[U-Boot] Possible bug in u-boot standalone examples

[U-Boot] Possible bug in u-boot standalone examples

[David George]

Hey Wolfgang,

Hmm, I think I might not have made my point clear enough ... my 
apologies! Either that or I have a fundamental misunderstanding of the 
issue (more likely).

> No.  See the FAQ entry.  The entry point address really depends on te
> code.  If you add code, it can be basically anywehre - there may other
> functions preceed the entry point, etc.

I had a look at this FAQ entry, and I think it's addressing a different 
point than the issue I'm seeing here. This part of the FAQ tells the 
user that the load point of their standalone function may differ due to 
the particular linker or link map they're using. This is totally fine 
and I understand that this is subject to change, and naturally if you're 
attempting to run a standalone application from an address other than 
its base address defined at link time, then the application won't 
execute correctly.

In my situation, I specifically told my linker to set my example 
application's start address to to be the same as the hello_world example 
(0x40000). Presumably the same problems would have occurred regardless 
of where I loaded it due to the fact that I was beginning execution four 
bytes into the code, thus skipping the first instruction (as outlined in 
the examples).

Of course, I understand that the situation becomes very different when 
alternative compilers and linkers are thrown into the mix.

I'd be very interested to know which incidents in the past had caused 
the standalone examples to fail to execute correctly from their base 
address yet work correctly from a 4 byte offset, because as a relative 
newcomer to development at this level of abstraction, this threw me a 
massive red herring until I realised that actually, for my example, the 
first instruction is -crucial- (stack frame allocation), and I can't 
work out in my head how the example actually works correctly when loaded 
at a 4 byte offset.

Thanks!

Dave George




Stay informed by joining the Rapita Systems mailing list
http://www.rapitasystems.com/contact/mailing_list

For real-time verifications issues and discussion, follow the Rapita Systems blog
http://www.rapitasystems.com/blog

[U-Boot] Possible bug in u-boot standalone examples

[Wolfgang Denk]

Dear David,

In message <54B7D34A.6090501@rapitasystems.com> you wrote:
> 
> I had a look at this FAQ entry, and I think it's addressing a different 
> point than the issue I'm seeing here. This part of the FAQ tells the 
> user that the load point of their standalone function may differ due to 
> the particular linker or link map they're using. This is totally fine 

No, please re-read, this is not the case.  The load address is always
a constant, defined only by your linker script / linker options.

But even with a fix load address (start address of the text segment)
the entry point address may vary depending on the code in your
program.

> In my situation, I specifically told my linker to set my example 
> application's start address to to be the same as the hello_world example 
> (0x40000). Presumably the same problems would have occurred regardless 

You misunderstand.  You told the linker where the text segment starts.
But this has nothing to do with the entry point address, i. e. where
the first instruction of your first code to be executed is.  Assume 3
different C sources, say:

main() { ... }
foo() { ... }
bar() { ...}

versus

foo() { ... }
main() { ... }
bar() { ...}

or

foo() { ... }
bar() { ...}
main() { ... }

Even if you link all for the same memory map, the address of main()
will be different in all 3 cases.

> I'd be very interested to know which incidents in the past had caused 
> the standalone examples to fail to execute correctly from their base 
> address yet work correctly from a 4 byte offset, because as a relative 

Forget about this offset thing, please.  It all depends how trhe
linker arranges the code within the text segment.  Yous, you can
probably write a linker script to make sure that the first executable
instruction of your code is at the beginning of the text segment, and
that the text segment is the first one in you binary image.  No such
measures are taken in the example link command, so the entry point
address is not fix, nor is it necessarily at, or even near to, the
start of the text segment.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
If something is different, it's either better or worse,  and  usually
both.                                                    - Larry Wall