From mboxrd@z Thu Jan  1 00:00:00 1970
From: sebastian.hesselbarth@gmail.com (Sebastian Hesselbarth)
Date: Mon, 27 Jan 2014 19:21:38 +0100
Subject: [PATCH 0/4] clk: mvebu: fix clk init order
In-Reply-To: <20140127153908.4f6f46c2@skate>
References: <1390673950-4521-1-git-send-email-sebastian.hesselbarth@gmail.com>
 <20140127153908.4f6f46c2@skate>
Message-ID: <52E6A3B2.6060301@gmail.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On 01/27/14 15:39, Thomas Petazzoni wrote:
> On Sat, 25 Jan 2014 19:19:06 +0100, Sebastian Hesselbarth wrote:
>> This patch set fixes clk init order that went upside-down with
>> v3.14. I haven't really investigated what caused this, but I assume
>> it is related with DT node reordering by addresses.
>>
>> Anyway, with v3.14 for MVEBU SoCs, the clock gating driver gets
>> registered before core clocks driver. Unfortunately, we cannot
>> return -EPROBE_DEFER in drivers initialized by clk_of_init. As the
>> init order for our drivers is always core clocks before clock gating,
>> we maintain init order ourselves by hooking CLK_OF_DECLARE to one
>> init function that will register core clocks before clock gating
>> driver.
>>
>> This patch is based on pre-v3.14-rc1 mainline and should go in as
>> fixes for it. As we now send MVEBU clk pull-requests to Mike directly,
>> I suggest Jason picks it up as a topic branch.
>
> I'm not sure I really like the solution you're proposing here. I'd very
> much prefer to keep one CLK_OF_DECLARE() per clock type, associated to
> one function registering only this clock type.

Have you ever had a look at e.g. clk-imx28.c? Not that I really like
the approach, but it is common practice to do so.

> Instead, shouldn't the clock framework be improved to *not* register a
> clock until its parent have been registered? If the DT you have the
> gatable clocks that depend on the core clocks, then the gatable clocks
> should not be registered if the core clocks have not yet been
> registered.
>
> Do you think this is possible? Am I missing something here?

As I said, clk_of_init does not care about return values from the
clock init functions. Without it, it cannot decide if a clock
driver failed horribly, failed because of missing dependencies, or
successfully installed all clocks. Also, it is early stuff and I guess
clk_of_init will have to build its own "defered_list" and loop over
until done.

BTW, this is a fix not an improvement. We should find an acceptable
solution soon. But I am still open for suggestions, too.

Sebastian