Re: releng/10.3 rc netwait disabled by ipfw dependency

[John Marshall]

On Mon, 14 Mar 2016, 08:04 -0600, Ian Lepore wrote:
> On Sat, 2016-03-12 at 16:46 +1100, John Marshall wrote:
> > Adding ipfw to the netwait REQUIRE list makes things work for me (see
> > Bug 207916) but I am sure there is more that ought to be taken into
> > account and perhaps the dependency ought to be resolved in a
> > different
> > way.
> 
> Just a quick followup for the archives, this was turned into PR 207916
> and a workaround was committed as r296807.  A better fix, I think,
> would be to separate waiting for the network interfaces to appear from
> waiting for network connectivity to be functional (ping test), but I
> ran into some circular-dependency problems, so I committed a "good
> enough for now" workaround.

Thank you Ian.  All works well for me in 10.3-RC3.

I am also of the opinion that separate scripts is the way to go.

-- 
John Marshall


signature.asc
Description: PGP signature

Re: releng/10.3 rc netwait disabled by ipfw dependency

[Ian Lepore]

On Sat, 2016-03-12 at 16:46 +1100, John Marshall wrote:
> I recently upgraded a server from 10.2-RELEASE to 10.3-RC2.  The
> server
> has ipfw configured and also has netwait enabled in rc.conf.  All was
> good in 10.2-RELEASE.
> 
> Now, rc pauses during netwait for the full ${netwait_timeout} failing
> to
> ping ${netwait_ip}, prints a WARNING about the network likely not
> being
> properly up, then proceeds to enable the firewall and start other
> components with varying degrees of success (now that the interfaces
> can
> pass packets and has started configuring itself - think IPv6 SLAAC).
> The netwait rc script used to "REQUIRE: NETWORKING", until r294680
> when
> (among other things) this happened...
> 
>  # PROVIDE: netwait
> -# REQUIRE: NETWORKING
> +# REQUIRE: devd routing
>  # KEYWORD: nojail
> 
> Adding ipfw to the netwait REQUIRE list makes things work for me (see
> Bug 207916) but I am sure there is more that ought to be taken into
> account and perhaps the dependency ought to be resolved in a
> different
> way.
> 

Just a quick followup for the archives, this was turned into PR 207916
and a workaround was committed as r296807.  A better fix, I think,
would be to separate waiting for the network interfaces to appear from
waiting for network connectivity to be functional (ping test), but I
ran into some circular-dependency problems, so I committed a "good
enough for now" workaround.

-- Ian

___
freebsd-stable@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"