Comment # 15 on bug 1196632 from Ivan Ivanov

(In reply to Stefan Wahren from comment #14)
> AFAIK the latest firmware has a workaround for the invalid index issue.
> 

Hm, not sure I get this. Currently RPI_FIRMWARE_GET_CLOCKS return
15 clocks. From 1 to 15 inclusive, so number 15 (VEC) clock is among them.
Except that we do out-of-bound access in clks array VEC clock is regitered 
successfully in the clock provider.

> Yes, the real issue comes from the downstream DTB. It uses the VEC clock
> from the raspberrypi clock and not from clk-bcm2835 driver. This is
> different from the mainline DTB.
>

I can't find at the moment where I found this sentence, or slightly
modified version, but:

"..
Platform driver clk-bcm2835 gets an inaccurate clock for VEC (107MHz).
Export VEC clock trough clk-raspberrypi which uses the right PLL to
get an accurate 108MHz.
..."

And somehow I think this was related around 4K60Hz support addition.
To me looks like we want to use clk-raspberrypi for VEC clock proivider.