Kinetis MicrocontrollersのトピックRe: K60fx512 flex memory

K60fx512 flex memory

kristies1 — Fri, 30 Nov 2012 18:41:42 GMT

on K60fx512 when I try to partition flex memory, the code hangs (in debugger) or mcu resets where shown below. Code is based on flexmem_demo.

If I put breakpoint at the hang location and then reset, the flexmem has actually partitioned and works as expected as an eeprom.

Also using MQX 3.8/Codewarrior 10.2 , if this is relevant.

#define DFLASH_SIZE_448 0x04

#define EEPROM_512_512 0x34 // subsystem A = 512 bytes, subsystem B = 512 bytes

void flexinit (void)

{

SCB_SHCSR|=SCB_SHCSR_BUSFAULTENA_MASK|SCB_SHCSR_MEMFAULTENA_MASK|SCB_SHCSR_USGFAULTENA_MASK;

/* Partition the memory to enable FlexMem mode */

if ( partition_flash( EEPROM_512_512, DFLASH_SIZE_448) )

{

/* Wait for the command to complete */

while(!(FTFE_FCNFG & FTFE_FCNFG_EEERDY_MASK))

{

}

int partition_flash(int eeprom_size, int dflash_size)

{

/* Test to make sure the device is not already partitioned. If it

* is already partitioned, then return with no action performed.

if ((SIM_FCFG1 & SIM_FCFG1_DEPART(0xF)) != 0x00000F00)

{

printf("\nDevice is already partitioned.\n");

return 0;

}

/* Write the FCCOB registers */

FTFE_FCCOB0 = FTFE_FCCOB0_CCOBn(0x80); // Selects the PGMPART command

FTFE_FCCOB1 = 0x00;

FTFE_FCCOB2 = 0x00;

FTFE_FCCOB3 = 0x00;

/* FCCOB4 is written with the code for the subsystem sizes (eeprom_size define) */

FTFE_FCCOB4 = eeprom_size;

/* FFCOB5 is written with the code for the Dflash size (dflash_size define) */

FTFE_FCCOB5 = dflash_size;

/* All required FCCOBx registers are written, so launch the command */

FTFE_FSTAT = FTFE_FSTAT_CCIF_MASK;

/* Wait for the command to complete */

while(!(FTFE_FSTAT & FTFE_FSTAT_CCIF_MASK))

{

// hangs or resets here

}

return 1;

}

Re: K60fx512 flex memory

deekay — Wed, 12 Dec 2012 13:51:29 GMT

I am also struggling with the flexNVM.

Do you have any core optimization enabled? I.e. -O compile flag?

I have managed to "destroy" a K10DX device when trying to block erase the flash from code, to be able to repartition the memory. I am not sure why it happen, but I think it is because an errata issue, or I did something very wrong....

I ran into an other problem when trying to write to an already partitioned EEEPROM, but I got a hard_fault problem when writing to other flexRAM address than 0x1400 0000. Do not know why. Something you have experienced?

Daniel

Re: K60fx512 flex memory

kristies1 — Wed, 12 Dec 2012 21:42:58 GMT

I had to add the code

asm("cpsid.n i"); //disable interrupts

before the first FCCOB write

and the enable interrupts

asm("cpsie.n i "); // enable interrupts

after the wait for the commands to complete.

This fixed the lock-up problems with MQX.

I use optimization level 1, and do not try block erase from code. I have no seen any problems in the flexRAM, and use addresses past 0x14000200. But depending on how its partitioned, you may get the hard_fault if you used flexRAM address outside the partitioned area.

Re: K60fx512 flex memory

deekay — Thu, 13 Dec 2012 06:14:43 GMT

Ok. Perhaps the partitioning went bad due to lack of interrupt disabling. That might explain why I couldn't access addresses > 0x1400 0000 without hard_fault.

Yes I did use the block erase command from inside code... That seems to have corrupted the flash so I cannot program in via JTAG anymore. As according to errata e3854 described in KINETIS_0M33Z.pdf.I guess I shall use Erase Flash Sector command instead to prepare flexNVM for re-partitioning, have you tried it?

Do you know any workarounds to make it "come back from the dead"? Or shall I consider it scrap....

Daniel

Re: K60fx512 flex memory

kristies1 — Thu, 13 Dec 2012 16:06:11 GMT

The dflash always gets erased by codewarrior, so I don't have to worry about erasing prior to partitioning. According to the datasheets, the app should only partition once for the lifetime of the device.