DEFINE RESET_ORG and DT Instant Interrupt

[pedja089]

Now I got it.
I have done bootloaders on top of main program. At address 64000. So in main program i call bootloader with @ GOTO 64000. No need to remap anything.


When you use reset org 3000h, PBP expect int vectors on 3008h, but CPU will jump at 08h.

So in your bootloader you need to have

@ ORG 8
	@ GOTO 3008h
@ ORG 4
	@ GOTO 3004h

Then CPU will jump to 08h, then jump to 3008h, then jump to ISR. Lot of jumping...
Also, when you power PIC it will enter to bootloader first, so you need exit from bootloader
eg

 @ GOTO 03000h

But this

@ ORG 8
	@ GOTO 3008h
@ ORG 4
	@ GOTO 3004h

will generate warning in ASM, overwriting prev address, or something like that. Then it depends, what was last, so it may assemble correctly, maybe not.
EDIT:
You could try something like this in bootloader:

Reset org 10h
@ ORG 0
	@ GOTO 10h
@ ORG 8
	@ GOTO 3008h
@ ORG 4
	@ GOTO 3004h

[Marcick]

Hi Pedja, thanks for your confirmation.
Yes, I was trying as you say placing the bootloader in the high memory, but then I faced with a "limit" of PBP that requires the library routines (assembly language for high-level commands like HSEROUT) be in memory below 64K.

Here I was asking help to Melabs http://support.melabs.com/forum/picb...modifier-issue

It can works if the code is very basilar, but not for instructions like Hserout, ArrayRead etc (and I need them).

[Marcick]

I'm experimenting this method:

Bootloader code uses DEFINE RESET_ORG 80h and Main program uses DEFINE RESET_ORG 3000h

When the main app need to call bootloader, I disable Interrupt, then remap vectrors, then @ RESET:

	Madr=$EF40: WriteCode 0,Madr		' --> GoTo $80 at address 0	
	Madr=$F000: WriteCode 2,Madr		
	Madr=$EF42: WriteCode 4,Madr		' --> GoTo $84 at address 4
	Madr=$F000: WriteCode 6,Madr		
	Madr=$EF44: WriteCode 8,Madr		' --> GoTo $88 at address 8
	Madr=$F000: WriteCode 10,Madr
	For Madr=12 to 126 step 2			' fill 128 bytes block for PIC18 to complete WriteCode
		WriteCode Madr,$FF
	Next Madr

When the bootloader has finished, same logic above to restore vector at 3000, 3004, 3008, then @ RESET

Not sure if it is a smart solution, but in my mind it should works .....

[pedja089]

Code for bootloader, generally need to be very basic. Then there is little space to make mistake.
My biggest bootloader is around 1,2K, smalest 278B.
I usually use 64K+ to store received HEX. So when I verify download in main code, check checksum, etc. Then I call bootloader, just to copy from 64K+ to 0-64K.
I also used I2C, and SPI flash to store hex.
Comm is handled by main code, sometimes I download hex from FTP, sometimes is over serial, or bluetooth(same as serial)...

[pedja089]

Not good solution at all. Because you need to erase FLASH before writing.
That code need to be in bootloader, as I explained in few post above.
GOTO need to be placed in while assembling, not at runtime.
EDIT:
Try this in bootloader

DEFINE RESET_ORG 80h
@ ORG 0
	@ GOTO 80h
@ ORG 8
	@ GOTO 3008h
@ ORG 4
	@ GOTO 3004h

[Marcick]

Yes, I have to erase just a 128 byte block before.
Well, you are right, but I have not enough resources to use your method. My main program is around 80K, I have not enough space in the external EEprom so I have to manage FTP communication to download the new HEX in the bootloader code. Get a block, flash, Get a block, flash .... no crc controll at all.
I agree there is much risk something goes wrong this way, but if it works I can try remotely update a firmware without having the device back to my lab from the customer as I'm doing now.

Anyway I have space in the GSM module to save a copy of the actual firmware before starting update and manage a restore if some check fails during the update process. It is a heavy bootloader, I know.....