public inbox for isar-users@googlegroups.com
 help / color / mirror / Atom feed
From: "Koch, Stefan" <stefan-koch@siemens.com>
To: "isar-users@googlegroups.com" <isar-users@googlegroups.com>,
	"Kiszka, Jan" <jan.kiszka@siemens.com>
Cc: "Storm, Christian" <christian.storm@siemens.com>,
	"cedric.hombourger@siemens.com" <cedric.hombourger@siemens.com>,
	"Adler, Michael" <michael.adler@siemens.com>,
	"amikan@ilbers.de" <amikan@ilbers.de>,
	"Sudler, Simon" <simon.sudler@siemens.com>,
	"ubely@ilbers.de" <ubely@ilbers.de>,
	"MOESSBAUER, Felix" <felix.moessbauer@siemens.com>,
	"Schmidt, Adriaan" <adriaan.schmidt@siemens.com>
Subject: Re: [PATCH v2] linux-custom: Speedup build of target specific linux-kbuild package
Date: Fri, 7 Jun 2024 15:37:43 +0000	[thread overview]
Message-ID: <64aee6fe6342deae4e2b5814ff9db9ed83022012.camel@siemens.com> (raw)
In-Reply-To: <6c75f8aa-41da-4fde-9883-09cd5b24a473@siemens.com>

On Fri, 2024-06-07 at 17:07 +0200, Jan Kiszka wrote:
> On 07.06.24 16:11, Stefan Koch wrote:
> > Avoids time expensive QEMU-emulated merging of kernel config by
> > using
> > the existing kernel config for the target specific linux-kbuild
> > 
> > Using DEPENDS instead of RDEPENDS ensures creation of
> > kernel (including config) before build of target specific linux-
> > kbuild
> > 
> > Signed-off-by: Stefan Koch <stefan-koch@siemens.com>
> > ---
> >  .../linux/files/debian/isar/configure.tmpl     | 18
> > ++++++++++++++----
> >  meta/recipes-kernel/linux/linux-custom.inc     | 10 +++++++---
> >  2 files changed, 21 insertions(+), 7 deletions(-)
> > 
> > diff --git a/meta/recipes-
> > kernel/linux/files/debian/isar/configure.tmpl b/meta/recipes-
> > kernel/linux/files/debian/isar/configure.tmpl
> > index 389c9a85..72eae448 100644
> > --- a/meta/recipes-kernel/linux/files/debian/isar/configure.tmpl
> > +++ b/meta/recipes-kernel/linux/files/debian/isar/configure.tmpl
> > @@ -10,10 +10,20 @@ do_configure() {
> >      # Trace what we do here
> >      set -x
> >  
> > -    # Process kernel config target and fragments
> > -    ${MAKE} O=${KERNEL_BUILD_DIR} ${KERNEL_CONFIG_TARGET}
> > -    ./scripts/kconfig/merge_config.sh -O ${KERNEL_BUILD_DIR}/ \
> > -        ${KERNEL_BUILD_DIR}/.config ${KERNEL_CONFIG_FRAGMENTS}
> > +    kernelconfig="$(find /boot -maxdepth 1 -name "config-${PV}*" -
> > print -quit)"
> 
> Can we additionally limit this to the case were we expect the config
> there (kbuild && !kernel)? Just to reduce the risk that some "this
> should never be there" config is ever silently pulled by the base
> package build. Only if our dependency is installed, that rather
> generic
> config-<version> file is from there, or we failed on installing the
> dependency.
Build profile can be checked here as second condition, too.

find with "config-${PV}*" is used because some kernel specify "-rt18"
or "-isar" or something else that is not included in PV:

/boot/config-5.10.209-audis-rtp-rt18
/boot/config-6.6.11-isar

Some kernel do not specify an suffix to ${PV}. Maybe it depends on
LOCALVERISON from the kernel config? But as kernel config is not
available, we cannot find out this information...

find returns the first match. When using the old isar-buildchroot
instead of sbuild/schroot: Multiple configs could exist - if multiple
kernels are built. But if ${PV} is identical dpkg should have a
conflict with overwriting an existing config with the same version? (as
long this "-rt18"/"-isar" suffix is not renamed for that build.)

See below for dependency problem.
> 
> > +    if [ -e "${kernelconfig}" ]; then
> > +        # Prefer existing kernel config
> > +        # So, very expensive QEMU-emulated merge_config.sh
> > +        # can be skipped for target specific linux-kbuild package
> > +        mkdir -p ${KERNEL_BUILD_DIR}
> > +        cp "${kernelconfig}" ${KERNEL_BUILD_DIR}/.config
> > +        ${MAKE} O=${KERNEL_BUILD_DIR} olddefconfig
> > +    else
> > +        # Process kernel config target and fragments
> > +        ${MAKE} O=${KERNEL_BUILD_DIR} ${KERNEL_CONFIG_TARGET}
> > +        ./scripts/kconfig/merge_config.sh -O ${KERNEL_BUILD_DIR}/
> > \
> > +            ${KERNEL_BUILD_DIR}/.config ${KERNEL_CONFIG_FRAGMENTS}
> > +    fi
> >  
> >      # Stop tracing
> >      set +x
> > diff --git a/meta/recipes-kernel/linux/linux-custom.inc
> > b/meta/recipes-kernel/linux/linux-custom.inc
> > index 647f09dd..dc354e13 100644
> > --- a/meta/recipes-kernel/linux/linux-custom.inc
> > +++ b/meta/recipes-kernel/linux/linux-custom.inc
> > @@ -25,6 +25,7 @@ KBUILD_DEPENDS ?= "build-essential:native, \
> >                     flex, \
> >                     git, \
> >                     kmod, \
> > +                   linux-image-
> > ${KERNEL_NAME_PROVIDED}:${DISTRO_ARCH} <kbuild !kernel>, \
That ensures installation of the dependency.
> >                     rsync,"
> >  
> >  KERNEL_DEBIAN_DEPENDS ?= "initramfs-tools | linux-initramfs-tool,
> > \
> > @@ -112,7 +113,7 @@ BUILD_PROFILES = "kernel kbuild"
> >  BBCLASSEXTEND:append:cross-profile = " kbuildtarget"
> >  
> >  # When cross-profile is active:
> > -# build only kernel with the default variant of the recipe
> > +# build only kernel (including config) with the default variant of
> > the recipe
> >  BUILD_PROFILES:cross-profile = "kernel"
> >  
> >  # -native: kbuild package for host
> > @@ -122,14 +123,17 @@ RECIPE_PROVIDES:class-native = " \
> >      linux-kbuild-${KERNEL_NAME_PROVIDED}"
> >  # Use pseudo target to pull in the base variant of the recipe.
> >  # Will be auto-extended with -native by multiarch.bbclass.
> > -RDEPENDS:class-native += "${BPN}-pseudo"
> > +# Using DEPENDS, see below
> > +DEPENDS:class-native += "${BPN}-pseudo"
> 
> Need to help me with remember my own code: Why do we also need this
> build dependency? I thought you are only optimizing the kbuildtarget
> case.
Kernel config is also read from linux-image for linux-kbuild-native.
When choosing to call merge_config.sh here again, this change would not
be needed. Then a different build profile must be added, because
<kernel !kbuild> won't match anymore.
> 
> >  
> >  # -kbuildtarget: kbuild package for target, enforcing non-cross-
> > build
> >  BUILD_PROFILES:class-kbuildtarget = "kbuild"
> >  RECIPE_PROVIDES:class-kbuildtarget = " \
> >      linux-headers-${KERNEL_NAME_PROVIDED} \
> >      linux-kbuild-${KERNEL_NAME_PROVIDED}"
> > -RDEPENDS:class-kbuildtarget = "${BPN}"
> > +# Using DEPENDS instead of RDEPENDS to ensure creation of kernel
> > including
> > +# pregenerated kernel config before target specific linux-kbuild
> > package build
> > +DEPENDS:class-kbuildtarget = "${BPN}"
> >  ISAR_CROSS_COMPILE:class-kbuildtarget = "0"
> >  
> >  # Make bitbake know we will be producing linux-image and linux-
> > headers packages
> 
> Fine-tuning - this generally looks like a nice optimization now!
> 
> Jan
> 

-- 
Stefan Koch
Siemens AG
www.siemens.com

      reply	other threads:[~2024-06-07 15:37 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-06-07 14:11 Stefan Koch
2024-06-07 15:07 ` Jan Kiszka
2024-06-07 15:37   ` Koch, Stefan [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=64aee6fe6342deae4e2b5814ff9db9ed83022012.camel@siemens.com \
    --to=stefan-koch@siemens.com \
    --cc=adriaan.schmidt@siemens.com \
    --cc=amikan@ilbers.de \
    --cc=cedric.hombourger@siemens.com \
    --cc=christian.storm@siemens.com \
    --cc=felix.moessbauer@siemens.com \
    --cc=isar-users@googlegroups.com \
    --cc=jan.kiszka@siemens.com \
    --cc=michael.adler@siemens.com \
    --cc=simon.sudler@siemens.com \
    --cc=ubely@ilbers.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox