Re: Advice for AT91SAM9620 sought

[Jake Knickerbocker <jacob at turbolinux dot co dot jp>]

Matthias Kaehlcke さんは書きました:
> i had a similar (or the same?) problem a while ago, the toolchain
> generated code for armv5t, but my cpu was an armv4t.
>
> martin guy kindly provided a patch that resolved the problem:
> http://sourceware.org/ml/crossgcc/2008-05/msg00009.html
> anyway i attach you a crosstool-ng .config using relativly up to date
> components that is working for me. it uses OABI not EABI, but maybe it
> can serve you as starting point.
>
>   

Ah fantastic. Thank you so much. This config worked for me and
I was able to do a couple of test compiles that appeared to
work just fine. I should have tried for ABI in the first place maybe.
Going to give it another shot with EABI today. In which case I will
also look at that other patch you posted.

But, I did have one problem, and that is with the kernel linking,
which seems like trouble with how binutils is being built. Basically
without this patch (see below), running

arm-none-linux-objcopy -O binary -R .note -R .comment -S vmlinux linux.bin

on the kernel will output a gigantic (3.2GB!) binary kernel image:

--------------------------------------------------------------------------
diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
index 5ff5406..538fcb5 100644
--- a/arch/arm/kernel/vmlinux.lds.S
+++ b/arch/arm/kernel/vmlinux.lds.S
@@ -94,6 +94,7 @@ SECTIONS
                         TEXT_TEXT
                         SCHED_TEXT
                         LOCK_TEXT
+                       *(.note.*)
  #ifdef CONFIG_MMU
                         *(.fixup)
  #endif

--------------------------------------------------------------------------------------------------------------

Which of course is the wrong way to be fixing this.
But it does the trick and I get a booting kernel. So I am
now looking for a cause and fix for this. Strange.

But, just to see a kernel compile and run on the board is a happy
thing. I really appreciate the help.

- Jake



--
For unsubscribe information see http://sourceware.org/lists.html#faq