Passing an initrd to a Linux kernel with Device Tree

Passing an initrd to a Linux kernel with Device Tree

[Olivier Martin]

Hello,

I tried recently to add support in UEFI to pass an initrd with FDT to a
Linux kernel. I have searched the Internet but I did not find any
information to do that.

FYI, this is the way it is done in ATAG:
typedef struct {
  UINT32  start;
  UINT32  size;
} LINUX_ATAG_INITRD2;

The boot loader loads the initrd in system memory and passes the memory
region information through ATAG.


Sorry, if it is the wrong mailing-list. In this case, can someone point me
on the right forum for any Linux FDT oriented questions.

Thanks,
Olivier

Re: Passing an initrd to a Linux kernel with Device Tree

[Grant Likely]

On Tue, Sep 20, 2011 at 06:46:22PM +0100, Olivier Martin wrote:
> Hello,
> 
> I tried recently to add support in UEFI to pass an initrd with FDT to a
> Linux kernel. I have searched the Internet but I did not find any
> information to do that.
> 
> FYI, this is the way it is done in ATAG:
> typedef struct {
>   UINT32  start;
>   UINT32  size;
> } LINUX_ATAG_INITRD2;
> 
> The boot loader loads the initrd in system memory and passes the memory
> region information through ATAG.
> 
> 
> Sorry, if it is the wrong mailing-list. In this case, can someone point me
> on the right forum for any Linux FDT oriented questions.

Hmmm, it may not be documented (the binding was implemented a long
time ago before we got strict about documenting bindings).  The magic
is two properties in the /chosen node; "linux,initrd-start" and
"linux,initrd-end".  Each contains the physical start and end
addresses of the initrd.

g.

RE: Passing an initrd to a Linux kernel with Device Tree

[Dave P Martin]

> -----Original Message-----
> From:
> devicetree-discuss-bounces+dave.martin=arm.com-uLR06cmDAlY8sQ00jmICUQ@public.gmane.org
g [mailto:devicetree-discuss-> bounces+dave.martin=arm.com-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org] On Behalf Of
> Grant Likely
> Sent: 20 September 2011 21:11
> To: Olivier Martin
> Cc: devicetree-discuss-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org
> Subject: Re: Passing an initrd to a Linux kernel with Device Tree
>
> On Tue, Sep 20, 2011 at 06:46:22PM +0100, Olivier Martin wrote:
> > Hello,
> >
> > I tried recently to add support in UEFI to pass an initrd
> with FDT to
> > a Linux kernel. I have searched the Internet but I did not find any
> > information to do that.
> >
> > FYI, this is the way it is done in ATAG:
> > typedef struct {
> >   UINT32  start;
> >   UINT32  size;
> > } LINUX_ATAG_INITRD2;
> >
> > The boot loader loads the initrd in system memory and passes the
> > memory region information through ATAG.
> >
> >
> > Sorry, if it is the wrong mailing-list. In this case, can someone
> > point me on the right forum for any Linux FDT oriented questions.
>
> Hmmm, it may not be documented (the binding was implemented a
> long time ago before we got strict about documenting
> bindings).  The magic is two properties in the /chosen node;
> "linux,initrd-start" and "linux,initrd-end".  Each contains
> the physical start and end addresses of the initrd.

What does "end address" mean here?

It could be read as meaning either:

a) address of the the final byte of the image
b) address of the first byte following the image (0 if the image ends at 4G)

Cheers
---Dave

-- IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium.  Thank you.

RE: Passing an initrd to a Linux kernel with Device Tree

[Nicolas Pitre]

On Wed, 21 Sep 2011, Dave P Martin wrote:

> 
> 
> > -----Original Message-----
> > From:
> > devicetree-discuss-bounces+dave.martin=arm.com-uLR06cmDAlY8sQ00jmICUQ@public.gmane.org
> g [mailto:devicetree-discuss-> bounces+dave.martin=arm.com-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org] On Behalf Of
> > Grant Likely
> > Sent: 20 September 2011 21:11
> > To: Olivier Martin
> > Cc: devicetree-discuss-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org
> > Subject: Re: Passing an initrd to a Linux kernel with Device Tree
> >
> > On Tue, Sep 20, 2011 at 06:46:22PM +0100, Olivier Martin wrote:
> > > Hello,
> > >
> > > I tried recently to add support in UEFI to pass an initrd
> > with FDT to
> > > a Linux kernel. I have searched the Internet but I did not find any
> > > information to do that.
> > >
> > > FYI, this is the way it is done in ATAG:
> > > typedef struct {
> > >   UINT32  start;
> > >   UINT32  size;
> > > } LINUX_ATAG_INITRD2;
> > >
> > > The boot loader loads the initrd in system memory and passes the
> > > memory region information through ATAG.
> > >
> > >
> > > Sorry, if it is the wrong mailing-list. In this case, can someone
> > > point me on the right forum for any Linux FDT oriented questions.
> >
> > Hmmm, it may not be documented (the binding was implemented a
> > long time ago before we got strict about documenting
> > bindings).  The magic is two properties in the /chosen node;
> > "linux,initrd-start" and "linux,initrd-end".  Each contains
> > the physical start and end addresses of the initrd.
> 
> What does "end address" mean here?
> 
> It could be read as meaning either:
> 
> a) address of the the final byte of the image
> b) address of the first byte following the image (0 if the image ends at 4G)

Looking at the code (remember this is always the ultimate 
documentation), end_address = start_address + size.  So (b) would be the 
right answer.

Obviously, it is adviseable not to locate your ramdisk so close to the 
end of memory so that its end precisely matches the end of the address 
space, and that is if your RAM is located at the top too.


Nicolas

Re: Passing an initrd to a Linux kernel with Device Tree

[Grant Likely]

On Wed, Sep 21, 2011 at 8:25 AM, Nicolas Pitre <nico-vtqb6HGKxmzR7s880joybQ@public.gmane.org> wrote:
> On Wed, 21 Sep 2011, Dave P Martin wrote:
>
>>
>>
>> > -----Original Message-----
>> > From:
>> > devicetree-discuss-bounces+dave.martin=arm.com-uLR06cmDAlY8sQ00jmICUQ@public.gmane.org
>> g [mailto:devicetree-discuss-> bounces+dave.martin=arm.com-uLR06cmDAlZ4lb6lR7t5LA@public.gmane.orgs.org] On Behalf Of
>> > Grant Likely
>> > Sent: 20 September 2011 21:11
>> > To: Olivier Martin
>> > Cc: devicetree-discuss-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org
>> > Subject: Re: Passing an initrd to a Linux kernel with Device Tree
>> >
>> > On Tue, Sep 20, 2011 at 06:46:22PM +0100, Olivier Martin wrote:
>> > > Hello,
>> > >
>> > > I tried recently to add support in UEFI to pass an initrd
>> > with FDT to
>> > > a Linux kernel. I have searched the Internet but I did not find any
>> > > information to do that.
>> > >
>> > > FYI, this is the way it is done in ATAG:
>> > > typedef struct {
>> > >   UINT32  start;
>> > >   UINT32  size;
>> > > } LINUX_ATAG_INITRD2;
>> > >
>> > > The boot loader loads the initrd in system memory and passes the
>> > > memory region information through ATAG.
>> > >
>> > >
>> > > Sorry, if it is the wrong mailing-list. In this case, can someone
>> > > point me on the right forum for any Linux FDT oriented questions.
>> >
>> > Hmmm, it may not be documented (the binding was implemented a
>> > long time ago before we got strict about documenting
>> > bindings).  The magic is two properties in the /chosen node;
>> > "linux,initrd-start" and "linux,initrd-end".  Each contains
>> > the physical start and end addresses of the initrd.
>>
>> What does "end address" mean here?
>>
>> It could be read as meaning either:
>>
>> a) address of the the final byte of the image
>> b) address of the first byte following the image (0 if the image ends at 4G)
>
> Looking at the code (remember this is always the ultimate
> documentation), end_address = start_address + size.  So (b) would be the
> right answer.

Yes, (b) is correct.  I lost 2 days of my life tracking down a bug
last year that turned out to be that I incorrectly assumed (a).

g.