Re: [PATCH] clk: don't call __of_clk_get_by_name() unnecessarily from clk_get()

From: David Lechner <david@lechnology.com>
To: Russell King - ARM Linux <linux@armlinux.org.uk>
Cc: Bartosz Golaszewski <brgl@bgdev.pl>,
	Bartosz Golaszewski <bgolaszewski@baylibre.com>,
	Sekhar Nori <nsekhar@ti.com>,
	linux-kernel@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	Kevin Hilman <khilman@baylibre.com>
Subject: Re: [PATCH] clk: don't call __of_clk_get_by_name() unnecessarily from clk_get()
Date: Mon, 12 Feb 2018 15:19:42 -0600	[thread overview]
Message-ID: <12e2aeb6-d1be-4c62-d48f-4d62b46dfc99@lechnology.com> (raw)
In-Reply-To: <20180212210053.GO9418@n2100.armlinux.org.uk>

On 02/12/2018 03:00 PM, Russell King - ARM Linux wrote:
> On Mon, Feb 12, 2018 at 02:51:57PM -0600, David Lechner wrote:
>> On 02/12/2018 08:24 AM, Bartosz Golaszewski wrote:
>>> From: Bartosz Golaszewski <bgolaszewski@baylibre.com>
>>>
>>> The way this function is implemented caused some confusion when
>>> converting the TI DaVinci platform to using the common clock framework.
>>>
>>> Current kernel supports booting DaVinci boards both in device tree as
>>> well as legacy, board-file mode. In the latter, we always end up
>>> calling clk_get_sys() as of_node is NULL and __of_clk_get_by_name()
>>> returns -ENOENT.
>>>
>>> It was not obvious at first glance how clk_get(dev, NULL) will work in
>>> board-file mode since we always call __of_clk_get_by_name(). Let's make
>>> it clearer by checking if of_node is NULL and skipping right to
>>> clk_get_sys().
>>>
>>> Cc: Sekhar Nori <nsekhar@ti.com>
>>> Cc: Kevin Hilman <khilman@baylibre.com>
>>> Cc: David Lechner <david@lechnology.com>
>>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
>>> ---
>>>   drivers/clk/clkdev.c | 2 +-
>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/clk/clkdev.c b/drivers/clk/clkdev.c
>>> index 7513411140b6..f394e8964909 100644
>>> --- a/drivers/clk/clkdev.c
>>> +++ b/drivers/clk/clkdev.c
>>> @@ -199,7 +199,7 @@ struct clk *clk_get(struct device *dev, const char *con_id)
>>>   	const char *dev_id = dev ? dev_name(dev) : NULL;
>>>   	struct clk *clk;
>>> -	if (dev) {
>>> +	if (dev && dev->of_node) {
>>>   		clk = __of_clk_get_by_name(dev->of_node, dev_id, con_id);
>>>   		if (!IS_ERR(clk) || PTR_ERR(clk) == -EPROBE_DEFER)
>>>   			return clk;
>>>
>>
>> Shouldn't you be sending this to the linux-clk mailing list and cc'ing
>> the clock maintainers?
> 
> No, I'm the maintainer for clkdev, as per MAINTAINERS.

Oops, I guess I should have looked before I said something.

> 
>> FWIW, it seems pretty clear to me that if we are using a board file
>> then we should expect clk_get_sys() to be called because there is
>> no device tree.
> 
> clk_get() pre-dates DT, and using it has no bearing on whether DT is
> in use or not.  The above change looks correct to me - if the
> struct device is not a DT device, then we shouldn't be trying to look
> up the clock in DT.
> 

Looks fine to me too.