Where to report kernel problems?

atelszewski · 09-22-2017, 05:33 PM

Hi,

How do I proceed about sorting out kernel problems?
I'm asking, because I have hardware related problems with past 4.4.x kernels and Slackware uses unmodified kernel, so this makes me think that this forum isn't the best place to discuss.
I guess it would be best to report to the upstream.
But I'm extremely scared of the amount of messages hitting kernel mailing lists ;-) (Or am I wrong?)

Thanks in advance!

--
Best regards,
Andrzej Telszewski

Emerson · 09-22-2017, 06:11 PM

Maybe discussing these hardware problems first would be the way to go. 4.4 has been out for long time and considering how widespread Linux is it is hard to believe you are the only one hit by a kernel bug. Perhaps it is something else. The way Slackware kernel is configured, maybe a hardware issue that wasn't exposed before?

atelszewski · 09-22-2017, 06:33 PM

Hi,

The bug is kinda known, but I would like to have an authoritative answer if it is going to be fixed or not.
This is something that Pat cannot give me.

Anyways, I'll might post it here soon enough.

But I think that my first question still holds.

And I have more bugs to talk about ;-)
They are related to TPM2 and DMAR/IOMMU on my NUC machine.

--
Best regards,
Andrzej Telszewski

atelszewski · 09-22-2017, 07:46 PM

Hi,

OK, for the brave out there ;-)

My machine is NUC5i3RYH (NUC5i3RYB).
I have the latest BIOS.
I'm on Slackware64-14.2.

When booting anything above 4.4.x (I think I first saw it on 4.6.x) I get:

Code:

$ dmesg | grep -i tpm
[    0.000000] ACPI: TPM2 0x00000000B43690F0 000034 (v03 INTEL  NUC5i3RY 0000016E AMI  00000000)
[    4.578877] tpm_crb MSFT0101:00: can't request region for resource [mem 0xb4fff000-0xb4fff02f]
[    4.578951] tpm_crb: probe of MSFT0101:00 failed with error -16

and the TPM /dev entry is not present.

This message actually comes from kernel version 4.13.3, but I'm certain it's the same for 4.9.x from -current.

I know that the TPM handling code has changed in-between the releases.
I don't know if the upstream will fix it.

Cheers :-)

--
Best regards,
Andrzej Telszewski

Emerson · 09-22-2017, 07:51 PM

https://www.kernel.org/doc/html/v4.1...ting-bugs.html

You do not need to subscribe to any mailing list.

atelszewski · 09-22-2017, 08:24 PM

Hi,

Quote:

Originally Posted by Emerson

https://www.kernel.org/doc/html/v4.1...ting-bugs.html

You do not need to subscribe to any mailing list.

Thanks.

I'm worried about bold:

Quote:

Once you know the subsystem that is causing the issue, you should send a bug report. Some maintainers prefer bugs to be reported via bugzilla (https://bugzilla.kernel.org), while others prefer that bugs be reported via the subsystem mailing list.

I'll try anyways.

--
Best regards,
Andrzej Telszewski

Emerson · 09-22-2017, 08:29 PM

You are welcome. BTW, error 16 is - resource busy. Have you tried to rmmod it and reload after the box is up and running?

atelszewski · 09-22-2017, 08:37 PM

Hi,

Quote:

Originally Posted by Emerson

BTW, error 16 is - resource busy. Have you tried to rmmod it and reload after the box is up and running?

Yes, it's not that simple ;-)
I think the important part is about requesting memory region.

--
Best regards,
Andrzej Telszewski

kjhambrick · 09-23-2017, 07:55 AM

atelszewski --

We run a few NUCs out in the world and there are more going out all the time so this is kinda-sorta interesting for me.

I googled your error and found this thread on Intel Communities where rguevara said in post #6 that the user's specific NUC did not support TPM.

I looked up the specs of your NUC5i3RYH and down in the Advanced Technologies Section, it says TPM = NO ...

Does this help ?

-- kjh

atelszewski · 09-23-2017, 01:21 PM

Hi,

@kjhambrick Something stinks here.

I have a BIOS setting named "Intel Platform Trust Technology". They call it PTT.
If I disable it, I have no /dev/tpm0, if I enable it, I have /dev/tpm0.

But more importantly, I'm using TPM based hardware random number generator and it works.

Anyways:

Code:

$ dmesg | grep -i tpm
[    0.000000] ACPI: TPM2 0x00000000B43690F0 000034 (v03 INTEL  NUC5i3RY 0000016E AMI  00000000)

If I change kernel to say 4.9.x, the /dev/tpm0 device disappears.

Something stinks, and I know it's close to me ;-)

EDIT:
Per TPM and Intel PTT Overview:

Quote:

PTT is a TPM2.0 implementation implemented as a FW application in ME

so I should have TPM.

--
Best regards,
Andrzej Telszewski

kjhambrick · 09-23-2017, 03:45 PM

Thanks atelszewski.

Interesting and informative slide show !

And the Specs for your NUC does say PTT = YES

This is what the Help Bubble for the PTT Line on the Spec Page says:

Quote:

Intel® Platform Trust Technology (Intel® PTT)

Intel® Platform Trust Technology (Intel® PTT) is a platform functionality for credential storage and key management used by Windows 8* and Windows® 10. Intel® PTT supports BitLocker* for hard drive encryption and supports all Microsoft requirements for firmware Trusted Platform Module (fTPM) 2.0.

Later in the thread that I linked from the Intel Community ( posts #9 and #10 ), John_Ewalt finally concludes that it was a BIOS bug ...

And there is a kernel patch on post #20 to work around the bug.

Finally on post #23, the "correct answer" is that BIOS Version 45 fixed the issue for John_Ewalt's NUC6i7KY ...

As for reporting the bug in your NUC5i3RYH and who to report it to ( Intel or Kernel Devs or ??? )... I am not sure either.

I've got an 'SPARE' Intel NUC6CAYH on my bench.

It has a fairly recent BIOS:

Code:

# dmesg |grep DMI:
[    0.000000] DMI:                  /NUC6CAYB, BIOS AYAPLCEL.86A.0040.2017.0619.1722 06/19/2017

It is running CentOS 7 ( 3.10.0 Kernel ), and it's got a different TPM Error

Code:

# dmesg |grep -i tpm
[    0.000000] ACPI: TPM2 00000000794ae670 00034 (v03  INTEL NUC6CAYB 00000001 AMI  00000000)
[    1.007574] IMA: No TPM chip found, activating TPM-bypass!
[    3.216331] tpm_crb MSFT0101:00: A TPM error (323) occurred continue selftest
[    3.216434] tpm_crb: probe of MSFT0101:00 failed with error 323

I didn't seem to miss TPM but your post reminded me about it.

Thanks again for the info about TPM and PTT !

-- kjh

atelszewski · 12-18-2017, 09:41 AM

Hi,

At some point in the distant future I plan to remove Intel ME.

Since the TPM seems to be implemented in ME, if ME disappears, then TPM disappears too.
And so does my problem ;-)

--
Best regards,
Andrzej Telszewski

ivanf · 02-10-2022, 09:09 AM

Hallo folks,

I am sorry to resurrect this old thread, but I am seeing the exact same error message after having upgraded from Slackware 14.2 to Slackware 15.0,

Feb 10 08:46:47 mybox kernel: tpm_crb MSFT0101:00: can't request region for resource [mem 0xa2fff000-0xa2fff02f]
Feb 10 08:46:47 mybox kernel: tpm_crb: probe of MSFT0101:00 failed with error -16

and the node /dev/tpm0 has disappeared. In 14.2 the fTPM / PTT worked throughout several stock (meaning coming from Slackware repos, not patched by me) kernels. I am using a NUC 5i3RYH with the latest available BIOS:

Hardware name: /NUC5i3RYB, BIOS RYBDWi35.86A.0385.2020.0519.1558 05/19/2020

Since the bug is supposed to have been fixed in the kernel some years ago, can it be that I am seeing a kernel regression?

Best
/ivanf

ivanf · 02-11-2022, 01:25 PM

Found a workaround under https://github.com/tpm2-software/tpm2-tools/issues/2148. The kernel needs a boot parameter:

memmap=0x1000%0xa2fff000+2

Best
/ivanf