Xeon Phi - Error in installation

[Julian_M_]

When installing the mpss-3.5.1 on CentOS 7 (kernel 3.10.0-229.el7.x86_64), I got an error when issue " micctrl --initdefaults" as below:

[Error] micrasrelmond: [Parse] START_WITH_SECURITY=false invalid config parameter

[Warning] micrasrelmond: Generating compatibility network config file /opt/intel/mic/filesystem/micrasrelmond/etc/sysconfig/network/ifcfg-mic0 for IDB.

[Warning]       This may be problematic at best and will be removed in a future release, Check with the IDB release.
micctrl(segv_handler+0x18) [0x407818]
/lib64/libc.so.6(+0x35650) [0x7f4aa9abd650]
/lib64/libmpssconfig.so.0.0.1(_add_miclist_not_present+0xe0) [0x7f4aaa05ce60]
/lib64/libmpssconfig.so.0.0.1(mpss_get_miclist+0x4d) [0x7f4aaa05d18d]
micctrl(gen_etchosts+0x15) [0x410e95]
micctrl(parse_config_args+0x4cc) [0x40af6c]
micctrl(main+0x246) [0x407a96]
/lib64/libc.so.6(__libc_start_main+0xf5) [0x7f4aa9aa9af5]
micctrl() [0x4076f9]

Same error in dealing with CentOS 6.5

Motherboard: Supermicro
Model: H8QG6
Xeon Phi 31s1p (x2)

Thanks, Julian

Sorry for my bad english
 

[JJK]

what does 'micinfo' report? is the card present at all and is the 'mic' driver loaded properly? what does 'lsmod | grep mic' return?

[Frances_R_Intel]

The two warning messages are "normal". They are there to let you know that the 'micctrl --initdefaults" command needed to make a copy of the file "/etc/sysconfig/network/ifcfg-mic0" in /opt/intel/mic/filesystem/micrasrelmond to support idb. In the future, that program will go away and the duplicate file will go away. That it did not also create one for mic1 is confusing, however.

The error message is an artifact of a change that was made in MPSS 3.4 to support micras security in RHEL 7.0 - although, you can see it in earlier RHEL - or CentOS - versions once things have gone wrong in 7.0. The parameter, START_WITH_SECURITY=false, is actually a valid parameter. As I understand it, you shouldn't see this error if the mic kernel module loaded correctly.

Step one would be, as JJK suggests, to make sure the kernel module loads correctly by checking 'lsmod | grep mic' for an entry called mic (not microcode). If it isn't loaded, try to load it by hand (see the install instructions in the readme.txt for directions). If it loads, try 'micctrl --initdefaults' again to see if it will create the mic*.conf files for you. If mic.ko won't load or if it does load but 'micctrl --initdefaults' still gets an error, try going through https://software.intel.com/en-us/forums/topic/393956 and if that doesn't help, try running micdebug.sh, look over the output to be sure there is no information you do not feel comfortable passing on to us and then send the output by attaching it as a private message to this post.

Let us know what happens.

[Julian_M_]

Hi @JJK and @Frances!
Where is the report micinfo?

[root@matrix mpss-3.5.1]# lsmod | grep mic
mic                   682463  0

mics appears in red

My bash output:

[root@matrix mpss-3.5.1]# lsmod | grep mic
mic                   682463  0
[root@matrix mpss-3.5.1]# micctrl --initdefaults
  [Error] micrasrelmond: [Parse] START_WITH_SECURITY=false invalid config parameter
[Warning] micrasrelmond: Generating compatibility network config file /opt/intel/mic/filesystem/micrasrelmond/etc/sysconfig/network/ifcfg-mic0 for IDB.
[Warning]       This may be problematic at best and will be removed in a future release, Check with the IDB release.
micctrl(segv_handler+0x18) [0x407818]
/lib64/libc.so.6(+0x35650) [0x7f1e1ab65650]
/lib64/libmpssconfig.so.0.0.1(_add_miclist_not_present+0xe0) [0x7f1e1b104e60]
/lib64/libmpssconfig.so.0.0.1(mpss_get_miclist+0x4d) [0x7f1e1b10518d]
micctrl(gen_etchosts+0x15) [0x410e95]
micctrl(parse_config_args+0x4cc) [0x40af6c]
micctrl(main+0x246) [0x407a96]
/lib64/libc.so.6(__libc_start_main+0xf5) [0x7f1e1ab51af5]
micctrl() [0x4076f9]
[root@matrix mpss-3.5.1]# sudo service mpss start
Starting mpss (via systemctl):                             [  OK  ]

[root@matrix mpss-3.5.1]# sudo /usr/bin/micinfo
MicInfo Utility Log
Created Sun Jun 21 19:38:48 2015

        System Info
                HOST OS                 : Linux
                OS Version              : 3.10.0-229.el7.x86_64
                Driver Version          : 3.5.1-1
                MPSS Version            : 3.5.1

                Host Physical Memory    : 32758 MB
[root@matrix mpss-3.5.1]#

Attached the output of micdebug.sh

It is really strange that before installing the driver could see lspci co-processors, but after installing the driver and restarting the computer does not.

Thanks and sorry for my bad english, Julian

[JJK]

Ah, your server is based on AMD Opterons. I do not have any experience with those in combination with a Xeon Phi.

As you state, the fact that the Phi disappears after the driver has been loaded is worrisome - can you verify that it comes back when you unload the driver?

[Julian_M_]

lift the module and uninstall drivers .... The Xeon are not listed in the lspci

[JJK]

this is worrisome, but there are a few things you can do:

• Uninstall the driver (if you installed it again), power down the box and bring it back up again. Is the card now listed in 'lspci' output?
• Power down the box, check that the cards are inserted properly
• Remove one of the cards and try again - it could be a power supply issue. if it works with one card but not with two then you know it's the power supply

[Julian_M_]

I contact the manufacturer of the mother, the motherboard does not support the Phi:
 

Hi Julian,

Thank you for contacting Supermicro tech support.

4042G-TRF server is with AMD platform. Xeon Phi is not compatible with it.

Thank you.

DH

Technical Support

Supermicro