dhcpcd-discuss

Hello Roy san,

I tried your patch. Unfortunately I cannot use this patch so far.
Because "dhcpcd -n eth0" causes dhcpcd to handle all existing network
interface even if I only allow both eth0 and wlan0 for dhcpcd in dhcpcd.conf:
  allowinterfaces wlan0,eth0
As the result, that confuses our system.

In addition, I needed to ask my designers to modify their code again if I would
decide to use this patch as well as I would take the time to check whether all
specifications could keep current stable behaviors.

I remember I also choose the way to use a pair of "dhcpcd -n" and "dhcpcd -k"
on my first integration. The above mentioned is one reason and I seemed to meet
more issues.

Hence, I would like to know your opinion for my modification although it's not
clean. My fundamental testing results on both IPv4 and IPv6 environment are OK
so far. If you don't have any problem for this modification, I'll use it so that
we can use latest dhcpcd version with both our reasonable cost and yours maybe.

Best Regards,
Koichi Okamoto

> -----Original Message-----
> From: Okamoto, Koichi
> Sent: Thursday, July 28, 2016 10:04 AM
> To: 'Roy Marples'; dhcpcd-discuss@xxxxxxxxxxxx
> Subject: RE: [dhcpcd-discuss] [dhcpcd-6.9.2] share these resources leak
> issues patch
> 
> Hello Roy san,
> 
> Thank you for your advice, test and solution with patch!
> 
> > I could not replicate this at all, at least, not on NetBSD. Which OS
> > are you using?
> 
> We use Linux 3.10.26.
> 
> LINK_UP may be one of condition when dhcpcd start with master mode and then
> "dhcpcd --exit <interface node name>. Second dhcpcd starts also under
> LINK_UP condtion. (It means that my step1 and my step2 under LINK_UP being
> kept.)
> 
> I am wondering if there is a mismatch between dhcpcd status and OS status.
> 
> > Also, I dislike the way you want to start dhcpcd .... it's not clean!
> 
> Me too. However I could not find out the solution when I studied to integrate
> dhcpcd on our system. This way has being able to work as our expectation
> on dhcpcd-6.9.2.
> 
> > So I made this patch to add an --inactive option:
> > http://roy.marples.name/projects/dhcpcd/info/6e2698370eaa3aad
> >
> > So you can do this:
> 
> I try this patch and tell you the results soon.
> 
> Your informations are very helpful for me!
> 
> Thanks and Best Regards,
> Koichi Okamoto
> 
> > -----Original Message-----
> > From: Roy Marples [mailto:roy@xxxxxxxxxxxx]
> > Sent: Thursday, July 28, 2016 1:53 AM
> > To: Okamoto, Koichi; dhcpcd-discuss@xxxxxxxxxxxx
> > Subject: Re: [dhcpcd-discuss] [dhcpcd-6.9.2] share these resources
> > leak issues patch
> >
> > Hi
> >
> > On 27/07/2016 13:18, Okamoto, Koichi wrote:
> > > I am trying to update dhcpcd-6.11.1 from 6.9.2 so that I can use
> > 6.11.2 short term
> > > when you will release dhcpcd-6.11.2.
> > >
> > > However I met one issue which always remained in waiting for carrier
> > condition.
> > >
> > > The attached patch
> > (0001-Candidated-modification-way-to-meet-our-operation.patch)
> > > can resolve this issue.
> > >
> > > If possible, could you give me your advice whether it is reasonable
> > modification?
> > >
> > > I try to explain what I would like to do with some assumption.
> > >
> > > [Background]
> > > As the cause of issue may be not appropriate our operation, we have
> > > to
> > control
> > > dhcpcd functionality with master mode which always stays in one
> > process during
> > > CPU on.
> > >
> > > In addition, we have to choose network interface one by one with
> > master mode.
> > > (Network interface are changed according to our system condition)
> > >
> > > To realize the above on dhcpcd-6.9.2, the following way is adopted:
> > >
> > > 1. First start up time
> > >    $ dhcpcd -M -b <network node name, e.g) eth0>
> > >    $ dhcpcd --exit <same as the above network node name, e.g) eth0>
> > >    Please note that we need only daemonised dhcpcd one process
> > > without
> > DHCP function
> > >
> > > 2. When DHCP function is needed
> > >    $ dhcpcd -M -b <network node name, e.g) eth0>
> > >
> > >
> > > 3. When DHCP function isn't needed
> > >    $ dhcpcd --exit <same as the above no.2 network node name, e.g)
> > > eth0>
> > >
> > > Repeat No.2 and No.3 accordingly. When CPU turns off, "$ dhcpcd
> > --exit" performs.
> >
> > I could not replicate this at all, at least, not on NetBSD. Which OS
> > are you using?
> >
> > Also, I dislike the way you want to start dhcpcd .... it's not clean!
> > So I made this patch to add an --inactive option:
> > http://roy.marples.name/projects/dhcpcd/info/6e2698370eaa3aad
> >
> > So you can do this:
> > # Start dhcpcd in master mode, but waiting for our instruction # to
> > start each interface.
> > dhcpcd -M -b --inactive
> >
> > # Start iwn0
> > dhcpcd -n iwn0
> >
> > # Wait some time
> > # We could use dhcpcd-online(8) instead sleep 10
> >
> > # Release all addresses from iwn0
> > dhcpcd -k iwn0
> >
> > # Restart
> > dhcpcd -n iwn0
> >
> > Hope this helps!
> >
> > Roy