topic Re: MC high priority command management in Layerscape

MC high priority command management

dmitri_lechtchinski — Sun, 28 Nov 2021 06:17:59 GMT

Hello,

According to DPAA2UM, the MC management command portal command includes "P" field for sending commands with high priority.

However from my understanding, if a specific client sets this to '1' for its commands, it can starve other clients which are trying to send commands with low priority.

Is it possible to disable or to manage usage of this field on system-wide level?

Thank you,
Dmitri

Re: MC high priority command management

andrei_skok — Mon, 29 Nov 2021 11:01:41 GMT

Could you please clarify which command / register / field / chapter /DPAA2UM document you are talking about?

Re: MC high priority command management

dmitri_lechtchinski — Mon, 29 Nov 2021 11:24:17 GMT

Document: DPAA2UM Rev 42

Chapter: 5.5 - Management command portal definition

Field: P - Priority (bit 15)

Re: MC high priority command management

andrei_skok — Mon, 29 Nov 2021 17:04:12 GMT

Dear Dmitri

This "P" bit field is used to specify High priority commands.So this option bit should be used with responsibility (in order to not starve other low priority commands...) This is a feature and it is the "user responsibility" to manage high priority commands vs low priority ones (according to their needs).For example if one specific command is very important and time critical, then that command can be marked as "High Priority" with this option bit.So that means if this option is not used, then it is equivalent as it would be disabled...It is user responsibility to manage High priority commands.Why would you like to disable it? Or how exactly would you like to be used at system-wide level? Can you provide more details about the usecase in order to understand this request?

Re: MC high priority command management

dmitri_lechtchinski — Mon, 29 Nov 2021 17:24:53 GMT

Assume we have a hypervisor OS with both safety-critical and non safety-critical partitions running on different cores, and each one uses its own MC portal.

So in this scenario, if a non safety-critical partition "decides" to overflow its portal with high priority commands, this could potentially starve a safety critical partition's functionality.

We are looking for a solution for preventing this from happening, assuming we have little or no control over the non safety-critical partition's behavior.

Thank you,
Dmitri