Hi,

There are previous posts with regard to the same problem, but I have tried all the solutions but they are not working for me, so I'm posting again.

I am working on a board with AT91SAM9G45 (designed based on AT91SAM9G45-EKES). I'm using Linux 2.6.32.8. My problem is that Linux does not use the second memory bank (128 MB on EBI_CS1 mapped at 0x20000000).

The memory mapping is as follows:
0x70000000 => 0xc0000000 (128 MB - DDRSDRC0)
0x20000000 => 0xc8000000 (128 MB - EBI_CS1)

While booting I see this message: "Ignoring RAM at 20000000-27ffffff (vmalloc region overlap)".

I have enabled SparseMem and also tried patches posted at:
http://blog.linuxconsulting.ro/2010/05/ ... -with.html

I have checked & verified my DDR2 initializations in AT91Bootstrap.

I have added uBoot boot args as:
mem=exactmap memmap=0x08000000@0x70000000 memmap=0x08000000@0x20000000

The exact/root problem as it seems to me now that linux does not get PHYS_OFFSET / PAGE_OFFSET dynamically depending on the memory bank in use.

Pls suggest the work arounds or any solutions, if someone has already solved it.

- PJ

Either you need to raise the vmalloc ceiling by changing VMALLOC_END in /arc/arm/mach-at91/include/mach/vmalloc.h
Or shrink the size of the vmalloc area so it does impinge on your upper memory block, by specifying the size of the command line using vmalloc=xxxM

Hello,

thanks for your reply.

I tried with vmalooc=64M & vmalloc=16M, the results were same.

I would like you to look into this file: arch/arm/mm/mmu.c
If you check function "sanity_check_meminfo()" where this error is printed, pls note the conditions of the memory check. The condition that fails is "__va(bank->start) < (void *)PAGE_OFFSET".

This happens because the linear memory mapping maps this bank at a wrong virtual address.

Now if you read my earlier post, you will be able to understand what I was trying to ask. (How to dynamically get PHYS_OFFSET with reference to the memory bank referenced)

If you think that I'm missing something, please point out to me.

Some values that you may need:
VMALLOC_END 0xFEE00000
VMALLOC_MIN 0xF6E00000
PAGE_OFFSET 0xC0000000
PHYS_OFFSET 0x70000000

-PJ

I've made the following changes to my Linux 2.6.32-8 to get 256MB of DDR2 working on at91sam9g45-ekes

(1)
added to arch/arm/mach-at91/include/mach/memeory.h
#if defined(CONFIG_ARCH_AT91SAM9G45) || defined(CONFIG_ARCH_AT91SAM9M10)
/*
* Non-linear mapping like so:
* phys => virt
* 0x70000000 => 0xc0000000
* 0x20000000 => 0xc8000000
*/

#define __phys_to_virt(p) \
(((p) & 0x07ffffff) + (((p) & 0x40000000) ? 0xc0000000 : 0xc8000000))


#define __virt_to_phys(v) \
(((v) & 0x07ffffff) + (((v) & 0x08000000) ? 0x20000000 : 0x70000000 ))

#define NODE_MEM_SIZE_BITS 27
#define MAX_PHYSMEM_BITS 32
#define SECTION_SIZE_BITS 27
#define HIGH_MEMORY_VIRT 0xd0000000

#endif


(2)
added to arch/arm/mm/init.c

#ifdef HIGH_MEMORY_VIRT
high_memory = HIGH_MEMORY_VIRT;
#else
high_memory = __va((max_low << PAGE_SHIFT) - 1) + 1;
#endif

(3)
The memory model selected is sparse mem & Highmem disabled.


Now when I boot my system then my kernel does not boot and when I analyze my kernel buffers I see this dump:


Linux version 2.6.32.8 (prasant@prasant-laptop) (gcc version4.2.0 20070413 (prerelease) (CodeSourcery Sourcery G++ Lite 2007q1-10)) #550 PREEMPT Thu Jul 28 16:26:14 IST 2011.

CPU: ARM926EJ-S [41069265] revision 5 (ARMv5TEJ), cr=00053177.
CPU: VIVTdata cache, VIVT instruction cache.................. AT91SAM9G45-EKES.
Ignoring unrecognised tag 0x54410008.
Memory policy:ECC disabled, Data cache writeback.
------------[ cut here ]------------.
WARNING: at mm/bootmem.c:545 alloc_arch_preferred_bootmem+0x34/0x64().
Modules linked in:.............................trace+0x0/0xdc) from [<c003d4b8>] (warn_slowpath_common+0x4c/0x68).
[<c003d4b8>] (warn_slowpath_common+0x4c/0x68) from [<c0010e50>] (alloc_arch_preferred_bootmem+0x34/0x64).
[<c0010e50>] (alloc_arch_preferred_bootmem+0x34/0x64) from [<c00113a0>] (___alloc_bootmem_nopanic+0x34/0xdc).
[<c00113a0>] (___alloc_bootmem_nopanic+0x34/0xdc) from [<c00116ac>] (___alloc_bootmem+0xc/0x30).
[<c00116ac>] (___alloc_bootmem+0xc/0x30) from [<c0012850>] (sparse_init+0x14/0x188).
[<c0012850>] (sparse_init+0x14/0x188) from [<c000c69c>] (bootmem_init+0x330/0x454).
[<c000c69c>] (bootmem_init+0x330/0x454) from [<c000d49c>] (paging_init+0x588/0x734).
[<c000d49c>] (paging_init+0x588/0x734) from [<c000b95c>] (setup_arch+0x378/0x634).
[<c000b95c>] (setup_arch+0x378/0x634) from [<c00087d0>] (start_kernel+0x60/0x394).
[<c00087d0>] (start_kernel+0x60/0x394) from [<70008034>] (0x70008034).
---[ end trace 1b75b31a2719ed1c ]---.

Any idea what could be the problems ?

-PJ

[quote="pj0585@gmail.com"]I've made the following changes to my Linux 2.6.32-8 to get 256MB of DDR2 working on at91sam9g45-ekes

(1)
added to arch/arm/mach-at91/include/mach/memeory.h
#if defined(CONFIG_ARCH_AT91SAM9G45) || defined(CONFIG_ARCH_AT91SAM9M10)
/*
* Non-linear mapping like so:
* phys => virt
* 0x70000000 => 0xc0000000
* 0x20000000 => 0xc8000000
*/

#define __phys_to_virt(p) \
(((p) & 0x07ffffff) + (((p) & 0x40000000) ? 0xc0000000 : 0xc8000000))


#define __virt_to_phys(v) \
(((v) & 0x07ffffff) + (((v) & 0x08000000) ? 0x20000000 : 0x70000000 ))

#define NODE_MEM_SIZE_BITS 27
#define MAX_PHYSMEM_BITS 32
#define SECTION_SIZE_BITS 27
#define HIGH_MEMORY_VIRT 0xd0000000

#endif


(2)
added to arch/arm/mm/init.c

#ifdef HIGH_MEMORY_VIRT
high_memory = HIGH_MEMORY_VIRT;
#else
high_memory = __va((max_low << PAGE_SHIFT) - 1) + 1;
#endif

(3)
The memory model selected is sparse mem & Highmem disabled.


Now when I boot my system then my kernel does not boot and when I analyze my kernel buffers I see this dump:


Linux version 2.6.32.8 (prasant@prasant-laptop) (gcc version4.2.0 20070413 (prerelease) (CodeSourcery Sourcery G++ Lite 2007q1-10)) #550 PREEMPT Thu Jul 28 16:26:14 IST 2011.

CPU: ARM926EJ-S [41069265] revision 5 (ARMv5TEJ), cr=00053177.
CPU: VIVTdata cache, VIVT instruction cache.................. AT91SAM9G45-EKES.
Ignoring unrecognised tag 0x54410008.
Memory policy:ECC disabled, Data cache writeback.
------------[ cut here ]------------.
WARNING: at mm/bootmem.c:545 alloc_arch_preferred_bootmem+0x34/0x64().
Modules linked in:.............................trace+0x0/0xdc) from [<c003d4b8>] (warn_slowpath_common+0x4c/0x68).
[<c003d4b8>] (warn_slowpath_common+0x4c/0x68) from [<c0010e50>] (alloc_arch_preferred_bootmem+0x34/0x64).
[<c0010e50>] (alloc_arch_preferred_bootmem+0x34/0x64) from [<c00113a0>] (___alloc_bootmem_nopanic+0x34/0xdc).
[<c00113a0>] (___alloc_bootmem_nopanic+0x34/0xdc) from [<c00116ac>] (___alloc_bootmem+0xc/0x30).
[<c00116ac>] (___alloc_bootmem+0xc/0x30) from [<c0012850>] (sparse_init+0x14/0x188).
[<c0012850>] (sparse_init+0x14/0x188) from [<c000c69c>] (bootmem_init+0x330/0x454).
[<c000c69c>] (bootmem_init+0x330/0x454) from [<c000d49c>] (paging_init+0x588/0x734).
[<c000d49c>] (paging_init+0x588/0x734) from [<c000b95c>] (setup_arch+0x378/0x634).
[<c000b95c>] (setup_arch+0x378/0x634) from [<c00087d0>] (start_kernel+0x60/0x394).
[<c00087d0>] (start_kernel+0x60/0x394) from [<70008034>] (0x70008034).
---[ end trace 1b75b31a2719ed1c ]---.

Any idea what could be the problems ?

-PJ[/quote]


Hi PJ,

I got the same scenario using a 2.6.30 and 2.6.34 kernels, ¿Do you have any news about the problem? Any solution?

I'm stuck, and I have located the same files as you with the same conclusion.

Hi rlobos,

I got the solution by switching to linux 3.0.0.

The details are on the link,
http://lists.arm.linux.org.uk/lurker/me ... 1d.en.html

You will have to read the entire thread, I've posted the entire solution there. Do read the links that I have provided in the thread.

Remember, you have to make changes to both Uboot and update linux.

-PJ