Hi everybody,
The good news is that I finally succeeded to boot over network
using the PXE-bootrom / ip dhcp autoconfig / nfsroot method.
But "ip=dhcp root=/dev/nfs nfsroot=%s" is not the way to go
with recent kernels. This results in a kernel panic caused by
the inability to find root. Things go wrong immediately after
rpc port lookup:
> NFS: Buggy server - nlink == 0!
> nfs_fhget failed
Well, adding ",v3" to the nfsroot parameter helps, forcing the
client not to use the default nfs version 2 but nfs version 3.
All machines use kernel 2.6.15-git7.
cu,
Knut
On Wed, 2006-02-01 at 07:07 +0100, Knut Petersen wrote:
> Hi everybody,
>
> The good news is that I finally succeeded to boot over network
> using the PXE-bootrom / ip dhcp autoconfig / nfsroot method.
> But "ip=dhcp root=/dev/nfs nfsroot=%s" is not the way to go
> with recent kernels. This results in a kernel panic caused by
> the inability to find root. Things go wrong immediately after
> rpc port lookup:
>
> > NFS: Buggy server - nlink == 0!
> > nfs_fhget failed
What kind of server is this that you are using? The above message
basically means that it is handing out inodes with a link count of 0.
Cheers,
Trond
On Wed, 2006-02-01 at 07:34 +0100, Knut Petersen wrote:
> Trond Myklebust schrieb:
>
> >What kind of server is this that you are using? The above message
> >basically means that it is handing out inodes with a link count of 0.
> >
> >Cheers,
> > Trond
> >
> >
>
> Server:
> =====
>
> AOpen i915GMm-HFS with Pentium M, linux kernel 2.6.15-git7
> running a system that startet as SuSE 9.2. Nfs-utils are still
> the original 1.0.6, grep -i nfs linuxbuild/.config gives
...and what kind of filesystem are you exporting?
Cheers,
Trond
Trond Myklebust schrieb:
>What kind of server is this that you are using? The above message
>basically means that it is handing out inodes with a link count of 0.
>
>Cheers,
> Trond
>
>
Server:
=====
AOpen i915GMm-HFS with Pentium M, linux kernel 2.6.15-git7
running a system that startet as SuSE 9.2. Nfs-utils are still
the original 1.0.6, grep -i nfs linuxbuild/.config gives
CONFIG_NFS_FS=y
CONFIG_NFS_V3=y
# CONFIG_NFS_V3_ACL is not set
# CONFIG_NFS_V4 is not set
# CONFIG_NFS_DIRECTIO is not set
CONFIG_NFSD=y
CONFIG_NFSD_V3=y
# CONFIG_NFSD_V3_ACL is not set
# CONFIG_NFSD_V4 is not set
CONFIG_NFSD_TCP=y
CONFIG_NFS_COMMON=y
Client
====
kernel build from the same source,
CONFIG_NFS_FS=y
CONFIG_NFS_V3=y
# CONFIG_NFS_V3_ACL is not set
# CONFIG_NFS_V4 is not set
# CONFIG_NFS_DIRECTIO is not set
CONFIG_NFSD=y
CONFIG_NFSD_V3=y
# CONFIG_NFSD_V3_ACL is not set
# CONFIG_NFSD_V4 is not set
CONFIG_NFSD_TCP=y
CONFIG_ROOT_NFS=y
CONFIG_NFS_COMMON=y
cu,
Knut
>>AOpen i915GMm-HFS with Pentium M, linux kernel 2.6.15-git7
>>running a system that startet as SuSE 9.2. Nfs-utils are still
>>the original 1.0.6, grep -i nfs linuxbuild/.config gives
>>
>>
>
>...and what kind of filesystem are you exporting?
>
>
>
linux:/src/k6bv3p # dmesg | grep sda
[4294667.296000] Kernel command line: root=/dev/sda3
video=intelfb:vram=8 vga=0x307
[ 46.848978] SCSI device sda: 488397168 512-byte hdwr sectors (250059 MB)
[ 46.850423] sda: Write Protect is off
[ 46.851871] sda: Mode Sense: 00 3a 00 10
[ 46.851884] SCSI device sda: drive cache: write back w/ FUA
[ 46.853385] SCSI device sda: 488397168 512-byte hdwr sectors (250059 MB)
[ 46.854868] sda: Write Protect is off
[ 46.856330] sda: Mode Sense: 00 3a 00 10
[ 46.856342] SCSI device sda: drive cache: write back w/ FUA
[ 46.857821] sda: sda1 sda2 sda3
[ 46.869751] sd 0:0:0:0: Attached scsi disk sda
[ 20.839566] ReiserFS: sda3: found reiserfs format "3.6" with standard
journal
[ 22.412053] ReiserFS: sda3: using ordered data mode
[ 22.417920] ReiserFS: sda3: journal params: device sda3, size 8192,
journal first block 18, max trans len 1024, max batch900, max commit age
30, max trans age 30
[ 22.419518] ReiserFS: sda3: checking transaction log (sda3)
[ 22.440517] ReiserFS: sda3: Using r5 hash to sort names
[ 24.597554] Adding 1052216k swap on /dev/sda1. Priority:42 extents:1
across:1052216k
(printk timestamp anomalities caused by cpu speed)
cu,
Knut
cu,
Knut
>>AOpen i915GMm-HFS with Pentium M, linux kernel 2.6.15-git7
>>running a system that startet as SuSE 9.2. Nfs-utils are still
>>the original 1.0.6, grep -i nfs linuxbuild/.config gives
>>
>>
>
>...and what kind of filesystem are you exporting?
>
>
>
I think it is _not_ related to reiserfs. Moving my
exported /tftpboot directory to a fresh ext2 partition
gave the same results - failing with nfs 2, succeeding with
nfs 3.
cu,
Knut
On Wed, 2006-02-01 at 08:13 +0100, Knut Petersen wrote:
> >>AOpen i915GMm-HFS with Pentium M, linux kernel 2.6.15-git7
> >>running a system that startet as SuSE 9.2. Nfs-utils are still
> >>the original 1.0.6, grep -i nfs linuxbuild/.config gives
> >>
> >>
> >
> >...and what kind of filesystem are you exporting?
> >
> >
> >
>
> I think it is _not_ related to reiserfs. Moving my
> exported /tftpboot directory to a fresh ext2 partition
> gave the same results - failing with nfs 2, succeeding with
> nfs 3.
Does it do the same if you mount the same partition normally (i.e. not
through nfsroot) in some other directory?
Cheers,
Trond
Trond Myklebust wrote:
>Does it do the same if you mount the same partition normally (i.e. not
>through nfsroot) in some other directory?
>
>
>
That also fails. Same error message.
Unlike the nfsroot code at least my version of mount
does not use nfs version 2 as the default. I had to force
a v2 nfs mount attempt with mount -o nfsvers=2 ...
Would it be a good idea to change the default nfs version
nfsroot uses? I think nfsroot and mount defaults should
be identical.
BTW: Google shows some related old threads, e.g. "Madhan" writes on
2 Aug. 2001 06:55 "There has been a change in the NFS Client
behaviour in Linux Kernel 2.4.3 onwards. There are 2 issues here,
1. as traces show new clients expect link count '1' and NetWare
NFS has been sending '0' for volumes. ..."
cu,
Knut