DMA during Flash write

Thanks for the reply, but from p. 320 of the MSP430F2618 user guide:

"Interrupts are automatically disabled during any flash operation when EEI = 0 and EEIEX = 0 and on
MSP430x20xx and MSP430G2xx devices where EEI and EEIEX are not present. After the flash operation
has completed, interrupts are automatically re-enabled. Any interrupt that occurred during the operation
has its associated flag set and generates an interrupt request when re-enabled."

Also, any thoughts on figuring out where exactly in memory my code resides?

Thanks again

Edit: Sorry, my previous response was cut short.

I should have said: "... not always automatically disabled ...". I was referring to F1xx,F2xx, Gxxs chip and also other chips with FLASH31 error.(;  It is prodent not to point you gun and pull the trigger even when you have the safety latched ;)

As for the Map File in IAR, you can use "Text" mode under [Project] => [Options] => [Linker] => [List] => [List] tab under [File Format]

Normally, Link will put all you code in Flash.To run some of your code in RAM, you need to (a) tell Linker to link that part of the code in the RAM addresses but load them in Flash first, and (b) copy that part from Flash into RAM yourself at run time. See IAR instructions.

Section 6.2.3.3 of the User's Guide mentions using DMA to write to flash memory, so it appears that DMA triggers are not disabled.

And what is the problem if you write a byte every 3 ms from RAM to TXBUF without DMA? Copy part of your program during runtime to RAM and execute it from there (flash block write up to 50 KByte/s). Don't use any interrupts. There will be plenty of free CPU cycles for manual testing of interrupt flags during flash write waiting for BUSY /  WAIT bits.

Thanks for all of your replies, I will respond to all of them in one place to hopefully decrease the chance that some of these questions get overlooked:

old_cow_yellow:

Thanks for the help with the memory mapping, and I'm outputting that file, but I'm having a little difficulty interpreting it; I've got a bunch of DATA16 segments, some <CODE> segments, some ISR_CODE segments, and then some miscellaneous stuff, e.g. INTVEC, CSTACK, CSTART, DATA20, RESET, etc.

Does all of that have to get copied over to RAM? If my code references variables, and I move code + variables into RAM, will the program know to read the variables in RAM, or do I have to update the pointers?

Lastly, when you say "See IAR instructions," can I find that on IAR's website somewhere?
I found this: https://e2e.ti.com/support/microcontrollers/msp430/f/166/p/72156/262251#262251
which seems to have some sample code.

Thanks so much for all the help, and sorry for the onslaught of questions

Clemens Ladisch:

Thanks, that's reassuring

zrno soli:

Without the DMA, I'd need timer-triggered interrupts to control the 3 ms timing of the transmissions from RAM to TXBUF. Also, if I don't use a DMA, I'd have to interleave writing to TXBUF with writing to flash, and even at 50 kB/s, I can transfer 128 bytes in 2.56 ms, which is cutting it pretty close.

The __ramfunc keyword makes a function execute in RAM.
The function is copied from ROM to RAM by cstartup

The keyword is specified before the return type:
__ramfunc void foo(void);

If the whole memory area used for code and constants is disabled—for example,
when the whole flash memory is being erased—only functions and data stored in RAM can be used.
Interrupts must be disabled unless the interrupt vector and the interrupt service routines are also stored in RAM

On the 5xx/6xx family devices, it is possible to re-allocate the interrupt vectors to the RAM by setting the SYSRIVECT of the SYSCTL register.

Peter McMenamin said:

Without the DMA, I'd need timer-triggered interrupts to control the 3 ms timing of the transmissions from RAM to TXBUF. Also, if I don't use a DMA, I'd have to interleave writing to TXBUF with writing to flash, and even at 50 kB/s, I can transfer 128 bytes in 2.56 ms, which is cutting it pretty close.

During flash write executed from RAM, there are plenty of free / spare CPU cycles between each double word flash write. Check flash writing function executed from RAM in 2xx family datasheet. BTW, maybe that his 3 ms can be also calculated on executed CPU cycles, without timer.  

Thanks everyone, this has all been very helpful, I just have two lingering questions:

1. Tony explained how to copy functions over to RAM, but how do I copy (possibly global) variables?

2. If I do copy a variable, and one of my functions modifies it, how do I make sure that the function modifies the variable that is located in RAM and not the original variable, which is still in Flash?

Thanks!

EDIT: Oh, and "RTFM" is also an acceptable answer