[U-Boot-Users] PPC 8260 Linux boot
I am trying to boot Linux 2.4.25 (code from denx linux-2-4-devel) The board is similar to MPC8260ADS What happens is that the kernel is starting but after a short while gets an exception and the board is reset again. We tried looking at the kernel sources to see where we get, we did this by lighting hardware LEDS we have on the board. We tried to find where in the kernel we get the exception, we could not find exactly where it happens, but we did find that it happens somewhere in the function prom_init called from early_init inside arch/ppc/kernel/setup.c
Does anyone have any idea what we should look next?
-- Ori Idan
In message 4105294D.5070508@helicontech.co.il you wrote:
I am trying to boot Linux 2.4.25 (code from denx linux-2-4-devel) The board is similar to MPC8260ADS
Arghh... A description like this is absolutley useless, and misleading. Your probably even fooling yourself. Similar? In that it uses the same processor? Or is everything IDENTICAL?
What happens is that the kernel is starting but after a short while gets an exception and the board is reset again.
Please define "after a short while". Do you see any boot messages on the console? Do you see any messages when you do a post mortem dump of the log_buf memory? [See http://www.denx.de/twiki/bin/view/DULG/LinuxPostMortemAnalysis]
We tried looking at the kernel sources to see where we get, we did this by lighting hardware LEDS we have on the board.
I see. Please remove ALL this code and try again. perhaps it is just your debugging code which crashes the system. [This has been discussed before, both here and in the linuxppc mailing lists. Search the archives.]
We tried to find where in the kernel we get the exception, we could not find exactly where it happens, but we did find that it happens somewhere in the function prom_init called from early_init inside arch/ppc/kernel/setup.c
Does anyone have any idea what we should look next?
OK, I'll give you two hints:
1) get yourself a BDI2000. You will need it.
2) Check your memory map, and the init sequence of your SDRAM [no, I'm not going to explain this here again. Search the archives.]
Best regards,
Wolfgang Denk
Thank you for your quick answer.
I am trying to boot Linux 2.4.25 (code from denx linux-2-4-devel) The board is similar to MPC8260ADS
Arghh... A description like this is absolutley useless, and misleading. Your probably even fooling yourself. Similar? In that it uses the same processor? Or is everything IDENTICAL?
If it was not IDENTICAL I would not have mentioned this board.
What happens is that the kernel is starting but after a short while gets an exception and the board is reset again.
Please define "after a short while". Do you see any boot messages on the console? Do you see any messages when you do a post mortem dump of the log_buf memory? [See http://www.denx.de/twiki/bin/view/DULG/LinuxPostMortemAnalysis]
I do not see any boot messages, I can't exactly define how long it takes until the reset but it is a very short time that during this time I expected at least one line saying this is the kernel.
We tried looking at the kernel sources to see where we get, we did this by lighting hardware LEDS we have on the board.
I see. Please remove ALL this code and try again. perhaps it is just your debugging code which crashes the system. [This has been discussed before, both here and in the linuxppc mailing lists. Search the archives.]
I tried removing all my code and tried again, that was the first thing I thought that maybe I crash the system. I tried again and again and the problem was the same.
We tried to find where in the kernel we get the exception, we could not find exactly where it happens, but we did find that it happens somewhere in the function prom_init called from early_init inside arch/ppc/kernel/setup.c
Does anyone have any idea what we should look next?
OK, I'll give you two hints:
get yourself a BDI2000. You will need it.
Check your memory map, and the init sequence of your SDRAM [no,
I'm not going to explain this here again. Search the archives.]
Unfortunatly the archives are not searchable, otherwise that would be the first thing I would do, could you give me some hints on what do you mean init sequence of SDRAM? As for the memory map, I made sure I have RAM starting from 0x0000, I have 128Mb of RAM. The IMMR is set to 0xF0000000.
-- Ori Idan
In message 4105FA9D.30400@helicontech.co.il you wrote:
I am trying to boot Linux 2.4.25 (code from denx linux-2-4-devel) The board is similar to MPC8260ADS
...
If it was not IDENTICAL I would not have mentioned this board.
So it is identical? Then why did you write "similar to" in the first place?
And if it's really identical I thinkt he standard MPC8260ADS code should work identical, too.
Please define "after a short while". Do you see any boot messages on the console? Do you see any messages when you do a post mortem dump of the log_buf memory? [See http://www.denx.de/twiki/bin/view/DULG/LinuxPostMortemAnalysis]
I do not see any boot messages, I can't exactly define how long it takes until the reset but it is a very short time that during this time I expected at least one line saying this is the kernel.
I repeat:
Do you see any messages when you do a post mortem dump of the log_buf memory? [See http://www.denx.de/twiki/bin/view/DULG/LinuxPostMortemAnalysis]
????
I tried removing all my code and tried again, that was the first thing I thought that maybe I crash the system. I tried again and again and the problem was the same.
You cannot make such a claim. The _symptoms_ may have been the same. The problem may have been a different one.
- Check your memory map, and the init sequence of your SDRAM [no,
I'm not going to explain this here again. Search the archives.]
Unfortunatly the archives are not searchable, otherwise that would be the first thing I would
The GMANE archive _is_ searchable.
do, could you give me some hints on what do you mean init sequence of SDRAM?
IO mean the initialization sequence that is required for the SDRAM chips. In 98% of all porting problems this is the culprit.
As for the memory map, I made sure I have RAM starting from 0x0000, I have 128Mb of RAM. The IMMR is set to 0xF0000000.
Did you read the FAQ? How about the "clocks_in_mhz" setting? Or the bd_info structure used by the Linux kernel?
Best regards,
Wolfgang Denk
I repeat:
Do you see any messages when you do a post mortem dump of the log_buf memory? [See http://www.denx.de/twiki/bin/view/DULG/LinuxPostMortemAnalysis]
????
I think I don't understand something in your explenation of post mortem, how can I do grep log_buf system.map when I have nothing but UBOOT prompt?
I tried removing all my code and tried again, that was the first thing I thought that maybe I crash the system. I tried again and again and the problem was the same.
You cannot make such a claim. The _symptoms_ may have been the same. The problem may have been a different one.
- Check your memory map, and the init sequence of your SDRAM [no,
I'm not going to explain this here again. Search the archives.]
Unfortunatly the archives are not searchable, otherwise that would be the first thing I would
The GMANE archive _is_ searchable.
What is the GNAME archive?
do, could you give me some hints on what do you mean init sequence of SDRAM?
IO mean the initialization sequence that is required for the SDRAM chips. In 98% of all porting problems this is the culprit.
I still do not understand what do you mean by init sequence of SDRAM, If I had any hint of what you mean by that I would check for it. All I know is that I have 128MB of RAM mapped from 0x0000 I have checked it again.
Did you read the FAQ? How about the "clocks_in_mhz" setting? Or the bd_info structure used by the Linux kernel?
Yes, I have thought about the clock value and made sure it is passed Ok, since I am using kernel 2.4.25 I understand that the clock should be in Hz and that is what U-boot transfers.
-- Ori Idan
In message 4106D6BF.6090600@helicontech.co.il you wrote:
Do you see any messages when you do a post mortem dump of the log_buf memory? [See http://www.denx.de/twiki/bin/view/DULG/LinuxPostMortemAnalysis]
????
I think I don't understand something in your explenation of post mortem, how can I do grep log_buf system.map when I have nothing but UBOOT prompt?
You do have a development system which you used to compile and link your Linux kernel, don't you? And you do have the "grep" command installed on that host, right?
The GMANE archive _is_ searchable.
What is the GNAME archive?
http://news.gmane.org/gmane.comp.boot-loaders.u-boot
[Why didn't a google search for "gmane u-boot" return useful links for you?]
I still do not understand what do you mean by init sequence of SDRAM, If I had any hint of what you mean by that I would check for it.
Read the documentation provided by the manufacturer of your RAM chips. It shosuld contain a section about the SDRAM init sequence.
Did you read the FAQ? How about the "clocks_in_mhz" setting? Or the bd_info structure used by the Linux kernel?
Yes, I have thought about the clock value and made sure it is passed Ok, since I am using kernel 2.4.25 I understand that the clock should be in Hz and that is what U-boot transfers.
And what about bd_info?
Best regards,
Wolfgang Denk
You do have a development system which you used to compile and link your Linux kernel, don't you? And you do have the "grep" command installed on that host, right?
Ok, I think it should be written clearly in the wiki that this is the host system.
The GMANE archive _is_ searchable.
What is the GNAME archive?
http://news.gmane.org/gmane.comp.boot-loaders.u-boot
[Why didn't a google search for "gmane u-boot" return useful links for you?]
I still do not understand what do you mean by init sequence of SDRAM, If I had any hint of what you mean by that I would check for it.
Read the documentation provided by the manufacturer of your RAM chips. It shosuld contain a section about the SDRAM init sequence.
Isn't the fact that I have RAM from 0x0000 means that RAM is initialized? I have checked in Uboot that I can write and read this RAM so I am sure I have it. What other initialization should be done?
Did you read the FAQ? How about the "clocks_in_mhz" setting? Or the bd_info structure used by the Linux kernel?
Yes, I have thought about the clock value and made sure it is passed Ok, since I am using kernel 2.4.25 I understand that the clock should be in Hz and that is what U-boot transfers.
And what about bd_info?
I have checked bd_info and it transfers the clock in Hz and also transfer the right RAM and IMMR information.
-- Ori Idan
In message 41079531.2000204@helicontech.co.il you wrote:
You do have a development system which you used to compile and link your Linux kernel, don't you? And you do have the "grep" command installed on that host, right?
Ok, I think it should be written clearly in the wiki that this is the host system.
You are the first person to misunderstand this.
Read the documentation provided by the manufacturer of your RAM chips. It shosuld contain a section about the SDRAM init sequence.
Isn't the fact that I have RAM from 0x0000 means that RAM is initialized?
No, it does not. Mapping some address space in the memory controller and actually initializing a device are completely different issues.
I have checked in Uboot that I can write and read this RAM so I am sure I have it.
Please read the archives. Simple read and write access are easy; this still does not give any guarantee that burst mode accesses will work.
What other initialization should be done?
I already told you:
READ THE DOCUMENTATION PROVIDED BY THE MANUFACTURER OF YOUR RAM CHIPS. IT SHOULD CONTAIN A SECTION ABOUT THE REQUIRED INIT SEQUENCE.
And _please_ read the mail list archive.
And what about bd_info?
I have checked bd_info and it transfers the clock in Hz and also transfer the right RAM and IMMR information.
Argh... You make it really difficult to help you.
Did you check ALL items of the FAQ "Linux hangs after uncompressing the kernel", i. e. did you really make sure that U-Boot and Linux are using the SAME version of bd_info?
Best regards,
Wolfgang Denk
Ok, it seems Linux is booting now or at least we think so There is not exception now but all we get is few unmeaningfull character on the terminal We belive Linux is booting but the kernel is miscalculating the baud rate. The clock information in bd_info seems to be Ok but I will double check this tomorow.
Does someone has any clue for what may cause this problem?
-- Ori Idan
Wolfgang Denk wrote:
In message 4105294D.5070508@helicontech.co.il you wrote:
I am trying to boot Linux 2.4.25 (code from denx linux-2-4-devel) The board is similar to MPC8260ADS
Arghh... A description like this is absolutley useless, and misleading. Your probably even fooling yourself. Similar? In that it uses the same processor? Or is everything IDENTICAL?
What happens is that the kernel is starting but after a short while gets an exception and the board is reset again.
Please define "after a short while". Do you see any boot messages on the console? Do you see any messages when you do a post mortem dump of the log_buf memory? [See http://www.denx.de/twiki/bin/view/DULG/LinuxPostMortemAnalysis]
We tried looking at the kernel sources to see where we get, we did this by lighting hardware LEDS we have on the board.
I see. Please remove ALL this code and try again. perhaps it is just your debugging code which crashes the system. [This has been discussed before, both here and in the linuxppc mailing lists. Search the archives.]
We tried to find where in the kernel we get the exception, we could not find exactly where it happens, but we did find that it happens somewhere in the function prom_init called from early_init inside arch/ppc/kernel/setup.c
Does anyone have any idea what we should look next?
OK, I'll give you two hints:
get yourself a BDI2000. You will need it.
Check your memory map, and the init sequence of your SDRAM [no,
I'm not going to explain this here again. Search the archives.]
Best regards,
Wolfgang Denk
participants (2)
-
Ori Idan -
Wolfgang Denk