GH1 firmware research volunteers required

.org can't be used in reality. forget about it.
as it specify offset inside current section.
so, org 0x4000000 could cause only memory overflow, and not any usable thing :-)

I need to be able to call function at any 32bit address from my routine located in some predeterminated 32bit address (not small :-) ) And it must be normal call instructions, not using any registers and other "magic".

Other important thing is ability to specify starting address for current section (no, org is not useful here).
 
found problem - call address on assembly for long addresses have bug when there also persist registers and stack size , that is not 0 ...

P.S.
about .org - you are not right
0x40000000 = .org 0
and NOT ".org 0x40000000" ... but about that i've already sayed to you (relative to begginning of file not absolute addressing) ... ignorance is bliss :) ...
 
Last edited:
1) About ".org" I am right, because I made many experiments and all works according to GNU assembler documentation. I can't use "org 0" as I need to use data inside my programs and all of them must have proper address. I don't see any way to specify correct starting address inside assembler file. Only in linker phase you can specify it, as far as I understand.
2) About long addresses bug I did not understand anything from your explanation.
 
Sorry, but all this is useless.

Get any large routine from firmware (with calls to 0x4.. part and 0x7.. parts).
And provide procedure to compile it without any modifications (other than quickly defining constants).
 
Sorry, I'll pass this discussion, as it is going nowhere.
Right now I see many rush answers without more deep thinking about actual problem.
 
I removed few messages from DaLiV as they can be misleading to anyone who want actually look at this problem solution.


Problem description:
We have assembler routines that must be located at
0x7C100000 (for example) address. After routines we have data that are referenced from this routines (so, address is slightly larger).
Routines must be able to call any routines located in 32bit address space (ala call 0x46000000) using normal call assembler opcode
ala
.equ my_routine, 0x46000000
call my_roitine;

Right now I found following solution:
Use GNU Assembler and treat each routine name as external symbols.
Make file with such symbols and use it at linker stage (--just-symbols=filename).
Use linker options to set loading address ( -Ttext address ) <-- this one is working perfectly .

Better solution can be good.
 
Last edited:
in one file only as in examples (that was in removed posts),
multiple files - you was not wanted before (was suggested to you at early stages /talked about splitting C, asm, makefile, other to different files /)
no other ways - use symbol-file

Again?
I'll remove this post after some time as it have no value as all previous.
You still do not understand the problem completely.
So, all the points here are false.
It is pity that I spent so much time on this discussion.
 
Better solution can be good.
According to http://stackoverflow.com/questions/495262/linking-symbols-to-fixed-addresses-on-linux it seems that --just-symbols is the way to go... You ask for a "better solution" but I don't understand what you don't like. All function addresses are listed in a external file and from what I understand it is not necessary to specify the .equ in the source file because they will be treated as external symbols and solved at link time.. I probably miss the point, please be more clear in what you don't like... Anyway, I asked one person that works on embedded systems if he uses gnu tools. I hope he will reply soon
 
Last edited:
If you refer to current solution, then you are right, .equ must be omitted.

Thanks to stackoverflow reference.
I know that my solution works.
But it requires to add few parts to IDE so, it'll be possible to define starting address and all symbols inside original file and symbol file will be formed automatically.
 
If there is no different answer in stackoverflow... there is no different answer - stop! :-)

One idea could be a converter injected in the assembler script. It reads the source file, extracts the .equ lines and creates the symbol file.

If the IDE calls for the "as" executable directly, it is possible to rename "as" into "_as" and create a batch file named "as" that performs this extra step and calls _as with the new parameter --just-symbols
 
I did not say that answer at SO is different :-)
I don't need any scripts as I can change IDE :-)
It already allows to compile inline asm sections inside C files without making them strings.
 
Important

We start searching proper guys to make hadrware based firmware dumping.
Someone who have necessary tools to remove BGA packages without damage and dump firmware contents. JTAG solution is also OK.
Sony NEX firmware is much more easy to dump as it contains normal flash chips, not BGA integrated flash/RAM package as Panasonic cameras.
 
I'm have necessary tools to remove BGA packages. And I'm searching flash programmer for MN103.
I view lot of interesting info in XBOX 360 hackers sites.

Regards
 
I'm have necessary tools to remove BGA packages. And I'm searching flash programmer for MN103.
I view lot of interesting info in XBOX 360 hackers sites.
Regards

You don't need flash programmer as flash do not have any relation to MN103 :-)

You can get cheap TZ5 or ZS3 for experiments.
Look at service documentation for all available Panasonic cameras to get some ideas about package pinouts (or see Elpida docs).
Main BGA package is made by Elpida (see photos in main thread), package ID is painted over by black on most cameras.
Each package contains flash and RAM in one package.
You'll need to read flash.

NEX5, btw is much easier target, as it has fully documented flash chip and not in BGA package.
 
Some info which I have found about chip:

manufacture: ELPIDA
name: EHB01C011MA-6C-F
package: 160FBGA
Version: V2.51t
Code: 3B09A2

But datasheet is it's riddle.
 
Back
Top