RFC: U-Boot OneNAND IPL TEXT_BASE
Hey,
I've been tinkering with OneNAND IPL in uboot. I found out it wan't to load itself to the address specified in board/$(BOARDDIR)/config.mk . That's fine in most cases, but in my case that wasn't possible.
In my case, SDRAM init didn't fit into the IPL, so I had to copy U-Boot into SRAM, then execute it and let it relocate itself into SDRAM. One more time:
IPL: OneNAND->SRAM
U-Boot: SRAM->SDRAM
It all works fine, but when compiling the IPL, I had to alter TEXT_BASE not to point into SDRAM but SRAM too (because of stack). I introduced a variable called IPL which allows using config.mk from (for example) onenand- ipl/board/$(BOARDDIR)/config.mk rather than the board/$(BOARDDIR)/config.mk one, which is in my opinion a correct behaviour.
Any opinions?
Thanks in advance.
config.mk +2 -0
@@ -130,9 +130,13 @@ BOARDDIR = $(VENDOR)/$(BOARD) else BOARDDIR = $(BOARD) endif +ifdef IPL +sinclude $(TOPDIR)/$(IPL)/board/$(BOARDDIR)/config.mk # include IPL specific
rules
+else
ifdef BOARD
sinclude $(TOPDIR)/board/$(BOARDDIR)/config.mk # include board specific rules endif
+endif
#########################################################################
Hi,
Which CPU do you use? In most ARM cpu, CPU load the IPL into its internal SRAM, and runs at here.
But in your case it's not. So you maybe redefine IPL address.
Instead of modifying the config.mk, how about to define IPL_TEXT_BASE.
e.g.,
In OneNAND IPL code,
#ifndef BOARD_IPL_TEXT_BASE
#define ONENAND_IPL_TEXT_BASE BOARD_IPL_TEXT_BASE
#endif
load IPL code ONENAND_IPL_TEXT_BASE instead of current code.
Thank you,
Kyungmin Park
Quoted 45 lines · Marek Vasut
Hey,
I've been tinkering with OneNAND IPL in uboot. I found out it wan't to load itself to the address specified in board/$(BOARDDIR)/config.mk . That's fine in most cases, but in my case that wasn't possible.
In my case, SDRAM init didn't fit into the IPL, so I had to copy U-Boot into SRAM, then execute it and let it relocate itself into SDRAM. One more time:
IPL:OneNAND->SRAM
U-Boot:SRAM->SDRAMIt all works fine, but when compiling the IPL, I had to alter
TEXT_BASEnot to point into SDRAM but SRAM too (because of stack). I introduced a variable called IPL which allows using config.mk from (for example) onenand- ipl/board/$(BOARDDIR)/config.mk rather than the board/$(BOARDDIR)/config.mk one, which is in my opinion a correct behaviour.Any opinions?
Thanks in advance.
config.mk +0
-0@@ -130,9 +130,13 @@ BOARDDIR = $(VENDOR)/$(BOARD)?else
?BOARDDIR = $(BOARD)
?endifconfig.mk +2
-0+ifdef ?IPL +sinclude $(TOPDIR)/$(IPL)/board/$(BOARDDIR)/config.mk ?# include IPL specificrules
+else
?ifdef ?BOARD
?sinclude $(TOPDIR)/board/$(BOARDDIR)/config.mk # include board specific rules ?endif
+endif?#########################################################################
_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot
Dne Ne 2. kv?tna 2010 05:54:41 Kyungmin Park napsal(a):
Quoted 4 lines
Hi,
Which CPU do you use? In most ARM cpu, CPU load the IPL into its internal SRAM, and runs at here.
PXA270 ... the BootRAM is mapped to 0x0 and the code runs from there.
Quoted 11 lines
But in your case it's not. So you maybe redefine IPL address.
Instead of modifying the config.mk, how about to define
IPL_TEXT_BASE.e.g.,
In OneNAND IPL code,
#ifndef
BOARD_IPL_TEXT_BASE
#defineONENAND_IPL_TEXT_BASEBOARD_IPL_TEXT_BASE
#endif
That won't help -- TEXT_BASE is used while compiling start.S (to setup stack under TEXT_BASE) and the config.mk prepares the gcc options for start.S (the TEXT_BASE is defined in the gcc options).So modifying config.mk is unavoidable I fear.
The thing we can do is define STACK_TOP_BASE and that'd certainly do the trick! This would allow me to point stack into SRAM, while running the IPL code from BootRAM. Of course, in case STACK_TOP_BASE wasn't defined, it'd fall back to the old code in start.S. This would of course work without modifying config.mk and I start to like such a solution.
Hm?
Quoted 5 lines
load IPL code
ONENAND_IPL_TEXT_BASEinstead of current code.Thank you,
Kyungmin Park
Thanks!
Quoted 49 lines · Marek Vasut
Quoted 47 lines
Hey,
I've been tinkering with OneNAND IPL in uboot. I found out it wan't to load itself to the address specified in board/$(BOARDDIR)/config.mk . That's fine in most cases, but in my case that wasn't possible.
In my case, SDRAM init didn't fit into the IPL, so I had to copy U-Boot into SRAM, then execute it and let it relocate itself into SDRAM. One more time: IPL:
OneNAND->SRAM
U-Boot:SRAM->SDRAMIt all works fine, but when compiling the IPL, I had to alter
TEXT_BASEnot to point into SDRAM but SRAM too (because of stack). I introduced a variable called IPL which allows using config.mk from (for example) onenand-
ipl/board/$(BOARDDIR)/config.mk rather than the board/$(BOARDDIR)/config.mk one, which is in my opinion a correct behaviour.Any opinions?
Thanks in advance.
config.mk +2
-0@@ -130,9 +130,13 @@ BOARDDIR = $(VENDOR)/$(BOARD) else BOARDDIR = $(BOARD) endif +ifdef IPL +sinclude $(TOPDIR)/$(IPL)/board/$(BOARDDIR)/config.mk # include IPLspecific rules
+else
ifdef BOARD
sinclude $(TOPDIR)/board/$(BOARDDIR)/config.mk # include board specific
rules endif
+endif########################################################################
#_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot
Quoted 14 lines · Marek Vasut
Dne Ne 2. kv?tna 2010 05:54:41 Kyungmin Park napsal(a):
Quoted 4 lines
Hi,
Which CPU do you use? In most ARM cpu, CPU load the IPL into its internal SRAM, and runs at here.
PXA270 ... the BootRAM is mapped to
0x0and the code runs from there.Quoted 8 lines
But in your case it's not. So you maybe redefine IPL address.
Instead of modifying the config.mk, how about to define
IPL_TEXT_BASE.e.g.,
In OneNAND IPL code,
Quoted 8 lines
Quoted 3 lines
#ifndef
BOARD_IPL_TEXT_BASE
#defineONENAND_IPL_TEXT_BASEBOARD_IPL_TEXT_BASE
#endifThat won't help --
TEXT_BASEis used while compilingstart.S(to setup stack underTEXT_BASE) and the config.mk prepares the gcc options forstart.S(theTEXT_BASEis defined in the gcc options).So modifying config.mk is unavoidable I fear.
Sorry my mistake. It's already supported at current drivers. Now I works on s5pc100 and s5pc110 and each define it's TEXT_BASE at onenand_ipl/board/samsung/${board_name}/config.mk. and it's used at start.S
No need to modify config.mk and Makefile. If you need I can't send a sample files to you.
Thank you,
Kyungmin Park
Quoted 66 lines
The thing we can do is define
STACK_TOP_BASEand that'd certainly do the trick! This would allow me to point stack into SRAM, while running the IPL code from BootRAM. Of course, in caseSTACK_TOP_BASEwasn't defined, it'd fall back to the old code instart.S. This would of course work without modifying config.mk and I start to like such a solution.Hm?
Quoted 5 lines
load IPL code
ONENAND_IPL_TEXT_BASEinstead of current code.Thank you,
Kyungmin ParkThanks!
Quoted 49 lines · Marek Vasut
Quoted 47 lines
Hey,
I've been tinkering with OneNAND IPL in uboot. I found out it wan't to load itself to the address specified in board/$(BOARDDIR)/config.mk . That's fine in most cases, but in my case that wasn't possible.
In my case, SDRAM init didn't fit into the IPL, so I had to copy U-Boot into SRAM, then execute it and let it relocate itself into SDRAM. One more time: IPL:
OneNAND->SRAM
U-Boot:SRAM->SDRAMIt all works fine, but when compiling the IPL, I had to alter
TEXT_BASEnot to point into SDRAM but SRAM too (because of stack). I introduced a variable called IPL which allows using config.mk from (for example) onenand-
ipl/board/$(BOARDDIR)/config.mk rather than the board/$(BOARDDIR)/config.mk one, which is in my opinion a correct behaviour.Any opinions?
Thanks in advance.
config.mk +0
-0@@ -130,9 +130,13 @@ BOARDDIR = $(VENDOR)/$(BOARD)?else
?BOARDDIR = $(BOARD)
?endifconfig.mk +2
-0+ifdef ?IPL +sinclude $(TOPDIR)/$(IPL)/board/$(BOARDDIR)/config.mk ?# include IPLspecific rules
+else
?ifdef ?BOARD
?sinclude $(TOPDIR)/board/$(BOARDDIR)/config.mk # include board specific rules endif
+endif?######################################################################## #
_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot
Dne Ne 2. kv?tna 2010 13:07:13 Kyungmin Park napsal(a):
Quoted 26 lines · Marek Vasut
Quoted 16 lines
Dne Ne 2. kv?tna 2010 05:54:41 Kyungmin Park napsal(a):
Quoted 4 lines
Hi,
Which CPU do you use? In most ARM cpu, CPU load the IPL into its internal SRAM, and runs at here.
PXA270 ... the BootRAM is mapped to
0x0and the code runs from there.Quoted 8 lines
But in your case it's not. So you maybe redefine IPL address.
Instead of modifying the config.mk, how about to define
IPL_TEXT_BASE.e.g.,
In OneNAND IPL code,
Quoted 8 lines
Quoted 3 lines
#ifndef
BOARD_IPL_TEXT_BASE
#defineONENAND_IPL_TEXT_BASEBOARD_IPL_TEXT_BASE
#endifThat won't help --
TEXT_BASEis used while compilingstart.S(to setup stack underTEXT_BASE) and the config.mk prepares the gcc options forstart.S(theTEXT_BASEis defined in the gcc options).So modifying config.mk is unavoidable I fear.
Quoted 5 lines
Sorry my mistake. It's already supported at current drivers. Now I works on s5pc100 and s5pc110 and each define it's
TEXT_BASEatonenand_ipl/board/samsung/${board_name}/config.mk. and it's used atstart.S
I noticed the config.mk in onenand-ipl/board/... were meant to be used, but in my case, those were not used. Instead, the board/... config.mk was used (and from what I saw in config.mk, that's understandable).
Also, see the Makefile in onenand-ipl/board/apollon/Makefile -- onenand- ipl/board/apollon/config.mk defines the TEXT_BASE in Makefile, but right before that it includes the system-wide config.mk, which configures the compiler options and amongst those is -DTEXT_BASE and that one is set to one defined in board/apollon/config.mk . Obviously when compiling start.S, that one is used.
Cheers
Quoted 69 lines
No need to modify config.mk and
Makefile. If you need I can't send a sample files to you.Thank you,
Kyungmin ParkQuoted 63 lines
The thing we can do is define
STACK_TOP_BASEand that'd certainly do the trick! This would allow me to point stack into SRAM, while running the IPL code from BootRAM. Of course, in caseSTACK_TOP_BASEwasn't defined, it'd fall back to the old code instart.S. This would of course work without modifying config.mk and I start to like such a solution.Hm?
Quoted 4 lines
load IPL code
ONENAND_IPL_TEXT_BASEinstead of current code.Thank you,
Kyungmin ParkThanks!
Quoted 48 lines · Marek Vasut
Quoted 47 lines
Hey,
I've been tinkering with OneNAND IPL in uboot. I found out it wan't to load itself to the address specified in board/$(BOARDDIR)/config.mk . That's fine in most cases, but in my case that wasn't possible.
In my case, SDRAM init didn't fit into the IPL, so I had to copy U-Boot into SRAM, then execute it and let it relocate itself into SDRAM. One more time: IPL:
OneNAND->SRAM
U-Boot:SRAM->SDRAMIt all works fine, but when compiling the IPL, I had to alter
TEXT_BASEnot to point into SDRAM but SRAM too (because of stack). I introduced a variable called IPL which allows using config.mk from (for example) onenand-
ipl/board/$(BOARDDIR)/config.mk rather than the board/$(BOARDDIR)/config.mk one, which is in my opinion a correct behaviour.Any opinions?
Thanks in advance.
config.mk +2
-0@@ -130,9 +130,13 @@ BOARDDIR = $(VENDOR)/$(BOARD) else BOARDDIR = $(BOARD) endif +ifdef IPL +sinclude $(TOPDIR)/$(IPL)/board/$(BOARDDIR)/config.mk # include IPLspecific rules
+else
ifdef BOARD
sinclude $(TOPDIR)/board/$(BOARDDIR)/config.mk # include board
specific rules endif
+endif#####################################################################
### #_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot