Yahoo Groups archive
Thread
2006-03-07 by ebarker123
Here is the scenario. I have an MCB2138 eval board. I have a menu talking on a serial port. Using the menu, I can set and print a variable called 'mytemp'. I set the value to say, 456, and print that out. I press reset and the value is changed to 0. Where/How is this variable getting changed. My application requires that we be able to press reset, and the SRAM contents not change. Any clues ? ebarker123
2006-03-07 by David Hawkins
ebarker123 wrote: > Here is the scenario. > > I have an MCB2138 eval board. I have a menu talking on a serial port. > > Using the menu, I can set and print a variable called 'mytemp'. > > I set the value to say, 456, and print that out. > > I press reset and the value is changed to 0. > Where/How is this variable getting changed. My application requires > that we be able to press reset, and the SRAM contents not change. > > Any clues ? The startup code zeros variables in the .data section. If you want the startup code to leave your variable alone, create a new section for it, and place it in that section. This question has been asked before, so take a look in the archives. Another option is both reset and power-down preservation, by writing the variable to flash when it changes. Look in the archives for threads on using the flash as EEPROM, there are a few discussions, and I believe there are examples in the files area. Dave
2006-03-07 by Tom Walsh
ebarker123 wrote:
>Here is the scenario.
>
>I have an MCB2138 eval board. I have a menu talking on a serial port.
>
>Using the menu, I can set and print a variable called 'mytemp'.
>
>I set the value to say, 456, and print that out.
>
>I press reset and the value is changed to 0.
>Where/How is this variable getting changed. My application requires
>that we be able to press reset, and the SRAM contents not change.
>
>
>
Look in your crt0.s file, bss sections are zeroed there. 'bss' data is
data which you declared globally and did not provide a value for. *ALL*
C code systems initialize the bss data section, all, this is a
requirement of the C language.
If you want to create memory which is part of the code but is not
initialized, add a section attribute to it and then place the section
into the linker ld script. For example:
char myString [8] __attribute__((section ("noinit")));
in the linker script, place a new section AFTER the bss. You want to
avoid placing this in front of "__bss_end__ = . ;"!
================= begin LPC2106ROM.ld ===============
/* .bss section which is used for uninitialized data */
.bss (NOLOAD) :
{
__bss_start = . ;
__bss_start__ = . ;
*(.bss)
*(COMMON)
. = ALIGN(4);
} > RAM
. = ALIGN(4);
__bss_end__ = . ;
PROVIDE (__bss_end = .);
.myvars (NOLOAD) :
{
*(noinit)
} > RAM
.stack ALIGN(256) :
{
. += STACK_SIZE;
PROVIDE (_stack = .);
} > RAM
_end = . ;
PROVIDE (end = .);
================== snip ====================
Note the new region called ".myvars" in the script? That is where I
will put the "noinit" section.
To make sure that all is well, look at the symbol file produced by the
linker. In my case it is main2106.sym, in it I look at the memory
location of '__bss_end__' and 'myString'. Here is a snippet of my
symbols file:
============= begin main2106.sym ===============
400092e2 B serial1ParityMode
400092e4 B tx1Buffer
40009ae8 B Words
40009dec B WordCnt
40009df0 A __bss_end__
40009df0 B myString
4000ae00 A end
4000ae00 A _end
4000ae00 B _stack
================= snip =====================
Actually, that is the very end of the symbol file. The variable
myString is placed at the end of the RAM. Adding more data to this new
section is quite easy now, simply assign the same attribute to each:
char myString [8] __attribute__((section ("noinit")));
struct BIG_DATA abort_info __attribute__((section ("noinit")));
int timeAborted __attribute__((section ("noinit")));
...
Hope this helps,
TomW
--
Tom Walsh - WN3L - Embedded Systems Consultant
http://openhardware.net, http://cyberiansoftware.com
"Windows? No thanks, I have work to do..."
----------------------------------------------------2006-03-07 by Guillermo Prandi
Excellent walkthrough, Tom!
I don't get why you're recommending this:
>> in the linker script, place a new section AFTER the bss.
>> You want to avoid placing this in front of "__bss_end__ = . ;"!
I would have put it *before*, and I can't think of a reason not to do
so (not that putting it after would be harmful).
Guille
--- In lpc2000@yahoogroups.com, Tom Walsh <tom@...> wrote:
>
> ebarker123 wrote:
>
> >Here is the scenario.
> >
> >I have an MCB2138 eval board. I have a menu talking on a serial
port.
> >
> >Using the menu, I can set and print a variable called 'mytemp'.
> >
> >I set the value to say, 456, and print that out.
> >
> >I press reset and the value is changed to 0.
> >Where/How is this variable getting changed. My application requires
> >that we be able to press reset, and the SRAM contents not change.
> >
> >
> >
> Look in your crt0.s file, bss sections are zeroed there. 'bss'
data is
> data which you declared globally and did not provide a value for.
*ALL*
> C code systems initialize the bss data section, all, this is a
> requirement of the C language.
>
> If you want to create memory which is part of the code but is not
> initialized, add a section attribute to it and then place the
section
> into the linker ld script. For example:
>
> char myString [8] __attribute__((section ("noinit")));
>
> in the linker script, place a new section AFTER the bss. You want
to
> avoid placing this in front of "__bss_end__ = . ;"!
>
> ================= begin LPC2106ROM.ld ===============
> /* .bss section which is used for uninitialized data */
> .bss (NOLOAD) :
> {
> __bss_start = . ;
> __bss_start__ = . ;
> *(.bss)
> *(COMMON)
> . = ALIGN(4);
> } > RAM
>
> . = ALIGN(4);
> __bss_end__ = . ;
> PROVIDE (__bss_end = .);
>
> .myvars (NOLOAD) :
> {
> *(noinit)
> } > RAM
>
> .stack ALIGN(256) :
> {
> . += STACK_SIZE;
> PROVIDE (_stack = .);
> } > RAM
>
> _end = . ;
> PROVIDE (end = .);
>
> ================== snip ====================
>
> Note the new region called ".myvars" in the script? That is where
I
> will put the "noinit" section.
>
> To make sure that all is well, look at the symbol file produced by
the
> linker. In my case it is main2106.sym, in it I look at the memory
> location of '__bss_end__' and 'myString'. Here is a snippet of my
> symbols file:
>
> ============= begin main2106.sym ===============
> 400092e2 B serial1ParityMode
> 400092e4 B tx1Buffer
> 40009ae8 B Words
> 40009dec B WordCnt
> 40009df0 A __bss_end__
> 40009df0 B myString
> 4000ae00 A end
> 4000ae00 A _end
> 4000ae00 B _stack
>
> ================= snip =====================
>
> Actually, that is the very end of the symbol file. The variable
> myString is placed at the end of the RAM. Adding more data to
this new > section is quite easy now, simply assign the same attribute to each:
>
> char myString [8] __attribute__((section ("noinit")));
>
> struct BIG_DATA abort_info __attribute__((section ("noinit")));
>
> int timeAborted __attribute__((section ("noinit")));
>
> ...
>
> Hope this helps,
>
> TomW
>
>
>
>
> --
> Tom Walsh - WN3L - Embedded Systems Consultant
> http://openhardware.net, http://cyberiansoftware.com
> "Windows? No thanks, I have work to do..."
> ----------------------------------------------------
>2006-03-07 by ebarker123
Thanks for you excellent reply. Is this for the GNU compiler ?
I failed to mention that I'm using the Keil compiler.
I've determined that the following directives do the trick.
#pragma NOINIT
char states[100]; /* Uninitialized Variable */
#pragma INIT
char array[100]; /* Initialized Variable */
Ed
--- In lpc2000@yahoogroups.com, Tom Walsh <tom@...> wrote:
>
> ebarker123 wrote:
>
> >Here is the scenario.
> >
> >I have an MCB2138 eval board. I have a menu talking on a serial port.
> >
> >Using the menu, I can set and print a variable called 'mytemp'.
> >
> >I set the value to say, 456, and print that out.
> >
> >I press reset and the value is changed to 0.
> >Where/How is this variable getting changed. My application requires
> >that we be able to press reset, and the SRAM contents not change.
> >
> >
> >
> Look in your crt0.s file, bss sections are zeroed there. 'bss' data is
> data which you declared globally and did not provide a value for.
*ALL*
> C code systems initialize the bss data section, all, this is a
> requirement of the C language.
>
> If you want to create memory which is part of the code but is not
> initialized, add a section attribute to it and then place the section
> into the linker ld script. For example:
>
> char myString [8] __attribute__((section ("noinit")));
>
> in the linker script, place a new section AFTER the bss. You want to
> avoid placing this in front of "__bss_end__ = . ;"!
>
> ================= begin LPC2106ROM.ld ===============
> /* .bss section which is used for uninitialized data */
> .bss (NOLOAD) :
> {
> __bss_start = . ;
> __bss_start__ = . ;
> *(.bss)
> *(COMMON)
> . = ALIGN(4);
> } > RAM
>
> . = ALIGN(4);
> __bss_end__ = . ;
> PROVIDE (__bss_end = .);
>
> .myvars (NOLOAD) :
> {
> *(noinit)
> } > RAM
>
> .stack ALIGN(256) :
> {
> . += STACK_SIZE;
> PROVIDE (_stack = .);
> } > RAM
>
> _end = . ;
> PROVIDE (end = .);
>
> ================== snip ====================
>
> Note the new region called ".myvars" in the script? That is where I
> will put the "noinit" section.
>
> To make sure that all is well, look at the symbol file produced by the
> linker. In my case it is main2106.sym, in it I look at the memory
> location of '__bss_end__' and 'myString'. Here is a snippet of my
> symbols file:
>
> ============= begin main2106.sym ===============
> 400092e2 B serial1ParityMode
> 400092e4 B tx1Buffer
> 40009ae8 B Words
> 40009dec B WordCnt
> 40009df0 A __bss_end__
> 40009df0 B myString
> 4000ae00 A end
> 4000ae00 A _end
> 4000ae00 B _stack
>
> ================= snip =====================
>
> Actually, that is the very end of the symbol file. The variable
> myString is placed at the end of the RAM. Adding more data to this
new > section is quite easy now, simply assign the same attribute to each:
>
> char myString [8] __attribute__((section ("noinit")));
>
> struct BIG_DATA abort_info __attribute__((section ("noinit")));
>
> int timeAborted __attribute__((section ("noinit")));
>
> ...
>
> Hope this helps,
>
> TomW
>
>
>
>
> --
> Tom Walsh - WN3L - Embedded Systems Consultant
> http://openhardware.net, http://cyberiansoftware.com
> "Windows? No thanks, I have work to do..."
> ----------------------------------------------------
>2006-03-07 by Tom Walsh
ebarker123 wrote: >Thanks for you excellent reply. Is this for the GNU compiler ? >I failed to mention that I'm using the Keil compiler. >I've determined that the following directives do the trick. > >#pragma NOINIT >char states[100]; /* Uninitialized Variable */ > >#pragma INIT >char array[100]; /* Initialized Variable */ > > > Yes, it is for the GNU gcc, I've tried this under the gcc-4.0.2 + binutils-2.16.1. TomW -- Tom Walsh - WN3L - Embedded Systems Consultant http://openhardware.net, http://cyberiansoftware.com "Windows? No thanks, I have work to do..." ----------------------------------------------------
2006-03-07 by Tom Walsh
Guillermo Prandi wrote: >Excellent walkthrough, Tom! > >I don't get why you're recommending this: > > > >>>in the linker script, place a new section AFTER the bss. >>>You want to avoid placing this in front of "__bss_end__ = . ;"! >>> >>> > >I would have put it *before*, and I can't think of a reason not to do >so (not that putting it after would be harmful). > > > Your choice, I just added it after. Consideration is this, I am using newlib stubs and the sbrk() function references the first free memory from the "end" or "_end" location specified in the linker script. Thus, the new section is "protected" from being part of the heap. TomW -- Tom Walsh - WN3L - Embedded Systems Consultant http://openhardware.net, http://cyberiansoftware.com "Windows? No thanks, I have work to do..." ----------------------------------------------------
2006-03-07 by Tom Walsh
ebarker123 wrote:
>Thanks for you excellent reply. Is this for the GNU compiler ?
>I failed to mention that I'm using the Keil compiler.
>I've determined that the following directives do the trick.
>
>#pragma NOINIT
>char states[100]; /* Uninitialized Variable */
>
>#pragma INIT
>char array[100]; /* Initialized Variable */
>
>
>
Well, that would work with Kiel, I guess. I don't use a Keil dev
system, my system is GNU compiled from source. YMMV.
TomW
>Ed
>
>
>--- In lpc2000@yahoogroups.com, Tom Walsh <tom@...> wrote:
>
>
>>ebarker123 wrote:
>>
>>
>>
>>>Here is the scenario.
>>>
>>>I have an MCB2138 eval board. I have a menu talking on a serial port.
>>>
>>>Using the menu, I can set and print a variable called 'mytemp'.
>>>
>>>I set the value to say, 456, and print that out.
>>>
>>>I press reset and the value is changed to 0.
>>>Where/How is this variable getting changed. My application requires
>>>that we be able to press reset, and the SRAM contents not change.
>>>
>>>
>>>
>>>
>>>
>>Look in your crt0.s file, bss sections are zeroed there. 'bss' data is
>>data which you declared globally and did not provide a value for.
>>
>>
>*ALL*
>
>
>>C code systems initialize the bss data section, all, this is a
>>requirement of the C language.
>>
>>If you want to create memory which is part of the code but is not
>>initialized, add a section attribute to it and then place the section
>>into the linker ld script. For example:
>>
>>char myString [8] __attribute__((section ("noinit")));
>>
>>in the linker script, place a new section AFTER the bss. You want to
>>avoid placing this in front of "__bss_end__ = . ;"!
>>
>>================= begin LPC2106ROM.ld ===============
>> /* .bss section which is used for uninitialized data */
>> .bss (NOLOAD) :
>> {
>> __bss_start = . ;
>> __bss_start__ = . ;
>> *(.bss)
>> *(COMMON)
>> . = ALIGN(4);
>> } > RAM
>>
>> . = ALIGN(4);
>> __bss_end__ = . ;
>> PROVIDE (__bss_end = .);
>>
>> .myvars (NOLOAD) :
>> {
>> *(noinit)
>> } > RAM
>>
>> .stack ALIGN(256) :
>> {
>> . += STACK_SIZE;
>> PROVIDE (_stack = .);
>> } > RAM
>>
>> _end = . ;
>> PROVIDE (end = .);
>>
>>================== snip ====================
>>
>>Note the new region called ".myvars" in the script? That is where I
>>will put the "noinit" section.
>>
>>To make sure that all is well, look at the symbol file produced by the
>>linker. In my case it is main2106.sym, in it I look at the memory
>>location of '__bss_end__' and 'myString'. Here is a snippet of my
>>symbols file:
>>
>>============= begin main2106.sym ===============
>>400092e2 B serial1ParityMode
>>400092e4 B tx1Buffer
>>40009ae8 B Words
>>40009dec B WordCnt
>>40009df0 A __bss_end__
>>40009df0 B myString
>>4000ae00 A end
>>4000ae00 A _end
>>4000ae00 B _stack
>>
>>================= snip =====================
>>
>>Actually, that is the very end of the symbol file. The variable
>>myString is placed at the end of the RAM. Adding more data to this
>>
>>
>new
>
>
>>section is quite easy now, simply assign the same attribute to each:
>>
>>char myString [8] __attribute__((section ("noinit")));
>>
>>struct BIG_DATA abort_info __attribute__((section ("noinit")));
>>
>>int timeAborted __attribute__((section ("noinit")));
>>
>>...
>>
>>Hope this helps,
>>
>>TomW
>>
>>
>>
>>
>>--
>>Tom Walsh - WN3L - Embedded Systems Consultant
>>http://openhardware.net, http://cyberiansoftware.com
>>"Windows? No thanks, I have work to do..."
>>----------------------------------------------------
>>
>>
>>
>
>
>
>
>
>
>
>Yahoo! Groups Links
>
>
>
>
>
>
>
>
--
Tom Walsh - WN3L - Embedded Systems Consultant
http://openhardware.net, http://cyberiansoftware.com
"Windows? No thanks, I have work to do..."
----------------------------------------------------