Bogus machine-type from bootloader

Bogus machine-type from bootloader

[Damjan Marion]

hi,

Is there any "proper" way if fixing bogus machine type sent by bootloader (ofcourse except fixing bootloader itself)?

I'm looking into code where author did following:


+++ a/arch/arm/boot/compressed/head-str9100.S	2009-01-20 04:26:26.000000000 +0100
@@ -0,0 +1,7 @@
+#include <asm/mach-types.h>
+
+		.section	".start", "ax"
+
+__str9100_start:
+		mov	r7, #(MACH_TYPE_STR9100 & 0xFF00)
+		orr	r7, r7, #(MACH_TYPE_STR9100 & 0x00FF)

I assume this is not something acceptable for having in official kernel source.


Damjan

Bogus machine-type from bootloader

[Russell King - ARM Linux]

On Fri, Dec 24, 2010 at 12:19:25PM +0100, Damjan Marion wrote:
> Is there any "proper" way if fixing bogus machine type sent by
> bootloader (ofcourse except fixing bootloader itself)?
> 
> I'm looking into code where author did following:
> 
> 
> +++ a/arch/arm/boot/compressed/head-str9100.S	2009-01-20 04:26:26.000000000 +0100
> @@ -0,0 +1,7 @@
> +#include <asm/mach-types.h>
> +
> +		.section	".start", "ax"
> +
> +__str9100_start:
> +		mov	r7, #(MACH_TYPE_STR9100 & 0xFF00)
> +		orr	r7, r7, #(MACH_TYPE_STR9100 & 0x00FF)
> 
> I assume this is not something acceptable for having in official kernel source.

Bogus machine-type from bootloader

[Damjan Marion]

On Dec 24, 2010, at 12:29 PM, Russell King - ARM Linux wrote:
> On Fri, Dec 24, 2010 at 12:19:25PM +0100, Damjan Marion wrote:
>> Is there any "proper" way if fixing bogus machine type sent by
>> bootloader (ofcourse except fixing bootloader itself)?
>> 
>> I'm looking into code where author did following:
>> 
>> 
>> +++ a/arch/arm/boot/compressed/head-str9100.S	2009-01-20 04:26:26.000000000 +0100
>> @@ -0,0 +1,7 @@
>> +#include <asm/mach-types.h>
>> +
>> +		.section	".start", "ax"
>> +
>> +__str9100_start:
>> +		mov	r7, #(MACH_TYPE_STR9100 & 0xFF00)
>> +		orr	r7, r7, #(MACH_TYPE_STR9100 & 0x00FF)
>> 
>> I assume this is not something acceptable for having in official kernel source.
> 
> From what I remember looking at the patch URL you posted, they also
> added entries for ID 2000 to their mach-types file, which have long
> since been taken by other people.

Yes, i fixed that

# bdinfo
arch_number = 0x00000CBA

but kernel still complains:

Error: unrecognized/unsupported machine ID (r1 = 0x00e6dca8).

Bogus machine-type from bootloader

[Alexander Clouter]

Damjan Marion <damjan.marion@gmail.com> wrote:
> 
> Is there any "proper" way if fixing bogus machine type sent by 
> bootloader (ofcourse except fixing bootloader itself)?
>
Proper no, as the bootloader is the 'proper' way.

The kludge fix that many distro's seem to use (Debian for example):

http://www.digriz.org.uk/ts78xx/booting-woes#unrecognized.2BAC8-unsupported_machine_ID

In short, what it does is get the first two instructions of a booting 
kernel to load the correct machine ID into r1.  The example there is for 
0x0674 (1652), you can adjust it for your own needs.

Cook the kernel, run the 'fixup' script and you should be okay.

Cheers

-- 
Alexander Clouter
.sigmonster says: Androphobia:
                  	Fear of men.