NFS root & shutdown/reboot

NFS root & shutdown/reboot

[Paul Eggleton]

Hi all,

I have been working on smoothing out some things to do with NFS root this week 
and I have come across one situation that I'm not quite sure of. 

At the moment when you shut down a Poky NFS root-based system or reboot it, it 
gets all the way to the end and then hangs, with an error "nfs: server 
myserver not responding, still trying". I tracked this down to the -i option 
that is passed to the final call to shutdown and reboot which is supposed to 
shut down all network interfaces - if you remove this then it works fine.

I'm not sure of the reason for having the -i option, but it was suggested to 
me that it might be in order to release DHCP leases where applicable; also 
some systems might otherwise leave their network adapters powered up when the 
system is shut down in order to receive wake-on-LAN requests, and if you're 
not using that feature you're just wasting power. I'm guessing for most 
situations we would want to retain -i. However, where the system is running 
from an NFS root it would seem that we would have to omit this option, unless 
there is another solution I'm not aware of.

I've done some googling and not come up with anything concrete in terms of how 
this should be handled correctly. Any suggestions?

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre (UK)

Re: NFS root & shutdown/reboot

[Bruce Ashfield]

On Fri, Apr 15, 2011 at 8:29 AM, Paul Eggleton
<paul.eggleton@linux.intel.com> wrote:
> Hi all,
>
> I have been working on smoothing out some things to do with NFS root this week
> and I have come across one situation that I'm not quite sure of.
>
> At the moment when you shut down a Poky NFS root-based system or reboot it, it
> gets all the way to the end and then hangs, with an error "nfs: server
> myserver not responding, still trying". I tracked this down to the -i option
> that is passed to the final call to shutdown and reboot which is supposed to
> shut down all network interfaces - if you remove this then it works fine.
>
> I'm not sure of the reason for having the -i option, but it was suggested to
> me that it might be in order to release DHCP leases where applicable; also
> some systems might otherwise leave their network adapters powered up when the
> system is shut down in order to receive wake-on-LAN requests, and if you're
> not using that feature you're just wasting power. I'm guessing for most
> situations we would want to retain -i. However, where the system is running
> from an NFS root it would seem that we would have to omit this option, unless
> there is another solution I'm not aware of.

I can't say exactly when this is required either, but I can say that we've
also removed -i from our NFS root boots in the past, for just what you
are describing above.

Shutting down and getting the interface down in the wrong order causes
the obvious problems you'd expect with a NFS root hanging and not
being able to complete the rest of the process.

>
> I've done some googling and not come up with anything concrete in terms of how
> this should be handled correctly. Any suggestions?

My only suggestion is to make it not present by default, and enable
it (-i) when a real reason for it is found (i.e. someone really wants
the nic to shutdown for a power profile). Detecting a NFS root
(and if the root continues to be NFS after boot) could be more error
prone that making -i the default.

Just my opinion,

Bruce

>
> Cheers,
> Paul
>
> --
>
> Paul Eggleton
> Intel Open Source Technology Centre (UK)
> _______________________________________________
> poky mailing list
> poky@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/poky
>



-- 
"Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end"

Re: NFS root & shutdown/reboot

[Darren Hart]

On 04/15/2011 06:11 AM, Bruce Ashfield wrote:
> On Fri, Apr 15, 2011 at 8:29 AM, Paul Eggleton
> <paul.eggleton@linux.intel.com> wrote:
>> Hi all,
>>
>> I have been working on smoothing out some things to do with NFS root this week
>> and I have come across one situation that I'm not quite sure of.
>>
>> At the moment when you shut down a Poky NFS root-based system or reboot it, it
>> gets all the way to the end and then hangs, with an error "nfs: server
>> myserver not responding, still trying". I tracked this down to the -i option
>> that is passed to the final call to shutdown and reboot which is supposed to
>> shut down all network interfaces - if you remove this then it works fine.
>>
>> I'm not sure of the reason for having the -i option, but it was suggested to
>> me that it might be in order to release DHCP leases where applicable; also
>> some systems might otherwise leave their network adapters powered up when the
>> system is shut down in order to receive wake-on-LAN requests, and if you're
>> not using that feature you're just wasting power. I'm guessing for most
>> situations we would want to retain -i. However, where the system is running
>> from an NFS root it would seem that we would have to omit this option, unless
>> there is another solution I'm not aware of.
> 
> I can't say exactly when this is required either, but I can say that we've
> also removed -i from our NFS root boots in the past, for just what you
> are describing above.
> 
> Shutting down and getting the interface down in the wrong order causes
> the obvious problems you'd expect with a NFS root hanging and not
> being able to complete the rest of the process.
> 
>>
>> I've done some googling and not come up with anything concrete in terms of how
>> this should be handled correctly. Any suggestions?
> 
> My only suggestion is to make it not present by default, and enable
> it (-i) when a real reason for it is found (i.e. someone really wants
> the nic to shutdown for a power profile). Detecting a NFS root
> (and if the root continues to be NFS after boot) could be more error
> prone that making -i the default.
> 
> Just my opinion,
> 

Agreed. Perhaps this needs some kind of a machine config setting that
feeds into how shutdown is called, sort of like formfactor feeds into
matchbox?

--
Darren

> Bruce
> 
>>
>> Cheers,
>> Paul
>>
>> --
>>
>> Paul Eggleton
>> Intel Open Source Technology Centre (UK)
>> _______________________________________________
>> poky mailing list
>> poky@yoctoproject.org
>> https://lists.yoctoproject.org/listinfo/poky
>>
> 
> 
> 

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel