Re: NFS on little-endian platform - Microblaze

On Wed, 2011-02-16 at 14:53 +0100, Michal Simek wrote: 

Trond Myklebust wrote:

quoted

On Wed, 2011-02-16 at 14:16 +0100, Michal Simek wrote: 

quoted

Hi again,

I forget to cc linux-nfs mailing list.

Michal

P.S.: Tested on kernels 2.6.38-rc4, 2.6.37 and 2.6.36

Michal Simek wrote:

quoted

Hi All,

I am trying to understand one problem which we have found.
The problem is that I can't on Microblaze little-endian platform
mount nfs without -o nolock options. (Log below)
Selecting tcp or udp has no effect.
I am using emaclite driver and there is no problem on big endian 
microblaze.

ping, telnet, http, ftp, iperf, netperf work well.

That's why I have a question if there is any endian specific option for 
NFS?

Thanks,
Michal

~ # mount -t nfs 192.168.0.101:/tftpboot/nfs /mnt
svc: failed to register lockdv1 RPC service (errno 13).
lockd_up: makesock failed, error=-13
svc: failed to register lockdv1 RPC service (errno 13).
~ # mount -t nfs -o nolock 192.168.0.101:/tftpboot/nfs /mnt
~ # mount
rootfs on / type rootfs (rw)
proc on /proc type proc (rw,relatime)
none on /var type ramfs (rw,relatime)
none on /sys type sysfs (rw,relatime)
none on /etc/config type ramfs (rw,relatime)
none on /dev/pts type devpts (rw,relatime,mode=600)
192.168.0.101:/tftpboot/nfs on /mnt type nfs 
(rw,relatime,vers=3,rsize=32768,wsize=32768,namlen=255,hard,nolock,proto=udp,port=65535,timeo=7,retrans=3,sec=sys,local_lock=all,addr=192.168.0.101) 

~ #
~ # ps
PID   USER     TIME   COMMAND
    1 root       0:02 init
    2 root       0:00 [kthreadd]
    3 root       0:00 [ksoftirqd/0]
    4 root       0:00 [kworker/0:0]
    5 root       0:00 [kworker/u:0]
    6 root       0:00 [khelper]
    7 root       0:00 [sync_supers]
    8 root       0:00 [bdi-default]
    9 root       0:00 [kblockd]
   10 root       0:00 [rpciod]
   11 root       0:00 [kworker/0:1]
   12 root       0:00 [kswapd0]
   13 root       0:00 [fsnotify_mark]
   14 root       0:00 [aio]
   15 root       0:00 [nfsiod]
   16 root       0:00 [kworker/u:1]
   58 root       0:00 udhcpc -R -n -p /var/run/udhcpc.eth0.pid -i eth0
   62 1          0:00 /bin/portmap
   64 root       0:00 /bin/inetd /etc/inetd.conf
   65 root       0:01 -sh
   66 root       0:00 /bin/syslogd -n
   67 root       0:00 /bin/flatfsd
   68 root       0:00 [kworker/0:2]
   91 root       0:00 ps

Where is rpc.statd? Without it, the above behaviour is 100% expected.

I see on BE that lockd is used but it is enabled on little endian too but hasn't started.

Enabled options:
CONFIG_NETWORK_FILESYSTEMS=y
CONFIG_NFS_FS=y
CONFIG_NFS_V3=y
CONFIG_LOCKD=y
CONFIG_LOCKD_V4=y
CONFIG_NFS_COMMON=y
CONFIG_SUNRPC=y

On Be lockd is up.
    69 root       0:00 /bin/flatfsd
    71 root       0:00 [lockd]
    73 root       0:00 ps

I have to look why.
How is it started?

Either rpc.bind or rpc.portmap and then rpc.statd need to be started
manually (in that order) before you may mount the NFS partition without
'-onolock'. The lockd daemon itself will be started by the kernel
whenever there is a need for it.

Please check your 'init' boot scripts to find out why they are not being
started as expected.

-- 
Trond Myklebust
Linux NFS client maintainer

NetApp
Trond.Myklebust@netapp.com
www.netapp.com