Hi, Geert,
On 08.10.2024 16:23, Geert Uytterhoeven wrote:
> Hi Claudiu,
>
> On Thu, Aug 22, 2024 at 5:28 PM Claudiu <claudiu.beznea@xxxxxxxxx> wrote:
>> From: Claudiu Beznea <claudiu.beznea.uj@xxxxxxxxxxxxxx>
>>
>> Now that we have a driver for SYSC driver for RZ/G3S move the SoC detection
>> for RZ/G3S in SYSC driver.
>>
>> Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@xxxxxxxxxxxxxx>
>
> Thanks for your patch!
>
>> --- a/drivers/soc/renesas/rzg3s-sysc.c
>> +++ b/drivers/soc/renesas/rzg3s-sysc.c
>> @@ -85,6 +97,39 @@ static int rzg3s_sysc_probe(struct platform_device *pdev)
>> sysc->dev = dev;
>> spin_lock_init(&sysc->lock);
>>
>> + compatible = of_get_property(dev->of_node, "compatible", NULL);
>> + if (!compatible)
>> + return -ENODEV;
>
> Please use of_match_device() and of_device_id.compatible instead.
OK.
>
>> +
>> + soc_id_start = strchr(compatible, ',') + 1;
>> + soc_id_end = strchr(compatible, '-');
>> + size = soc_id_end - soc_id_start;
>> + if (size > 32)
>> + size = 32;
>> + strscpy(soc_id, soc_id_start, size);
>> +
>> + soc_dev_attr = devm_kzalloc(dev, sizeof(*soc_dev_attr), GFP_KERNEL);
>> + if (!soc_dev_attr)
>> + return -ENOMEM;
>> +
>> + soc_dev_attr->family = "RZ/G3S";
>> + soc_dev_attr->soc_id = devm_kstrdup(dev, soc_id, GFP_KERNEL);
>> + if (!soc_dev_attr->soc_id)
>> + return -ENOMEM;
>> +
>> + devid = readl(sysc->base + RZG3S_SYS_LSI_DEVID);
>> + revision = FIELD_GET(RZG3S_SYS_LSI_DEVID_REV, devid);
>> + soc_dev_attr->revision = devm_kasprintf(dev, GFP_KERNEL, "%u", revision);
>> + if (!soc_dev_attr->revision)
>> + return -ENOMEM;
>> +
>> + dev_info(dev, "Detected Renesas %s %s Rev %s\n", soc_dev_attr->family,
>> + soc_dev_attr->soc_id, soc_dev_attr->revision);
>> +
>> + soc_dev = soc_device_register(soc_dev_attr);
>> + if (IS_ERR(soc_dev))
>> + return PTR_ERR(soc_dev);
>> +
>> return rzg3s_sysc_reset_probe(sysc, "reset", 0);
>> }
>
> My first thought was "oh no, now this is handled/duplicated in two
> places", but if you later migrate the chip identification support for
> the rest of RZ/G2L devices to here, it may start to look better ;-)
Yes, this is how I see it going forward.
>
> One caveat is that soc_device_match() can be called quite early in
> the boot process, hence renesas_soc_init() is an early_initcall().
> So registering the soc_device from a platform_driver might be too late,
> especially since fw_devlinks won't help you in this particular case.
> However, I think all real early calls to soc_device_match() are gone
> since the removal of the support for R-Car H3 ES1.x, and all remaining
> calls impact only R-Car and RZ/Gx (not G2L) SoCs.
That is good to know. I get that we should be safe going forward with this
approach.
Thank you,
Claudiu Beznea
>
> Gr{oetje,eeting}s,
>
> Geert
>
