[meta-freescale] [PATCH] wks: add a kickstart file with vfat boot and ext4 rootfs partition

Otavio Salvador otavio.salvador at ossystems.com.br
Thu Sep 28 05:46:50 PDT 2017


On Thu, Sep 28, 2017 at 9:31 AM, Ed Bartosh <ed.bartosh at linux.intel.com> wrote:
> On Fri, Sep 22, 2017 at 10:21:57AM -0300, Otavio Salvador wrote:
>> Hello Max,
>>
>> On Fri, Sep 22, 2017 at 8:48 AM, Max Krummenacher <max.oss.09 at gmail.com> wrote:
>> > 2017-09-21 22:51 GMT+02:00 Otavio Salvador <otavio.salvador at ossystems.com.br>:
>> >> I am adding Ed, Richard and Ross on Cc here so we can discuss it further.
>> >>
>> >> On Thu, Sep 14, 2017 at 5:08 AM, Max Krummenacher <max.oss.09 at gmail.com> wrote:
>> >>> Hi
>> >>>
>> >>>
>> >>>
>> >>> 2017-09-13 23:20 GMT+02:00 Stefan Agner <stefan.agner at toradex.com>:
>> >>>>
>> >>>> I guess / is a bit different here since it just uses /dev/root, in the end the mmcblk comes from the boot arguments (root=).
>> >>>>
>> >>>> One could probably just define the proper mmcblkX with the --disk argument, but that does not work since mmcblk devices are not stable. E.g. if you use the second SDHC controller for an eMMC, and the first SDHC controller is available externally for SD cards, you root device is mmcblk0 if there is no SD card available, and mmcblk1 if there is one available.. At least on mainline that is the case. There are patches floating around which provide a more stable block device name. But the official answer is PARTUUID. So I guess one would have to teach wks PARTUUID support for fstab generation...
>> >>>>
>> >>>> Personally I also like to have /boot mounted to the partition where the kernels are, even if that is an external FAT partition... That is also the case on regular Linux distros booting with UEFI.
>> >>>>
>> >>>> But Max approach would work around that issue for now until PARTUUID support is available...
>> >>>>
>> >>>> --
>> >>>> Stefan
>> >>>
>> >>> As said in the patch this commit to oe-core changed the behaviour of wic.
>> >>> http://cgit.openembedded.org/openembedded-core/commit/scripts/lib/wic/plugins/imager/direct.py?id=2376b05512ddb8c4ec3aaf1df11071f536a76bd9
>> >>>
>> >>> Before that wic added a line to fstab for each partition which had a
>> >>> mountpoint set if the mountpoint was not '/' or '/boot'. With the
>> >>> patch only '/' is not added to to fstab.
>> >>> And the commit text clearly states that this exception is removed
>> >>> because a user who does not want that entry can simply omit the
>> >>> mountpoint in the kickstarter file.
>> >>>
>> >>> Which is why I created a new kickstarter file which does exactly what
>> >>> the commit text proposes.
>> >>
>> >> I understand the need here for a new wks file however it is a little
>> >> confusing for end user that sdcard-image.wks does not work anymore for
>> >> general users.
>> >>
>> >> Before we start adding new wks files, I think we have to discuss this.
>> >> We did propose https://patchwork.openembedded.org/series/6418/# which
>> >> seemed to be the best approach for reproducible builds as we provide a
>> >> final fstab which is ready for use by the device and keep the wks
>> >> files compatible.
>> >>
>> >> What you guys think?
>> >
>> > I like this generic solution, thanks.
>>
>> So do I.
>>
>> > I tried it out, with Fabio's patch my issue with using
>> > sdimage-bootpart.wks is solved.
>>
>> :-) Good news. Ed?
>>
>
> I'm not against it after your explanation while ago.
>
> +1
>
> I think the patchset should be rebased on current master and resent for
> review.

It was merged already.


-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750


More information about the meta-freescale mailing list