devkit8000: Move CONFIG_SYS_TEXT_BASE out of bss

This moves CONFIG_SYS_TEXT_BASE one MB after beginning of SD-RAM. Move CONFIG_SYS_SPL_MALLOC_START to have one MB of free space for the u-boot image.

CONFIG_SYS_TEXT_BASE was in the middle of the bss-section. This was the reason for the problems with MMC boot described here:
http://article.gmane.org/gmane.comp.boot-loaders.u-boot/118711

Signed-off-by Simon Schwarz <[email protected]>

Cc s-paulraj at ti.com

Cc wd at denx.de

Cc trini at ti.com

---
?include/configs/devkit8000.h | ? ?5 +++--
?1 files changed, 3 insertions(+), 2 deletions(-)

include/configs/devkit8000.h +0 -0

index e1743dc..b29d564 100644
@@ -37,7 +37,8 @@

?#define CONFIG_OMAP3430 ? ? ? ? ? ? ? ?1 ? ? ? /* which is in a 3430 */ ?#define CONFIG_OMAP3_DEVKIT8000 ? ? ? ?1 ? ? ? /* working with DevKit8000 */

include/configs/devkit8000.h +2 -1

-#define ? ? ? ?CONFIG_SYS_TEXT_BASE ? ?0x80008000
+/* 1 MB into SD-RAM to leave place for .bss-section */
+#define ? ? ? ?CONFIG_SYS_TEXT_BASE ? ?0x80100000

?#define CONFIG_SDRC ? ?/* The chip has SDRC controller */

include/configs/devkit8000.h +0 -0

@@ -359,7 +360,7 @@

?#define CONFIG_SYS_NAND_U_BOOT_OFFS ? ?0x80000
?#define CONFIG_SYS_NAND_U_BOOT_SIZE ? ?0x200000

include/configs/devkit8000.h +1 -1

-#define CONFIG_SYS_SPL_MALLOC_START ? ?0x80108000
+#define CONFIG_SYS_SPL_MALLOC_START ? ?0x80200000

?#define CONFIG_SYS_SPL_MALLOC_SIZE ? ? 0x100000 ? ? ? ?/* 1 MB */

?#endif /* __CONFIG_H */
--
1.7.4.1

This moves CONFIG_SYS_TEXT_BASE one MB after beginning of SD-RAM. Move CONFIG_SYS_SPL_MALLOC_START to have one MB of free space for the u-boot image.

CONFIG_SYS_TEXT_BASE was in the middle of the bss-section. This was the reason for the problems with MMC boot described here:

This moves CONFIG_SYS_TEXT_BASE one MB after beginning of SD-RAM. Move CONFIG_SYS_SPL_MALLOC_START to have one MB of free space for the u-boot image.

CONFIG_SYS_TEXT_BASE was in the middle of the bss-section. This was the reason for the problems with MMC boot described here:
http://article.gmane.org/gmane.comp.boot-loaders.u-boot/118711

Hi Simon,

Quoted 6 lines · Simon Schwarz

This moves CONFIG_SYS_TEXT_BASE one MB after beginning of SD-RAM. Move CONFIG_SYS_SPL_MALLOC_START to have one MB of free space for the u-boot image.

CONFIG_SYS_TEXT_BASE was in the middle of the bss-section. This was the reason for the problems with MMC boot described here:
http://article.gmane.org/gmane.comp.boot-loaders.u-boot/118711

Does this affect only boards having the SPL support? Can you, please, shed some more light on this?

--
Regards,
Igor.

Hi Igor,

this only affects Boards with SPL support - and only devkit8000.

The SPL has its .bss section in SD-RAM while the rest is in SRAM - CONFIG_SYS_TEXT_BASE was in the same area as .bss. This caused trouble as the header of the u-boot.img was written before CONFIG_SYS_TEXT_BASE. In the described case this messed with the pointer to the mcc datastructure - therefore the u-boot.img file failed to be read.

Quoted 10 lines · Simon Schwarz

Hi Igor,

this only affects Boards with SPL support - and only devkit8000.

The SPL has its .bss section in SD-RAM while the rest is in SRAM - CONFIG_SYS_TEXT_BASE was in the same area as .bss. This caused trouble as the header of the u-boot.img was written before CONFIG_SYS_TEXT_BASE. In the described case this messed with the pointer to the mcc datastructure - therefore the u-boot.img file failed to be read.

Thank you very much for the explanation.

So as to my understanding, what happens is:
1) The SPL initializes the SDRAM
2) Copies its .bss section to SDRAM
3) Starts loading the U-Boot image and overwrites the .bss section Is the above correct?

If it is, then why do we need the .bss in SDRAM?

Aren't there enough space in SRAM?

Also, is this what X-Loader did in first place?

Hi Igor,

Quoted 14 lines · Igor Grinberg

Quoted 10 lines · Simon Schwarz

Hi Igor,

this only affects Boards with SPL support - and only devkit8000.

The SPL has its .bss section in SD-RAM while the rest is in SRAM - CONFIG_SYS_TEXT_BASE was in the same area as .bss. This caused trouble as the header of the u-boot.img was written before CONFIG_SYS_TEXT_BASE. In the described case this messed with the pointer to the mcc datastructure - therefore the u-boot.img file failed to be read.

Thank you very much for the explanation.

Quoted 7 lines

So as to my understanding, what happens is:
1) The SPL initializes the SDRAM
2) Copies its .bss section to SDRAM
3) Starts loading the U-Boot image and overwrites the .bss section Is the above correct?

If it is, then why do we need the .bss in SDRAM?

U-Boot FAT implementation uses 3 buffers each of size 64 KB.

Quoted 1 line

Aren't there enough space in SRAM?

SRAM doesn't have space for such huge buffers.

Quoted 1 line

Also, is this what X-Loader did in first place?

Yes, x-loader had these buffers in SDRAM too. But instead of putting the .bss in SDRAM, x-loader made these huge arrays into pointer variables that were then made to point to the SDRAM.

This moves CONFIG_SYS_TEXT_BASE one MB after beginning of SD-RAM. Move CONFIG_SYS_SPL_MALLOC_START to have one MB of free space for the u-boot image.

CONFIG_SYS_TEXT_BASE was in the middle of the bss-section. This was the reason for the problems with MMC boot described here:
http://article.gmane.org/gmane.comp.boot-loaders.u-boot/118711

Signed-off-by Simon Schwarz <[email protected]>

Cc s-paulraj at ti.com

Cc wd at denx.de

Cc trini at ti.com

---
 include/configs/devkit8000.h |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

include/configs/devkit8000.h +2 -1

index e1743dc..b29d564 100644
@@ -37,7 +37,8 @@
 #define CONFIG_OMAP3430		1	/* which is in a 3430 */
 #define CONFIG_OMAP3_DEVKIT8000	1	/* working with DevKit8000 */
 
-#define	CONFIG_SYS_TEXT_BASE	0x80008000
+/* 1 MB into SD-RAM to leave place for .bss-section */
+#define	CONFIG_SYS_TEXT_BASE	0x80100000
 
 #define CONFIG_SDRC	/* The chip has SDRC controller */
 

This moves CONFIG_SYS_TEXT_BASE one MB after beginning of SD-RAM. Move CONFIG_SYS_SPL_MALLOC_START to have one MB of free space for the u-boot image.

CONFIG_SYS_TEXT_BASE was in the middle of the bss-section. This was the reason for the problems with MMC boot described here:
http://article.gmane.org/gmane.comp.boot-loaders.u-boot/118711

Signed-off-by Simon Schwarz <[email protected]>

---
V2 changes:
CHG added a more comprehensive comment on CONFIG_SYS_TEXT_BASE CHG Moved CONFIG_SYS_SPL_MALLOC_START from 0x80200000 to 0x80208000 to be
consistent with beagle board

Cc s-paulraj at ti.com

Cc trini at ti.com

Cc albert.u.boot at aribaud.net

---
 include/configs/devkit8000.h |   10 ++++++++--
 1 files changed, 8 insertions(+), 2 deletions(-)

include/configs/devkit8000.h +7 -1

index e1743dc..a2e1d2d 100644
@@ -37,7 +37,13 @@
 #define CONFIG_OMAP3430		1	/* which is in a 3430 */
 #define CONFIG_OMAP3_DEVKIT8000	1	/* working with DevKit8000 */
 
-#define	CONFIG_SYS_TEXT_BASE	0x80008000
+/*
+ * 1MB into the SDRAM to allow for SPL's bss at the beginning of SDRAM
+ * 64 bytes before this address should be set aside for u-boot.img's
+ * header. That is 0x800FFFC0--0x80100000 should not be used for any
+ * other needs.
+ */
+#define	CONFIG_SYS_TEXT_BASE	0x80100000
 
 #define CONFIG_SDRC	/* The chip has SDRC controller */