On 04/06/2016 11:05 PM, Sergei Shtylyov wrote:
I've got a Porter board (Revision B) which stopped showing
serial
output since the patch mentioned in the subject. The terrible
clearly
wrong hack below gets serial back (reverting to just the fck
clock
for
scif0) on both linux-next and linux master.
Thanks for your report!
I assume something is wrong with the SCIF_CLK input.
Hence I guess the issue is also fixed by removing the "&scif_clk"
block in
arch/arm/boot/dts/r8a7791-porter.dts?
Confusingly it didn't, see the patch i just sent out:
ARM: dts: r8a7791: Don't disable referenced optional clocks
That patch looks, er... strange/incomplete. Perhaps we should
fix
clk_get() instead?
Not sure how it's incomplete?
First, R8A7791 is not the only SoC having disabled clocks in its .dtsi
file. And if you remove the status = "disabled" in the SoC file, please also
remove "enabled" in the board files as that is the default node status anyway.
>
But see also Geerts comment on the
relevant thread where he nicely pinpoint what triggered that behaviour.
That said I do find it a bit odd to refer to disabled clock nodes, but
that might be me.
I don't find it odd. What I found odd is clk_get() returning -EPROBE_DEFER.
After looking into this issue myself, I must admit that disabled clocks
were a bad idea. The kernel can't differ disabled device node from no node at
all (on this level, at least).
So, will you respin your patch and fix the other SoCs?
[...]
MBR, Sergei
