> +++ b/package/xenomai/README
> @@ -0,0 +1,48 @@
> +****************
> +* Xenomai *
> +****************
> +
> +Xenomai comes with 2 parts : Kernel Part and User-space part.
> +Kernel part patch Linux with ADEOS(i-pipe)
> +(http://home.gna.org/adeos/) and some Xenomai kernel space stuff.
> +
> +From Adeos main site :
> +"The purpose of Adeos is to provide a flexible environment for sharing
> +hardware resources among multiple operating systems, or among multiple
> +instances of a single OS."
> +
> +From Xenomai main site :
> +"Xenomai is a real-time development framework cooperating with the
> +Linux kernel, in order to provide a pervasive, interface-agnostic,
> +hard real-time support to user-space applications, seamlessly
> +integrated into the GNU/Linux environment."
> +
> +****************
> +* Installation *
> +****************
> +
> +* Buildroot Kernel: *
> + - Check that your version of Linux Kernel is supported by the
> + version of xenomai you choose.
> + - Download Adeos patch at
> + http://download.gna.org/adeos/patches/v2.6/$(ARCH)/, or if your
> + are sure, let the Xenomai 'prepare-kernel.sh' script do the work.
> + - Check the Xenomai Kernel option in Buildroot :
> + -> Kernel
> + -> Linux Kernel (BR2_LINUX_KERNEL [=y])
> + -> Linux Kernel Extensions
> + -> [*] Adeos/Xenomai Real-time patch
> + ..And add the path to Adeos patch. (or leave it empty if you let
> + xenomai script do the work)
> +
> +* Non-Buildroot Kernel AND Buildroot Kernel *
> + - Make sure your Kernel is patched with Xenomai
> + - Check the option in Buildroot menu :
> + -> Package Selection for the target
> + -> Real-time
> + -> [*] Xenomai Userspace
> + - You can download a custom version of Xenomai that fit with your
> + Kernel.
> + - If you use static /dev and the default device_table, Uncomment
> + Xenomai part in 'target/generic/device_table_dev.txt'. (rtheap,
> + rtscope and rtp[0-31])
> diff --git a/package/xenomai/xenomai-do-not-install-devices.patch b/package/xenomai/xenomai-do-not-install-devices.patch
> new file mode 100644
> --- /dev/null
> +++ b/package/xenomai/xenomai-do-not-install-devices.patch
> @@ -0,0 +1,11 @@
> +--- a/Makefile.in 2011-06-24 16:11:23.591449817 +0200
> ++++ b/Makefile.in 2011-06-24 16:20:29.638129633 +0200
> +@@ -719,7 +719,7 @@
> + $$sudo mknod -m 666 $(DESTDIR)/dev/rtheap c 10 254 ; \
> + fi
> +
> +-install-exec-local: devices
> ++install-exec-local:
> +
> + install-user:
> + $(MAKE) SUDO=false install
I used a patch which solves the underlying issue (at some point, sudo is
called instead of $SUDO) and I use
XENOMAI_INSTALL_STAGING_OPT = DESTDIR=$(STAGING_DIR) install-user
XENOMAI_INSTALL_TARGET_OPT = DESTDIR=$(TARGET_DIR) install-user
> diff --git a/package/xenomai/xenomai.mk b/package/xenomai/xenomai.mk
> new file mode 100644
> --- /dev/null
> +++ b/package/xenomai/xenomai.mk
> @@ -0,0 +1,88 @@
> +#############################################################
> +# Xenomai
> +# URL : http://xenomai.org
> +# NOTE : Real-Time Framework for Linux
> +#
> +#############################################################
> +
> +XENOMAI_VERSION = $(call qstrip,$(BR2_PACKAGE_XENOMAI_VERSION))
> +ifeq ($(XENOMAI_VERSION),)
> +XENOMAI_VERSION = 2.5.6
> +endif
> +
> +XENOMAI_SITE = http://download.gna.org/xenomai/stable/
> +XENOMAI_SOURCE = xenomai-$(XENOMAI_VERSION).tar.bz2
> +
> +XENOMAI_INSTALL_STAGING = YES
> +
> +ifeq ($(BR2_arm),y)
> +XENOMAI_CPU_TYPE = $(call qstrip,$(BR2_PACKAGE_XENOMAI_CPU_TYPE))
> +# Set "generic" if not defined
> +ifeq ($(XENOMAI_CPU_TYPE),)
> +XENOMAI_CPU_TYPE = generic
> +endif
> +XENOMAI_EXTRA_CONFIGURE_OPTS += "--enable-$(BR2_ARCH)-mach=$(XENOMAI_CPU_TYPE)"
> +endif #BR2_arm
> +
> +define XENOMAI_CONFIGURE_CMDS
> + (cd $(@D); rm -rf config.cache; \
> + $(TARGET_CONFIGURE_OPTS) \
> + $(TARGET_CONFIGURE_ARGS) \
> + CCFLAGS_FOR_BUILD="$(HOST_CFLAGS)" \
> + ./configure \
> + $(XENOMAI_EXTRA_CONFIGURE_OPTS) \
> + --host=$(BR2_ARCH)-linux \
> + $(if $(shell grep "^CONFIG_SMP*=*y" \
> + $(LINUX_DIR)/.config ),--enable-smp,) \
> + )
> +endef
> +
> +define XENOMAI_INSTALL_STAGING_CMDS
> + $(MAKE) DESTDIR=$(STAGING_DIR) -C $(@D) install
> +endef
> +
> +define XENOMAI_INSTALL_TARGET_CMDS
> + $(MAKE) DESTDIR=$(TARGET_DIR) -C $(@D) install
> + rm -rf $(TARGET_DIR)/usr/xenomai/share/doc
> + rm -rf $(TARGET_DIR)/usr/xenomai/include
> + # Make sure /etc/ld.so.conf exists
> + touch $(TARGET_DIR)/etc/ld.so.conf
> + # Add /usr/xenomai/lib only if it is not yet present
> + (grep "/usr/xenomai/lib" $(TARGET_DIR)/etc/ld.so.conf \
> + &>/dev/null) || (echo "/usr/xenomai/lib" >> $(TARGET_DIR)/etc/ld.so.conf)
> +endef
> +
> +# If you use static /dev creation don't forget to update your
> +# device_table_dev.txt
No need for this if we can apply the pkg-device patches that were posted by
Maxime a week ago.
> +ifeq ($(BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_UDEV),y)
> +XENOMAI_DEPENDENCIES += udev
> +
> +define XENOMAI_UDEV_HOOKS
> + if test -d $(TARGET_DIR)/etc/udev/rules.d ; then \
> + for f in $(@D)/ksrc/nucleus/udev/*.rules ; do \
> + cp $$f $(TARGET_DIR)/etc/udev/rules.d/ ; \
> + done ; \
> + fi;
> +endef
> +
> +XENOMAI_POST_INSTALL_TARGET_HOOKS += XENOMAI_UDEV_HOOKS
> +endif # udev
Call me crazy, but I would make this:
UDEV_POST_INSTALL_TARGET_HOOKS += XENOMAI_UDEV_HOOKS
without the condition or the dependency on udev. I always considered this
the main power of the hooks setup.
> +
> +define XENOMAI_CLEAN_CMDS
> + -$(MAKE) DESTDIR=$(TARGET_DIR) -C $(@D) clean
> +endef
> +
> +define XENOMAI_UNINSTALL_TARGET_CMDS
> + if test -d $(TARGET_DIR)/etc/udev/rules.d ; then \
> + for f in $(@D)/ksrc/nucleus/udev/*.rules ; do \
> + rm -f $(TARGET_DIR)/etc/udev/rules.d/$$f ; \
> + done ; \
> + fi;
> + $(MAKE) DESTDIR=$(TARGET_DIR) -C $(@D) uninstall
> +endef
> +
> +define XENOMAI_UNINSTALL_STAGING_CMDS
> + $(MAKE) DESTDIR=$(STAGING_DIR) -C $(@D) uninstall
> +endef
> +
> +$(eval $(call GENTARGETS,package,xenomai))
AUTOTARGETS works fine for me. You just need to override the configure.
Perhaps it's even possible to override the default prefix in the
XENOMAI_CONF_OPT variable, but I never tried that.
> diff --git a/target/generic/device_table_dev.txt b/target/generic/device_table_dev.txt
> --- a/target/generic/device_table_dev.txt
> +++ b/target/generic/device_table_dev.txt
> @@ -141,3 +141,8 @@
>
> # I2C device nodes
> /dev/i2c- c 666 0 0 89 0 0 1 4
> +
> +# Xenomai
> +#/dev/rtheap c 666 0 0 10 254 0 0 -
> +#/dev/rtscope c 666 0 0 10 253 0 0 -
> +#/dev/rtp c 666 0 0 150 0 0 1 32
As I mentioned, unnecessary with Maxime's patch.
Regardless of all my comments, this patch set is in a much better shape
than the hacks I am using at the moment!
Regards,
Arnout
--
Arnout Vandecappelle arnout at mind be
Senior Embedded Software Architect +32-16-286540
Essensium/Mind http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint: 31BB CF53 8660 6F88 345D 54CC A836 5879 20D7 CF43
More information about the buildroot
mailing list