FATAL: Module mic not found.

[Mourad_B_2]

I am trying to install mpss in my server and I have this error for more than 2 days!!!

sudo service mpss start
WARNING: Deprecated config file /etc/modprobe.conf, all config files belong into /etc/modprobe.d/.
FATAL: Module mic not found.
Starting MPSS Stack: [FAILED]

sudo micctrl -r
WARNING: Deprecated config file /etc/modprobe.conf, all config files belong into /etc/modprobe.d/.
FATAL: Module mic not found.

Unable to load MIC driver. The 'micctrl' command, when
called with no MIC card name list requires the MIC driver

to be loaded in order to numerate the card list.

If you are performing remote configuration provide the MIC
The MIC card name list.

configuration : 2 x Xeon E5-2630 2.30GHz, RHEL Server 6.4, Linux 2.6.32-358.6.1.el6

need help. thanks

-mb 

[Frances_R_Intel]

The mic kernel module provides the driver for the coprocessor's PCIe connection. So if it doesn't load into the kernel, you are completely cut of from the coprocessor. I thing you will find that the configuration for the mic kernel module is located in /etc/modprobe.d. That is the default location. As your error message shows, your system is trying to use /etc/modprobe.conf to load the kernel modules. That is why the mic kernel module is not loading. Are you intentionally using modprobe.conf? If so, you will need to change this file to allow for the mic kernel module to load. If it is not intentional, can you change to using the modprobe.d directory?

[Mourad_B_2]

thanks Frances,

I fixed the "modprobe" issue. I have the fatal error though:

sudo micctrl --initdefaults
FATAL: Module mic not found.

Unable to load MIC driver. The 'micctrl' command, when
called with no MIC card name list requires the MIC driver
to be loaded in order to numerate the card list.

If you are performing remote configuration provide the MIC
The MIC card name list.

any other thoughts or steps for checking?

Thanks.

--mb

[Frances_R_Intel]

The mic kernel module, mic.ko, should have been installed in /lib/modules/{kernel_version_number}/extra. Is it there? 

Also, what Linux operating system are you using?

[Mourad_B_2]

thanks Frances

yes: /lib/modules/2.6.32-358.el6.x86_64/extra/mic.ko 

my OS: RHEL Server 6.4

Thanks.

[BelindaLiviero]

Hi, would you mind trying out this "flow" that we posted yesterday -- http://software.intel.com/sites/default/files/forum/393956/intelr-mpss-for-linux-troubleshoot-flow-chart.pdf 

and work through the steps (running lspci, etc), to help narrow down the issue?

[Mourad_B_2]

Hi Belinda,

the chart is taking me to check with Intel.

I think my Xeon Phi is ok. I am sure it's just a driver issue.

Frances, please do you have any idea to avoid this WARNING: Deprecated config file /etc/modprobe.conf, all config files belong into /etc/modprobe.d/. I dont want to use modprobe.conf so how to use /etc/modprobe.d/?

Thanks.

[Frances_R_Intel]

Apologies - I misunderstood you to say that this modprobe.conf error had been corrected. It was because we believed the configuration files were now correct that Belinda suggested going though that flow chart to separate out hardware vs software errors. When you went through the flow chart, was the path you followed: "Did the Intel® MPSS service start correctly?"->no->"Is the coprocessor enumerated by the host?"->yes->... If so, then let's get back to addressing the modprobe problem.

You are using a RHEL 6.4 operating system which no longer uses the /etc/modprobe.conf file by default. To get rid of the message "WARNING: Deprecated config file /etc/modprobe.conf, all config files belong into /etc/modprobe.d/.", you will need to remove the file /etc/modprobe.conf. If the file is empty (except for comments) you can just remove it. If it is actually being used to configure some kernel module, you will need to figure out whether you need to add a file to /etc/modprobe.d to handle that module or whether there is already a configuration file in /etc/modprobe.d to deal with it. This is a much more general Linux question that I can give advice about here.

As for the mic kernel module, there should be a file /etc/modprobe.d/mic.conf and /lib/modules/2.6.32-220.el6.x86_64/extra/mic.ko on your host system. If they are not there, then there was most likely an error in your installation and, painful as it my be, you should probably do that installation over.

If those files are there, do 'lsmod |grep mic' and look for a line that starts with 'mic' (not micro - that is something else). If you see a line that starts out mic, then the kernel module actually did load, and I may need you to repeat some of the tests from the flow chart Belinda pointed you to and send us the output.

If you do not find a line that starts out with 'mic', execute 'modprobe -C /etc/modprobe.d/mic.conf mic'. If this produces an error message, please let me know what the message was. If you do not get an error message, repeat the 'lsmod |grep mic' and look again for a line that starts out 'mic'. If you find that then you should be able to start the mpss service and use your coprocessor. You can use this as a work around until you remove the /etc/modprobe.conf file.

[Mourad_B_2]

Hi Frances,

Now I have this error:

No MIC cards found in the system
The MIC driver has been determined to be loaded. Use the
'lspci' utility to verify cards are installed. 

My lsmod is :

# lsmod |grep mic
mic 587320 0
microcode 112653 0

lspci | grep 2250
08:00.0 Co-processor: Intel Corporation Device 2250 (rev 11)

I do have my mic.ko in: 

cat /lib/modules/2.6.32-279.5.2.xxx/extra/mic.ko

service mpss start
Starting MPSS Stack: cannot open logfile '/var/log/mpssd'
[FAILED]

so please help me, where/ what's the problem?

Thanks

[Frances_R_Intel]

In your message you said you executed:

cat /lib/modules/2.6.32-279.5.2.xxx/extra/mic.ko

mic.ko is an executable file and not cat-able (at least not without producing garbage on your screen). So, this is a typo?

When you went through the chart at http://software.intel.com/en-us/forums/topic/393956, dmesg didn't show any errors saying "(can't reserve [mem 0x00000000-0x1ffffffff 64bit pref])" or "AER: Multiple Uncorrected (Fatal) error received: id=8010" and 'lspci -vvv' didn't show the message "LnkSta: Speed 5GT/s, Width x16, TrErrTrain- SlotClk+ DLActive- BWMgmt- ABWMgmtN" ?

You ran 'micctrl --initdefaults' and 'micctrl --resetconfig' when you did your install? There is a link '/sys/class/mic/mic0' and the directory it points to is populated with files such as state, serialnumber and sku? (Should have been automatically created for you.) And the directory /var/log exist and is it writeable?