I'd like to get some feedback on our idea as well as check
if somebody faces similar situations and if so what would be the best
way to implement some generic solution that suits everyone.

So that's our problem:
1. On power-on hardware might start clocking CPU with either
   too high frequency (such that CPU may get stuck at some point)
   or too low frequency.

   That all sounds stupid but let me elaborate a bit here.
   I'm talking about FPGA-based devboards firmware for which
   (here I mean just image loaded in FPGA with CPU implementation
   but not some software yet) might not be stable or be even experimental.

   For example we may deal with dual-core or quad-core designs.
   Former might be OK running @100MHz and latter is only usable
   @75MHz and lower. The simplest solution might be to use some safe
   value before something like CPUfreq kicks in. But we don't yet have
   CPUfreq for ARC (we do plan to get it working sometime soon) which
   means simple change of CPU frequency once time-keeping infrastructure
   was brought-up is not an option... I.e. we'll end up with the system running
   much slower compared what could have been possible.

2. Up until now we used to do dirty hacks in early platform init code.
   Namely (see axs103_early_init() in arch/arc/plat-axs10x/axs10x.c):
    1) Read CPU's "clock-frequency" from .dtb (remember we're on very early
       boot stage still so no expanded DevTree yet exists).
    2) Check how many cores we have and which freq is usable
    3) Update PLL settings right in place if new freq != existing in PLL.

   Even though it is proven to work but with more platforms in the pipeline
   we'll need to copy-paste pretty much the same stuff across all affected
   plats. Which is not nice.

   Moreover back in the day we didn't have a proper clk driver for CPU's PLL.
   Thus acting on PLL registers right in place was the only thing we were able
   to do. Now with introduction of normal clk driver
   (see drivers/clk/axs10x/pll_clock.c in linux-next) we'd like to utilize
   it and have a cleaner and more universal solution to the problem.

   That's how it could be done - http://patchwork.ozlabs.org/patch/801240/
   Basically in architecture's time_init() we check if there's explicitly
   specified "clock-frequency" parameter in cpu's node in Device Tree and
   if there's one we set it via just instantiated clk driver.

We may indeed proceed with mentioned solution for ARC but if that makes
sense for somebody else it might worth getting something similar in generic
init code. Any thoughts?

All comments and suggestions are more than welcome.

linux-snps-arc mailing list

Reply via email to