[PATCH 2/2] cpufreq: mediatek: Add MT8173 cpufreq driver

[Pi-Cheng Chen]

Hi Viresh,

On Wed, Jun 24, 2015 at 8:57 AM, Viresh Kumar <viresh.kumar at linaro.org> wrote:
> On 23-06-15, 23:25, Pi-Cheng Chen wrote:
>> On Mon, Jun 22, 2015 at 7:45 PM, Viresh Kumar <viresh.kumar at linaro.org> wrote:
>> >> +static struct mtk_cpu_dvfs_info *mtk_cpu_dvfs_info_get(int cpu)
>> >
>> > A very bad name to a routine with very specific functionality.
>>
>> Would get_mtk_cpu_dvfs_info() or cpu_to_mtk_cpu_dvfs_info() be better?
>>
>> >
>> >> +{
>> >> +     struct mtk_cpu_dvfs_info *info;
>> >> +     struct list_head *list;
>> >> +
>> >> +     list_for_each(list, &cpu_dvfs_info_list) {
>> >> +             info = to_mtk_cpu_dvfs_info(list);
>> >> +
>> >> +             if (cpumask_test_cpu(cpu, info->cpus))
>> >> +                     return info;
>> >> +     }
>
> Firstly there is no need of extra long name for static routines. Just
> keep them simple. i.e. no need of mtk_cpu_.
>
> And then I was a bit confused about the functionality earlier and your
> initial name wasn't that bad. Sorry :(

That's ok. :)

>
> So maybe just, get_dvfs_info().

Yes. Thanks. I'll pick this one.

>
>> >> +     freq_hz = freq_table[index].frequency * 1000;
>> >
>> > A blank line here.
>>
>> Will remove it.
>>
>
> I was asking you to add it :)

I thought you were asking to remove the previous one originally.
Sorry for my misunderstanding.
Will add it.

>
>> >> +     rcu_read_lock();
>
>
>> >> +static int mtk_cpu_dvfs_info_init(int cpu)
>> >> +{
>> >> +     struct device *cpu_dev;
>> >> +     struct regulator *proc_reg = ERR_PTR(-ENODEV);
>> >> +     struct regulator *sram_reg = ERR_PTR(-ENODEV);
>> >> +     struct clk *cpu_clk = ERR_PTR(-ENODEV);
>> >> +     struct clk *inter_clk = ERR_PTR(-ENODEV);
>> >> +     struct mtk_cpu_dvfs_info *info;
>> >> +     struct cpufreq_frequency_table *freq_table;
>> >> +     struct dev_pm_opp *opp;
>> >> +     unsigned long rate;
>> >> +     int ret;
>> >> +
>> >> +     cpu_dev = get_cpu_device(cpu);
>> >> +     if (!cpu_dev) {
>> >> +             pr_err("failed to get cpu%d device\n", cpu);
>> >> +             return -ENODEV;
>> >> +     }
>> >> +
>> >> +     ret = of_init_opp_table(cpu_dev);
>> >> +     if (ret) {
>> >> +             pr_warn("no OPP table for cpu%d\n", cpu);
>> >> +             return ret;
>> >> +     }
>> >> +
>> >> +     cpu_clk = clk_get(cpu_dev, "cpu");
>> >> +     if (IS_ERR(cpu_clk)) {
>> >> +             if (PTR_ERR(cpu_clk) == -EPROBE_DEFER)
>> >> +                     pr_warn("cpu clk for cpu%d not ready, retry.\n", cpu);
>> >> +             else
>> >> +                     pr_err("failed to get cpu clk for cpu%d\n", cpu);
>> >> +
>> >> +             ret = PTR_ERR(cpu_clk);
>> >> +             goto out_free_opp_table;
>> >> +     }
>> >> +
>> >> +     inter_clk = clk_get(cpu_dev, "intermediate");
>> >> +     if (IS_ERR(inter_clk)) {
>> >> +             if (PTR_ERR(inter_clk) == -EPROBE_DEFER)
>> >> +                     pr_warn("intermediate clk for cpu%d not ready, retry.\n",
>> >> +                             cpu);
>> >> +             else
>> >> +                     pr_err("failed to get intermediate clk for cpu%d\n",
>> >> +                            cpu);
>> >> +
>> >> +             ret = PTR_ERR(cpu_clk);
>> >> +             goto out_free_resources;
>> >> +     }
>> >> +
>> >> +     proc_reg = regulator_get_exclusive(cpu_dev, "proc");
>> >> +     if (IS_ERR(proc_reg)) {
>> >> +             if (PTR_ERR(proc_reg) == -EPROBE_DEFER)
>> >> +                     pr_warn("proc regulator for cpu%d not ready, retry.\n",
>> >> +                             cpu);
>> >> +             else
>> >> +                     pr_err("failed to get proc regulator for cpu%d\n",
>> >> +                            cpu);
>> >> +
>> >> +             ret = PTR_ERR(proc_reg);
>> >> +             goto out_free_resources;
>> >> +     }
>> >> +
>> >> +     /* Both presence and absence of sram regulator are valid cases. */
>> >> +     sram_reg = regulator_get_exclusive(cpu_dev, "sram");
>> >> +
>> >> +     info = kzalloc(sizeof(*info), GFP_KERNEL);
>> >> +     if (!info) {
>> >> +             ret = -ENOMEM;
>> >> +             goto out_free_resources;
>> >> +     }
>> >> +
>> >> +     ret = dev_pm_opp_init_cpufreq_table(cpu_dev, &freq_table);
>> >> +     if (ret) {
>> >> +             pr_err("failed to init cpufreq table for cpu%d: %d\n",
>> >> +                    cpu, ret);
>> >> +             goto out_free_mtk_cpu_dvfs_info;
>> >> +     }
>> >> +
>> >> +     if (!alloc_cpumask_var(&info->cpus, GFP_KERNEL))
>> >
>> > Why getting into such trouble? What about just saving policy's pointer
>> > in info and use it for freq_table, cpus mask, etc ?
>>
>> This function is called from driver probe function. At this moment, cpufreq
>> driver is not yet registered and the policy struct for each cluster is not yet
>> allocated.
>
> This should be called from ->init() callback of the cpufreq driver.

Some reasons I did this in probe(). Please see below.

>
>> Also, the mtk_cpu_dvfs_info structs for all clusters are stored in a list
>> after probe done. cpus mask is used to find out the proper info struct
>> for corresponding cpu.
>
> Yeah, but this will be exactly same to policy->related_cpus, just use
> that. Why allocate another mask for same purpose?

Yes. I agree. Please see the reasons why I did it below.

>
>> For freq_table, I simply don't want to free and allocate the table
>> every time when cpu is unplugged and plugged so I just allocate it
>> in the probe function call path, and cache it in the info struct.
>
> ->exit() will be called only when all the CPUs from a policy are
> hotplugged out. Not on every hotplug. So you probably have 4 big and 4
> little cores. So policy for big cluster will be removed only when all
> the CPUs are down. And so no point storing things from probe.
>
> Or is there anything special with your usecase?

One reason to put those initialization and resource allocation in probe is
that it's easier to handle the return value -PROBE_DEFER from clock
and regulator framework when trying to get clocks and regulators
consumed by cpufreq driver.

The other reason is when the whole system is resuming from suspend,
the other subsystem e.g. i2c on which cpufreq driver depends might not
ready yet during cpufreq driver initialization. In this case, the cpufreq
driver will be blocked when trying to get resources e.g. regulator on i2c
bus, and the whole system will stuck for seconds.

For the two reasons above, I try to do all resource initialization and
allocation in probe(), and leave less things to do in init() as possible.
Do you have any suggestion for this?

>
>> But for module_platform_driver(mt8173_cpufreq_platdrv), I saw many
>> built-in drivers in kernel call module_platform_driver() also so I use it.
>
> That was wrong for them as well :)
>
>> Would it be better to use platform_driver_register() instead?
>
> Hmm..

Will fix it.

Thanks.
Pi-Cheng
>
> --
> viresh