SPRZ575A March   2024  – April 2025 AM67 , AM67A , TDA4AEN-Q1 , TDA4VEN-Q1

 

  1.   1
  2. 1Modules Affected
  3. 2Nomenclature, Package Symbolization, and Revision Identification
    1. 2.1 Device and Development-Support Tool Nomenclature
    2. 2.2 Devices Supported
    3. 2.3 Package Symbolization and Revision Identification
  4. 3Silicon Revision 1.0 Usage Notes and Advisories
    1. 3.1 Silicon Revision 1.0 Usage Notes
      1.      i2134
    2. 3.2 Silicon Revision 1.0 Advisories
      1.      i2049
      2.      i2062
      3.      i2097
      4.      i2120
      5.      i2137
      6.      i2160
      7.      i2189
      8.      i2190
      9.      i2196
      10.      i2199
      11.      i2208
      12.      i2242
      13.      i2243
      14.      i2249
      15.      i2253
      16.      i2278
      17.      i2279
      18.      i2310
      19.      i2311
      20.      i2312
      21.      i2326
      22.      i2330
      23.      i2351
      24.      i2362
      25.      i2366
      26.      i2372
      27.      i2383
      28.      i2399
      29.      i2401
      30.      i2407
      31.      i2409
      32.      i2410
      33.      i2419
      34.      i2424
      35.      i2431
      36.      i2436
      37.      i2457
      38.      i2478
  5.   Trademarks
  6. 4Revision History

i2399

C7x: CPU NLC Module Not Clearing State on Interrupt

Details:

Data corruption will occur when:

  1. An application is running that involves task switching. In this case there are at least 2 tasks that may use NLC.
  2. There is a NLCINIT issued that would be followed by a TICK when an interrupt comes in for Task A. This action ends up setting some internal state in the NLC module that says we need to reload the ILCNT_INIT value to ILCNT on the next TICK since the forwarded case it computed was flushed. This state is not being properly cleared when the interrupt is taken.
  3. The ISR performs a task switch to Task B, which is also running NLC code. The NLC code being returned to needs to be in-progress and have a different ILCNT_INIT value than the NLC loop in the original task.
  4. After returning from the ISR, the next TICK will end up setting ILCNT to the wrong value (ILCNT_INIT - 2) due to the corrupted state.

At this point the ILCNT is corrupted and the NLC loop will execute the wrong number of iterations, leading to data corruption.

Workaround(s):

Issue a NLCINIT (parameters don't matter and there's no need for TICK's/BNL afterwards) in ISR's as part of the context saving. There is no performance impact due to the work-around.