Re: [PATCH 0/5] Generic clocks framework

[ian <spyro@xxxxxxx>]

On Thu, 2008-02-07 at 03:58 +0300, Dmitry Baryshkov wrote:
> Hi,
> 
> Here is the current version of clocklib patchset.

Just a thought on this, but it might be nice if we could use the
resource APIs to handle passing clocks to devices?

Not saying we should, or that its intrinsically better, just that its an
alternative method.

Im just musing over the case of the tc6387 SoC which has only one
subdevice but still needs to use the MFD API because it uses the
tmio_mmc driver.

In the case of that device, its core doesnt even gate the MMC core clock
(32kHz), so with the currently planned way of doing things we'd have
(mostly because I remember we were told not to pass struct clk * to
drivers)

1) Declare GPIO12_CLK clock in pxa25x.c  (this is OK, it happens already
2) Declare an alias of GPIO12_CLK as CLK_32K for the tc6387 core
3) the tc6387 core creates another clock alias - CLK_MMC_32K for the
TMIO MMC driver.

To me, it seems like it'd be simpler to have my platform code for the
tc6387 core 'get' the struct clk* and simply pass it to the tc6387 core,
which would then just pass it to the tmio_mmc driver.

I wonder which would be smaller, creating loads of clock aliases or
creating a bunch of IORESOURCE_CLK entries in the mfd core drivers.

-Ian


-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php