[OE-core] [PATCH 2/2] kernel: user defined KERNEL_VERSION_PKG_NAME

Razvan Heghedus razvan.heghedus at ni.com
Mon Jul 3 14:36:40 UTC 2017



On 06/29/2017 04:06 PM, Bruce Ashfield wrote:
>
>
> On Thu, Jun 29, 2017 at 5:10 AM, Razvan Heghedus 
> <razvan.heghedus at ni.com <mailto:razvan.heghedus at ni.com>> wrote:
>
>
>
>     On 06/28/2017 04:29 AM, Bruce Ashfield wrote:
>>
>>
>>     On Tue, Jun 27, 2017 at 5:15 AM, Razvan Heghedus
>>     <razvan.heghedus at ni.com <mailto:razvan.heghedus at ni.com>> wrote:
>>
>>
>>
>>         On 06/26/2017 06:52 PM, Bruce Ashfield wrote:
>>>
>>>
>>>         On Wed, Jun 21, 2017 at 8:00 AM, Heghedus Razvan
>>>         <razvan.heghedus at ni.com <mailto:razvan.heghedus at ni.com>> wrote:
>>>
>>>             Add possibility to set KERNEL_VERSION_PKG_NAME to a user
>>>             defined value.
>>>
>>>             Signed-off-by: Heghedus Razvan <razvan.heghedus at ni.com
>>>             <mailto:razvan.heghedus at ni.com>>
>>>             ---
>>>              meta/classes/kernel.bbclass | 8 ++++++--
>>>              1 file changed, 6 insertions(+), 2 deletions(-)
>>>
>>>             diff --git a/meta/classes/kernel.bbclass
>>>             b/meta/classes/kernel.bbclass
>>>             index 605c101e62..02728d5a86 100644
>>>             --- a/meta/classes/kernel.bbclass
>>>             +++ b/meta/classes/kernel.bbclass
>>>             @@ -28,12 +28,16 @@ INITRAMFS_IMAGE_BUNDLE ?= ""
>>>              # LINUX_VERSION which is a constant.
>>>              KERNEL_VERSION_NAME = "${@d.getVar('KERNEL_VERSION') or
>>>             " <mailto:$%7B at d.getVar%28%27KERNEL_VERSION%27%29or>"}"
>>>              KERNEL_VERSION_NAME[vardepvalue] = "${LINUX_VERSION}"
>>>             -KERNEL_VERSION_PKG_NAME =
>>>             "${@legitimize_package_name(d.getVar('KERNEL_VERSION'))}"
>>>             -KERNEL_VERSION_PKG_NAME[vardepvalue] = "${LINUX_VERSION}"
>>>
>>>              python __anonymous () {
>>>                  import re
>>>
>>>             +    if d.getVar('USER_KERNEL_VERSION_PKG') is None :
>>>             + d.setVar('KERNEL_VERSION_PKG_NAME',
>>>             "${@legitimize_package_name(d.getVar('KERNEL_VERSION'))}")
>>>             + d.setVar('KERNEL_VERSION_PKG_NAME[vardepvalue]',
>>>             "${LINUX_VERSION}")
>>>             +    else:
>>>             + d.setVar('KERNEL_VERSION_PKG_NAME',
>>>             "${@legitimize_package_name(d.getVar('USER_KERNEL_VERSION_PKG'))}")
>>>
>>>
>>>         This is introducing yet another variable that tweaks the
>>>         already complex setting of
>>>         the kernel version. Not to mention this code is already
>>>         touchy with respect to
>>>         parse time and rebuilding of the kernel.
>>>
>>>         My concern is that if this is set, we are completely
>>>         disassociated with the source
>>>         code of the kernel.
>>>
>>>         Where did you think this would be set ? local.conf ? distro
>>>         config ? somewhere else ?
>>>
>>>         If we had a way to simply override KERNEL_VERSION, we
>>>         wouldn't need any extra
>>>         variables.
>>>
>>>         Bruce
>>>
>>>             +
>>>                  # Merge KERNEL_IMAGETYPE and KERNEL_ALT_IMAGETYPE
>>>             into KERNEL_IMAGETYPES
>>>                  type = d.getVar('KERNEL_IMAGETYPE') or ""
>>>                  alttype = d.getVar('KERNEL_ALT_IMAGETYPE') or ""
>>>             --
>>>             2.13.1
>>>
>>>             --
>>>             _______________________________________________
>>>             Openembedded-core mailing list
>>>             Openembedded-core at lists.openembedded.org
>>>             <mailto:Openembedded-core at lists.openembedded.org>
>>>             http://lists.openembedded.org/mailman/listinfo/openembedded-core
>>>             <http://lists.openembedded.org/mailman/listinfo/openembedded-core>
>>>
>>>
>>>
>>>
>>>         -- 
>>>         "Thou shalt not follow the NULL pointer, for chaos and
>>>         madness await thee at its end"
>>         I have setting the variable in the kernel recipe. I need a
>>         way to override the KERNEL_VERSION because I want the kernel
>>         packages name to contain only a part of the version or
>>         nothing at all.
>>         I need this for the the kernel upgrade stuff, because if the
>>         package name is something like: kernel-4.9.8-{static_string}
>>         then I couldn't upgrade to a version like:
>>         kernel-4.9.10-{static_string}, because they are two different
>>         packages. I wanted a simple way to be able to have the
>>         package name : kernel-4.9-{static_string}, then I could do
>>         the upgrade for the new minor updates of the kernel.
>>
>>
>>     I could have sworn this (upgrading) was already possible via the
>>     version string we
>>     are currently using.  i.e. the PV is already picked up from the
>>     kernel source, and
>>     that should be doing the job.
>>
>>     i.e. when I unpack my
>>     kernel-image-bzimage-4.10.15-yocto-standard_4.10.15+git0+4d929fac34_d2c1ed3c0c-r0_qemux86_64.ipk
>>     package, I see that it has:
>>
>>      Version: 4.10.15+git0+4d929fac34_d2c1ed3c0c-r0
>>      but obviously has the general provides: Provides:
>>     kernel-image-bzimage
>>
>>     So that should be upgradable based on the version ... sure they
>>     have different names, but the provides
>>     and versions take care of things.
>>
>>     The versioning, ability to install multiple kernels, upgrades,
>>     etc, have really churned
>>     these variables making them a mess to read.
>>
>>     I'm probably misunderstanding your use case and error, can you
>>     elaborate for me
>>     and/or provide a log ? I'm more of a kernel guy than a package
>>     format guy .. so
>>     I'm probably missing something obvious.
>>
>>     Bruce
>>
>     If I build a genericx86_64 core-image_sato without this patch,
>     only with the other patch which add the versions(only the parts
>     that are in round parenthesis) I have the following packages:
>
>     Package: kernel-4.10.9-yocto-standard
>     Version: 4.10.9+git0+ad2e885015_fe0fb8da3d-r0
>     Depends: kernel-image-4.10.9-yocto-standard (=
>     4.10.9+git0+ad2e885015_fe0fb8da3d-r0)
>     Provides: kernel-4.10.9-yocto-standard, kernel-base
>
>     Package: kernel-image-4.10.9-yocto-standard
>     Version: 4.10.9+git0+ad2e885015_fe0fb8da3d-r0
>     Depends: kernel-image-bzimage-4.10.9-yocto-standard (=
>     4.10.9+git0+ad2e885015_fe0fb8da3d-r0)
>     Provides: kernel-image
>
>     Package: kernel-module-6lowpan-4.10.9-yocto-standard
>     Version: 4.10.9+git0+ad2e885015_fe0fb8da3d-r0
>     Depends: kernel-4.10.9-yocto-standard (=
>     4.10.9+git0+ad2e885015_fe0fb8da3d-r0)
>     Provides: kernel-module-6lowpan
>
>     So the problem are the Depends field. We have the 4.10.9 part in
>     the dependents that is messing with the upgrade. If we want to do
>     only some minor update that doesn't change this value than
>     everything is ok. It works. But if we have a new value e.g.
>     4.10.10 than we couldn't do the upgrade.
>     So that's why I came up with this patch to be able to modify the
>     Depends value to something like: kernel-image-4.10-yocto-standard,
>     or even kernel-image.
>
>
> That's the point I'm trying to make.
>
> Patch 1/2 adds the depends, but there's no way to disable it. Without 
> defining the variable
> that you have in patch 2/2, you can't upgrade the kernel packages .. 
> which makes it not only
> a new variable, but a new requirement when working with the kernel 
> recipes/packages.
>
> People have been upgrading the kernel before either of these patches 
> (I can't say that
> I do very often .. but I know that users of the commercially supported 
> distros do) .. unless
> it has been inadvertently broken, or there are more patches floating 
> around that I've not
> seen.
>
> what is the upgrade behaviour with neither of these patches applied ? 
> .. that's the behaviour that
> I'm most concerned about maintaining.
>
> Also, if we were to introduce something like this, the series needs to 
> have documentation
> updates along with it .. or we'll surely forget to document it and 
> catch people by surprise :(
>
> Bruce

You misunderstood the first patch. It only appends the version to the 
packages in the Depends field. E.g (from the pyro branch):

Package: kernel-4.10.17-yocto-standard
Version: 4.10.17+git0+4b89f331d4_6648a34e00-r0
Depends: kernel-image-4.10.17-yocto-standard
Provides: kernel-4.10.17-yocto-standard, kernel-base

Package: kernel-image-4.10.17-yocto-standard
Version: 4.10.17+git0+4b89f331d4_6648a34e00-r0
Depends: kernel-image-bzimage-4.10.17-yocto-standard
Provides: kernel-image

Package: kernel-module-6lowpan-4.10.17-yocto-standard
Version: 4.10.17+git0+4b89f331d4_6648a34e00-r0
Depends: kernel-4.10.17-yocto-standard
Provides: kernel-module-6lowpan

As you can see the difference from the last email was only the versions 
in the Depends field.
About the upgrades, I only manage to upgrade the kernel only for new 
revisions version, from kernel-4.10.17-yocto-standard version 
4.10.17+git0+4b89f331d4_6648a34e00-r0 to kernel-4.10.17-yocto-standard 
version 4.10.17+git0+4b89f331d4_6648a34e00-r1. I tried to do an upgrade 
from kernel-4.10.15-yocto-standard to kernel-4.10.17-yocto-standard, but 
it failed (For the upgrade tests I used ipk format, might be the new 
opkg solver).

About the other distros based on OE, maybe they have some custom recipe 
for the kernel and somehow achieve the the same behaviour as this patch, 
or use rpm/deb, or the have custom scripts for the kernel and modules 
upgrades, or something else, idk.

The idea of these patches is to have a simple and easy way to do a 
kernel upgrade(with the kernel image and installed kernel modules), 
because now, when you want to upgrade the kernel you need to manually do 
the upgrade for the kernel-image and the modules.

>>
>>         This was the simple and cleanest way I could think of to
>>         achieve the my scenario. But if there is a better idea for
>>         this, let me know.
>>
>>         -- 
>>
>>         Razvan
>>
>>
>>
>>
>>     -- 
>>     "Thou shalt not follow the NULL pointer, for chaos and madness
>>     await thee at its end"
>
>     -- 
>
>     Razvan
>
>
>
>
> -- 
> "Thou shalt not follow the NULL pointer, for chaos and madness await 
> thee at its end"

-- 

Razvan

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20170703/8c3b9106/attachment-0002.html>


More information about the Openembedded-core mailing list