On 07/12/2012 08:08 AM, Sebastian Hesselbarh wrote:
> On 07/12/2012 02:14 PM, Rob Herring wrote:
>>> +Required child properties:
>>> +- reg : should contain the individual bit and polarity to control
>>> + the clock gate. A polarity of 0 means that by setting the
>>> + bit to 1 the clock passes through the clock gate while
>>> + setting the bit to 0 disables the clock. Any other value
>>> + for polarity inverts the meaning of the control bit.
>>
>> This is a bit of overloading reg to specify the polarity.
>
> Well, yes it is overloading but still matches reg somehow, as the
> extra information is required to access the resource. But I agree,
> expecially wrt more-than-one-bit clk-gate (see below).
>
You can define your own property names.
>>> + /* SATA clock gate with different parent clock */
>>> + cg_sata: clockgate@3 {
>>> + reg =<3 0>; /* register bit 3, normal polarity */
>>> + clocks =<&sata_clk>;
>>> + };
>>
>> I'm not sure I like the node per bit. What about a bit mask for valid
>> bits and polarities. Then add a clock cell to specify the bit or index.
>>
>> i.MX has 2-bit enable fields for its leaf clocks, so how and if you
>> would support that is something to think about.
>
> Yeah, I thought of "what if the clk_gate needs to be enabled with more
> than 1 bit" already. But this is a short-comming of the current clk-gate
> implementation.
What's implemented in Linux should not define the binding. The binding
should describe the hardware.
> Just to get it right, i.MX requires to set more than one bit to change
> the state of the gate for one leaf clock?
It's basically ON, OFF, and "ON in run/OFF in wfi".
Perhaps the iMX case is unique enough we don't try to make it use a
common binding.
> If this is true, that would require a change of the generic clk-gate
> anyway.
True, but not your problem to implement. A binding doesn't necessarily
mean there is a full Linux implementation. We just don't want to create
something only to find others need something completely different.
Rob
> I had a look at pinctrl-bindings.txt maybe this is the way to go for
> clock gating control, too. That would require clk-gate to handle an
> 'active' and 'gated' state and leave it to a clock gate control to
> actually set the required bits in any registers. This would allow
> other special implementations of clock gating controllers to reuse
> clk-gate DT description. Additionally, there could be a
> simple-clock-gating-control that can set states by reg address and
> for each controlled gate a mask, enable value, and disable value.
>
> Sebastian
--
To unsubscribe from this list: send the line "unsubscribe linux-doc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
