[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <9365247a-8aa0-bad5-c619-9d5a984b17de@linaro.org>
Date: Tue, 10 May 2022 14:54:05 +0200
From: Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
To: Andrew Lunn <andrew@...n.ch>
Cc: Chris Packham <Chris.Packham@...iedtelesis.co.nz>,
"catalin.marinas@....com" <catalin.marinas@....com>,
"will@...nel.org" <will@...nel.org>,
"robh+dt@...nel.org" <robh+dt@...nel.org>,
"krzysztof.kozlowski+dt@...aro.org"
<krzysztof.kozlowski+dt@...aro.org>,
"gregory.clement@...tlin.com" <gregory.clement@...tlin.com>,
"sebastian.hesselbarth@...il.com" <sebastian.hesselbarth@...il.com>,
"kostap@...vell.com" <kostap@...vell.com>,
"robert.marko@...tura.hr" <robert.marko@...tura.hr>,
"linux-arm-kernel@...ts.infradead.org"
<linux-arm-kernel@...ts.infradead.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"devicetree@...r.kernel.org" <devicetree@...r.kernel.org>
Subject: Re: [PATCH v5 1/2] arm64: dts: marvell: Add Armada 98DX2530 SoC and
RD-AC5X board
On 10/05/2022 14:37, Andrew Lunn wrote:
> On Tue, May 10, 2022 at 09:08:08AM +0200, Krzysztof Kozlowski wrote:
>> On 10/05/2022 06:14, Chris Packham wrote:
>>>
>>> Based on the information I have (which isn't much) there is a ref_clk
>>> input that is connected to a 25MHz oscillator and then I'm assuming
>>> these are all generated from that with various dividers. 25MHz is the
>>> only documented option.
>>>
>>> There doesn't appear to be any documented register where I can read out
>>> the divider ratios. It might be nice I could have the fixed osc node and
>>> have these 3 clocks derived with fixed divisors but I don't see any what
>>> of achieving that.
>>
>>
>> OK, but where are the dividers? The ref_clk is outside of SoC, so should
>> be defined in board DTS (at least its rate). If the rest is in the SoC,
>> they are usually part of clock controller, because usually they belong
>> to some power domain or have some clock gating.
>
> 25MHz is a 'magic value' in Ethernet, nearly everything is based
> around it. And remember this SoC is basically an Ethernet switch with
> a small CPU glued on one side. If you gated clocks derived from the
> 25MHz reference clock, probably part of your Ethernet switch stops
> working, which is the whole point of this SoC. So i doubt there are
> gates on the derived clocks. If there is any gating and power domains,
> it is generally at a different level. You can power down individual
> ports of the Ethernet switch. But generally, there is one bit in a
> register somewhere to do that, and you don't have direct control over
> clocks and regulators etc.
The 25 MHz input clock I understand, it was about other clocks, like
spi, axi and core. These clearly look like part of SoC, so defining them
with a "stubs" (uncontrollable fixed-clock) is not the best way of
modelling an SoC. Although maybe this SoC does not have a proper clock
controller and even SPI and AXI clocks are always on?
Best regards,
Krzysztof
Powered by blists - more mailing lists