SPRZ574 December   2023 AM62P , AM62P-Q1

 

  1.   1
  2. 1Usage Notes and Advisories Matrices
    1. 1.1 Devices Supported
  3. 2Silicon Usage Notes and Advisories
    1. 2.1 Silicon Usage Notes
      1.      i2351
      2.      i2372
    2. 2.2 Silicon Advisories
      1.      i2049
      2.      i2062
      3.      i2097
      4.      i2137
      5.      i2189
      6.      i2190
      7.      i2196
      8.      i2208
      9.      i2249
      10.      i2253
      11.      i2278
      12.      i2279
      13.      i2310
      14.      i2311
      15.      i2312
      16.      i2366
      17.      i2383
      18.      i2401
      19.      i2407
      20.      i2409
      21.      i2410
  4.   Trademarks

i2049


ECC_AGGR: Potential IP Clockstop/Reset Sequence Hang due to Pending ECC Aggregator Interrupts

Details:

The ECC Aggregator module is used to aggregate safety error occurrences (which are rare) and generate interrupts to notify software. The ECC Aggregator provides software control over the enabling/disabling and clearing of safety errors interrupts.

When software is performing a clockstop/reset sequence on an IP, the sequence can potentially not complete because the IP's associated ECC Aggregator instance is not idle. The ECC Aggregator idle status is dependent upon any pending safety error interrupts either enabled or disabled, which have not been cleared by software. As a result, the IP's clockstop/reset sequence may never complete (hang) if there are any pending safety errors interrupts that remain uncleared.

The affected ECC_AGGRs can be determined by the value listed in the Technical Reference Manual (TRM) for their REV register at Register Offset 0h. The REV register encodes the ECC_AGGR version in its fields as follows:

v[REVMAJ].[REVMIN].[REVRTL]

ECC_AGGR versions before v2.1.1 are affected. ECC_AGGR versions v2.1.1 and later are not affected.

Affected Example:

REVMAJ = 2

REVMIN = 1

REVRTL = 0

The above values decode to ECC_AGGR Version v2.1.0, which is Affected.

Not Affected Example:

REVMAJ = 2

REVMIN = 1

REVRTL = 1

The above values decode ECC_AGGR Version v2.1.1, which is Not Affected.

Workaround(s):

General Note:

Clockstopping the ECC Aggregator is not supported in functional safety use-cases.

Software should use the following workaround for non-functional safety use-cases:

  1. Enable all ECC Aggregator interrupts for the IP
  2. Service and clear all Pending interrupts
  3. Step 3:
    1. Disable all interrupt sources to the ECC Aggregator, followed by performing Clockstop/reset sequence.
    2. Perform Clockstop/reset sequence, while continuing to service/clear pending interrupts.

Due to interrupts being external stimuli, software has two options for step 3:

  1. Disable all interrupt sources (EDC CTRL checkers) that can generate pending ECC_AGGR interrupts prior to performing the clockstop/reset sequence
  2. Continue to service/clear pending interrupts that occur while performing the clkstop/reset sequence. The sequence would proceed when all interrupts are cleared.

Software in general may need to detect pending interrupts that continuously fire during this entire sequence (ex. in the case of a stuck-at fault scenario), and disable their associated EDC CTRL safety checkers to allow the clockstop/reset sequence to progress towards completion.