CONFIG_BOOTBINFUNC for AT91RM9200

U-Boot-Users Steven Scholz 2 participants 12 messages Nov 23, 2004 – Nov 24, 2004

lore

Share & tools

lore

Steven Scholz

author

2004-11-23 14:11 UTC #0 lore raw

Dear Wolfgang,

Quoted 3 lines · you

Instead of doing all this which is really non-trivial and also ineffficient you could just boot from flash. I've just checked in the necessary patches for the AT91RM9200DK and CMC_PU2 boards ...

IIUC then part of this code is in cpu/at91rm9200/start.S wrapped in #ifdef CONFIG_BOOTBINFUNC.

As you know I'd like to move the at91rm9200 stuff into cpu/arm920t/at91rm9200. The start.S already calls memsetup if CONFIG_INIT_CRITICAL is defined.

Do you mind moving out the at91 specific stuff?
You already call something like lowlevelinit. Clock init could be done there.

So I'd like to move out the clock specific stuff into
a) board depended memsetup.S or clocksetup.S or
b) a cpu dependend cpu/(arm920t/)at91rm9200/clock.{c,S}.

What do you think?

And:
I don't think CONFIG_BOOTBINFUNC is a good name!
U-Boot should not "include the functionality of boot.bin in u-boot" but include lowlevel, hardware dependend init functionality!
I don't know why Atmel introduced this two stage boot process with gzipped compressed u-boot images without clock or memory setup functions. And actually I don't wanna know.
There already is CONFIG_INIT_CRITICAL. We can add CONFIG_INIT_CLOCKS or CONFIG_INIT_MEMORY or CONFIG_INIT_SDRAM or CONFIG_INIT_LOWLEVEL...

regards,

Steven Scholz

author

2004-11-23 15:42 UTC #1 lore raw

Dear Wolfgang,

is there a standard way in U-Boot to prevent to the startup code from relocating into RAM?

I know that arm920t figures out the currebt address. compares it to _TEXT_BASE and decides wether to relocate or not.

relocate: /* relocate U-Boot to RAM */
adr r0, _start /* r0 <- current position of code */
ldr r1, _TEXT_BASE /* test if we run from flash or RAM */
cmp r0, r1 /* don't reloc during debug */

So there might be no need for a define like CFG_DONT_RELOCATE. But how about other cpus?

(I am asking cause you wrapped the relocate code in cpu/at91rm9200/start.S with CONFIG_BOOTBINFUNC instead of using the above technique.

Thanks,

Steven

Wolfgang Denk

2004-11-23 16:42 UTC #2 lore raw

Quoted 3 lines · In message <[email protected]> you wrote:

is there a standard way in U-Boot to prevent to the startup code from relocating into RAM?

No. U-Boot will always relocate itself to RAM. Well, nearly always - in any halfway sane implementation.

Quoted 7 lines

I know that arm920t figures out the currebt address. compares it to _TEXT_BASE and decides wether to relocate or not.
relocate: /* relocate U-Boot to RAM */
adr r0, _start /* r0 <- current position of code */
ldr r1, _TEXT_BASE /* test if we run from flash or RAM */
cmp r0, r1 /* don't reloc during debug */

I wish that code was never written.

Quoted 2 lines

So there might be no need for a define like CFG_DONT_RELOCATE. But how about other cpus?

All CPUs always relocate the code to RAM. This is the general case. Anything else is not supported.

Quoted 2 lines

(I am asking cause you wrapped the relocate code in cpu/at91rm9200/start.S with CONFIG_BOOTBINFUNC instead of using the above technique.

Please consider the CONFIG_BOOTBINFUNC as an optimization of effort and effect (read: quick & dirty hack). Feel free to clean up as long the result is working code, too.

Best regards,

Wolfgang Denk

Signature · 5 lines

Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Fools ignore complexity. Pragmatists suffer it. Some  can  avoid  it.
Geniuses remove it.
     - Perlis's Programming Proverb #58, SIGPLAN Notices, Sept.  1982

Steven Scholz

author

2004-11-23 16:51 UTC #3 lore raw

Quoted 12 lines · Wolfgang Denk

Quoted 2 lines · In message <[email protected]> you wrote:
is there a standard way in U-Boot to prevent to the startup code from relocating into RAM?
No. U-Boot will always relocate itself to RAM. Well, nearly always - in any halfway sane implementation.
Quoted 3 lines
I know that arm920t figures out the currebt address. compares it to _TEXT_BASE and decides wether to relocate or not.

Quoted 7 lines

Quoted 4 lines
relocate: /* relocate U-Boot to RAM */
adr r0, _start /* r0 <- current position of code */
ldr r1, _TEXT_BASE /* test if we run from flash or RAM */
cmp r0, r1 /* don't reloc during debug */
I wish that code was never written.

Why is that so bad? It makes it possible to debug U-Boot with just loading the image into RAM using BDI2000.

Quoted 2 lines

All CPUs always relocate the code to RAM. This is the general case. Anything else is not supported.

But how would you debug U-Boot?
I though the common way is to disabled relocation and memsetup, load that U-Boot image with an BDI2000 into RAM and execute it - skipping the init code...

--
Steven

Wolfgang Denk

2004-11-23 17:18 UTC #4 lore raw

Quoted 4 lines · In message <[email protected]> you wrote:

Quoted 2 lines
I wish that code was never written.
Why is that so bad? It makes it possible to debug U-Boot with just loading the image into RAM using BDI2000.

Which is an unsupported mode of operation which works for a handfull of experts and causes confusion with many, many newbees.

Quoted 1 line

But how would you debug U-Boot?

As I always do. Attach the BDI, burn to flash, start in GDB.

Quoted 2 lines

I though the common way is to disabled relocation and memsetup, load that U-Boot image with an BDI2000 into RAM and execute it - skipping the init code...

No. This is not the way things should be done. You can do this if you know exactly what you're doing, but it does not help you nor will it save you any time. On contrary. It will cost additional time.

Anyway, this is a FAQ, so please RTFM.

Best regards,

Wolfgang Denk

Signature · 4 lines

Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Consultants are mystical people who ask a company for  a  number  and
then give it back to them.

Wolfgang Denk

2004-11-23 15:45 UTC #5 lore raw

In reply to Steven Scholz

Quoted 6 lines · In message <[email protected]> you wrote:

So I'd like to move out the clock specific stuff into
a) board depended memsetup.S or clocksetup.S or
b) a cpu dependend cpu/(arm920t/)at91rm9200/clock.{c,S}.
What do you think?

If you submit a patch, and I find that it works on all systems I can test, then it is OK with me.

Quoted 4 lines

And:
I don't think CONFIG_BOOTBINFUNC is a good name!
U-Boot should not "include the functionality of boot.bin in u-boot" but include lowlevel, hardware dependend init functionality!

OK, ok, ok.

Quoted 2 lines

There already is CONFIG_INIT_CRITICAL. We can add CONFIG_INIT_CLOCKS or CONFIG_INIT_MEMORY or CONFIG_INIT_SDRAM or CONFIG_INIT_LOWLEVEL...

The current setup allows to change a single #define to switch between two different modes of operation: using boot.bin, and not using boot.bin. I don't really care about the name of theis variable as long as it's decoumented and works. But please don;t split it into 15 different names that must be set in a consistent way.

Best regards,

Wolfgang Denk

Signature · 5 lines

Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"We have the right to survive!"
"Not be killing others."
	-- Deela and Kirk, "Wink of An Eye", stardate 5710.5

Steven Scholz

author

2004-11-23 15:51 UTC #6 lore raw

Quoted 3 lines · Wolfgang Denk

Quoted 2 lines
And:
I don't think CONFIG_BOOTBINFUNC is a good name!
OK, ok, ok.

;-)

Quoted 8 lines

Quoted 2 lines
There already is CONFIG_INIT_CRITICAL. We can add CONFIG_INIT_CLOCKS or CONFIG_INIT_MEMORY or CONFIG_INIT_SDRAM or CONFIG_INIT_LOWLEVEL...
The current setup allows to change a single #define to switch between two different modes of operation: using boot.bin, and not using boot.bin. I don't really care about the name of theis variable as long as it's decoumented and works. But please don;t split it into 15 different names that must be set in a consistent way.

Then I think CONFIG_INIT_CRITICAL is the best choice!

--
Steven

Wolfgang Denk

2004-11-23 16:52 UTC #7 lore raw

Quoted 2 lines · In message <[email protected]> you wrote:

Then I think CONFIG_INIT_CRITICAL is the best choice!

Definitely not - CONFIG_INIT_CRITICAL is such an ugly name that I wonder who accepted a patch to introduce such a name ;-)

There is no documentation what it does or means; there is not even a
definitian what "sys-critical inits" might be - all the
initializations are "system critical". This is bullsh*i.

We should not use this name, but try to get rid of it.

Best regards,

Wolfgang Denk

Signature · 4 lines

Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"To take a significant step forward, you must make a series of finite
improvements." - Donald J. Atwood, General Motors

Steven Scholz

author

2004-11-23 17:01 UTC #8 lore raw

Quoted 5 lines · Wolfgang Denk

Quoted 1 line · In message <[email protected]> you wrote:
Then I think CONFIG_INIT_CRITICAL is the best choice!

Quoted 8 lines

Definitely not - CONFIG_INIT_CRITICAL is such an ugly name that I wonder who accepted a patch to introduce such a name ;-)
There is no documentation what it does or means; there is not even a
definitian what "sys-critical inits" might be - all the
initializations are "system critical". This is bullsh*i.
We should not use this name, but try to get rid of it.

Hmm. So I suggest replacing it with CONFIG_INIT_LOWLEVEL or CONFIG_LOWLEVEL_INIT.

But then: these init steps (clocks and memory) are vital for U-Boot to run. So why do we need this define anyway if these steps should be done always... ?

Signature · 1 line

Steven Scholz

Wolfgang Denk

2004-11-23 17:23 UTC #9 lore raw

Quoted 2 lines · In message <[email protected]> you wrote:

Hmm. So I suggest replacing it with CONFIG_INIT_LOWLEVEL or CONFIG_LOWLEVEL_INIT.

Hmmm.. CONFIG or CFG? And what exactly is low (vs. high?) level here?

Quoted 2 lines

But then: these init steps (clocks and memory) are vital for U-Boot to run. So why do we need this define anyway if these steps should be done always... ?

Good question. The only reason I can think of is to allob such a two-stage boot loader as used by Atmel on the AT91RM9200DK.

Best regards,

Wolfgang Denk

Signature · 4 lines

Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
There's no sense in being precise  when  you  don't  even  know  what
you're talking about.                             -- John von Neumann

Steven Scholz

author

2004-11-24 12:54 UTC #10 lore raw

Quoted 8 lines · Wolfgang Denk

Quoted 1 line · In message <[email protected]> you wrote:
Hmm. So I suggest replacing it with CONFIG_INIT_LOWLEVEL or CONFIG_LOWLEVEL_INIT.
Hmmm.. CONFIG or CFG? And what exactly is low (vs. high?) level here?

Quoted 6 lines

Quoted 2 lines
But then: these init steps (clocks and memory) are vital for U-Boot to run. So why do we need this define anyway if these steps should be done always... ?
Good question. The only reason I can think of is to allob such a two-stage boot loader as used by Atmel on the AT91RM9200DK.

So we check if all other arm920t boards have an implementation of memsetup(). Then kick out CONFIG_INIT_CRITICAL and introduce CONFIG_DO_NOT_MAKE_LOWLEVEL_INIT or something like that !? ;-)

Signature · 1 line

Steven Scholz

Wolfgang Denk

2004-11-24 15:06 UTC #11 lore raw

Quoted 4 lines · In message <[email protected]> you wrote:

So we check if all other arm920t boards have an implementation of memsetup(). Then kick out CONFIG_INIT_CRITICAL and introduce CONFIG_DO_NOT_MAKE_LOWLEVEL_INIT or something like that !? ;-)

Sounds like a plan to me - but let's use a better name, please.

Best regards,

Wolfgang Denk

Signature · 4 lines

Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Why can you only have two doors on a chicken coop? If it had four  it
would be a chicken sedan.