[PATCH 1/5] USB: ehci-atmel: rework clk handling

[Douglas Gilbert]

On 15-03-18 08:28 AM, Greg Kroah-Hartman wrote:
> On Wed, Mar 18, 2015 at 12:16:22PM +0100, Nicolas Ferre wrote:
>> Le 17/03/2015 20:01, Alan Stern a écrit :
>>> On Tue, 17 Mar 2015, Boris Brezillon wrote:
>>>
>>>> The EHCI IP only needs the UTMI/UPLL (uclk) and the peripheral (iclk)
>>>> clocks to work properly. Remove the useless system clock (fclk).
>>>>
>>>> Avoid calling set_rate on the fixed rate UTMI/IPLL clock and remove
>>>> useless IS_ENABLED(CONFIG_COMMON_CLK) tests (all at91 platforms have been
>>>> moved to the CCF).
>>>>
>>>> This patch also fixes a bug introduced by 3440ef1 (ARM: at91/dt: fix USB
>>>> high-speed clock to select UTMI), which was leaving the usb clock
>>>> uninitialized and preventing the OHCI driver from setting the usb clock
>>>> rate to 48MHz.
>>>> This bug was caused by several things:
>>>> 1/ usb clock drivers set the CLK_SET_RATE_GATE flag, which means the rate
>>>>     cannot be changed once the clock is prepared
>>>> 2/ The EHCI driver was retrieving and preparing/enabling the uhpck
>>>>     clock which was in turn preparing its parent clock (the usb clock),
>>>>     thus preventing any rate change because of 1/
>>>>
>>>> Signed-off-by: Boris Brezillon <boris.brezillon at free-electrons.com>
>>>
>>> Acked-by: Alan Stern <stern at rowland.harvard.edu>
>>
>> Acked-by: Nicolas Ferre <nicolas.ferre at atmel.com>
>> and:
>> Fixes: 3440ef169100 ("ARM: at91/dt: fix USB high-speed clock to select UTMI")
>>
>> Alan, Greg,
>>
>> Can you please take this patch (only this patch 1/5 of the series) as a fix
>> for the 4.0-rc? It would solve the issue that we see on at91sam9x5/at91sam9n12.
>> I'll take care of the rest of the series for 4.1.
>>
>> If you want me to take it of to re-send the patch, tell me.
>
> I'll queue it up now, thanks.  You can take the rest of the series :)

Hi,
I reported this error to Nicolas based on tests with
lk 4.0-rc4. With the same USB WiFi dongle and platform
this patch fixes my problem.

Tested-by: Douglas Gilbert <dgilbert at interlog.com>