Bug 389672

Created attachment 214607 [details]
run_all.sh

resulting run_all.sh

Maybe YaST fell into the trap of the changed dhcpcd-<if>.info file format?

This is issue for Steffen:
>> mkinitrd is unable ...
>> ... has no /etc/resolv.conf during first stage

yast2-network comes later ;-)

re comment #3: YaST don't read that file directly

I hoped so, but grep showed a few matches.

Olaf, can you check if linuxrc leaves a resolv.conf? I see no problem on
my test machine.

I said that name resolving works on in inst-sys.

But the target directory has no resolv.conf.
So I assume its currently not well defined how name resolving should be done inside the target chroot when mkinitrd creates the initrd with nfsroot support.
install.inf states that dhcp was used, but mkinitrd falls back to static ip for some reason.

then it's not linuxrc

>> But for some reason, mkinitrd did fall back to static interface configuration.

then reassigned to mkinitrd maintainer

f200 is currently at the end of first stage, 11.0beta3-i386.

I havent verified if mkinitrd did the right thing.
booted from dvd with netsetup=1, nfsroot is boettger.suse.de:/data/inst/nfs/110b3-i386

tested with beta3 on snowberry, js20:




...

preping 03-storage.sh
running 03-storage.sh
preping 04-udev.sh
running 04-udev.sh
Creating device nodes with udev
tg3.c:v3.91 (April 18, 2008)
PCI: Enabling device 0000:11:01.0 (0140 -> 0142)
eth0: Tigon3 [partno(none) rev 2003 PHY(serdes)] (PCIX:133MHz:64-bit) 1000Base-SX Ethernet 00:0d:60:1e:cc:f4
eth0: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[1] WireSpeed[0] TSOcap[0]
eth0: dma_rwctrl[769f4000] dma_mask[64-bit]
PCI: Enabling device 0000:11:01.1 (0140 -> 0142)
eth1: Tigon3 [partno(none) rev 2003 PHY(serdes)] (PCIX:133MHz:64-bit) 1000Base-SX Ethernet 00:0d:60:1e:cc:f5
eth1: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[0] WireSpeed[0] TSOcap[1]
eth1: dma_rwctrl[769f4000] dma_mask[64-bit]
preping 05-blogd.sh
running 05-blogd.sh
blogd: can not open pty/tty pair: No such file or directory
preping 11-network.sh
running 11-network.sh
FATAL: Module af_packet not found.
[NETWORK] using interface eth1
[NETWORK] using static config based on ip=10.10.2.220::10.10.0.8:255.255.0.0:snowberry:eth1:none
Ignoring mode none, using static configuration
preping 11-usb.sh
running 11-usb.sh
preping 21-devinit_done.sh
running 21-devinit_done.sh
preping 21-nfs.sh
running 21-nfs.sh
RPC: Registered udp transport module.
RPC: Registered tcp transport module.
preping 81-kdump.sh
running 81-kdump.sh
preping 82-resume.userspace.sh
running 82-resume.userspace.sh
preping 83-resume.kernel.sh
running 83-resume.kernel.sh
preping 84-mount.sh
running 84-mount.sh
Mounting root whortleberry.suse.de:/nfsroot/110b3-ppc
mount.nfs: can't get address for whortleberry.suse.de

could not mount root filesystem -- exiting to /bin/sh
$ tg3: eth1: Link is up at 1000 Mbps, full duplex.
tg3: eth1: Flow control is off for TX and off for RX.

and before that:

...Freeing unused kernel memory: 324k freed
doing fast boot
mount used greatest stack depth: 11792 bytes left
mknod used greatest stack depth: 11088 bytes left
boot/02-start.sh: line 70: cmd_--login=--login: command not found
SCSI subsystem initialized
libata version 3.00 loaded.
pata_amd 0000:00:04.1: version 0.3.10
scsi0 : pata_amd
scsi1 : pata_amd
ata1: PATA max UDMA/133 cmd 0x17400 ctl 0x16c00 bmdma 0x17c00 irq 17
ata2: PATA max UDMA/133 cmd 0x17800 ctl 0x17000 bmdma 0x17c08 irq 17
ata1.00: ATA-6: TOSHIBA MK6026GAXB, PB204E, max UDMA/100
ata1.00: 117210240 sectors, multi 16: LBA
ata1.00: limited to UDMA/33 due to 40-wire cable
ata1.00: configured for UDMA/33
RTAS: event: 44, Type: Unknown, Severity: 2
scsi 0:0:0:0: Direct-Access     ATA      TOSHIBA MK6026GA PB20 PQ: 0 ANSI: 5
modprobe used greatest stack depth: 8976 bytes left
....

$ cat /proc/cmdline 
debug panic=12 sysrq=1 init=/bin/bash --login  
$ 

nothing past init= can be a kernel cmdline option. 
everything past init= is passed as option to init

Created attachment 217251 [details]
ignore-args-after-init=

Ignore all commandline arguments after 'init='.

Submitted updated mkinitrd rpm to autobuild.

still not fixed with current factory.

if rootdev=boettger:/nfsroot/factory, boettger cant be resolved (appearently).
Maybe dhcpcd should be ran with the -M option

I see that mount.nfs is missing, and also /etc/resolv.conf is not created during boot.

boettger:/data/inst/nfs/factory/tmp/x # cat ../../var/log/YaST2/y2logmkinitrd 

Kernel image:   /boot/vmlinuz-2.6.25.4-8-pae
Initrd image:   /boot/initrd-2.6.25.4-8-pae
Kernel Modules: processor thermal dock scsi_mod libata ata_piix fan edd af_packet tg3 usbcore ohci-hcd uhci-hcd ehci-hcd ff-memless hid usbhid sunrpc nfs_acl lockd nfs 
Features:       network usb nfs resume.userspace resume.kernel
21880 blocks
boettger:/data/inst/nfs/factory/tmp/x # find . -name "*mount*"
./bin/mount
./bin/umount
./boot/85-remount.sh
./boot/84-mount.sh
./sbin/mount.nfs
./config/mount.sh
boettger:/data/inst/nfs/factory/tmp/x # find . -name "*resolv*"
./lib/libresolv.so.2
./lib/libresolv-2.8.so
boettger:/data/inst/nfs/factory/tmp/x # 

dhcpcd was called with wrong options, -p is missing for example
resolv.conf creation fails because the dhcpcd info fileformat changed, IFS should be a space instead of a comma.
There may be more bugs to resolv to get nfsroot finally working.

no idea when nfsroot became ship stopper for openSUSE. Possibly for sle11

Created attachment 220455 [details]
mkinitrd.nfsroot.patch

nss files missing, dhcpcd info file format changed, dhcpcd path never taken, dhcp timeout too short (not configureable)

Updated mkinitrd rpm submitted to STABLE.

*** Bug 398558 has been marked as a duplicate of this bug. ***

Actually, init= does _not_ terminate the argument parsing. At least not for the normal kernel. So introducing that in our initrd would be incompatible behaviour.
Thus reverting the change.

The init= issue will be tracked in bug #403995.

*** Bug 406493 has been marked as a duplicate of this bug. ***