[RFC V2 6/8] ARM i.MX: prepare common clk support

[richard.zhao@linaro.org (Richard Zhao)]

On Fri, Dec 16, 2011 at 03:34:14PM -0800, Mike Turquette wrote:
> On Wed, Dec 14, 2011 at 1:23 AM, Richard Zhao <richard.zhao@linaro.org> wrote:
> > From: Sascha Hauer <s.hauer@pengutronix.de>
> >
> > Add static clock help macros, clock register spinlock.
> >
> > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> > Signed-off-by: Richard Zhao <richard.zhao@linaro.org>
> > ---
> > ?arch/arm/plat-mxc/clock.c ? ? ? ? ? ? ?| ? ?8 ++++++
> > ?arch/arm/plat-mxc/include/mach/clock.h | ? 42 +++++++++++++++++++++++++++++--
> > ?2 files changed, 47 insertions(+), 3 deletions(-)
> >
> ...
> >
> > +extern spinlock_t imx_ccm_lock;
> > +
> > +#define DEFINE_CLK_GATE2B(_name, _parent, _reg, _shift) \
> > + ? ? ? struct clk_gate2b _name = { \
> > + ? ? ? ? ? ? ? .clk = { \
> > + ? ? ? ? ? ? ? ? ? ? ? .name = #_name, \
> > + ? ? ? ? ? ? ? ? ? ? ? .ops = &clk_gate2b_ops, \
> > + ? ? ? ? ? ? ? ? ? ? ? .parent = _parent, \
> > + ? ? ? ? ? ? ? }, \
> > + ? ? ? ? ? ? ? .reg = (_reg), \
> > + ? ? ? ? ? ? ? .shift = (_shift) * 2, \
> > + ? ? ? ? ? ? ? .val_en = 0x3, \
> > + ? ? ? ? ? ? ? .val_dis = 0x0, \
> > + ? ? ? ? ? ? ? .lock = &imx_ccm_lock, \
> > + ? ? ? }
> > +
> > +#define IMX_DEFINE_CLK_DIVIDER(_name, _parent, _flags, _reg, _shift, _width) \
> > + ? ? ? DEFINE_CLK_DIVIDER(_name, _parent, _flags, _reg, _shift, _width, \
> > + ? ? ? ? ? ? ? ? ? ? ? &imx_ccm_lock)
> > +
> > +#define IMX_DEFINE_CLK_MUX(_name, _reg, _shift, _width, _clks) \
> > + ? ? ? DEFINE_CLK_MUX(_name, _reg, _shift, _width, _clks, &imx_ccm_lock)
> 
> gate2b, basic div and basic mux clks all use the imx_ccm_lock.  How
> many registers are we talking about here?  If the number is very high
> then maybe per-register locking makes sense for you?
per-register/per-register group locking make me feel complicated. How do I group the
registers? and register layout may change for different SoCs.
Considering prepare_lock and enable_lock, only gate register that share with other
non-gate clocks needs a lock. Keeping the rule in mind is strange, and easy to come
with faults.
Another way is to use a global lock. Do normal register read/write in the lock. but for
status busy wait checking, it's read only, we can release the lock temporarily.

Thanks
Richard
> 
> Also, do you need locking added to the basic gated clk type, or any of
> the others?  I'll probably add it anyways but just curious...
I think yes. Give usr an option to use the lock.

Thanks
Richard
> 
> Regards,
> Mike