Re: [PATCH] drivers/of: add option to load a default Device Tree

[Frank Rowand <frowand.list-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

On 06/30/16 13:44, Franck Jullien wrote:
> 2016-06-30 19:45 GMT+02:00 Frank Rowand <frowand.list-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>:
>> Hi Franck,
>>
>> (This could be confusing, with Franck and Frank -- feel free to
>> address me as rowand in this thread if it is less confusing.)
>>
>> On 06/22/16 08:05, Franck Jullien wrote:
>>> Even if a platform doesn't use a device tree during its
>>> boot process it can be useful to enable CONFIG_OF and get
>>> an empty device tree.
>>>
>>> Then, devices can use device tree overlays to populate this
>>> default tree.
>>>
>>> Signed-off-by: Franck Jullien <franck.jullien-EduPiq9onwdphFt5fpzH3laPQRlvutdw@public.gmane.org>
>>> ---
>>>  drivers/of/Kconfig     |  9 +++++++++
>>>  drivers/of/Makefile    |  3 +++
>>>  drivers/of/base.c      | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++
>>>  drivers/of/default.dts |  4 ++++
>>>  drivers/of/unittest.c  | 33 +++++----------------------------
>>>  include/linux/of.h     |  2 ++
>>>  6 files changed, 73 insertions(+), 28 deletions(-)
>>>  create mode 100644 drivers/of/default.dts
>>
>> < snip >
>>
>> For context, in a later reply, you mention that this is for x86_64.
>>
>> My current inclination is to prefer not to solve the problem this way.
>> I was going to ask why you didn't just add the compiled default.dts
>> to the kernel as an appended device tree blob, until I found out
>> this was for x86_64.
>>
>> Can you add unflatten_devicetree() to x86_64, then just use the
>> appended device tree blob method?
>>
> 
> You're saying I should implement the ability for x86_64 to detect and
> load a devicetree that is appended to the binary image like ARM does [1] ?

Yes, that is what I was referring to.


> Would you keep CONFIG_DEFAULT_DTB at all ? Or appending an
> empty (or not) devicetree would be after the binary image is built ?
> 
> I like the idea. I think I could remove CONFIG_DEFAULT_DTB and
> let the user append a blob to the image.

There should be a config option to add the Makefile entry for the
default.dts so that default.dtb is created.

And yes, the .dtb would be appended to the kernel image after the
kernel image was built.

One description of this is at:

http://community.cadence.com/cadence_blogs_8/b/sd/archive/2013/10/09/combining-the-linux-device-tree-and-kernel-image-for-arm


> 
> By the way, why having an empty devicetree built-in is not your
> preferred choice ?

The issue is not whether the .dtb is build into the kernel image or
appended to the kernel image.  The issue is whether the normal
unflatten_devicetree()  path is used to load the .dtb or whether
an additional method of creating the root device tree was added.


> Franck.
> 
> [1] http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=e2a6a3aafa9862c4a4b59f2a59b8f923d64a680e
> .
> 

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html