Re: [PATCH] rtc: omap: Support scratch registers

From: Keerthy <j-keerthy@ti.com>
To: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Cc: Sekhar Nori <nsekhar@ti.com>, <linux-rtc@vger.kernel.org>,
	<linux-kernel@vger.kernel.org>,
	Linux OMAP List <linux-omap@vger.kernel.org>
Subject: Re: [PATCH] rtc: omap: Support scratch registers
Date: Wed, 8 Nov 2017 14:01:37 +0530	[thread overview]
Message-ID: <79f52c27-1c6a-c4ec-9635-d05eec635661@ti.com> (raw)
In-Reply-To: <20171108082145.kpd3c45x5skce5wd@piout.net>

On Wednesday 08 November 2017 01:51 PM, Alexandre Belloni wrote:
> On 08/11/2017 at 13:36:15 +0530, Keerthy wrote:
>>
>>
>> On Wednesday 08 November 2017 12:46 PM, Alexandre Belloni wrote:
>>> On 08/11/2017 at 12:38:05 +0530, Keerthy wrote:
>>>>
>>>>
>>>> On Wednesday 08 November 2017 11:57 AM, Alexandre Belloni wrote:
>>>>> Hi,
>>>>>
>>>>> On 08/11/2017 at 11:30:45 +0530, Keerthy wrote:
>>>>>>>>> +static int omap_rtc_scratch_read(void *priv, unsigned int offset, void *_val,
>>>>>>>>> +				 size_t bytes)
>>>>>>>>> +{
>>>>>>>>> +	struct omap_rtc	*rtc = priv;
>>>>>>>>> +	u32 *val = _val;
>>>>>>>>> +	int i;
>>>>>>>>> +
>>>>>>>>> +	for (i = 0; i < bytes / 4; i++)
>>>>>>>>> +		val[i] = rtc_readl(rtc,
>>>>>>>>> +				   OMAP_RTC_SCRATCH0_REG + offset + (i * 4));
>>>>>>
>>>>>> Can the offset be the Scratch register number instead of bytes offset?
>>>>>> More intuitive to me.
>>>>>>
>>>>>> So that one can request using offset as 0, 1, 2 instead of 0, 4, 8?
>>>>>>
>>>>>
>>>>> Well, the offset is coming from the nvmem core, itself getting it from
>>>>> the Linux file API (and it is in bytes). However, you have the guarantee
>>>>> that it will be aligned on a word, see:
>>>>> http://elixir.free-electrons.com/linux/latest/source/drivers/nvmem/core.c#L88
>>>>
>>>> Okay Alexandre. Thanks for clarifying. Looks good to me.
>>>> I have tested on AM437X-GP-EVM.
>>>>
>>>
>>> If needed, you can define nvmem cells (and I guess that is what you
>>> want):
>>> http://elixir.free-electrons.com/linux/latest/source/Documentation/devicetree/bindings/nvmem/nvmem.txt
>>
>> With this patch we cannot still use nvmem_device_write and
>> nvmem_device_read as the nvmem_device is still not registered.
>>
>> How can a driver write to scratch pad registers? Should we be calling
>> nvmem_register in the probe?
> 
> It is not needed, it is registered by rtc_nvmem_register().

Let me go through the chain:

rtc_register_device -> __rtc_register_device --> rtc_nvmem_register ->
rtc_nvram_register

I am seeing that for rtc-omap driver there is rtc_register_device call
and i see that rtc_nvmem_register is not getting called.

Should we add rtc_register_device in probe of rtc-omap?

> 
> nvmem_device_read/write are working because that is what I'm using to
> implement the old sysfs ABI with rtc_nvram_read/write.
> 