Is there no auto-recovery for slicehang issue on c4xxx (qat-1.8) device?

[DAS__SANJIB1]

With jumbo packets (sent to QAT for encryption) we see these logs.

[ 822.153243] c4xxx 0000:f4:00.0: SliceHang detected in Cipher2 unit of accel1
[ 822.239335] c4xxx 0000:f4:00.0: SliceHang detected in Cipher2 unit of accel7

Following this when we reload our network device, additional failure logs appear.

[ 2195.124086] c4xxxvf 0000:f4:00.1: Failed to send Shutdown event to PF
[ 2198.009087] c4xxxvf 0000:f4:00.2: Failed to send Shutdown event to PF

[ 2544.568097] c4xxx 0000:f4:00.0: Failed to send restarting msg to VF0
[ 2547.442100] c4xxx 0000:f4:00.0: Failed to send restarting msg to VF1
[ 2911.247110] c4xxx 0000:f4:00.0: Some VFs are still running

Qat-1.8 code flow shows, slicehang is enabled for c4xxx only. For other variants its kept disabled.

Example:

adf_c62x_hw_data.c

adf_enable_ints

ALLOW_SLICE_HANG_INTERRUPT => not enabled for c6xx

Why is it explicitly done for c4xxx & what is the impact if we keep it disabled here as well?

Thanks,

Sanjib

[BrusC_Intel]

Hello, DAS__SANJIB1.

Thank you for posting on the Intel Community Support forums.

We received your ticket regarding this particular QAT inquiry, and I will be assisting you with this.

Please allow me to review this and I will contact you back as soon as possible.

Regards,

Bruce C.

Intel Customer Support Technician

[BrusC_Intel]

Hello, DAS__SANJIB1.

Thank you for waiting.

Please check this behavior using the Intel® QuickAssist Technology Driver for Linux* – HW Version 1.X. This driver supports QAT Hardware Version 1.8 excluding the inline features. If you are not using the inline features of QAT Hardware Version 1.8, then you can use this QAT driver.

The Release Notes are available in the following link in case you want to check more details: Intel® QuickAssist Technology Software for Linux* - Release Notes - Customer Enabling Release.

I will follow up on July 3rd in case you need additional time.

Regards,

Bruce C.

Intel Customer Support Technician

[DAS__SANJIB1]

Considering our software architecture, simply replacing the qat package is not feasible. Its tightly integrated with our software infra, requiring a considerable amount of time & effort.

Moreover, I exactly dont know whats the difference between HW Version 1.X. & my current version. If you could kindly point out specific changes/differences in that "slice hang" context, it would be really helpful.

Intel shared this package with us.

PACKAGE_TYPE=QAT18
PACKAGE_OS=L
PACKAGE_VERSION_MAJOR_NUMBER=1
PACKAGE_VERSION_MINOR_NUMBER=6
PACKAGE_VERSION_PATCH_NUMBER=0
PACKAGE_VERSION_BUILD_NUMBER=00024

Thanks,

Sanjib

[BrusC_Intel]

Hello, DAS__SANJIB1.

Thank you for the details.

I'm currently working on this and I will get back to you as soon as possible.

Regards,

Bruce C.

Intel Customer Support Technician

[BrusC_Intel]

Hello, DAS__SANJIB1.

Thank you for waiting.

While we were consulting with the Engineering Team about this issue, we noticed they currently have an Intel Premier Support (IPS) ticket from you regarding the same topic. Since the IPS ticket is being worked directly by the Engineering Team, we would like to confirm if we can close this community thread to prevent duplicate information.

Additionally, if you have access to IPS, submitting your inquiries via that channel would be preferable over submitting them via the public forum, as IPS tickets go directly to the QAT Engineering Team.

If you have any comments, please let me know, but if that is not the case I will close this thread on July 11th.

Regards,

Bruce C.

Intel Customer Support Technician

[DAS__SANJIB1]

Yes I do have an IPS ticket open. You can close this thread/post.

[BrusC_Intel]

Hello, DAS__SANJIB1.

Thank you for the confirmation.

The thread will be closed right now, and if you need any type of assistance from Intel in the future, you can always contact us back.

Regards,

Bruce C.

Intel Customer Support Technician