[OE-core] [PATCH] connman: get the correct network interface name from dmesg during NFS booting

[Christopher Larson]

On Tue, Oct 11, 2016 at 12:03 PM, Jagadeesh Krishnanjanappa <
jkrishnanjanappa at mvista.com> wrote:

> On Tue, Oct 11, 2016 at 9:06 PM, Christopher Larson <clarson at kergoth.com>
> wrote:
>
>>
>> On Tue, Oct 11, 2016 at 5:11 AM, Jagadeesh Krishnanjanappa <
>> jkrishnanjanappa at mvista.com> wrote:
>>
>>> Thanks for reviewing the patch.
>>>
>>>
>>>> I think this is too fragile to land in OE-Core. What happens if a
>>>> network driver prints "device=eth0"? Or if other kernel messages make
>>>> the string disappear from dmesg and connman gets restarted later?
>>>>
>>>
>>> True. If device=eth0 gets disappeared/corrupted, then we may have
>>> problem.
>>>
>>>>
>>>> FWIW, I'm attaching my current wrapper script for connmand. I don't
>>>> think it's perfect, but at least it doesn't depend on any init manager
>>>> and works across restarts.
>>>
>>> The wrapper script attached by you, takes care of the missing scenarios.
>>> Seems to be more complete.
>>>
>>>>
>>>>
>>> Add these lines to connman's do_install, in case you'd like to test
>>>> and/or submit it:
>>>>
>>>> mv ${D}${sbindir}/connmand ${D}${sbindir}/connmand.real
>>>> install -m 755 ${WORKDIR}/connmand-nfsroot.in ${D}${sbindir}/connmand
>>>> sed -e 's, at sbindir@,${sbindir},g' -i ${D}${sbindir}/connmand
>>>>
>>> I think it would be good idea to integrate your changes into the already
>>> existing OE-core's connman script, instead of a calling original connman
>>> script from the wrapper script.
>>>
>>
>> I threw together https://github.com/openembedded/openembedded-core/
>> compare/master...kergoth:connman-systemd-nfs last night, only limited
>> testing. I think using a script in libexecdir is rather cleaner than
>> wrapping connmand in place, for something like this, personally.
>>
>
> In the above commit,  the last line of connman/connman/start-connman file,
> should be exec @SBINDIR@/connmand "$@" instead of exec @SBINDIR@/connman
> "$@" ; as the actual binary file name is connmand.


Thanks, didn’t notice that. Obviously it needs polish and testing before
submission, I mainly linked it to solicit thoughts on the approach :)
-- 
Christopher Larson
clarson at kergoth dot com
Founder - BitBake, OpenEmbedded, OpenZaurus
Maintainer - Tslib
Senior Software Engineer, Mentor Graphics
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20161011/0bbd0c30/attachment-0002.html>