Re: [PATCH v5 7/7] clk: Add floor and ceiling constraints to clock rates

[Tomeu Vizoso]

On 14 November 2014 08:50, Stephen Boyd <sboyd@xxxxxxxxxxxxxx> wrote:
> On 10/30, Tomeu Vizoso wrote:
>> @@ -2145,6 +2218,16 @@ struct clk *__clk_register(struct device *dev, struct clk_hw *hw)
>>  }
>>  EXPORT_SYMBOL_GPL(__clk_register);
>>
>> +static void __clk_free_clk(struct clk *clk)
>> +{
>> +     struct clk_core *core = clk->core;
>> +
>> +     hlist_del(&clk->child_node);
>
> Does this race with aggregation in clk_core_set_rate()? I would think
> we need to hold the prepare lock here so we don't traverse the list
> while it's being modified?

Yes, thanks.

>> +     kfree(clk);
>> +
>> +     clk_core_set_rate(core, core->req_rate);
>> +}
>> +
>>  /**
>>   * clk_register - allocate a new clock, register it and return an opaque cookie
>>   * @dev: device that is registering this clock
>> diff --git a/include/linux/clk.h b/include/linux/clk.h
>> index c7f258a..0d55570 100644
>> --- a/include/linux/clk.h
>> +++ b/include/linux/clk.h
>> @@ -302,6 +302,24 @@ long clk_round_rate(struct clk *clk, unsigned long rate);
>>  int clk_set_rate(struct clk *clk, unsigned long rate);
>>
>>  /**
>> + * clk_set_floor_rate - set a minimum clock rate for a clock source
>> + * @clk: clock source
>> + * @rate: desired minimum clock rate in Hz
>> + *
>> + * Returns success (0) or negative errno.
>> + */
>> +int clk_set_floor_rate(struct clk *clk, unsigned long rate);
>> +
>> +/**
>> + * clk_set_ceiling_rate - set a maximum clock rate for a clock source
>> + * @clk: clock source
>> + * @rate: desired maximum clock rate in Hz
>> + *
>> + * Returns success (0) or negative errno.
>> + */
>> +int clk_set_ceiling_rate(struct clk *clk, unsigned long rate);
>> +
>
> I still don't see anything to do with clk_round_rate()?

Yeah sorry, i don't really have any good ideas, and was kind of hoping
that Mike would comment.

> It's
> possible that whatever is constrained at this user level goes
> down to the hardware driver and then is rounded up or down to a
> value that is outside of the constraints, in which case the
> constraints did nothing besides control the value that the
> hardware driver sees in the .round_rate() op. I doubt that was
> intended.

Indeed. Wonder what can be done about it with the least impact on
existing code. I see the situation as clk implementations being able
to apply arbitrary constraints in determine_rate() and round_rate(),
and they would need to take into account the per-user constraints so
they can all be applied consistently.

> I also wonder what we should do about clocks that are in the
> middle of the tree (i.e. not a leaf) and have constraints set on
> them. It looks like if a child of that clock can propagate rates
> up to the parent that we've constrained with the per-user
> constraints, those constraints won't be respected at all, leading
> to a hole in the constraints.

True. Do we want to support per-user constraints on non-leaf clocks?

> I imagine both of these points don't matter to the emc clock
> scaling patch

That's right.

> (BTW is there some pointer to that and the usage of
> these APIs?)

There's the proposed ACTMON driver, that currently is the solely user
of the EMC clock, so it's using clk_set_rate for now:

http://thread.gmane.org/gmane.linux.kernel/1816846/focus=1826037

In the future, it would set a floor constraint, and drivers such as
thermal and battery (when we have them) would set ceiling rates.

This is the last version of the EMC clock for Tegra124:

https://lkml.org/lkml/2014/11/18/272

> because that is only dealing with a leaf clock that
> doesn't care about clk_set_rate() being used along with
> constraints and the rounding behavior doesn't violate a floor?
>
> I'm all for having a clk_set_rate_range() API (and floor/ceiling
> too), but it needs to be done much deeper in the core framework
> to actually work properly. Having a range API would make all the
> confusion about how a particular clock driver decides to round a
> rate go away and just leave an API that sanely says the rate will
> be within these bounds or an error will be returned stating it
> can't be satisfied.

This sounds like a good way forward, but TBH I don't understand what
you are proposing. Would you care to elaborate on how the API that you
propose would look like?

Thanks,

Tomeu

> This would be useful so we don't have a bunch
> of drivers littered with code that loops on clk_round_rate() to
> figure out what their hardware actually supports or having some
> hard-coded frequency table per driver because the hardware can't
> generate some frequency that's part of a spec (but still lies
> within some acceptable tolerance!). It would also make
> clk_round_rate() mostly obsolete because we know the rate is
> within whatever acceptable bounds we've chosen. Eventually
> clk_set_rate() could be become a small wrapper on top of
> clk_set_rate_range(), constraining the rate to be exactly
> whatever the clock driver returns as the rounded rate.
>
> --
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
> a Linux Foundation Collaborative Project
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/