Re: [PATCH 3/4] tty: serial: bcm63xx_uart: add support for DT probing

[Florian Fainelli <f.fainelli@gmail.com>]

Hi Arnd,

Le 20/02/2014 04:16, Arnd Bergmann a écrit :
> On Thursday 20 February 2014 11:59:04 Jonas Gorski wrote:
>> On Thu, Feb 20, 2014 at 2:29 AM, Florian Fainelli <f.fainelli@gmail.com> wrote:
>>> 2014-02-19 16:00 GMT-08:00 Jonas Gorski <jogo@openwrt.org>:
>>>> On Thu, Feb 20, 2014 at 12:22 AM, Florian Fainelli <f.fainelli@gmail.com> wrote:
>>>>> @@ -857,6 +861,12 @@ static int bcm_uart_remove(struct platform_device *pdev)
>>>>>          return 0;
>>>>>   }
>>>>>
>>>>> +static const struct of_device_id bcm63xx_of_match[] = {
>>>>> +       { .compatible = "brcm,bcm63xx-uart" },
>>>>
>>>>  From my understanding, this should be "brcm,bcm6345-uart", because
>>>> this kind of uart appeared first on bcm6345 (well, maybe bcm6335, no
>>>> idea which one of these two was first, but the latter was never
>>>> supported in mainline anyway).
>>>
>>> That's right, in fact, I think it might be desirable to handle both
>>> compatible string, just as a hint that it is compatible with the
>>> entire bcm63xx family. Would that work for you?
>>
>> I think using a "generic" compatible string is rather frowned upon
>> (what do you do if there is eventually a bcm63xx chip with an
>> incompatible uart?), but I'm no device tree expert.
>
> It's ok to have a generic name, it's wildcards like the xx above
> that we try to avoid, since that breaks down when you get another
> device in the same SoC family that is not compatible. This is different
> from the Linux way of naming things.
>
> brcm,bcm6345-uart sounds good, if that is the closest we can get
> to a generic name, working under the assumption that it's the oldest
> implementation of this UART. Ideally we'd find someone with access
> to the design documents of the SoC to tell us what the UART is really
> called by whoever designed it (which may not even be Broadcom).

This is just called an uart, it does not have any specific project name 
as far as I could see, and BCM6345 was indeed the very first SoC using it.

>
> If we think there may be some level of variation between the UARTS
> in the various bcm63xx SoCs, it would be good to list both the
> specific model of the SoC as well as the generic name in "compatible",
> so the driver can later detect those differences without requiring
> an updated DT.

Makes sense. So far this UART has been not modified (as far as a 
programmer may be concerned) since its very first design in a BCM6345.

>
> A BCM63138 for instance could list this as
>
> 	compatible = "brcm,bcm62138-uart", "brcm-bcm6345-uart";
>
> 	Arnd
>
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html