From carlo.caione at gmail.com  Tue May 15 22:33:57 2012
From: carlo.caione at gmail.com (Carlo Caione)
Date: Wed, 16 May 2012 00:33:57 +0200
Subject: [arm-gnu] undefined instruction
Message-ID: <CAOQ7t2ZxttqTjx8XOJu9=vOHOpAG1WO_6N47AU225LCP0fovoQ@mail.gmail.com>

Hi,
In a big project (barebox) I have this simple snippet of code in string.c

size_t strlen(const char * s)
{
	const char *sc;

	for (sc = s; *sc != '\0'; ++sc)
		/* nothing */;
	return sc - s;
}

I'm using Sourcery CodeBench Lite 2011.09-69 to compile the project.
In particular for string.c:

arm-none-eabi-gcc -Wp,-MD,lib/.string.o.d  -nostdinc -isystem
/opt/CodeSourcery/Sourcery_CodeBench_Lite_for_ARM_EABI/bin/../lib/gcc/arm-none-eabi/4.6.1/include
-D__KERNEL__ -D__BAREBOX__ -Iinclude
-I/home/carlo/projects/barebox/arch/arm/include
-I/home/carlo/projects/barebox/arch/arm/include -include
include/generated/autoconf.h -include
/home/carlo/projects/barebox/include/linux/kconfig.h -fno-builtin
-ffreestanding -D__ARM__ -fno-strict-aliasing -marm -mlittle-endian
-mabi=aapcs-linux -mno-thumb-interwork -D__LINUX_ARM_ARCH__=7
-march=armv7-a  -msoft-float  -Iarch/arm/mach-omap/include
-DTEXT_BASE=0x40200000 -fdata-sections -ffunction-sections -Wall
-Wundef -Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing
-fno-common -Os -pipe -g  -fno-stack-protector
-Wno-unused-but-set-variable -Wno-trampolines
-Wdeclaration-after-statement -Wno-pointer-sign
-D"KBUILD_STR(s)=#s" -D"KBUILD_BASENAME=KBUILD_STR(string)"
-D"KBUILD_MODNAME=KBUILD_STR(string)" -c -o lib/string.o lib/string.c

The problem is that the disassembly looks like:

40205c84: strlen+0              beq 0x40205d18 <memcpy>
40205c88: strlen+4               ; <UNDEFINED> instruction: 0xf03fe03a
40205c8c: strlen+8              andeq r0, r0, r6
40205c90: strlen+12             andeq r0, r0, r4
40205c94: strlen+16             andeq r0, r0, r7
40205c98: strlen+20             andeq r0, r0, sp, lsl r0
40205c9c: strlen+24             ldrdmi r0, [r0], -r3
40205ca0: strlen+28              ; <UNDEFINED> instruction: 0xffffffff

Why the undefined instruction?

Thank you,

--
Carlo


From carlos_odonell at mentor.com  Tue May 15 23:04:06 2012
From: carlos_odonell at mentor.com (Carlos O'Donell)
Date: Tue, 15 May 2012 19:04:06 -0400
Subject: [arm-gnu] undefined instruction
In-Reply-To: <CAOQ7t2ZxttqTjx8XOJu9=vOHOpAG1WO_6N47AU225LCP0fovoQ@mail.gmail.com>
References: <CAOQ7t2ZxttqTjx8XOJu9=vOHOpAG1WO_6N47AU225LCP0fovoQ@mail.gmail.com>
Message-ID: <4FB2E0E6.7090400@mentor.com>

On 5/15/2012 6:33 PM, Carlo Caione wrote:
> Hi,
> In a big project (barebox) I have this simple snippet of code in string.c
> 
> size_t strlen(const char * s)
> {
> 	const char *sc;
> 
> 	for (sc = s; *sc != '\0'; ++sc)
> 		/* nothing */;
> 	return sc - s;
> }
> 
> I'm using Sourcery CodeBench Lite 2011.09-69 to compile the project.
> In particular for string.c:
> 
> arm-none-eabi-gcc -Wp,-MD,lib/.string.o.d  -nostdinc -isystem
> /opt/CodeSourcery/Sourcery_CodeBench_Lite_for_ARM_EABI/bin/../lib/gcc/arm-none-eabi/4.6.1/include
> -D__KERNEL__ -D__BAREBOX__ -Iinclude
> -I/home/carlo/projects/barebox/arch/arm/include
> -I/home/carlo/projects/barebox/arch/arm/include -include
> include/generated/autoconf.h -include
> /home/carlo/projects/barebox/include/linux/kconfig.h -fno-builtin
> -ffreestanding -D__ARM__ -fno-strict-aliasing -marm -mlittle-endian
> -mabi=aapcs-linux -mno-thumb-interwork -D__LINUX_ARM_ARCH__=7
> -march=armv7-a  -msoft-float  -Iarch/arm/mach-omap/include
> -DTEXT_BASE=0x40200000 -fdata-sections -ffunction-sections -Wall
> -Wundef -Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing
> -fno-common -Os -pipe -g  -fno-stack-protector
> -Wno-unused-but-set-variable -Wno-trampolines
> -Wdeclaration-after-statement -Wno-pointer-sign
> -D"KBUILD_STR(s)=#s" -D"KBUILD_BASENAME=KBUILD_STR(string)"
> -D"KBUILD_MODNAME=KBUILD_STR(string)" -c -o lib/string.o lib/string.c
> 
> The problem is that the disassembly looks like:
> 
> 40205c84: strlen+0              beq 0x40205d18 <memcpy>
> 40205c88: strlen+4               ; <UNDEFINED> instruction: 0xf03fe03a
> 40205c8c: strlen+8              andeq r0, r0, r6
> 40205c90: strlen+12             andeq r0, r0, r4
> 40205c94: strlen+16             andeq r0, r0, r7
> 40205c98: strlen+20             andeq r0, r0, sp, lsl r0
> 40205c9c: strlen+24             ldrdmi r0, [r0], -r3
> 40205ca0: strlen+28              ; <UNDEFINED> instruction: 0xffffffff
> 
> Why the undefined instruction?

What did you use to get this disassembly?

When you compile with -save-temps what does the output assembly look like before being assembled?

It might be that the mapping symbols are missing and your disassembling constant pool data.

What are the .gnu_attribute flags for the object file?

Cheers,
Carlos.
-- 
Carlos O'Donell
Mentor Graphics / CodeSourcery
carlos_odonell at mentor.com
carlos at codesourcery.com
+1 (613) 963 1026


From graub at hotmail.com  Wed May 16 05:46:14 2012
From: graub at hotmail.com (Bill HG)
Date: Wed, 16 May 2012 05:46:14 +0000
Subject: [arm-gnu] building the lite package in Ubuntu or MAC OS X
Message-ID: <DUB113-W16BA9A4A40B46DC87E52ACAC180@phx.gbl>


Hi,

I subscribed to your arm-gnu forum list...

This is the issue I am encountering:

the source code package 

arm-2009q3-68-arm-none-eabi

requires that I run the script:

arm-2009q3-68-arm-none-eabi.sh

However every time I attempt to run the script, it fails with the following error message:

./arm-2009q3-68-arm-none-eabi.sh: line 136: sed: command not found
./arm-2009q3-68-arm-none-eabi.sh: line 136: grep: command not found

This happens in Ubuntu 11.10 and MAC OS X 10.6.8

grep and sed are both installed and in the path:

/bin in Ubuntu

and

/usr/bin in MAC OS X

I will appreciate if someone publishes a workaround to this...

Incidentally, the .bin installation goes very smoothly on Ubuntu, but it targets only Linux flavours, and not MAC OS X...
I prefer my MAC as I am more used to it, and have put all my resources there...

thanks...

Billy G
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sourcerytools.com/pipermail/arm-gnu/attachments/20120516/42bafdb4/attachment.html>

From carlo.caione at gmail.com  Wed May 16 19:02:07 2012
From: carlo.caione at gmail.com (Carlo Caione)
Date: Wed, 16 May 2012 21:02:07 +0200
Subject: [arm-gnu] building the lite package in Ubuntu or MAC OS X
In-Reply-To: <DUB113-W16BA9A4A40B46DC87E52ACAC180@phx.gbl>
References: <DUB113-W16BA9A4A40B46DC87E52ACAC180@phx.gbl>
Message-ID: <CAOQ7t2Zs3F48T6UqHJhoODCDpKDWaR8cD5Wc5y0tPFbnPnS3xA@mail.gmail.com>

On Wed, May 16, 2012 at 7:46 AM, Bill HG <graub at hotmail.com> wrote:
> Hi,
>
> I subscribed to your arm-gnu forum list...
>
> This is the issue I am encountering:
>
> the source code package
>
> arm-2009q3-68-arm-none-eabi
>
> requires that I run the script:
>
> arm-2009q3-68-arm-none-eabi.sh
>
> However every time I attempt to run the script, it fails with the following
> error message:
>
> ./arm-2009q3-68-arm-none-eabi.sh: line 136: sed: command not found
> ./arm-2009q3-68-arm-none-eabi.sh: line 136: grep: command not found
>
> This happens in Ubuntu 11.10 and MAC OS X 10.6.8
>
> grep and sed are both installed and in the path:
>
> /bin in Ubuntu
>
> and
>
> /usr/bin in MAC OS X
>
> I will appreciate if someone publishes a workaround to this...
>
> Incidentally, the .bin installation goes very smoothly on Ubuntu, but it
> targets only Linux flavours, and not MAC OS X...
> I prefer my MAC as I am more used to it, and have put all my resources
> there...
>
> thanks...
>
> Billy G

This worked well for me:

https://github.com/jsnyder/arm-eabi-toolchain

Enjoy :)


From carlo.caione at gmail.com  Wed May 16 19:43:15 2012
From: carlo.caione at gmail.com (Carlo Caione)
Date: Wed, 16 May 2012 21:43:15 +0200
Subject: [arm-gnu] undefined instruction
In-Reply-To: <4FB2E0E6.7090400@mentor.com>
References: <CAOQ7t2ZxttqTjx8XOJu9=vOHOpAG1WO_6N47AU225LCP0fovoQ@mail.gmail.com>
	<4FB2E0E6.7090400@mentor.com>
Message-ID: <CAOQ7t2aLGp1C-jd+EkiYSPrEtapuYcx153HSf1WM3=JegHso_w@mail.gmail.com>

On Wed, May 16, 2012 at 1:04 AM, Carlos O'Donell
<carlos_odonell at mentor.com> wrote:

> What did you use to get this disassembly?

arm-none-aebi-gdb in eclipse. Without eclipse is (obviously :) the same:

(gdb) disassemble
Dump of assembler code for function strlen:
=> 0x40205c84 <+0>:    beq    0x40205d18 <memcpy>
   0x40205c88 <+4>:            ; <UNDEFINED> instruction: 0xf03fe03a
   0x40205c8c <+8>:    andeq    r0, r0, r6
   0x40205c90 <+12>:    andeq    r0, r0, r4
   0x40205c94 <+16>:    andeq    r0, r0, r7
   0x40205c98 <+20>:    andeq    r0, r0, sp, lsl r0
   0x40205c9c <+24>:    ldrdmi    r0, [r0], -r3
   0x40205ca0 <+28>:            ; <UNDEFINED> instruction: 0xffffffff
End of assembler dump.

> When you compile with -save-temps what does the output assembly look like before being assembled?

.section        .text.strlen,"ax",%progbits
        .align  2
        .global strlen
        .type   strlen, %function
strlen:
.LFB83:
        .loc 1 223 0
        .cfi_startproc
        @ args = 0, pretend = 0, frame = 0
        @ frame_needed = 0, uses_anonymous_args = 0
        @ link register save eliminated.
.LVL34:
        .loc 1 223 0
        mov     r3, r0  @ ivtmp.163, s
.LVL35:
.L36:
        .loc 1 226 0 discriminator 1
        mov     r2, r3  @ sc, ivtmp.163
.LVL36:
        add     r3, r3, #1      @ ivtmp.163, ivtmp.163,
.LVL37:
        ldrb    r1, [r2, #0]    @ zero_extendqisi2      @ MEM[base:
sc_1, offset: 0B], MEM[base: sc_1, offset: 0B]
        cmp     r1, #0  @ MEM[base: sc_1, offset: 0B],
        bne     .L36    @,
        .loc 1 229 0
        rsb     r0, r0, r2      @, s, sc
.LVL38:
        bx      lr      @
        .cfi_endproc
.LFE83:
        .size   strlen, .-strlen

This is the disassebly of the object file:

Disassembly of section .text.strlen:

00000000 <strlen>:
   0:   e1a03000        mov     r3, r0
   4:   e1a02003        mov     r2, r3
   8:   e2833001        add     r3, r3, #1
   c:   e5d21000        ldrb    r1, [r2]
  10:   e3510000        cmp     r1, #0
  14:   1afffffa        bne     4 <strlen+0x4>
  18:   e0600002        rsb     r0, r0, r2
  1c:   e12fff1e        bx      lr


> It might be that the mapping symbols are missing and your disassembling constant pool data.
>
> What are the .gnu_attribute flags for the object file?

I don't have any flags (If I have correctly understood what you mean).

--
Carlo


From carlos_odonell at mentor.com  Wed May 16 19:55:54 2012
From: carlos_odonell at mentor.com (Carlos O'Donell)
Date: Wed, 16 May 2012 15:55:54 -0400
Subject: [arm-gnu] building the lite package in Ubuntu or MAC OS X
In-Reply-To: <DUB113-W16BA9A4A40B46DC87E52ACAC180@phx.gbl>
References: <DUB113-W16BA9A4A40B46DC87E52ACAC180@phx.gbl>
Message-ID: <4FB4064A.6030502@mentor.com>

On 5/16/2012 1:46 AM, Bill HG wrote:
> However every time I attempt to run the script, it fails with the following error message:
> *./arm-2009q3-68-arm-none-eabi.sh: line 136: sed: command not found
> ./arm-2009q3-68-arm-none-eabi.sh: line 136: grep: command not found*
> I will appreciate if someone publishes a workaround to this...

Please read the *.sh file comment header to learn about the file and what purpose it serves.

> Incidentally, the .bin installation goes very smoothly on Ubuntu, but it targets only Linux flavours, and not MAC OS X...
> I prefer my MAC as I am more used to it, and have put all my resources there...

Unfortunately we don't don't currently support Mac OS X.

I would suggest running a VM and using the Linux or Windows tools from the VM.

Using a VM for development is very popular on Windows.

Cheers,
Carlos.
-- 
Carlos O'Donell
Mentor Graphics / CodeSourcery
carlos_odonell at mentor.com
carlos at codesourcery.com
+1 (613) 963 1026


From carlos_odonell at mentor.com  Wed May 16 20:05:45 2012
From: carlos_odonell at mentor.com (Carlos O'Donell)
Date: Wed, 16 May 2012 16:05:45 -0400
Subject: [arm-gnu] undefined instruction
In-Reply-To: <CAOQ7t2aLGp1C-jd+EkiYSPrEtapuYcx153HSf1WM3=JegHso_w@mail.gmail.com>
References: <CAOQ7t2ZxttqTjx8XOJu9=vOHOpAG1WO_6N47AU225LCP0fovoQ@mail.gmail.com>
	<4FB2E0E6.7090400@mentor.com>
	<CAOQ7t2aLGp1C-jd+EkiYSPrEtapuYcx153HSf1WM3=JegHso_w@mail.gmail.com>
Message-ID: <4FB40899.7090705@mentor.com>

On 5/16/2012 3:43 PM, Carlo Caione wrote:
> On Wed, May 16, 2012 at 1:04 AM, Carlos O'Donell
> <carlos_odonell at mentor.com> wrote:
> 
>> What did you use to get this disassembly?
> 
> arm-none-aebi-gdb in eclipse. Without eclipse is (obviously :) the same:
> 
> (gdb) disassemble
> Dump of assembler code for function strlen:
> => 0x40205c84 <+0>:    beq    0x40205d18 <memcpy>
>    0x40205c88 <+4>:            ; <UNDEFINED> instruction: 0xf03fe03a
>    0x40205c8c <+8>:    andeq    r0, r0, r6
>    0x40205c90 <+12>:    andeq    r0, r0, r4
>    0x40205c94 <+16>:    andeq    r0, r0, r7
>    0x40205c98 <+20>:    andeq    r0, r0, sp, lsl r0
>    0x40205c9c <+24>:    ldrdmi    r0, [r0], -r3
>    0x40205ca0 <+28>:            ; <UNDEFINED> instruction: 0xffffffff
> End of assembler dump.

Either gdb thinks this is wrong architecture/endian, or you didn't load anything and it's disassembling garbage from memory, or the probe is returning garbage for that memory location. You need to rule these out one at a time.

Switch to the gdb console and issue `show architecture', then issue `show endian', what does it show?

>> When you compile with -save-temps what does the output assembly look like before being assembled?

That looks good.

> This is the disassebly of the object file:

This looks good.

>> It might be that the mapping symbols are missing and your disassembling constant pool data.
>>
>> What are the .gnu_attribute flags for the object file?
> 
> I don't have any flags (If I have correctly understood what you mean).

Run `*-readelf -Wa` on your file (there is an run external tool for this in the IDE), and provide the output of the "Attribute section"

e.g.

Attribute Section: aeabi
File Attributes
  Tag_CPU_name: "ARM7TDMI"
  Tag_CPU_arch: v4T
  Tag_ARM_ISA_use: Yes
  Tag_THUMB_ISA_use: Thumb-1
  Tag_FP_arch: VFPv2
  Tag_ABI_PCS_wchar_t: 4
  Tag_ABI_FP_denormal: Needed
  Tag_ABI_FP_exceptions: Needed
  Tag_ABI_FP_number_model: IEEE 754
  Tag_ABI_align_needed: 8-byte
  Tag_ABI_enum_size: small
  Tag_DIV_use: Not allowed


Cheers,
Carlos.
-- 
Carlos O'Donell
Mentor Graphics / CodeSourcery
carlos_odonell at mentor.com
carlos at codesourcery.com
+1 (613) 963 1026


From carlo.caione at gmail.com  Wed May 16 21:27:41 2012
From: carlo.caione at gmail.com (Carlo Caione)
Date: Wed, 16 May 2012 23:27:41 +0200
Subject: [arm-gnu] undefined instruction
In-Reply-To: <4FB40899.7090705@mentor.com>
References: <CAOQ7t2ZxttqTjx8XOJu9=vOHOpAG1WO_6N47AU225LCP0fovoQ@mail.gmail.com>
	<4FB2E0E6.7090400@mentor.com>
	<CAOQ7t2aLGp1C-jd+EkiYSPrEtapuYcx153HSf1WM3=JegHso_w@mail.gmail.com>
	<4FB40899.7090705@mentor.com>
Message-ID: <CAOQ7t2aZpf-J-HP+aLkQZxSuU4U9xNLX65gW-ZmLrMo9wBar4w@mail.gmail.com>

On Wed, May 16, 2012 at 10:05 PM, Carlos O'Donell
<carlos_odonell at mentor.com> wrote:
[cut]

Ok, I think I have spotted the bug. And actually the problem seems to
me related to qemu-linaro (yes, I'm emulating a Cortex-A8 with qemu)
and not to CodeSourcery.
In fact in the code an exact point does exist in which the strlen
routine gets corrupted.


(gdb) target remote :1234
Remote debugging using :1234
0x40014000 in ?? ()
(gdb) disassemble __mmu_cache_flush,+48
Dump of assembler code from 0x402000d0 to 0x40200100:
   0x402000d0 <__mmu_cache_flush+0>:	mrc	15, 0, r10, cr0, cr1, {5}
   0x402000d4 <__mmu_cache_flush+4>:	tst	r10, #983040	; 0xf0000
   0x402000d8 <__mmu_cache_flush+8>:	mov	r10, #0
   0x402000dc <__mmu_cache_flush+12>:	beq	0x402000e8 <__mmu_cache_flush+24>
   0x402000e0 <__mmu_cache_flush+16>:	mcr	15, 0, r10, cr7, cr14, {0}
   0x402000e4 <__mmu_cache_flush+20>:	b	0x40200178 <iflush>
   0x402000e8 <__mmu_cache_flush+24>:	mcr	15, 0, r10, cr7, cr10, {5}
   0x402000ec <__mmu_cache_flush+28>:	push	{r0, r1, r2, r3, r4, r5,
r6, r7, r9, r10, r11}
   0x402000f0 <__mmu_cache_flush+32>:	mrc	15, 1, r0, cr0, cr0, {1}
   0x402000f4 <__mmu_cache_flush+36>:	ands	r3, r0, #117440512	; 0x7000000
   0x402000f8 <__mmu_cache_flush+40>:	lsr	r3, r3, #23
   0x402000fc <__mmu_cache_flush+44>:	beq	0x4020016c <finished>
End of assembler dump.
(gdb) b *0x402000ec
Breakpoint 1 at 0x402000ec
(gdb) c
Continuing.

Breakpoint 1, 0x402000ec in __mmu_cache_flush ()
(gdb) disassemble strlen
Dump of assembler code for function strlen:
   0x40205c84 <+0>:	mov	r3, r0
   0x40205c88 <+4>:	mov	r2, r3
   0x40205c8c <+8>:	add	r3, r3, #1
   0x40205c90 <+12>:	ldrb	r1, [r2]
   0x40205c94 <+16>:	cmp	r1, #0
   0x40205c98 <+20>:	bne	0x40205c88 <strlen+4>
   0x40205c9c <+24>:	rsb	r0, r0, r2
   0x40205ca0 <+28>:	bx	lr
End of assembler dump.
(gdb) stepi
0x402000f0 in __mmu_cache_flush ()
(gdb) disassemble strlen
Dump of assembler code for function strlen:
   0x40205c84 <+0>:	beq	0x40205d18 <memcpy>
   0x40205c88 <+4>:			; <UNDEFINED> instruction: 0xf03fe03a
   0x40205c8c <+8>:	andeq	r0, r0, r6
   0x40205c90 <+12>:	andeq	r0, r0, r4
   0x40205c94 <+16>:	andeq	r0, r0, r7
   0x40205c98 <+20>:	andeq	r0, r0, sp, lsl r0
   0x40205c9c <+24>:	ldrdmi	r0, [r0], -r3
   0x40205ca0 <+28>:			; <UNDEFINED> instruction: 0xffffffff
End of assembler dump.
(gdb)

So, is this a qemu-linaro bug?

Thank you,

--
Carlo


From carlos_odonell at mentor.com  Thu May 17 14:25:18 2012
From: carlos_odonell at mentor.com (Carlos O'Donell)
Date: Thu, 17 May 2012 10:25:18 -0400
Subject: [arm-gnu] undefined instruction
In-Reply-To: <CAOQ7t2aZpf-J-HP+aLkQZxSuU4U9xNLX65gW-ZmLrMo9wBar4w@mail.gmail.com>
References: <CAOQ7t2ZxttqTjx8XOJu9=vOHOpAG1WO_6N47AU225LCP0fovoQ@mail.gmail.com>
	<4FB2E0E6.7090400@mentor.com>
	<CAOQ7t2aLGp1C-jd+EkiYSPrEtapuYcx153HSf1WM3=JegHso_w@mail.gmail.com>
	<4FB40899.7090705@mentor.com>
	<CAOQ7t2aZpf-J-HP+aLkQZxSuU4U9xNLX65gW-ZmLrMo9wBar4w@mail.gmail.com>
Message-ID: <4FB50A4E.8090607@mentor.com>

On 5/16/2012 5:27 PM, Carlo Caione wrote:
> Ok, I think I have spotted the bug. And actually the problem seems to
> me related to qemu-linaro (yes, I'm emulating a Cortex-A8 with qemu)
> and not to CodeSourcery.
> In fact in the code an exact point does exist in which the strlen
> routine gets corrupted.
...
> 0x402000f0 in __mmu_cache_flush ()
> (gdb) disassemble strlen
> Dump of assembler code for function strlen:
>    0x40205c84 <+0>:	beq	0x40205d18 <memcpy>
>    0x40205c88 <+4>:			; <UNDEFINED> instruction: 0xf03fe03a
>    0x40205c8c <+8>:	andeq	r0, r0, r6
>    0x40205c90 <+12>:	andeq	r0, r0, r4
>    0x40205c94 <+16>:	andeq	r0, r0, r7
>    0x40205c98 <+20>:	andeq	r0, r0, sp, lsl r0
>    0x40205c9c <+24>:	ldrdmi	r0, [r0], -r3
>    0x40205ca0 <+28>:			; <UNDEFINED> instruction: 0xffffffff
> End of assembler dump.
> (gdb)
> 
> So, is this a qemu-linaro bug?

It is most likely either a bug in the cache handling routines or a bug in QEMU's MMU emulation for the Cortex-A8.

I would suggest you test upstream QEMU to see if this is fixed.

If it is then you can then track down the fix and ask Linaro to backport.

If it's not fixed you need to put together a test case for upstream QEMU and report the issue there.

Cheers,
Carlos.
-- 
Carlos O'Donell
Mentor Graphics / CodeSourcery
carlos_odonell at mentor.com
carlos at codesourcery.com
+1 (613) 963 1026


From g4 at novadsp.com  Thu May 24 21:31:20 2012
From: g4 at novadsp.com (g4 at novadsp.com)
Date: Thu, 24 May 2012 22:31:20 +0100
Subject: [arm-gnu] GCC and NXP 1769 bare metal
Message-ID: <4FBEA8A8.1010300@novadsp.com>

Greetings.

Thanks for the tooling. Invaluable.

I'm trying to assemble the right spell for compiling and linking bare 
metal apps (FreeRTOS/lwip/FatFS) on the NXP 1769 using the arm-none-eabi 
compiler/linker.

Currently the main problem is 'C' runtimes. By way of example the 
compilation command line arguments are (eliding is mine):

arm-none-eabi-gcc.exe ./.../mmc.o -c -fmessage-length=0 -fno-builtin 
-ffunction-sections -fdata-sections -mcpu=cortex-m3 -mthumb -MMD -MP 
-Ir:/data/...

arm-none-eabi-ld -nostdlib -Map="rbmap" -Trb_debugld  
-or:/data.../rb.axf -L R:/apps/Mentor/lib/gcc/arm-none-eabi/4.6.1/thumb

The problem here is that I end up with a handsome (edited) listing of 
undefined references as shown below:

./obj/system_lpc17xx.o: In function `SystemCoreClockUpdate':
system_lpc17xx.c:(.text.SystemCoreClockUpdate+0xb4): undefined reference 
to `__aeabi_uldivmod'
system_lpc17xx.c:(.text.SystemCoreClockUpdate+0xd8): undefined reference 
to `__aeabi_uldivmod'
system_lpc17xx.c:(.text.SystemCoreClockUpdate+0x154): undefined 
reference to `__aeabi_uldivmod'
system_lpc17xx.c:(.text.SystemCoreClockUpdate+0x178): undefined 
reference to `__aeabi_uldivmod'
system_lpc17xx.c:(.text.SystemCoreClockUpdate+0x1ee): undefined 
reference to `__aeabi_uldivmod'
./obj/system_lpc17xx.o:system_lpc17xx.c:(.text.SystemCoreClockUpdate+0x212): 
more undefined references to `__aeabi_uldivmod' follow
./obj/main.o: In function `prvTaskStatsCommand':
main.c:(.text.prvTaskStatsCommand+0x1a): undefined reference to `strcpy'
main.c:(.text.prvTaskStatsCommand+0x20): undefined reference to `strlen'
./obj/main.o: In function `prvRunTimeStatsCommand':
main.c:(.text.prvRunTimeStatsCommand+0x1a): undefined reference to `strcpy'
main.c:(.text.prvRunTimeStatsCommand+0x20): undefined reference to `strlen'
./obj/main.o: In function `prvManuallyPlaceLargeDataInAHBRAM':
main.c:(.text.prvManuallyPlaceLargeDataInAHBRAM+0x7c): undefined 
reference to `memset'
./obj/spi-interface-to-sd-card.o: In function `prvWriteFilesToDisk':
spi-interface-to-sd-card.c:(.text.prvWriteFilesToDisk+0x8c): undefined 
reference to `memset'
./obj/spi-interface-to-sd-card.o: In function `prvReadBackCreatedFiles':
spi-interface-to-sd-card.c:(.text.prvReadBackCreatedFiles+0x94): 
undefined reference to `memset'
.//obj/queue.o: In function `prvCopyDataToQueue':
api_msg.c:(.text.err_tcp+0x3a): undefined reference to `fflush'
api_msg.c:(.text.err_tcp+0x3e): undefined reference to `abort'
queue.c:(.text.prvCopyDataToQueue+0x44): undefined reference to `memcpy'
queue.c:(.text.prvCopyDataToQueue+0x7a): undefined reference to `memcpy'

My first questions are these:

1. Which runtime libraries should be referenced when linking?
2. Linker scripts are a bit of a mystery - the NXP has 512K contiguous 
flash and 64K RAM in 2 possibly contiguous banks. Are there pro-forma 
scripts I can use?

Any help greatly appreciated.

ATB

Jerry




From jamesk at edmi.com.au  Fri May 25 03:40:36 2012
From: jamesk at edmi.com.au (James Kehl)
Date: Fri, 25 May 2012 03:40:36 +0000
Subject: [arm-gnu] GCC and NXP 1769 bare metal
In-Reply-To: <4FBEA8A8.1010300@novadsp.com>
References: <4FBEA8A8.1010300@novadsp.com>
Message-ID: <F8805965C237CB48B563CD31BD5762E83D7D68F5@Eclipse.edmi.local>

1) You're using -nostdlib. Why? That's why standard libraries (like libgcc for __aeabi_uldivmod, or libc/newlib for strlen) aren't being pulled in. 
Note you may need to add some OS wrapper functions (like sbrk) for newlib, later - if FreeRTOS doesn't have those already.

2) Using ld directly, instead of using gcc for the final linking step, means you get to deal with all the joys of selecting the right multilib, etc. etc. Using the gcc driver is generally easier.

3) Sure FreeRTOS doesn't already have a linker file for your platform? Google suggests:
http://websvn.hylands.org/filedetails.php?repname=Projects&path=%2Ffree-rtos%2Flpc17xx%2Flpc17xx.ld&rev=318&sc=0

> -----Original Message-----
> From: arm-gnu-bounces at codesourcery.com [mailto:arm-gnu-
> bounces at codesourcery.com] On Behalf Of g4 at novadsp.com
> Sent: Friday, 25 May 2012 7:31 AM
> To: arm-gnu at codesourcery.com
> Subject: [arm-gnu] GCC and NXP 1769 bare metal
> 
> Greetings.
> 
> Thanks for the tooling. Invaluable.
> 
> I'm trying to assemble the right spell for compiling and linking bare
> metal apps (FreeRTOS/lwip/FatFS) on the NXP 1769 using the arm-none-eabi
> compiler/linker.
> 
> Currently the main problem is 'C' runtimes. By way of example the
> compilation command line arguments are (eliding is mine):
> 
> arm-none-eabi-gcc.exe ./.../mmc.o -c -fmessage-length=0 -fno-builtin
> -ffunction-sections -fdata-sections -mcpu=cortex-m3 -mthumb -MMD -MP
> -Ir:/data/...
> 
> arm-none-eabi-ld -nostdlib -Map="rbmap" -Trb_debugld
> -or:/data.../rb.axf -L R:/apps/Mentor/lib/gcc/arm-none-eabi/4.6.1/thumb
> 
> The problem here is that I end up with a handsome (edited) listing of
> undefined references as shown below:
> 
> ./obj/system_lpc17xx.o: In function `SystemCoreClockUpdate':
> system_lpc17xx.c:(.text.SystemCoreClockUpdate+0xb4): undefined
> reference
> to `__aeabi_uldivmod'
> system_lpc17xx.c:(.text.SystemCoreClockUpdate+0xd8): undefined
> reference
> to `__aeabi_uldivmod'
> system_lpc17xx.c:(.text.SystemCoreClockUpdate+0x154): undefined
> reference to `__aeabi_uldivmod'
> system_lpc17xx.c:(.text.SystemCoreClockUpdate+0x178): undefined
> reference to `__aeabi_uldivmod'
> system_lpc17xx.c:(.text.SystemCoreClockUpdate+0x1ee): undefined
> reference to `__aeabi_uldivmod'
> ./obj/system_lpc17xx.o:system_lpc17xx.c:(.text.SystemCoreClockUpdate+0
> x212):
> more undefined references to `__aeabi_uldivmod' follow
> ./obj/main.o: In function `prvTaskStatsCommand':
> main.c:(.text.prvTaskStatsCommand+0x1a): undefined reference to `strcpy'
> main.c:(.text.prvTaskStatsCommand+0x20): undefined reference to `strlen'
> ./obj/main.o: In function `prvRunTimeStatsCommand':
> main.c:(.text.prvRunTimeStatsCommand+0x1a): undefined reference to
> `strcpy'
> main.c:(.text.prvRunTimeStatsCommand+0x20): undefined reference to
> `strlen'
> ./obj/main.o: In function `prvManuallyPlaceLargeDataInAHBRAM':
> main.c:(.text.prvManuallyPlaceLargeDataInAHBRAM+0x7c): undefined
> reference to `memset'
> ./obj/spi-interface-to-sd-card.o: In function `prvWriteFilesToDisk':
> spi-interface-to-sd-card.c:(.text.prvWriteFilesToDisk+0x8c): undefined
> reference to `memset'
> ./obj/spi-interface-to-sd-card.o: In function `prvReadBackCreatedFiles':
> spi-interface-to-sd-card.c:(.text.prvReadBackCreatedFiles+0x94):
> undefined reference to `memset'
> .//obj/queue.o: In function `prvCopyDataToQueue':
> api_msg.c:(.text.err_tcp+0x3a): undefined reference to `fflush'
> api_msg.c:(.text.err_tcp+0x3e): undefined reference to `abort'
> queue.c:(.text.prvCopyDataToQueue+0x44): undefined reference to
> `memcpy'
> queue.c:(.text.prvCopyDataToQueue+0x7a): undefined reference to
> `memcpy'
> 
> My first questions are these:
> 
> 1. Which runtime libraries should be referenced when linking?
> 2. Linker scripts are a bit of a mystery - the NXP has 512K contiguous
> flash and 64K RAM in 2 possibly contiguous banks. Are there pro-forma
> scripts I can use?
> 
> Any help greatly appreciated.
> 
> ATB
> 
> Jerry
> 
> 
> _______________________________________________
> arm-gnu mailing list
> arm-gnu at codesourcery.com
> http://sourcerytools.com/cgi-bin/mailman/listinfo/arm-gnu


From freddie_chopin at op.pl  Fri May 25 15:13:16 2012
From: freddie_chopin at op.pl (Freddie Chopin)
Date: Fri, 25 May 2012 17:13:16 +0200
Subject: [arm-gnu] GCC and NXP 1769 bare metal
In-Reply-To: <4FBEA8A8.1010300@novadsp.com>
References: <4FBEA8A8.1010300@novadsp.com>
Message-ID: <4FBFA18C.8060309@op.pl>

W dniu 2012-05-24 23:31, g4 at novadsp.com pisze:
> My first questions are these:
>
> 1. Which runtime libraries should be referenced when linking?
> 2. Linker scripts are a bit of a mystery - the NXP has 512K contiguous
> flash and 64K RAM in 2 possibly contiguous banks. Are there pro-forma
> scripts I can use?

You can try my template projects which have all the necessary files - 
Makefile, startup, linker script, vector table, ...

http://www.freddiechopin.info/en/download/category/6-przyklady

Adding FreeRTOS to them is just putting the files in some directory and 
adding the dirs in Makefile. You will usually need to add 3 definitions 
that assign FreeRTOS system handlers to correct vectors in the vector table.

In case of problems - just ask, I've done that recently.

4\/3!!


From g4 at novadsp.com  Fri May 25 15:29:41 2012
From: g4 at novadsp.com (g4 at novadsp.com)
Date: Fri, 25 May 2012 16:29:41 +0100
Subject: [arm-gnu] GCC and NXP 1769 bare metal
In-Reply-To: <4FBFA18C.8060309@op.pl>
References: <4FBEA8A8.1010300@novadsp.com> <4FBFA18C.8060309@op.pl>
Message-ID: <4FBFA565.3030501@novadsp.com>

Hello Freddie,

Grand. Many thanks.

On 25/05/2012 16:13, Freddie Chopin wrote:
> W dniu 2012-05-24 23:31, g4 at novadsp.com pisze:
>> My first questions are these:
>>
>> 1. Which runtime libraries should be referenced when linking?
>> 2. Linker scripts are a bit of a mystery - the NXP has 512K contiguous
>> flash and 64K RAM in 2 possibly contiguous banks. Are there pro-forma
>> scripts I can use?
>
> You can try my template projects which have all the necessary files -
> Makefile, startup, linker script, vector table, ...
>
> http://www.freddiechopin.info/en/download/category/6-przyklady
>
> Adding FreeRTOS to them is just putting the files in some directory and
> adding the dirs in Makefile. You will usually need to add 3 definitions
> that assign FreeRTOS system handlers to correct vectors in the vector
> table.
>
> In case of problems - just ask, I've done that recently.
>
> 4\/3!!
> _______________________________________________
> arm-gnu mailing list
> arm-gnu at codesourcery.com
> http://sourcerytools.com/cgi-bin/mailman/listinfo/arm-gnu
>




From kalpesh.bafna at einfochips.com  Tue May 29 07:28:14 2012
From: kalpesh.bafna at einfochips.com (Kalpesh Bafna)
Date: Tue, 29 May 2012 12:58:14 +0530
Subject: [arm-gnu] Application crashes in vsprintf . build with
	arm-none-linux-gnueabi
Message-ID: <008301cd3d6c$9c3a9660$f2026e0a@EIC.COM>

Hi 

I have a crash problem with my application running on ARM-11.

I am using arm-none-linux-gnueabi-gcc (Sourcery G++ Lite 2009q1-203) 4.3.3 compiler to build the application.

When i build the application with -O2 optimization it crashes in vsprintf.

But when i build with -O1 it does not crashes.

Does anybody have idea on this?

It's urgent. Please help.

Regards
Kalpesh Bafna
Enginner (Embedded)
eInfochips Limited
Pune
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sourcerytools.com/pipermail/arm-gnu/attachments/20120529/1fa2f022/attachment.html>

From list-bastian.schick at sciopta.com  Tue May 29 16:02:52 2012
From: list-bastian.schick at sciopta.com (42Bastian)
Date: Tue, 29 May 2012 18:02:52 +0200
Subject: [arm-gnu] Application crashes in vsprintf . build with
	arm-none-linux-gnueabi
In-Reply-To: <008301cd3d6c$9c3a9660$f2026e0a@EIC.COM>
References: <008301cd3d6c$9c3a9660$f2026e0a@EIC.COM>
Message-ID: <4FC4F32C.10001@sciopta.com>

Hi

> I have a crash problem with my application running on ARM-11.

"crash" is a imprecise term to get any help.

Explain more what kind of crash. Even better, debug into vprintf to find
the place where it "crashes".

-- 
42Bastian
+
| http://www.sciopta.com
| Fastest direct message passing kernel.
| IEC61508 certified.
+


From ricardo_anguiano at mentor.com  Tue May 29 18:24:46 2012
From: ricardo_anguiano at mentor.com (Ricardo Anguiano)
Date: Tue, 29 May 2012 11:24:46 -0700
Subject: [arm-gnu] Application crashes in vsprintf . build with
	arm-none-linux-gnueabi
In-Reply-To: <008301cd3d6c$9c3a9660$f2026e0a@EIC.COM> (Kalpesh Bafna's message
	of "Tue, 29 May 2012 12:58:14 +0530")
References: <008301cd3d6c$9c3a9660$f2026e0a@EIC.COM>
Message-ID: <yxj8pq9mj15d.fsf@mentor.com>

Kalpesh Bafna <kalpesh.bafna at einfochips.com> writes:
> I am using arm-none-linux-gnueabi-gcc (Sourcery G++ Lite 2009q1-203)
> 4.3.3 compiler to build the application.

Sourcerh G++ was rebranded to Sourcery CodeBench.  The most recent
Sourcery CodeBench Lite ARM GNU/Linux release (2011.09-70) is two years
newer than what you are using.  You can register and download the newest
version here:

http://www.mentor.com/embedded-software/sourcery-tools/sourcery-codebench/editions/lite-edition/

> When i build the application with -O2 optimization it crashes in
> vsprintf.  But when i build with -O1 it does not crashes.  Does
> anybody have idea on this?

Sourcery CodeBench Lite is unsupported but someone on this list maybe
able to help if you submit a more detailed bug report.  You may find
these pages helpful in generating such a report to this list.

https://sourcery.mentor.com/GNUToolchain/kbentry216
http://gcc.gnu.org/bugs/

Thanks,
-- 
Ricardo Anguiano
Mentor Graphics
+1-510-354-6774


From krzysztof.wesolowski at rainlabs.pl  Tue May 29 18:40:15 2012
From: krzysztof.wesolowski at rainlabs.pl (=?ISO-8859-2?Q?Krzysztof_Weso=B3owski?=)
Date: Tue, 29 May 2012 20:40:15 +0200
Subject: [arm-gnu] Application crashes in vsprintf . build with
	arm-none-linux-gnueabi
In-Reply-To: <yxj8pq9mj15d.fsf@mentor.com>
References: <008301cd3d6c$9c3a9660$f2026e0a@EIC.COM>
	<yxj8pq9mj15d.fsf@mentor.com>
Message-ID: <CAPrKjpbeREk-V01xXs5ttcm5x-XCiKhQv-P6i4uG9OtwmP2UBA@mail.gmail.com>

Reporting bugs without event checking what "crash" means? Usually
compiler bug is not the first thing too look at ;)

First check if nothing more obvious happens, for example different
optimization levels use different amount of stack space.

Regards,
Krzysztof Weso?owski,
tel. +48 721 337 238


On Tue, May 29, 2012 at 8:24 PM, Ricardo Anguiano
<ricardo_anguiano at mentor.com> wrote:
> Kalpesh Bafna <kalpesh.bafna at einfochips.com> writes:
>> I am using arm-none-linux-gnueabi-gcc (Sourcery G++ Lite 2009q1-203)
>> 4.3.3 compiler to build the application.
>
> Sourcerh G++ was rebranded to Sourcery CodeBench. ?The most recent
> Sourcery CodeBench Lite ARM GNU/Linux release (2011.09-70) is two years
> newer than what you are using. ?You can register and download the newest
> version here:
>
> http://www.mentor.com/embedded-software/sourcery-tools/sourcery-codebench/editions/lite-edition/
>
>> When i build the application with -O2 optimization it crashes in
>> vsprintf. ?But when i build with -O1 it does not crashes. ?Does
>> anybody have idea on this?
>
> Sourcery CodeBench Lite is unsupported but someone on this list maybe
> able to help if you submit a more detailed bug report. ?You may find
> these pages helpful in generating such a report to this list.
>
> https://sourcery.mentor.com/GNUToolchain/kbentry216
> http://gcc.gnu.org/bugs/
>
> Thanks,
> --
> Ricardo Anguiano
> Mentor Graphics
> +1-510-354-6774
> _______________________________________________
> arm-gnu mailing list
> arm-gnu at codesourcery.com
> http://sourcerytools.com/cgi-bin/mailman/listinfo/arm-gnu