CC2538: Sometimes (very rare) a device “stops to communicate”

Hi Nalves,

We are looking into this, I will post any updates here.

Hi Nalves,

In Z3.0 Release notes says there was a fix on the Nv items for Nwk Key on CC2538, does symptom A happens upon flashing the device, if so, then Z3.0 fix may solve this.

The related change for this fix is in the osal_nv.c

static uint8 writeItem( uint8 pg, uint16 id, uint16 len, void *buf, uint8 flag )
{

 ...

      if ( chk == calcChkF( pg, datOff, len ) )
      {
        hdrData[0] = chk;
        flashWrite(OSAL_NV_PAGE_TO_PTR(pg) + hdrOff + OSAL_NV_HDR_CHK,
                   OSAL_NV_HDR_ITEM, (uint8 *)(hdrData));
        readHdr( pg, hdrOff, (uint8 *)(&hdr) );

        if ( chk == hdr.chk )
        {
          hotItemUpdate(pg, datOff, hdr.id);
          rtrn = TRUE;
        }
      }
    }

...

}

Hello Luis,

Glad to know that someone that supposed had the same problem. Did you has the same problem to have that conclusion? I test it but the problem happens, althout with less frequency. When the problem happens, the NV read result for ZCD_NV_NWK_ACTIVE_KEY_INFO (0x003A) was 0x00, 0xFF .... and i don't know why this value...

Best regards
Nalves

Hi Nalves,

Are you saying that even with the fix that Luis proposed, you're still seeing the same issue with the NWK key in NV?

Hi Jason,

The fix proposed by Luis, didn’t solve the problem of Symptom A, but I felt improvements because it became harder to reproduce the problem.

However, with this propose, I understood that z-stack doesn’t work well, when osal_nv_item_init() and osal_nv_write() returns NV_OPER_FAILED when OSAL_NV_CHECK_BUS_VOLTAGE is lower than the minimal value.

To fix it, i changed the condition, “if ( !OSAL_NV_CHECK_BUS_VOLTAGE ) return NV_OPER_FAILED;” for:

while ( !OSAL_NV_CHECK_BUS_VOLTAGE );

 //if ( !OSAL_NV_CHECK_BUS_VOLTAGE )

 //{

 //  return NV_OPER_FAILED;

 //}

in the first line of the functions: osal_nv_item_init(), osal_nv_write() and osal_nv_read(). Whith this exchange, any call to NV memory only starts, if the minimal voltage "OSAL_NV_CHECK_BUS_VOLTAGE" is checked.

With this change, I'm not able to reproduce any more problems described in the first post, however, I don’t understand how a bad start, with a wrong NV item initialized, only have impact in a few days after the start-up of device. I’m saying this, because at the test field, the devices that had problems, only show the malfunction after a few days of normal operation.

best regards,

nalves

Hi Jason,

Adicional to this information, I checked that in ZDSecMgrNwkKeyInit(), if some the inicialization of the following items fail (return NV_OPER_FAILED)

- osal_nv_item_init( ZCD_NV_NWKKEY, sizeof(nwkActiveKeyItems), (void *)&keyItems );
- osal_nv_item_init( ZCD_NV_NWK_ACTIVE_KEY_INFO, sizeof(nwkKey), &nwkKey);
- osal_nv_item_init( ZCD_NV_NWK_ALTERN_KEY_INFO, sizeof(nwkKey), &nwkKey );

that are the hot items(hotPg[], hotOff[]), they will be set with 0x00.

The NV memory can be currupted with hotPg[], hotOff[], iqual 0x00, and as this values never more are inicializaded and device never recover the comunication with network, without a restart!

Best regards
Nalves

Hi Nalves,

Can you clarify what outstanding issues you still have? Just want to make sure I fully understand before I answer again.

Hello Jason,

Ok, let me clarify you. We are developing a system based in Zigbee comunication between devices with HA profile. e are using the C2538 and Z-Stack 1.2.0. We have this solution working almost fine at 3 years. At a few week ago we detect that some devices apparentlly stop comunicate with other devices in network (Stop sending packets and stop responding to remote requests). In our laboratories we ry to reproduce the syntoms of issues reported and we can reproduced turn it on and turn it off constantlly at the same time sending packets requests, and after few secounds, the device stop communicating. internally the device Works fine, responde at key pressed turn on and turn off local actuator, etc, but in network it is dead.

We do it some changes in Z-Stack, one of them was the Minimum power voltage to initialize and write information in the NV memory. We change it from 2.0V to 2.5V.

Our perception, is that exist some problems with NV memory operation because we have diferente symptoms (described at 1st post) for the same problem and we thought that was is related with parameters that are readed / writed from NV. 
Hope this help. If you doubt persists, please feel free to ask me.

Best regards
Nalves

Hi Nalves,

We have one other patch available that deals with low voltage situations and NV operations. I will need to send it to you via a private message, so I will send you an e2e friendship request so we can proceed.