scb9328: Add ARM relocation support

PATCH New Torsten Koschorrek 2 participants 6 messages Jun 9, 2011 – Jul 4, 2011 1 message not in the archive

lore

Share & tools

lore

Torsten Koschorrek

author

2011-06-09 11:36 UTC #0 lore raw

PATCH 1/1 scb9328: Add ARM relocation support

This patch fixed compiler errors due to missing definitions of CONFIG_SYS_SDRAM_BASE and CONFIG_SYS_INIT_SP_ADDR.

The board doesn't start, though. A v2 of this patch or a seperate patch will fix the error when it was found.

Signed-off-by Torsten Koschorrek <[email protected]>

---
  board/scb9328/config.mk   |    2 +-
  board/scb9328/scb9328.c   |   10 +++++++++-
  include/configs/scb9328.h |    3 +++
  3 files changed, 13 insertions(+), 2 deletions(-)

board/scb9328/config.mk +1 -0

index 7c5e067..6e411de 100644
@@ -7,4 +7,4 @@
  # the code in RAM device only.
  #
  -CONFIG_SYS_TEXT_BASE = 0x08f00000
+CONFIG_SYS_TEXT_BASE = 0x10000000

board/scb9328/scb9328.c +9 -1

index 428e8c9..631f4e3 100644
@@ -39,6 +39,15 @@ int board_init (void)
   int dram_init (void)
  {
+	/* dram_init must store complete ramsize in gd->ram_size */
+	gd->ram_size = get_ram_size((volatile void *)SCB9328_SDRAM_1,
+				    SCB9328_SDRAM_1_SIZE);
+
+	return 0;
+}
+
+void dram_init_banksize (void)
+{
  #if ( CONFIG_NR_DRAM_BANKS > 0 )
  	gd->bd->bi_dram[0].start = SCB9328_SDRAM_1;
  	gd->bd->bi_dram[0].size = SCB9328_SDRAM_1_SIZE;

@@ -55,7 +64,6 @@ int dram_init (void)
  	gd->bd->bi_dram[3].start = SCB9328_SDRAM_4;
  	gd->bd->bi_dram[3].size = SCB9328_SDRAM_4_SIZE;
  #endif
-	return 0;
  }
   /**

include/configs/scb9328.h +2 -0

index 3da214e..c610ed1 100644
@@ -127,6 +127,9 @@
  #define SCB9328_SDRAM_1		0x08000000	/* SDRAM bank #1	   */
  #define SCB9328_SDRAM_1_SIZE	0x01000000	/* 16 MB		   */
  +#define CONFIG_SYS_SDRAM_BASE	SCB9328_SDRAM_1
+#define CONFIG_SYS_INIT_SP_ADDR	SCB9328_SDRAM_1 + 0xf00000
+
  /*
   * Configuration for FLASH memory for the Synertronixx board
   */

Signature · 1 line

1.7.2.5

Stefano Babic

2011-06-09 14:34 UTC #1 lore raw

PATCH 1/1 scb9328: Add ARM relocation support

Quoted 18 lines · Torsten Koschorrek

This patch fixed compiler errors due to missing definitions of CONFIG_SYS_SDRAM_BASE and CONFIG_SYS_INIT_SP_ADDR.
The board doesn't start, though. A v2 of this patch or a seperate patch will fix the error when it was found.
Signed-off-by Torsten Koschorrek <[email protected]>
---
 board/scb9328/config.mk   |    2 +-
 board/scb9328/scb9328.c   |   10 +++++++++-
 include/configs/scb9328.h |    3 +++
 3 files changed, 13 insertions(+), 2 deletions(-)
board/scb9328/config.mk +0 -0
index 7c5e067..6e411de 100644

board/scb9328/config.mk line 10

Quoted 4 lines

board/scb9328/config.mk +1 -0

@@ -7,4 +7,4 @@
 # the code in RAM device only.
 #
 -CONFIG_SYS_TEXT_BASE = 0x08f00000
+CONFIG_SYS_TEXT_BASE = 0x10000000

Try to get rid of config.mk, too. It is obsolete, and you can move CONFIG_SYS_TEXT_BASE inside scb9328.h.

board/scb9328/scb9328.c int board_init (void) · line 45

Quoted 12 lines

board/scb9328/scb9328.c +5 -0

index 428e8c9..631f4e3 100644
@@ -39,6 +39,15 @@ int board_init (void)
  int dram_init (void)
 {
+    /* dram_init must store complete ramsize in gd->ram_size */
+    gd->ram_size = get_ram_size((volatile void *)SCB9328_SDRAM_1,
+                    SCB9328_SDRAM_1_SIZE);
+
+    return 0;

Not understood. If you have more as one bank, the size is the sum of all banks. In dram_init_banksize() you setup more as one bank if CONFIG_NR_DRAM_BANKS > 1. Then you should set in dram_init:

size1 = get_ram_size((volatile void *)SCB9328_SDRAM_1,

	SCB9328_SDRAM_1_SIZE);
#if ( CONFIG_NR_DRAM_BANKS > 1 )
size2 = get_ram_size((volatile void *)SCB9328_SDRAM_2,
	SCB9328_SDRAM_2_SIZE);

.....

and then:

	gd->ram_size = size1 + size2 + size3 + size4;

include/configs/scb9328.h line 130

Quoted 10 lines

board/scb9328/scb9328.c +0 -0

/**

include/configs/scb9328.h +1 -0

index 3da214e..c610ed1 100644
@@ -127,6 +127,9 @@
 #define SCB9328_SDRAM_1        0x08000000    /* SDRAM bank #1       */
 #define SCB9328_SDRAM_1_SIZE    0x01000000    /* 16 MB           */
 +#define CONFIG_SYS_SDRAM_BASE    SCB9328_SDRAM_1
+#define CONFIG_SYS_INIT_SP_ADDR    SCB9328_SDRAM_1 + 0xf00000

In most boards/processors the stack pointer is set to the internal RAM, that is available before the SDRAM controller is activated. When not available, cache can be used. Is there no internal memory available for this processor ? It could be that the stack is used before the SDRAM is initialized, and then it hangs.

Best regards,
Stefano Babic

Signature · 5 lines

=====================================================================
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-0 Fax: +49-8142-66989-80  Email: office at denx.de
=====================================================================

A message that is not in the archive (sent off-list or not yet received): [email protected]

Stefano Babic

2011-06-10 07:24 UTC #3 lore raw

PATCH 1/1 scb9328: Add ARM relocation support

Quoted 1 line · Torsten Koschorrek

Hello,

Hi Torsten,

send your answer to the ML, too. Someone else can help you ;-)

Quoted 11 lines

Quoted 10 lines

size1 = get_ram_size((volatile void *)SCB9328_SDRAM_1,

    SCB9328_SDRAM_1_SIZE);
#if ( CONFIG_NR_DRAM_BANKS>  1 )
size2 = get_ram_size((volatile void *)SCB9328_SDRAM_2,
    SCB9328_SDRAM_2_SIZE);

.....

and then:

    gd->ram_size = size1 + size2 + size3 + size4;

Quoted 5 lines

Yes, I thought about it. The thing is, we only have one bank on the system. So, dram_init_banksize() needs a cleanup, but that's a problem for another cleanup-patch, which will be committed in a next step.
(If it's needed I could do a minor cleanup first...)

You decide. However, the code in the patch is wrong. If you have only one bank, you could directly simplify your code, I think.

Quoted 19 lines

Quoted 18 lines
include/configs/scb9328.h line 130
Quoted 11 lines
   /**
include/configs/scb9328.h +1 -0
index 3da214e..c610ed1 100644
@@ -127,6 +127,9 @@
  #define SCB9328_SDRAM_1        0x08000000    /* SDRAM bank #1       */
  #define SCB9328_SDRAM_1_SIZE    0x01000000    /* 16 MB           */
  +#define CONFIG_SYS_SDRAM_BASE    SCB9328_SDRAM_1
+#define CONFIG_SYS_INIT_SP_ADDR    SCB9328_SDRAM_1 + 0xf00000
In most boards/processors the stack pointer is set to the internal RAM, that is available before the SDRAM controller is activated. When not available, cache can be used. Is there no internal memory available for this processor ? It could be that the stack is used before the SDRAM is initialized, and then it hangs.

Quoted 2 lines

I suppose, here is the cause of the hanging problem, yes. While I was trying to solve it I saw some board configs which use sdram, too.

Take into account that a lot of new processors set automatically the RAM controller before u-boot is started. I am thinking to the newer i.MX processors, but this is true for other architectures, too. And in some cases u-boot runs as second or third bootloader, such as for at91sam, davinci and omap, without touching the RAM controller that is already programmed by the first-stage bootloader. In your case the SDRAM controller is probably not yet programmed when the stack pointer is set.

Quoted 5 lines

Internal RAM came to my mind, too, but I didn't had the time to investigate further.
Unfortunately I have to work on another project today and next week and I think I'm not able to solve the hanging problem.

Understood, I tried only to give you some hints where to check ;-)

Quoted 2 lines

Minor fixes (such as
config.mk) for the above patch should be possible, though.

Ok, agree. Fix first the problem to make MAKEALL happy and build the board again.

Best regards,
Stefano Babic

Signature · 5 lines

=====================================================================
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-0 Fax: +49-8142-66989-80  Email: office at denx.de
=====================================================================

Torsten Koschorrek

author

2011-06-10 17:55 UTC #4 lore raw

PATCH 1/1 scb9328: Add ARM relocation support

Hello,

Quoted 5 lines · Stefano Babic

On 06/10/2011 07:44 AM, Torsten Koschorrek wrote:
Hi Torsten,
send your answer to the ML, too. Someone else can help you ;-)

Oh, yes, right. This little 'Reply All' Button, sorry :-)

Quoted 11 lines

Quoted 11 lines
Quoted 10 lines
size1 = get_ram_size((volatile void *)SCB9328_SDRAM_1,
     SCB9328_SDRAM_1_SIZE);
#if ( CONFIG_NR_DRAM_BANKS>   1 )
size2 = get_ram_size((volatile void *)SCB9328_SDRAM_2,
     SCB9328_SDRAM_2_SIZE);
.....
and then:
     gd->ram_size = size1 + size2 + size3 + size4;

Quoted 9 lines

Quoted 5 lines
Yes, I thought about it. The thing is, we only have one bank on the system. So, dram_init_banksize() needs a cleanup, but that's a problem for another cleanup-patch, which will be committed in a next step.
(If it's needed I could do a minor cleanup first...)
You decide. However, the code in the patch is wrong. If you have only one bank, you could directly simplify your code, I think.

Good point.

Quoted 6 lines

Quoted 3 lines
Unfortunately I have to work on another project today and next week and I think I'm not able to solve the hanging problem.
Understood, I tried only to give you some hints where to check ;-)

... and I appreciate that very much :-) Above all, your answers showed me, that I was looking in the right direction so far.

Quoted 6 lines

Quoted 2 lines
Minor fixes (such as
config.mk) for the above patch should be possible, though.
Ok, agree. Fix first the problem to make MAKEALL happy and build the board again.

OK, I just tested it again, MAKEALL is happy.

'include/configs/scb9328.h' is the only file that has to be patched with the patch already send to the ml. 'board/scb9328/config.mk' and 'board/scb9328/scb9328.c' definately need a cleanup, but compilation is ok. The cleanup of those two files 'll be done next week. And hopefully I find some time next week to work on the hangup problem, too.

Quoted 3 lines

Best regards,
Stefano Babic

Thanks
Torsten

Torsten Koschorrek

author

2011-07-04 09:34 UTC #5 lore raw

PATCH 1/1 scb9328: Add ARM relocation support

Quoted 51 lines · Torsten Koschorrek

Hello,
Quoted 5 lines · Stefano Babic
On 06/10/2011 07:44 AM, Torsten Koschorrek wrote:
Hi Torsten,
send your answer to the ML, too. Someone else can help you ;-)
Oh, yes, right. This little 'Reply All' Button, sorry :-)
Quoted 11 lines
Quoted 11 lines
Quoted 10 lines
size1 = get_ram_size((volatile void *)SCB9328_SDRAM_1,
      SCB9328_SDRAM_1_SIZE);
#if ( CONFIG_NR_DRAM_BANKS>    1 )
size2 = get_ram_size((volatile void *)SCB9328_SDRAM_2,
      SCB9328_SDRAM_2_SIZE);
.....
and then:
      gd->ram_size = size1 + size2 + size3 + size4;
Quoted 9 lines
Quoted 5 lines
Yes, I thought about it. The thing is, we only have one bank on the system. So, dram_init_banksize() needs a cleanup, but that's a problem for another cleanup-patch, which will be committed in a next step.
(If it's needed I could do a minor cleanup first...)
You decide. However, the code in the patch is wrong. If you have only one bank, you could directly simplify your code, I think.
Good point.
Quoted 6 lines
Quoted 3 lines
Unfortunately I have to work on another project today and next week and I think I'm not able to solve the hanging problem.
Understood, I tried only to give you some hints where to check ;-)
... and I appreciate that very much :-) Above all, your answers showed me, that I was looking in the right direction so far.
Quoted 6 lines
Quoted 2 lines
Minor fixes (such as
config.mk) for the above patch should be possible, though.
Ok, agree. Fix first the problem to make MAKEALL happy and build the board again.

Quoted 8 lines

OK, I just tested it again, MAKEALL is happy.
'include/configs/scb9328.h' is the only file that has to be patched with the patch already send to the ml. 'board/scb9328/config.mk' and 'board/scb9328/scb9328.c' definately need a cleanup, but compilation is ok. The cleanup of those two files 'll be done next week. And hopefully I find some time next week to work on the hangup problem, too.

What's the status of this patch? Do I need to resend a modified version without config.mk and scb9328.c? Or I could do a small cleanup as discussed in this thread. But as I said, MAKEALL is happy with the modified scb9328.h

thanks
Torsten

Stefano Babic

2011-07-04 09:43 UTC #6 lore raw

PATCH 1/1 scb9328: Add ARM relocation support

On 07/04/2011 11:34 AM, Torsten Koschorrek wrote:

Hi Torsten,

Quoted 4 lines

What's the status of this patch? Do I need to resend a modified version without config.mk and scb9328.c? Or I could do a small cleanup as discussed in this thread. But as I said, MAKEALL is happy with the modified scb9328.h

The patch is set to "Changed requested", that means, you have to fix the issues we discuss in the thread. It will not be merged as it is.

You can alwasy check the status of the patches you submit on patchworks.ozlabs.org.

Best regards,
Stefano Babic

Signature · 5 lines

=====================================================================
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-0 Fax: +49-8142-66989-80  Email: office at denx.de
=====================================================================