[PATCH] arm: imx: correct the hardware clock gate setting for shared nodes

[u.kleine-koenig@pengutronix.de (Uwe Kleine-König)]

Hallo Sascha,

On Tue, Dec 09, 2014 at 09:46:53AM +0100, Sascha Hauer wrote:
> On Tue, Dec 09, 2014 at 03:58:05PM +0800, Anson Huang wrote:
> > For those clk gates which hold share count, since its is_enabled
> > callback is only checking the share count rather than reading
> > the hardware register setting, in the late phase of kernel bootup,
> > the clk_disable_unused action will NOT handle the scenario of
> > share_count is 0 but the hardware setting is enabled, actually,
> > uboot normally enables all clk gates, then those shared clk gates
> > will be always enabled until they are used by some modules.
> > 
> > So the problem would be: when kernel boot up, the usecount cat
> > from clk tree is 0, but the clk gates actually is enabled in
> > hardware register, it will confuse user and bring unnecessary power
> > consumption, take i.MX6SX for example, the ESAI clk info
> > is as below, the use count is 0, but the hardware register read
> > from CCM_CCGR1_CG8 is ungated.
> 
> I believe the problem is that the shared count is not increased
> correctly during registration. When it would be increased during
> registration for each clock that is enabled in hardware clock_disable_unused
> could do its job like intended.
> 
> Could you test the attached patch?
> 
> -------------------------------8<--------------------------
> 
> From 8095ee6e407f33887b2afa504767d00fcca4f10c Mon Sep 17 00:00:00 2001
> From: Sascha Hauer <s.hauer@pengutronix.de>
> Date: Tue, 9 Dec 2014 09:39:06 +0100
> Subject: [PATCH] ARM: i.MX: clk-gate2: Fix shared clock enable counting
> 
> When during registration of a shared clock the clock is enabled
> in hardware the shared_count must be increased. This makes sure
> that during clk_disable the shared_count is decreased first before
> actually diabling the clock. When done like this the is_enabled
> callback can return the hardware state of the clock, like it is
> intended.
> 
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
>  arch/arm/mach-imx/clk-gate2.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/arm/mach-imx/clk-gate2.c b/arch/arm/mach-imx/clk-gate2.c
> index 5a75cdc..a673364 100644
> --- a/arch/arm/mach-imx/clk-gate2.c
> +++ b/arch/arm/mach-imx/clk-gate2.c
> @@ -96,10 +96,7 @@ static int clk_gate2_is_enabled(struct clk_hw *hw)
>  {
>  	struct clk_gate2 *gate = to_clk_gate2(hw);
>  
> -	if (gate->share_count)
> -		return !!__clk_get_enable_count(hw->clk);
> -	else
> -		return clk_gate2_reg_is_enabled(gate->reg, gate->bit_idx);
> +	return clk_gate2_reg_is_enabled(gate->reg, gate->bit_idx);
>  }
>  
>  static struct clk_ops clk_gate2_ops = {
> @@ -129,6 +126,9 @@ struct clk *clk_register_gate2(struct device *dev, const char *name,
>  	gate->lock = lock;
>  	gate->share_count = share_count;
>  
> +	if (share_count && clk_gate2_reg_is_enabled(reg, bit_idx))
> +		(*share_count)++;
> +
This is equivalent to calling clk_enable on this clk. So the result is
that clk_disable_unused now succeeds to disable if all related clks are
unused. But it introduces a new problem I think.

Consider that the bootloader starts linux with the clk on, so
assuming 3 shared clks share_count ends at 3 after the clocks are
registered. Then a driver makes use of one of them, calls
clk_prepare_enable on it, share_count becomes 4, enable_count = 1.
clk_disable_unused disables the two unused clks making share_count = 2.
After the driver now calls clk_disable the clk stays on. -> bad

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |