Re: [PATCH 0/5] da8xx USB clocks

From: Sergei Shtylyov
Date: Wed Mar 16 2016 - 08:37:03 EST

On 3/16/2016 1:35 AM, David Lechner wrote:

I've implemented some proper clocks for the USB PHY clocks on da8xx. For those
not familiar with the architecture, the SoC has one USB 1.1 OHCI port and one
USB 2.O OTG port. The USB 1.1 PHY clock can optionally be supplied by the PLL
in the USB 2.0 PHY.

I have seen some comments in the past that these clock details don't belong in
the USB drivers and I agree with that. So, I have moved the handling of the
clocks out of the USB drivers to the mach code with the rest of the SoC clocks.

This code has been tested on LEGO MINDSTORMS EV3 (AM1808/da850 family). Here is
an output of the davinci clock debug to give you a better idea of how clocks
are related.

root@ev3dev:~# cat /sys/kernel/debug/davinci_clocks
ref_clk users=23 24000000 Hz
pll0 users=20 pll 300000000 Hz
pll0_aux_clk users= 4 pll 24000000 Hz
usb20_phy users= 2 24000000 Hz
usb11_phy users= 1 24000000 Hz
usb_ref_clk users= 0 48000000 Hz

usb20_phy and usb11_phy can optionally be children of usb_ref_clk instead.

I'm planning on adding device tree bindings for the ohci driver, but I need to
get some things sorted out with the regulator subsystem first. I see that Petr
has been working on device tree support for the musb driver. This should take
care of some of the concerns related to his changes too, for example, the
ti,usb2-phy-refclock-hz device tree property is no longer needed because it
is now taken care of in the clock code. I've actually included one of Petr's
patchs here since one of my patches depends on it.

I'm also working on device tree bindings for davinci clocks, but it will take
me a while to get there. But that should not hold up the device tree bindings
for da8xx ohci and musb.

Nice to see that someone still cares about this code. With the ending of the MontaVista's efforts, it largely seemed like abandonware...

MBR, Sergei