
Hi Tom,
On 5/3/19 4:07 PM, Tom Rini wrote:
On Fri, May 03, 2019 at 01:28:02PM +0200, Niel Fourie wrote: [snip]
diff --git a/arch/arm/dts/am33xx-u-boot.dtsi b/arch/arm/dts/am33xx-u-boot.dtsi index 78f5e2c4d3..a58734e441 100644 --- a/arch/arm/dts/am33xx-u-boot.dtsi +++ b/arch/arm/dts/am33xx-u-boot.dtsi
Sorry, I had not fully understood the implications of moving these changes to am33xx-u-boot.dtsi, so will move them back to am335x-wega-rdk-u-boot.dtsi where they belong. As Marek also pointed out, they belong in a seperate patch from this one anyways.
@@ -6,6 +6,17 @@
/ { ocp {
u-boot,dm-pre-reloc;u-boot,dm-spl;
First off, this is wrong. In the code (see drivers/core/ofnode.c and drivers/core/util.c) everywhere we check for u-boot,dm-spl we've already checked for u-boot,dm-pre-reloc and returned true. So all of these kind of changes above need to be fixed.
Thanks, now I understand the mechanism. Will fix!
};
- memory@80000000 {
u-boot,dm-spl;
u-boot,dm-pre-reloc;
- };
+};
+&scm {
- u-boot,dm-spl;
- u-boot,dm-pre-reloc; };
Second, where are you seeing that we need these nodes to be moved to be available earlier? Thanks!
In my case, I needed "ocp" because it is the parent of mmc1, and I used DM wherever I could, including for MMC in the SPL. I needed "memory" because the RAM size is specified in the device tree for RAM initialisation. After double checking, it appears that scm is in fact redundant (I included it for DDR3 and pin muxing). But, as this is likely not the case for everyone, I will move them back to am335x-wega-rdk-u-boot.dtsi as mentioned before.
Thanks for your feedback.
Best regards, Niel Fourie