Re: UBSAN error in dmesg after OneAPI installation (Ubuntu 20.04, non-Intel CPU)

TiborGY · ‎10-27-2022

Dear all,

I have recently installed the OneAPI BaseKit (l_BaseKit_p_2022.3.0.8767) on a server running Ubuntu 20.04 (with HWE kernel, version: 5.15.0-46-generic #49~20.04.1-Ubuntu) with an AMD EPYC 7443P CPU. The installer was left at the default settings, and all components were installed.

After the installation, I have found these messages in the dmesg:
[4819887.770431] UBSAN: array-index-out-of-bounds in /opt/intel/oneapi/vtune/2022.4.0/sepdk/src/socperf/src/pmu_list.c:392:60
[4819887.770710] index -1 is out of range for type 'PMU_INFO_NODE [10]'

Please see attachment for complete dmesg log of the event.

To me this looks like a bug in the Intel Vtune kernel module, it should not be running into undefined behavior even if it is installed/started on a non-Intel CPU. It is perfectly reasonable that it only supports Intel CPUs, but it should still handle the situation gracefully and not run into UB.

This is especially a problem because the installer installs the kernel module by default even on non-Intel CPUs.

Soumya_M_Intel · ‎10-28-2022

Hi TiborGY,

Thanks for posting your concern on the forum. We understand your issue regarding non-Intel® CPUs. However, please know, the issue is occurring since it is not a supported system for the product to be installed on.

Kindly note that our engineering team has tested and documented the Hardware Requirements for the Intel® oneAPI Base Toolkit. We recommend users to check the system requirements via the below link.

Link: https://www.intel.com/content/www/us/en/developer/articles/system-requirements/intel-oneapi-base-toolkit-system-requirements.html

The AMD EPYC 7443P CPU is not among them. Due to that it is out of our scope of support.

However, please feel free to let us know if the issue persists on a supported system. Looking forward to your update.

Regards,

Soumya

TiborGY · ‎10-30-2022

Dear Soumya,

I have found that a similar error message is also issued on a computer with a Xeon W-2295 CPU, running a slightly older version of the Toolkit (2022.0.0).

Please find attached the relevant portion of the kernel log.

Best,

Tibor

Soumya_M_Intel · ‎11-01-2022

Hi Tibor,

Thanks for getting back to us with the update on the thread. We are checking on the issue with our resources. Please wait for 2-3 business days. We will get back to you as soon as possible.

Regards,

Soumya

Soumya_M_Intel · ‎11-07-2022

Hi Tibor,

Thanks for your patience while we were checking on the concern with our higher team. They have suggested to try installing the Intel® oneAPI Base Toolkit 2022.3 on the supported system where the old version (2022.0.0) of the toolkit is installed.

To make sure it works for you, we have replicated it on a similar environment and there were no issues. However, kindly let us know if you have any concerns.

Regards,

Soumya

Soumya_M_Intel · ‎11-09-2022

Hi Tibor,

Please let us know if there are any further concerns.

Regards,

Soumya

TiborGY · ‎11-10-2022

Dear Soumya,

I will try updating the OneAPI installation on the computer in question next week.

I will post the results here.

Best,

Tibor

Soumya_M_Intel · ‎11-10-2022

Hi Tibor,

Thanks for getting back to us on the thread. We will await your update.

Regards,

Soumya

Soumya_M_Intel · ‎11-21-2022

Hi Tibor,

Please feel free to let us know if you had come across any queries or concerns after trying out the installation last week.

Regards,

Soumya

Soumya_M_Intel · ‎11-23-2022

Hi Tibor,

We understand that you might be looking into the issue. However, this is to update you that we have not heard from you on our previous responses. Please reply within the next 24 business hours to keep this thread active.

Regards,

Soumya

TiborGY · ‎11-23-2022

Dear Soumya,

Sorry for the delay. I have installed the latest toolkit, and the UBSAN warning is still issued in dmesg when the VTune kernel module loads. Please see the attached log.

Best,

Tibor

Soumya_M_Intel · ‎11-28-2022

Hi Tibor,

Thanks for your patience while we were checking with our higher team. As per their update, we are routing your query to the 'Analyzers' forum as they will be able to assist you further regarding the concern.

Our internal team will get back to you there promptly.

Hope you have a nice day ahead!

Regards,

Soumya

JyothisV_Intel · ‎12-01-2022

Hi,

Good day to you.

Sorry for the delay.

Unfortunately, we were unable to replicate this from our side. Can you let us know in which of the below cases are you observing the dmesg logs being generated:

Installation/Repair/Update of Intel oneAPI Base Toolkit
Building/Loading the Intel VTune Sampling Drivers using binaries like build_driver or insmod-sep, etc
Opening Intel VTune Profiler
Performing a specific type of analysis using Intel VTune Profiler
Any other instances (kindly mention the steps to replicate it)

It would be extremely helpful if you can try clearing the dmesg logs by running the below command and let us know when exactly you are getting these UBSAN messages to investigate this further

sudo dmesg -c

N.B: Alternatively, you can also perform the above tasks while watching the dmesg logs using the below command in another terminal windows to get the same:

watch dmesg

Thanks and Regards,

Jyothis V James

TiborGY · ‎12-08-2022

Dear Jyothis,

The error message appears in dmesg when the kernel module is loaded. I will try and get back to you with more exact steps to reproduce soon, but one thing I want to add is that I am lead to believe that this error can only be reproduced on Intel CPUs that do not have a GPU. Eg. I would not expect this to be reproducable on a 12700k or a typical laptop.

Best,

Tibor

JyothisV_Intel · ‎12-07-2022

Hi,

Good day to you.

We have not heard back from you. Are you still getting those messages? If yes, do get back to us with the requested information so that we can assist you better.

Thanks and Regards,

Jyothis V James

JyothisV_Intel · ‎12-26-2022

Hi,

Good day to you.

We have not received any response from you. Intel will no longer monitor this thread. If you need further assistance, please post a new question.

Thanks and Regards,

Jyothis V James