From mboxrd@z Thu Jan 1 00:00:00 1970 From: Harald Hoyer Subject: Re: Netboot fails with multiple interfaces, wireless Date: Wed, 20 May 2009 15:14:23 +0200 Message-ID: <4A14022F.8000809@redhat.com> References: <4A137E3C.1040204@redhat.com> <4A13AD3B.4010606@bfh.ch> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4A13AD3B.4010606-omB+W0Dpw2o@public.gmane.org> Sender: initramfs-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: Content-Type: text/plain; charset="us-ascii"; format="flowed" To: Seewer Philippe Cc: Warren Togami , "initramfs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" On 05/20/2009 09:11 AM, Seewer Philippe wrote: > > > Warren Togami wrote: >> Harald, >> >> I tried your root=dhcp netboot code. >> >> Is it intended that you need both root=dhcp and ip=dhcp? Surely the >> first implies the second? > > I guess a simple patch could set ip=dhcp if root=dhcp is supplied... > >> >> So it does successfully mount and NFS root boot, but as I feared it >> has trouble due to the presence of a wireless device. One thin client >> I have is Intel Atom with e1000e NIC and iwlagn wireless. It appears >> to dhcp once successfully on the eth0 ethernet, but its event-driven >> nature also tries to launch dhclient again on the wlan0. dhclient >> complains that dhclient is already running and fails. > > That's weird it shouldn't do that, since dhclient is started with an > interface option. > >> dhclient wouldn't have succeeded at all on the iwlagn interface >> because the firmware is lacking from the initrd, and it isn't >> associated. Also I don't see when it is ever desired to netboot over a >> wireless interface. >> >> As for excluding wireless interfaces, we should do a few things: >> - Do not copy wireless drivers into the initrd. You suggested using >> /etc/modprobe.d/something.conf to blacklist the driver from loading. >> We would need a way to blacklist the driver during the initrd without >> blacklisting it on the host system where the initrd is generated. How? >> - Perhaps we should exclude any interfaces named "wlan*" from ifup and >> dhclient. This works for most devices, but some wireless drivers call >> the wireless interfaces "eth*". >> - I have no idea if it is attempting to ifup and dhclient other >> devices that appear like wmaster0. > > I agree there. Wireless drivers shouldn't be included (yet). As for > separating ethernet from wireless drivers, why not use 'nm' and > disregard those drivers that contain 'ieee802011_register_hw'? > >> >> A separate but related matter is what to do if a machine has more than >> one ethernet interface. It seems the only sane thing you CAN do is >> ifup and dhclient only on the first interface? > > That depends on your definition of sane. See the nfs/networking patches > I sent in march. They're be able to handle all kinds of weird combinations. > > Regards, > Philippe Philippe, could you try to rebase your patches on the latest git? If not, I will do so next week. Warren, nfs now works for me for: ip=dhcp root=dhcp, root=192.168.1.1:/mnt/root, root=nfs://192.168.1.1/mnt/root -- To unsubscribe from this list: send the line "unsubscribe initramfs" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html