topic Hi in Software Archive

NFS as RootDevice

Hirotaka_N_ — Thu, 16 Oct 2014 15:05:08 GMT

Hi,

Since we tried to use nfs export dir as RootDevice of Xeon Phi card.
According to https://software.intel.com/sites/default/files/article/373934/system-administration-for-the-intel-xeon-phi-coprocessor.pdf, we made changes as below.

It seems booting, however /dev/* is not usable at mic0 so we can't login via SSH nor screen /dev/ttyMIC0

[host] # micctrl --rootdev=NFS -v -c -d -t [host_ip_addr]:/srv/nfs/mic0  mic0
[host] # ls -al /srv/nfs/mic0
total 32
drwxr-xr-x 18 root    root    4096 Oct 16 23:18 .
drwxr-xr-x  4 root    root      28 Oct 16 23:17 ..
drwxr-xr-x  2 root    root    4096 Oct 16 23:06 bin
drwxr-xr-x  2 root    root       6 Oct 16 23:06 boot
drwxr-xr-x  2 root    root    4096 Oct 16 23:06 dev
drwxr-xr-x 31 root    root    4096 Oct 16 23:18 etc
drwxr-sr-x  4 root    root      31 Oct 16 23:06 home
-rwxr-xr-x  1 root    root    3946 Oct 16 23:06 init
drwxr-xr-x  3 root    root      20 Oct 16 23:06 lib
drwxr-xr-x  5 root    root    4096 Sep 19 07:51 lib64
drwxr-xr-x 10 root    root      94 Oct 16 23:06 media
drwxr-xr-x  2 root    root      57 Oct 16 23:06 mnt
drwxr-xr-x  3 root    root      18 Oct 16 23:06 opt
drwxr-xr-x  2 root    root       6 Oct 16 23:06 proc
-rw-r--r--  1 polkitd polkitd   62 Oct 16 23:17 .profile
drwx------  2 root    root      21 Oct 16 23:17 root
drwxr-xr-x  2 root    root    4096 Oct 16 23:18 sbin
drwxr-xr-x  2 root    root       6 Oct 16 23:06 sys
lrwxrwxrwx  1 root    root       8 Oct 16 23:18 tmp -> /var/tmp
drwxr-xr-x 12 root    root     123 Sep 19 08:32 usr
drwxr-xr-x  8 root    root     142 Sep 19 07:51 var
[host] # cat /etc/exports
/srv/nfs/mic0	[mic0_ip_addr]/32(rw,sync,no_root_squash)
[host] # micctrl -s mic0
mic0: online (mode: linux image: /usr/share/mpss/boot/bzImage-knightscorner)
[host] # service mpss start
[host] # tail -n f /var/log/messages
Oct 16 23:17:55 ccx10 systemd: Starting Intel(R) MPSS control service...
Oct 16 23:17:56 ccx10 kernel: mic0: Transition from state ready to booting
Oct 16 23:17:56 ccx10 kernel: mic image: /usr/share/mpss/boot/bzImage-knightscorner
Oct 16 23:17:56 ccx10 kernel: MIC 0 Booting
Oct 16 23:18:01 ccx10 kernel: Waiting for MIC 0 boot 5
Oct 16 23:18:06 ccx10 kernel: Waiting for MIC 0 boot 10
Oct 16 23:18:11 ccx10 kernel: Waiting for MIC 0 boot 15
Oct 16 23:18:16 ccx10 kernel: Waiting for MIC 0 boot 20
Oct 16 23:18:21 ccx10 kernel: Waiting for MIC 0 boot 25
Oct 16 23:18:22 ccx10 kernel: MIC 0 Network link is up
Oct 16 23:18:22 ccx10 kernel: br0: port 1(mic0) entered forwarding state
Oct 16 23:18:22 ccx10 kernel: br0: port 1(mic0) entered forwarding state
Oct 16 23:18:22 ccx10 rpc.mountd[30856]: authenticated mount request from [mic0_ip_addr]:969 for /srv/nfs/mic0 (/srv/nfs/mic0)
[host] # tail -f /srv/nfs/mic0/var/log/messages 
Oct 16 23:53:58 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 16 23:54:08 ccx11 user.warn kernel: [ 2164.337626] Host state not PC0
Oct 16 23:54:08 ccx11 daemon.info init: Id "0" respawning too fast: disabled for 5 minutes
Oct 16 23:59:09 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 16 23:59:19 ccx11 user.warn kernel: [ 2475.812577] Host state not PC0
Oct 16 23:59:19 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 16 23:59:29 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 16 23:59:40 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 16 23:59:50 ccx11 user.warn kernel: [ 2506.029530] Host state not PC0
Oct 16 23:59:50 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 17 00:00:00 ccx11 auth.err getty: /dev/hvc0: No such file or directory
[host] # ping -c 5 [mic0_ip_addr]
PING [mic0_ip_addr] ([mic0_ip_addr]) 56(84) bytes of data.
64 bytes from [mic0_ip_addr]: icmp_seq=1 ttl=64 time=0.524 ms
64 bytes from [mic0_ip_addr]: icmp_seq=2 ttl=64 time=0.424 ms
64 bytes from [mic0_ip_addr]: icmp_seq=3 ttl=64 time=0.398 ms
64 bytes from [mic0_ip_addr]: icmp_seq=4 ttl=64 time=0.268 ms
64 bytes from [mic0_ip_addr]: icmp_seq=5 ttl=64 time=0.226 ms

--- [mic0_ip_addr] ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4000ms
rtt min/avg/max/mdev = 0.226/0.368/0.524/0.108 ms

Could someone know the correct way to setup NFS export as a root of mic card?

Thanks!

Can you tell me how you

Frances_R_Intel — Thu, 16 Oct 2014 21:59:00 GMT

Can you tell me how you captured the files that you put in /srv/nfs/mic0?

I know that the current (MPSS 3.x) recommendation for making a cpio file to use for the StaticRAM boot option is boot the coprocessor using the default RAM file system, make any changes you want, then on the host, execute:

ssh root@mic0 "cd / ; find . /dev -xdev ! -path "./etc/modprobe.d*" ! -path "./var/volatile/run*" | cpio -o -H newc | gzip -9" > /usr/share/mpss/boot/custom.cpio.gz

Notice that in addition to . (which is / in this case), you list /dev as a search directory (then exclude the runtime files and current kernel module loads). By default, I don't think find walks /dev. Could this be your problem?

If I were going to make an NFS file system for the coprocessor's root, I think I would boot the coprocessor with the default RAM root file system, log into the coprocessor, make a directory /mnt, mount /srv/nfs/mic0 under /mnt then:

find . /dev -xdev ! -path "./etc/modprobe.d*" ! -path "./var/volatile/run*" | cpio -pdumv /mnt

And be very careful not to forget the -xdev on the find, since I am copying to a mount point under /.

But we probably have cluster administrators out there who have better solutions.

Hi

Hirotaka_N_ — Mon, 20 Oct 2014 09:51:51 GMT

Thanks for the comment.
I captured the log from NFS-server.
<pre>tail -f /srv/nfs/mic0/var/log/messages</pre>

I followed
<pre>find . /dev -xdev ! -path "./etc/modprobe.d*" ! -path "./var/volatile/run*" | cpio -pdumv /mnt</pre>
However after applied this command and reboot, nothing has changed.

# tail -n 100 /srv/nfs/mic0/var/log/messages
Oct 20 18:45:05 ccx11 user.warn kernel: [   29.638606] Module blcr_imports loaded at 0xffffffffa0024000
Oct 20 18:45:05 ccx11 user.warn kernel: [   29.669338] Module blcr loaded at 0xffffffffa00a2000
Oct 20 18:45:05 ccx11 user.info kernel: [   29.690853] blcr: vmadump: (from bproc-"4.0.0pre8") Erik Hendriks <erik@hendriks.cx>
Oct 20 18:45:05 ccx11 user.info kernel: [   29.690887] blcr: vmadump: Modified for blcr 0.8.5 <http://ftg.lbl.gov/checkpoint>
Oct 20 18:45:05 ccx11 user.info kernel: [   29.690912] blcr: Berkeley Lab Checkpoint/Restart (BLCR) module version 0.8.5.
Oct 20 18:45:05 ccx11 user.info kernel: [   29.690933] blcr:   Parameter cr_io_max = 0x4000000
Oct 20 18:45:05 ccx11 user.info kernel: [   29.690948] blcr:   Supports kernel interface version 0.10.3.
Oct 20 18:45:05 ccx11 user.info kernel: [   29.690967] blcr:   Supports context file format versions 8 though 9.
Oct 20 18:45:05 ccx11 user.info kernel: [   29.690986] blcr: http://ftg.lbl.gov/checkpoint
Oct 20 18:45:05 ccx11 user.warn kernel: [   29.805207] MPSSBOOT Boot acknowledged
Oct 20 18:45:05 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:45:05 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:45:16 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:45:16 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:45:26 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:45:26 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:45:37 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:45:37 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:45:47 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:45:47 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:45:57 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:45:58 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:46:08 ccx11 user.warn kernel: [   92.697559] Host state not PC0
Oct 20 18:46:08 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:46:08 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:46:18 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:46:19 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:46:29 ccx11 user.warn kernel: [  113.647744] Host state not PC0
Oct 20 18:46:29 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:46:29 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:46:39 ccx11 daemon.info init: open(/dev/tty0): No such file or directory
Oct 20 18:46:40 ccx11 auth.err getty: /dev/hvc0: No such file or directory
Oct 20 18:46:50 ccx11 user.warn kernel: [  134.585774] Host state not PC0
Oct 20 18:46:50 ccx11 daemon.info init: Id "0" respawning too fast: disabled for 5 minutes
Oct 20 18:46:50 ccx11 daemon.info init: no more processes left in this runlevel
Oct 20 18:47:32 ccx11 user.notice shutdown[4936]: shutting down for system halt
Oct 20 18:47:33 ccx11 daemon.info init: Switching to runlevel: 0
Oct 20 18:47:33 ccx11 daemon.info automount[4885]: autofs stopped
Oct 20 18:47:33 ccx11 syslog.info syslogd exiting

It would be great if anyone have succeeded to let the Phi card to have NFS as a root and tell us how to do this.

Thanks in advance.

Can you post the content of

Erich_C_Intel — Thu, 23 Oct 2014 16:52:07 GMT

Can you post the content of /var/mpss/mic0/etc/fstab after the configuration of rootdev with micctrl ?

Hi,

Hirotaka_N_ — Thu, 23 Oct 2014 16:59:47 GMT

Hi,
Thanks for the comment.

/var/mpss/mic0/etc/fstab doesn't exist. I found /srv/nfs/mic0/etc/fstab but not /var/mpss/.. Here's the content of /srv/nfs/mic0/etc/fstab

# cat /srv/nfs/mic0/etc/fstab 
# stock fstab - you probably want to override this with a machine specific one

rootfs               /                    auto       defaults              1  1
proc                 /proc                proc       defaults              0  0
devpts               /dev/pts             devpts     mode=0620,gid=5       0  0
usbdevfs             /proc/bus/usb        usbdevfs   noauto                0  0
tmpfs                /var/volatile        tmpfs      defaults              0  0
tmpfs                /media/ram           tmpfs      defaults              0  0

# uncomment this if your device has a SD/MMC/Transflash slot
#/dev/mmcblk0p1       /media/card          auto       defaults,sync,noauto  0  0

Thanks!

It worked if I create /var

Hirotaka_N_ — Thu, 23 Oct 2014 17:18:52 GMT

It worked if I create /var/mpss/mic0/etc/fstab and set nfs mount point as a /.
Here's /var/mpss/mic0/etc/fstab

# cat /var/mpss/mic0/etc/fstab 
#rootfs		/		auto		defaults		1  1
[host_ip_addr]:/srv/nfs/mic0	/	nfs	rw,hard,intr,tcp,nfsvers=3 1 1
proc		/proc		proc		defaults		0  0
devpts		/dev/pts	devpts		mode=0620,gid=5		0  0

Thanks for the all comments!!!