Re: [meta-raspberrypi][PATCH 1/5] devicetree: auto-disable dts for old kernels

["Petter Mabäcker" <petter@technux.se>]

[-- Attachment #1: Type: text/plain, Size: 4957 bytes --]

On 05/10/2015 04:24 PM, Andrei Gherzan wrote:
>
>
> On Sun, May 10, 2015 at 4:20 PM, Petter Mabäcker <petter@technux.se 
> <mailto:petter@technux.se>> wrote:
>
>     Hi Andrei,
>
>     See answers inline.
>
>     BR,
>     Petter
>
>
>     On 05/10/2015 12:16 PM, Andrei Gherzan wrote:
>>     Hello Petter,
>>
>>     On Fri, May 8, 2015 at 11:49 PM, Petter Mabäcker
>>     <petter@technux.se <mailto:petter@technux.se>> wrote:
>>
>>         After '6392a63 rpi-base.inc: Use KERNEL_DEVICETREE by
>>         default' was
>>         introduced, kernel versions < 3.18 might not be buildable.
>>         Since full
>>         device tree support was introduced in 3.18 this change
>>         ensures that all
>>         kernel < 3.18 will automatically disable device tree.
>>
>>         Signed-off-by: Petter Mabäcker <petter@technux.se
>>         <mailto:petter@technux.se>>
>>         ---
>>          classes/linux-raspberrypi-base.bbclass  | 29
>>         +++++++++++++++++++++++++++++
>>          classes/sdcard_image-rpi.bbclass  |  6 ++++--
>>          recipes-kernel/linux/linux-raspberrypi.inc |  4 +++-
>>          3 files changed, 36 insertions(+), 3 deletions(-)
>>          create mode 100644 classes/linux-raspberrypi-base.bbclass
>>
>>         diff --git a/classes/linux-raspberrypi-base.bbclass
>>         b/classes/linux-raspberrypi-base.bbclass
>>         new file mode 100644
>>         index 0000000..4bcadd0
>>         --- /dev/null
>>         +++ b/classes/linux-raspberrypi-base.bbclass
>>         @@ -0,0 +1,29 @@
>>         +inherit linux-kernel-base
>>         +
>>         +
>>         +def get_dts(d, ver):
>>         +    staging_dir = d.getVar("STAGING_KERNEL_BUILDDIR", True)
>>         +    dts = d.getVar("KERNEL_DEVICETREE", True)
>>         +
>>         +    # d.getVar() might return 'None' as a normal string
>>         +    # leading to 'is None' check isn't enough.
>>         +    # TODO: Investigate if this is a bug in bitbake
>>         +    if ver is None or ver == "None":
>>         +        ''' if 'ver' isn't set try to grab the kernel version
>>         +        from the kernel staging '''
>>         +        ver = get_kernelversion_file(staging_dir)
>>         +
>>         +    if ver is not None:
>>         +        min_ver = ver.split('.', 3)
>>         +    else:
>>         +        return dts
>>         +
>>         +    # Always turn off device tree support for kernel's < 3.18
>>         +    try:
>>         +        if int(min_ver[0]) <= 3:
>>         +            if int(min_ver[1]) < 18:
>>         +                dts = ""
>>         +    except IndexError:
>>         +        min_ver = None
>>         +
>>         +    return dts
>>         diff --git a/classes/sdcard_image-rpi.bbclass
>>         b/classes/sdcard_image-rpi.bbclass
>>         index 1ff664d..ca94566 100644
>>         --- a/classes/sdcard_image-rpi.bbclass
>>         +++ b/classes/sdcard_image-rpi.bbclass
>>         @@ -1,4 +1,5 @@
>>          inherit image_types
>>         +inherit linux-raspberrypi-base
>>
>>          #
>>          # Create an image that can by written onto a SD card using dd.
>>         @@ -88,7 +89,8 @@ IMAGE_CMD_rpi-sdimg () {
>>                 ROOTFS_SIZE_ALIGNED=$(expr ${ROOTFS_SIZE_ALIGNED} -
>>         ${ROOTFS_SIZE_ALIGNED} % ${IMAGE_ROOTFS_ALIGNMENT})
>>                 SDIMG_SIZE=$(expr ${IMAGE_ROOTFS_ALIGNMENT} +
>>         ${BOOT_SPACE_ALIGNED} + ${ROOTFS_SIZE_ALIGNED})
>>
>>         -       echo "Creating filesystem with Boot partition
>>         ${BOOT_SPACE_ALIGNED} KiB and RootFS ${ROOTFS_SIZE_ALIGNED} KiB"
>>
>>
>>     Why do you remove this line? Maybe by mistake?
>     Yes this is by mistake, thanks for finding it. Will send up a new
>     version with this fixed.
>
> Don't bother. I'll fix it before merging.
Ok, thanks.
>
>>         +      # Check if we are building with device tree support
>>         +       DTS="${@get_dts(d, None)}"
>>
>>
>>     Aren't we able to use the same variable name, KERNEL_DEVICETREE?
>>     What that be a little clearer?
>     No, at least i couldn't get this working. When trying to set the
>     environment variable KERNEL_DEVICETREE in this context the values
>     was only ignored. Don't know if the image recipe (or at least the
>     IMAGE_CMD func) are doing something special. Perhaps it's possible
>     to use 'eval' or something similar to enforce the expansion of the
>     variable. If you have any good advice please share them, otherwise
>     I can at least write some comment about this.
>
>
> I'll try it a little. But I didn't expect this. Will see. Otherwise 
> this is fine.
Neither did I, let me know if you find anything interesting about this.
>
>
> -- 
> *Andrei Gherzan*
> *e: */andrei@gherzan.ro <mailto:andrei@gherzan.ro>/
> /*w:* /www.gherzan.ro <http://www.gherzan.ro>


[-- Attachment #2: Type: text/html, Size: 13108 bytes --]