[U-Boot-Users] Re: Problem with bootm command

[U-Boot-Users] Re: Problem with bootm command

[Mirco Fuchs]

I have one more question about using the bootm command.

I searched the do_bootm_linux() function to find out what u-boot is doing when booting linux using the bootm command, because the kernel patch i found in the mailing list (which makes the kernel able to copy an initrd directly from flash to ram) has to be used with kernel version 2.4.18, but i have 2.6.12.

It seems as if u-boot checks for an enviroment string called "initrd_high" and if it exists, u-boot does not copy the initrd to ram -> the kernel will fail to load the ramdisk. But there is no such string in my enviroment, so shouldn't the bootm command copy initrd to it's designated sdram location? It seems as it does not because the start address of the initrd, which u-boot delivers to the linux kernel is located in flash (one can read it in the kernel booting sequence; the error is that the initrd extends physical memory, but it's quite clear why this error occures).

Is it possible to force u-boot to copy the image to ram? Of course i could use the cp.b command before using bootm (i tested this and it works fine), but thats not what i want to do.

Regards

Mirco
_________________________________________________________________________
Mit der Gruppen-SMS von WEB.DE FreeMail k?nnen Sie eine SMS an alle 
Freunde gleichzeitig schicken: http://freemail.web.de/features/?mc=021179

[U-Boot-Users] Re: Problem with bootm command

[Wolfgang Denk]

In message <4221395@web.de> you wrote:
> 
> I searched the do_bootm_linux() function to find out what u-boot is doing
> when booting linux using the bootm command, because the kernel patch i fou
> nd in the mailing list (which makes the kernel able to copy an initrd dire
> ctly from flash to ram) has to be used with kernel version 2.4.18, but i h
> ave 2.6.12.

It would be definitely helpful if you expolained  which  architecture
you're working on. I speculate that this is ona ARM system?

> It seems as if u-boot checks for an enviroment string called "initrd_high"
>  and if it exists, u-boot does not copy the initrd to ram -> the kernel wi

No, this is not what it does. Please read the documentation.
The behavious is documented both in the manual and in the README.

> ll fail to load the ramdisk. But there is no such string in my enviroment,
>  so shouldn't the bootm command copy initrd to it's designated sdram locat
> ion? It seems as it does not because the start address of the initrd, whic
> h u-boot delivers to the linux kernel is located in flash (one can read it
>  in the kernel booting sequence; the error is that the initrd extends phys
> ical memory, but it's quite clear why this error occures).

I cannot parse this. Isn't your flash mapped in  a  physical  address
range? Or what exacty do you mean?

All this has been discussed many  times  before.  May  I  suggest  to
search the archives?

> Is it possible to force u-boot to copy the image to ram? Of course i could

Why should it do that? It's just a waste of boot time. THe kernel can
poick it up from flash, so why bother with copying it to ram first?

>  use the cp.b command before using bootm (i tested this and it works fine)
> , but thats not what i want to do.

And it's not needed.

Best regards,

Wolfgang Denk

-- 
Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"Ada is PL/I trying to be Smalltalk.                 - Codoso diBlini

[U-Boot-Users] Re: Problem with bootm command

[Mirco Fuchs]

>in message <4221395@web.de> you wrote: 
>> 
>> I searched the do_bootm_linux() function to find out what u-boot is doing 
>> when booting linux using the bootm command, because the kernel patch i fou 
>> nd in the mailing list (which makes the kernel able to copy an initrd dire 
>> ctly from flash to ram) has to be used with kernel version 2.4.18, but i h 
>> ave 2.6.12. 
>
>It would be definitely helpful if you expolained which architecture 
>you're working on. I speculate that this is ona ARM system? 

 It's an AT91RM9200

>
>> ll fail to load the ramdisk. But there is no such string in my enviroment, 
>> so shouldn't the bootm command copy initrd to it's designated sdram locat 
>> ion? It seems as it does not because the start address of the initrd, whic 
>> h u-boot delivers to the linux kernel is located in flash (one can read it 
>> in the kernel booting sequence; the error is that the initrd extends phys 
>> ical memory, but it's quite clear why this error occures). 
>
>I cannot parse this. Isn't your flash mapped in a physical address 
>range? Or what exacty do you mean? 

The point is the following. The flash is mapped from 0x10000000 to 0x10800000.
u-boot delilveres the start address of the initrd to the kernel which is 0x10160000, if the image is located in flash. But the Kernel expects an address located in the sdram which is mapped from 0x2000000 to 0x24000000. Thats why the kernel is not able to mount the initrd. It tells me that the initrd extends physical memory

>All this has been discussed many times before. May I suggest to 
>search the archives? 
>
>> Is it possible to force u-boot to copy the image to ram? Of course i could 
>
>Why should it do that? It's just a waste of boot time. THe kernel can 
>poick it up from flash, so why bother with copying it to ram first? 

Of course, it's a waste of time, but obviously my kernel is not able to pick it up from flash

Regards

Mirco
_________________________________________________________________________
Mit der Gruppen-SMS von WEB.DE FreeMail k?nnen Sie eine SMS an alle 
Freunde gleichzeitig schicken: http://freemail.web.de/features/?mc=021179

[U-Boot-Users] Re: Problem with bootm command

[Wolfgang Denk]

In message <4428425@web.de> you wrote:
> 
>  It's an AT91RM9200

OK, so it's an ARM as I expected.

> The point is the following. The flash is mapped from 0x10000000 to 0x10800000.
> u-boot delilveres the start address of the initrd to the kernel
> which is 0x10160000, if the image is located in flash. But the Kernel
> expects an address located in the sdram which is mapped from
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Here is your problem. The kernel shouldn't do this. Patches have been
posted several times before.

> 0x2000000 to 0x24000000. Thats why the kernel is not able to mount
> the initrd. It tells me that the initrd extends physical memory

[Can you please stick to the Netiquette and use shorter lines? Thanks.]

> >Why should it do that? It's just a waste of boot time. THe kernel can 
> >poick it up from flash, so why bother with copying it to ram first? 
> 
> Of course, it's a waste of time, but obviously my kernel is not able to pick it up from flash

Then fix the problem where it happens: in your kernel.  As  mentioned
before,  all  this has been discusses so many times before, including
the required modifications.

Best regards,

Wolfgang Denk

-- 
Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"Unibus timeout fatal trap program lost sorry"  -  An  error  message
printed by DEC's RSTS operating system for the PDP-11