Re: [PATCHv2 0/4] ACPI / LPSS: Solution for two issues seen on Asus T100

["Li, Aubrey" <aubrey.li@linux.intel.com>]

On 2014/5/16 0:11, Mika Westerberg wrote:
> On Thu, May 15, 2014 at 11:59:46PM +0800, Li, Aubrey wrote:
>> On 2014/5/15 22:53, Andy Shevchenko wrote:
>>> On Thu, 2014-05-15 at 22:35 +0800, Li, Aubrey wrote:
>>>> On 2014/5/15 21:40, Heikki Krogerus wrote:
>>>>> Changes since v1:
>>>>> - now using do_div() in clk_fd_recalc_rate() as suggested by Andy
>>>>> - NULL checks for clk_name allocation in acpi_lpss.c
>>>>>
>>>>> This combines two patch sets for LPSS that I had already send for
>>>>> review separately. They conflicted with each other.
>>>>>
>>>>> The first two patches will fix a problem were the context of the
>>>>> private LPSS registers is lost when entering D3. The last two will add
>>>>> support for the M/N dividers on LPSS by adding a new basic clock type
>>>>> for fractional dividers. The UART driver needs support for it in order
>>>>> to get clock rates that suit the requested baud rates.
>>>>
>>>> The major issue in my mind is, this proposal makes a couple between I2C
>>>> designware, HSUART, or probably DMA driver as well with LPSS driver.
>>>
>>> acpi_lpss driver creates platform devices for each found and enumerated
>>> device.
>>
>>> If there no acpi_lpss enabled the drivers work as supposed without it.
>>
>> This is not true.
> 
> The drivers work fine on non-LPSS platform. If you make them depend on
> acpi_lpss, you break that.
> 

People don't know the relationship between LPSS driver and I2C/HSUART,
there is nowhere to describe that. If LPSS driver is unchecked, they
will encounter a weird issue which is very hard to figure out what's
going on.

Thanks,
-Aubrey
>