On 09/05/2017 03:04 PM, Rob Herring wrote:
On Tue, Sep 5, 2017 at 10:37 AM, Alexey Brodkin
<Alexey.Brodkin@xxxxxxxxxxxx> wrote:
Hello,
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)
But even if we had cpufreq driver going - I don't think it would be usable for
doing large freq switches, since in current implementations of SoCs (or fpga), the
clk/pll etc driving core (and all timers etc) are not fixed like say ARM. And as
discussed before (and pointed to by tglx), timer subsys can't tolerate (on
purpose) such large drifts.
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 - https://urldefense.proofpoint.com/v2/url?u=http-3A__patchwork.ozlabs.org_patch_801240_&d=DwICAg&c=DPL6_X_6JkXFx7AXWqB0tg&r=c14YS-cH-kdhTOW89KozFhBtBJgs1zXscZojEZQ0THs&m=wuUcceY8Cz5EhVklWLAgj7RzU3rvpanujvQ3qTJK0Gw&s=N5IBjq_eCyOf_GRkZskzqGhczBPTbxLJW_MUfauKvuA&e=
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.
The patch looks generally okay. I'd move all the logic to the clock
driver unless perhaps how to set the cpu freq is defined by the
architecture.
But the above patch is clk driver agnostic and it would have to be added each clk
driver (axs10x, hsdk - both in linux-next) ?
Also note that this code is using a new / adhoc DT binding cpu-freq in cou node to
do the override - is that acceptable ?
-Vineet
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?
If any ARM platforms are doing something similar, then it's done in
their clock driver via of_clk_init. Or they just wait for cpufreq to
kick in and expect the bootloader has initialized things to a
reasonable frequency.
