• State Resolved
  • Locked Locked
  • Replies 2 replies
  • Subscribers 99 subscribers
  • Views 438 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AM5716: ARM Core Stall

Kuerbis
Expert 4145 points
Part Number: AM5716
Other Parts Discussed in Thread: AM5K2E02,

Hi Team.

 

 

There is E2E post about ARM Core Stall which is 798870: A memory read can stall indefinitely in the L2 cache as below.

AM5K2E02: ARM Core Stall

https://e2e.ti.com/support/processors/f/791/t/704294

 

Is AM5716 affected by this eratta ID 798870?

If yes, what is a workaround for it?

 

AM5716 is Cortex-A15 MPU Core r2p2. But it also show the following configurations affected.

So I would like to confirm if AM5716 is affected.

 

Configurations affected:

This erratum does not affect the processor if the revision and variant reported by the MIDR is r2p0, r2p1, r2p2, r2p3 or r2p4 and REVIDR[2] is set to 1.

To be affected by this erratum, the “L2 arbitration register slice” configuration option must be included

 

Thanks and Best regards,

 Kuerbis

  • 0
    TI__Mastermind 23735 points

    Hello Kuerbis,

    The A15 in the AM5716 does not implement the "L2 arbitration register slice" config option.  As such it is not an exact match for the errata 798870 as described by ARM.  The other e2e thread seemed to indicate they saw some behavioral changes when setting this option but that likely was helping avoid some other traffic specific condition and not the documented one in the errata.  We have seen customers A15 systems hang due to various root causes. In some cases it was an unimplemented ARM errata, in some cases, it was SW mismanagement of MMU attributes causing bad access into IO regions and yet others things like an external PCIe master stopping its clock before completing an ongoing transaction which left the ARM waiting for completion.  In such hangs when possible is useful to attach to the DAP target via JTAG (not a hung cpu) and check the A15's memory-mapped DBGPCSR@ensd:0x54140084 (its PC sometimes reveals why its hung) and the L3 FLAGMUX registers which can help isolate which slave might be timed out. The TRMs interconnect chapter shows the flow chart on how to decode these registers. 

    Regards,

    Richard W.

  • 0 in reply to Richard Woodruff
    TI__Expert 4145 points

    Hi Richard,

     

    Thank you for your reply.

    I understand that AM5716 is not affected by this eratta ID 798870.

     

    Thanks and Best regards,

     Kuerbis