LPC54618: Hard Fault When entering DeepSleep

[AleemS]

When trying to enter deepsleep mode on the LPC54618 an HardFault occurs. Going through step by step with gdb the instruction

0x200058e0 <+0>: push {r4, r5, r6, r7, lr}

in the function relocMem.8033 causes the HardFault. The Main stack starts at 0x200044a8 and has Size 0x1000. I have tried calling POWER_EnterDeepSleep with arguments uint64_t(-1), 0, and

(uint64_t(SYSCON_PDRUNCFG_PDEN_SYSOSC_MASK | SYSCON_PDRUNCFG_PDEN_EEPROM_MASK) << 32) |

        (SYSCON_PDRUNCFG_PDEN_SRAMX_MASK | SYSCON_PDRUNCFG_PDEN_SRAM0_MASK |

         SYSCON_PDRUNCFG_PDEN_SRAM1_2_3_MASK | SYSCON_PDRUNCFG_PDEN_VD6_MASK), which expands to 0x28'2000'E000.

Context:

++ MbedOS Fault Handler ++

FaultType: HardFault

Context:
R 0: FFFFFFFF
R 1: FFFFFFFF
R 2: 00000000
R 3: 200058E1
R 4: 20002BC4
R 5: 00000008
R 6: 20005474
R 7: 000084A1
R 8: 200030C4
R 9: 00000001
R 10: 00021744
R 11: 00000000
R 12: 200058E0
SP : 20005438
LR : 000088A5
PC : 200058E0
xPSR : 610F0000
PSP : 200053D0
MSP : 20027FC0
CPUID: 410FC241
HFSR : 40000000
MMFSR: 00000001
BFSR : 00000000
UFSR : 00000000
DFSR : 00000000
AFSR : 00000000
Mode : Thread
Priv : Privileged
Stack: PSP

-- MbedOS Fault Handler --

++ MbedOS Error Info ++
Error Status: 0x80FF013D Code: 317 Module: 255
Error Message: Fault exception
Location: 0x200058E0
Error Value: 0x20005538
Current Thread: main Id: 0x20004460 Entry: 0x111A1 StackSize: 0x1000 StackMem: 0x200044A8 SP: 0x20005438
For more info, visit: https://mbed.com/s/error?error=0x80FF013D
-- MbedOS Error Info --

[AleemS]

Unfortunately it still HardFaults just at another address (due to the build being a different one) but the same instruction. Running the underlying function DeepSleepReloc directly does not cause a hard fault.

[xiangjun_rong]

Hi,

As the following figure of LPC546xx memory map, the address 0x2000_xxxx are SRAM, for the error code:0x200058e0 <+0>: push {r4, r5, r6, r7, lr}, it appears that you allocate/execute the code in SRAM instead of Flash, while you allocate main stack in address 0x200044a8, as you see the code address(0x200058e0) and stack address(0x200044a8) are approaching, I guess that the code is overlapped with stack, that is why you get the hard-fault error, but it is only guess.

Pls try to allocate the application code in flash instead of SRAM or the other SRAM address and have a try.

BR

XiangJun Rong

[AleemS]

Is it possible, that calling the function directly from flash, corrupts the system state and makes interrupts impossible?

Because after calling the function from Flash i cannot cause the chip to wake up from deepsleep.

[xiangjun_rong]

Hi

As you know that the LPC546xx uses interrupt mechanism to wake-up the processor from deep-sleep state, so you have to set up the ISR of the waking-up  source. I suggest you run all the code in flash.

Anyway, in SDK, NXP provides power_manager_lpc example, which is fully tested,  If you run the power_manager_lpc example directly, do you have any issue?

BR

XiangJun Rong

[AleemS · Accepted solution]

Hello,

i have solved the problem. In Order to call POWER_EnterDeepSleep in an mbed-os Program you have to create an mbed::ScopedRamExecutionLock inside the Scope that needs to execute from RAM. Otherwise the execution will cause a Hardfault because of the MemoryProtectionUnit. It might be similarly handled in other Real-Time-Operating-Systems.

This also seems to eliminate all Problems with the Interrupts.

[xiangjun_rong]

Hi,

If you run the power_manager_lpc example directly, do you have any issue?

This is the code in above example when you enter deep sleep:

#define APP_EXCLUDE_FROM_DEEPSLEEP \
(SYSCON_PDRUNCFG_PDEN_SRAMX_MASK | SYSCON_PDRUNCFG_PDEN_SRAM0_MASK | SYSCON_PDRUNCFG_PDEN_SRAM1_2_3_MASK)

case kPmu_Deep_Sleep: /* Enter deep sleep mode. */

POWER_EnterDeepSleep(APP_EXCLUDE_FROM_DEEPSLEEP);

break;

Pls have a try

BR

XiangJun Rong