linux-wrs: improve BSP-bootstrap process
cover linux-wrs: improve BSP-bootstrap process
I've been looking to improve the introduction of new boards to the supported kernel for some time, and this takes us most of the way there. Now with a machine.conf and a defconfig you can create an initial working environment for BSP development with no manual/workaround steps.
When creating a new BSP there isn't always an upstream branch to validate SRCREVs against. Therefore, creating a new BSP required extra manipulations of the git repository, even though the branch for the build would be dynamically created.
To fix this, provide a fallback to the 'standard' branch if WRMACHINE is not assigned a specific value for the given MACHINE. The fallback saved in KBRANCH, and represents the branch that we should build, even if no fallback is required. Some substeps of the recipe have been updated to use KBRANCH rather than enforcing MACHINE-KERNELTYPE.
There will be one additional change to not track the SRCREV of wrs_meta as a fallback, but I wasn't able to get the standard branch tracking working, so I'm sending this as-is for now.
I've tested this with my kernel development layer, and with the stock meta directories only. Either way, I was able to build and test a 'qemumips-variant' quickly.
Cheers,
Bruce
Pull URL: git://git.pokylinux.org/poky-contrib.git
Branch: zedd/kernel
Browse: http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=zedd/kernel
Thanks,
Bruce Ashfield <[email protected]>
---
Bruce Ashfield (2):
linux-wrs: implement BSP bootstrapping via fallback branches
linux-wrs: BSP bootstrap
.../conf/distro/include/poky-default-revisions.inc | 2 +- meta/recipes-kernel/linux/linux-wrs_git.bb | 40 ++++++++++++++++---- 2 files changed, 33 insertions(+), 9 deletions(-)
PATCH 1/2 linux-wrs: implement BSP bootstrapping via fallback branches
When creating a new BSP there isn't always an upstream branch to validate SRCREVs against. Therefore, creating a new BSP required extra manipulations of the git repository, even though the branch for the build would be dynamically created.
To fix this, provide a fallback to the 'standard' branch if WRMACHINE is not assigned a specific value for the given MACHINE. The fallback saved in KBRANCH, and represents the branch that we should build, even if no fallback is required. Some substeps of the recipe have been updated to use KBRANCH rather than enforcing MACHINE-KERNELTYPE.
--- meta/recipes-kernel/linux/linux-wrs_git.bb | 40 ++++++++++++++++++++++----- 1 files changed, 32 insertions(+), 8 deletions(-)
meta/recipes-kernel/linux/linux-wrs_git.bb +32 -8
@@ -10,10 +10,10 @@ PV = "2.6.34+git${SRCPV}" # To use a staged, on-disk bare clone of a Wind River Kernel, use a # variant of the below # SRC_URI = "git://///path/to/kernel/default_kernel.git;fullclone=1" -SRC_URI = "git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;fullclone=1;branch=${WRMACHINE}-${LINUX_KERNEL_TYPE};name=machine \ +SRC_URI = "git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;fullclone=1;branch=${KBRANCH};name=machine \ git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;noclone=1;branch=wrs_meta;name=meta" -WRMACHINE = "${MACHINE}" +WRMACHINE = "UNDEFINED" WRMACHINE_qemux86 = "common_pc" WRMACHINE_qemux86-64 = "common_pc_64" WRMACHINE_qemuppc = "qemu_ppc32"
@@ -24,11 +24,33 @@ WRMACHINE_routerstationpro = "routerstationpro" WRMACHINE_mpc8315e-rdb = "fsl-mpc8315e-rdb" WRMACHINE_beagleboard = "beagleboard" +# Determine which branch to fetch and build. Not all branches are in the +# upstream repo (but will be locally created after the fetchers run) so +# a fallback branch needs to be chosen. +# +# The default machine 'UNDEFINED'. If the machine is not set to a specific +# branch in this recipe or in a recipe extension, then we fallback to a +# branch that is always present 'standard'. This sets the KBRANCH variable +# and is used in the SRC_URI. The machine is then set back to ${MACHINE}, +# since futher processing will use that to create local branches +python __anonymous () { + import bb, re + + bb.data.setVar("KBRANCH", "${WRMACHINE}-${LINUX_KERNEL_TYPE}", d) + mach = bb.data.getVar("WRMACHINE", d, 1) + if mach == "UNDEFINED": + bb.data.setVar("KBRANCH", "standard", d) + bb.data.setVar("WRMACHINE", "${MACHINE}", d) + # track the global configuration on a bootstrapped BSP + bb.data.setVar("SRCREV_machine", "${SRCREV_meta}", d) + bb.data.setVar("BOOTSTRAP", "t", d) +} + COMPATIBLE_MACHINE = "(qemuarm|qemux86|qemuppc|qemumips|qemux86-64|atom-pc|routerstationpro|mpc8315e-rdb|beagleboard)" LINUX_VERSION = "v2.6.34" LINUX_VERSION_EXTENSION = "-wr-${LINUX_KERNEL_TYPE}" -PR = "r12" +PR = "r13" S = "${WORKDIR}/linux" B = "${WORKDIR}/linux-${WRMACHINE}-${LINUX_KERNEL_TYPE}-build"
@@ -69,7 +91,7 @@ do_patch() { } validate_branches() { - branch_head=`git show-ref -s --heads ${WRMACHINE}-${LINUX_KERNEL_TYPE}` + branch_head=`git show-ref -s --heads ${KBRANCH}` meta_head=`git show-ref -s --heads wrs_meta` target_branch_head="${SRCREV_machine}" target_meta_head="${SRCREV_meta}"
@@ -132,14 +154,16 @@ IFS=' cd ${S} # checkout and clobber and unimportant files - git checkout -f ${WRMACHINE}-${LINUX_KERNEL_TYPE} - - validate_branches + git checkout -f ${KBRANCH} + + if [ -z "${BOOTSTRAP}" ]; then + validate_branches + fi # this second checkout is intentional, we want to leave ourselves # on the branch to be built, but validate_branches could have changed # our initial checkout. So we do it a second time to be sure - git checkout -f ${WRMACHINE}-${LINUX_KERNEL_TYPE} + git checkout -f ${KBRANCH} } do_wrlinux_checkout[dirs] = "${S}"
Signature · 1 line
1.7.0.4
PATCH 2/2 linux-wrs: BSP bootstrap
In reply to Bruce AshfieldUpdating the kern-tools SRCREV to pickup changes required for BSP bootstrapping:
commit 9722d8decacd2b750f079b3fde7918810700f80e
Author: Bruce Ashfield <[email protected]> Date: Thu Nov 11 01:28:33 2010 -0500
createme: improve BSP bootstrapping
To streamline BSP bootstrapping, createme now actually creates the required branch rather than only dumping it in a .scc file. Later phases of the build will create the .scc file, so it isn't required here.
Also, a BSP with a '-' in the name would break the parent branch detection. Improving the parsing of this allows more flexibile branch naming.
commit c7d222b4cffb5108d9b14298fc4dc4cae31b784f
Author: Bruce Ashfield <[email protected]> Date: Thu Nov 11 01:27:33 2010 -0500
updateme: improve parent and target parsing
A BSP with a '-' in the name would break the parent branch detection. Improving the parsing of this allows more flexibile branch naming.
--- .../conf/distro/include/poky-default-revisions.inc | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
meta/conf/distro/include/poky-default-revisions.inc +1 -1
@@ -55,7 +55,7 @@ SRCREV_pn-gypsy ??= "147" SRCREV_pn-inputproto ??= "7203036522ba9d4b224d282d6afc2d0b947711ee" SRCREV_pn-inputproto-native ??= "7203036522ba9d4b224d282d6afc2d0b947711ee" SRCREV_pn-inputproto-nativesdk ??= "7203036522ba9d4b224d282d6afc2d0b947711ee" -SRCREV_pn-kern-tools-native ??= "140693a6b0d81b7ba7175b6cfce11c6c22f81e24" +SRCREV_pn-kern-tools-native ??= "9722d8decacd2b750f079b3fde7918810700f80e" SRCREV_pn-libdrm ??= "3f3c5be6f908272199ccf53f108b1124bfe0a00e" SRCREV_pn-libfakekey ??= "2031" SRCREV_pn-libgdbus ??= "aeab6e3c0185b271ca343b439470491b99cc587f"
Signature · 1 line
1.7.0.4
Re: [PATCH 1/2] linux-wrs: implement BSP bootstrapping via fallback branches
In reply to Bruce AshfieldHey Bruce,
Dropping yocto list as this is a development discussion (and the webpage says those take place on [email protected]). Someone correct me if I've got this wrong.
Quoted 8 lines · Bruce Ashfield
When creating a new BSP there isn't always an upstream branch to validate SRCREVs against. Therefore, creating a new BSP required extra manipulations of the git repository, even though the branch for the build would be dynamically created.
To fix this, provide a fallback to the 'standard' branch if WRMACHINE is not assigned a specific value for the given MACHINE. The fallback saved in KBRANCH, and represents the branch that
It seems that KBRANCH_FALLBACK might be a bit more self-descriptive. Unless I'm misunderstanding it's purpose...
OK reading the patch, KBRANCH is basically the new variable we store the kernel branch in, and it is set to standard if WRMACHINE is not set - KBRANCH itself is not the fallback. Code looks good, commit message was ambiguous (to me).
Quoted 14 lines
we should build, even if no fallback is required. Some substeps of the recipe have been updated to use KBRANCH rather than enforcing MACHINE-KERNELTYPE.
Signed-off-by Bruce Ashfield<[email protected]>--- meta/recipes-kernel/linux/linux-wrs_git.bb | 40 ++++++++++++++++++++++----- 1 files changed, 32 insertions(+), 8 deletions(-)meta/recipes-kernel/linux/linux-wrs_git.bb +0
-0
meta/recipes-kernel/linux/linux-wrs_git.bb PV = "2.6.34+git${SRCPV}" · line 16
Quoted 9 lines
meta/recipes-kernel/linux/linux-wrs_git.bb +2
-2@@ -10,10 +10,10 @@ PV = "2.6.34+git${SRCPV}" # To use a staged, on-disk bare clone of a Wind River Kernel, use a # variant of the below # SRC_URI = "git://///path/to/kernel/default_kernel.git;fullclone=1" -SRC_URI = "git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;fullclone=1;branch=${WRMACHINE}-${LINUX_KERNEL_TYPE};name=machine \ +SRC_URI = "git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;fullclone=1;branch=${KBRANCH};name=machine \ git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;noclone=1;branch=wrs_meta;name=meta" -WRMACHINE = "${MACHINE}" +WRMACHINE = "UNDEFINED"
Not a critique, just a question - why "UNDEFINED" instead of just not set?
Signature · 2 lines
Darren Hart Yocto Linux Kernel
Re: [PATCH 1/2] linux-wrs: implement BSP bootstrapping via fallback branches
Quoted 17 lines · Darren Hart
Hey Bruce,
Dropping yocto list as this is a development discussion (and the webpage says those take place on [email protected]). Someone correct me if I've got this wrong.
Quoted 9 lines · Bruce Ashfield
When creating a new BSP there isn't always an upstream branch to validate SRCREVs against. Therefore, creating a new BSP required extra manipulations of the git repository, even though the branch for the build would be dynamically created.
To fix this, provide a fallback to the 'standard' branch if WRMACHINE is not assigned a specific value for the given MACHINE. The fallback saved in KBRANCH, and represents the branch that
Quoted 7 lines
It seems that
KBRANCH_FALLBACKmight be a bit more self-descriptive. Unless I'm misunderstanding it's purpose...OK reading the patch, KBRANCH is basically the new variable we store the kernel branch in, and it is set to standard if WRMACHINE is not set - KBRANCH itself is not the fallback. Code looks good, commit message was ambiguous (to me).
Right. The fallback is actually only in the phython chunk. KBRANCH will be used shortly to separate what you checkout and build from what it validated by the fetchers. I needed this for the branch renaming and minimization changes that will follow this shortly.
Quoted 27 lines
Quoted 27 lines
we should build, even if no fallback is required. Some substeps of the recipe have been updated to use KBRANCH rather than enforcing MACHINE-KERNELTYPE.
Signed-off-by Bruce Ashfield<[email protected]>---
meta/recipes-kernel/linux/linux-wrs_git.bb | 40
++++++++++++++++++++++-----1 files changed, 32 insertions(+), 8 deletions(-)meta/recipes-kernel/linux/linux-wrs_git.bb +0
-1@@ -10,10 +10,10 @@ PV = "2.6.34+git${SRCPV}" # To use a staged, on-disk bare clone of a Wind River Kernel, use a # variant of the below # SRC_URI = "git://///path/to/kernel/default_kernel.git;fullclone=1" -SRC_URI ="
git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;fullclone=1;branch=${WRMACHINE}-${LINUX_KERNEL_TYPE};name=machine\
+SRC_URI = "git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;fullclone=1;branch=${KBRANCH};name=machine
\
Quoted 6 lines
meta/recipes-kernel/linux/linux-wrs_git.bb PV = "2.6.34+git${SRCPV}" · line 16
Quoted 4 lines
git://git.pokylinux.org/linux-2.6-windriver.git;protocol=git;noclone=1;branch=wrs_meta;name=meta"meta/recipes-kernel/linux/linux-wrs_git.bb +1
-1-WRMACHINE = "${MACHINE}" +WRMACHINE = "UNDEFINED"Not a critique, just a question - why "UNDEFINED" instead of just not set?
It was easier for me to test :) and let me make it more explicit in the python chunk if someone was grepping around .. but outside of that, nothing particularly deep.
Cheers,
Bruce
Quoted 9 lines
--
Darren Hart
Yocto Linux Kernel
_______________________________________________
poky mailing list
[email protected]
https://lists.yoctoproject.org/listinfo/poky
Signature · 2 lines
"Thou shalt not follow the NULL pointer, for chaos and madness await thee at its end"