Bug#385226: initramfs-tools: no need to set SERVER_IP in scripts/nfs
On Wed, 30 Aug 2006, vagrant+bugs@freegeek.org wrote:
> On Wed, Aug 30, 2006 at 08:36:58AM +0200, maximilian attems wrote:
> > On Tue, Aug 29, 2006 at 03:32:50PM -0700, Vagrant Cascadian wrote:
> > > i believe ipconfig defines ROOTSERVER when it is passed via the ip=
> > > parameter, and thus it's not necessary to re-implement this code with
> > > SERVER_IP in initramfs-tools/scripts/nfs
> >
> > thanks for your check!
> > need to see why it doesn't do it for NEW_DEVICE?
>
> well, the NEW_DEVICE code i wrote was merely a way so that when it hits:
>
> . /tmp/net-${DEVICE}.conf
ah indeed that's nice forgot about that.
> it's actually sourcing the right file... though maybe we could just do
> something like:
>
> . /tmp/net-*.conf
>
> some risk of getting some other device that way, even though *usually*
> there will only be one...
yes nack aboves star expansion is ugly.
> alternately, maybe we could get ipconfig to write to a different
> filename that doesn't include the device in it?
hmm let me think about it,
postboned for now.
> > > attached is a patch which removes this extra code.
> >
> > i'll double check and will merge - reduces complexity.
>
> not only complexity, but using shell parameter expansion instead of cut
> is definitely prone to error, too. so we should avoid it when possible
> :)
sure but quite solid if you test
> > very appreciated.
> > i assume that the current retry thing is working for you?
>
> yes, it works quite well, although i still think it needs a call to
> "sleep" to slow down the retry attempts(especially when not using dhcp
> to configure).
ok i'll retest that and can put an sleep 0.1 inside before it retires,
i have no idea to what timing the current delay of 300 corresponds,
especially if you get nasty dhcp timeouts..
> overall, the nfsroot support is looking pretty good now.
>
> live well,
> vagrant
thanks a lot for the input.
--
maks
Reply to: