sd_ble_gap_device_identities_set hardfault

[testsweng]

SDK15.0 whitelist hardfault

when setting the sd_ble_gap_device_identities_set using the iOS nrfConnect the non DK board hardfaults.

On an android device all works fine.

Any ideas of where to start.  The whitelist code from the PM_EVT_PEER_DATA_UPDATE_SUCCEEDED event is straight out of ble_peripheral_hids_keyboard.

static uint32_t          m_whitelist_peer_cnt = 0;                      /**< Number of peers currently in the whitelist. */
static pm_peer_id_t      m_whitelist_peers[1];                       /**< List of peers currently in the whitelist. */

                if ((p_evt->params.peer_data_update_succeeded.flash_changed)
                        && (p_evt->params.peer_data_update_succeeded.data_id == PM_PEER_DATA_ID_BONDING))
                {
                    //NRF_LOG_WARNING("New Bond, add the peer to the whitelist if possible");
                    //NRF_LOG_WARNING("\tm_whitelist_peer_cnt %d, MAX_PEERS_WLIST %d", m_whitelist_peer_cnt + 1, NRF_SDH_BLE_PERIPHERAL_LINK_COUNT);
                    // Note: You should check on what kind of white list policy your application should use.
                    if (m_whitelist_peer_cnt < 1)
                    {
                            // Bonded to a new peer, add it to the whitelist.
                            m_whitelist_peers[m_whitelist_peer_cnt++] = p_evt->peer_id;

                            // The whitelist has been modified, update it in the Peer Manager.
                            err_code = pm_device_identities_list_set(m_whitelist_peers, m_whitelist_peer_cnt);
                            if (err_code != NRF_ERROR_NOT_SUPPORTED)
                            {
                                APP_ERROR_CHECK(err_code);
                            }

                            err_code = pm_whitelist_set(m_whitelist_peers, m_whitelist_peer_cnt);
                            APP_ERROR_CHECK(err_code);
                    }
                }

I am looking thru the iOS Accessory Design guildelines and can not find a reason as to what I need to be looking at

Thanks for any help you might give.

[Einar Thorsrud]

Hi,

Can you confirm that it is a hardfault and not another error (SoftDevice assert or an error code returned by pm_device_identities_list_set())? What is the CPU registers in this state? What is the output from "nrfjprog.exe --readregs"? I suggest you disable the WDT while debugging this issue.

[testsweng]

1. Yes it is definitely a hardfault as I put a breakpoint there.

2.  Yes I turned off WDT. 

3.  Procedure now:

     A. nrfjprog -f NRF52 --eraseall;nrfjprog -f NRF52 --program s140_nrf52_6.0.0_softdevice.hex;

     B. load app thru the debugger

     c. run app and it hardfaults.  (this is right after it bonds)

     d. power cycle the board

     e.  Now the code runs fine and the phone connects/disconnect as many times as I want.

[Einar Thorsrud]

Hi,

testsweng said:

1. Yes it is definitely a hardfault as I put a breakpoint there.

I don't understand. Can you check the CPU registers (using "nrfjprog --readregs") to verify that it is a hardfault?

[testsweng]

Sorry: here they are

$ nrfjprog --readregs
R0:   0x00000004
R1:   0x000497B6
R2:   0x00026000
R3:   0x0003C11B
R4:   0x0000000A
R5:   0x0000000A
R6:   0x00000000
R7:   0x2003F8D8
R8:   0x00000000
R9:   0x00000000
R10:  0x20030000
R11:  0x00000000
R12:  0x5002B0D4
SP:   0x2003F8B0
LR:   0xFFFFFFF1
PC:   0x0003C11A
xPSR: 0x21000003
MSP:  0x2003F8B0
PSP:  0x00000000

[Einar Thorsrud]

Hi,

Perfect. It is definitely a hardfault. There is nothing here indicating what caused it, though. It could be a buffer overflow or similar somewhere in your application which corrupts memory, leading to this. 

Can you use the hardfault handler from components\libraries\hardfault\? This will print debug information that should point us in the right direction.

[testsweng]

how to I invoke it properly as it is in the makefile (hardfault_implementation) and the map file but set to zero as it not used.

[testsweng]

Ok:  I enable it in sdk_config and it shows to be in the map file. I dont seem to hit any bp's I set. I am using gcc

[Einar Thorsrud]

Hi,

You need to set HARDFAULT_HANDLER_ENABLED to 1 in sdk_config.h, and define DEBUG and DEBUG_NRF in your Makefile. You also need to add hardfault_handler_gcc.c in addition to hardfault_implementation.c. Then enable logging if you have not, and you should get a printout similar to this:

<error> hardfault: HARD FAULT at 0x00031CC4
<error> hardfault:   R0:  0x00050003  R1:  0xE000E100  R2:  0x00100000  R3:  0xBADEBA11
<error> hardfault:   R12: 0xFFFFFFFF  LR:  0x00028165  PSR: 0x01000000
<error> hardfault: Cause: Data bus error (PC value stacked for the exception return points to the instruction that caused the fault).
<error> hardfault: Bus Fault Address: 0xBADEBA11

Looking at the PC from the output from when you called "nrfjprog --readregs" it seems the PC is within the application (0x0003C11A is above the SoftDevice). What is there if you check the .mapfile or check with arm-none-eabi-addr2line?