[PATCH 2/7] ARM: dts: skeleton: add unit name to memory node

[Rob Herring]

On Thu, Mar 31, 2016 at 11:34 AM, Joachim  Eastwood <manabian at gmail.com> wrote:
> On 31 March 2016 at 17:21, Joachim  Eastwood <manabian at gmail.com> wrote:
>> On 31 March 2016 at 12:38, Mark Rutland <mark.rutland at arm.com> wrote:
>>> On Wed, Mar 30, 2016 at 10:45:06PM +0200, Joachim Eastwood wrote:
>>>> On 30 March 2016 at 19:06, Mark Rutland <mark.rutland at arm.com> wrote:
>>>> > On Wed, Mar 30, 2016 at 06:15:35PM +0200, Joachim Eastwood wrote:
>>>> >> I used the following script to check for the memory node in all built dtb's.
>>>> >>   make ARCH=arm CONFIG_OF_ALL_DTBS=y dtbs
>>>> >>   for i in $(ls arch/arm/boot/dts/*.dtb); do
>>>> >>          m=$(scripts/dtc/dtc -I dtb -O dts $i | grep -m1 'memory.*{')
>>>> >>          if [ -z "$m" ]; then
>>>> >>                  echo "Missing memory node in $i"
>>>> >>           fi
>>>> >>   done
>>>> >>
>>>> >> So it should be pretty safe to just remove the memory node entry in
>>>> >> the skeleton files. Unless I have missed something with the script
>>>> >> above.
>>>> >
>>>> > The above might match reserved-memory nodes; it might be better to check
>>>> > for 'device_type\s*=\s*"memory"'.
>>>>
>>>> I did check the output of the grep and it looks good. But there are
>>>> indeed DTs that are missing the 'device_type = "memory"' parameter.
>>>> Actually; _a lot_ or 438 of 741 to be exact. ugh...
>>>>
>>>> I guess all those should be fixed up before we can remove the memory
>>>> node from skeleton. :/
>>>
>>> Ouch, yes. :(
>>>
>>> That said, the cahnges don't need to be an atomic operation. We could
>>> start adding device_type = "memory" to dts immediately (in as whatever
>>> size batches maintainers are happy with), as a duplicate device_type
>>> shouldn't be problematic.
>>
>> Yes, that is true.
>>
>>
>>> When we hit critical mass, we could then remove the skeleton memory
>>> nodes, fixing up any remaining fallout.
>>>
>>> As for the mechanical changes, it sounds like we need coccinelle for DT.
>>>
>>> That, or a laptop, a long flight, and a gin and tonic.
>>
>> :-)
>>
>> I'll see if I can cook up something with awk.
>
> Something like this should do it:
> #!/usr/bin/gawk -f
> # find arch/arm/boot/dts/ -type f -name *.dts* | xargs
> ./mem_node_add_dev_type.awk -i inplace
> BEGIN { go = 0 }
> /[^-]memory ?{/ { go = 1; idx = 0; device_type = 0 }
>
> go == 1 {
>         buf[idx++] = $0
>         if ($0 ~ /device_type/)
>                 device_type = 1
>
>         if ($0 ~ /};/) {
>                 print buf[0]
>
>                 if (!device_type) {
>                         buf[0] = buf[1]
>                         gsub(/[^\t]*/, "", buf[0])
>                         buf[0] = buf[0] "device_type = \"memory\";"
>                 } else {
>                         delete buf[0]
>                 }
>
>                 for (i in buf)
>                         print buf[i]
>                 delete buf
>
>                 go = 0
>                 device_type = 0
>         }
>         next
> }
>
> { print }
>
> Gives me: 249 files changed, 249 insertions(+), 1 deletion(-) on v4.6-rc1.
>
> Running the my check script after this reveals that 83 DTs lack the
> device_type parameter, but these seem to also lack the memory node as
> well. Wonder why my script didn't pick up the missing memory node in
> those... I also doubt that the memory node is set by the boot loader
> in some of these boards.

I don't think this script is right. It will not work if the memory
node is spread across dtsi files. You would need to compiler
everything to dtb and then back to a flat dts file before running thru
this script.

I don't see the rest of the series in -next. Is that going to happen soon?

Rob

>
>
> regards,
> Joachim Eastwood